The Engineering of Sport 7 Vol. 1
Springer Paris Berlin Heidelberg New York Hong Kong Londres Milan Tokyo
Margaret Estivalet Pierre Brisson
The Engineering of Sport 7 Vol. 1
Margaret Estivalet ESTIA Technopole Izarbel 64210 Bidart France
Pierre Brisson UTC 66, avenue de Landshut 60200 Compiègne France
ISBN-13: 978-287-09410-1 Springer Paris Berlin Heidelberg New York
© Springer-Verlag France, Paris, 2008 Printed in France Springer-Verlag France is member of groupe Springer Science + Business Media Apart from any fair dealing for the purposes of the research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1998, this publication may only be reproduced, stored or transmitted, in any forrn or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms of licenses issued by the copyright. Enquiry concerning reproduction outside those terms should be sent to the publishers. The use of registered names, trademarks, etc., in this publication does not imply, even in the absence of a specific statement that such names are exempt from the relevant laws and regulations and therefore free for general use.
Cover design: Jean-François Montmarché
ISEA 2008, just before the summer olympic games! What a fantastic opportunity to present a compilation of more than 160 articles talking about sports engineering, analysing the coefficients of friction between the balls and the rim and back-board for leather and synthetic basket balls, extracting the aerodynamic force data during real ski jumping flights, optimizing new prosthesis of the lower human leg, analysing the golf ball spin rate after impact, analysing the most common injury in sport climbing using eight fresh dozen cadaver fingers, describing the heat transfer in footwear using finite elements, measuring the aerodynamic performance of cycling time trial helmets, etc, … What a challenge too to be honnest1 A huge diversity of articles, top level contributions to sports engineering. Today the world is convinced sport is not only fun but economically a sector, a multi sector, which is not only growing if you only take into acccount the total turn over but is becoming one of the fast growing business. Sport is not any more reserved for top sporters who want to maintain a certain level in some disciplines, it became a new philosophy of life, a new trend, a way to cope with aging population, with the reality of the society today. Our every day life is concerned with sport or sport derived products or services, it is in our shoes, our suits, our car, our bike, at home, when we eat, when we drink, when we sleep, relax, when we look at TV for international events, when we listen, watch the news, for fun.The sports engineering community as it was noted two years ago keeps growing. We have to admit it was a very difficult task to review all the contributions and to come down to 150 articles; It was very difficult too to allocate reviewers to contributions because a lot of articles were proposing not only scientific contributions but also engineering solutions and methodologies. Some groups of articles could have been selected as a basis for a workshop in itself! In front of such a diversity of contributions we have decided not to group the articles by families, by themas, by keywords, by branches, by sports, by subjects, by numbers of contributions but we decided to regroup it in two different volumes without any introduction which we thought would not bring anything to the readers, just proposing the articles in a natural order creating of course some surprises, but it was a choice! Of course there is a table listing the articles with their authors and co-authors and the programme will indicate evey time the article number. Complex to read? Difficult to apprehend? We thought it would give the best way to understand the complexity of sports engineering today; An article about football in a ball section of proceedings, in the shoes section, in the field surface section, in the injury section, in the training part, in the video group, in the sliding effect paragraph, in the referee point of view chapter, in the leather section may be, why not the aerodynamics or the finite elements analysis, may be in the professional sports section or the leisure, the TV business, the star system, … So many possibilities, we just did it in the way we were convinced would be the most open!
VI The Engineering of Sport 7 - Vol. 1 What we wanted to do is to provide the readers with the best sports engineering contributions in 2008, before the biggest sports event on earth, the olympic games, in front of 5 billions telespectators who will enjoy the show and for many of them start again sporting, or just start a new sport, realising what they can do, discover a new passion, using in any case the brain storming of the world of engineering contributors to improve our every day life. This is the magic of sport1 Margaret ESTIVALET & Pierre BRISSON
Contents
Effects of Body Weight on Ski Jumping Performances under the New FIS Rules (P3) ..............................................................................................
1
Luca Oggiano, Lars Sætran
Calculated Golf Ball Performance Based on Measured Visco-hyperelastic Material Properties (P5) ......................................................
11
Khairul Ismail, Bill Stronge
Interaction of Flexor Tendons and Pulleys in Sport Climbing (P6) .................................................................................................................
19
Andreas Schweizer, Beat Moor, Hans-Peter Bircher
Friction Between Players’ Hands and Sports Equipment (P7) .....
27
S.E. Tomlinson, R. Lewis, M.J. Carré
Development of a Comfort Model for Cricket Leg Guards (P9)
35
James Webster, Jonathan Roberts, Roy Jones
Enabling Technologies for Robust Performance Monitoring (P10) .................................................................................................................................
45
Laura Justham, Sian Slawson, Andrew West, Paul Conway, Michael Caine, Robert Harrison
An Objective Performance and Quality Comparison of Drivers from Different Market Sectors (P11) ............................................
55
Jeff Brunski, John Rae
Defining Strategies for Novel Snowboard Design (P12) .......................
65
Aleksandar Subic, Patrick Clifton, Jordi Beneyto-Ferre
Business Process Modelling and its Use Within an Elite Training Environment (P15) ...........................................................................
73
Laura Justham, Sian Slawson, Andrew West, Paul Conway, Michael Caine, Robert Harrison
Accelerometer Profile Recognition of Swimming Strokes (P17) .
81
S.E. Slawson, L.M. Justham, A.A. West, P.P. Conway, M.P. Caine, R. Harrison
Evaluation of Start Techniques in Sports Swimming by Dynamics Simulation (P18).................................................................................................................................... 89 Thomas Härtel, Axel Schleichardt
VIII The Engineering of Sport 7 - Vol. 1 A simulation of outrigger canoe paddling Performance (P19) .............................................................................................................................
97
Nicholas Caplan
The Dynamic Compaction of Cricket Soils for Pitch Preparation (P20) ................................................................................................................................. 107 Peter Shipton, Iain James
Experimental Validation of a Finite-element Model of a Tennis Racket String-bed (P21)...................................................................................... 115 Tom Allen, Simon Goodwill, Steve Haake
Experimental Validation of a Tennis Ball Finite-element Model for Different Temperatures (P22) ..................................................................................... 125 Tom Allen, Simon Goodwill, Steve Haake
Nonlinear Dynamics of a Simplified Skateboard Model (P24) ..... 135 Alexander S. Kuleshov
Cricket Batting Stroke Timing of a Batsman When Facing a Bowler and a Bowling Machine (P26) ..................................................................... 143 Alex Cork, Laura Justham, Andrew West
Estimation of a Runner’s Speed Based on Chest-belt Integrated Inertial Sensors (P27) ...................................................................................................................... 151 Rolf Vetter, Emanuel Onillon, Mattia Bertschi
Design and Construction of a Custom-made Lightweight Carbon Fibre Wheelchair (P28) .......................................................................................... 161 Marc Siebert
Design and Implementation of a Rugby-specific Garment Evaluation Trial (P30)..................................................................................................................... 169 Bryan C. Roberts, Gareth Williams, Mike P. Caine
Open Rotator Cuff Surgery in Swiss Elite Rock Climbers (P31) . 177 Hans-Peter Bircher, Christoph Thür, Andreas Schweizer
A Quantitative Analysis Of Beach Casting (P33) ........................................... 183 Benjamin Charles, Darryl P Almond, Aki I T Salo, Presented by Alan N Bramley
An Assessment of Sensing Technologies to Monitor The Collision of a Baseball and Bat (P34)............................................................... 191 Lawrence Fallon, James Sherwood, Michael Donaruma
Correlation Between the Linear Impulse and Golf Ball Spin Rate (P35) ........................................................................................................................................ 199 Woo-Jin Roh, Chong-Won Lee
The Engineering of Sport 7 - Vol. 1
IX
Dynamics-based Force Sensor Using Accelerometers-application of Hammer Throw Training Aid- (P37) .................................................................... 207 Ken Ohta, Koji Umegaki, Koji Murofushi, Ayako Komine, Chikara Miyaji
Influence of Pedal Foot Position on Muscular Activity during Ergometer Cycling (P39) .............................................................................................................. 215 Stefan Litzenberger, Sandrina Illes, Martin Hren, Martin Reichel, Anton Sabo
Accurate Trajectory and Orientation of a Motorcycle derived from low-cost Satellite and Inertial Measurement Systems (P42)............. 223 Adrian Waegli, Alain Schorderet, Christophe Prongué, Jan Skaloud
Wireless Impact Measurement for Martial Arts (P43) ............................ 231 J.I. Cowie, J.A. Flint, A.R. Harland
A Comparative Study of Ball Launch Measurement Systems; Soccer Case Study (P44) ................................................................................................................ 239 Jouni Ronkainen, Chris Holmes, Andy Harland, Roy Jones
Testing Protocol for Quantitative Comparison of Top of the Range Soccer Boots (P45) ........................................................................................................... 247 Jouni Ronkainen, Dan Toon, Joe Santry, Tom Waller
Development of a Measurement-Prosthesis for a Ski Boot Test Bench (P48)..................................................................................................................................... 255 M. Reichel, A. Haumer, H. Schretter, A. Sabo
Development of Multi-platform Instrumented Force Pedals for Track Cycling (P49) ................................................................................................................. 263 Jean-Marc Drouet, Yvan Champoux, Sylvain Dorel
In-Situ Measurement of Clipless Cycling Pedal Floating Angles (P51) ................................................................................................................................................ 273 Yvan Champoux, Daniel Paré, Jean-Marc Drouet, Denis Rancourt
Correlation Between Treadmill Acceleration, Plantar Pressure, and Ground Reaction Force During Running (P52) .................................. 281 Alex, J. Y. Lee, Jia-Hao Chou, Ying-Fang Liu, Wei-Hsiu Lin, Tzyy-Yuang Shiang
Development of Immediate Feedback Software for Optimising Glide Performance and Time of Initiating Post-Glide Actions (P56) .............................................................................................................. 291 Roozbeh Naemi, Serdar Aritan, Simon Goodwill, Steve Haake, Ross Sanders
Rod Response Analysis to Fish Bite Based on Multi-link Model Solved by Lower Triangularization of Sparse Symmetric Coefficient-matrix (57) .................................................................................................................. 301 Shigeyuki Yamabe, Hiromitsu Kumamoto, Shingo Nishioka
X
The Engineering of Sport 7 - Vol. 1
Design and Manufacture of Customised Orthotics for Sporting Applications (P62) ............................................................ 309 Paul Crabtree, Vimal Dhokia, Martin Ansell, Stephen Newman
Analysis of Snowboard Stiffness and Camber Properties for Different Riding Styles (P65) .................................................................................................. 319 Aleksandar Subic, Patrick Clifton, Jordi Beneyto-Ferre, Arnaud LeFlohic, Yoshiki Sato, Victor Pichon
The Fluctuating Flight Trajectory of a Non-Spinning Punted Ball in Rugby (P67) .......................................................................................................................................... 329 Kazuya Seo, Osamu Kobayashi, Masahide Murakami
Aerodynamics of Bicycle Helmets (P68) ................................................................... 337 Firoz Alam, Aleksandar Subic, Aliakbar Akbarzadeh
Aerodynamics of Cricket Ball-an Effect of Seams (P70) ....................... 345 Firoz Alam, Roger La Brooy, Aleksandar Subic, Simon Watkins
Numerical Modelling of the Flow Around Rowing Oar Blades (P71) .................................................................................................................................... 353 Anna Coppel, Trevor Gardner, Nicholas Caplan, David Hargreaves
The Acute Response to a Garment-based Elastic Thoracic Load, Applied During Exercise on Inspiratory Muscle Strength and Pulmonary Function (P72) ....................................................................................................... 363 Ashley R. Gray, Dr Tom M. Waller, Prof Mike P. Caine
Aerodynamic Performance of Cycling Time Trial Helmets (P76)............................................................................................................................. 371 Kim B. Blair, Ph.D., Stephanie Sidelko
Physical Motion Analysis of Nordic Walking (P77) .................................... 379 Takayuki Koizumi, Nobutaka Tsujiuchi, Masaki Takeda, Yusuke Murodate
Driving Performance Variability Among Elite Golfers (P79).......... 387 Ian C. Kenny, Eric S. Wallace, Steve R. Otto
Power Measurement in Cycling using inductive Coupling of Energy and Data (P80) ............................................................................................................ 397 Reinhardt Tielert†, Norbert Wehn, Thomas Jaitner, Roland Volk
Online-Monitoring of Multiple Track Cyclists During Training and Competition (P81) ....................................................................................... 405 Thomas Kuhn, Thomas Jaitner, Reinhard Gotzhein
A Model Predictive Controller for Sensor-based Training Optimization of a Cyclist Group (P82) ...................................................................... 413 Ankang Le, Lothar Litz, Thomas Jaitner
The Engineering of Sport 7 - Vol. 1
XI
A Dynamic Heart Rate Prediction Model for Training Optimization in Cycling (P83) ............................................................................................. 425 Ankang Le, Thomas Jaitner, Frank Tobias, Lothar Litz
Stability Training and Measurement System for Sportsperson (P84) ................................................................................................................... 435 S.N. Omkar, D.K. Ganesh
SRM Torque Analysis of Standing Starts in Track Cycling (P85) 443 Paul Barratt
Aerodynamic Study of Ski Jumping Flight Based on High-Speed Video Image (P86) ............................................................................................ 449 Masahide Murakami, Nobuyuki Hirai, Kazuya Seo, Yuji Ohgi
The Role of Materials and Construction on Hockey Ball Performance (P88)............................................................................................................................... 457 Dan Ranga, James Cornish, Martin Strangwood
Shape Optimization of Golf Clubface using Finite Element Impact Models (P90)......................................................................................................................... 465 Willem Petersen, John McPhee
An Examination of Cricket Bat Performance (P92) ................................... 475 Lloyd Smith, Harsimranjeet Singh
Forces Applied on Rowing Ergometer Concept2®: a Kinetic Approach for Development (P94) ......................................................... 483 Nicolas Découfour, Franck Barbier, Philippe Pudlo, Philippe Gorce
JUMPICUS – Computer Simulation in Ski Jumping (P95)................ 491 Heike Hermsdorf, Falk Hildebrand, Norman Hofmann, Sören Müller
Kinematic Response to Variations in Natural Turf During Running (P96)........................................................................................................................................... 499 Stiles, V. H., Dixon, S.D., Guisasola, I.N., James, I.T
Finite Element Simulation of Ice Pick Torquing (P97) ........................... 509 Rae S. Gordon, Kathryn L. Franklin
A Sociological Analysis of a Controversy in French Sport Science Field: How to Manage Teams Specialising in Technological Innovation (P99).................................................................................................................................... 519 Philippe Terral, Cécile Collinet
Biomechanical Ingredients Measurement: A New Vision-Based Approach (P102) .................................................................................................................................... 529 Mohammad Reza Mohammadi, Hadi Sadoghi Yazdi
XII The Engineering of Sport 7 - Vol. 1 How Optimal Baseball Swings Change for Three Levels of Play (P103) ............................................................................................................................................ 539 Ann Chase, Mont Hubbard, Chris Ray
Graduated Compression Stockings and Delayed Onset Muscle Soreness (P105) ....................................................................................................................................... 547 Stéphane Perrey, Aurélien Bringard, Sébastien Racinais, Kostia Puchaux, Nicolas Belluye
A Study of Knuckling Effect of Soccer Ball (P106) ....................................... 555 Takeshi Asai, Kazuya Seo, Yousuke Sakurai, Shinichiro Ito, Sekiya Koike, Masahide Murakami
Ball and Racket Movements Recorded at the 2006 Wimbledon Qualifying Tournament (P109) ........................................................................................... 563 Simon B Choppin, Simon Goodwill, Steve Haake, Stuart Miller
Ball Spin Generation at the 2007 Wimbledon Qualifying Tournament (P110) ............................................................................................................................ 571 John Kelley, Simon Goodwill, Jamie Capel-Davies, Steve Haake
Analysis and Optimization of the Sliding Properties of Luge Steel Blades on Ice (P111) ........................................................................................................... 579 Mathieu Fauve, Hansueli Rhyner
Brake Induced Vibration in Mountain Bikes (P112) ................................. 587 Robin C. Redfield
Aerodynamic Optimization and Energy Saving of Cycling Postures for International Elite Level Cyclists (P114) ............................. 597 Luca Oggiano, Stig Leirdal, Lars Sætran, Gertjan Ettema
A Comparison of Test Methodologies to Enable the Improved Understanding of Soccer Boot Traction (P115) ............................................... 605 J.D. Clarke, M.J. Carré, R.F. Kirk
How to Build an Optimized Movement Analysis Laboratory for High Performance Athletes of Various Sport Disciplines (P116) ................................................................................................................................ 613 Lars Janshen
Analysis of the Wobble of a Spinning Disc at Launch (P117) ......... 623 William Rae, Mont Hubbard
A Study of the Influence of the Environmental Condition and the Garment in Skin Temperature in Sport Activity (P119) ............ 631 Natividad Martínez, David Rosa, Javier Gámez, Juan Carlos González, Carlos Chirivella, José María Gutiérrez, Jaime Prat, José Javier Sánchez
The Engineering of Sport 7 - Vol. 1
XIII
Compression Sleeves Significantly Counteracts Muscular Fatigue During Strenuous Arm Exercise (P124) ............................................. 641 Thibaud Thedon, Nicolas Belluye, Stéphane Perrey
Development of a New System for Measuring Tennis Court Pace (P126) ................................................................................................................................ 649 Simon Goodwill, Steve Haake, James Spurr, Jamie Capel-Davies
A Feedback System for Coordination Training in Double Rowing (P127) .......................................................................................................................................... 659 Arnold Baca, Philipp Kornfeind
Modelling the Oblique Impact of Golf Balls (P128) ................................... 669 James Cornish, Steve Otto, Martin Strangwood
Modelling and Stability Analysis of a Recumbent Bicycle with Oscillating Leg Masses (P131) ............................................................................................... 677 Brendan Connors, Mont Hubbard
Computerised Games for Balance Training: A Pilot Study on Collegiate Females (P135) ................................................................................................. 687 Jonathan S. Wheat, Ben Heller, Stephanie Lovick
Effects of Turbo-jav Release Conditions on Distance of Javelic Throw (P136) ............................................................................................................................................. 697 M. Maeda
Differences Between Leather and Sybthetic NBA Basketballs (P137) ................................................................................................................................ 705 Hiroki Okubo, Mont Hubbard
Subject Index .............................................................................................................................................. 713
Effects of Body Weight on Ski Jumping Performances under the New FIS Rules (P3) Luca Oggiano1, Lars Sætran2
Topics: Aerodynamics. Abstract: Based on the results of several different experiments, it has been concluded that the weight of a ski jumper is crucial in performing a long ski jump. In response to this conclusion, many of the best ski jumpers in the world began dieting to reduce their weight, resulting in many underweight athletes and some incidents of anorexia. In order to deal with this problem the International Ski Federation (FIS) introduced a new rule where the ski length is determined by both the jumper’s height and weight. An athlete with a Body Mass Index (BMI) of less than 20 must reduce the length of his or her skis. To evaluate the effect of the new rules a numerical and experimental investigation on the effects of the BMI on ski jumper's performances has been done. A numerical model has been built in order to evaluate the effects of BMI on the final speed in the in-run path. The numerical results obtained from the model match experimental data present in the literature. Experiments in the wind tunnel have been made in order to evaluate the aerodynamic forces acting on the ski jumper and on the skis during the flight path according to the new FIS rules. Experiments have been carried out on a doll mounted on a 6 components balance and different positions and ski length have been tested. The data acquired have been introduced into a numerical model and the final jump length has been then estimated. In conclusions it has been found out that the current FIS rules do reduce the problem addressed but experiments shows that it is still more advantageous to lose weight and consequently cut the skis, compared to gaining weight in order to keep the full ski length. Keywords: Aerodynamic, Ski jumping, Drag, Lift.
1. Norwegian University of Science and Technology, Faculty of Engineering Science and Technology, N-7491 Trondheim, Norway - E-mail:
[email protected] 2. Norwegian University of Science and Technology, Faculty of Engineering Science and Technology, N-7491 Trondheim, Norway - E-mail:
[email protected]
2
The Engineering of Sport 7 - Vol. 1
1- Introduction Ski Jumping is a sport discipline which involves different engineering fields. In this paper we will focus on the effect that BMI (and then the body weight) has on ski jumper’s performances and especially on the jump length. The increase the BMI has 2 main effects on ski jumpers performances. It has a positive effect during the in-run (higher weight gives a higher speed at take-off) and it has a negative effect during the flight path (the higher the weight is, the shorter the jump is. The 2 effects are not balanced. The negative effect during the flight is much stronger than the positive one during the in run. The final effect of BMI increasing is a shorter jump length. [SM1]. Because of this advantage of being light in terms of jump length, athletes began to lose weight, and many cases of underweight and some of anorexia athletica [S] came up. This alarming trend forced FIS to create new rules in order to reduce the problem with underweight ski jumpers. Under the old rules, an athlete's ski length was determined by the athlete’s height only but, in 2004, the rules changed, and today the ski length is determined by both height and BMI. Under the new rules, an athlete with a BMI of less than 20 must reduce the length of his skis according to a table made by FIS. Any athlete who has a BMI below 17.5 is not allowed to participate in the competitions. The rules change required underweight ski jumpers to use shorter skis than the jumpers’ height had formerly allowed. The intention was to reduce the positive lift forces acting on a ski jumper and his equipment during the flight, hence reducing the positive effect of being light. When the new rules became operative, there was a general belief within the ski jumping community that it would be beneficial for the athletes to gain weight by building up their thigh muscles. The weight gained would cause an increase in BMI and the athletes could then keep their original ski length and, theoretically, increase the power generated at the jump. The trend that the athletes followed it has not been the same that FIS expected. The average BMI among all the athletes present in the Olympic Games in Turin 2006 has been 19.41, 0.5 less than the average BMI measured in 2000 [SM1]. In order to determine whether the changes to the rules are justified, experiments have been conducted in a wind tunnel with 1:1 model of skis and ski jumper and a numerical model for In-run and Flying-path has been made and used to compute the final jump length. Different positions and angles for both jumper and skis have been analyzed and tested. Experimental data from previously studies and wind tunnel experiments [SM2] have been used in order to determine the position of skis and body angle. An adaptive numerical model has been built trying to describe as close as possible the real flight path.
The Engineering of Sport 7 - Vol. 1
3
1.1 BMI The BMI is a relation between a person’s weight and height. It’s defined as: (1)
Table 1 - BMI compared to weight status.
From results regarding the decrease of ski jumpers BMI acquired from 1973 to 2000 and shown by Vaverka [V] and Müller [SM1] it has been noticed that the average BMI among ski jumpers dropped from a value above 23 in the past to a value under 20 in 2000 (see Fig. 1). The last data about 106 ski jumpers who participated to the Olympic Games in Turin 2006 give an average BMI of 19.41 (ca. 0.5 less than what was measured in 2000). This shows that the trend of losing weight for obtaining better performances has not been stopped with the introduction of the new FIS rules.
Figure 1 - BMI trend in ski jumping competitions during the last 40 years.
1.2 Forces acting on a ski jumper Several forces act on a ski jumper during the three phases of the jump (in-run, take off, flight and landing). The combination of these forces will decide whether the ski jump is successful or not.
4
The Engineering of Sport 7 - Vol. 1
Roughly dividing the jump into 2 parts, the in-run and the flight, the main forces acting during these parts are drag, gravity, ski-snow friction during the in-run and drag, lift and gravity during the flight.
2- Experimental Setup For the experiments, the wind tunnel of NTNU (Norwegian University of Science and Technology) in Trondheim has been used. The test section of the wind tunnel is 12.5 meters long, 1.8 m high, and 2,7 m wide. The wind tunnel is equipped with a 220KW fan that can produce a variation of speed between 0.5 - 30 m/s.The balance (Carl Schenck AG) used is a six component balance capable to measure the three forces and the three momentums around the three axes. Variations of forces and moments are measured using strain gauges glued to the balance body. The voltage outputs are measured by a LABVIEW based PC program.
3- Numerical simulation of the in-run In order to evaluate the effect of BMI on the take off speed a numerical model has been built. The mathematical model here presented includes friction forces between ski and snow, mass forces due to gravity and aerodynamic drag forces acting on the ski jumper. The simulation has been divided in three parts, following the different shapes of the in run path. The path (see Fig. 2) is divided in three parts one straight part with a constant heeling-angle 1 (no curvature), one curved part with a constant radius of curvature r and then another straight part with constant heeling angle 2.
Figure 2 - In-run path. The in-run path is divided into 3 different parts: 2 straight parts and a curved one between the 2 straight ones.
The results obtained applying the model show that the BMI does not have a huge influence on the speed at take off. The difference between the take-off speed calculated for an athlete with a BMI around 30 (weight 85kg and height 170cm) and the speed calculated for an athlete with a BMI around 15 (weight 45kg and height 170cm) is about
The Engineering of Sport 7 - Vol. 1
5
1m/s (see Fig. 3). Considering that most of the ski-jumpers have a BMI between 17 and 21, this difference is reduced and it has been calculated to be about 0.2 m/s.
Figure 3 - Effect of increase of BMI in the speed at take off calculated for the Granåsen jumping hill in Trondheim (Norway). The curve which reaches the lower speed has been calculated for a speed skater with BMI 16 while the curve which shows the higher speed has been simulated for a ski jumper with BMI 25. L1=50m, L2=6.8m, r=110m, 1=34.5˚, 2=11˚, TOTlength=101.9m.
4- Experimental investigation on the aerodynamic forces 4.1 Skis and doll position in the wind tunnel
Figure 3 - Different angles between ski jumper and wind direction. is the angle between wind direction and skis, is the angle between wind direction and ski jumper’s body and is the angle between wind direction and horizon line.
6
The Engineering of Sport 7 - Vol. 1
Skis and the doll have been tested separately since they can be considered as two separate systems. The ski’s wake does not in fact interact with the ski jumper’s body. This means that the 2 aerodynamic parameters (Lift and Drag) can be measured separately. They were both connected to a shielded support connected to the 6 component balance. The doll used for the test is 170cm tall and his position has been varied from 10degrees to 60degrees. The suit used is the same suit used by the Norwegian skijumpers during the Olympic Winter Games in 2006. The correct angles were adjusted using the joints on the support. Trying to model a ski jump in the most realistic way possible, the angles were adjusted in relation to the flying path, wind direction and the tilt of a skier’s ankle. In the experiment, 3 different velocity levels have been used:13 m/s, 20 m/s and 27 m/s, respectively. The ski-length is about 268cm and it has been tested at 6 angles relatives to 6 different flight positions. In order to evaluate the effect of the new FIS rules, the skis got cut in the back end for 5cm at a time and the same test has been done until a ski length of 248cm has been reached.
4.2 The flight path In order to evaluate the effects of the new FIS rules on the aerodynamic forces, data regarding the positions assumed by a ski jumper during his flying path were needed. These data have been acquired by Schmölzer, & Müller [SM2]. The simplified flight path used for the model here presented has been divided in five different parts, assuming aero dynamical forces to be constant in each part. Forces during the flight path have been decomposed in vertical and horizontal forces [R]. - Take off 1 (t<0,21s) (=-5, = 58) - Take off 2 (0,21
2.71) (=35, = 52)
5- Results The main target of the aerodynamic test on skis is to evaluate the effects of the new rules on Lift and Drag. Lift and drag has been measured per each position a, with different ski length, from 248 to 268cm. Table 2 – Experimental data for drag and lift.
The Engineering of Sport 7 - Vol. 1
7
Table 3 – Experimental data and FIS table.
Figure 8 - Comparison between the FIS rules and the data obtained with the wind tunnel experiments.
The ski jumper system has been obtained by summing the aerodynamic forces calculated separately for skis and doll. Wakes and possible interferences between doll and skis have been neglected. Comparing the curve obtained for the ski-jumper with the length-weight obtained from the new FIS rules (for a 180cm tall ski jumper) it has been obtained, for =30, =45 (which is the position assumed for the longest period during the flight path). According to the flight path angles measured by Schmölzer and. Müller [SM2] two angles have been chosen. =40 and =30 which represent the angles where the aerodynamic forces count more in the y direction and where f is the angle between the wind direction acting on the skier and the vertical axis parallel to the gravity force (fig. 8). The flight path has been divided in 5 parts. Per each part, drag and lift of the ski jumper have been calculated using the experimental data. A vertical velocity at take off of 2.5m/s has been considered [VKK]. To evaluate the effect of the new FIS rules on a ski jumper’s flight path, two different ski lengths, (248cm and 268cm) have been considered. A ski length of 248 cm correspond to a ski jumper (180cm tall) who weighs 58kg while 268 correspond to a ski jumper (180 cm tall) who weighs 65kg.
8
The Engineering of Sport 7 - Vol. 1
Drag and Lift has been considered constant during each part of the flying path and the correction relative to the take off speed illustrated in par.1 has been taken into account. The 2 curves presented in fig. 9 represent the flight path for 55kg ski jumper and a 67 kg ski jumper calculated using the new FIS rules. The solid curve is the path calculated for a 67kg ski jumper and the dotted curve is the curve obtained for a 55kg ski jumper. The difference in jump length between the 2 jumpers has been estimated to be about 7 meters and the longest jump length has been obtained by the 55kg jumper. The lighter jumper has then a big advantage due to his low body weight and this advantage is not enough compensated by the smaller aerodynamic forces acting on him during his flight. It has been also estimated that, in order to compensate the disadvantage due to the higher weight, the ski jumper who weight 67kg has to increase his vertical take-off speed at up to 3m/s.
Figure 9 - Simulated flight path for a 55kg ski jumper and a 67kg ski jumper in the Venlångerung jumping hill.
6- Conclusions The new rules imposed by FIS have not solved the problem of low BMI in ski jumping. Athletes are in fact keeping loosing weight in order to improve their performances. It has been demonstrated that the negative effects due to higher weight can not be compensated with the adjustment imposed by the new rules. Furthermore, the increase of speed at take off is not enough to compensate the negative effect due to increase of BMI. The possible solution proposed by some trainers, consisting on only increase the BMI by building up muscles in the thigh in order to obtain a higher vertical speed at take off could work but the vertical speed at take off sufficient to compensate the weighteffect should be around 3m/s, 20% higher than the normal vertical speed estimated by Virmavirta in 2.5m/s [VKK].
The Engineering of Sport 7 - Vol. 1
9
A better solution could be obtained by reducing the width of the skis instead of the length. This would not affect the stability of the jumper and at the same time will have a negative effect in terms of aerodynamic forces reducing the advantage of having a low BMI. New tests will then be carried out in order to evaluate this solution.
7- References [R] - Remizov L. Biomechanics of optimal flight in ski-jumping. Journal of biomechanics. 3. 167171, 1984. [S] - Smith, N.J. Excessive weight loss and food aversion in athletes simulating anorexia nervosa. Pediatrics 66, 139–143. 1980 [SM1] - Schmölzer, B., W. Müller. The importance of being light: aerodynamic forces and weight in ski jumping. Journal of biomechanics 35, 1059-1069, 2002. [SM2] - Schmölzer, B., W. Müller. Individual flight styles in ski jumping: results obtained during Olympic Games competitions. Journal of biomechanics 38, 1055-1065, 2004. [V] - Vaverka, F. Somatic problems associated with the flight phase in ski jumping. Studia I monografia AWF we Wroclawiv; 40: 123 – 128, 1994 [VKK] - Virmavirta, Kivekas, J. & M., Komi, P. Take off aerodynamics in ski jumping. Journal of Biomechanics, 34, 465-470, 2001.
Calculated Golf Ball Performance Based on Measured Visco-hyperelastic Material Properties (P5) Khairul Ismail1, Bill Stronge
Abstract: Dynamic behaviour of golf balls during oblique impact is investigated using Abaqus (FEA software) to represent the ball structure and measured material properties. The properties of the mantle layer were measured from independent tests of material properties. Materials of golf balls were shown to possess viscoelastic and hyperelastic (or visco-hyperelastic) behaviours. The merit of this approach is that characterisation of new materials and modelling impact of a golf ball can be performed before construction of a prototype golf ball. This paper compares calculated and experimental coefficients of restitution (COR) of a hollow, multilayer golf ball. Key words: Golf ball; Finite element analysis; Coefficient of restitution; Hyperelastic; Viscoelastic.
1- Introduction Dynamic behaviour of both bodies is important in analysing impact responses between a golf ball and a club. In this paper we investigate modelling of the impact between a golf club head and a hollow golf ball - a ball designed to have a larger polar moment of inertia about the centre. This ball was struck at an angle of obliquity by a rigid body simulating the club head and dynamic behaviour during impact was analysed. Experimental study on dynamic behaviour of golf balls can involve impacting a golf ball on a plate, either at normal or oblique angle. Results of such experiment were usually compared with an analytical solution from rigid body theory (Arakawa et al., 2006) or with the solution of a dissipative linear compliance impact model (Ujihashi 1994). FEA software such as Abaqus can be used as a tool in designing new golf equipments. Recent publications on golf balls tend to compare experimental results with the FEA results (for a few examples, see Tavares et al., 2006 and Tanaka et al., 2006). This paper investigates dynamic behaviour of golf balls during oblique impact using Abaqus to represent the ball structure and measured material properties. Previous FEA 1. Department of Engineering, University of Cambridge, Cambridge, CB2 1PZ, UK. - E-mail: [email protected]
12 The Engineering of Sport 7 - Vol. 1 studies on the subject used either a) measured properties of the ball as a whole (not the constituents) that were obtained from independent static and dynamic tests (Tavares et al., 2006) or b) predictions of properties obtained from rigorous parametric studies by comparing FEA results with independent dynamic tests of the ball (Tanaka et al., 2006). These investigations were phenomenological in that they selected properties which matched a measured impact outcome. In contrast, the present paper obtains properties for the mantle layer from independent tests of material properties. Materials of golf balls were shown to possess viscoelastic and hyperelastic (or visco-hyperelastic) behaviour. The merit of this approach is that characterisation of new materials and modelling impact of a golf ball can be performed before construction of a prototype golf ball. This paper compares calculated and experimental coefficients of restitution (COR) and initial spin of a hollow, multilayer golf ball where the calculation uses measured material properties.
2- Material properties of a hollow golf ball The golf ball to be investigated consists of three layers; namely a hollow Ti alloy core, a polymeric mantle and an ionomer resin cover. The study concentrates on the effect of differences between phenomenologically determined coefficients and measured properties of the mantle materials.
2.1 Phenomenological coefficients The core of the hollow ball is a thin-walled titanium alloy (Ti-6Al-4V) spherical shell. Since this core is quite stiff and shell deformation is anticipated to be below a yielding point, the core was modelled as an elastic material. The cover material for the golf ball is made of ionomer resin. Here, the cover was defined as a hyperelastic material using coefficients of Mooney-Rivlin strain energy function. The mantle material was defined as hyperelastic and viscoelastic. The properties for mantle and cover layers were taken from the phenomenologically determined coefficients given by Tanaka et al. (2006). These properties were selected based on maximum load, contact time, deformation history and rebound velocity by comparing dynamic test results with FEA simulations. Table 1 gives the material properties and dimensions for each layer of the golf ball. Table 1 - Dimensions and properties of layers of the hollow golf ball.
The Engineering of Sport 7 - Vol. 1
13
2.2 Measurements of visco-hyperelastic properties Measurements were made on the mantle material only. The mantle layer consists of polybutadiene rubber material. A static compression test was conducted on a dog-bone sample of the mantle material. The deformation of this material is highly nonlinear (Fig.1).
Figure 1 - Nominal stress-strain curve of mantle materials.
The rate-dependent or viscoelastic behaviour of the mantle material was measured using the Dynamic Mechanical Analysis (DMA) technique which is a forced vibration test of a cantilever specimen with a maximum strain of 5%. It was conducted at a constant temperature on the sample of mantle material to obtain elastic storage and elastic loss moduli during a frequency sweep of 0-20 Hz. Tests were then repeated at different temperatures. The time-temperature shift function such as the William-Landel-Ferry is known to work well with most viscoelastic materials (Oyadiji 2004). Using this, the timetemperature superposition principle was applied to the DMA test results to extend the frequency range (Fig. 2).
14 The Engineering of Sport 7 - Vol. 1
Figure 2 - Variations in the elastic storage and loss moduli with frequency .
3- Modelling impact using FEA The Abaqus software was employed to model impact of a deformable hollow, multilayer golf ball with a rigid club face. Since the deformations will be asymmetric in an oblique impact, a 3D model was required. Construction of the FEA impact model can be categorised into the following key stages;
3.1 Geometrical construction A deformable spherical body with hollow centre and thick outer layer was constructed. The thick layer was then sectioned into three different sub-layers with different material properties. This method of subsection removes the possibility of error in defining contact interaction of different layers of material. However, this method neglects the energy losses due to shear traction at the interface. A club face was modelled as a rectangular rigid plate using rigid shell elements.
The Engineering of Sport 7 - Vol. 1
15
3.2 Material definition Material properties are specified in Abaqus for each layer - these properties were obtained either using test data or coefficients for a chosen constitutive equation of the material e.g. elastic, elastic-plastic, hyperelastic or viscoelastic. The constitutive behaviour of hyperelastic material is defined in terms of ‘strain energy potential’, U() in Abaqus software. Abaqus computes the strain energy potential based on the user’s input of the nominal stress-strain test data. There are several forms of strain energy potential e.g. Mooney-Rivlin, Ogden and Marlow, where each form is suitable for different type of test data (uniaxial, biaxial, planar or volumetric test). The hyperelastic materials are normally nearly incompressible, hence modelling the dynamic behaviour using Abaqus requires accurate material definition especially if it is confined between stiffer components, as the case for this golf ball. During collision, the golf ball experienced compressive strain due to the contact force. Thus, Marlow form of the strain energy potential was chosen since only uniaxial compression test data (Fig. 1) was used for describing the hyperelastic behaviour. This form produces a reliable calibration for the case where only one type of test data is available (e.g. compression test data). Alternatively, a strain energy function can be selected with pre-defined coefficients. The broad-band test data of viscoelastic behaviour were expressed in terms of storage and loss moduli. These data were calibrated by Abaqus in order to derive Prony series, which were used to define viscoelastic properties in time domain. Otherwise, the coefficients of Prony series were defined using the viscoelastic material definition.
3.3 Contact interaction Interaction between the rigid plate and the golf ball was defined using a contact algorithm that allows slip motion once the shear stress on the interface reached a critical value, which is based on the Coulomb friction model. If the shear surface traction is below the friction limit, stick occurs where the elements vibrate in the direction parallel to the interface. The coefficient of friction between the ball and the plate was taken as 0.3 (Tanaka et al., 2006)1.
3.4 Meshing of elements An eight node solid element was used to model the ball with three layers; core, mantle and cover. The colliding rigid plate was modelled as discrete rigid element. Meshing was made easier by partitioning the balls into 8 quadrants. The density of meshing can be controlled with fine mesh was concentrated on the area with significant deformations. An explicit formulation was used since this impact involves a transient dynamic event.
3.5 Initial and boundary conditions An initial uniform velocity field was defined for the golf ball at the start of analysis (t=0). Reference node of the rigid plate which represents the motion of a golf club head was fixed for the entire analysis duration. 1. Alternative friction theories may be appropriate for the polymeric golf ball cover sliding on a metal club face.
16 The Engineering of Sport 7 - Vol. 1
3.6 Output The measured material properties of the mantle layer were compared to the visco-hyperelastic coefficients of Tanaka et al. (2006). The core and cover material properties were kept similar in all simulations (as in Table 1). Hence, the simulations were divided according to the definitions of the mantle properties in Abaqus; i) T - phenomenologically determined Mooney-Rivlin and three-element viscoelastic material properties for mantle - refer to Table 1. ii) M - measured hyperelastic and viscoelastic material properties for mantle refer to Fig. 1 and 2. Simulations were conducted for an impact speed V0 of 45 m/s and three different loft angles of 10°, 20° and 30°; the purpose was to investigate the effect of mantle properties on calculated values of contact force, spin and COR. The contact force between ball and plate was measured at the reference node. The velocity of the ball after impact was approximated as an average velocity of nodes in the region at two opposite diameter in the core layer. This layer has the highest stiffness; consequently, this measure minimises the effect of post-impact radial and tangential vibrations on the rebound velocity.
4- FEA results and discussion
Figure 3 - Maximum normal contact force at different loft angle (V0=45 m/s). An inset shows the impact configuration of a golf ball and a rigid plate.
The Engineering of Sport 7 - Vol. 1
17
The comparisons of calculated maximum normal force for simulations using Tanaka (T) and measured (M) material properties are shown in Fig. 3. The maximum normal force reduces with increasing angle. The differences are about 16-19% at these different angles. The measurement of spin shows differences up to 10% between simulations T and M (Fig. 4). Spin of the ball increases almost linearly with the loft angle.
Figure 4 - Spin resulting from different loft angle (V0=45 m/s). An inset shows the cross-section mesh around the contact region of a hollow golf ball.
Although significant, these disagreements in the results of simulations using material properties obtained from phenomenological tests and measurements are attributed to differences in the properties of polybutadiene rubber of the mantle layer; e.g. variation in the elastic modulus of the mantle layer will significantly affect the maximum normal contact force. A range of different polybutadiene rubbers are used in golf balls. The results of simulation T (Tanaka et al., 2006) do not necessarily represent experimental values since, with the exception of mantle density, the parameters in Table 1 were obtained to match performance of a solid rather than a hollow golf ball. The Tanaka mantle density has been adjusted to obtain the correct total mass of 45g for the hollow golf ball. Comparisons of COR and spin for impact at 10 deg club face loft angle are shown in Table 2. Table 2 - Comparisons of FEA results with experiment (V0=45 m/s; =10°).
contact time (ms) spin (rpm) COR
calculated phenomenological, T
calculated measured, M
experiment
0.345 2766 0.855
0.384 2498 0.878
4217 0.814
18 The Engineering of Sport 7 - Vol. 1 The experimental results were conducted on a hollow ball that was struck by a golf club (Table 2). In our simulation, at an impact speed of 45 m/s the mantle of the hollow golf ball suffered a maximum strain of 0.2. As shown in Fig.1, the mantle material has an elastic modulus that decreases with increasing strain for strains in the range 0<<0.4. Hence, the present simulation has used small strain, linear viscoelastic material properties somewhat outside their range of validity. An elastic modulus that decreases from the value in Table 1 with increasing strain would decrease the calculated COR. On the other hand, a flexible golf club face can increase the COR in comparison with that obtained from impact of a rigid golf club. The simulation results show that a preliminary study of golf ball design and material selection can be conducted by FEA using properties of the constituents measured by independent tests of material properties. The scope of the FEA study can be extended to model dynamic behaviour of a flexible golf club striking a deformable golf ball using measured properties of both constituents.
- Acknowledgement Hollow golf balls compliments of Nanodynamics, Buffalo, NY.
- References Arakawa, K., Mada, D., Komatsu, H., Shimizu, T., Satou, M., Takehara, K. and Etoh, G. (2006). Dynamic contact behaviour of a golf ball during oblique impact. Experimental Mechanics, 46, 691-697. Oyadiji, S. O. (2004). How to analyse the static and dynamic response of viscoelastic components. The International Association for the Engineering Analysis Community, Nafems Ltd, p48-52. Tanaka, K., Sato, F., Oodaira, H., Teranishi, Y., Sato, F. and Ujihashi, S. (2006) Construction of the finite-element models of golf balls and simulations of their collisions. Proc. IMechE Vol.220 Part L: J. Materials: Design and Applications, 13-22. Tavares, G., Sullivan, M. and Nesbitt, D. (2006) Use of finite element analysis in design of multilayer golf balls. Science and Golf III - Proceedings of the World Scientific Congress of Golf, (ed. Farrally, M. R. and Cochran, A. J.), Human Kinetics Publishers, Chap. 59. p473-480. Ujihashi, S. (1994) Measurement of dynamic characteristics of golf balls and identification of their mechanical models. Science and Golf II - Proceedings of the World Scientific Congress of Golf, (ed. Cochran, A. J. and Farrally, M. R.), E&FN Spon, London, Chap. 46., p302-308.
Interaction of Flexor Tendons and Pulleys in Sport Climbing (P6) Andreas Schweizer1, Beat Moor, Hans-Peter Bircher2
Abstract: The most common injury in sport climbing is the disruption of a flexor tendon pulley of the middle or ring finger, mostly the A2 pulley, during the so called crimp grip position, where the proximal interphalangeal joint is flexed about 90 and the distal interphalangeal joint is hyperextended. This study aimed to investigate the frictional properties of the interaction of the flexor tendons and pulleys in vitro. Eight fresh frozen cadaver fingers were used. After dissection of the flexor tendons the fingers were fixated at the proximal phalanx and mounted on a isokinetic movement device (centre of rotation through the proximal interphalangeal joint, kinetic device/resistance at the finger tip, flexor tendons loaded with 20-100N) simulating the crimp grip position. Maximum flexion torque was at 85° of flexion. Strength applied at the flexor tendons was 3 times higher as at the tip of the finger. The highest amount of eccentric-concentric strength deficit of 12% (SD 0.8) was also at 85° of flexion which indicates that there is a substantial amount of friction. Strength deficit showed a rise of up to 28% when the tension of the superficial flexor tendon was increased. This indicates a synergism of both flexor tendons (superficial surrounds profundus tendon) during eccentric work. There was no correlation of friction and speed of motion up to 210°/s when smooth tendon gliding became an interrupted staggering movement. There was a proportional linear increase of friction with increasing load to flexor tendons from 20-100N. An interesting “memory effect” of the tendon pulley interaction was found. Static flexion torque was influenced by the preceding movement and was substantially higher (11%, SD 2.2) after an eccentric movement compared to a concentric movement. The difference did almost not level out even after several minutes which indicates an elastic tendon internal energy saving mechanism or static friction. Keywords: sport climbing, flexor tendon, pulley, crimp grip.
1- Introduction The mechanical stress between finger flexor tendons and the distal edge of the A2 pulley is especially high in rock climbers using the crimp grip (Bollen 1990; Schweizer 2001) where the proximal interphalangeal (PIP) joint is held flexed from 90° to 100° and the 1. Balgrist Clinic, Forchstrasse 340, 8008 Zürich, Switzerland - E-mail: [email protected], 2. Orthopaedic Clinic, Hospital of Zug, 6300 Zug, Switzerland - E-mail: [email protected]
20 The Engineering of Sport 7 - Vol. 1 distal interphalangeal (DIP) joint slightly hyperextended. It is used by up to 90% of rock climbers (Bollen 1988; marco et al. 1998). Accordingly, closed rupture of the A2 and the A4 pulley is one of the most frequent climbing-related inju-ries (Bollen 1990; Gabl et al. 1992). The causative injury mechanism of pulley rupture not only arises from a bowstringing effect but probably also from high friction in near-static high loads (Schweizer et al. 2003). Friction and interaction between flexor tendons and pulleys has been investigated differently in vitro on human and animal cadaver studies An et al. 1993; Uchiyama et al. 1995) and was suggested to be of functional value in a sense that it serves as a support of the finger muscles during eccentric load of the finger joints (Schweizer et al. 2003; Walbeehm et al. 1995) to either improve finger muscle strength or decrease energy consumption. A highly specialised mechanism with high friction between tendons and pulleys has been described in bats and climbing mammals (Schaffer 1905) and was named tendon locking mechanism. The contact areas of their tendons and pulleys are covered with ridges and knobs which engage one another that no active muscular contraction is necessary for them to dangle with the whole body weight on their fingers. As the microstructure of the inner surface of the human pulleys and tendons show much smaller but similar cross running ridges as in bats an analogous mechanism in the human tendon pulley interaction was postulated (Walbeehm 1995). Besides that the flexor tendons themselves are known to be able to store elastic energy (Alexander 1984; Ker et al. 1986) and to serve as a natural spring which was observed in different animals like horses, deers and donkeys (Alexander 1984; Bennett et al. 1986; Riemersa and Schamhardt 1985) as well as in humans (Maganaris and Paul 2002). To quantify friction between human flexor tendons and the A2 pulley following in vitro investigation with cadaver fingers was performed. A similar method described by Schweizer (2003) using an isokinetic device was used to determine friction measuring the difference of eccentric and concentric maximum strength at the fingertip (Fig. 1). In this study the influence of different load (test 1) to the tendon, speed of movement (test 2), static and dynamic flexion torque (test 3) according to a preceding eccentric or concentric movement and different tensions of the superficial flexor tendon (test 4) on friction was determined.
Figure 1 - Quantification of force of friction between the A2 pulley and the flexor tendons (left) in the crimp grip like position (right) of the finger during rock climbing: FR= (FE ecc r2 - FE con r2)/2 r1. Force (FE) at the fingertip during eccentric (ecc) and concentric (con) movement. Friction (FR) between flexor tendons (gray) and A2 pulley, simulated muscular force at the flexor tendons (FsM).
The Engineering of Sport 7 - Vol. 1
21
2- Material and Methods Eight long fingers were obtained from two fresh-frozen cadaver hands of two male adults, performing a transmetacarpal ray resection. The entire finger including the metacarpophalangeal (MP) joint was left untouched to preserve the physiological milieu surrounding the flexor tendons and flexor sheath. The tendons were cut distal to the musculotendinous junction of the flexor digitorum superficialis (FDS) and profundus (FDP) muscle. Each tendon was armed separately with a conventional braided polyester filament (Ethibond Excel 2) using a Krakow suture technique. Two 2.7 mm Schanz screws were inserted dorsally into the proximal phalanx which allowed a rigid fixation of the specimen to the testing device. A previously described testing apparatus (Schweizer et al. 2003) which was modified for this cadaver study was used (Fig. 2). It consisted of two box-like modules which could be coupled to each other. The first module, the so called gear box contained a Zuerrer TFVB9-55/2 three-phase electric cage motor (630 Watt), coupled to a worm gear (Zuerrer 2/1MH) with an output torque of at least 30 Nm. The rotational speed was controlled via a frequency converter (Hardmeier control VF61M R722).
Figure 2 - Testing apparatus (left) with the device simulating PIP joint flexion (1), fixator to finger, force transducer (3), frequency converter (4), engine (5) and force transmission. Depiction of a finger fixated in the device (right).
The specimen was anchored at the proximal phalanx through the external fixator system. A device was incorporated which simulated the PIP joint flexion movement and was powered by means of the gear box. A piezoelectric force transducer (Kistler 9301A SN488642) was positioned at finger tip of the test digit. The signal was reinforced by a charge amplifier (Kistler Type 5011) and stored by a personal computer supported storage oscilloscope (Voltcraft DSO 2100). The eccentric and concentric flexion movements in the PIP joint were produced ranging from 10 to 105° flexion. The position of flexion was recorded with a custom made electronic torsion angle measurement device. To simulate active muscular force acting on the flexor tendons (FDP and FDS) they were
22 The Engineering of Sport 7 - Vol. 1 loaded together. This resulted in a PIP joint flexion working against the external force generated from the isokinetic movement device. Because of a minimal difference of excursion of the FDS and FDP tendons both were connected by a pulley. The load was applied then at the pulley to ensure that both tendons were loaded equally. The DIP joint remained always in extension (crimp grip like position) and therefore there was no movement between FDP tendon and the A4 pulley to consider (crimp grip position like). Following tests were performed: Test 1: To investigate possible load dependency on friction, the tendon force (FsM) was investigated at 20, 60, 80 and 100 N, in addition to the default 40N load, all at a rotational speed of 30°/s. Test 2: To evaluate if the friction force has a dependency on flexion velocity, the rotational speed was increased to 60, 90, 120, 150 and finally 210 degrees per second, all at a load of the flexor tendons of 40 N. Test 3: Static and dynamic torque during eccentric and concentric movement was recorded stepwise in steps of approximately 10° increments after a preceding eccentric and a preceding concentric movement was recorded. Static measurements of flexion torque at about 85° of flexion during 6 min. were recorded. This test was performed with 6 specimens. Test 4: To determine the influence of different tensions of the FDS tendon on friction (between FDP and FDS and A2 pulley) the FDP tendon was detached from its insertion at the distal phalanx and mounted with linear traction at the movement device (90° flexion to the proximal phalanx) with the transducer intercalated. The PIP joint was fixed at 90° of flexion with a Schanz screw so that different static tension of 20, 40, 60, 80 and 100 N could be applied at the proximal end of the FDS tendon. This test was performed preliminary with only one specimen.
3- Results Test 1: The average maximum moment of all fingers was 0.58±0.04 Nm (mean±SD) for eccentric and 0.50±0.04 Nm for concentric flexion movement. The mean difference between eccentric and concentric force was 12.9±0.8%, corresponding to a calculated friction force of 4.98±0.56 N. There was no statistically significant change (Fig. 3) in the force deficit upon different tendon load magnitudes (the coefficient of correlation was 0.189, p=0.085). Test 2: The rise in rotational speed appeared not to have any influence (Fig. 3) on the friction force (coefficient of correlation -0.036, p=0.739). At rotational velocities above 180 degrees per second there was a transition of the smooth gliding to a jerky tendon movement.
The Engineering of Sport 7 - Vol. 1
23
Figure 3 - Eccentric to concentric force deficit during different loads at the flexor tendons (left) and different rotational speed (right) with 30° (2), 60° (4), 90° (6), 120° (8), 150° (10) and 210° (12). There was a linear increase of force deficit (remaining percentage of force deficit) in both tests.
Test 3: The maximum post-concentric static torque was 0.587 Nm (SD 0.163) and the maximum post-eccentric static torque was 0.658 Nm (SD.173). The maximum posteccentric static torque was therefore 11% (SD 2.2) greater than the post-concentric static torque. In the static situation the flexion force acting at the A2 pulley flexor tendon interface was 9.52 N (SD 2.26) greater during the post-eccentric static torque compared to the post-concentric static torque (Fig. 4). The post-eccentric static torque was greater than the post-concentric static torque over the whole range of motion with a peak at about 83° of Flexion of the PIP joint. There was a decrease of the difference between post-eccentric static torque compared to the post-concentric static torque during the initial 6 minutes of 19% (SD 2.2). Test 4: Eccentric-concentric strength deficit increased from 19% with 20 N traction on the FDS tendon to 28% with 100 N tension on the FDS tendon.
4- Discussion Friction between pulleys and flexor tendons has been differently investigated mostly to assess friction after trauma (Lane et al. 1976) and friction of different suture techniques and tendon grafts (Coert et al. 1995; Nishida et al. 1998; Peterson et al. 1986; Uchiyama et al. 1995; Williams and Amis 1995; Woo et al. 1981). Therefore, various indirect measurement techniques and models (Fowler and Nicol 2000) and measurements in animals have been described (Lane et al. 1976). Uchiyama (1995) and An (1993) developed a method to measure friction in vitro only between the human A2 pulley and suggested that friction was significantly higher than in diarthrodial joints. This indicates that friction is apparent during tendon and tendon sheath interaction. In a previous in vivo study (Schweizer et al. 2003) friction was tried to be calculated indirectly and to be responsible for 9.1% (SD 2.6) of the maximum possible eccentric flexion force in the PIP joint. Coefficient of friction (quasi static) was estimated to be 0.075 (+/- 0.021). The exact muscular strength of the finger flexors was not known and was estimated by the measurement of the strength of the wrist flexors.
24 The Engineering of Sport 7 - Vol. 1
Figure 4 - Single measurement of a concentric and eccentric phase (left) and the summary of the joint torques of static and dynamic measurements (right) according to PIP joint flexion.
In the present study cadaver finger were used to estimate friction between finger flexor tendons and pulleys in static and dynamic situations as well as the influence of different loads and speed on the behaviour of friction. We found that highest amount of eccentric-concentric strength deficit of 12% (SD 0.8) was also at 85° of flexion of the PIP joint which indicates that there is a substantial amount of friction of 17 N during eccentric and 10 N during concentric movement with 40 N pull at the flexor tendons. Strength deficit showed a rise of up to 28% when the tension of the superficial flexor tendon was increased to 100 N. This indicates a synergism of both flexor tendons (superficial surrounds profundus tendon) during eccentric work which increase the maximum eccentric strength substantially by pure friction. This mechanism particularly favours to sustain high loads during eccentric or near eccentric static movements of the fingers. There was a linear proportional increase of fric-tion with speed of motion (Fig. 3) up to 240°/s. Thereafter smooth tendon gliding became an interrupted staggering move-ment and could not be evaluated further. There was also a linear proportional increase (Fig.3) of friction with increasing load to flexor tendons from 20-100N. An interesting “memory effect” of the tendon pulley interaction was found. Static flexion torque was influenced by the preceding movement and was substantially higher (11%, SD 2.2) after an eccentric movement compared to a concentric movement. The difference of the post eccentric to the post concentric torque may be regarded as the pure static friction between pulleys and flexor tendons. The difference did almost not level out (19% decrease) even after sev-eral minutes and may be regarded as an elastic tendon internal energy saving mechanism with viscoelastic similar behaviour. Bats, some climbing mammalians and birds display a so called tendon locking mechanism (Schutt 1993). Bats for instance may dangle on their digits without muscular contraction. The mechanism sustains flexion by interlocking the flexor tendon with the corresponding pulleys. They may hang all night long, during hibernation or still after they have died. Schaffer (Schaffer 1905) was the first to describe a “SperrHemmvorrichtung” in bats where the flexor tendon interacts and locks with the fibrous tendon sheath. Schutt (1993) investigated the mechanism more precisely in bats and
The Engineering of Sport 7 - Vol. 1
25
Quinn and Baumel (1993) finally called it TLM (Tendon locking mechanism) and compared it in different bat species (chiropterans). According to the findings in this study, a similar but far less distinct mechanism is present in the human pulley – flexor tendon interaction. Walbeehm and McGrouther (1995) compared the TLM with the anatomy and function of the human flexor tendon sheath and described a tendon compressing mechanism (TCM) where the FDP tendon is compressed circularly by the chiasma of the FDS tendon and the A2 pulley which is supported by the results of the present findings. Using SEM they showed transverse ridges on the inner surface of the A2 pulley and on the palmar surface of the FDP tendon which favours an interdigitation of the ten-dons and pulleys rather than encourage gliding notably during extension. The A2 pulley and the A4 pulley are stressed during rock climbing in a way that tenosynovitis and disruption of the pulleys are one of the most common injuries in this sport (Bollen 1990; Cartier et al. 1985; Tropet et al 1990). Foremost the so called crimp grip position where the PIP joint is flexed 90° or more and the DIP joint is hyperextended results in a distinct bow-stringing (Schweizer 2001) and stresses the distal edge of the A2 pulley mostly where disruption most commonly occurs. Zhao et al. (2000) showed that gliding resistance of the flexor tendon increased as the PIP joint angle increased. The presence of friction between tendons and pulleys during flexion movement (Schweizer et al. 2003; Uchiyama et al. 1995) and the fact that also in static situations after an eccentric movement of the PIP joint higher torque is generated further support the hypothesis of high friction in the crimp grip position or during a flexed PIP joint. Friction and a TLM similar mechanism may be an advantage for rock climbers in a way that it increases the holding force of their flexor muscles. According to the present results a short initial eccentric movement of the PIP joints rather than a concentric one before holding a grip increases the maximum strength of the holding force considerably and may be advised as technical improvement to rock climbers. The PIP joint then is best positioned in 80 to 85° a flexion where the TLM analogue is most distinct. On the other hand the TLM analogue may also increase the susceptibility to pulley injuries. When the climber performs a dynamic move or if the foot abruptly comes off the rock, the fingers experience a sudden rise of load which may lead to injury of the flexor tendon pulleys. This is probably not only due to the bowstringing effect but also by an increased adhesion between flexor tendons and pulleys in a sense of friction as a physiologic function.
References Alexander RM. Elastic energy stores in running vertebrates. Am Zool, 24: 85-94, 1984 An KN, Berglund L, Uchiyama S, Coert JH. Measurement of friction between pulley and flexor tendon. Biomed Sci Instrum, 29: 1-7, 1993 Bennett MB, Ker RF, Dimery NY, Alexander RM. Mechanical properties of various mammalian tendons. J Zool, 209: 537-48, 1986 Bollen SR. Injury to the A2 pulley in rock climbers. J Hand Surg [Br], 15(2): 268-70, 1990 Bollen SR. Soft tissue injury in extreme rock climbers. Br J Sports Med, 22(4): 145-7, 1988 Cartier J-L, Toussaint B, Darlot P, Herry J-P, Allieu Y, Bousquet G. Approche d'une nouvelle pathologie de la main liée à la praqtique de l'escalade. Journal of Traumatol Sport, 2: 35-9, 1985
26 The Engineering of Sport 7 - Vol. 1 Coert JH, Uchiyama S, Amadio PC, Berglund LJ, An KN. Flexor tendon-pulley interaction after tendon repair. A biomechani-cal study. J Hand Surg [Br], 20(5): 573-7, 1995 Fowler NK, Nicol AC. Interphalangeal joint and tendon forces: normal model and biomechanical consequences of surgical reconstruction. J Biomech, 33(9): 1055-62, 2000 Gabl M, Lener M, Pechlaner S, Judmaier W. [Isolated injuries of the pulley of the finger flexor tendon sheath--injuries in ex-treme climbing sports]. Sportverletz Sportschaden, 6(3): 119-22, 1992 Ker RF, Dimery NY, Alexander RM. The role of tendon elasticity in hopping in a wallaby. J Zool, 208: 417-28, 1986 Lane JM, Black J, Bora FW, Jr. Gliding function following flexor-tendon injury. A biomechanical study of rat tendon function. J Bone Joint Surg Am, 58(7): 985-90, 1976 Maganaris CN, Paul JP. Tensile properties of the in vivo human gastrocnemius tendon. J Biomech, 35(12): 1639-46, 2002 Marco RA, Sharkey NA, Smith TS, Zissimos AG. Pathomechanics of closed rupture of the flexor tendon pulleys in rock climbers. J Bone Joint Surg Am, 80(7): 1012-9, 1998 Nishida J, Amadio PC, Bettinger PC, An KN. Excursion properties of tendon graft sources: interaction between tendon and A2 pulley. J Hand Surg [Am], 23(2): 274-8, 1998 Peterson WW, Manske PR, Kain CC, Lesker PA. Effect of flexor sheath integrity on tendon gliding: a biomechanical and histologic study. J Orthop Res, 4(4): 458-65, 1986 Quinn TH, Baumel JJ. Chiropteran tendon locking mechanism. J Morphol, 216(2): 197-208, 1993 Riemersa DJ, Schamhardt HC. In vitro mechanical properties of equine tendons in relation to cross-sectional area and collagen content. Res Vet Sci, 39: 263-70, 1985 Schaffer J. Anatomisch histologische Untersuchung über den Bau der Zehen bei Fledermäusen und einigen kletternden Säuge-tieren. Zeitschrift für wissenschaftliche Zoologie, 83: 231-84, 1905 Schutt WA, Jr. Digital morphology in the Chiroptera: the passive digital lock. Acta Anat (Basel), 148(4): 219-27, 1993 Schweizer A. Biomechanical properties of the crimp grip position in rock climbers. J Biomech, 34(2): 217-23, 2001 Schweizer A, Frank O, Ochsner PE, Jacob HA. Friction between human finger flexor tendons and pulleys at high loads. J Biomech, 36(1): 63-71, 2003 Tropet Y, Menez D, Balmat P, Pem R, Vichard P. Closed traumatic rupture of the ring finger flexor tendon pulley. J Hand Surg [Am], 15(5): 745-7, 1990 Uchiyama S, Coert JH, Berglund L, Amadio PC, An KN. Method for the measurement of friction between tendon and pulley. J Orthop Res, 13(1): 83-9, 1995 Walbeehm ET, McGrouther DA. An anatomical study of the mechanical interactions of flexor digitorum superficialis and profundus and the flexor tendon sheath in zone 2. J Hand Surg [Br], 20(3): 269-80, 1995 Williams RJ, Amis AA. A new type of flexor tendon repair. Biomechanical evaluation by cyclic loading, ultimate strength and assessment of pulley friction in vitro. J Hand Surg [Br], 20(5): 578-83, 1995 Woo SL, Gelberman RH, Cobb NG, Amiel D, Lothringer K, Akeson WH. The importance of controlled passive mobilization on flexor tendon healing. A biomechanical study. Acta Orthop Scand, 52(6): 615-22, 1981 Zhao CF, Amadio PC, Berglund L, An KN. The A3 pulley. J Hand Surg [Am], 25(2): 270-6, 2000
Friction between Players’ Hands and Sports Equipment (P7) S E Tomlinson1, R Lewis2, M J Carré3
Topics: Rugby balls, Friction. Abstract: This paper tests the friction between a popular design of rugby ball contacting the palm and three different designs of gloves. The gloves used were from the same manufacturer; one was made from synthetic leather, one had a round, pronounced silicon pattern of pimples and the final one had a silicon pattern with oval shapes cut into it. The friction of the contact was measured in dry conditions, with a small amount of water added to the palm/glove, a larger amount of water added to the palm/glove and finally a dry palm/glove contacting wet material. These tests showed the silicon designs of gloves gave a consistent measurement of coefficient of friction across all moisture levels. However, for the bare palm and the synthetic leather glove the coefficient of friction increased with a small amount of water and then decreased when more water was added. The synthetic leather glove was shown to have the highest coefficient of friction across all levels of water, with the silicon oval patterned glove having the lowest coefficient of friction in the majority of cases. Keywords: Rugby, Friction, Skin, Grip.
1- Introduction The ease of handling a rugby ball is extremely important to a player and is one of the central areas of research for rugby ball manufacturers. Balls can be difficult to handle “in the wet or when you play night-time rugby and there is dew on the ball, but also on very dry days when your hands are very sweaty, which can be a lot worse.” (a quote from Jonny Wilkinson (Palmer, 2007)). Moisture is therefore a major issue on both dry and wet days and the ball needs to perform well in these circumstances. Manufacturers work hard to achieve this through the surface pattern and material of the ball, and some players wear gloves to try and aid grip. 1. Department of Mechanical Engineering, University of Sheffield, Mappin Street, Sheffield, S1 3JD E-mail: [email protected] 2. E-mail: [email protected] 3. E-mail: [email protected]
28 The Engineering of Sport 7 - Vol. 1 The International Rugby Board (IRB), only states “The ball should be manufactured from leather or suitable synthetic material. It may be treated to make it water resistant and easier to grip” (International Rugby Board, 2006a). This means there are a number of different ball surface materials and patterns available on the market, some of which are designed specifically with the aim of increased grip. From previous studies it is known that increasing the moisture content of the skin by a small amount will increase the coefficient of friction, when considering water applied to the surface of the skin; this is due to liquid bridging (Dinç et al., 1991; Tomlinson et al., 2007a). The formation of the liquid bridges increases the amount of viscous shear, which in turn increases the friction of the contact. This increase continues to a point, past which, the coefficient of friction will start to decrease due to the water starting to separate the surfaces, making it easier for them to move over each other. Bobjer (Bobjer et al., 1993) carried out work looking at the effect of water added to ridged polycarbonate specimens. This work showed that when water was added to the ridged material, there was more of an increase for the wider ridged surfaces than for the narrow ridged surfaces. It was also shown that for non-textured surfaces the coefficient of friction decreased. Therefore, if pimples behave in a similar way to ridges the coefficient of friction should increase when water is added, probably up to a point (as found by Dinç et al., 1991). However, the difference in this case is that not only are they pimples not ridges, the rugby ball material is viscoelastic. Some players have started to use gloves to improve the grip between their hands and the ball. The IRB regulations on gloves state that “Coverage of the fingers and thumbs is permitted to the outer joint but no further. The mitt zone of coverage should not continue beyond the wrist. The body of the mitt should be of a stretch type material with the grip material being made of a soft rubber/synthetic compound not exceeding a depth of 1mm. No part of a mitt should contain buttons or potentially dangerous items.” (International Rugby board, 2006b). This again gives rise to the possibility of numerous designs of gloves; however there are 2 main designs on the market; silicon patterned and synthetic leather, however there are many ways in which the silicon pattern can be printed on the glove. The initial testing done in this study was to examine the contact of the ball when catching and throwing a ball to ascertain the amount of contact, if any, with the palm and the ball. This was done to highlight the significance of knowing the friction between the rugby ball material and the gloves/palm. Once contact with the ball was established, friction tests were carried out. These tests were done using a finger friction rig designed at the University of Sheffield (Lewis et al., 2007; Tomlinson et al., 2007b), the use of which was modified to test the skin of the palm. The study examined the contact of three types of gloves and the bare palm contacting a popular design of rugby ball, with varying amounts of water introduced to the contact.
The Engineering of Sport 7 - Vol. 1
29
2- Testing 2.1 Materials tested The ball material tested is shown in Figure 1. This material was chosen because it is an established design, which although is now no longer used at world cup standard, it is still a popular design for training and also with amateur teams. It is also the design seen replicated by many other manufacturers.
Density of pimples = 0.58 Mppm2 Figure 1 - Rugby ball material used in tests. Mppm2 refers to millions of pimples per m2.
Figure 2 shows three examples of the gloves, used in this study, with a schematic (not to scale) showing the dimensions of the design. Two different silicon designs were tested and one synthetic leather design.
Synthetic leather
Silicon oval pattern (white = silicon)
Silicon round pimple pattern (white =silicon)
Figure 2 - Gloves used in tests
2.2 High Speed video footage of handling situations High speed video tests were carried out in two separate sessions, with all filming being done indoors. Six university rugby players, who train and play regularly, aged 21-23 and of the field positions prop, second row, back row, fly half and centre, were asked to throw and catch the ball with good aim, over a distance of 10m. The speed of pass was not specified. High speed video recordings were then used to examine both the passes and
30 The Engineering of Sport 7 - Vol. 1 catches. These tests were done using the ball shown in Figure 1. This ball was tested both wet and dry. The ball was made wet by dunking it in a bucket of water and then shaking off the excess before handing it to the player; in both the wet and dry instances the players’ hands were dry. The main catch seen, involved contact with the palm and was observed in both wet and dry conditions. In this catching mechanism the top of the palm (part adjoining the fingers) was the first point of contact with the ball. This contact with the palm slows the ball down, and then the fingers waver over the top of the ball until choosing a fixed position to clamp down and secure the ball. Figure 3 shows the first instance of contact with the ball and the final position of the hands, once the ball is secured. This video footage shows that the tests of friction between glove and ball are relevant, because in this catching mechanism the top of the palm plays a large role in controlling the ball. The second type of catch, seen only on one occasion, was where the fingertips contacted the ball first. The ball then slowed down smoothly until the palm was in contact with the ball. Once in contact with the palm the ball was secured in the hands of the player. Although, the palm contacts the ball, this is not until the ball has been controlled, so the friction of the ball – palm/glove contact is less important. It is also not important for throwing the ball because there is no contact with the palm; the whole throwing action involves only the fingertips.
Palm makes the initial contact with the ball
Final position: Fingertips and palm in contact with the ball
Figure 3 - First and last positioning of hands in main recorded catching mechanism.
2.3 Method of friction tests The friction tests were carried out in lab conditions using a skin friction rig. The rig is shown in Figure 4. It consists of two load cells, one measuring the normal force and one the friction force. The rugby ball specimens are attached to the upper metal plate, using double-sided adhesive tape. The top of the palm is then moved from one end of the specimen to the other, in a direction away from the body (to simulate the contact when catching a ball), and the voltages of each load cell recorded. This voltage is then correlated to the correct force, using a calibration factor. The tests were carried out on one person’s palm and then on three different types of gloves. Four different tests were carried out; dry ball material against dry palm/glove, dry ball material against ‘slightly wet’ palm/gloves, dry ball against ‘wet’ palm/glove and wet ball material against dry palm/glove. In the ‘slightly wet’ case the water was applied to
The Engineering of Sport 7 - Vol. 1
31
the skin/glove using a wet paper towel and in the ‘wet’ condition the water was sprayed on to the palm/glove using a water spray held 15 cm away. The ball material was wet using the same spray, however four sprays were used across the surface. The moisture level of the palm was monitored using a Moistsense device (Moritex Europe). This device measures the moisture using a capacitance measurement; it then transposes it to a scale from 1- 99 based on measurements of 300 people. In these tests the average reading for the dry tests was 52 and the average reading for the partially wet test was 89, the wet condition was too moist for the monitor to read it. The device is also not capable of measuring the moisture of the gloves, so this was not monitored, but the same procedure was followed so it is expected to be a similar increase in moisture. In the case of the dry tests and all the skin tests the procedure was repeated five times, in all other cases they were carried out once.
Figure 4 – Skin friction rig.
2.4 Friction test results Figure 5, shows a summary of the results from this test. The ‘slightly wet’ condition refers to where the palm/glove was wet with a paper towel, the ‘wet’ condition is where one spray of water was applied to the hand/glove and ‘wet material’ is where the hand was dry and the material was wet using 4 sprays of water, as described in the method. These results provide both a comparison of the glove performance, to the bare palm and other glove designs, in different moisture conditions and also provides picture of how consistently a single glove, or the bare palm, performs across each moisture condition. When comparing the results of each glove it shows that in all cases the synthetic leather glove has the highest coefficient of friction. They also show that generally the oval, silicon patterned glove had the lowest coefficient of friction. In terms of comparing the performance of the gloves to the bare hand, in dry or slightly wet conditions the hand has a higher coefficient of friction than both the silicon designs of glove. However, in the wet conditions, both wet hand and wet material, the round silicon patterned glove performed better than the bare palm. The oval, silicon patterned glove had a slightly higher coefficient of friction for the wet hand condition than the bare palm, but
32 The Engineering of Sport 7 - Vol. 1
Figure 5 – Average friction coefficient for all hands/gloves in each moisture condition.
performed worse for the wet material condition. The synthetic leather glove was better than the bare palm across all conditions. The silicon patterned gloves are the most consistent across the different levels of moisture; showing only a small variation in coefficient of friction between each moisture condition. More variation was seen for the bare palm and the synthetic leather glove. Both of these display a small increase in coefficient of friction when a small amount of water is added to the hand/glove, consistent with the findings of Bobjer (Bobjer et al., 1993). Then with more water added (the wet hand case), the coefficient of friction significantly reduces, more for the wet hand than the synthetic leather. The wet material coefficient of friction is only slightly higher than the wet glove coefficient of friction for the synthetic leather glove, again this is the case for the bare palm, but the difference in coefficients of friction is more significant.
3- Discussion In all cases the synthetic leather glove had the highest coefficient of friction. The reason for this is probably two-fold; the leather probably has a higher coefficient of friction when tested against rugby ball material than silicon, secondly the surface is flat, so there is a larger area of contact with the ball, therefore more adhesion is possible. The synthetic leather has a higher coefficient of friction than the hand; this is probably due to the leather having a higher coefficient of friction than skin. There is no data on the coefficient of friction for the materials with out surface patterns, so it is difficult to know whether the material or the texture is having a larger effect, except for the case of the two silicon designs, which can be compared.
The Engineering of Sport 7 - Vol. 1
33
The silicon designs generally had the lowest coefficient of friction. From comparison of the results of the two designs it can be assumed that silicon on its own against rugby ball material will probably have a lower coefficient of friction than the human skin against rugby ball material. However, the texture can be useful for improving grip in wet conditions. In wet conditions the material the silicon pattern is mounted on can absorb the moisture, so there is less water in the actual contact (as this part of the glove will not contact in most places), which also explains why the silicon designs are the most consistent across all moisture conditions. This can happen to a lesser extent with the oval pattern, which is why it is more consistent across all conditions than the round pimple pattern. In the case of the wet ball the round pimple patterned glove performs well compared to the hand and oval patterned glove, this is thought to be due to its ability to cut through the water, so it is less effected by the presence of the water than the other silicon glove and bare palm. This effect cannot occur with the oval pattern, due to it being a bulk pattern. The coefficient of friction for the round pimpled ball in all cases is higher than that of the oval pattern because the round pimples mean there is a larger pressure in the contact, so more adhesion possible, the pimples may also get caught in the pattern on the ball, further increasing the friction. Whereas the oval pattern is a large flat pattern, so it is more like a smooth sheet, with a lower coefficient of friction than the other materials, sliding across the ball. There is variation between different moisture conditions for both the synthetic leather glove and the hand. In the case of the leather glove the initial increase in coefficient of friction is most probably due the material absorbing the moisture. This causes the material to become more supple, so more adhesion can occur. When more water is added the water starts to separate the surface, as is also the case when the material is wet and the glove is dry, so the coefficient of friction is reduced. In the case of the hand the increase in coefficient of friction is due to liquid bridging (Dinç et al., 1991), these liquid bridges increase the friction force due to an increase in viscous shear forces. The amount of water that can be added before the coefficient of friction starts to decrease has a maximum limit, after this the coefficient of friction will decrease due to the water starting to separate the two surfaces, as seen with the wet hand and wet material conditions in these tests. These tests only give a view of the performance of the gloves and hand contacting one type of rugby ball material, the performance of each glove/hand could vary with different materials and textures, so this should be investigated further. It also does not look at the effects of any of the other contaminants found in a game of rugby, (other than water), which will also affect the grip on the ball, so should be investigated further. The results gained in the glove wet conditions are limited as only one test run was carried out, these could be repeated to check consistency. Finally, conclusions about whether it is better to wear a glove or not can not be based solely on these results as they do not take into account the fact that the hands are warmed by the glove which could increase the dexterity of the fingers, this is a factor to be investigated to gain a full picture of whether or not gloves improve grip.
34 The Engineering of Sport 7 - Vol. 1
4- Conclusions When recording the processes of throwing and catching a rugby ball, the main mechanism of catching a ball showed that the contact between the palm and ball plays a large role in decelerating and controlling the ball, therefore the friction between the skin of the palm/glove and the ball is an important factor to optimise. These results have shown that with the ball tested there are advantages of using the synthetic leather glove in all situations and the round pimpled silicon glove also had a higher coefficient of friction than the hand in the two wet conditions. The silicon glove with the oval pattern performed worse than the bare hands in all conditions, other than the condition where the hand was sprayed with water, in this instance it had a slightly higher coefficient of friction than the bare hand, but this difference was not significant. The variation in the coefficient of friction with the addition of water was seen most with the synthetic leather glove and the bare hand. There was little variation for the silicon patterned gloves because the material the pattern was mounted on, which does not contact the ball, absorbed most of the water. A small amount of water increased the coefficient of friction for the synthetic leather glove and the bare hand, however more water added to either the hand or the ball decreased the coefficient of friction. There are several ways in which this study should be continued. It should be furthered to look at more balls, more tests in wet conditions should also be done to check for consistency, other contaminants could be looked at, and before making decisions about whether or not players should wear gloves, the effect the glove has on keeping the hand warm and the possible increased dexterity from this should be investigated.
5- References [BJ1] Bobjer, O., Johansson, S. E., and Piguet, S. Friction between hand and handle. Effects of oil and lard on textured and non textured surfaces; perception of discomfort. In Applied Ergonomics 24: 190-202, 1993 [DE1] Dinç, O.S., Ettles, C.M., Calabrese, S.J., and Scarton, H.A. Some parameters affecting tactile friction. In Journal of Tribology 113: 512-517, 1991 [IRB1] International Rugby board. Law 2 The ball, Playing Charter. International Rugby board, 2006 [IRB2] International Rugby board. Regulations relating to the game, 2006 [LM1] Lewis, R., Menardi, C., Yoxall, A., and Langley, J. Finger Friction: grip and opening packaging. In Wear 263: 1124-1132, 2007 [P1] Palmer, B. Wilkinson frustrated by cup balls. In BBC Sport, 2007 [TL1] Tomlinson, S.E., Lewis, R., and Carre, M.J. Review of the frictional properties of fingerobject contact when gripping. In Proceedings of the IMechE Part J, Journal of Engineering Tribology 221: 841-850, 2007 [TL2] Tomlinson, S.E., Lewis, R., and Carré, M.J. Improving the understanding of grip. In The Impact of Technology on Sport II. Fuss, F.K., Subic, A. and Ujihashi, S. (eds.): Taylor & Francis Group, London, 129-134, 2007
Development of a Comfort Model for Cricket Leg Guards (P9) James Webster1, Jonathan Roberts, Roy Jones
Topics: Safety, Lawn sports (Hockey, Cricket). Abstract: Within sport, personal protective equipment (PPE) is becoming increasingly important due to the intense schedules of sports performers and the cost of injury. Traditional PPE often restricts the movement of the user and is uncomfortable to wear because it is usually cumbersome, heavy, and ill-fitting. As a result, wearing PPE has been found to adversely effect both cognitive and physical performance, as well as having a detrimental affect on players’ well being (Fox et al., 1966, Adams and Keyserling, 1996). Comfort and fit are, therefore, considered important factors in the design of PPE. The majority of research within PPE has focused on the prevention of injury through mechanical impact tests, neglecting psychological factors such as comfort. Using cricket leg guards as an example, the aim of this study is to develop a better understanding of factors influencing players’ perceptions of comfort and performance so that, in future, products can be developed that are more suitable for the user. In this study, three methods were used to elicit players’ perceptions of cricket pads; these were co-discovery, focus groups following practical testing, and interviews. Twenty one cricketers currently playing at county first or second team level, participated in testing six different pads, covering a range of styles. Once the testing was complete the data was transcribed and the specific characteristics of PPE which influence and determine comfort emerged through using an inductive analysis. This analysis procedure allows the characteristics of importance to the player emerge from the data rather than from previous literature, which is beneficial in exploratory research. From the analysis six main themes were identified - protection, aesthetics, sensorial comfort, thermal comfort, weight and fit. The knowledge gained can be used in this, and future, projects to develop better product design specifications and to evaluate PPE. Keywords: Sport; Inductive analysis; Perception; Cricket.
1. Loughborough University, Sports Technology Institute, Loughborough Science & Enterprise Park, 1 Oakwood Drive, Loughborough, LE11 3QF - E-mail: {J.M.F.Webster, J.R.Roberts, R.Jones}@lboro.ac.uk
36 The Engineering of Sport 7 - Vol. 1
1- Introduction Modern day sport is becoming faster and more intensive due to improved training and professionalism of athletes. This transformation within sport has resulted in an increased risk of injury, costing the industry, individual and community an estimated ten billion euros per annum within the European Union alone (Stevenson et al., 2003). Consequently numerous sports governing bodies have made the use of personal protective equipment (PPE) mandatory to limit the risk of injury. Traditional designs of PPE however have been bulky, cumbersome and ill-fitting, resulting in decreased range of motion and inhibited performance, causing players to discard equipment due to discomfort (Akbar-Khanzadeh et al., 1995). The majority of research within PPE has focused on the prevention of injury through mechanical impact tests, neglecting psychological factors such as comfort. Recently however, it has been shown that it is not just a luxury for the user to feel comfortable, but a key safety feature (Stull, 2000, Zimmerli, 1998) and can have a positive affect on both cognitive and physical performance (Fox et al., 1966, Adams and Keyserling, 1996). Although comfort has been studied with more frequency over the past thirty years, the majority of research has been based within industrial settings, utilising human responses to comfort. This form of investigation has resulted in a lack of theoretical models, causing the comfort construct to be plagued with undefined, idiosyncratic terminology, generating confusion about attributes relevant to comfort (Civille and Dus 1990). Existing models within the comfort construct have stemmed from Fourt and Hollies (1970) “Comfort Triad”, where comfort was seen to be determined by the interaction of three components, the individual, their clothing, and the environment they are in. The conceptualisation of comfort into this triad provided quantities and units for the description of each of the components. Physical variables were identified for the clothing and environmental factors and physiological variables for the individual (Branson, Sweeney 1991). This delineation of the variables in the components was seen as the greatest contribution to the understanding of comfort using the Fourt and Hollies model. This model was then refined through three further developmental stages, these were Pontrelli’s (1977) “Comfort’s Gestalt”, Sontag’s (1986) “Human ecological approach” and finally Branson and Sweeney’s (1991) comfort model. Each developmental stage incorporated but further developed the previous model furthering understanding of how perceptions of comfort are formulated; however, the appropriateness of one model representing all clothing items and all situations is questionable, due to the inherent differences of each sport, and the way in which the individual and piece of equipment interact. All situations will differ with regards to the features that influence perceived comfort and the relative contribution of each component, which current comfort models appear to neglect. End user involvement in the design process is becoming more salient, leading to better designed products from the users perspective. However, within comfort, techniques utilised have focused on symptoms rather than causes of discomfort, limiting their impact on product development. Therefore the aims of this paper were to identify
The Engineering of Sport 7 - Vol. 1
37
characteristics of importance to the player when determining comfort, and to evaluate the pad characteristics that influence comfort.
2- Method Twenty one participants took part in the testing with a mean age of 19.9 years (±1.7), playing at either county first or second team level or an equivalent standard. Initial data collection was through co-discovery and focus groups with a practical testing session inbetween to help elicit responses. Individual interviews were then conducted, completing a triangulation of methods, increasing the reliability and validity of the test results. These three qualitative methods were utilised to achieve a balance between the strengths and weaknesses of each technique, as demonstrated in Table 1. The validation process of individual interviews was used to ensure data saturation had been achieved and to compensate for any group effects present within the results from the co-discovery and focus group sessions. Table 1 - Characteristics of three data collection methods.
Opportunities for probing Investigator influence Group effect Freedom of participant
Co-discovery
Focus group
Individual interviews
None Limited Extensive Extensive
Extensive Extensive Extensive Average
Extensive Extensive None Limited
At the beginning of each test, a group of six players participated in a co-discovery session. The players were asked to evaluate six pairs of pads, which were chosen as a representative sample of the market, by discussing the positive and negative attributes of the pads and the features they would change to increase comfort. Following the co-discovery session, all players participated in a practical test using the different pads, this was done to stimulate thoughts increasing quality of response. Each participant used two different types of pad (allocated at random) for 5-10 minutes, facing a minimum of 10 balls wearing each pad. The players were encouraged to run between the wickets when they felt they had hit a scoring shot and to pad balls away to allow for a complete assessment of the pad’s performance. No control pads were used to reduce kinaesthetic after-effects preventing fixations on certain stimuli and augmentation/reduction of sensations (Ashdown and DeLong 1995). The participants then participated in a focus group, where a naturalistic approach was used, due to this research being of an explorative nature. Naturalistic inquiry is deemed suitable due to theory being posteriori rather than priori which is beneficial in exploratory research (Erlandson et al. 1993), minimising investigator bias and restraint on subjects’ responses, allowing for a more detailed and representative reflection of subjects’ perceptions. Open ended questions were used because of the greater understanding gained through them (Takemura et al. 2005), whilst capturing their view point without predetermined biases influencing the results (Patton 2001).
38 The Engineering of Sport 7 - Vol. 1 Transcriptions were produced from the recordings of the co-discovery and focus group discussions which were then analysed using an inductive process. The aim of this stage of testing was to build a greater understanding of the comfort construct, and a more comprehensive model for PPE. Inductive analysis was seen as an appropriate way to do this as the method : 1. Condenses extensive and varied raw text data into a brief, summary format 2. Establishes clear links between the research objectives and the summary findings derived from the raw data 3. Develops models or theories about the underlying structure of experiences or processes which are evident in the raw data (Thomas 2006) The inductive analysis process began with the verbatim transcription of the data allowing for familiarisation with the results. Once familiar with the data, important quotes were highlighted and became the basic unit for analysis. Quotes based around a similar concept were then grouped together forming base themes, these were then grouped together into higher level themes, this process continued until further categorisation was not possible (Roberts et al., 2001). Four individual interviews were conducted to further validate the results. The interviews were conducted whilst participants used the pads, facing between 10 and 15 balls per pad, running a minimum of 6 runs, and padding different deliveries away. A bowling machine (Bola) was used to deliver the ball at 50mph representing an average spin delivery. Four of the six pads were used by each participant. Prior to testing each pad, the participants were asked for their initial assessment of the equipment. Throughout the use of the pad, feedback on it’s performance was recorded and a final evaluation was gathered post use. This method was then repeated for all pads. The data collected was transcribed and analysed through a deductive process, where quotes were grouped into the themes that emerged from the inductive analysis. By utilising a deductive approach, the applicability of the model developed through the inductive analysis process previously outlined was assessed. No new themes emerged, therefore it can be concluded that a data saturation point had been reached.
3- Results The aims of this study were to illustrate how end users’ opinions can be obtained and utilised within product development, and to identify the characteristics that influence their comfort, allowing for a greater understanding of perceived comfort and its determinants. Six general dimensions emerged from the data, these were “Fit”, “Aesthetics”, “Weight”, “Protection”, “Thermal comfort” and “Sensorial comfort”. As the complete set of results are too extensive for the confines of this paper the example of “Protection” will be presented to illustrate the results. The tree structure for this dimension is shown in Figure 1, and demonstrates how the analysis process progressed from the initial quotes on the left hand side to the general dimensions on the right hand side, through the use of clustering.
The Engineering of Sport 7 - Vol. 1
39
The level of protection was found to influence a player’s willingness to use the pads, their confidence and overall satisfaction, which ultimately influenced perceived comfort. Higher order sub-themes of “Amount of Padding “and “Weaknesses” were grouped together within the general dimension of protection as illustrated within Figure 1. Within the sub-theme of amount of padding players often described the pads as ‘too thick’, ‘over padded’ or ‘bulky’ through statements such as “They are bigger, which can be negative, as too much padding gets in the way” illustrating the negative perceptions associated with too much padding. At the other end of the scale players described pads which were perceived to have less padding as ‘flimsy’, ‘thin’ and ‘unsafe’, demonstrating the players reluctance to wear pads which fit within this category. “Compared to my pads, they look a bit flimsy… if I got hit on the leg I wouldn’t be 100% confident” “Weaknesses” depicted areas of the leg exposed to the ball, perceived level of protection and specific areas with reduced protection. Players consistently identified the knee roll as an area of weakness, commenting on it’s ‘reduced protection’, and the ‘sensitivity to impact’ around the joint. “Lacking in protection at the knee roll definitely… I got one on the front of the knee, it didn’t hurt but jolted me a bit, there is nothing there behind the knee roll, or just below it,… If you get hit just below the knee roll you know how sensitive it is, and there is nothing there.” The other dimensions encompassed a variety of themes, with “Fit” being the most complex containing seven higher order sub-themes. It emerged, that in general players prefer a pad that shapes to the leg, with increased flexibility allowing for a greater range of motion. Dimensions such as “Thermal”, “Sensorial” and “Weight” had fewer contributory factors, however they illustrated the players desire for a pad to have increased breathability, reducing sweat rate, whilst being light. Finally the “Aesthetics” dimension clearly indicated the influence of both physical and psychological factors influencing the user’s perceived comfort, attributing feelings of discomfort to self image, pad shape and look of the pad.
40 The Engineering of Sport 7 - Vol. 1
Figure 1 - Tree structure for ‘Protection’ dimension.
4- Relationship modelling Analysis of the data revealed many inter-dimension relationships, demonstrating that the complete analysis could no longer be represented by simple tree-structures; therefore relationship modelling (Roberts et al., 2001) was used to identify links between different themes through further analysis of the data collected. There were 13 links identified within the data, representing the participant’s perceived influence of one theme on another and the relationship between them. For example, the sub-theme ‘Amount of padding’ was found to be associated with three other themes through various quotes as illustrated in Figure 2:
The Engineering of Sport 7 - Vol. 1
41
Figure 2 - Relationship modelling for “Amount of Padding”.
Movement – Amount of padding “The pads were a bit over padded stopping you being able to move properly” “Others can be over padded and you are not able to move in them.” Size – Amount of padding “A lot of padding so it’s really comfortable, but they look quite big, they look bulky” Flexibility – Amount of padding “Flexibility is key, but you’ve got to have the padding as well.” “One thing I always want is flexibility, but with the padding,” These examples demonstrate how the perceived increase in the amount of padding can affect the players’ perception of size and flexibility of the pad as well as their ability to move whilst using the equipment, which will ultimately influence their comfort and performance. These relationships highlight the users’ perception of each category and its relationship with other sub-components, but it does not show the relative importance of each characteristic. Further knowledge is needed to be able to fully understand the relationships between dimensions and the resultant affect each theme has on overall comfort, allowing designers to understand where tradeoffs will be beneficial in terms of user comfort and satisfaction.
5- Discussion Current literature has focused on symptoms of discomfort rather than causes, preventing the findings from being utilised in product development to optimise user comfort. Through the use of subjective methods, this study has identified several factors that influence the perceived comfort of cricket leg guards. It was found that perceived comfort can be influenced prior to wearing the pad due to the appearance and the initial interaction between the user and the equipment. This initial interaction can include picking the pad up, bending it and hitting it, as well as analysing how the pad looks. This
42 The Engineering of Sport 7 - Vol. 1 initial assessment then combines with the wearer’s experience whilst using the pad to produce their overall perceived comfort. From the analysis, six main attributes have been identified that influence comfort, through both the initial interaction and wearing of the PPE. These attributes are believed to interact and influence one another to determine the comfort, as demonstrated by the structured relationship model. The links between different themes can be either positively or negatively correlated. One example of a positive correlation is between the amount of padding and weight of the pad, as it emerged that as the amount of padding increased so did the weight. Within the model negatively correlated links were also highlighted where increasing one component is perceived to decrease another. An example of this is between material tactile feel and sweat, where increased sweat decreases tactile feel of the material. Therefore, if sweat production is decreased by improving thermal comfort, material tactile feel will increase. Another example of this is between weight and movement, where a decrease in weight was found to increase player movement. This was highlighted through quotes such as: “Also when I was running in them they seemed dead light, and obviously the less weight you are carrying when you are out there the better” “If they are too heavy you will not be able to run in them, or be able to move” These findings suggest a need for further understanding of the relative affect each attribute has on the perceived comfort, allowing for tradeoffs to optimise comfort. The understanding of these relationships will be vital in developing comfortable equipment. This procedure also demonstrated that different methods to those employed in previous literature within comfort can be more beneficial, improving understanding of the causes of discomfort and how these interact. Through this testing the benefits of using a practical element within the test procedure were identified. Feedback from the participants indicated that the practical testing allowed them to gain “a good feel of different pads” and identify factors they found to be beneficial or detrimental to comfort. The methods utilised within this study are also transferable to other sports and pieces of equipment, where the results from these qualitative studies can be used to develop further research instruments to provide both qualitative and quantitative data regarding the product, which can be utilised within product development. These results provide an opportunity for further investigation into the construct of comfort allowing for a greater knowledge base to be gained and for information to be provided to the design team through the use of a product development specification (PDS). The information from this investigation could be utilised in developing a hierarchy of needs for the performer, giving each dimension a relative importance, allowing for designers to focus on the key design aspects of the product as perceived by the end user. From the development of the hierarchy of needs objective measurements of the highlighted design elements can be used to provide vital information within the PDS, allowing for a more comfortable/ satisfactory product to be developed.
The Engineering of Sport 7 - Vol. 1
43
6- Conclusion The aim of this study was to identify factors that influence the perceived comfort of cricket leg guards. A combination of qualitative methods were utilised to elicit the players’ perceptions; the data was analysed using an inductive approach, then validated through deductive analysis of individual interviews. The analysis identified six core dimensions - ‘Sensorial comfort’, ‘Thermal comfort’, ‘Aesthetics’, ‘Fit’, ‘Weight’ and ‘Protection’. The testing also identified several inter-dimension relationships. These links highlighted a need for further analysis of each dimension particularly the relative importance of each one, in order to maximise player comfort. This study focused on cricketers but could easily be adapted for other sports, allowing for the development of comfort models specific to each piece of equipment.
7- References [AK1] Adams, P., and Keyserling, W. (1996). Performance of Protective Clothing: Fifth Volume, ASTM 1237, Philadelphia, PA: American Society for Testing and Materials. [AB1] Akbar-Khanzadeh, F., Bisesi, M.S. & Rivas, R.D. (1995). Comfort of personal protective equipment. Applied Ergonomics, 26(3), 195-198. [AD1] Ashdown, S.P. & DeLong, M. (1995). Perception testing of apparel ease variation. Applied Ergonomics, 26(1), 47-54 [BS1] Branson, D.H. & Sweeney, M.M. (1991). Clothing Comfort Conceptualization and Measurement: Toward a Metatheory. In Critical Linkages in Textiles and Clothing: Theory, Methods and Practice. (edited by S. Kaiser and M.L. Damhorst), ITAA Publishers. [CD1] Civille, G.V. & Dus, C.A. (1990). Development of Tern-linology to Describe the Handfeel Properties of Paper and Fabrics. Journal of Sensory Studies, 5, 19-32. [EH1] Erlandson, D.A., Harris, E.L., Skipper, B.L. & Allen, S.D. (1993). Doing naturalistic inquiry: A guide to methods. USA: Sage publications. [FH1] Fourt, L. & Hollies, N. R. S. (1970). Clothing; Comfort and function. USA: Marcel Dekker. [FM1] Fox, E.L., Mathews, D.K., Kaufman, W.S. & Bowers, R.W. (1966). Effects of football equipment on thermal balance and energy cost during exercise. Research Quarterly, 37, 332-339. [P1] Patton, M.Q. (2001). Qualitative Research and Evaluation Method. London: Sage Publications [P2] Pontrelli, G.J. (1977). Partial Analysis of Comforts Gestalt. In Clothing Comfort (edited by Hollies, N. R. S. and R.F. Goldman), 71-80. USA: Ann Arbor Science [RJ1] Roberts, J.R., Jones, R., Harwood, C.G., Mitchell, S.R. & Rothberg, S.J. (2001). Human Perceptions of Sports Equipment under Playing Conditions. Journal of Sports Sciences, 19(7), 485-497. [S1] Sontag, M.S. (1986). Comfort dimensions of actual and ideal insulative clothing for older woman. Clothing and Textiles Research Journal, 4, 9-17 [SF1] Stevenson, M., Finch, C., Hamer, P., and Elliott, B. (2003) The Western Australian sports injury study. British Journal of Sports Medicine, 37, 380-381 [S1] Stull, J. (2000). Cooler fabrics for Protective Apparel. Industrial Fabric Products Review, 3, 62-68.
44 The Engineering of Sport 7 - Vol. 1 [TS1] Takemura, Y., Sakurai, Y., Yokoya, S., Otaki, J., Matsuoka, T., Ban, N., Hirata, I., Miki, T. & Tsuda, T. (2005). Open-Ended Questions: Are They Really Beneficial for Gathering Medical Information from Patients? The Tohoku Journal of Experimental Medicine, 206(2), 151-154 [T1] Thomas, D.R. (2006). A General Inductive Approach for Analyzing Qualitative Evaluation Data. American Journal of Evaluation, 27(2), 237-246. [Z1] Zimmerli, T. (1998). Schutz und Komfort von Feuerwehrbekleidung (Protection and Comfort of Firefighters’ clothing). Textilveredlung, 33(3), 52-56.
Enabling Technologies for Robust Performance Monitoring (P10) Laura Justham1, Sian Slawson1, Andrew West2, Paul Conway2, Michael Caine1, Robert Harrison2
Topics: Sailing/Water Sports, Modelling. Abstract: Monitoring and performance analysis of elite swimmers during training and competition is an under-developed area of research due to difficulties in collecting data while the athletes are in the water. A monitoring system which can provide timely feedback to the swimmer, coach and sports scientist regarding the performance and physiological capabilities of an athlete is critical for the development of optimised personal training plans that could ensure a swimmer’s continued improvement and enhance their ability to win medals in major competitions. This paper is focused on the development of the requirements of such a monitoring system. A critical review of current monitoring systems that are in use in swimming training has been carried out and has identified areas where technology is currently being used. These technologies include video analysis, pressure and force measurements, velocity and acceleration measurements and physiological monitoring. Suitable technologies for a novel integrated and distributed monitoring system have been identified and a prototype demonstrator, which is as non-invasive and non-encumbering as possible, is currently under development at Loughborough University. This prototype provides both inter and intra stroke information as well as video augmentation and a comprehensive web-based training database to provide a comprehensive monitoring system for swimming. Keywords: Swimming, Performance Monitoring, Sensor Networks, Image Processing.
1. Loughborough University Sports Technology Institute, Loughborough Science and Enterprise Park, 1 Oakwood Drive, Loughborough, LE11 3QF, United Kingdom - E-mail: {L.Justham, S.E.Slawson, M.P.Caine}@Lboro.ac.uk 2. Wolfson School of Mechanical and Manufacturing Engineering, Loughborough University, Loughborough, LE11 3TU United Kingdom - E-mail: A.A.West, P.P.Conway, [email protected]
46 The Engineering of Sport 7 - Vol. 1
1- Introduction Sports monitoring systems which provide timely feedback to a coach and their athlete are critical at an elite level. Athletes require information regarding their technique, physiological capabilities and past performance in competition to develop a personalised training plan that is focused upon continued improvement in their chosen discipline. To this end the expertise from different professional practitioners, such as sports scientists and physiologists, is required to address the needs of each individual athlete and quantitatively assess their progression. A large number of commercial monitoring systems are available to support these activities in various sporting disciplines. The systems which are available focus upon both direct (i.e. equipment attached to the athlete) and remote (i.e. equipment remote to the athlete) monitoring techniques. The most commonly used direct monitoring techniques are sensors attached to the athlete. Inertial measurement units (IMU) which contain combinations of accelerometers, gyroscopes and magnetometers can be used to track human limb movement (Luinge 2002; Luinge 2005; Rootenberg 2006; Xsens 2008) and in outdoor sports GPS receivers can be used to measure details such as position, velocity and distance travelled (Garmin 2008). Data collected from these devices may be logged using on-board memory or transmitted wirelessly to a host computer and analysed with proprietary software. The most commonly used remote monitoring techniques are digitisation of video, motion tracking using vision systems, force plates and pressure mats. Video digitisation requires manual post-processing of video sequences (Dartfish 2008; Quintic 2008). Vision based motion tracking requires the athlete to wear reflective markers to allow their movement to be accurately tracked (Vicon 2008). Force plates and pressure mats are often wired directly to a computer to provide real-time feedback to the athlete regarding the force exerted during a predefined movement (Kistler 2008; RSScan 2008). These systems can provide informative and quantitative information however they can be limited by their set-up time and they often require the athlete to perform in a constrained environment such as in the laboratory. Both direct and remote monitoring tools are widely used in many sports however it is unusual that they are used in swimming due to the harsh environment where the athlete competes. It is the purpose of the current research to identify and develop an integrated monitoring system which is suitable for use in water. In order to carry this out, a critical evaluation of the current state of the art monitoring systems in swimming has been carried out and is presented in the following section. The prototype demonstrator, which is under development at Loughborough University, is being designed to deliver non-invasive, non encumbering real time feedback to the athletes and coaching personnel during both training and competition scenarios. Preliminary specification and design of the integrated monitoring system have been detailed in this paper.
The Engineering of Sport 7 - Vol. 1
47
2- Current State of the Art Performance Analysis Technologies A detailed literature review has been conducted to identify the current state of the art systems used for performance analysis in swimming. Seven types of monitoring have been identified, see Figure 1, and are (i) video analysis, (ii) physiological analysis, (iii) pressure or force analysis, (iv) velocity or acceleration analysis, (v) measurement of drag, (vi) ergometer analysis and (vii) theoretical analysis. These techniques have been divided into three distinct measurement approaches: 1. Remote Analysis where the required measurement equipment is separate to the athlete in the water 2. Direct Analysis where the required measurement equipment is attached to the athlete in the water 3. Modelled Analysis where testing is undertaken out of the water or where theoretical modelling is used to analyse the athlete Video or image processing is a popular remote performance monitoring technique as it can be used in both training and competition. Visual information is gathered and analysed using post-processing and digitisation. Qualitative analysis can be undertaken to discuss the athlete’s perception of their performance, whereas quantitative analysis can be performed by manually or automatically digitising images to allow information such as velocity, joint angle and body position with respect to elapsed time to be derived. To ensure that meaningful quantitative results are obtained, video hardware must be calibrated carefully to enable them to interface with software applications. Manual digitisation is currently undertaken using software such as Quintic (Quintic 2008) and Dartfish (Dartfish 2008) and automated analysis is possible for elements of the stroke, for example when considering the glide (Sanders 2007). In addition to digitisation and video analysis software, human motion capture can also be carried out using systems such as Vicon (Vicon 2008). Markers are applied to the swimmer’s body and tracked through the water that might inhibit or interfere with their swimming stroke. Direct monitoring forms the bulk of current analysis techniques in swimming and involves the swimmer wearing markers or equipment or being tethered to monitoring equipment on poolside. Velocity has been directly measured using tethered systems such as the commercially available SPEED system which was developed by AP Labs (APLabs 2008) and in research (Dekerle 2002). The swimmer is attached to a light cable and asked to swim as normal. The cable is tethered to a poolside tachometer, which provides a measurement of velocity. It should also be possible to integrate accelerometer data to calculate velocity; however this is felt to be an unreliable method under normal circumstances due to errors accruing from the integration process (Mackintosh 2008). Generally, information gathered from accelerometers is used to develop an understanding of stroke technique by positioning the devices on the wrists (Ohgi 1999; Ohgi 2002) or on the small of the back (James 2004). Specific and detailed information over multiple stroke cycles may be logged and represented graphically to visualise the movement of an
48 The Engineering of Sport 7 - Vol. 1 athlete through space. The method of using accelerometer devices in swimming is still in its infancy in terms of development and complexity, however there is the potential for these types of devices to be developed further such that more thorough and detailed analysis techniques to monitor and understand stroke technique can be developed. Pressure and force sensors are used widely in land-based applications to measure the forces exerted by the body during various forms of exercise. Within water, the measurement of pressure or force becomes more complex due to the effect of the water. AP Labs have developed a hand pressure measurement system, KZ (APLabs 2008) which uses a differential pressure transducer to monitor the pressure applied by each hand as the swimmer moves through the water. This device requires the swimmer to wear a waterproofed pressure transducer on the small of the back which is connected to the palms of the hands via air filled tubes. Aquanex is a semi-tethered hand pressure measurement system where the swimmer is attached to a boom which is a fixed link to the pressure measurement device on poolside. The swimmer is asked to swim 15m and the pressure exerted by the hands on the water is monitored (Aquanex 2008). Not only is it possible to monitor the pressure exerted by the hands or feet during swimming (Berger 1999; Takagi 2002), it is also possible to monitor the force or pressure exerted on the blocks at the start or on the wall during the turns (Blanksby 1996; Lyttle 1999; ATTRU 2008). Drag on the swimmer through the water has also been considered using the MAD system (measurement of active drag) (Strojnik 1999; Toussaint 2002; MAD 2008), which measures the push off force applied by the swimmer during front crawl swimming. Physiological monitoring is the fourth main area of interest in swimming performance analysis and is concerned with the continual process of monitoring the overall health of an athlete as well as the outcomes of targeted training sessions. For example the weight, body fat, general health and fitness of an athlete is continually under scrutiny whereas heart rate might only be investigated during heart rate specific training sets where chest strap monitors are used. In addition to active monitoring, where the swimmer is in the water, techniques such as land-based ergometer testing and computer based modelling have been used to model and approximate how the swimmer moves through the water. Computational Fluid Dynamics (CFD) using software such as Fluent (Fluent 2008) has been used to model how a swimmer moves through the water. This is particularly successful during the underwater phases of swimming (i.e the glide) but has been less successful during free swimming due to the complexities of the water-air interface (Lyttle 2006). Theoretical models of the hands have also been generated and analysed to calculate the forces produced as the swimmer moves through the water (Berger 1999; Sanders 1999). In addition to modelling the swimmer in the water dry land monitoring using swimming ergometers, such as those manufactured by Vasa (Vasa 2007) and Weba (Weba 2007), which are designed to imitate the actions required for swimming have been carried out. The types of measurements which are possible out of the water are consistent with the types of measurements that are desirable in water, however it is likely that there are differences in the swimming action on the ergometer with respect to in the water. Ergometer testing has been carried out looking into arm and leg power and also the cardiopulmonary response during simulated swimming (Swaine
The Engineering of Sport 7 - Vol. 1
Figure 1 - Monitoring techniques currently used in swimming performance analysis.
49
50 The Engineering of Sport 7 - Vol. 1 1996; Swaine 1997; Swaine 2000). These tests are not ideal as they are laboratory based simulations rather than real training however the provision of physiological as well as biomechanical information has the potential of providing a thorough data set which is otherwise not possible to obtain during training.
3- Requirements of an Integrated Performance Monitoring System A research project being undertaken at Loughborough University is concerned with the design and development of an integrated monitoring system which is suitable for use in a swimming training environment. The review of current performance analysis technologies in swimming (above), has led to a detailed requirements specification for an ideal training system. A prototype monitoring system is currently under development with an aim of providing a distributed monitoring system which is as discrete and non-invasive as possible. The system aims to provide both indirect (remote) and direct (continuous) monitoring, which results in instrumentation being fixed onto the swimmer in training and into the pool environment, see Figure 2.
Figure 2 - The identified requirements of an integrated distributed monitoring system. Indirect monitoring takes place remote from the swimmer, within the pool environment, whereas direct monitoring involves equipment being attached to the swimmer in the water.
In swimming, video and image processing appears to be the most favoured method of performance analysis as it provides a non-invasive and non-encumbering approach. This type of performance monitoring is an important element within swimming as it is possible for data to be collected during training and competition and the swimmers are able to see themselves on screen. Generally video analysis is used for qualitative discussion rather than quantitative analysis because it is manually intensive and time consu-
The Engineering of Sport 7 - Vol. 1
51
ming to provide meaningful quantitative data to the swimmer. For the prototype performance monitoring system the integration and synchronisation of video analysis software is an important aspect of the complete system. This is because video monitoring is such an established and trusted technique which can be used to augment the sensors and devices which are integrated into the complete distributed monitoring system. In addition to image processing and video analysis techniques, force and pressure sensors which are instrumented into the pool surround, such as within the start blocks or timing touchpads, are also required for a fully integrated system. These devices can be used to provide quantitative information to the swimmer regarding the position and time spent on the wall during the turn and the time taken before their first movement off the blocks. The third remote monitoring approach is land based physiological monitoring which includes weight, body fat, general health and cardiovascular fitness observations. Continuous monitoring is possible when the swimmer is instrumented with sensor devices and it is envisaged that the distributed monitoring system should include both physiological and physical measurements. This instrumentation should allow inter and intra stroke information about a swimmer’s technique to be monitored as well as differences in technique between individual swimmers. Initial investigations have involved the swimmer using a single tri-axis accelerometer device which logs their motion through the water (Slawson 2008), however inertial measurement units which are capable of logging data for post-processing or transmitting wirelessly to a poolside computer are required to provide continuous quantitative information to the coach and sports scientists based on poolside. Real-time data transmission, processing, analysis and presentation are required to allow the coach to disseminate quantitative feedback to the swimmer during their training session. This will allow the swimmer to make changes in their technique and measure whether these changes are successful in improving their stroke. To ensure that any data collected from the distributed monitoring system provides comprehensive and useful information to the swimmer and coach it is necessary for the data to be linked and archived via a comprehensive web-based training database, using a system oriented approach, to provide a complete performance assessment package (Justham 2008). Data which is collected is required to be collated with any other data from that session and for that swimmer and stored for comparison or retrieval at a later date. In addition data must be analysed and presented in a format which is accessible for the swimmers, coaches and sports scientists to provide in-depth and useful information in a timely manner.
4- Discussion and Conclusions In conclusion the distributed system which is under development at Loughborough University is being designed to provide a novel state of the art performance analysis tool for swimming training to provide timely feedback on all aspects of the swimmer who’s technique is being monitored. This type of tool is critical for the elite swimmer who requires accurate information to develop the most effective training programs. The current research has been focused upon the assessment of existing state of the art tech-
52 The Engineering of Sport 7 - Vol. 1 nologies and the identification of novel technologies which might be viable within the current application. Existing state of the art technologies provide limited feedback to the swimmer and are restricted by their accuracy, set-up time and obtrusiveness to the athlete. It is envisaged that the system under development will deliver non-invasive, nonencumbering real-time feedback to athletes, coaches and sports scientists in training and competition. In this initial phase, the system is being designed with the elite athlete in mind, however it is envisaged that a successful monitoring system should be capable of being disseminated and used by swimmers of all standards from recreational enthusiasts to the international athlete.
5- Acknowledgements The authors would like to thank the sports scientists, coaches and swimmers at Loughborough University who have taken part in the information gathering stages of this research. Funding for this work has been provided by the Innovative Manufacturing and Construction Research Centre (IMCRC) based at Loughborough University, part of the Engineering and Physical Sciences Research Council of Great Britain (EPSRC), and UK Sport.
6- References [A1] APLabs (2008) "The website of AP Labs which is an Italian research company who have developed the KZ and SPEED systems for swimming: : http://www.aplab.it/homeeng.html, accessed 28th Feb 2008. [A2] Aquanex (2008) "Website for the Aquanex pressure measurement system: http://swimmingtechnology.com/Aquanex.htm, accessed 28th Feb 2008. [A3] ATTRU (2008). Media press release detailing the features of the AIS Recovery and Swimming Centre in Canberra Australia http://catalogue.ausport.gov.au/fulltext/2006/ascmedia/2006.10.16.asp, accessed 28th Feb 2008. [BH1] Berger, M. A. M., Hollander, A.P., Groot, G.D. (1999). "Determining propulsive force in front crawl swimming: A comparison of two methods.." Journal of Sports Sciences 17: 97-105. [BG1] Blanksby, B. A., Gathercole, D.G., Marshall, R.N. (1996). "Force plate and video analysis of the tumble turn by age-group swimmers." Journal of Swimming Research 11: 40-45. [D1] Dartfish (2008) "Manufacturers website of the Dartfish video software solutions company: http://www.dartfish.com/en/index.htm Accessed 27.02.08." [DS1] Dekerle, J., Sidney, M., Hespel, J.M., Pelayo, P. (2002). "Validity and Reliability of Critical Speed, Critical Stroke Rate and Anaerobic Capacity in Relation to Front Crawl Swimming Performances.." International Journal of Sports Medicine 23(93-98). [F1] Fluent (2008). Website for the Fluent CFD modelling software comapny http://www.fluent.com/. This company uses its ANSYS software for modelling the swimmer's motion in the water. Date accessed 27-02-08. [G1] Garmin (2008) "Manufacturers Website for the Garmin GPS fitness products: http://www.garmin.com/garmin/cms/site/uk/ontofitness/ Accessed 27.02.08." [JD1] James, D. A., Davey, N., Rice, T. (2004). "An Accelerometer Based Sensor Platform for Insitu Elite Athlete Performance Analysis." Sensors 2004, Proceedings of IEEE 3: 1373-1376.
The Engineering of Sport 7 - Vol. 1
53
[JS1] Justham, L. M., Slawson, S.E., West, A.A., Conway, P.P. (2008). Business Process Modelling and its use within an Elite Training Environment. International Sports Engineering Association (ISEA) 7th International Conference, Biarritz, France. [K1] Kistler (2008). "Company Website for the Kistler force measurement company: http://www.kistler.com/do.content.gb.en-gb?content=KistlerCountryHome_KIL Accessed 27.02.08." [L1] Luinge, H. J. (2002). Inertial Sensing of Human Movement, University of Twente. PhD Thesis. [LV1] Luinge, H. J., Veltink, P. H. (2005). "Measuring orientation of human body segments using miniature gyroscopes and accelerometers." Medical and Biological Engineering and Computing 43: 273-282. [LK1] Lyttle, A., Keys, M. (2006). The application of computational fluid dynamics for technique prescription in underwater kicking. Biomechanics and Medicine in Swimming X (Portuguese Journal of Sport Sciences), Porto. [LB1] Lyttle, A. D., Blanksby, B.A., (1999). "Investigating Kinetics in the Freestyle Flip Turn PushOff." Journal of Applied Biomechanics 15: 142-152. [MJ1] Mackintosh, C., James, D., Grenfell, R., Zhang, K. (2008). Monitoring Sport and Swimming Publish Number: US 2008/0018532 A1. U. S. P. Office. Australia: 18. [M1] MAD (2008). Measurement of Active Drag (MAD) system website: http://web.mac.com/htoussaint/SwimSite/Welcome.html, accessed 28th Feb 2008. [O1] Ohgi, Y. (2002). "Microcomputer-based Acceleration Sensor Device for Sports Biomechanics ~ stroke evaluation using swimmers wrist acceleration." Proceedings of IEEE, Sensors 2002 1(699704). [OI1] Ohgi, Y., Ichikawa, H., Miyaji, C. (1999). "Characteristics of the forearm acceleration in swimming." Biomechanics and Medicine in Swimming VIII. [Q1] Quintic (2008) "Manufacturers website of Quintic Consultancy Ltd.: http://www.quintic.com/ Accessed 27.02.08." [R1] Rootenberg, D. (2006). Inertial and Magnetic Sensing of Human Motion, University of Twente. PhD Thesis. [R2] RSScan (2008) "Manufacturers website for the RS Scan company: http://www.rsscan.com/ Accessed 27.02.08." [SH1] Sanders, R., Haake, S.J. (2007). EPSRC Funded research: EP/F006128/1 Improving Swim Performance by Optimising Glide Efficiency and Time of Initiating Post-Glide Actions. 12 Month project running from July 2007. [S1] Sanders, R. H. (1999). "Hydrodynamic Characteristics of a Swimmer’s Hand." Journal of Applied Biomechanics 15: 3-26. [SJ1] Slawson, S. E., Justham, L.M., West, A.A., Conway, P.P. (2008). Accelerometer Profile Recognition of Swimming Strokes. International Sports Engineering Association (ISEA) 7th International Conference, Biarritz, France. [SB1] Strojnik, V., Bednarik, J., Strmbelj, B. (1999). "Active and passive drag in swimming." Biomechanics and Medicine in Swimming VIII. [S2] Swaine, I. L. (1997). "Cardiopulmonary Responses to Exercise in Swimmer Using a Swim Bench and a Leg Kicking Ergometer " International Journal of Sports Medicine 18: 359-362. [S3] Swaine, I. L. (2000). "Arm and leg power output in swimmers during simulated swimming." Medicine and Science in Sports and Exercise 32(7): 1288-1292.
54 The Engineering of Sport 7 - Vol. 1 [SZ1] Swaine, I. L., Zanker, C.L. (1996). "The Reproducibility of Cardiopulmonary Responses to Exercise Using a Swim Bench. " International Journal of Sports Medicine 17: 140-144. [T1] Takagi, H. (2002). Measurement of propulsion by the hand during competitive swimming. The Engineering of Sport 4, Oxford: Blackwell Publishing. [T2] Toussaint, H. B. (2002). "The “Fast-Skin” Body Suit: Hip, hype, but does it reduce drag during front crawl swimming?" Scientific Proceedings – Applied Program – XXth International Symposium on Biomechanics in Sports – Swimming: 15-24. [V1] Vasa (2007). Vasa swim ergometer website: http://www.vasatrainer.com/index.php?page= Swim%20Training%20with%20Vasa%20Trainer%20Vasa%20Ergometer. Accessed 23/04/2007. [V2] Vicon (2008) "Manufacturers website of the Vicon motion capture company: http://www.vicon.com/applications/life_sciences.html accessed 27.02.08." [W1] Weba (2007). Weba Sport Website for their Swim Ergometer: http://www.webasport.com/weba/swim_ergo.html. Accessed 23/04/2007. [X1] Xsens (2008) "Manufacturers website of Xsens Motion Technologies: http://www.xsens.com/ Accessed 27.02.08."
An Objective Performance and Quality Comparison of Drivers from Different Market Sectors (P11) Jeff Brunski1, John Rae2
Abstract: In the golf club market, there is a wide discrepancy in consumer cost between clubs made by premium, professional-line OEMs (original equipment manufacturers) and those offered by mass-merchant, so-called “value” brands. In an attempt to rationalize such cost differences, this study comprehensively compares the difference in quality and performance between drivers from these two disparate market sectors. The drivers chosen for the study represent the 2007 best-selling OEM in the woods category and a typical massmerchant brand (Datatech, 2007). The performance of each club was measured using a combination of robotic and human testing from which ball flight data was collected by radar and image-based launch monitors. Durability was measured with simulated repetitive use and exposure to extreme conditions. Lastly, the physical properties of each club were measured in order to quantify consistency and to help identify the origin of any difference in performance. Keywords: Golf Clubs, Drivers, Comparison, Performance, Premium.
1- Introduction This paper is an in-depth look at the driver portion of “An Objective Performance and Quality Comparison of Professional Line, Premium Golf Clubs to Value Sets Sold Through Mass Merchants” (Brunski and Rae, 2007). Though that previous study outlines much of the methodology and results of the drivers, this paper aims to further elaborate and detail those methods, results, and their likely causes. The driver, often hailed as the flagship of a product line, is deserving of this additional attention due to its high visibility and marketing prominence.
2- Method Unless otherwise specified, testing methodology followed the procedures described by Brunski and Rae (2007). 1. 5601 Skylab Rd. Huntington Beach, CA 92647 - E-mail: [email protected] 2. E-mail: [email protected]
56 The Engineering of Sport 7 - Vol. 1
2.1 Club Selection and Description The value driver was one of 12 clubs in a set purchased from a large mass-merchant store for $149.99 USD. This values it at $12.99, though the club would likely retail for more if offered on an individual basis. The premium product cost $300 USD.
2.2 Static Static measurements (the physical properties of the clubs) were taken with standard measurement instruments that were all calibrated for accuracy. When mentioned, averages and standard deviations represent a population of three products. Industry averages come from product specs provided by the leading five OEMs in woods sales (Datatech, 2007). Frequency is the club’s first mode of vibration when clamped at the grip. Sweet-spot is where the center of gravity orthogonally projects on the face.
2.3 Player & Robot The player test consisted of a sampling of 25 Southern-California golfers with an average handicap of 14. Seven players reported handicaps of 10 and under, 15 fell between 11 and 19, and the last three had handicaps between 20 and 30. Players were asked to hit each driver until they had recorded what they considered to be three typical shots and only these shots were used in calculating results. A radar based launch-monitor (Trackman) was used to collect ball flight data. Furthermore, the players were asked to rank the clubs subjectively in a number of categories. Player’s responses were on a 1-5 scale (5 being excellent, 1 being awful) for all topics, save trajectory (1-5 describing increasing height with 3 being ideal) and flex (1-5 describing increasing stiffness with 3 being ideal). The subjective player feedback results can be compared against the objective results to help understand the origins of perceived performance. The drivers were not disguised in any manner. This testing methodology does not take into consideration how many swings it took each player to accomplish three good shots. One could argue that this ignores a certain playability factor. Playability is a difficult parameter to quantify and it’s probable that it would correlate strongly with how well the test club matched the player’s gamer. This would be an interesting study in and of itself. Moreover, playability surely held influence on subjective feedback.
2.4 Robot A Golf-Labs® robot was used to acquire more objective performance results. The swing chosen was selected for its launch conditions matching historical player data. Impacts were made in a 12mm-spaced, nine-point square grid around the center of the club-face as found by USGA protocol (USGA 2005). Each impact location was hit 5 times in order to eliminate differences from ball to ball. Ball flight characteristics were captured using an image-based launch monitor and carry/dispersion data was collected as the balls landed within a marked grid. Dispersion is defined as the area of the ellipse formed by the carry and offline standard deviations. Face angle was set in the orientation required
The Engineering of Sport 7 - Vol. 1
57
in order to have center-face shots land within ±4.6m of the center-line. Relevant environmental conditions – wind speed, temperature, etc. – were consistent for both clubs.
2.5 Durability Both drivers were put through extensive durability testing in order to simulate repetitive use. Four tests were run: air-cannon impacts, exposure to UV light, paint adhesion, and vibration scratch-resistance. For air-cannon testing, each club was clamped in place at the grip and subjected to 500 center-face, 56m/s impacts. This ball speed was chosen based on the market’s average swing speed and a factor-of-safety (T. Mase, personal communication, 2003). Bulge and roll (which describe the curvature of the face) measurements were taken before and after testing to determine if any face-flattening had occurred, and the clubs were also inspected for signs of visible damage. The UV test consisted of 16 hours of UV light exposure, simulating extended exposure to sunlight (non-ASTM test). The paint was then visually inspected for fading. Paint adhesion was also tested using a standardized cross-hatch test (ASTM D3359). Lastly, we placed the drivers in a fully-loaded golf bag and left the bag on a vibration plate. This is not a standardized test, but it does simulate the rattling that occurs within most bags. After 30 minutes, the drivers were visually inspected for damage.
3- Results Table 1 - Static measurements of the club and shaft for both Drivers tested. *Beyond instrument’s limit.
58 The Engineering of Sport 7 - Vol. 1 Table 2 - Player and Robot testing performance results.
Figures 1 & 2 - Player (1) and robot (2) test results. Average carry and offline distance as well as overall dispersion indicated.
Figure 3 - Subjective player feedback (5 = best, *3 = ideal).
Figure 4 - Internal view of premium and value driver.
4- Discussion There are results in this study that are supported by static measurements and there are results that are not intuitive or readily explained. The scope of this section is to present the results, discuss likely causes for them, point out conflicting data, and draw any logical conclusions.
The Engineering of Sport 7 - Vol. 1
59
4.1 Specifications and Averages The specs of the products tested closely matched market averages in length, swing weight, loft and lie. Driver specs are basically similar across the market. The value product is flatter (lower degree) in lie angle, which would contribute to players hitting shots to the right. The premium is heavier in swing weight, which may have adversely affected head speed in player testing. These dynamic effects will be discussed more in later sections. As far as consistency, the swing weight, club weight, and loft data indicate a clear difference between the two products. The standard deviations for these measurements were much smaller for the premium drivers. The most glaring differences were in swing weight and loft – both of which have significant influence on playability. Based upon our measurements of three products, the data supports what most people would assume: premium products are more consistently made. This can be attributed to increased manufacturing precision and quality control, both of which add cost to the OEM, and in turn, to the consumer.
4.2 Launch Conditions Launch conditions are not explicitly apparent to a golfer; without a launch monitor, a player must extrapolate these values from sight, feel, carry and offline distance. However, considering the increasing availability of launch monitors, as well as the fact that launch conditions are the building blocks of ball flight, these parameters are of utmost importance. The most relevant launch conditions affecting ball flight are ball speed, launch angle, and backspin. Azimuth and sidespin also contribute but will not be discussed here outside of their resulting effect on dispersion. One of the most influential factors on ball speed is clubhead speed. We measured head speed in player testing using the radar-based monitor. For this group of players, the lower swing weight and softer flex of the value product produced nearly 3% higher head speed. This is in spite of the club being slightly shorter and heavier – two parameters that would typically be expected to cause lower head-speeds. This suggests the relative significance of these variables on head speed. The faster head speed of the value product should have accounted for more ball speed, all else being equal. This was not the case; the value driver actually showed inferior ball speed in both robot and player testing. The explanation for this result lies in other parameters that influence ball speed: COR, MOI, CG-SS depth, and sweet-spot location on the face. The premium product appears to be shaped and engineered to achieve values for these parameters that are beneficial to ball speed; the value product does not (See Section 4.6). The large COR difference between the two products should have accounted for about 1.34m/s less ball speed for the value driver (Beer et al 2004). This is just an estimate, however, since the COR was measured at center-face only while impacts occurred all over the face. The value driver’s lower MOI relative to that of the premium – about 14% to 19% depending on the axis of measurement – also certainly contributed to greater ball speed losses for impacts occurring off the sweet-spot. Lastly, ball speed is aided by a larger CG-SS depth (greater translational force exerted on the ball) and a more centered sweet-spot (less distance available for off-center impacts). The premium driver has superior properties in both of these parameters.
60 The Engineering of Sport 7 - Vol. 1 Launch angle and backspin also varied between the value and premium products throughout testing. This can be attributed to mass-properties such as CG-depth relativeto-the-hosel and sweet-spot height relative-to-center-face. The CG-depth affects launch angle and backspin by helping to determine how much the club dynamically lofts during the swing. The SS-height helps to determine whether an impact behaves like a high-face or low-face shot. Relative to the value clubhead, the premium product has values for these mass properties that better promote higher-launching, lower-spinning shots. However, in the robot test especially, these mass-property effects are overshadowed by shaft effects. Considering that all three launch conditions – ball speed, launch angle, and backspin – are tied at the hip, it is worth understanding the shaft properties before continuing this discussion.
4.3 Shafts The shafts played a major role in producing the launch conditions seen in testing. Ultimately, there was one main issue: the value shaft was much too weak, both in flex and torque (See Figure 1). This resulted in the clubhead not dynamically lofting in the robot swing and the club-face not coming square in the player swing. Nearly all launch conditions support this theory. Players exhibited similar launch angles with the two drivers, indicating that they were able to dynamically loft each driver nearly equally (guided by the feel of the shaft unloading). However, players hit the value driver 10 meters further to the right and with much less ball speed, both of which indicate an open club-face. In the robot test, where we forced center shots to the target line, the value driver showed a large drop-off in launch angle, despite having the same loft as the premium product. This low launch angle was accompanied by high ball speed and low backspin, all of which indicate that the value shaft was not dynamically lofting the clubhead. Figure 5 shows the EI value of both shafts throughout their lengths. The EI value quantifies stiffness and is a function of material properties (E being modulus) and geometry/thickness (I being inertia). It is derived by direct thickness measurements combined with deflection values of the shaft under a known load. This graph is further evidence that the value shaft is weaker than the premium – perhaps by as much as a “flex” or two depending on which shaft and club OEM you standardize by. The value shaft was simply too flexible for our players and even more so for the robot.
Figure 5 - Combined EI curves. ––– Value EI; - - - Premium EI.
Figure 6 - Individual E and I plots. –– Premium E; - - - Value E; Premium I; ▲▲▲ Value I.
The Engineering of Sport 7 - Vol. 1
61
The E and I curves also speak to the overall quality of the shafts. High performance shafts achieve the desired flex profile with minimal use of weight. The value shaft, being weaker and heavier than the premium shaft, is clearly an inferior product. The I curves in Figure 6 show that the value shaft is achieving stiffness via wall thickness and geometry more so than the premium shaft. This is typical of shafts made from inexpensive, low-modulus materials. The E portion of Figure 6 explicitly shows that the premium shaft is made from a higher modulus material than the value shaft. These quality differences are subsidiary to the actual measured performance results, but they speak to how performance is achieved.
4.4 Carry, Offline Distance, and Dispersion Though launch conditions are important, ultimately people tend to place more emphasis on overall carry distance. In our robot testing, the value driver had about 92% of the premium driver’s distance, and that number grew to about 96% in player testing. These percentages look impressive, but the raw numbers – 16.5 meters for the robot and 8.2 meters for players – are quite significant in the golfing world. On average, an 8.2meter improvement in driving distance would move a PGA player 53 positions in overall ranking, and this number grows to 84 if he improves 16.5 meters (PGAtour.com 2007). So the relative strength of the premium product is subject to interpretation. Offline distance was almost identical for the two drivers in robot testing, which is a function of forcing center-face impacts to the center-line. In player testing, the value driver was 10 meters more right than the premium – a significant difference. This is a function of the shaft, as discussed, as well as mass properties. For example, the value product’s relatively lower heel-toe MOI and more heel-biased CG both contributed to the face opening more at impact, causing shots to go right. Dispersion is the second ultimate performance variable desired in a driver. Dispersion actually encompasses the carry parameter since it is mathematically a function of the standard deviation of carry distance. In this study, the premium driver displayed tighter dispersion in both tests: 7.5% with players and 232% with the robot. Clearly, the player test was much closer than the robot; the players hit the value product into a similarly sized area as the premium, just shorter and more to the right. The robot dispersion area, however, was the largest difference in performance between the two clubs. From Figure 2, it is clear that this is a result of differences in distance control rather than offline control. Furthermore, per Table 2, it is clear that the distance variation was dictated by launch angle differences, not backspin or ball speed. The value driver had low-face shots launching at less than 4 degrees. At typical ball speeds, this trajectory simply cannot achieve very much carry. Ultimately, the glaring performance difference between these products was distance-control due to launch angle consistency. An interesting fact here is that the value and premium drivers were somewhat similar in terms of left-right dispersion. Though the value product was hit further offline nominally, both drivers sprayed shots equally left and right of their average offline distance. This happened in spite of MOI differences between the two drivers, indicating that dispersion is dictated by more than just pure MOI values, as some marketing may suggest.
62 The Engineering of Sport 7 - Vol. 1
4.5 Player Reactions The premium driver was ranked higher than the value in every category of our survey (Figure 3). The closest margin was 0.5 points in the sound category; the largest difference was nearly 2 points, coming in the design category. Of 22 players (3 abstained from selecting), 21 responded that their overall preference was for the premium. Considering the similar dispersion, this overwhelming player preference suggests that a driver’s appeal is very much subjective. In trajectory and flex, the premium was rated as having just the right ball flight but feeling slightly weak. The value driver was rated 0.6 points less stiff (matches data showing it was more flexible) and was rated as having a ball flight slightly too high. This is interesting since the value driver actually launched lower on average, but was probably due to the higher ball flight ultimately achieved due to greater backspin.
4.6 Durability & Construction Summary From inspection, the value driver is made from four individually stamped pieces of aluminum welded together. The premium club, alternatively, is composed of a milled titanium face welded to a cast titanium body. Both manufacturing techniques have their advantages and can ultimately achieve high-performance. The value driver is certainly a less-expensive clubhead to manufacture: stamping is a lower-cost process than casting and aluminum is a lower-cost material than titanium. The negative to stamping is that it limits the head to uniform thickness in each region. The value driver also exhibits imprecise welding at the joints and a large pool of weld material at the hosel region (see Figure 4). The large welds and uniform thickness regions reduce the available weightbudget for other advantageous locations, ultimately hurting mass properties. The premium product exhibits varying thickness throughout the clubhead and has no extraneous material build-up due to welding. This helps explain why it is able to achieve mass properties that are more conducive to launch conditions producing longer carry. As far as durability, neither clubhead massively failed or showed signs of visible damage after 500 center-face impacts. However, the value club-face fared better in this test – deforming 83% less than the premium’s. The value club’s bulge and roll barely moved (1.0 cm each), while the premium’s bulge and roll change of 3.6 cm and 8.6 cm respectively is borderline acceptable. The value driver’s face, made from a weaker material, is nearly twice the thickness as the premium club-face. Since the value driver could have achieved a higher COR at the cost of more face deformation over the lifetime of use, one might conclude that durability was a higher priority than ball speed for this product. Both drivers fared equally in ultra-violet testing, paint adhesion, and resistance to scratching. Considering cost differences, this could be seen as a victory for the value product.
5- Application This study reveals an interesting relationship between cost and performance – two characteristics that are not entirely proportional in this instance. The premium product
The Engineering of Sport 7 - Vol. 1
63
did out-perform the value driver in distance, dispersion, and it dominated player preference. The premium product was also more consistent in terms of manufacturing. However, the mass-merchant product was actually more durable. This perhaps speaks to the primary design focus of each product – performance in the premium product and durability in the mass-merchant. The value of each is a subjective combination of these characteristics. It would be necessary to test a larger number of products from each price-point in order to make more confident assertions about each market segment.
References [A1] Avallone E.A. & Baumeister T. (1996). Marks’ standard handbook for mechanical engineers, (10th Ed.) New York, Mcgraw Hill, pp. 3-19. [B1] Baker, K., Schempp, P. G., Hardin, B., & Clark, B. (1999). The routines and rituals of expert golf instruction. In M. R. Farrally and A. J. Cochran (Eds.) Science and Golf III; Proceedings of the World Scientific Congress of Golf, pp. 271-281. [B2] Beer, F., Johnston, E., & Clausen, W. Vector Mechanics for Engineers: Dynamics. New York, McGraw Hill, pp. 821-823. [B3] Brunski, J. and Rae, J. (2007). An Objective Performance and Quality Comparison of Professional Line, Premium Golf Clubs to Value Sets Sold Through Mass Merchants. Manuscript submitted for publication. [C1] Chou, A. (2004) The Business of Golf Technology. In M. Hubbard, R.D. Mehta, & J.M. Pallis (Eds.), The Engineering of Sport 5, Vol. 1. Sheffield, ISEA, pp. 21-32. [G1] Golf Datatech. (2007). U.S. Market Share Analysis of Top 10 Manufacturers by Product and Market Segment (in Units). http://www.golfdatatech.com/ [U1] United Stated Golf Association. (2005). Procedure for measuring the flexibility of a golf clubhead (Revision 2.0). Far Hills, NJ.
Defining Strategies for Novel Snowboard Design (P12) Aleksandar Subic1, Patrick Clifton, Jordi Beneyto-Ferre
Topics: Sports Technology and Design. Abstract: There is considerable anecdotal evidence that suggests snowboarders relate the perceived “feel” of a snowboard to its on-snow performance. The Snowboard Research Group at RMIT University in Melbourne thus set out to fully characterise the “feel” of snowboards for the main riding styles. By correlating subjective evaluations to objective laboratory and field based data, the relevant matrices of parameters leading to the desired “feel” of the board can be determined. This article deals with the front-end of the characterisation process by focusing on the identification of potential design innovation opportunities through a benchmarking analysis of modern snowboards. The qualitative data associated with snowboard “feel” has been obtained through a range of surveys and interviews, conducted online and in person (on-snow). A Quality Function Deployment (QFD) method was used to process the information collected, relating the subjective customer requirements to relevant objective technical attributes of snowboards for the selected riding styles. From the market research, user surveys and QFD, a comprehensive gap analysis was completed resulting in the identification of innovation opportunities and preferred design features for modern snowboards. The research determined bending and torsional stiffness distribution as well as camber as the key design characteristics influencing the “feel” and performance of snowboards for both freestyle and freeride riding styles. Keywords: Snowboards; Benchmarking; Feel; Performance; QFD.
1- Introduction Snowboarding is one of the fastest growing sports in the world today, with an estimated 3.4 million boarders on the slopes annually (MaxLifestyle.net 2006). As a result the snowboard equipment market is expanding at a phenomenal rate, with total industry sales in excess of $280 million dollars per annum in the US alone (SIA 2006). Unit sales of snowboards also rose by 23% in US chain stores between 2005 and 2006 (Ibid.). Given the relatively short history of this sport and the extent of expansion within the snow1. School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University, Melbourne, Australia E-mail: [email protected]
66 The Engineering of Sport 7 - Vol. 1 board market, there exists significant scope for potentially lucrative technological development and innovation, which to date has been primarily conducted on a trial and error basis by hobbyists and enthusiasts (Shah 2006). Considerable anecdotal evidence suggests that riders relate a snowboard’s on-snow performance to its perceived “feel”, or the physical and psychological feedback given to the rider whilst snowboarding. Such feedback may be visual, aural, kinaesthetic or vibrational (Roberts et al 2005), all having an effect on the muscular inputs applied to the board by the rider and the resultant movement and control achieved on the slope. Manufacturers currently spend significant time and money trialling new designs, relying heavily on the feedback of professional riders in attempting to design-in the “feel” and optimise the performance of the board. A systematic user-centred design procedure could provide the intelligence required to alleviate the trial and error approach, resulting in higher customer satisfaction as well as cost and time savings. The Snowboard Research Group at RMIT University in Melbourne set out to fully characterise the “feel” of snowboards for the main riding styles. By correlating subjective evaluations to objective laboratory and field based data, the relevant matrices of parameters leading to the desired “feel” of the board can be determined. This article deals with the front-end of the characterisation process by focusing on the identification of potential design innovation opportunities through a benchmarking analysis of modern snowboards.
2- Identification of Snowboard Requirements 2.1 Qualitative Analysis The qualitative data associated with snowboard “feel” used in this research has been obtained through a range of surveys and interviews, conducted online and in person (on-snow). Participants in an initial mass circulated survey (115 completed responses) were asked to identify, rate and analyse their current board against the set criteria. A secondary follow up survey with a smaller focus group (9 experts) allowed more specific questions about pinpoint “feel” and performance to be discussed leading to further refinement of the qualitative parameter list, which in its final form is as follows: Stability = How stable the rider feels on the board. Feedback = The amount of stress felt on the rider's body including the effects of board chatter. Speed = The gliding speed of the board compared to other boards of similar length. Accuracy = The precision of board movement in response to rider input. Forgiveness = The tolerance of the board to errors from the rider. Edge Grip = The level of grip exhibited during turns. Manoeuvrability = How easily the board responds to rider inputs. Transition smoothness = How easily the board flows from edge to edge. Board Liveliness = The level of 'pop' or spring in the board when performing a jump.
The Engineering of Sport 7 - Vol. 1
67
To finalise the qualitative analysis, on-snow testing and interviews using a range of high quality test boards were conducted to obtain subjective ratings with a strong statistical basis, and determine the interrelationships between each of the qualitative parameters. This data will also be used for eventual correlation with objective laboratory based measurements. Eight experienced testers (snowboarding instructors) were employed to ride and rate three best-in-class new snowboards that spanned the freeride-freestyle board spectrum. One highly freeride oriented and one specialist freestyle board were chosen to represent the end points of the spectrum, and a third versatile board was selected mid-way. The rating system utilised in the analysis was based on the approach used by the BMW Group and the University of Bath investigating steering “feel” for BMW vehicles (Harrer et al. 2006). Each parameter was subjectively rated between 1 and 10, with a rating of 1 representing very low levels of the parameter present in the board’s on-snow performance (and 10 the opposite). Furthermore, a secondary rating between 1 and 10 of the user’s perceived ideal level of the parameter was given, to allow determination of whether the board exhibited too little, too much or the correct amount of each parameter, and if applicable, the margin by which the board was sub-optimal. An importance rating between 1 and 10 was also sought for each parameter to give them a relative weighting. Tables 1-3 below display the importance weightings, user ideal levels and test board ratings for both freeride and freestyle boards. The versatile test board ratings are also shown for completeness.
Table 1 - Freeride Ratings.
Table 2 - Freestyle Ratings.
Table 3 - Versatile Ratings.
2.2 – Quantitative Analysis The quantitative parameters used in the research were based primarily on the ASTM Standard F1107-1995 – Standard Terminology Relating to Snowboarding, although several other objectively measurable parameters relating to material properties were defined to cover all relevant aspects of the snowboard design. All parameters were measured in the laboratory or obtained from published data sheets for each of the test boards purchased, and form a strong basis for the Quality Function Deployment (QFD) methodology described in the following section. Table 4 shows the key quantitative data collected in
68 The Engineering of Sport 7 - Vol. 1 this research. Note that the edge sharpness data is displayed as a range due to the ability of the rider to modify the edge angle at any time through tuning and detuning.
Table 4 - Quantitative Data.
3- Benchmarking Analysis Using Quality Function Deployment In order to identify the preferred design features for each of the riding styles based on fulfilling the identified customer requirements, a Quality Function Deployment (QFD) method was used to process the information collected. Using various correlations between the subjective and objective parameters (on a scale from strong positive to strong negative), the QFD analysis identified the key objective design features associated with the desired “feel” of the board. The graph shown in Figure 1 below compares the resulting parameter importance values for the two major riding styles calculated by the QFD processor. It was noted that the body stiffness and body material parameters were significantly more important to both styles than any other parameter, implying that the bending stiffness distribution and mass of the main body section are crucial in any optimal feel design. These parameters were also slightly more important to freestyle designs (as was the self-weighted camber), whereas the major length parameters and sidecut radius were considerably more important to freeride designs than their freestyle counterparts. These
The Engineering of Sport 7 - Vol. 1
69
results were unsurprising given that stability was the paramount consideration to freeride designs (length/sidecut and stability are strongly related) whilst forgiveness, manoeuvrability and board liveliness were the key user rated considerations for freestyle designs (all highly dependent on camber and stiffness).
Figure 1 - QFD Results Comparison.
4- Snowboard Design Innovation Opportunities A comprehensive gap analysis has been completed to identify possible design innovation and product development opportunities for modern snowboards. Using the ratings from the first survey of board models manufactured between 2004 and 2007 (40 out of the 67 total different models), the snowboard’s cumulative performance under the prescribed qualitative headers was plotted against its published style, within a range between pure freeride and pure freestyle. The performance measure for each model was calculated using the weighted average of ratings, compared to the ideal levels of each parameter within the prescribed style, as follows: Performance = 1/Average (Importance x_Parameter Rating – Ideal Parameter Level_) (1) The averages were reciprocated to give low ratings for high average differences from the ideal levels, and the results normalised to fit a scale between +5 and -5. Considering that the ideal levels for both styles were unique, for snowboard models located between the pure freeride and pure freestyle endpoints of the spectrum, a cumulative weighted difference was used to compare ratings with both sets of ideal levels. The resulting Market Opportunity Map (MoM), shown in Figure 2 aimed to identify gaps in the overall snowboard market with respect to the freeride-freestyle riding style domains.
70 The Engineering of Sport 7 - Vol. 1 The MoM showed that the performance levels of snowboard models manufactured between 2004 and 2007 were highly variable across the entire spectrum, involving both low and high performing pure freeride and freestyle models. It was noted that there were practically no high performing versatile boards at present (board models half-way between freeride and freestyle), or in other words, models that satisfy both sets of ideal levels. This result appeared at odds with the desires of modern snowboarders identified through the various surveys and interviews. It was readily apparent that riders desired versatile boards that exhibit high level performance within both styles. This confirmed that a gap in the current snowboard marketplace exists, which provides potential design innovation opportunities for high performing, versatile snowboards.
Figure 2 - Market Opportunity Map.
In order to realise the identified design innovation opportunity, it was important to pinpoint the objective design parameters that affect the versatility of snowboards. This has been achieved using a combination of existing qualitative and quantitative data. A versatility value was formulated as a measure of the extent a variation in an objective design parameter will drive the feel and performance from freestyle to freeride or vice versa. It was defined as the product of the average Relative Importance (RI) factor from the QFD chart and the normalised range of objective technical assessment data between the three test boards. In simpler terms it combines how important each factor is to the feel and performance of a snowboard and the variation of that factor between freestyle and freeride designs. Versatility Value = Avg. Relative Importance x Normalised Range
(2)
The Engineering of Sport 7 - Vol. 1
71
Figure 3 - Versatility Values.
The graph above (Figure 3) shows how the Versatility Value varies with each objective parameter. It was noted that several features appear to be crucial to the versatility of a modern snowboard. The self-weighted camber, bending/torsional stiffness in the body and the body stiffness/weight ratio all possessed exceptionally high values, however the low value of the mass parameter indicates that stiffness is of key concern as the mass does not vary to any significant extent between the test boards. Furthermore, the source of the high values varies between the three design features. The bending stiffness value was primarily the result of a very high Relative Importance value from the QFD chart, indicating its importance to the overall feel for both styles. The normalised range value was only of the order of 20%, implying that small changes in stiffness result in strong feel and performance variation. The camber and torsional stiffness on the other hand showed the opposite trend, where between the test boards, the normalised ranges were approximately 45% and 40% respectively, and the Relative Importance values were notably lower. Overall, varying the bending and torsional stiffness distributions and the camber would appear to be the key approach to altering the feel and performance of a snowboard across the major riding styles. Thus a design opportunity exists to create a versatile board design based on the described approach that would potentially result in higher levels of customer satisfaction.
4- Conclusion The research presented in this article identified bending and torsional stiffness distributions as well as camber as the key design characteristics influencing the “feel” and performance of the snowboard for both freeride and freestyle riding styles. Furthermore, the
72 The Engineering of Sport 7 - Vol. 1 use of a Quality Function Deployment method to analyse and correlate objective and subjective parameters has enabled the identification of key design parameters that have the potential to meet customer requirements better than the existing snowboard designs. This analysis also provided a better understanding of the relative importance of different snowboard features for different riding styles in relation to the identified customer requirements. The formulation of an overall performance measure using the qualitative data collected has allowed the ranking and comparison of existing board models between the formalised styles via a Market Opportunity Map (MoM), which identified gaps within the current snowboard marketplace. The MoM confirmed that there is a potential design innovation opportunity for high performing, versatile snowboards. In order to realise this opportunity for a novel snowboard design, the key objective parameters that drive versatility were identified using a Versatility Value, which was derived in this research using a combination of objective data ranges between test boards and relative importance values of each parameter obtained from the QFD analysis. Overall, the analysis has paved the way for the creation of new high performance, versatile snowboard designs which should satisfy the identified market opportunity and customer requirements.
5- References [A1] ASTM International. Standard Terminology Relating to Snowboarding. Annual book of ASTM Standards, F1107-1995. [G1] Graf Snowboards. Design, http://www.grafsnowboards.com, 2006. [HP1] Harrer M., Pfeffer P. and Johnston N. Steering Feel – Objective Assessment of Passenger Cars Analysis of Steering Feel and Vehicle Handling. FISITA World Automotive Congress 2006, Yokohama, Japan, October 2006. [M1] MaxLifestyle.net. ABC-of-Snowboarding. http://www.abc-of-snowboarding.com, 2006. [RJ1] Roberts J.R., Jones R., Rothberg S.J., Mansfield N.J. and Meyer C. The feel of a golf shot: a major factor in golf equipment selection. The Impact of Technology on Sport, pp. 20-27, Australasian Sports Technology Alliance Pty Ltd, Melbourne, 2005. [S1] Shah K. From Innovation to Firm Foundation in the Windsurfing, Skateboarding and Snowboarding Industries. 6th International Conference on Sports Engineering, Munich, Germany, July 2006. [S2] Snowsports Industries America. US Ski and Snowboard Industry Retail Audit Topline Report, March 2006. [SC1] Subic A., Clifton P., Beneyto-Ferre J. Identification of innovation opportunities for snowboard design through benchmarking. Sports Technology, Vol 1, No 1, 2008.
Business Process Modelling and its Use Within an Elite Training Environment (P15) Laura Justham1, Sian Slawson1, Andrew West2, Paul Conway2, Michael Caine1, Robert Harrison2
Topics: Sailing/Water Sports, Modelling. Abstract: In this paper the Computer Integrated Manufacture Open System Architecture (CIMOSA) business process modelling approach has been used in the development of a module to support coaches and athletes in swimming. A review of existing modelling approaches was carried out and it was decided that the CIMOSA approach was the most appropriate for this application as it is a stand alone, integrated technique which can be used in both system planning and analysis. The requirements specification for the coaching support module was developed based on findings from interviews with elite level coaches, swimmers and support staff. It has been represented in this paper using the four CIMOSA modelling constructs from the high level context diagram to the low level activity diagram. Processes and activities occurring within the development of training schedules for elite athletes have been identified and the flow of information and resources have been formally documented. The results obtained from the development of these structured diagrams have been used to design an aid to coaching which allows the coach to develop, monitor and record all aspects of a swimmer’s training diary. It is envisaged that such a system could help to revolutionise the management of an athlete’s development and ensure training is effective and focused towards an individual’s needs. Key words: Enterprise Modelling; Systems Integration; swimming; elite athlete monitoring; requirements specification.
1- Introduction Swimming is a popular sport and is regularly enjoyed by approximately 20% of the population of the UK with individuals taking part for enjoyment, to reap the health benefits associated with regular exercise, or for competition at various standards from local club galas to international events (Mintel 2005). There are over 1200 swimming clubs, that are affiliated to the Amateur Swimming Association (ASA), in the UK (British_Swimming 2007). Within these clubs coaches are responsible for planning and 1. Loughborough University Sports Technology Institute, Loughborough Science and Enterprise Park, 1 Oakwood Drive, Loughborough, LE11 3QF, United Kingdom - E-mail: L.Justham, S.E.Slawson, [email protected] 2. Wolfson School of Mechanical and Manufacturing Engineering, Loughborough University, Loughborough, LE11 3TU United Kingdom - E-mail: A.A.West, P.P.Conway, [email protected]
74 The Engineering of Sport 7 - Vol. 1 implementing training cycles to ensure the improvement of technique and development of skills of the swimmers. Generally coaches will generate training schedules by hand, without the use of specialist computer software to develop, record, store and evaluate each session. At an elite level swimmers will train towards two or three major competitions per year, and will undertake either water or land-based training and conditioning at least once per day. As such, the coach is required to carry out detailed session planning and organisation to meet the needs of each individual swimmer within their squad. At the beginning of the year annual training goals are chosen, which include the identification of two to three major competitions. The dates of these competitions then form the basis of the training cycles undertaken throughout the year. Each training cycle is segmented into medium term goals such as identifying minor events to monitor progress and short term aims which include determining the objective of each week’s training whilst focusing on building towards the major competitions. The weekly training objectives are further decomposed into individual session aims which are then divided into the composite parts of each training session. The final stage of the training planning process is an ongoing evaluation of each training session and training cycle to monitor how the swimmer is improving. This can take the form of recording data collected during the session (i.e. time trials) or a qualitative review of a swimmer’s improvement. This complete process is illustrated in Figure 1.
Figure 1 - At an elite level coaches develop annual training plans for their swimmers that is broken down into two or three training cycle to coincide with major competitions and events. The training year is progressively broken down into shorter term aims and schedule requirements.
The purpose of the current research has been to identify and apply a suitable enterprise modelling technique for the development of a training planning and monitoring module to support both coaches and athletes throughout every stage of the training year, as identified in Figure 1. Enterprise Modelling is an established technique which has been applied across a diverse range of business scenarios, in generic and specific applications, to aid the requirements definition, design, implementation and test of a system (Monfared 2002; Rahimifard 2007). The approach is used to build a competitive advantage by fulfilling customer requirements and improving the product development process whilst maintaining economic feasibility. A complete enterprise model is comprised of a set of purposeful and complementary models which describe the various aspects of an enterprise according to specific modelling constructs and semantics (Vernadat 1996). For this application the enterprise modelling approach has been developed in line with specifications and requirements outlined by British Swimming’s elite level coaches and sports scientists who are based at Loughborough University.
The Engineering of Sport 7 - Vol. 1
75
Within enterprise modelling manufacturing paradigms have been developed. These are computational representations of reality and are used to represent the structure, processes, resources, information, goals and constraints of a business (Gruninger 1996; Fox 1998). Real-life applications are decomposed into a set of input processes which interact with the system’s components to form the required output processes. An example is the Computer Integrated Manufacturing (CIM) paradigm which is used to integrate activities within an enterprise and can be applied to a broad range of applications depending on the requirements of a specific situation (Vernadat 1996). In addition a number of different modelling frameworks have been developed, see Figure 2, and each has strengths depending on the specific requirements of an application. According to Aguilar-Saven, four categories describe the purposes of business process models within a modelling framework: (i) descriptive models for learning about a system, (ii) descriptive and analytical models for decision support, process development and design (iii) enactable or analytical models for decision support during process execution and control and (iv) enactment models for support in information technology (Aguilar-Saven 2004). These four categories make up the x-axis of Figure 2 and the y-axis separates those techniques which may be interacted with to allow changes without complete remodelling (active models) and those which may not (passive models). For the purpose of the current research an active model is required to ensure that modifications can be made as the training cycles are developed and varied, depending on the requirements of each individual swimmer. Therefore only the approaches in the top half of Figure 2 are suitable. A modelling approach which lies within each of the four purpose categories outlined by Aguilar-Saven is also required to ensure that each component of the swimmer’s training year may be adequately modelled. Therefore the Computer Integrated Manufacture Open System Architecture (CIMOSA) is the only suitable approach to provide a stand alone integrated method.
Figure 2 - An example of some of the available Business Process Modelling Frameworks and how they may be used in a practical application (adapted from (Aguilar-Saven 2004)).
76 The Engineering of Sport 7 - Vol. 1
2- Design of the Business Process Model for Coaching The representation of the training planning module was developed using graphically represented CIMOSA models, based on an abstraction method developed at Loughborough University (Aguiar 1995). This involves the development of four sets of diagrams which represent each level of decomposition. The context diagram comprises the first stage in the decomposition process. The purpose of this diagram is to identify the domains which are fundamental to the development of a focused training cycle. Five domains have been identified and are planning of the macro-cycle (MP1), detailed weekly planning (MP2), focused session planning (MP3), development of session summaries (MP4) and weekly summaries (MP5), as illustrated in Figure 3. These five domains are based upon the scheduling requirements of coaches, as outlined in Figure 1. For example, included within the macro-cycle planning domain, MP1, are the identification of annual training goals (major and minor competitions) and the development of more focused training plans.
Figure 3 - The Context diagram which has been used to identify the fundamental domain processes required for the development of a structured training cycle in swimming.
The interaction diagram is the second stage in the decomposition and is used to identify how the domains are related to one another. It is concerned with the flow of information and resources as well as events occurring within and between the domains. Information about the swimmer’s training goals, which are identified within MP1, are transferred to MP2 (week by week session planning) and subsequently MP3 (individual session planning) to assist the coach in developing balanced and progressive training sessions. Target and actual outcomes from these three domains are then transferred into MP4 and MP5 which are concerned with recording a summary and evaluation of the outcomes of training sessions, see Figure 4, to monitor progress and ensure the original training goals are being fulfilled.
The Engineering of Sport 7 - Vol. 1
77
Figure 4 - The interaction diagram, used to detail events occurring and the flow of resources and information between the domains identified within the concept diagram.
The structure diagram comprises the third level of the decomposition procedure and is used to determine the core processes and activities occurring within each domain. The domains are first divided into processes, see Figure 5, and are then further decomposed into activities (not shown in the diagram). These processes and activities represent the design and functionality required to detail each domain in the system. In MP1 the development of long term training goals is the only process, but it is vital to ensure that the training year is divided to allow sufficient time to recover from one event, work on areas of weakness to improve technique and prepare for the next event between major competitions. Within MP2 the processes are focused on shorter term training aims from a weekly perspective. This includes activities such as identifying when sessions will run, the focus of each session and when the swimmer should rest. Within an individual swim session there are three processes: (1) the warm-up, (2) the main set and (3) the recovery or cool down. The activities contained within these processes include the distance to be swum, the allowed time for each set, holding or target times to be met by the swimmer, details of kick, pull or drill sets and any additional equipment which is required. In addition to swim sessions there are also other types of training sessions, which make up the fourth process within MP3. These sessions include dry-land conditioning such as circuit training or weights workouts and must be planned and evaluated in a similar manner to the water-based training. The processes in MP4 and MP5 are similar, and are concerned with documenting session outcomes and evaluating the performance of a swimmer. Activities such as recording total distances swum, recording split times from key sets within the session, making note if a swimmer were feeling unwell or had to get out early, recording the filename of any videos and recording any generic feelings about a session. The final diagrams of the CIMOSA decomposition process are the activity diagrams which are developed to show the time-based flow of information for each domain process. As these are low-level diagrams they provide detailed information for each process in turn, which leads to numerous separate diagrams being developed. For the
78 The Engineering of Sport 7 - Vol. 1 purpose of the current paper only the activity diagram for MP3 has been shown, see Figure 6. The longer term training goals of MP1 and the weekly training aims of MP2 are used by the coach to identify the requirements of an individual session. The session is then broken up into the warm-up, main set and recovery segments and within these information such as the number of repeats of a set, the number of sets, target times and any additional equipment is recorded. Each swimmer is provided with a personalised training plan and at the end of the session a breakdown of the distances swum and any significant outcomes are recorded into MP4 and MP5.
Figure 5 - A compacted version of the structure diagram, which is the third stage in the decomposition process, developed to include the domain processes. Each process may be further decomposed into the individual activities contained within them.
Figure 6 - The activity diagram for MP3 (Individual session planning). These diagrams are the lowest level in the decomposition process and are used to detail the time-based flow of information within each domain. Information is fed into MP3 from the coach and from MP1 and MP2. This information is used to develop a personalised session plan for each swimmer, which is evaluated after completion.
The Engineering of Sport 7 - Vol. 1
79
3- Discussion and Conclusions Business Process Modelling is an established technique which is regularly used to aid the design and implementation of a system. For this specific application the CIMOSA modelling approach was selected as the most appropriate because it can be used during both the planning and analysis phases of an application. This is particularly important as the elite level coach is required to plan long and short term training cycles whilst carrying out day to day session development and athlete monitoring. In addition they are required to evaluate session outcomes and monitor an athlete’s progression over a prolonged time period. The CIMOSA approach has been used to represent the requirements specification and to facilitate the design of a coaching support module which aids the coach in each aspect of his job. The Context Diagram has been used to formalise the high-level domains which form the basis of the training year. These have been decomposed into an interaction diagram to understand how the domains are interrelated and how events, information and resources are passed within and between them. The structure diagram has been used to identify the processes and activities occurring within each domain and finally the low-level activity diagrams have been used to organise the system into timedependent sequences. The finalised module specification is under development at Loughborough University. It is envisaged that the module will be used by the coach, the athlete and auxiliary support staff. It will enable the coach to monitor and record all aspects of the swimming training diary from determining training plans to recording split times logged during a training session. It will also allow the swimmer to monitor their own progress and take a more active role in their continued development as an athlete. Finally it will facilitate support staff linking additional information to the swimmer’s profile such as physiological or medical information and results from biomechanical or video analysis carried out during training.
4- Acknowledgements The authors would like to thank the sports scientists, coaches and swimmers at Loughborough University who have taken part in the information gathering stages of this research. Funding for this work has been provided by the Innovative Manufacturing and Construction Research Centre (IMCRC) based at Loughborough University, part of the Engineering and Physical Sciences Research Council of Great Britain (EPSRC), and UK Sport.
5- References [A1] Aguiar, M. W. C. (1995). PhD Thesis: An approach to enacting business process models in support of the life cycle of integrated manufacturing systems. Manufacturing Engineering Department, Loughborough University. [A2] Aguilar-Saven, R. S. (2004). “Business process modelling: Review and framework.” International Journal of Production Economics 90: 129-149.
80 The Engineering of Sport 7 - Vol. 1 [B1] British_Swimming (2007). Homepage for the Amateur Swimming Association (ASA) and British Swimming http://www.britishswimming.org. [FG1] Fox, M. S., Gruninger, M. (1998). Enterprise Modelling. AI Magazine - American Association for Artificial Intelligence. FALL 1998: 109-121. [GF1] Gruninger, M., Fox, M.S. (1996). The Logic of Enterprise Modeling. IFIP TC5 Working Conference on Models and Methodologies for Enterprise Integration, Queensland, Australia, International Federation for Information Processing. [J1] Justham, L. M. (2007). The Design and Development of a Novel Training System for Cricket. Loughborough, PhD Thesis Awarded from the Wolfson School of Mechanical and Manufacturing Engineering, Loughborough University. [JW1] Justham, L. M., West, A.A., Cork, A.E.J. (2007). “Use of the Quality Function Deployment Methodology in the Development of a Novel Training System for Cricket.” Submitted to the IMechE Part B - Engineering Manufacture. [M1] Mintel (2005). The Mintel Corporation UK Sports Participation Report July 2005 http://www.mintel.com [MW1] Monfared, R. P., West, A.A., Harrison, R., Weston, R.H. (2002).“An implementation of the business process modelling approach in the automotive industry.” Proceedings of the Institute of Mechanical Engineers Part B: Journal of Engineering Manufacture 216: 1413-1427. [RW1] Rahimifard, A., Weston, R. (2007).“The enhanced use of enterprise and simulation modelling techniques to support factory changeability.” International Journal of Computer Integrated Manufacturing 20(4): 307-328. [V1] Vernadat, F. B. (1996). Enterprise Modelling and Integration: Principles and Applications, Chapman and Hall.
Accelerometer Profile Recognition of Swimming Strokes (P17) S.E. Slawson1, L.M. Justham1, A.A. West2, P.P. Conway2, M.P. Caine1, R. Harrison2
Topics: Sailing/Water Sports, Measurement Systems. Abstract: The use of technology in sports performance analysis is a rapidly increasing practise. Tools for analysis aim to provide useful information to supplement coach knowledge and improve feedback in the development of athletes. In swimming the use of subjective video analysis is wide-spread, however, unlike some other sports, there are few quantitative measures of performance. Quantitative measures, such as intra cyclic variations of stroke characteristics, have the potential to provide more specific performance metrics from which to make improvements. Such measures are currently not widely available to coaches, support staff and swimmers, due to the infancy or lack of sufficiently developed technologies. The research outlined in this paper has explored the relationship between stroke characteristics and how they are represented by accelerometer data. Each of the four competition swimming strokes has been investigated, where swimmers instrumented with a portable sensor were asked to perform their normal swimming strokes. During each swimming trial, sensor and video data were recorded. Preliminary testing has shown that accelerometer data can be useful in the determination of simple stroke characteristics, for example stroke rate and duration, and that differences in profiles can be attributed to a certain stroke or swimmer. Future work is required to broaden this understanding and support the outcomes already generated from this preliminary study. Key words: swimming, accelerometer profile, stroke recognition.
1- Introduction Performance monitoring and analysis in sport plays a key role in the development of athletes at any level. Swimming is a discipline built from a number of skill sets and movement profiles, which together contribute to overall performance. The ability to scrutinise each of these skills in detail could enable significant improvements in performance to be made. In swimming, performance analysis techniques are predominantly visually1. Loughborough University Sports Technologies Institute Loughborough Science and Enterprise Park 1 Oakwood Drive, Loughborough, LE11 3QF, United Kingdom - E-mail: [email protected]; [email protected] 2. Wolfson School of Mechanical and Manufacturing Engineering, Loughborough University, Loughborough, LE11 3TU, United Kingdom - E-mail: [email protected]; [email protected]
82 The Engineering of Sport 7 - Vol. 1 based and require post processing of recorded images to derive quantitative and qualitative measures of performance. There is currently a gap in direct, (i.e. non derived), quantitative analysis methods developed specifically for swimming performance. Analysis tools based upon the integration of distributed sensors e.g. accelerometers may be able to give valuable information regarding the breakdown of swimming strokes and their performance related characteristics ([DA1], [DJ1], [IO1], [JD1], [O1], [OY1], [OI2], [T1]). These methods are not yet widely used amongst the swimming community as tools for analysis. This can be attributed to a combination of lack of development and refinement of the technologies used and a lack of understanding regarding how to process the data into a useful and useable format. Preliminary testing has been undertaken to start to address these gaps in knowledge. Testing was carried out to establish the types of relationships found between accelerometer trace characteristics and swim stroke characteristics. Results were broken into two areas of interest addressing two research questions: – Can accelerometer data be used to identify swimming stoke characteristics such as stroke count and stroke duration? – Can accelerometer trace characteristics be attributed to a certain stroke or swimmer?
2- Experimental Protocol Two swimmers were instrumented with a three-axis accelerometer (Freescale Semiconductors) attached to the small of the back with the axes oriented as marked in Figure 1. In the prone position the x-axis represented forward motion, the y-axis was sideways roll and the z-axis was up and down undulation. Swimmer 1 was a university swimmer of national competition standard and swimmer 2 was a national team squad swimmer of international competition standard. A Panasonic AG-HV200E camera was used to digitally record both swimmers performing each of the four competition strokes that were performed in an individual medley format (IM), Figure 2. Data from the accelerometer was sampled at 100Hz and downloaded to a central computer for post processing.
Figure 1 - Orientation of accelerometer on the swimmer.
The Engineering of Sport 7 - Vol. 1
83
Figure 2 - Testing set up.
The three axes of acceleration were measured in relation to gravity (g) however the orientation of the sensor during the trial was not known, due to the absence of an inclinometer or similar complimentary sensor. Therefore the influence of gravity on the sensor varied throughout the trials. Despite this limitation the same accelerometer unit and the same set up for both trials was used to ensure consistency in testing such that relative measures of acceleration could be confidently assumed.
3- Results Video and accelerometer data were synchronised and comparisons were made between the stroke count and duration as measured from the video and accelerometer data. Data from the x-axis was used to evaluate the butterfly and breaststroke and from the y-axis for backstroke and freestyle, as a strong cyclic pattern was found in the generated data. Stroke rate was distinguishable on the accelerometer traces as cyclic patterns of acceleration, see Figure 3. For each trial the stroke count per length was measured using manual observation of the video and cyclic counting techniques from the raw accelerometer data. Total stroke counts measured on the video and accelerometer were found to be equivalent throughout all trials. Stroke duration was measured for both swimmers. A distinct point in the stroke cycle was identified, e.g. hand entry for butterfly, and used to measure the start of each new stroke when analysing stroke duration from the video data. Accelerometer data was divided similarly using recurrent peaks to establish stroke cycle, Figure 3.The average difference between video and accelerometer measured stroke duration was 0.056s and 0.077s for swimmer 1 and 2, respectively. This equated to an overall average of 0.067s, or less than two frames difference, when operating a standard video camera, i.e. running at 25 frames per second (fps).
84 The Engineering of Sport 7 - Vol. 1
Figure 3 - Graph showing backstroke acceleration profile, broken into individual strokes using cyclic patterns for recognition.
Further analysis was carried out to establish whether the shape of the accelerometer data profile could be related to a given stroke or swimmer. Key areas of interest were identified as: – Magnitude of acceleration (amplitude) –Duration of stroke (wavelength) – Range of acceleration values – Standard deviation of accelerations – Profile shape The raw accelerometer data shown in Figure 4 demonstrates with clarity the variability of the four IM strokes in terms of acceleration characteristics. Each stroke can be identified by a unique combination of x,y,z accelerations from the raw data collected. Similar comparisons can also be made by considering differences in profile parameters for individual axes of motion, when considering the magnitude of acceleration, stroke rate and major axis of acceleration.
Figure 4 - Graph representations for accelerations in x for breaststroke and freestyle trials. ‘Typical profile pattern’ shown to the right of the graphs.
The Engineering of Sport 7 - Vol. 1
85
The variability within stroke acceleration characteristics between the different swimmers was also considered. Using two examples from the trial, comparisons of each swimmer’s stroke characteristics were analysed and used to provide assumptions about their technique. The analysis of accelerometer data for the butterfly trial showed that swimmer 1 produced a standard deviation of 4.85 m/s/s, conversely swimmer 2 had a lower standard deviation of 3.96 m/s/s, which signified a smaller variability in intra-stroke acceleration in the forwards direction (x-axis). This may imply that swimmer 2 has a smoother stroke technique. In addition, this suggests that swimmer 1 experiences a greater contrast in forward acceleration during the stroke cycle whereas swimmer 2 presents a more constant acceleration throughout the stroke. A conclusion that may be drawn from this, is that swimmer 2 demonstrates a greater ability to maintain propulsion during the stroke whereas swimmer 1 suffers from large decelerations during the recovery phase implying a less efficient stroke technique. Breaststroke trials showed strong similarities between the two swimmers, with respect to the range of accelerations being measured and the standard deviation values (28.14 m/s/s and 26.63 m/s/s for the range of acceleration measured and 2.95 m/s/s and 2.87 m/s/s for the standard deviations for swimmer 1 and 2 respectively). This indicates a greater consistency in their stroke characteristics for breaststroke from an acceleration perspective. Graphical representation of the two traces, see Figure 5, has also demonstrated a clear correlation in the acceleration profile shape. Despite these close parametric similarities the two swimmers may be distinguished from one another by considering variables such as the stroke duration and the magnitude of accelerations. The increased convergence in acceleration data for breaststroke specifically may be attributed to it being the preferred stroke for swimmer 1 and therefore their standard in this discipline may be closer to the generally higher overall standard of swimmer 2.
Figure 5 - Graph showing a comparison of breaststroke traces for two different swimmers.
86 The Engineering of Sport 7 - Vol. 1
4- Conclusions The results presented within this paper have been taken from testing that was carried out to establish the types of relationships found between accelerometer trace characteristics and swim stroke characteristics for different swimmers. The results have been broken into two areas of interest addressing the research questions identified in the introduction. Accelerometer traces established from IM trials were analysed with relation to these research questions. It was found that data collected for both swimmers in all trials could be used to extract simple stroke characteristics, namely, stroke rate and stroke duration. The resolution of these measures was found to be within 0.067 seconds of equivalent measures taken from manual analysis of the corresponding videos images. Further analysis was carried out to ascertain whether specific accelerometer characteristics could be attributed to a given stroke. Simple visual representations of data collected from the trials demonstrated clearly distinguishable characteristic variability that could be accredited to the four IM strokes. Further analysis of data is required to correlate phases of the stroke cycle with patterns in acceleration cycles. This may be used to better determine features that can be exclusively qualified to a specific stroke. Comparisons of data were made between the two swimmers for each of the IM strokes. Profiles for each swimmer were distinguished from one another by considering variables such as magnitude of acceleration, stroke duration and standard deviation of accelerations. Discrepancies between the data collected for each swimmer may correspond to differences in competency in terms of technique and can help to provide information from which improvements can be made in the future. Preliminary testing has shown that accelerometer data can be useful in the determination of stroke characteristics. Future work is required to broaden this understanding and support the assumptions generated from this preliminary study. Development of hardware such that orientation of the sensor is known should also be addressed to enable a more detailed understanding of data produced. Broader testing with a larger number of swimmers, of all standards, will also generate further confidence in any conclusions which have been drawn.
5- Acknowledgements The authors would like to thank the sports scientists, coaches and swimmers at Loughborough University who have taken part in the information gathering stages of this research. Funding for this work has been provided by the Innovative Manufacturing and Construction Research Centre (IMCRC) based at Loughborough University, part of the Engineering and Physical Sciences Research Council of Great Britain (EPSRC), and UK Sport.
6 - References [DA1] Davey NP, Anderson ME, James DA. An accelerometer based system for elite athlete swimming performance analysis. Proceedings of SPIE – Smart Structures, 5649(1) : 409-415
The Engineering of Sport 7 - Vol. 1
87
[DJ1] Davey N, James DA, Zhang K, Grenfell R, Mackintosh C, inventors. Monitoring Sports and Swimming. 2006 19/10. [IO1] Ichikawa H, Ohgi Y, Miyaji C. Analysis of stroke of the freestyle swimming using accelerometer. Biomechanics and Medicine in Swimming VIII 1999 28/06-2/07 1998. [JD1] James DA, Davey N, Rice T. An Accelerometer Based Sensor Platform for In-situ Elite Athlete Performance Analysis. Sensors 2004, Proceedings of IEEE 2004;3:1373-1376. [O1] Ohgi Y. Microcomputer-based Acceleration Sensor Device for Sports Biomechanics ~ stroke evaluation using swimmers wrist acceleration. Proceedings of IEEE, Sensors 2002 2002;1:699704. [OY1] Ohgi Y, Yasumura M. Analysis of stroke technique using acceleration sensor IC in freestyle swimming. Engineering of Sport 2000. [OI2] Ohgi Y, Ichikawa H, Miyaji C. Characteristics of the forearm acceleration in swimming. Biomechanics and Medicine in Swimming VIII 1999 28/06-2/07 1998. [T1] Traqua – Swimming Computer Technology. Available at: http://www.crca.asn.au/resource_materials/reapingbenefits.pdf. Accessed 10th January, 2007.
Evaluation of Start Techniques in Sports Swimming by Dynamics Simulation (P18) Thomas Härtel1, Axel Schleichardt2
Topics: Biomechanics, Modelling, Performance Sports, Virtual Reality & Computer application in Sports. Abstract: Winning and losing times in swimming competitions sometimes differ by only a few hundredths of a second. Marginal improvements in executing the start phases may give a significant advantage. Two basic techniques are performed: grab and track start including significant parameters like reaction time, take-off velocity, flying width etc. Modern methods of modelling and simulation of human motion supply new knowledge on efficiency in execution of the start process. The model of a swimmer describes the essential anthropometrical and kinematical properties. The motion is simulated based on video capturing including the free flight phase. The differences of the techniques, advantages of high level execution and the influence of the anthropometric data are shown in a comparison of animations of the motion and the parameter based evaluation as well as time histories of joint angles. Keywords: Biomechanics, Simulation, Modelling, Swimming, Start.
1- Introduction The performance development in sports swimming shows tendencies of decreasing competition time and lags between the athletes. Slight differences are strongly influenced by start and turning techniques (Küchler 1994). This non-cyclic and springiness oriented phases of motion are crucial factors for lost and won races. An increasing role is assigned to the configuration of the start motion of the athlete (Küchler and Graumnitz, 2006, Vilas-Boas et al. 2000, Breed and McElroy, 2000, Krüger et al. 2003). Disadvantages of the swing start are described more then 30 years ago (Lowell 1979, Roffer 1972).
1. Institute of Mechatronics, Reichenhainer Str. 88, 09126 Chemnitz, Germany - E-mail: [email protected] 2. Institute for Applied Training Science, Marschnerstraße 29, 04109 Leipzig, Germany E-mail: [email protected]
90 The Engineering of Sport 7 - Vol. 1
Figure 1 - Start techniques.
The currently favoured start techniques differ in the initial position. In case of the grab start the motion is characterized by a symmetric motion of legs. The track start is distinguished by a forward or backward shifted body mass and asymmetric initial position of the legs (see figure 1).
2- Modelling Based on the multibody simulation tool alaska, an integrated development environment for modelling of mechanical systems (alaska 2007), a specific modelling interface for the analysis of start techniques in sports swimming has been developed. The functionality of alaska is applied in biomechanics in sports, rehabilitation, ergonomics, as well as mechanical and automotive engineering and robotics. The three-dimensional biomechanical human model DYNAMICUS (2007) has been used for this investigations. It consists of a highly granular template library of parts of the body including joints, anthropometrical data set, joint limits, damping properties etc. These elements are used to set up the model and to simulate the motion. The hierarchical structure of the model includes the components head and cervical, trunk, left and right arm with hand, left and right leg as well as left and right foot. Every component consists of rigid bodies connected by joints. The coupling between the components is realized by joints in a selectable granularity (Härtel and Hermsdorf, 2006). The anthropometric data are computed either with respect to total mass, length and gender based on regression formula or with a body segment measurement method for length, width and circumference. All necessary model parameters like length and inertia properties of the segments are used to define an individual anthropometric data set of swimmers and to define a database of swimmer properties. The swimmer model contains 21 body segments with a maximal total degree of freedom of 57. The environment is parameterized to adapt the different calibration conditions and compare different executions in the same reference frame. It is embedded in a special simulation environment for application in sports swimming. The user can apply predefined models for grab or track start technique, choosing a 2D- or 3D- motion and only has to specify the anthropometrical data set. All essential parameters for the simulation of the motion can be defined in a simple way using the input dialogs. Modelling is supported by a visualisation of components in different kind of characteristics in an arbitrary combination (see figure 2).
The Engineering of Sport 7 - Vol. 1
91
Figure 2 - Human body model with body markers.
3- Pre-processing The real motion of the swimmer is captured by digital cameras and analysed from the beginning with the start signal to the touch of the hands on the water surface. Special marker points on joints or specific points of the body surface are labelled. For the planar analysis in case of grab start 10 motion marker points and in case of track start 13 marker point are used (see figure 3), while up to 19 marker points for a 3D-motion can be analyzed. Using the measure and analyse program “Mess2DDV” of the IAT, the time history of the marker positions with respect to the inertial reference frame are computed (Drenk and Hildebrand, 2002). For a calibration of the motion area special markers of the environment are utilized. In a following process the motion markers and body fixed markers are joined. For every motion marker a weight parameter exists and joint limits bound the range of motion in every joint. In this process of inverse kinematics the minimum of the distance between all motion and body markers will be computed in an optimization process. As a result of this pre-processing the reference motion of the inner joint angles of the human model and the time history of motion in the inertial reference frame of the body pelvis can be extracted.
92 The Engineering of Sport 7 - Vol. 1
Figure 3 - Motion and body markers.
Additional knowledge for the analyses of start techniques is generated by the reaction forces at the blocks (Kibele et al. 2007). Therefore, in addition to video based analyses a dynamometric block (see figure 4) was applied. With usual dimensions and a forward tilted surface of 5°, the block, developed at the IAT, is able to measure vertical and horizontal forces separately for each leg and total vertical forces of the hands (Knoll and Wagner, 2006, see figure 4). The measured forces are represented as vector arrows in the simulation environment and serve as comparison to the computed ground reaction forces.
Figure 4 - Starting block dynamometer.
4- Simulation Process and Results For the predefined time of take-off the best possible initial conditions of flight phases are determined by multicriterial optimization with regard to minimal deviation between motion markers and body markers. The simulation of the entire process is based on a dynamic control which guides the actual motion to the reference motion of joint angles.
The Engineering of Sport 7 - Vol. 1
93
During the whole simulation, except the flight phase, the pelvis is dynamically controlled in the inertial reference frame by a so-called Lyapunov-stable control law for underactuated systems (Härtel et al. 2006). For the evaluation of time parameters the block time (from the start signal to the take-off) and the flight time (from take-off to initial water contact) is used. Basic simulation results of every body segment are position, angular velocity, linear and angular momentum etc. Motion specific results are system parameters like the velocity and angle of take-off, jump width and the position of the center of mass. Other results are the time history of linear and angular momentum, angular velocity and all joint angles, which are very useful for motion analyses. The time history of the results is given by curves or text files. Figure 5 shows the graphs of knee and hip angles of two track starters. There are two reasons for swimmer S1, represented by the solid curve, to achieve a shorter block time versus swimmer S2 (dotted line) with the same reaction time (time between start signal and first movement of the swimmer): First, swimmer S1 starts flexing the knee joint of his front leg immediately and secondly, a faster stretching of the back leg knee joint. Furthermore, the greater amplitudes in the knee and hip of swimmer S1 provide a higher take-off velocity.
Figure 5 - Hip and knee joint angles of track start variants (solid line conforms to a high level execution)
Further graphical topics are the representation of the environment, different kinds of visualisation of the swimmer, several views with fixed or moving camera and visualisation of mechanical parameters like total angular velocity and angular momentum. The trajectory of body points or the total center of mass can be represented in a spatial track in the animation. It is possible to shown additionally the projection point on the water
94 The Engineering of Sport 7 - Vol. 1 surface of the hand and of the center of mass. The motion of several swimmers can be compared using a special animation viewer in order to explain the different techniques of the swimmers and the execution quality of top level swimmers. Figure 6 is representing the comparison of male swimmers in the initial position and the time point of the first water contact. Each two swimmers use the grab start, track start with a forward or a backward shifted body mass. The point of first water contact differs up to 0,85 m. Angle and velocity of the take-off differ significantly. In this comparison a swimmer with a track start and a backward shifted body mass reached the best start parameters and flight width.
Figure 6 - Comparison of national and top level start techniques.
To analyse the influence of take-off parameters or segment motion a method to manipulate the movement in the flight phase is implemented. The velocity and angular velocity of the body pelvis, the total linear momentum or total angular momentum may be given explicitly by value as initial conditions of this phase. Furthermore, it is possible to change the motion of selected body segments, e.g. the leg or the arm.
5- Conclusion The application-specific simulation environment enables the user to carry out an efficient analysis of different start techniques without specific modelling and simulation knowledge. Many results of the simulation allow a comprehensive analysis of the start execution. The tool offers several options for manipulating the conditions in the start phase in sports swimming. So the user can easily analyse the different interrelations of biomechanical parameters during the start phase. Variations of initial parameters of the take-off generate new knowledge for the understanding of dominating factors. Animations can be used in cooperation with the swimmer to explain cause-effectconnections in the execution. Analyses show, that both techniques lead to top start performances (Graumnitz et al. 2007). The asymmetric track start requires higher movement coordination. On the other hand for the horizontal acceleration and shorter block times the initial position, based
The Engineering of Sport 7 - Vol. 1
95
on the placed back leg, is more beneficial. Furthermore the effort of the back leg can be used to produce a larger vertical impulse and advanced angular momentum. Even so there cannot be a recommendation for one or the other technique unless taking individual physical conditions of the swimmer into account. This project is supported by the Bundesinstitut für Sportwissenschaft (BISp), Bonn, Germany.
6- Reference [A1] alaska 5.1. Reference Manual, Institute of Mechatronics, Chemnitz, 2007. [BM1] Breed, R. V. P. and McElroy, G. K. A biomechanical comparison of the grab, swing and trackstarts in swimming. Journal of Human Movements Studies, 39, 277-293. 2000. [DH1] Drenk, V. and Hildebrand, F. Plane-based camera calibration for 3D-Videogrammetry for canoeing and rowing. Proceedings of the XX. International Symposium on Biomechanics in Sports, 349, 2002. [D1] DYNAMICUS. Reference Manual, Institute of Mechatronics, Chemnitz, 2007. [GK1] Graumnitz, J., Küchler, J. and Drenk, V. Greifstart oder Schrittstart - Fakten und Tendenzen aus Analysen bei internationalen Meisterschaften im Sportschwimmen. In W. Leopold (Hrsg.), Schwimmen: Lernen und Optimieren, Band 28. (S. 90-101). Beucha: Deutsche SchwimmtrainerVereinigung e. V., 2007 [HH1] Härtel, T. and Hermsdorf, H. Biomechanical Modelling and Simulation of Human Body by means of DYNAMICUS, Journal of Biomechanics; Volume 39, Supplement 1, Abstracts of the 5th World Congress of Biomechanics, S549, Elsevier, 2006. [HH2] Härtel, T., Hildebrand, F. and Knoll, Ka. Methods of Simulation and Manipulation for the Evaluation of Figure Skating Jumps, In E.F. Moritz, S. Haake: The Engineering of Sport 6, Volume 2, Developments for Disciplines, Springer Science+Business Media, New York, pp. 179-184, 2006. [K1] Kibele, A. et al. Biomechanische Leistungsdiagnostik zum Startsprung im Schwimmen. In: Leistungssport. Heft 4, p.5157. 2007. [K2] Krüger et al. Biomechanics of the grab and track start technique. Biomechanics and Medicine in Swimming IX. Proceeding of the IX International Symposium on Biomechanics and Medicine in Swimming (pp. 219-223). University of Saint-Etienne, France. 2003. [K3] Küchler, J. Mechanische Analyse des Startabschnitts im Sportschwimmen. In W. Freitag (Hrsg.), Schwimmen lernen und optimieren, Bd. 85 (S. 73-85). Rüsselsheim: Deutsche Schwimmtrainer-Vereinigung e.V. 1994. [KG1] Küchler, J. and Graumnitz, J. Ergebnisse aus einer Wettkampfbeobachtung bei den XI.Weltmeisterschaften im Schwimmen. In W. Leopold (Hrsg.): Schwimmen ? Lernen und Optimieren,26 (S. 7-38). Rüsselsheim: DSTV. 2006. [KW1] Knoll, K. and Wagner, K. Requirements and Solution Concepts fort he Development of Sport-Specific Measuring Units in High Performance Sports, In E.F. Moritz, S. Haake: The Engineering of Sport 6, Volume 3, Developments for Innovation, Springer Science+Business Media, New York, pp. 295-300, 2006. [L1] Lowell, J.C. Analysis of the grab start and the conventional start. Swimming Technique. Los Angeles 12 (1979), S. 66-69. [R1] Roffer, N. The grab start is faster. Swimming Technique. Los Angeles 8 (1972), S.714. [VB1] Vilas-Boas, J. P., Cruz, M.J., Sousa, F. and Conceicao, F. Integrated kinematic and dynamic analysis of two track-start techniques. Proceedings of XVIIIth International Symposium on
96 The Engineering of Sport 7 - Vol. 1 Biomechanics in Sports: Application of biomechanical study in swimming, Hong Kong / Department of Sports Science and Physical Education (Ed.), S. 113 – 117. 2000.
A Simulation of Outrigger Canoe Paddling Performance (P19) Nicholas Caplan1
Topics: Modelling; Biomechanics; Sailing/water sports. Abstract: The main performance variable in outrigger canoeing is boat velocity. The aims of this study were to derive a model of the mechanics of the outrigger canoeing (O1) stroke and to perform a sensitivity analysis to determine the key variables for increasing boat velocity. The model considered the fluid dynamic interaction between the paddle and water, the oscillatory motions of the paddler’s upper body relative to the boat, the mass of the paddler and boat, and stroke rate. Due to the identical hulls between the O1 and K1 boats, measured hydrodynamic drag relationships for a K1 kayak were used. The aerodynamic drag acting on the paddler was assumed to be similar to that of a seated man. The model was solved using a fourth order Runge-Kutta solver with a fixed time step of 0.001s. The model was validated against the mean boat velocity for an elite female paddler over a 500m race distance. The paddler was filmed in order to determine the linear and angular trajectories of the paddle and the paddler’s centre of mass through the stroke. A sensitivity analysis was then performed by increasing/decreasing each input variable by 1%. A difference of less than 3% was observed between modelled and measured mean boat velocity showing the model to be valid. The sensitivity analysis showed that the key variable influencing boat velocity was horizontal paddle velocity relative to the boat. The degree to which this variable can be increased is influenced by the strength of the paddler. The next most sensitive variables were the paddle’s size and maximum drag coefficient. The magnitude of maximum drag coefficient is a function of the paddle design. Future research should now focus on determining the underlying mechanisms contributing to the increases in boat velocity elicited by changes to these key variables. Keywords: Outrigger canoe; performance; model; sensitivity analysis.
1. School of Psychology and Sport Sciences, Northumbria University, Newcastle upon Tyne, UK. NE1 8ST E-mail: [email protected]
98 The Engineering of Sport 7 - Vol. 1
1 - Introduction Performance in outrigger canoe paddling can be indicated by a high mean boat velocity and hence a low race time. In order to achieve a high mean boat velocity, the paddler must generate large forces which are transmitted to the water through the paddle. These are achieved through the correct sequencing of muscle contractions to move the trunk and arms relative to the boat so as to accelerate the paddle with respect to the boat and water. The paddle stroke can be split into the drive and recovery. During the drive, the paddle is inserted deep into the water and pulled backwards through the water, developing propulsive forces due to the fluid dynamic interaction between the paddle and water. At the end of the drive, the paddle is extracted from the water. The paddler then moves the paddle forwards, out of the water, in order to return it to the correct position to commence the next drive phase. Humphries et al. (2000) found high correlations between force output on a canoe ergometer and 250m race time, as well as finding similar correlations against race time for a range of anthropometric characteristics. Standon et al. (2002) investigated the role of stroke rate in determining paddling performance, suggesting that stroke rate has an important influence on the rate of force development. Although these studies make a valuable contribution to our understanding of what contributes to outrigger canoe paddling performance, they are specific to tethered canoe and ergometer paddling. Investigating on-water paddling, however, is more difficult and has received little attention in the literature. It is useful, therefore, to use mathematical models to investigate this area in more detail. The aim of this study, therefore, was to develop a model of on-water outrigger canoe paddling mechanics in order to determine the key variables leading to an increased propulsive force, and hence a high mean boat velocity. The model will first be derived, and once validated, a sensitivity analysis performed.
2 - Model derivation According to Newton’s second law, a linear force,, generated by a system can be given by (1) where m is the mass of the system and a is its acceleration. If input forces and masses of a system are known, then the velocity of that system can be calculated as the integral of the system acceleration, such that (2) In the case of the outrigger canoe stroke, we can develop equation (1), thus giving (3) where P is the total propulsive force, D is the total drag force, m is the mass of the boat, M is the combined mass of the boat, paddlers and paddles, vboat is the velocity of
The Engineering of Sport 7 - Vol. 1
99
the boat relative to the earth and vcrew is the velocity of the paddler’s centre of mass relative to the boat. As the lower body remains almost stationary relative to the boat through the stroke, it is the movements of the centre of mass of only the upper body that were assumed to influence boat velocity. In order to determine the propulsive force at the paddle, the fluid dynamic properties of the paddle and the kinematics of the paddle through the stroke must be modelled. The paddle will generate both a lift and a drag force. However, it is assumed that the paddle shaft only moves in the sagittal plane, so the lift forces generated will be vertical. Thus it is assumed that only drag forces contribute to propulsion. The drag force at the paddle is given by (4) where is the fluid density, A is the projected area of the paddle, VP is the relative velocity between the paddle and water and CD is a dimensionless drag force coefficient that will depend on the angle of attack between the paddle and oncoming fluid flow, and the design of the paddle. The relative velocity, Vp-rel, between the paddle and water is then given by (5) where vpadlle is the velocity of the paddle relative to the boat. This velocity was measured from video data, as described below. Sumner et al. (Sumner et al., 2003) presented drag coefficients for a range of kayak paddles held static at a range of angles of attack in a wind tunnel. It was shown that the drag coefficient, CD, could be modelled by (6) where CDO is the drag coefficient at an angle of attack, , of zero degrees, or when the direction of fluid flow is perpendicular to the face of the paddle. The angle of attack, , was taken from the measured relationship (see below) for the paddle orientation with respect to the water through the stroke. During each stroke, the location of the paddler’s upper body centre of mass will oscillate back and forth. In a similar way as has been done in rowing (Brearley & de Mestre, 1996; Caplan & Gardner, 2007) this movement was modelled as a single mass moving with half of simple harmonic motion, although the direction of motion was opposite to that seen for rowing. The position of the paddler’s upper body centre of mass during the drive phase, x1, was therefore modelled as (7) where a is the amplitude of the rower movements back and forth in the shell. During the recovery phase this relationship changes such that the upper body centre of mass position of the paddler, x2, is given by
100 The Engineering of Sport 7 - Vol. 1 (8) where 2 is the duration of the recovery phase, and 1 t 1+2. The upper body centre of mass position of the paddler throughout the stroke is then differentiated to give the velocity of the upper body centre of mass, such that, (9) and (10) during the drive and recovery phases of the stroke, respectively. Finally, the drag forces acting on the boat and paddler must be calculated. The O1 single paddler outrigger canoe uses the same hull as a K1 kayak. Although no data was available for the O1, data was presented by Dansprint ApS (Dansprint ApS, 2004) for a K1 with an added paddler mass of 50 kg, similar to the mass of the paddler used here for the model validation. From this data, the hydrodynamic drag acting on the boat was given by (11) The aerodynamic drag acting on the surfaced portion of the boat and the paddler is assumed to be modelled by the relationship presented by Hoerner (Hoerner, 1965) for a seated man (Lazauskas, 1997; Caplan & Gardner, 2007) such that (12) where n is the number of paddlers in the boat. The total drag, D, was then the sum of the aerodynamic and hydrodynamic drag forces.
3 - Model validation For any model to be applicable to the real world situation being modelled, it must be appropriately validated. For this purpose, the race time for an elite female British outrigger canoe paddler was used, along with kinematic data from the same paddler, from which the centre of mass oscillations were derived. The paddler, whose body mass was 52 kg, recorded a 500m race time of 175 s, equating to a mean boat velocity of 2.86 m.s-1. A typical stroke rate of 77 strokes per minute was also observed. The same paddler was filmed from a lateral view by a miniDV camera (HDR-AV1, Sony, Japan) with a frame rate of 50Hz and a shutter speed of 1/1000th second. Scale points on the boat indicated a horizontal distance of 1m. A 7 point, 6 segment spatial model was used for the upper body, with an additional 2 points used to define the upper and lower limits of the paddle shaft. The video data was digitized using Peak Motus (Vicon Motion Systems, Oxford). The percentage mass of each segment with respect to total body mass, and the centre of mass location for each segment were defined according to Winter (Winter, 1990), and the centre of mass coordinates deter-
The Engineering of Sport 7 - Vol. 1
101
mined using Peak Motus. Only the horizontal oscillations of the centre of mass were needed for the model as any vertical movements were assumed not to contribute to propulsion. The video data showed the change in angle of attack through the stroke to be well modelled (r = 0.9995) by a linear function, such that =175.7t+3.5 (13) The input variables used for model validation are shown in Table 1. The model was setup in Simulink (Mathworks, MA) and solved using a 4th order Runge-Kutta fixed step solver, with a step size of 0.001 s. The model ran for 10 seconds, by which time a steady state velocity had been achieved. Mean steady state velocity was subsequently determined for the last stroke. Table 1 - Model validation input variables.
Variable
Value
Drive time (s) Recovery time (s) Stroke time (s) Stroke rate (1.min-1) Rower amplitude (m) Paddler mass (kg) Boat mass (kg) mean blade velocity (m.s-1) Fluid density (kg.m-3) blade area (m2) Peak drag coefficient
0.38 0.4 0.78 76.9 0.13 52 20 3.8 999 0.1233 1.7
The model was shown to closely simulate the actual race time for the elite paddler suggesting the validity of the model. The modelled velocity of 2.78 m.s-1 was within 3 % of the measured mean boat velocity (2.86 m.s-1). Figure 1 shows the modelled boat velocity for the validation simulation, showing the expected oscillations of boat velocity above and below the mean, due to the cyclical nature of paddling technique. The boat velocity was seen to achieve the maximum velocity within 5 strokes, and to fluctuate by approximately 0.5 m.s-1 above and below the mean during each stroke. The velocity was seen to increase during the drive phase and decrease during the recovery phase.
102 The Engineering of Sport 7 - Vol. 1
Figure 1 - Modelled boat velocity is shown. Race velocity is reached within five strokes and velocity is seen to oscillate above and below the mean by approximately 0.5 m.s-1.
4- Sensitivity analysis In order to determine the key variables contributing to the magnitude of the mean boat velocity, a sensitivity analysis was performed. Each variable was, in turn, increased or decreased in magnitude by 1 %, and its influence on boat velocity recorded. Table 2 shows the results of this analysis for the variables tested. Only the change (increase or decrease) to each variable that elicited an increase in mean boat velocity is shown. The variables are ranked according to the magnitude of the change in boat velocity observed through the change to each variable. The velocity of the blade relative to the boat was seen to have the greatest influence on mean boat velocity. The next most important variables were the drag coefficient, blade area and fluid density. However, the magnitude of the influence of these variables was approximately ten times smaller than for blade velocity. The movement of the paddler within the boat was seen to influence the mean boat velocity to a similar degree as drag coefficient, blade area and fluid density. Reducing recovery duration, which would act to increase stroke rate whilst maintaining the same force production characteristics of the stroke through the drive phase, was seen to have only a small relative influence on mean boat velocity, whilst increasing boat mass and decreasing paddler mass had little effect on increasing boat velocity.
The Engineering of Sport 7 - Vol. 1
103
Table 2 - Sensitivity analysis results. Each variable is ranked based on its importance in influencing boat velocity.
Variable Validation simulation blade velocity () D () blade area () fluid density () rower amplitude () recovery duration () boat mass () paddler mass ()
Mean boat velocity
Increase from validation
Rank
2.7809 2.8103 2.7836 2.7836 2.7836 2.7834 2.7823 2.7816 2.7816
n/a 0.0294 0.0027 0.0027 0.0027 0.0025 0.0014 0.0007 0.0007
n/a 1 2= 2= 2= 5 6 7= 7=
5- Discussion The aim of this study was to develop a mathematical model of outrigger canoe paddling mechanics, in order to determine the key variables that influence performance through a sensitivity analysis. The model was developed based on simple Newtonian mechanics, and was shown to be valid against real world data for an elite female outrigger canoe paddler. The sensitivity analysis revealed that the key variable, in terms of increasing mean boat velocity, was the relative velocity between the paddle and boat. The importance of this variable is not surprising considering the velocity squared relationship for the drag force generated by the blade (see equation (4)). The squared relationship of paddle velocity to the drag coefficient, paddle area and fluid density supports the reduced influence of the latter three variables on mean boat velocity. A limiting factor in increasing the velocity of the paddle with respect to the boat is the ability of the paddler to pull the paddle through the water. If paddle area was maintained, an increase in paddle velocity would only be possible through an increase in paddler strength. Alternatively, paddle velocity could be increased by reducing paddle area. Although paddle area was shown to be important in increasing boat velocity, the larger influence of increasing paddle velocity may offset the influence of reducing paddle area in order to achieve an increased boat velocity. Further research should investigate this in more detail. In order to increase boat velocity, the amplitude of paddler motion with respect to the boat had to be reduced. During the drive phase the motion of the paddler within the boat will act to increase boat velocity due to the exchanges in momentum that occur between the paddler and boat. During the recovery phase, the movement of the paddler forwards in the boat will cause a concomitant reduction in boat velocity by the momentum exchanges effectively “pushing” the boat backwards relative to the paddler’s upper body. This is similar, but opposite, to the effect of athlete movement seen in rowing, where the boat is seen to increase in velocity during the recovery phase (Martin & Bernfield, 1980). The fluctuations in boat velocity caused by paddler motion with
104 The Engineering of Sport 7 - Vol. 1 respect to the boat will be detrimental to boat performance due to the velocity squared relationship for boat drag. Any reductions in velocity below the mean boat velocity will reduce drag. However, these decreases will be smaller than the increases in drag experienced when boat velocity increases above mean boat velocity. By reducing the amplitude of paddler motion, the oscillations in boat velocity will reduce, thus resulting in smaller boat velocity fluctuations through the stroke. In rowing, it is typical for a rower to reduce the duration of the recovery phase in order to increase stroke rate and velocity (McGregor et al., 2004). When the same occurs for canoe paddling, the sensitivity analysis suggested that boat velocity will increase, although this increase was small in comparison to the other variables tested. The results suggested that the recovery phase duration is not a key variable influencing performance in outrigger canoe paddling. This is likely due to the smaller influence of paddler movement on boat velocity, compared to rowing. Changing the mass of either the boat or the paddler, will influence the momentum exchanges occurring through the stroke, in a similar way to changing the amplitude of paddler motion with respect to the boat. The sensitivity analysis revealed that boat mass should be increased and paddler mass should be decreased in order to elicit an increase in boat velocity. Although it might seem counterintuitive to increase boat mass, this is necessary to negate the influence of paddler motions back and forth in the boat, and the negative effect these motions have during the recovery when no propulsive force is developed. Similarly, reducing the mass of the paddler has an equal effect on boat velocity, thus supporting the relationship between boat and paddler mass, and their opposite influences on momentum exchanges occurring through the stroke. In conclusion, this study has presented a valid model of outrigger canoe paddling mechanics and revealed that the key variable influencing boat velocity in outrigger canoe paddling is the velocity between the paddle and boat. A higher paddle velocity relative to the boat will result in an increased relative velocity between the blade and the water, thus increasing propulsive force generated through the fluid dynamic interaction between the paddle and water. The degree to which this variable influences boat velocity is approximately ten times greater than any other variable tested. Future research should be conducted to investigate how paddle velocity can be increased most effectively, either through increasing paddler strength or through paddle design.
6- References Brearley, M. N. & de Mestre, N. J. (1996). Modelling the rowing stroke and increasing its efficiency. 3rd conference on mathematics and computers in sport, Bond University, Queensland, Australia, 35-46. Caplan, N. & Gardner, T. N. (2007). A mathematical model of the oar blade-water interaction in rowing. Journal of Sports Sciences, 25, 1025-1034. Dansprint ApS (2004). Kayak ergo - Technical info for kayak simulator. http://www.dansprint.com/1/12/technical-info.html [Accessed: 25/1/08]. Hoerner, S. F. (1965). Fluid-dynamic drag. Albuquerque, New Mexico: Hoerner Fluid Dynamics. Humphries, B., Abt, G.A., Stanton, R. & Sly, N. (2000). Kinanthropometric and physiological characteristics of outrigger canoe paddlers. Journal of Sports Sciences, 18, 395-399.
The Engineering of Sport 7 - Vol. 1
105
Lazauskas, L. (1997). A performance prediction model for rowing races. University of Adelaide Department of Applied Mathematics Technical Report, L9702. Martin, T. P. & Bernfield, J. S. (1980). Effect of stroke rate on velocity of a rowing shell. Medicine and Science in Sports and Exercise, 12, 250-256. McGregor, A. H., Bull, A. M. J., & Byng-Maddick, R. (2004). A comparison of rowing technique at different stroke rates: a description of sequencing, force production and kinematics. International Journal of Sports Medicine, 25, 465-470. Standon, R., Evens, G., Dascombe, B. & Peddle, M. (2002). Biometric and biomechanical correlates to outrigger canoe paddling. Strength and Conditioning Coach, 10, 19-26. Sumner, D., Sprigings, E. J., Bugg, J. D., & Heseltine, J. L. (2003). Fluid forces on kayak paddle blades of different design. Sports Engineering, 6, 11-20. Winter, D. A. (1990). Biomechanics and motor control of human movement. Canada: John Wiley & Sons
The Dynamic Compaction of Cricket Soils for Pitch Preparation (P20) Peter Shipton1, Iain James1
Topics: Cricket pitch preparation. Abstract: In cricket, where a hard leather ball is bounced on to a natural turf, clay soil pitch between the 'bowler' and the 'batsman' at speeds of 20 to 160 km h-1, the ball-surface interaction is critical. This interaction is a function of the mechanics of the surface, which in turn are affected by soil packing (bulk density) and moisture content in the critical top 100 mm of the profile. To achieve the required mechanical properties of the pitch, the surface is consolidated with a smooth wheeled, steel roller at optimum moisture for compaction. This paper describes initial results from laboratory scale tests using an apparatus designed to investigate the effect of key design and operating variables for the roller (mass, roller diameter, forward speed) and environmental factors (soil type, moisture content, grass) on surface consolidation with depth through the facilitate profile. Keywords: cricket; soil mechanics; playability.
1- Introduction The nature of the importance of the ball-surface interaction in cricket is described in James, Carré, and Haake, (2004). This interaction is controlled by the mechanical properties of the clay loam soil surface, which are manipulated using a smooth-wheeled steel roller (Shipton, James, and Vickers, 2006). The aim of this process is to achieve a flat, consolidated surface that maximizes soil strength by increasing soil particle packing and minimizing moisture content, but still permits grass growth to aid the playing performance of the pitch, reinforce the soil with the grass root network and facilitate drying of the soil through transpiration. The critical soil depth for playability of a soil profile ranges from the top 75 to the top 100 mm (Adams & Gibbs, 1994; Baker et al., 2003). The rolling of cricket pitches is carried out prior to the start of the playing season and then immediately before a pitch is used for play. A survey of practices across England and Wales revealed a variation of 1000% in terms of the time spent rolling at different 1. School of Applied Sciences, Cranfield University, UK - E-mail: [email protected]; [email protected]
108 The Engineering of Sport 7 - Vol. 1 cricket facilities, highlighting the need for an improved understanding and an optimization of the rolling process (unpublished data, Cranfield University). The aim when rolling with a smooth wheeled roller is to increase dry bulk density (b). This is achieved in conjunction with shrinkage of the soil during natural drying of the soil profile. The soil moisture content required for efficient compaction is greater than for optimum playability (Shipton, James, and Vickers, 2006) thus a drying period after rolling and before play is necessary. To achieve the required soil density the load applied to the soil needs to be sufficient to overcome the soil structural resistance to the applied stress. This resistance is largely a result of soil mineralogy, density and moisture content. The magnitude and direction of the applied stress is a function of a number of roller design parameters: contact area – governed by roller diameter and width, roller mass and operating speed. This study provides data from laboratory-scale experiments investigating different roller design parameters using an instru-mented rolling rig. This approach provides optimized design parameters for roller design and operation.
2- Materials and Methods 2.1 Instrumented Rolling Rig The rolling rig (Figure 1) comprises a 3500 mm x 500 mm x 350 mm steel tank, filled with the test soil. At the midway posi-tion along the length of the tank a 600 mm x 500 mm viewing window allows investigation of soil movement through the soil profile (y:z plane). Roller diameters from 200 mm to 1000 mm can be mounted in the frame which allows the roller to float and can be towed in both directions using a linear transmission and positioning device (HepcoMotion, Devon, UK) powered by an AC servomotor/drive system (Baldor, Bristol, UK). Forward speed can be controlled at 20 mm/s to 1000 mm/s. Figure 1 - Cricket rolling rig – (a) 3-D schematic and (b) side view
The Engineering of Sport 7 - Vol. 1
109
The most common type of roller used in cricket has two rollers of 600 mm diameter and a mass of 2000 kg. The results pre-sented in this paper are for a roller with a diameter of 600 mm driven at 84 mm/s and operated at four passes of five ballasts (130; 220; 310; 400; 490; 580 kg) applied consecutively. A period of 10 minutes between roller passes was allowed to facili-tate elastic recovery and to emulate cricket pitch preparation requirements. The maximum load is equivalent to a 3000 kg, 2 drum roller. The soil used was a clay loam soil from Essex, UK (30% sand; 40% silt; 30% clay). It is typical of soils used at elite and well resourced recreational levels of the game. The soil was passed through a 5 mm sieve and water added to give the required gravimetric moisture content of 20% and was evenly packed to give an initial density of 1.2 g cm-3 to a depth of 250 mm. The prepared profile was left for 24 hrs to ensure equilibration of moisture. Applied pressure at depth was recorded using three ceramic membrane pressure transducers of 18 mm diameter mounted in 25 mm diameter tubular aluminium housings at depths of 25 mm, 50 mm and 80 mm in a method similar to James, Dixon, Blackburn, and Pettican (2006). Each transducer was logged at 10 Hz.
Figure 2 - Cross section of soil profile as imaged through the viewing window.
An array of 6 mm diameter markers was placed in the soil to be viewed through the window at depths of 25 mm, 50 mm, 80 mm, 100 mm and 140 mm, with seven markers at each depth (Figure 2). Their movement under load was recorded at 30 fps using a Fujifilm finepix S9600 digital camera. The horizontal (y) and vertical (z) components of
110 The Engineering of Sport 7 - Vol. 1 movement of the marker centroids was determined using image processing scripts in Matlab (Mathworks, Cambridge, UK). Mean data for pressure and displacement with depth are presented for four consecutive passes of the roller at 490 kg load and for the first pass of each roller load. Displacement data is presented as R, max dy and max dz as defined in Figure 3.
Figure 3 - Clockwise trace of marker movement with left-to-right rolling (25 mm depth, 220 kg load). The three recorded parameters are shown, R: the final permanent displacement; max dy, the maximum horizontal displacement; max dz, the maximum vertical displacement.
3- Results and Discussion 3.1 Pressure – Load – Depth relationship At 25 mm, pressure ranged from 150 to 450 kPa and increased with load (Figure 4). At 50 mm, a similar relationship between pressure and load was observed but the magnitude was between 100 and 150 kPa for all ballast loads. At 80 mm, a constant pressure of 50 kPa was observed, irrespective of load. These results are as expected and tie with the data presented in Shipton, James and Vickers (2006) which determined little effective compaction below 50 mm with similar roller configurations.
The Engineering of Sport 7 - Vol. 1
111
Figure 4 - Mean pressure with increasing roller load, at depths of 25, 50 and 80 mm within a clay loam soil
3.2 Displacement –depth – load Generally, the cumulative resultant displacement from the initial condition to the final loading, i.e. the permanent displacement of the marker from O to R increased with roller load, and with depth (Figure 5). At 130 and 220 kg, displacement was minimal; at greater loads the soil was moved downwards and forwards to a greater extent. For example at 25 mm depth, with the 580 kg ballast, the soil was moved downwards by -4.5 mm and forwards by 0.3 mm. As depth increased for the same load, forward displacement was similar, but vertical displacement was reduced to -2 mm. The maximum movement of the markers in any pass reflects the elasticity of the bulk soil with maximum vertical displace-ments of 30 mm and 18 mm horizontally (Figure 5). Again, the maximum displacements increased with load, and decreased significantly with depth.
112 The Engineering of Sport 7 - Vol. 1
Figure 5 - Displacement at depths related to sensor depths (25, 50 and 80 mm). Left hand column: position of R in y:z for first pass of different loads: + 130 kg, ⵦ 220 kg, 310 kg, 얖 400 kg, • 490 kg (4 passes), 䉫580 kg. Right hand column max dy (•) and max dz (+).
3.3 Implications for rolling Results show pressure and displacement reduce with depth, with minimal displacement at the lowest depths measured. Increas-ing load indicates large increases in pressure at the top of the profile but a reduced increase with depth resulting in a relatively greater compaction near to the surface. Even compaction from rolling throughout a 100 mm cricket pitch profile is therefore unlikely with this diameter of roller and a maximum 3t load on a typical roller. Displacement reduces with consecutive roller passes indicating that after four passes the gain in b is relatively insignificant. The current practice by most cricket groundsmen of multiple roller passes would appear to be ineffectual in achieving greater b and additional roller passes would only be appropriate after an increase in roller weight or a change in soil moisture. These results also indicate that heavier cricket rollers than is currently normal i.e. >2.5 ton, would be advantageous in increasing b to a greater depth. The horizontal component of movement is important as this is thought to cause shearing of roots in the soil, leading to the adverse phenomenon of 'root breaks' where a sheared layer between 25 and 50 mm allows dense horizontal root growth, creating a 'spring' in the cricket pitch profile. This technique will be used to explore this hypothesis further, and future studies will include vertical reinforcement from live grass roots,
The Engineering of Sport 7 - Vol. 1
113
which is expected to reduce this horizontal component. Future work will also investigate moisture content and roller diameter variables.
4- Conclusions The instrumented rolling rig provides a tool for the development of steel smoothwheeled rollers for cricket pitch preparation in a controlled environment. The resultant data, which in this case show that pressure at 80 mm is negligible, even with relatively heavy rollers is significant in understanding the role of rolling in pitch consolidation. To increase density throughout the whole profile requires effective drying by the plant – it cannot be achieved by rolling alone. The horizontal component of deformation is considered significant and should be minimised in the rolling of cricket pitches. Further investigation will combine this data with field data in the form of a model to deliver optimised parameters for cricket pitch rolling and preparation.
5- References [AG1] Adams, W. A. & Gibbs, R. J. Natural Turf for Sport and Amenity. CAB International, UK, 1994. [BH1] Baker, S.W., Hammond, L.K.F., Owen, A.G. and Adams, W.A. Soil physical properties of first class cricket pitches in England and Wales. 1. Classification system for soil characteristics. In Journal of Turfgrass and Sports Surface Science, 79: 2-11, 2003. [JC1] James, D.M., Carre, M.J. and Haake, S.J. The playing performance of county cricket pitches. In Sports Engineering, 7: 1-14, 2004. [JD1] James I; Dixon S; Blackburn K; Pettican N. The Measurement of Applied Pressure at Depth with Two Natural Soil Surfaces at Different Densities. In: The Engineering of Sport 6, Volume 2: Developments for Disciplines. E F Moritz & S Haake (Eds.), 2006, Springer, NY, USA. ISBN-10: 0387-34678-3, p 173-178, 2006 [SJ1] Shipton, P.M.R., James, I.T. and Vickers, A. The mechanical behaviour of cricket soils during preparation by rolling. In: The Engineering of Sport 6, Volume 1: Developments for Sports. E F Moritz & S Haake (Eds.), Springer, NY, USA: 229-234, 2006.
Experimental Validation of a Finite-element Model of a Tennis Racket String-bed (P21) Tom Allen, Simon Goodwill, Steve Haake1
Topics: Tennis & other Rackets Sports, Modelling. Abstract: An explicit finite-element (FE) model of a tennis racket string-bed was produced in Ansys/LS-DYNA 10.0. This model was used to simulate a range of impacts between a tennis ball and string-bed, which were validated against experimental data. The laboratory validation was undertaken by firing balls, with backspin in the range from 0 to 600 rad·s-1, from a pitching machine onto a head-clamped tennis racket. Inbound velocities and angles in the range from 20 to 30 m·s-1 and approximately 20 to 60° respectively were tested. Results were obtained for rebound spin, angle and velocity, with good agreement between the model and experiment. Keywords: Tennis, Finite Element Analysis, String-bed, Spin.
1- Introduction The tennis ball-rigid surface impact has been modelled successfully using the finiteelement (FE) technique (GK1, AG1). A number of authors have attempted to include a string-bed in an FE model of a tennis racket with varying success (WM1, WM2, KN1). Kanda et al. (2002) produced an FE model of a tennis ball impacting perpendicular to a freely suspended strung tennis racket. The ball was modelled as a pressurised rubber core, with linear material properties. In addition the felt cover was not incorporated into the model even though it has been found to influence rebound spin for impacts on a rigid surface (GK1); there was also no reference of the ball being independently validated. A number of impact locations were simulated and the coefficient of restitution (COR) was found to decrease with increasing string tension and to be highest between the geometric centre of the string-bed and the throat, in agreement with other authors (BC1, GH1). However, as with previous publications (WM1, WM2), the strings were assumed to be fixed at their intercepts, effectively ignoring the effect of string to string friction. This is clearly not a realistic representation of reality and subsequent errors would become apparent if simulating an oblique impact. Sports Engineering Research Group, Sheffield Hallam University, UK - E-mail: [email protected]; s.r.groodwill,[email protected]
116 The Engineering of Sport 7 - Vol. 1 Cross (2000) produced a mathematical model of a ball impacting obliquely on a string-bed at 10 m·s-1, to analyse the effect of the sliding friction between the ball and the strings. Cross experimentally obtained sliding and rolling friction coefficients for 5 different strings, which were 0.27-0.42 and 0.05, respectively. However, he calculated sliding friction by placing a 10 kg mass on a ball and dragging it across the string-bed at a constant velocity, a method not representative of a typical high speed collision. The critical value of sliding friction between the ball and string-bed was found to be 0.3; below this the ball’s rebound angle and range drops significantly, which would result in a detrimental effect on the player’s performance. As well as not validating the model against experimental data, Cross made a number of assumptions and simplifications. His results were based on the assumption that the ball impacts at the centre of the stringbed, which is not representative of a shot during play (CG1). He didn’t consider the effects of friction between the strings and their movements. However, currently there is no published data which relates to the method of obtaining string to string fiction. Goodwill and Haake (2004) experimentally analysed the impact of an oblique spinning ball on a head-clamped racket. They tested inbound velocities of 23 and 31 m·s-1 at an angle of 39° to the normal, with backspin in the range from 0 to 420 rad·s-1. It was concluded that a rigid body mathematical model under-predicted the rebound spin of the ball. Furthermore, the experimentally measured spin was found to be higher than that associated with rolling. A deformable ball mathematical model was produced and it was concluded that the ball starts to over-spin at the mid point of the impact, resulting in the friction force reversing direction, which hence produces an increase in the horizontal velocity. This reversal of the friction force acting on the ball was in agreement with the findings of numerous authors for oblique impacts on a rigid surface (C2, HC1, HC2). However, Goodwill and Haake’s (2004) model did not have the capacity to calculate the rebound spin of the ball. Further testing including higher spin rates and a range of angles is required to gain insight into the ball’s characteristics when impacting with a string-bed. In this investigation an FE model of a tennis racket string-bed will be validated against experimental data for nominal impact angles of 20°, 40° and 60°, relative to the racket normal. The ball’s rebound velocity, angle and spin will be compared against inbound spin, for a set angle and velocity. For a set material and gauge (diameter), string-bed stiffness will be determined by string tension, which has a typical range of 220-310 N (BC1). The agreement of the FE model with the experimental data for rebound spin will provide an initial indication as to whether the ball-string friction is in the correct range.
2- FE model An explicit FE model of a tennis racket string-bed, consisting of 16 main and 19 cross strings, was created in Ansys/LS-Dyna 10.0 (Fig. 1). The overall dimensions of the stringbed were 0.331 0.253 m, with the strings having a diameter of 1.32 10-3 m, MAT_ELASTIC was selected as the linear material model for the strings. The method for obtaining the dynamic stiffness of a tennis string, using a hammer strike, was deve-
The Engineering of Sport 7 - Vol. 1
117
loped by Cross et al. (2000); in this FE model, the Young’s modulus of 7.2 GN·m-2 was obtained using KL/A, where K is the dynamic stiffness of the string and L and A are the length and cross sectional area of the string, prior to the hammer strike. The value of K used (30822 N·m-1) was taken from Lindsey (2006), which was a value consistent with the particular nylon strings used in this model. The density of 1100 kg·m-3 was provided by the manufacturer and the Poisson’s ratio was assumed to be 0.3. SOLID164 3-D 8 node bricks, with single point integration or constant stress, were used to mesh the main and cross strings, which consisted of 19,624 and 17,712 elements, respectively. This particular value of mesh density was selected as it produced elements of a similar size to those in the ball and hence prevented contact instabilities. Solid elements were selected, above 1-dimensional elements, as they were a better representation of reality; they have the ability to accurately represent the shape and structure of the string-bed, in particular the 3D deformation characteristics of individual strings.
Figure 1 - String-bed model.
Contact between the ball and string-bed was defined as CONTACT_SURFACE_TO_SURFACE with a friction coefficient of 0.4 (C1). The same technique was used for string to string contact, except the friction coefficient was set to 0.1. A rigid cylinder 1 10-3 m in length and 1.32 10-3 m in diameter was attached to both ends of every string. This FE model consists of two phases: the dynamic relaxation phase, in which the static pre-loads are applied to structures (string tension and internal ball pressure), which occurs prior to the transient stage of the model, in which the ball is set in motion. The end of one phase and the beginning of the next is triggered when the displacements of the structures, following their loads, start to “settle” to within their convergence tolerances, which is a value that denotes how closely the displacements must reach the value they are converging to. A load of 150 N was applied to each of the rigid cylinders, in the required direction during the dynamic relaxation phase, to produce a total tension on every string of 300 N. These rigid volumes were then fixed in position during the transient phase of the simulation, effectively resulting in a headclamped racket. The convergence tolerance for dynamic relaxation was 0.01. Applying constraints to rigid cylinders as opposed to directly to the ends of the nylon strings, prevented element distortion, hence resulting in a more stable model. A pressurised rubber core with a felt cover was used to simulate the ball; the material models were MAT_OGDEN_RUBBER and MAT_LOW_DENSITY_FOAM, respectively. Details of the ball model, including the material properties and its validation, can
118 The Engineering of Sport 7 - Vol. 1 be found in Allen et al. (2007). The ball’s initial velocity and spin were assigned using INITIAL_VELOCITY_GENERATION. Due to the ball having three different inbound angles, the string-bed was translated in the horizontal direction, prior to the simulation, in order to ensure the ball contacted at the correct string-bed location. This was achieved using a piece of software, specially developed in MS Visual Basic 2005, which modifies the DYNA text file by selecting the nodes corresponding to the string-bed and updating their locations. Translating the ball, rather than the string-bed, to allow for the correct contact positions, would make calculating spin difficult, thus this method enabled the ball to be kept in the same location, with its initial position at the intercept of the x, y and z axes.
3- Method Tennis balls were projected against a head-clamped racket (Fig. 2) using a modified pitching machine device. An aluminium tube was fitted to the pitching machine to provide greater consistency for the balls inbound angle. A carbon fibre tennis racket with a head size of 0.063 m2 was used for all tests. Two groups of four rackets were used in the investigation, strung at 200 N and 289 N, respectively. To ensure consistent and accurate results the string-bed deflection of the rackets was measured directly before and after testing using a Babolat RDC, the mean of these 2 values are quoted in this paper as opposed to the stringing tension. Ball inbound angles and velocities in the range from 20 to 60° and 20 to 30 m·s-1 were analysed. The inbound angle was adjusted by tilting the racket as opposed to adjusting the cannon and flight path of the ball. Around 20 impacts were undertaken for every racket at each angle; the inbound backspin was varied from 0 to 600 rad·s-1.
Figure 2 - Experimental setup.
The flight of the balls was recorded as a series of bitmap files, using a Phantom V4.1 high-speed video camera, positioned 8 m from the racket in the direction of its longitudinal axis and recording at 1000 fps. Inbound and rebound velocities, angles and spins were measured from the recorded images using Richimas (bespoke image analysis software). The mean inbound angles were found to have deviated slightly from those
The Engineering of Sport 7 - Vol. 1
119
predicted from the experimental set-up; the calculated values are shown in table 1. For simplicity the inbound conditions will be referred to by their nominal values. The mean horizontal distances from the ball’s impact location to the geometric centre of the stringbed, were calculated for each set of impacts. This was achieved by estimating each ball’s impact position relative to the string-bed centre, from its initial location upon exiting the pitching machine and the calculated mean angle corresponding to the experimental set-up (Table 1). FE simulations were undertaken with inbound velocities, angles and impact locations identical to those in the laboratory experiment. For each angle and velocity pair, simulations were undertaken with backspin ranging from 0 to 600 rad·s-1, at 200 rad·s-1 increments. Table 1 - Inbound angles, velocities and impact locations relative to the centre of the string-bed.
4- Results Figure 3a-f shows that the model results for rebound velocity are in good agreement with the experimental data. Although the model marginally under-estimated the rebound velocity of the ball at 20° and 20 m·s-1, for a backspin of 200 rad·s-1 (Fig. 3a). The model also appears to slightly over-calculate rebound velocity at 60° and 30 m·s-1 (Fig. 3f). Overall the results from both the model and experiment show that rebound velocity decreases as the inbound angle, relative to the racket normal, increases. Rebound velocity can also be seen to decrease with increasing inbound backspin; with the decrease becoming more pronounced as the inbound angle increases. Close inspection of figures 3e and 3f show that the rate of decrease in rebound velocity drops significantly for inbound backspins greater than around 400 rad·s-1. This non-linear relationship of the data appears to be evident in both the model and experiment.
120 The Engineering of Sport 7 - Vol. 1
Figure 3 - Rebound velocity against inbound spin at a) 20°, 20 m/s, b) 20°, 25 m/s, c) 40°, 20 m/s, d) 40°, 30 m/s, e) 60°, 20 m/s, f) 60°, 30 m/s
Figure 4 shows that the model results for rebound angle are in good agreement with the experimental data. However, the rebound angles for the 20° simulations are slightly higher than the experimental values (Fig. 4a and b). The general trend is that the rebound angle of the ball increases with the inbound angle. The results also show that rebound angle decreases as inbound backspin increases. However, figure 5e and f show that the reduction in rebound angle with increasing inbound backspin appears to become less pronounced in the experimental data, for backspins greater than approximately 350 rad·s-1. This non-linearity, which is in agreement with the results for rebound velocity, can be observed in the FE model with a nominal inbound angle of 60° at 20 m·s-1 (fig 3e) but not at 30 m·s-1 (fig 3f).
Figure 4 - Rebound angle against inbound spin at a) 20°, 20 m/s, b) 20°, 25 m/s, c) 40°, 20 m/s, d) 40°, 30 m/s, e) 60 °, 20 m/s, f) 60°, 30 m/s.
The Engineering of Sport 7 - Vol. 1
121
Figure 5 shows that all the model results for rebound spin are in good agreement with the experimental data. The results show that rebound spin decreases with increasing backspin, whilst it increases with the inbound angle. Again a non-linearity can be observed, for both model and experiment, in the rebound characteristics of the ball for inbound backspins above approximately 350 rad·s-1 at a nominal inbound angle of 60° (Fig. 5e and f).
Figure 5 - Rebound spin against inbound spin at a) 20°, 20 m/s, b) 20°, 25 m/s, c) 40°, 20 m/s, d) 40°, 30 m/s, e) 60 °, 20 m/s, f) 60°, 30 m/s.
Figure 6a shows that the horizontal force acting on the ball, comprising of friction and a string-bed horizontal reaction force, switches direction just after the midpoint (2.85 ms) of the impact. The initial horizontal force is negative which means that the force is acting in the opposite direction to the ball motion. At a time of approximately 2.85 ms the horizontal force becomes positive, implying that it is in the same direction as the ball motion. This causes an increase and decrease in the horizontal and angular velocities of the ball, respectively (Fig.6b and c). The horizontal force switches direction again at around 4.2 ms, resulting in a very slight decrease in the horizontal velocity and an increase in spin. As the horizontal force acting on the ball switches direction during impact then there will be an instance at which the coefficient of friction between the ball and strings will equal zero. This illustrates that any model that assumes a simple, linear relationship between a friction and vertical reaction force is invalid.
122 The Engineering of Sport 7 - Vol. 1
Figure 6 - Inbound velocity 30 m·s-1, angle 40° and backspin 200 m·s-1 a) Vertical and horizontal force, b) Horizontal velocity and c) Spin.
5- Discusssion The FE model has been found to be in good agreement with the experimental data for rebound velocity, angle and spin, albeit a few marginal discrepancies. It was found that the horizontal force acting on the ball, switches direction at approximately the midpoint of the impact when the inbound velocity, angle and backspin were 30 m·s-1, 40° and 200 rad·s-1, respectively. This implies that the ball spin rate has exceeded that associated with rolling and is over-spinning, as found by Goodwill and Haake (2004). The fact that the friction force becomes negative again indicates a tennis ball impacting obliquely to a string-bed converges towards a rolling state. Further research would be required to quantify this hypothesis. This investigation has shown that the current FE model is capable of replicating a single impact location on a string-bed for different inbound velocities, angles and spins. This model has also allowed realistically-strung strings to move independently, providing a more realistic representation of string to string friction. However, further research should be undertaken in order to determine a more precise value for both ball, and string, to string friction. In reality, a ball will impact at a variety of locations on a stringbed during play (CG1). Therefore, to ensure that the model accurately represents reality it must be validated for different impact positions on the string-bed. These impact positions should be extremes in the longitudinal and lateral directions, to allow the largest possible area of the string-bed to be validated. This would also allow the string-bed to be encompassed into a frame to create an accurate FE model of a racket, which could be used to simulate all the typical impacts encountered during a game of tennis. This model could be used by manufacturers to evaluate the performance of different frame and string constructions, to complement their existing experimental testing.
6- Conclusion An FE model of a tennis racket string-bed has been produced and successfully validated against experimental data for different impacts. These impacts had inbound backspins in the range from 0 to 600 rad·s-1, with nominal velocities and angles from 20 to 30 m·s-1
The Engineering of Sport 7 - Vol. 1
123
and 20 to 60°, respectively. The experimental validation was undertaken using a pitching machine to project balls onto a head-clamped tennis racket. It was found that a ball will enter into an over-spinning stage during the impact, except when there is a combination of a high inbound angle (60°) and backspin (>400 rad·s-1). Extending the validation to include different impact positions on the string-bed would increase its applicability. This would also allow for a realistic FE model of a tennis ball-racket impact to be produced.
7- Acknowledgements The authors would like to thank the International Tennis Federation (ITF), for the use of their testing facilities. They would also like to thank Prince for sponsorship of the project.
8- References [AG1] Allen, T., Goodwill, S. & Haake, S. Experimental validation of a tennis ball finite-element model. Tennis Science and Technology 3, 1, 21-30, 2007. [BC1] Brody, H., Cross, R. & Lindsey, C. The Physics and Technology of Tennis. 1st edn. Racquet Tech Publishing, 2002. [CG1] Choppin, S., Goodwill, S., Haake, S. & Miller, S. 3D player testing results from the Wimbledon Qualifing Tournament. Tennis Science and Technology 3, 1, 341-348, 2007. [C1] Cross, R. Effects of friction between the ball and strings in tennis. Sports Engineering, 3, 8597, 2000. [C2] Cross, R. Grip-slip behavior of a bouncing ball. American Journal of Physics, 70, 1093-1102, 2002. [CL1] Cross, R., Lindsey, C. & Andruczyk, D. Laboratory testing of tennis strings. Sports Engineering, 3, 219-230, 2000. [L1] Lindsey, C. String Selection Map 2006. Racket Tech, 1, 22-28, 2006. [GH1] Goodwill, S. and Haake, S. Modelling of an impact between a tennis ball and racket. Tennis Science and Technology 2, 1, 79-86, 2003. [GH2] Goodwill, S.R. & Haake, S.J. Ball spin generation for oblique impacts with a tennis racket. Experimental Mechanics, 44, 195-206, 2004. [GK1] Goodwill, S.R., Kirk, R. & Haake, S.J. Experimental and finite element analysis of a tennis ball impact on a rigid surface. Sports engineering (Sheffield, England), 8, 145-158, 2005. [HC1] Haake, S., Carre, M. & Goodwill, S. Modelling of oblique impacts of tennis ball impacts on tennis surfaces. Tennis Science and Technology 2, 1, 133-137, 2003. [HC2] Haake, S.J., Carre, M.J., Kirk, R. & Goodwill, S.R. Oblique impact of thick walled pressurized spheres as used in tennis. Proceedings of the Institution of Mechanical Engineers C, Journal of Mechanical Engineering Science, 219, 1179-1189, 2005. [KN1] Kanda, Y., Nagao, H. & Naruo, T. Estimation of tennis racket power using three-dimensional finite element analysis. 218, 2002. [WM1] Widing, M. & Moeinzadeh, M. Finite Element Modeling of a Tennis Racket with Variable String Patterns and Tensions. International Journal of Sport Biomechanics, 6, 78-91, 1990. [WM2] Widing, M. & Moeinzadeh, M. Nonlinear finite element analysis of a frame stiffened with tension members. Computers and structures, 33, 233-240, 1989.
Experimental Validation of a Tennis Ball Finite-element Model for Different Temperatures (P22) Tom Allen, Simon Goodwill, Steve Haake1
Topics: Tennis & other Rackets Sports, Modelling. Abstract: An explicit finite-element (FE) model of a pressurised tennis ball was produced in Ansys/LS-DYNA 10.0 and validated at room temperature. This model was successfully updated to simulate temperatures of 283.15 and 313.15 K (10 and 40 ºC), by adjusting the internal pressure and material properties of the ball’s rubber core. The validation experiment was undertaken using an impact rig in a climate chamber, for perpendicular impacts on a rigid surface with inbound velocities in the range from 15 to 30 m·s-1. The impact rig consisted of an air-cannon, for firing the balls, a set of light gates for calculating coefficient of restitution (COR), and a force plate for measuring contact time. The model was found to be in good agreement with the experimental data across the entire range of temperatures tested. Keywords: Tennis, Finite Element Analysis, Ball, Temperature.
1- Introduction A number of authors have successfully modeled ball impacts using finite-element (FE) technique (CS1, HS1, GK1, AG1, Goodwill et al. (2005) created and validated an FE model of a pressurized tennis ball and pressurized rubber core, for impacts with a rigid surface at a range of velocities. Allen et al. (2007) modified this model to simulate a tennis ball from a different manufacturer, whilst extending the validation process to include pressurized and punctured, cores and balls. This was to independently validate the three separate parts of the model; the felt cover, rubber core and internal pressure. However, both of these studies only resulted in the production of FE models, with the capacity to simulate impact characteristics of a tennis ball at room temperature. Rose and Coe (2000) experimentally tested the effect of temperature on the static stiffness and COR of tennis balls for perpendicular impacts up to 45 m·s-1. Within the range of 273.15 to 313.15 K, they found static stiffness to effectively remain constant, whilst COR increased with temperature. Downing (2007a) experimentally measured the Sports Engineering Research Group, Sheffield Hallam University, UK - Email: [email protected]; s.r.groodwill,[email protected]
126 The Engineering of Sport 7 - Vol. 1 effect of temperature, in the range of 283.15 to 313.15 K using a climate chamber, on COR and contact times for perpendicular impacts. Downing used an impact rig, consisting of an air-cannon to fire the balls, a set of light gates to measure velocity and a force plate to record contact times, testing inbound speeds in the range from 15 to 30 m•s-1. COR was found to increase with temperature, which agreed with Rose and Coe (2000). Contact times were also found to increase with temperature, indicating a reduction in the ball structural stiffness (BC1, CL1, GK1), which was concluded to be caused by a change in the material properties of the rubber. However, neither of these experimental investigations provided an insight into how the individual changes in internal pressure and rubber properties with temperature, could effect the rebound characteristics of a tennis ball. As rubber increases in temperature, its stiffness decreases. However, it is currently not known to what extent the material properties of a tennis ball core could change in the range from a typical, to an extreme, playing temperature. In addition to the change in rubber properties, the internal pressure of a tennis ball increases with temperature. It is possible to accurately calculate the change in internal pressure with temperature, using Gay-Lussac’s law by assuming the volume to remain constant. Therefore, it is feasible to produce an FE model of a tennis ball at different temperatures, by updating the internal pressure and then adjusting the material properties, until the rebound characteristics are in good agreement with the experimental data. Downing (2007b) examined the effect of temperature on surface pace rating (SPR) for an acrylic and synthetic carpet surface. SPR was found to decrease with temperature, indicating an increase in the coefficient of friction (COF) between the ball and court. Although, due to only two surfaces being tested this investigation does not provide an overall picture of how the SPR of tennis courts changes with temperature. A reliable FE model, which simulates an oblique impact between a tennis ball and different surfaces, could be used to determine how SPR changes with temperature. The first stage of producing such a model would be the creation and validation of a tennis ball for different temperatures. This paper aims to create and validate explicit FE models of pressurized tennis balls, for temperatures of 283.15 and 313.15 K. An initial investigation will be undertaken on the effect of only modifying the internal pressure of the model, to simulate the two temperatures. Following this, a separate study will be conducted on independently adjusting the static rubber material properties. Finally, the internal pressure and material properties, both static and dynamic, of the tennis ball in the FE model will be updated to simulate the different temperatures. The FE models will all be validated against the experimental data from Downing (2007a).
2- FE model The initial FE model of a tennis ball, used in this investigation, was identical to the one produced by Allen et al. (2007). The temperature, at which this original FE model’s material testing or validation was undertaken, was not recorded. For the purpose of this investigation both of these temperatures were considered to have been 295.15 K (22 ºC); assuming this to be a realistic estimation of room temperature. In the current FE model,
The Engineering of Sport 7 - Vol. 1
127
the internal pressure is simulated as an airbag. The relationship between the pressure (P) and volume (V) during impact was assumed to be adiabatic and defined by PV1.4 equal to a constant. This adiabatic assumption, for which there is no heat transfer between the air enclosed by the ball and its surrounding, was based on the high rate at which the volume changes during impact, combined with the insulating properties of the rubber and felt. During the experiment, the balls were acclimatised to each temperature before testing (refer to D1 for details); hence there was sufficient time for heat transfer between the enclosed volume of air and the surroundings. Therefore, in the modified models the initial pressure was calculated at each temperature (T) by PV/T, which is equal to a constant. For simplicity the volume was assumed to remain constant. As with the original model the relationship between the pressure and volume during impact was assumed to be adiabatic. Figure 1a shows the relationship between the internal pressure and relative volume of the ball, for temperatures of 283.15, 295.15 and 313.15 K. The relative volume of the ball is defined as the actual volume divided by the original volume. The ball was simulated in the model using a pressurised rubber core with a felt cover; the material models were MAT_OGDEN_RUBBER and MAT_LOW_DENSITY_FOAM, respectively. Details of the ball model, including the material properties and its validation, can be found in Allen et al. (2007). The effect of altering the static stiffness of the rubber core was achieved by adjusting the uniaxial data in the MAT_OGDEN_RUBBER material model (Fig. 1b). To produce models, which simulated the full effects of adjusting temperature, both the internal pressure (Fig. 1a) and static and dynamic material properties of the rubber were modified. The static modulus of the rubber was altered by ±10 % for the models at 283.15 and 313.15 K, respectively (Fig. 1b). The dynamic modulus of the rubber in the model was increased by 75 % to simulate a temperature of 283.15 K and it was decreased by 20 % for 313.15 K. The apparatus required to obtain the material properties of the rubber at different temperatures was not available to the authors; therefore, the stated changes in the material properties of the rubber were used as they were in good agreement with the experimental data, in terms of contact times and COR, obtained by Downing (2007a), refer to figures 6 and 7 for more details. The properties of the ball, in both the model and the experiment, were measured at each temperature in terms of contact time and COR. The material properties of the felt were not modified with temperature, as any changes were assumed to have an insignificant effect on the rebound characteristics of the ball.
Figure 1 - a) Tennis ball internal pressure against relative volume, for temperatures in the range from 283.15 to 313.15 K, b) Static rubber core material properties.
128 The Engineering of Sport 7 - Vol. 1
3- Method The FE models were validated by comparing COR and contact times, for perpendicular impacts with a rigid surface, with the experimental results published by Downing (2007a). Details of this experimental procedure can be found in Downing (2007a), whilst the original validation of the FE model can be found in Allen et al. (2007). The first stage of this investigation was to analyse the effect of only adjusting the internal pressure of the model for temperatures of 283.15 and 313.15 K. Following this, an analysis was undertaken to identify the effect of increasing and decreasing the static stiffness of the rubber by 10 %, whilst keeping the original internal pressure. Finally, the internal pressure and static and dynamic material properties of the rubber were all updated to simulate the two temperatures under investigation.
4- Results The results from Downing (2007a) at 298.15 K were initially compared to the FE model and experimental data published by Allen et al. (2007). This was necessary due to different balls being used in each experiment, in addition to discrepancies in the test temperatures. Figure 2 shows that COR and contact times were found to be in good agreement between the experimental data from Downing (2007), Allen et al. (2007) and the original FE model. However, contact times were marginally higher at inbound velocities below 20 m·s-1 for the validation data and the FE model in comparison to Downing (2007a).
Figure 2 - a) COR and b) contact time, comparison between the experimental data from Allen et al. (2007) (295.15 K) and Downing (2007) (298.15 K) and the original FE model.
Figure 3 shows that when only the internal pressure was adjusted, the model overpredicted COR when the temperature was 283.15 K. At 313.15 K the model is in good agreement with the experimental data, although it marginally under-predicted COR for inbound velocities below 20m·s-1. The FE model and experimental data both show increasing COR with temperature.
The Engineering of Sport 7 - Vol. 1
129
Figure 3 - COR for adjusted internal pressure at temperatures of a) 283.15 K and b) 313.15 K.
Figure 4b shows that that when only the internal pressure was adjusted, the FE model was in relatively good agreement with the experimental data at both temperatures, for contact time. However, the experimental data shows increasing contact time with temperature, whilst the FE model has the opposite trend.
Figure 4 - Contact time for adjusted internal pressure at temperatures of a) 283.15 K and b) 313.15 K.
Figure 5a shows that increasing the static stiffness of the rubber in the FE model by 10 % results in a marginal increase in the COR (dynamic stiffness kept constant). The opposite was the case for a 10 % reduction in static stiffness. Figure 5b shows that reducing the static material stiffness by 10 % results in a significant increase in contact time. Again, the opposite was the case for a 10 % increase in stiffness. The range in both COR and contact time between both FE models (20 % change in rubber material stiffness) is approximately equal to the range of scatter in the experimental data, for a set inbound velocity.
130 The Engineering of Sport 7 - Vol. 1
Figure 5 - Effect of the static rubber material stiffness on a) COR and b) contact time.
Figure 6 shows that the FE model, with the updated internal pressure and static and dynamic rubber material properties, is in good agreement with the experiment for COR at 283.15 K (static and dynamic rubber modulus 10 and 75 % higher) and 313.15 K (static and dynamic rubber modulus 10 and 20 % lower). The model and experiment both show increasing COR with temperature.
Figure 6 - Complete model - COR a) 283.15 K and b) 313.15 K.
Figure 7 shows that the complete model is also in good agreement with the experimental data for contact time at both temperatures. Both the model and experiment show increasing contact times with temperature.
Figure 7 - Complete model - Contact time a) 283.15 K and b) 313.15 K.
The Engineering of Sport 7 - Vol. 1
131
Figure 8 shows that the maximum impact force is marginally higher at 283.15 K in comparison to 313.15 K for all the inbound velocities under investigation. It can also be observed that the maximum force occurs earlier into the impact for the model at 283.15 K for inbound velocities of 15 and 20 m·s-1. At inbound velocities of 25 and 30 m·s-1 the maximum impact force at 283.15 K lags that of the model at 313.15 K.
Figure 8 - Complete model force plot a) 15 m·s-1, b) 20 m·s-1, c) 25 m·s-1, b) 30 m·s-1
5- Discussion Increasing the internal pressure of a tennis ball, in isolation, raises its structural stiffness, thus reducing its contact time. Hence, when only the internal pressure of the tennis ball was adjusted, contact times were found to decrease with temperature. This was in contradiction to the experimental data, where contact times increased with temperature. When the effect of changing the material properties with temperature is added into the model, contact time and COR are in good agreement with the experimental data. Therefore, in agreement with Downing (2007a), it is concluded that the change in a ball’s material properties with temperature, have a greater influence on its rebounds characteristics than the alteration in internal pressure. In this paper it was found that a greater change in the dynamic material properties of the rubber was required to simulate a temperature of 283.15 K (75 % increase), in comparison to 313.15 K (20 % decrease). This was also found by Downing (2007a), who demonstrated that for normal impacts there is a greater difference in both COR and contact times between 25 and 10 ºC, in comparison to 25 and 40 ºC.
132 The Engineering of Sport 7 - Vol. 1 Independently adjusting the static material properties of the rubber resulted in an increase in COR and a reduction in contact time, with increasing stiffness. A tennis ball’s static structural stiffness is predicted to be affected by temperature in two ways; i) the material and hence structural stiffness of the ball is reduced with increasing temperature and ii) the lower static stiffness of the rubber at higher temperatures results in the ball expanding more from the applied internal pressure, which in turn increases its volume lowering its initial internal pressure. The diameter of the ball in the FE model was found to increase from 0.066004 to 0.066204 m when the simulated temperature was increased from 283.15 to 313.15 K. This equated to a 0.62 % increase in the frontal area or drag force acting on the ball during flight. Further research is required to determine whether temperature has a significant effect on tennis ball’s cross sectional area and hence flight characteristics. The results show that when just the internal pressure of the ball was updated to simulate a temperature of 283.15 K, the model over-predicted COR. It was also found that increasing the static stiffness of the rubber, to simulate a drop in temperature, resulted in an increase in the COR, making the model over-predict the COR even more. Therefore, if only the internal pressure and static stiffness of the rubber were adjusted the model would over-predict COR at 283.15 K, whilst the opposite would be the case at 313.15 K. An increase in damping results in a decrease in COR (DH1); hence including the change in the dynamic material properties of the rubber resulted in strong agreement with the experimental data. This investigation has provided an indication as to how the static and dynamic material properties of a tennis ball rubber core change with temperature. However, the intention of this study was only to provide a gauge of the extent to which the material properties of a tennis ball change with temperature. In this investigation the static material properties were assumed to adjust with temperature, whilst following the trend of the original data. In reality, it is very unlikely that this would be case, especially in the range from 298.15 to 283.15 K which experienced the largest change in rebound characteristics. In addition, more precise material testing would be required to determine how the static properties of rubber cores change with temperature. Indeed, a more accurate methodology needs to be developed in order to experimentally obtain both the static and dynamic material properties of the ball at different temperatures. The change in felt material properties with temperature may also have an influence on the rebound characteristics of a tennis ball. If the felt were to become less stiff with increasing temperature, then the ball would stretch more from the internal pressure. This would, in turn, increase its initial volume, reducing its internal pressure and hence structural stiffness. However, it is predicted that the stiffness of the felt would change by a very marginal amount within the temperature range used in this investigation. In-depth material testing would be required to quantify how the felt properties change with temperature.
The Engineering of Sport 7 - Vol. 1
133
6- Conclusion An FE model of a tennis ball, validated at room temperature, has been updated to include simulations at temperatures of 283.15 and 313.15 K for inbound velocities in the range from 15 to 30 m·s-1. This was achieved by modifying the internal pressure in accordance with the laws of thermodynamics, whilst simultaneously estimating the change in the rubber core material properties. The model was found to be in good agreement with the experimental data for the entire range of velocities under investigation, at both temperatures. Overall, the change in rubber properties with temperature was found to have a more significant effect on the rebound characteristics of a tennis ball than the change in internal pressure (Downing 2007a). In-depth material testing would be required to determine precisely how the rubber core and felt cover properties change with temperature.
7- Acknowledgements The authors would like to thank the International Tennis Federation (ITF), in particular Matt Downing, for providing the experimental data. They would also like to thank Prince for sponsorship of the project.
8- References [AG1] Allen, T., Goodwill, S. & Haake, S. Experimental validation of a tennis ball finite-element model. Tennis Science and Technology 3, 1, 21-30, 2007. [BC1] Brody, H., Cross, R. & Lindsey, C. The Physics and Technology of Tennis. 1st edn. Racquet Tech Publishing, 2002. [CS1] Calder, C. & Sandmeyer, B. Modelling The Bat-ball Impact Using Finite Element Analysis. 158-159, 1997. [CL1] Cross, R. & Lindsey, C. Technical tennis. 1st edn. Racquet Tech Publishing, 2005. [DH1] Dignall, R.J. & Haake, S.J.. Analytical Modelling of the Impact of Tennis Balls on Court Surfaces. In Tennis Science and Technology. (edited by S.J.Haake and A.O. Coe), First edn. pp. 155162. Blackwell science Ltd., 2000. [D1] Downing, M. The effects of climate changes on the properties of tennis balls. Tennis Science and Technology 3, 1, 49-55, 2007a. [D2] Downing, M. The effect of temperature on the court pace rating of tennis surfaces. Tennis Science and Technology 3, 1, 81-86, 2007b. [GK1] Goodwill, S.R., Kirk, R. & Haake, S.J. Experimental and finite element analysis of a tennis ball impact on a rigid surface. Sports engineering (Sheffield, England), 8, 145-158, 2005. [HS1] Hubbard, M. & Stronge, W.,J. Bounce of hollow balls on flat surfaces. Sports Engineering, 4, 49-61, 2001. [RC1] Rose, P. & Coe, A. The variation of static and dynamic tennis ball properties with temperature. Tennis Science and Technology, 1, 169-174, 2000.
Nonlinear Dynamics of a Simplified Skateboard Model (P24) Alexander S. Kuleshov1
Topics: Skate & other Urban Sports; Extreme Sports; Modelling. Abstract: In this paper the further investigation and development for the simplified mathematical model of a skateboard with a rider are obtained. This model was first proposed by Mont Hubbard (Hubbard 1979, Hubbard 1980). It is supposed that there is no rider’s control of the skateboard motion. To derive equations of motion of the skateboard the Gibbs-Appell method is used. The problem of integrability of the obtained equations is studied and their stability analysis is fulfilled. The effect of varying vehicle parameters on dynamics and stability of its motion is examined. Keywords: Skateboard; Nonholonomic Constraints; Integrability; Stability of Motion.
1- Introduction Skateboarding is one of the most popular extreme sports of today. In 2003 skateboarding had over 11 million participants in the U.S. alone (Frederick et al. 2006), putting the sport on a par with tennis and volleyball in terms of participant levels. However, despite of the growing number of participants skateboarding is poorly represented in the scientific literature. At the late 70th – early 80th of the last century Mont Hubbard (Hubbard 1979, Hubbard 1980) proposed the two mathematical models describing motions of a skateboard in the presence of a rider. To derive equations of motion he used the principal theorems of dynamics. In this paper the further development of the simplest skateboard model proposed by Hubbard is given.
Figure 1- The Skateboard Side View. 1. Department of Mechanics and Mathematics, Moscow State University, Main building of MSU, Leninskie Gory, Moscow, 119991, Russia - E-mail: [email protected]
136 The Engineering of Sport 7 - Vol. 1 The skateboard typically consists of a board, two trucks and four wheels (Figure 1). The modern boards are usually from 78 to 83cm long, 17 to 21cm wide and 1 to 2cm thick. The most essential elements of a skateboard are the trucks, connecting the axles to the board. Angular motion of both the front and rear axles is constrained to be about their respective nonhorizontal pivot axes, thus causing a steering angle of the wheels whenever the axles are not parallel to the plane of the board. The vehicle is steered by making use of this kinematic relationship between steering angles and tilt of the board. In addition, there is a torsional spring, which exerts the restoring torque between the wheelset and the board proportional to the tilt of the board with respect to the wheelset. We denote the stiffness of this spring by k1.
2- Formulation of the Problem. Equations of Motion. The simplest skateboard model assumes that the rider, modeled as a rigid body, remains perpendicular with respect to the board. Therefore, when the board tilts through γ, the rider tilts through the same angle relative to the vertical (Figure 2). Let us introduce an inertial coordinate system OXYZ in the ground plane. Let FR = a is the distance between two axle centers F and R of a skateboard. The position of the line FR with respect to the OXYZ - system is defined by X and Y coordinates of its centre and by the angle θ between this line and the OX - axis (Figure 3). The tilt of the board is accompanied by rotation of the front wheels clockwise through δf and rotation of the rear wheels anticlockwise through δr. The wheels of the skateboard are assumed to roll without lateral sliding. This condition is modeled by the constraints, which are nonholonomic as can be proved:
Figure 2 - The Skateboard Rear and Top View.
(1)
The Engineering of Sport 7 - Vol. 1
137
. . Equations (1) can be solved with respect to X and Y: (2)
Thus the velocities of the points F and R will be directed horizontally and perpendicularly to the axles of the wheels and there is a point P on the line FR which has zero lateral velocity and hence only forward speed u. It may be shown, that (3) Using the results obtained in (Hubbard 1979, Osterling 2004) it is easy to find the following relations between the tilt of the board and the steering angles f and r (4) where f and r are the fixed angles which the front and rear axes make with the horizontal. . . According to (4) equations (2) for x and y can be rewritten as follows: (5)
Figure 3 - The Basic Coordinate Systems.
138 The Engineering of Sport 7 - Vol. 1 Expressions (3) become (6)
Suppose that the board of the skateboard is located a distance h above the line FR. The length of the board is also equal to a. The board center of mass is located at its center. The rider’s center of mass is not be located above the board center of mass, but it is located over the central line of the board a distance d from the front truck. Let l be the height of the rider’s center of mass above the point P. Other parameters for the problem are: mb is the mass of the board; mr is the mass of the rider; Ibx, Iby, Ibz are the principal central moments of inertia of the board; Irx, Iry, Irz are the principal central moments of inertia of the rider. The following parameters will be used below:
It can be proved (Osterling 2004, Kuleshov 2007) that the variables u and satisfy the following differential equations (7)
Here A, …, E, K are the functions of the parameters, namely
Thus, equations (5)-(7) form the close system of equations of the skateboard motion. Moreover equations (7) admit the energy integral
The Engineering of Sport 7 - Vol. 1
139
Equations (7) have the particular solution (8) which corresponds to uniform straight-line motion of a skateboard. Consider the problem of stability of this particular solution. Setting u = u0 + and keeping for its notation we write the equations of the perturbed motion (9)
. Here and are functions depending on , and whose development as the series in powers of said variables starts with terms of at least the second order. Moreover, . these functions identically vanish with respect to when = 0 and = 0:
The characteristic equation corresponding to the linearized system (9) has the form (10) When conditions (11) are fulfilled equation (10) has one zero-root and two roots in the left-half plane. . Since the functions and identically vanish for = 0, = 0, then under conditions (11) the critical case of one zero-root takes place and solution (8) is stable with respect . . to , and u (asymptotically stable with respect to and ). Since a condition is valid for all values of parameters then stability conditions (11) can be finally written in the form (12)
(13) If at least one of the conditions (12)-(13) is not fulfilled then equations (9) has the root in the right-half plane and solution (8) will be unstable.
140 The Engineering of Sport 7 - Vol. 1 One can make now simple conclusions about stability of a straight-line motion of the skateboard using conditions (12)-(13). Note, first of all, that the expression on the left-hand side of inequality (12) contains u0 as a multiplier. This means that the stability of motion depends on its direction. If one direction of motion is stable the opposite direction is necessary unstable. Such behavior is peculiar to many nonholonomic systems. First of all this effect exists the classical problem of motion of a rattleback (aka wobblestone or celtic stone (Bondi 1986, Garcia and Hubbard 1988)). In this problem the stability of permanent rotations of a rattleback also depends on the direction of rotation. Suppose that u0 > 0, f = r = and condition (13) is valid. In this case the stability of motion (8) depends on the location of the rider. If the rider stands closer to the front truck d < a/2, the motion will be stable and when the rider stands closer to the rear truck d > a/2 it will be unstable. When u0 = 0, the skateboard is in the equilibrium position on the plane. In this case the characteristic equation (10) has one zero-root and two pure imaginary roots under the condition (14) It can be proved (Hubbard 1979, Osterling 2004, Kuleshov 2007) that condition (14) is necessary and sufficient condition for stability of the skateboard equilibrium position.
3- Integrable case The characteristic equation (10) can have one zero-root and two pure imaginary roots also in the case when B = 0. In particular, this condition is valid when the skateboard is symmetric f = r and the rider stands in the center of the board d = a/2. One can see that in this case equations (5)-(7) can be completely solved in terms of quadratures. Indeed, in the case B = 0 equations (7) will have a form (15) with the energy integral (16) By introducing a new independent variable in the first of equations (15) this equation can be transformed to the form (17) Equation (17) is differential equation with separate variables. Its solution gives the variable u as a function of : u = u(). Substituting this function into the energy integral
The Engineering of Sport 7 - Vol. 1
141
. (16) and solving it with respect to one obtains the following differential equation for : (18) Equation (18) is also differential equation with separate variables. The solution of this equation gives the angle as a function of time: = (t). By substituting this function in the expression for u = u() this expression can be rewritten as follows: u = u((t)) = u(t). Having the functions = (t) and u = u(t) it is easy to find all the remaining variables as functions of time. Indeed the angle satisfies the following differential equation
and therefore (19) The coordinates X and Y of the board center of mass satisfy the following differential equations
Since the functions u = u(t), = (t), = (t) are already known, it is easy to obtain by integration (20)
Thus, in the case B = 0 all unknown functions in the problem can be expressed as functions of time by formulas (17)-(20).
4- Conclusions In this paper the problem of motion of a skateboard with a rider was examined. This problem has many common features with other problems of nonholonomic dynamics. In particular it was shown that the stability of motion of the skateboard depends on the direction of motion. The similar effects have been found earlier in the classical problem of a rattleback dynamics. It was found also the integrable case in the problem. Note that the integrable problems are very rare in nonholonomic mechanics and therefore these results seem to be interesting and helpful.
142 The Engineering of Sport 7 - Vol. 1
5- Acknowledgements This research was supported financially by the Russian Foundation for Basic Research (grant 08-01-00363).
6- References [B1] Bondi H. The rigid body dynamics of unidirectional spin. In Proceedings of the Royal Society of London. Series A, 405: 265-274, 1986. [FD1] Frederick E.C., Determan J.J., Whittlesey S.N. & Hamill J. Biomechanics of skateboarding: Kinetics of the Ollie. In Journal of Applied Biomechanics, 22: 33-40, 2006 [GH1] Garcia A. and Hubbard M. Spin reversal of the rattleback: theory and experiment. In Proceedings of the Royal Society of London. Series A, 418: 165-197, 1988. [H1] Hubbard M. Lateral Dynamics and Stability of the Skateboard. In ASME Journal of Applied Mechanics, 46: 931-936, 1979. [H2] Hubbard M. Human Control of the Skateboard . In Journal of Biomechanics, 13 : 745-754, 1980. [K1] Kuleshov A.S. Mathematical Model of a Skateboard with One Degree of Freedom. In Doklady Physics, 52(5): 283-286, 2007. [O1] Osterling A.E. MAS 3030. On the skateboard, kinematics and dynamics. School of Mathematical Sciences. University of Exeter, United Kingdom. 2004. http://www.longboard.nu/files/theSkateboard.pdf
Cricket Batting Stroke Timing of a Batsman When Facing a Bowler and a Bowling Machine (P26) Alex Cork1, Laura Justham1, Andrew West1
Topics: Lawn Sports (Hockey, Cricket). Abstract: Cricket batsmen must evaluate each delivery and select a shot to play from temporal and spatial information gained from the movements and anatomical position of the bowler (pre-release cues), initial ball flight and their own previous experience. Current batting training methods often make use of bowling machines; however these machines offer the batsman no pre-release visual cues pertaining to ball type. Previous research has suggested that highly skilled batsmen are superior to less skilled players in the pick-up of pre ball flight information and that the bowling arm provides referential information regarding ball type. A pilot study has been conducted to establish the different approaches adopted by a batsman when faced with a human bowler and a bowling machine. The movements and reaction times of an amateur (English premier club level) batsman are compared for both scenarios. Front and side on high speed video footage was recorded of the batsman when facing a random selection of deliveries from a human bowler and a bowling machine. Front on high speed footage of the bowler/bowling machine was simultaneously recorded allowing the responses of the batsman to be analysed when facing a human bowler (with visual cues) and a bowling machine (without visual cues). Keywords: Cricket Batting, Bowling Machine, Movement.
1- Introduction An express paced delivery (90mph and higher) as classed by Abernethy (1981) takes, approximately, 450msecs to travel from the bowler’s hand to striking the bat. Previous experimentation has shown that combined choice reaction time and movement time for a choice of just four possible strokes would equal approximately 700msecs, so, effective batsmen must use some form of anticipation when facing even moderately fast bowling (Gibson and Adams, 1989). The ability to accurately predict the length, direction and pace of the delivery from the movements of the bowler prior to delivery is therefore a 1. Sports Technology Research Group, Sports Technology Institute, Loughborough Science and Enterprise Park, 1 Oakwood Drive, Loughborough, Leicestershire, LE11 3QF, UK - Email: A.E.J.Cork, L.Justham, [email protected]
144 The Engineering of Sport 7 - Vol. 1 skill that will greatly enhance a batsman’s ability. This skill becomes increasingly important when facing express paced bowlers where the time constraints imposed by ball velocity will typically exceed the time available to process information or high class spin bowlers who are able to disguise the changes in their action that lead to subtly different deliveries. The movements made by batsmen in cricket are the result of feedback gained from the movements of the bowler and the ball in the air during the delivery. Information about an individual delivery arising from the movements and anatomical positions of the bowler will be added to prior knowledge the batsman has about the pace and style of the bowler, the pitch conditions and their own form, culminating in a decision being made as to which stroke they elect to play. Using existing bowling machines coaches are unable to present the batsman with a realistic match environment in with to train.“Coaches operating bowling machines typically provide some cue as to the time at which they place the ball in the machine” (Gibson and Adams, 1989) usually by raising their hand prior to dropping a ball into the machine. This clearly does not provide batsmen any pre-release information pertaining to ball type, neither does it offer batsmen the opportunity to look for such information. Some bowling machines operate with warning lights on the front of them; this is usually in the form of a “traffic-light style” countdown to release. However in a match or when facing a bowler in the nets a batsman must recognis the moment of release during the delivery stride, there are no warning lights or raised hands to help. Further to this the direction and length of the delivery can be ascertained from viewing the angle of the head of the bowling machine as recognised by Gibson and Adams (1989). This is clearly not a match realistic pre release cue and could cause deficiencies in technique to develop in batsmen. Abernethy has deduced that “for skilled players, information to help predict the length of a bowled ball is available prior to release” (Abernethy et al., 2005). Thus if a batsman is using the angle of the bowling machine head to determine the line and length of delivery, when presented with a match situation he is poorly prepared to face deliveries as he has little exposure to the pre release cues emanating from a bowler’s action. In cricket expert batsmen have shown a persistent capability to use early sources of information to aid shot selection which other skill groups are not attuned to (Muller et al, 2006). There are three key abilities encompassed within the anticipatory skill demonstrated by expert batsmen; (i) visual search: selecting the areas the eyes will focus upon during the delivery stride and release, (ii) selective attention: picking out the key events within the bowling action that relate to ball type and finally, (iii) discrimination ability: being able to recognise the movements of the bowler and interpret them into the resultant ball type. It is therefore reasonable to assume that each of these abilities is aided by experience. Perhaps the area that is most enhanced by experience is discrimination ability where players are provided with a wider store of potentially relevant memories and a rapid and automatic access to these (Schneider and Fisk, 1983). In conclusion; does training with bowling machines add to this store of relevant memories and aid batsman in match scenarios when attempting to evaluate an approaching delivery from the movements of the opposing bowler?
The Engineering of Sport 7 - Vol. 1
145
The pilot study described in this paper focuses upon the movement patterns of a premier league club cricketer when facing a human bowler and a typical bowling machine. The aim of this study was to see whether there were differences in the timing of movements of the batsman when faced with the two different scenarios. While there is no doubting the benefit of using bowling machines to teach and practice the technique of particular shots, they do not provide batsmen with relevant information about approaching deliveries, neither do they add to the memory store of previously faced deliveries available to batsmen. From the results of this study it is possible to make an initial judgement on the validity of using bowling machines to train for match scenarios. While there have been previous studies conducted in this area, notably Gibson and Adams (1989), it was felt that with highly accurate, modern equipment available such as a number of high speed video cameras offering a high image resolution, the HawkEye ball tracking system and a realistic training environment for testing to be conducted within, additional knowledge could be developed within this area of research.
2- Method Player testing was conducted at the ECB National Cricket Centre, Loughborough. An amateur batsman, who plays for a premier league club side was filmed batting against a non-familiar human bowler of the same standard and a Bola bowling machine. The movements of both the batsman and bowler/bowling machine operator were recorded from the initiation of the bowler’s run up/machine operator’s signal to the completion of the batsman’s shot. The batsman was filmed from both a front on and side on perspective and the bowler/bowling machine was filmed from a front on perspective. The cameras were synchronised and recording was initiated using an SV TTL trigger when the bowler broke a laser beam positioned at the beginning of the run up. For the deliveries bowled by the bowling machine, recording was initiated when the bowling machine operator broke a laser beam when signalling an imminent delivery to the batsman by raising his arm. A specifically designed test bed was assembled to house two high speed video cameras, one was focussed upon the batsman and the second was focussed upon the bowler, both from a front on perspective. The test bed was positioned in the centre of the designated pitch at a distance of 6 metres from the stumps at the bowler’s end. Both cameras were Photron SA1 High Speed Video Cameras. The cameras were set up to sample at 500 frames per second which enabled data to be recorded at resolution of 1024 x 1024 pixels. A third high speed camera was positioned orthogonally to the pitch, in line with the batsman’s crease on a standard Manfrotto tripod at a height of 1.2 metres, focussed upon the batsman. This camera was a Photron Ultima APX high speed video camera which also sampled at 500 frames per second at a resolution of 1024 x 1024 pixels. Additional data was recorded for each delivery using the HawkEye system installed within the indoor facility. HawkEye is a computer based ball tracking system that uses three orthogonal cameras recording at 140 frames per second to capture the trajectory of the ball before and after it bounces. Mathematical algorithms are then used to predict the complete ball trajectory. These data were used to give an overview of each delivery faced by the batsman and enabled the ball flight characteristics of the deliveries bowled to be quantified.
146 The Engineering of Sport 7 - Vol. 1
3- Procedure The batsman was allowed a short period to warm up and faced two deliveries from the human bowler prior to the commencement of recording to test the synchronisation of the equipment and to acclimatise the batsman to the style and pace of the bowler. The batsman then faced twelve recorded deliveries. He was asked to bat normally and without any irregular movements that may have been intended to affect the bowler. The bowler was asked to bowl consistently, at his normal pace, line and length. After a prolonged break testing resumed as before with a bowling machine replacing the human bowler. The batsman faced two deliveries, as with the human bowler, to check the synchronisation of equipment and to allow acclimatisation to the pace and style of the bowling machine deliveries. The average speed of the human bowler’s deliveries were determined from the HawkEye data and the bowling machine was set to replicate this. The bowling machine operator warned the batsman of an imminent delivery by raising his hand with the ball prior to inserting it into the machine; this was considered to be the most commonly used method adopted by coaches. Between deliveries, the angle of the head of the bowling machine was marginally altered in an attempt to reproduce some of the variability in length seen in human bowling, without any short balls or Yorkers. The line of the deliveries was kept consistent; on the stumps.
4- Results Each of the video cameras sampled at 500 frames per second, therefore each frame equated to 0.002 seconds. It was subsequently possible to establish the timings of events during each ball faced. Mean times were calculated for each event during the batting stroke, and the mean times for foot movements were also recorded. Figure 1 is a series of still images taken from the front and side on high speed footage of the batsman during the study at six key points identified during the batting stroke.
Figure 1 - Still images taken from the high speed video of the batsman from both front and side on at six key positions during the batting stroke (from left to right); Bat back lift, Bat held level, Movement upwards, Bat reaches top of backswing, Movement down, Bat and ball contact.
The Engineering of Sport 7 - Vol. 1
147
The mean times for each of the events during the batting stroke with respect to ball release for all of the deliveries faced from a human bowler and a bowling machine are displayed in table 1. The events are chronologically ordered. The point of release was considered to be the first frame in which the bowler no longer had contact with the ball or the moment the ball emerged from the “mouth” of the bowling machine, examples of these points can be seen in figure 2. On the occasion that there was no bat and ball contact made then the frame in which the ball was level with the batsman’s bottom hand was assumed to be the point of contact. The sixth column of table 1 gives the calculated t-values for each of the events. This value indicates whether there are significant differences when the batsman faces the bowler and bowling machine. It should be noted that against the bowling machine there was only a secondary front foot movement seen on nine of the twelve deliveries faced. When facing the bowler, the batsman had a secondary movement of the front foot on all twelve of the deliveries faced. Thus the mean and t values for the secondary front foot movement have been calculated for nine deliveries, with the corresponding delivery numbers being selected from each of the data sets for analysis. Table 1 - The mean times, in seconds, of specified events during the batting stroke with respect to ball release by a bowler or bowling machine. * Denotes a significant difference between the two sets of data. Critical t: (22df, two tailed, of 0.05) = ±2.074, (16df, two tailed, of 0.05) = ± 2.120.
From the data in table 1 it is possible to say that there are significant differences in the timings of the batsman’s bat back lift and front foot movement up and down when facing a bowler and a bowling machine. Against a bowling machine the batsman’s bat back lift was significantly earlier than against a bowler. His front foot movement, both up and down was significantly later against a bowling machine than against a bowler. Once the front foot has been planted, there is little difference in the timings of the batsman’s movements up until bat and ball contact. The discrepancy between the values for secondary movement down of the front foot can be explained by the fact that on two occasions, when facing deliveries from the human bowler, the batsman played shots with only his back leg on the ground and only planted his front foot once he had played the shot. The timings of events during the operation of the bowling machine and the delivery stride of the human bowler are displayed in table 2. These events are highlighted in figure 2. From the figures it can be said that there is more variation in the times of events for
148 The Engineering of Sport 7 - Vol. 1 the operation of the bowling machine than the delivery stride of the human bowler. The initiation of arm movement upwards by the bowling machine operator commenced, on average, 1.131 seconds before the ball was released. In contrast to this, the human bowler began the bound of the delivery stride 0.697 seconds prior to releasing the ball.
Table 2 - The mean times, in seconds, of specified events during the operation of the bowling machine and run up and delivery stride of the human bowler during testing with respect to ball release.
Figure 2 is a series of still images of both the human bowler and the bowling machine taken at key points during the delivery stride and operation of the machine respectively. The images have been arranged sequentially with respect to release. It should be noted that there are only three images for the bowling machine; this is because there was no discernable difference between the signal and release stages from the batsman’s perspective.
Figure 2 - The visual information available to the batsman from a human bowler (top) and a bowling machine (bottom) (from left to right); Top: The Bound, Rear foot impact, Front foot impact, Release. Bottom: Initiation of arm movement, Signal to batsman, Release.
The Engineering of Sport 7 - Vol. 1
149
The data displayed in table 3 are the mean values for the physical features of the deliveries bowled by both the human bowler and the bowling machine. The data is concerned with the release and pitching position of the deliveries and their release speed.
Table 3 - Mean values of features of the deliveries bowled by the human bowler and bowling machine during testing. Features are measured in metres unless stated, negative values indicate distances towards the off stump.
5- Discussion The results gained from this pilot study offer an insight into the different batting approaches employed by a batsman when facing a bowler and a bowling machine. Against a bowling machine the batsman takes longer to reach the top of his backswing and although he reaches the top at a similar time with respect to ball release, he initiated the movement considerably earlier. The timings of the machine operator’s signal and the bound of the bowler’s delivery stride are very similar. It could therefore be assumed that the batsman is taking his cue to initiate bat pick up from the earlier occurring initial arm movement of the machine operator rather than waiting for his signal. The data in table 3 confirm that there were differences between the deliveries that the batsman faced from the human bowler and the bowling machine. The human bowler released the ball from wider in the crease; he also pitched the ball wider than the bowling machine and considerably shorter. Both the bowler and the bowling machine were consistent in the line they bowled, however there was considerable variation in the length that the human bowler bowled. There was also greater variation in the position from which the ball was released. Despite this the batsman was still more consistent in the early part of the batting stroke, through to the top of the backswing, when facing the human bowler. The batsman commented after the test that he felt that he was waiting at the top of his backswing in anticipation of the ball being released by the bowling machine. This is reflected in the earlier pick up of his bat and the prolonged time it took him to reach the top of his backswing in contrast to facing a human bowler. A very clear indication of the visual differences between each scenario from a batsman’s perspective can be seen in figure 2. One of the clearest indicators of pre-release information being used is the consistently earlier movement of the batsman’s front foot when facing the human bowler. The technique observed by most batsmen is to move your front foot forward, towards the pitch of the ball if it is a full delivery, or to move your feet backwards if the delivery is short. For the batsman to be moving his feet before the ball has been released, he must
150 The Engineering of Sport 7 - Vol. 1 be interpreting information received from the movements of the bowler and predicting the length of the imminent delivery.
6- Conclusion It is clear that there is a need to conduct this testing with a larger player collection of varying standards, before it will be possible to form a clear judgement on the differences observed in players batting against a human bowler and a bowling machine. It is however possible to conclude from this initial pilot study and previous research conducted, that there appears to be a different approach adopted by batsmen when facing bowling machines to that seen against human bowlers. Further work should include the testing of players of different abilities; this may allude to the importance of visual cues to higher skilled players with the hypothesis that more experienced players are more reliant on earlier visual cues to interpret an imminent delivery. They will gain information later and therefore have to react more quickly to the bowling machine where the only information is available post release. Further to this it would be of benefit to monitor the batsmen’s eye movements from the start of the bowler’s run up to the point of bat and ball contact either through the use of an eye tracking system or a screen based test.
7- Acknowledgements The authors would like to thank the ECB for the use of the National Cricket Centre. They would also like to thank the participants in this study for their time.
8- References [A1] Abernethy, B. (1981) Mechanisms of Skill in Cricket Batting. Australian Journal of Sports Medicine 13: 3-10. [AM1] Abernethy, B. Muller, S. Farrow, D. Guy, W. Barras, N. (2005) Dealing With Natural Constraints: The Timing of Information Pick Up by Cricket Batsmen of Different Skill Levels. In Proceedings of the ISSP 11th World Congress of Sports Psychology (CD). Sydney: International Society of Sports Psychology. [GA1] Gibson, A.P. and Adams, R.D. (1989) Batting Stroke timing with a Bowler and a Bowling Machine: A Case Study. The Australian Journal of Science and Medicine in Sport 21(2): 3-6. [MA1] Muller, S. Abernethy, B. Farrow, D. (2006) How Do World Class Cricket Batsmen Anticipate a Bowler’s Intention? The Quarterly Journal of Experimental Psychology 2006, 59(12): 2162-2186. [SF1] Schneider, W. Fisk, A.D. (1982) Degree of consistent Training: Improvements in Search Performance and Automatic Process Development. Perspectives in Psychophysics, 31(2): 160-168.
Estimation of a Runner’s Speed Based on Chest-belt Integrated Inertial Sensors (P27) Rolf Vetter, Emanuel Onillon, Mattia Bertschi1
Topics: Athletics, biomechanics, fitness. Abstract: In long distance running, real time monitoring and performance optimization has been recently rendered possible through the commercialization of a large variety of running computers. Generally, such computers simultaneously monitor several parameters such as heart rate, running speed, stride frequency and stride length. More precisely, stride and speed information is obtained using foot located inertial sensors. Unfortunately, due to its leg extremity location, the foot inertial sensor presents some disadvantages: it may be perceived as cumbersome; it requires supplementary telecommunication facilities as well as local power supply; it may increase a runner’s energy expenditure even though it weights only a few tens of grams. To overcome the above-mentioned drawbacks we have developed a method for the estimation of a runners speed, stride frequency and stride length based on inertial sensors which may directly be integrated in classical chest-belt heart-rate-monitors. The method processes various kinetic features from chest located accelerometers. The instantaneous speed is then estimated through a linear mapping of the most reliable kinetic features. The determination of the most reliable kinetic features as well as the mapping coefficient is achieved through a minimal variance approach on a calibration run. A validation has been performed on 9 subjects on various running surfaces and over durations of 1 year in order to highlight the consistency of the proposed approach. We have found that different runners necessitate different optimal weighting factors of the kinetic features. This may be related to a runner’s stride efficiency, that is, more or less energy wasting and bouncing trajectories for similar forward speed. However, for a given runner we obtained consistent results namely a speed estimation accuracy of about 7% in a range of 60% to 110% of his maximal aerobic speed. Keywords: running speed; stride length; stride frequency; inertial sensor; chest-belt. 1. Swiss Center of Electronics and Microtechnology, Neuchâtel, Switzerland - E-mail: rve, eon, [email protected]
152 The Engineering of Sport 7 - Vol. 1
1- Introduction Recent progresses in micro-electronics, biomedical signal processing, biomechanics, and telecommunications have allowed the development of various portable devices for monitoring vital parameters for outdoor activities. Whereas during the past the monitoring abilities of such devices focussed mainly on cardiovascular parameters such as heart rate, there is actually a tendency to extend their features to kinetic properties such as acceleration, velocity, movement frequency, and position of the subject. The motivations of such a tendency include subject specific personal interest, training and performance considerations of serious athletes, rehabilitation of disabled, injury prevention, and the design and analysis of footwear. In running or jogging activities, there is also a demand for monitoring heart rate, speed and step frequency in view of performance analysis and optimization [ML1]. Various portable devices for running speed estimation have been recently developed. They can be mainly categorized in the following classes with respect to the underlying technologies: inertial sensors, GPS, Doppler-microwave, and pedometers. Devices based on Doppler-microwave techniques and pedometers are not currently able to provide sufficient accuracy for jogging or running activities. GPS techniques provide high accuracies in open field use but they are not able to provide information about step frequency which is an important parameter in running performance optimization. A very attractive solution is based on inertial sensors. Polar has brought recently such a solution on the market [FR1]. Experimental evidence has shown that the estimated speed is highly accurate if a preliminary calibration is performed. At the calibration speed one obtains an accuracy of about 2%. This accuracy degrades slightly as one runs much slower or faster than the calibration speed. An inconvenient of such a device resides in the fact that it requires an accelerometer to be fixed on the running shoe [RF1]. This is somewhat cumbersome and the supplementary weight may decrease a runner’s performance. Moreover, such a solution requires a supplementary telecommunication link between the foot-located sensor and the processing device and thus increased power consumption. To avoid the need of a supplementary accelerometer device on the lower leg extremity we have developed a method based purely on an accelerometer integrated in an existing chest belt [V2].
Figure 1 - Illustration of variations in trunk inclination over a whole running cycle of two strides. The dashed lines illustrate at each phase of the cycle the inclination of the accelerometer plane located in a classical commercial chest belt. On the right side we show the two accelerometer signals a1 and a2 with respect to the dashed plane.
The Engineering of Sport 7 - Vol. 1
153
2- Methods 2.1 Developed System and data acquisition In order to develop and validate an algorithm for the estimation of a runner’s speed we have developed a portable system consisting of a 3d accelerometer integrated in a commercial chest belt and an associated wire-linked Holter. The Holter records the accelerometer data at a sampling rate of 256 Hz after anti-aliasing filtering and 12 bit analogto-numeric conversion. The recorded data were then uploaded on a PC where data processing was achieved offline.
2.2 Background The focus of the proposed approach is to exploit acceleration measured by an inertial sensor located in a chest belt to estimate a runner’s speed. Trunk located acceleration measures in experimental studies have been used to assess the transmission and attenuation of heel-strike since the early nineties [V1]. However, they were not exploited for speed estimation. In contrast, studies using force platforms have focused on various aspects of a runner’s kinetic behavior [C1,C2,M1]. Notably, it has been shown that the integral of the vertical ground reaction force over the whole stance period is related to a runner’s speed [M1]. More recently Hunter et al have highlighted the relative importance of propulsive, vertical and breaking impulse in sprinting [H1]. The different impulses have been processed using ground reaction forces obtained through force platforms. They concluded that speed estimation could be performed using the propulsive and breaking impulse with a relative importance of 57% and 7% respectively. The main differences between ground reaction forces and chest located accelerometer measurements reside in the stability of their measurement referential. Ground reaction forces are measured by force platforms which have time invariant measurement referential. In contrast, chest located accelerometer change their inclination during a whole running cycle as illustrated in Figure 2. Changes may depend on a runner’s stride efficiency, his technique, and his core stability [C1]. Therefore, chest-located accelerometers grasp a runner’s vertical or forward acceleration up to an instantaneous rotation. If the trunk inclination angle was known the instantaneous rotation could be compensated through matrix processing. However, the rotation angle is not known and we measure on the two acceleration signals a1 and a2 mixtures of a runner’s vertical and frontal acceleration. In long distance running, stride wise vertical acceleration variations are much larger than variations in frontal acceleration. Thus, rotational artifacts on accelerometer signal a1 may be neglected while this is not the case accelerometer signal a2.
2.2 Developed method The global concept of the algorithm is represented in Figure 2 and operates as follows. At each stride 3 features which may convey information about a runner’s speed are extracted from the acceleration signals. Firstly, we used a feature whose correlation with speed has been assessed during previous studies on force platforms. The feature F1 is obtained
154 The Engineering of Sport 7 - Vol. 1 through the average value over the stance phase of the accelerometer signal a1 such as to mimic the stance wise average value of the vertical ground reaction forces [M1]. Our tentative to process features from the acceleration signal a2 to mimic breaking and propulsive impulses as they have been used on force platforms [H1] was unsuccessful. As highlighted in the previous section, this may be due to rotational artifacts on the signal a2 mediated to a runner’s varying trunk inclination throughout a stride.
Figure 2 - Bloc scheme of proposed algorithmic concept for the estimation of a runner’s speed.
Secondly, we processed a feature whose relation to speed became apparent in the present study. Figure 1 illustrates this experimental evidence on the average stance stride pattern of accelerometer a1 at various running speeds. One can observe that the slope of the signal at toe-off, namely at t = 0 sec, depends on a runner’s speed. Thus, the feature F2 was processed as the minimum value of the instantaneous derivate of the accelerometer signal a1 around toe-off.
Figure 3 - Typical vertical acceleration pattern for the stance phase recorded by the proposed portable device. The pattern we show have been obtained through averaging subsequent instantaneous stride patterns over 800m run on a track at the indicated speed.
The Engineering of Sport 7 - Vol. 1
155
A third feature was processed exploiting knowledge of advanced signal processing together with kinetic considerations. Indeed, a runner’s instantaneous speed reaches its maximum during the flight phase. The distance he covers during the flight time depends mainly on his total speed and thus on his impulse at toe-off. Stride wise impulses are related to stride wise speed variability. Thus, in order to grasp the total impulse in the vectorial sense we proceeded to a stride wise principal component analysis (PCA) of variability of speed, namely the integral of accelerometer signals a1 and a2 after highpass filtering at 0.5 Hz. The instantaneous PCA achieves a rotation of the observation space [H2] and may thus compensate rotational artifacts due to varying trunk inclination of a runner. Furthermore, PCA’s rotation is performed in such a way that the first principal component yields in our case the norm of the maximum speed variation in the vectorial sense, that is, a variable related to the norm of the vectorial impulse over the stance period. Further features have been processed and tested but not retained due to their inability to represent accurately a runner’s speed. The developed algorithm processes in a subsequent step for each feature and at each stride an estimation of a runner’s speed through a linear affine mapping. It is important that the map is linear and affine such as to minimize the unknown parameters. Indeed, in a linear affine function only one parameter is unknown. Therefore, it can be estimated during one calibration run. The calibration run should typically be performed on a distance of 800m to 2000m. The average calibration speed is processed as the ratio of calibration distance to calibration run duration. For each feature we obtain the associated calibration coefficient as follows. (1)
Three speed estimates are finally obtained by multiplying the three features by their respective calibration coefficient. During the calibration run we selected the most reliable feature for a given runner through a minimum variance approach. Indeed, the variance of the stride wise estimates of the three speeds provides us information about their reliability. The feature providing speed estimates of minimum variance over the calibration run is automatically the best according to statistical considerations [H1]. Finally, to further improve the statistical reliability of the estimated speed a median filter over 40 strides was applied.
3- Validation The validation of the proposed algorithm has been achieved in two phases: in a first phase we have assessed whether features and running speed could be related through linear relationship during 800m running intervals on a track; in a second phase we focussed on the assessment of instantaneous speed estimation in off-track running. The features processed by our method were as explained in the previous section: F1 the average of the accelerometer signal a1 over the stance segment; F2 the slope of the acce-
156 The Engineering of Sport 7 - Vol. 1 lerometer signal a1 at toe-off; F3 the maximum Eigenvalue of a principal component analysis of the two dimensional stance wise speed variability obtained through integration of accelerometer signal a1 and a2.
3.1 Assessment on 800m track intervals This validation was performed on a 400m track on 17 subjects ranging form occasional joggers to elite runners. Subjects were requested to run five 800m intervals at regular speeds raging from 60% to 110% of their maximal aerobic speed. The covered speed range of the total database extended from 7 to 24km/h. The duration of each interval was recorded to evaluate the average speed. The average value of feature F1 was then processed for each interval and the accuracy of its linear relationship with the average interval speed assessed. We found that a linear relationship could provide average speed estimates with mean absolute relative error over the whole database of 3%. Subject specific results of the mean absolute relative error ranged from 1% to 5%. The mean correlation coefficient over the whole database was 0.96 with a subject specific minima and maxima of 0.94 and 0.99. This validation allows us to ascertain that the proposed method allows an estimation of an average speed using the feature F1 and a linear mapping. However, it does not provide any information about the ability of the proposed method to track variations in speed as they may be observed in long distance runners. Moreover, it is important to evaluate the invariance of the mapping function and the performance on various running surfaces.
3.2 Assessment on off-track running In order to gain further insights on the performance and to highlight the consistency of the proposed approach a second validation has been performed on nine subjects, various running surfaces and over a period of 1 year. Runners achieved during the test period 3 to 15 evaluation sessions including a calibration run of 800m and, subsequently a validation run of 20 to 60 minutes. As a reference for the calibration and the evaluation run we used a previously calibrated RS800 of Polar. Results for the runners are shown in Table 1. A brief analysis of the mean absolute relative error (MARE) between the estimations of the proposed algorithm and the RS800 points out that MARE are considerably higher than in the previous validation on the track. Notably, the MARE for feature F2 ranges from 13% to 24% whereas we obtained 1% to 5% on the track. The reason for this degradation may be related to the nature of this feature. Indeed, F2 is a derivative feature which implies that it is very sensible to noise. In the validation on the track an average value over 800m has been computed and noise related inaccuracies may thus be diminished through averaging. In contrast, in the present validation short term estimation has been achieved and therefore estimator noise may appear clearly.
The Engineering of Sport 7 - Vol. 1
157
Table 1 - Detailed speed estimation results for the nine subjects and various recordings over a duration of one year. We show mean ± standard deviation of the mean absolute relative error (MARE) and the mapping coefficients of the three features.
A further analysis highlights that best results in terms of MARE are obtained for feature F3 in 6 out of 9 subjects. For the subjects No 3, 6 and 7 the feature F1 performs better. However, even if for each subject the best feature was selected we end up with MARE with respect to the RS800 of Polar ranging from 5% to 8% (mean=7.1%, standard deviation=1.2%). The causes for this high values may be numerous. Firstly, the experimental protocol may have been too ambitious. The evaluation runs have been executed on a large variety of surfaces such as tar, track, gravel, grass and hilly forest trails. It is obvious that the different surfaces and profiles may alter a runner’s stride and thus the kinetic features of a runner’s chest [C1]. Secondly, it is questionable if the RS800 of Polar yields highly accurate estimations for such a large variety of running surfaces.
Figure 3 - An example of speed estimation during a progressive run.
Finally, an analysis of the mapping coefficient of the most reliable results, namely, c1 or c3 shows that they are subject-dependent. Values for coefficient c1 are distributed in uniform manner from 62 to 99. Coefficient c3 cover also a range from 53 to 85. However, one can observe that they are mainly grouped in two clusters: the first around 82 is put in evidence in the table through bold fonts; the second around 53 is highlighted through
158 The Engineering of Sport 7 - Vol. 1 italic fonts. The fact that different runners necessitate different mapping coefficients may be related to a runner’s stride efficiency [CO1,V2]. On one hand, stride cycles are accomplished by runners with more or less energy wasting, bouncing trajectories for similar forward speeds [C1]. The amplitudes of the bouncing trajectories are related to vertical speed changes or impulses [H1]. On the other hand, the instantaneous forward speed of a runner changes slightly during each stride cycle. It slows down at touch down during the early stance phase due to the breaking impulse and increases again during the late stance phase towards toe-off due to the propulsive impulse [H1]. Elite runners with high running efficiencies are able to come along with lower vertical, breaking and propulsive impulses at a given frontal speed than novice runners [CO1]. Features F1 and F3 are a function of stride wise speed variations in the plane a1-a2 (see Figure 1) and the mapping coefficient are therefore related to the relationship between mean forward speed and stride wise speed variations. Thus, runners with lower speed variations for a given forward speed and which may be considered as more efficient in accordance with the above considerations, have larger mapping coefficients c1 and c3. In Figure 4 we show a typical example of a speed estimation using the proposed method. The device has been calibrated in a run over 800m on the track. After a recuperation of 10 minutes the runner started off for two progressive segments of about six minutes. One can observe that the speed estimated by the proposed approach follows with small errors the speed estimated by the Polar RS800 as long as the running speed is not too slow. Notably, during the walking segments in the beginning and at time t=600 seconds, the estimation errors become considerable. A reason for this erroneous behavior of the algorithm resides in the fact that walking and running are completely different form the biomechanical point of view [N1]. In walking, potential and kinetic energy reach their maximum in phase opposition while in running both of them are in phase.
4- Conclusions We have developed a new method for the estimation of a runner’s speed based on a chest located accelerometer. The method has been validated in two different phases. Feasibility and linear mapping characteristics have been assessed on a track where average speed could be estimated with mean absolute relative errors of 1% to 5% in 17 subjects and for a large speed range. The ability of the proposed method to provide instantaneous speed estimation for a large variety of outdoor running surface was assessed on 9 subjects and over duration of 1 year. Higher mean absolute errors of about 5% to 9% have been obtained when compared with the commercial system RS800 of Polar. However, taking into account of the large variety of running surfaces and the lack of an absolutely accurate reference system for such conditions, the results are very promising. A major limitation of the proposed solution is the mandatory user specific calibration. The advantages of the proposed solution, when compared with classical foot located sensing, are manifold: The accelerometer can be integrated directly in the existing chest belt used for the pulse measurement; The power consumption is reduced since no telecommunication facilities are required for a supplementary link with the foot located sensor; Compactness and effectiveness for the runner are improved since no foot located sensor is required any more.
The Engineering of Sport 7 - Vol. 1
159
5- References [C1] P.R. Cavanagh. Biomechanics of distance running. Human kinetics books, 1990. [C2] G.A. Cavangna. Force platforms as ergometers. Journal of applied physiology, vol. 39 pages 174-179, 1985. [CO1] J. Cornuz, E. Onillon, R. Vetter. Procédé et dispositif de mesure de l’efficacité d’un geste sportif, EP1586353, 2005. [FR1] K.R. Fyfe, J.K. Rooney, K.W. Fyfe. Motion analysis system, US 2002/0040601, 2002. [H1] J.P. Hunter, R.N. Marshall, P.J. McNair. Relationships between ground reaction force impulse and kinematics of sprint-running. Journal of applied biomechanics pages 31-43, 2005. [H2] S. Haykin. Neural networks. Prentice Hall, 1994. [M2] D.I. Miller. Ground reaction forces in distance running. Chapter 8, Biomechanics of distance running, Human Kinetics Books, pages 203–224., 1990. [M3] C.F. Munro. Vertical ground reaction forces in running. Journal of biomechanics, vol. 20, pages 147-155, 1987. [ML1] M. Marquardt, C. Loeffelholz, B. Gustafsson. Die Laufbibel. Spomedis, Hamburg, Germany, 2006. [N1] T.F. Novacheck. The biomechanics of running. Gait and posture, vol. 7, pages 7795., 1998. [V1] G.A. Valinat. Transmission and attenuation of heel-strike accelerations. Chapter 9, Biomechanics of distance running, Human kinetics books, pages 225-247, 1990. [V2] Rolf Vetter. Procédé et dispositif de détermination de la vitesse d’un coureur. WO2007017471, 2007. [RF1] J.K. Rooney, K.W. Fyfe, K.R. Fyfe, W. Bortz. Shoe clip. US 2003/0000053, 2003.
Design and Construction of a Custom-made Lightweight Carbon Fibre Wheelchair (P28) Marc Siebert1
Abstract: Mobility can be equated with quality of life. For handicapped (paraplegic) people, a lightweight wheelchair means an enormous facility. Particularly with regard to independent mobility the weight of the wheelchair is very important, because the user has to lift the wheelchair very often especially when getting into his vehicle [S1]. The aim was to create an extreme light and perfect fitted wheelchair to increase the mobility of handicapped people. Therefore in a first step, dynamic and static tests have been done to learn more about the loads of a wheelchair. In a second step, an adjustable measurement wheelchair has been developed. With the measurement wheelchair it is possible to adjust the seating position separately from the ride characteristics. It is not a static process and not only the measurement of body lengths, the handicapped person can ride the wheelchair, feel the seating position and the driving behaviour and make sure for example, if the breakover point is perfect. After the fitting process in a third step the computer aided construction starts by using the data of the measurement chair to create a 3D-model of the wheelchair. The last step is the manufacturing of the wheelchair and therefore a special flexible manufacturing system has been developed. To get a high performance and extreme lightweight wheelchair, carbon fibre reinforced plastic (cfrp) has been used. A couple of different proceedings for the parts, such as filament winding, hand wrapping and compression methods have been used. For the wheelchair frame the tube-to-tube-process and special adhesive joining techniques have been used. As a result of the application of high tech materials and joining methods it was possible to realize an unique lightweight 5.9 kg carbon wheelchair for the daily use. Keywords: design, measuring wheelchair, manufacturing system, carbon fibre wheelchair.
1. Universität Kassel Mönchebergstr. 3, 34125 Kassel, Germany - E-mail: [email protected]
162 The Engineering of Sport 7 - Vol. 1
1 - Mechanical loads The aim of this project was to develop an extreme lightweight and perfect fitted wheelchair. For the optimum dimensioning it is necessary to know as much as possible about the mechanical loads. To learn more about the load types, several wheelchair users were accompanied during their daily routine. As a first result it can be said, that the maximum (impact) load was caused by the jump off the kerb [VCRB1] (Figure 1). The height of a standard kerb according to DIN EN 1340 and DIN EN 483 is about 120 mm. To simulate the jump off the kerb a vertically adjustable stage (100-440 mm) was built. To collect the data (vertical force) a Kistler load cell was used (Figure 2). Two wheelchair users were recruited, a male and a female. Tyres, pressure of the tyres and height of the stage were changed. Before the dynamic test the static loads and the load distribution were measured (Table 1). The biggest part of the static loads was found on the rear wheels (77-93%). To determine the dynamic loads in connection with the stage height, the subjects were instructed to do five jumps from 100 mm, 200 mm and 300 mm stage height. The male subject, the winner of three gold medals at the paralypic winter games in Torino 2006, the professional monoski rider Martin Braxenthaler was able to jump off 440 mm, so that it was possible to get extreme data.
Figure 1 - Jump off the kerb.
Figure 2 - Kistler load cell.
The Engineering of Sport 7 - Vol. 1
163
Table 1 - Wheelchair characteristics and static loads.
The vertical force is about 2300-2500 N caused by the jump off the kerb. With a 200 mm stage height the vertical force is higher than 3800 N. The maximum force was collected at 440 mm stage height (5000 N). Figure 3 shows the results of the measurements.
Figure 3 - Vertical force in connection to stage height.
Further dynamic loads appear as a result of collisions between barriers and the front wheels (Figure 4). Depending on the moving direction the collisions between barriers and the front wheels involves planar stresses or torsion in the structure. Simulations of the collisions between barriers and the front wheels were realized by the fixation of different rails (20 mm and 30 mm). To determine the dynamic loads in connection with the rail height, the subjects were instructed to do five high speed collisions (straight and rotating) with 20 mm and 30 mm rail height. The maximum horizontal force (straight collision) is about 2500 N.
164 The Engineering of Sport 7 - Vol. 1
Figure 4 - Collision test.
2- Measurement and adaptation process Normally the fitting process and the selection of a wheelchair result from the measurement of body data [HLMSZW1]. The advisor measures the femoral and lower leg length and some more body data and selects the wheelchair model and size. Rarely it is possible for the customer/patient to try the selected wheelchair to check the seating position and the driving behaviour. In many cases it is impossible to change the seating position without changing the driving behaviour. The perfect wheelchair offers the optimum seating position, an adequate driving behaviour and minimum weight. To achieve the aim of the best possible seating position and driving behaviour a multi adjustable measurement/fitting wheelchair has been developed. The measuring/fitting chair was realized as a lightweight construction. To be as close as possible to the geometry and handling behaviour of the customised wheelchair, the same moulded cfrp-parts has been used. With the new measuring/fitting wheelchair (Figure 5) it is possible, to adapt and optimize the seating position separately from the driving behaviour. An enormous advantage is that the customer can feel the seating position and check it with his therapist concerning to medical requirements. The customer can also verify the driving behaviour, especially the break over characteristics.
The Engineering of Sport 7 - Vol. 1
165
Figure 5 - Measuring/fitting wheelchair.
3- Data collection and design The next step after the fitting process is the collection of the adjustment parameters. For a definite acquisition of the seating position and the chassis, several parameters like seat width, seat height, incline of the seat, height of the backrest, incline of the backrest, position of the axle (horizontal), distance between front and rear wheels, wheel size and the camber of the rear wheels are necessary. The adjustments represent the input for special parametrised three-dimensional computer model. With this computer model it is possible to get a true to scale design of the customised wheelchair. The designer is able to do an interference check (free rotatability of the front wheels without touching the foot plate or tubes). Furthermore the three-dimensional is the basis for computational analysis, such as finite element analysis.
4- Manufacturing As just noted, the determination of the optimum seating position and handling behaviour on the basis of measured body data is imprecise. The manufacturing of custom made wheelchairs is state of the art, but imprecise too. The tolerance of commercially available custom made wheelchairs is about ±1-1.5 cm. To reduce the tolerance and to facilitate an exact transfer of the measured data, a flexible manufacturing device has been developed (Figure 6). The manufacturing device consists of aluminium T-slot profiles and special fixation elements for the tubular wheelchair structure. The link-up of the three-dimensional computer model and the manufacturing device allows the exact adjustment of the fixation elements and with it the manufacturing of custom made wheelchairs with a tolerance about ±1 mm.
166 The Engineering of Sport 7 - Vol. 1
Figure 6 - Manufacturing device with three-dimensional model of the wheelchair (true to scale).
5- Ultralight wheelchair The explanations about measuring/fitting, design and the manufacturing system are independent of the used material of the custom made wheelchair. The aim of this project was to create a perfect fitted and extreme light wheelchair. We decided not to give too detailed information about dimensioning, Finite Elemente Analysys and other very technical data. The following informations should give a review about materials, manufacturing processes and joining technology. Lightweight materials like carbon fibre reinforced plastic (cfrp), aluminium and titan were used. For the frame (tubular structure) cfrp was used. To manufacture the different components, a wide range of processes were used. The formed components were manufactured via blow-moulded process, tubular components were manufactured via filament winding or wrapping processes. Planar structures like the fenders were manufactured vacuum assisted. The joining of the components was very complex. For the wheelchair frame the tube-to-tube-process and special adhesive joining techniques have been used.
The Engineering of Sport 7 - Vol. 1
167
6 - Conclusions The system, consisting of measuring/fitting chair, three-dimensional design tool and manufacturing device enables for the first time the determination of the optimum seating position, adequate handling, computer aided design and exact manufacturing. A special feature is that the system is independent form material. Exclusive application of steel, aluminium or carbon fibre reinforced plastic is as possible as multi material design to manufacture custom made wheelchairs. As a result of the application of high tech materials and joining methods it was possible to realize a unique lightweight 5.9 kg carbon fibre wheelchair for the daily use (Figure 7). With such a lightweight wheelchair it is possible to increase the mobility and reduce the physical strain of handicapped people.
Figure 7 - Carbon fibre wheelchair.
7- References [HLMSZW1] Harfich, K.-H., Lex, W., Mertsch, S., Sharma, J., Zurheide, M., Weege, R.-D. Leitfaden zur Rollstuhlversorgung, Manuelle Greifreifenrollstühle, 2. Auflage, 2001 [S1] Sonderhüsken, H.: Rollstuhltauglichkeit von Pkw’s. In the journal „Der Paraplegiker“, Heft 3/2001 [VCRB1] VanSickle, D.P., Cooper, R.A., Robertson, R.N., Boninger, M.L. Determination of wheelchair dynamic load data for use with finite element analysis. In journal of Rehabilitation Engineering, Volume: 4: 161-170, 1996
Design and Implementation of a Rugbyspecific Garment Evaluation Trial (P30) Bryan C. Roberts1, Gareth Williams1, Mike P. Caine1
Topics: Apparel and rugby. Abstract: A structured rugby-specific wearer trial to extract meaningful on-field garment performance data does not exist. To address this, a novel rugby-specific garment evaluation protocol was developed and trialled using a rugby-union international prototype shirt. Three wearer trial elements were investigated namely, rugby-specific wear-service conditions, players’ perceptions of the shirt and, the determination of garment performance on the field of play. Methods: A field test mimicking the demands of the game was devised using published international match-play time-motion analysis. Fifteen non-professional club players (age 26.1 yrs ± 5.2 SD, height 1.83 m ± 0.05 SD, pre-trial body mass 95.8 kg ± 10.7 SD) participated and completed the aforementioned field test. Heart rate was recorded throughout and protocol intensity determined. Garment performance was assessed using controlled visual assessment techniques and dimensional stability measurements. Structured questionnaires were administered during and post-trial to determine player perceptions. Results: Forwards’ mean and peak heart rates (153.3 bpm ± 24.1SD and 181.9 bpm ± 10.0 SD) were lower than match-play target values. Backs’ mean and peak heart rates (147.2 bpm ± 26.8 SD and 185.0 bpm ± 10.7 SD) were similar to target values. Dimensional changes and defects were identified and quantified successfully. Players conformed well to the protocol and responded favourably to the questionnaire. Conclusion: The protocol described represents the first rugby-specific garment evaluation protocol to be documented. It is hoped that this will be adopted and refined thereby adding structure to the sports-specific garment development process. Keywords: time-motion analysis, garment durability, wearer trial, player perception, heart rate.
1. Sports Technology Research Group, Loughborough University Sports Technology Institute, 1 Oakfield Avenue, Loughborough University, Loughborough, Leicestershire, UK, LE11 3QF - E-mail: [email protected]
170 The Engineering of Sport 7 - Vol. 1
1- Introduction Innovation in the rugby union (rugby) team-wear industry is cyclic in nature; predominately driven by major tournaments. Competition at the highest level is seen as a visible platform to promote brand identity through innovative product design. Prior to 1999 Rugby World Cup, rugby team-wear design was predominately fashion led. Focus has since shifted to the improvement of athletic performance via enhanced garment functionality therefore prototype evaluation is critical to assess functionality whilst also ensuring garments can withstand the rigours of the game. Rugby-specific wearer trials are often used to assess the performance of, and wearers’ physiological response to, apparel during training sessions and simulated match-play. However, personal experience and interviews with manufacturers suggest that the current wearer trial process is ad-hoc and has limitations. Test time with elite players is often limited, end-use conditions may not replicate the intensity of match-play and, feedback may be biased. Post-trial interviews are typically informal and unstructured to create a rapport with the athlete rather than maximise opportunity for comprehensive user insight capture. Finally, when tested over a number of trials, the process is timeconsuming and expensive. Wearer trials enable the validation of expected wear performance of a textile in its intended end-use conditions (ASTM D3181 1995) and are defined as any controlled tests carried out on an item of textile apparel in which the item is worn (BSI 7754 1994). Two international standards summarise generalised wearer trial procedures (BS 7754 1994 and ASTM D3181 1995). To date, there are no published sports-wear performance evaluation protocols. The development of a rugby-specific wearer trial was dependent on the development of a rugby-specific exercise protocol to elicit an applicable physiological response to exercise, a thorough knowledge of player perception techniques and structured garment assessment protocols to predict on-field performance. This was to ensure that the shirts and players were subjected to demands similar to that of the game. It is evident that a structured, controlled and repeatable rugby-specific wearer trial is necessary to extract meaningful garment performance and player perception data over a short period of time. The current paper details a pilot study detailing the design and implementation of a rugby-specific wearer trial, however the principles outlined may be applied to other sports.
2- Methods Fifteen international match-play prototype rugby shirts, worn by fifteen non-professional competitive club players (age 26.1 yrs ± 5.2 SD, height 1.83 m ± 0.05 SD, pre-trial body mass 95.8 kg ± 10.7 SD) from each rugby union position, were evaluated using a rugby-specific wearer trial protocol. Eight additional players acted as opposition to the tasks where necessary and all tasks were performed on an outdoor turf rugby pitch. All players provided written informed consent in accordance with generic clearance from the University’s ethical advisory committee. Ambient temperature and humidity were 15.8°C ± 2.1 SD and 55% ± 10 SD respectively.
The Engineering of Sport 7 - Vol. 1
171
Figure 1 - Wear-service conditions.
The prototype shirts were constructed from 78% polyester 22% cotton COOLMAX® weave blend (main body panels), 65% nylon, 21% polyester, 14% LYCRA® blend (arms and side panels) and, 100% cotton twill (collar). Grip panels were positioned on the chest, shoulders and upper back. Time-motion analysis data, used to derive the exercise portion of the rugby-specific wearer trial (Figure 1), were extracted from televised game analysis of the Rugby World Cup 2003 (IRB 2003). Wear-service conditions focussed primarily on non-running intensive exertion (rucking/mauling, lineouts and scrummaging) and game-specific activities (kicking) rather than movement patterns (standing still, walking, jogging, cruising, sprinting and utility) as equipment to classify individual movement patterns throughout the trial were not available. The exercise section of the trial was completed in 60 minutes.
172 The Engineering of Sport 7 - Vol. 1 Protocol intensity was estimated through determination of sweat loss and heart rate which were compared to target values. Nude body mass was measured pre- and posttrial using digital weighing scales (2204-Duo Model, Tanita UK Ltd; accuracy ± 0.2 kg) to derive an approximation of sweat rate however, players were allowed to consume water freely as in match-play. Heart rate (HR) was monitored every 5 seconds using wireless heart rate monitors (Vantage NV, Polar Electro, Finland) to determine mean and peak heart rates. Robergs & Landwehr (2002) conducted a review of age-predicted HRmax and concluded that, to date, the most suitable for the general population was proposed by Inbar et al., (1994) which was used in this study (Equation 1). Mean percentages of time in each HR zone (<75%, 75-85%, 85-95% and >95% HRmax) were also determined. Scrum-half data were not included in the backs category due to the distinctive nature of the position and for comparison with previous time-motion analysis studies. The movement patterns of a scrum-half are distinctly different from the forward and backs category (Duthie et al., 2005). HRmax = 205.8 - (0.685*age) (1)
Figure 2 - Measurements taken to assess dimensional stability of rugby shirts and garment defect location.
Garments were measured pre-trial, post-trial and post-laundering to assess dimensional stability (accuracy ± 0.5 mm) using a steel rule. Measurement points, as specified by the manufacturer, were 1) half chest width (measured 40 mm vertically downward from the arm-hole), 2) half hem, 3) sleeve length, 4) half sleeve opening, 5) underarm length, 6) neck width, 7) front neck drop, and 8) centre back length, as shown in Figure 2. Visual garment assessment techniques were devised from previous rugby shirt durability testing, literature reviews and interviews with the manufacturer. Scales and failure criteria were assigned to each defect category. The reader is directed to Saville (1999) for further classification and examples of dense, distinct, moderate and slight defects. Defects were described by their defect category, size, severity, shirt location, specific location (anatomical or shirt feature) and failure criteria (Table 1). Failure in a particular defect category is fulfilled when classification matches the number denoted by ‘failure = x’ in Table 1.
The Engineering of Sport 7 - Vol. 1
173
Questionnaires, devised from interviews with players, manufacturer consultation and previous wearer trial information, were distributed during the position-specific drills contained predominately quantitative questions. The aim was to gauge opinion, in a short period of time, regarding the effects of the shirt on performance during a particular activity. Post-trial questions were concerned with the shirt performance throughout the complete trial. Four discrete areas of interest were identified to aid analysis. Firstly the player’s history was gauged to understand personal experience, and perhaps, aptitude for the test. Secondly, general garment functionality was assessed by determining player preference for fit, style, strength, thermal comfort, thermal sensation and, weight. Then rugby-game specificity was assessed such as the need for grip and its positioning, difficulty to tackle, ease of binding, affects during scrummaging and, range-ofmovement whilst handling the ball. Finally, the players were given an opportunity to give additional comments on any aspect of the shirt design.
Table 1 - Visual assessment defect classification post-trial.
3- Results 3.1 Physiological profiles Mean body mass losses were 1.23 kg ± 0.70 SD, 1.70 kg ± 0.37 SD, 0.69 kg ± 0.59 SD for the whole squad, forwards and backs respectively with a mean fluid deficit of 1.3 % initial body mass. Mean heart rate, mean peak heart rates and, percentage of time spent in each HR zone are shown in Table 2.
174 The Engineering of Sport 7 - Vol. 1
Table 2 - Estimated maximum heart rate and mean (SD) participant heart rate data.
3.2 Garment performance Few major defects were observed post-trial; there were three instances of slight seam puckering, two instances of slight aesthetic deterioration, one instance of moderate aesthetic deterioration and two instances of slight seam slippage across all shirts. However, all shirts exhibited staining post-laundering, particularly soil retention. The garments initially increased in size post-trial; the main body material (78% polyester 22% cotton COOLMAX® weave blend) exhibited the most dimensional instability where half hem differences varied 20 mm ± 2.2 SD post-trial. Differences were outside the limits as specified by manufacturers (± 1 mm) therefore the garments failed this aspect of the test.
3.3 Players’ perception of the rugby shirt design Of the 15 respondents, eight felt there were no fit-related issues. Of the other seven, four indicated a preference for a looser fit around the abdomen. The majority of players thought the shirt was stylish and both the fabric and seam construction appeared strong. Ten players felt warm during the trial but only one player felt uncomfortably warm. Three players described particular areas of discomfort; the side, chest and armhole of the shirt respectively. Players stated that the shirt did not restrict passing, tackling, running, scrummaging, or mauling. Six players believed the shirt was too light whereas the rest felt it was “just right” in both wet (after sweating) and dry (initially donned) conditions respectively. Nine players considered the shirt difficult to tackle but none thought it was easy. Finally, the majority of players found the grip, a hexagonal raised silicon transfer measuring 0.5 mm in height on the front and back of the shirt, beneficial when catching and ball carrying (11 players) but felt no benefit during tackling (9 players), binding (9 players), driving (10 players) or lifting (12 players).
4- Discussion A rugby-specific garment evaluation trial was devised to replicate the demands of rugby union using time-motion analysis of game-specific activities and non-running intensive exertion. The aim of the study was to apply the knowledge gained from a number of wearer trials conducted by rugby manufacturers to create a time-saving, structured and controlled rugby-specific wearer trial. This was achieved in this pilot study by assessing the functionality of fifteen international prototype shirts using structured garment evaluation and player perception techniques. The first aim of the investigation was to determine protocol (or wear-service condition) physical intensity. Although overall intensity was lower than a real-life match situation, the exercises were deemed structured and easy to replicate in further trials;
The Engineering of Sport 7 - Vol. 1
175
particularly since the trial was administered by team coaches through a single verbal and written communication prior to exercise. Lower than target heart rates may also be attributed to a lack of competitiveness during the trial, lack of opposition and, periods of rest to respond to questionnaires. In this study, mean heart rates recorded during the study were 153.3 ± 24.1 SD and 149.0 ± 26.7 SD compared to 175 ± 5 SD and 150 ± 5 for forwards and backs respectively (Doutreloux et al. 2002). Players also exhibited a mean fluid deficit of 1.3 % initial body mass whilst still consuming fluid freely throughout the 60 minute trial. Burke (1997) suggests the mean total sweat losses, for the whole squad, during a game of rugby union (~80 minutes) lie between 1.7 to 2.3 L per match corresponding to a mean deficit of 1.5 % initial body mass. Future experiments should incorporate maximal exercise tests to accurately calculate HRmax, thus enabling the percentage of time spent in each heart rate zone to be derived more accurately. It is hoped that the exercise protocol be further refined to elicit a more accurate forwards physiological response and that repeatability be assessed through testing a number of rugby clubs. Two methods of physical garment assessment were employed namely, dimensional stability and visual defect assessment. The pass-fail criteria for both were determined from a series of match-play trials held prior to this study. Defect category and classification were also determined from interviews with manufacturers and analysis of prototype garment standards. In this study, few defects were observed post-trial, however, all shirts exhibited a moderate degree of staining, particularly soil retention. Seam slippage, aesthetic deterioration, and seam puckering were minimal. Only one shirt failed completely due to logo deterioration on the back of the collar. The main body material exhibited signs of garment instability however, according to players’ opinion, performance was not affected. In future, garment features or defects should be given a weighting of importance and with further research, garment defects should be compared to visual grades to more accurately classify each defect. Player perceptions were gauged during and post-trial through position-specific qualitative and quantitative questionnaires. Questions were determined from previous international player interviews, match-play analysis and interviews with rugby-shirt designers. Personal or group interviews were not appropriate for this process due to the limited amount of time on the field of play therefore short quantitative questionnaires were optimal. In this study, players’ opinions were favourable, mostly approving of the shirt design features and style. Player perceptions, in this test, were used to increase understanding and aid product development of rugby shirt design.
5- Conclusions Current garment evaluation trials are often unstructured, time-consuming and potentially biased. A review of the rugby-specific wearer trial development process has been detailed in the current paper. A rugby union specific field test was developed to replicate the non-running intensive exertion and game-specific activities of the game. Wearservice conditions and players’ physiological response to exercise were monitored to assess game intensity, which is invaluable for the reliable assessment of sports garment
176 The Engineering of Sport 7 - Vol. 1 performance. Structured garment assessment techniques were successfully implemented post-trial. All trial prototype shirts passed the wearer trial with few negative player perceptions or functional defects, however, the test procedure was able to identify and distinguish even very minor defects. It is hoped that testers will apply the principles outlined to develop additional sports-specific wearer and equipment trials.
6- References [ASTM1] ASTM D3181-95 Standard guide for conducting wear tests on textiles, ASTM. [BS1] BS 7754:1994 Code of practice for garment evaluation by wearer trials. BSI. [B1] Burke, L.M. Fluid balance during team sports. Journal of Sports Sciences, 15, 287-295, 1997. [DW1] Docherty, D., Wenger, H. A. & Neary, P. Time motion analysis related to the physiological demands of rugby. Journal of Human Movement Studies, 14, 269-277, 1988. [DT1] Doutreloux, J.P., Tepe, P., Demont, M., Passelergue, P. & Artigot, A. Exigences énergétiques estimées selon les postes de jeu en rugby. Science and Sports, 17(4), 189-197(9), 2002. [DP1] Duthie, G., Pyne, D. & Hooper, S. Time motion analysis of 2001 and 2002 super 12 rugby. Journal of Sports Sciences, 23(5), 523 – 530, 2005. [IO1] Inbar, O. Oten, A., Scheinowitz, M., Rotstein, A., Dlin, R. and Casaburi, R. Normal cardiopulmonary responses during incremental exercise in 20-70-yr-old men. Medicine and Science of Sport and Exercise, 26(5), 538-546, 1994. [IRB1] IRB. RWC 2003: Statistical review and match analysis. Dublin: International Rugby Board (IRB), 2003. [M1] McLean, D.A. Field testing in rugby union football. In: Macleod et al., [Eds.] Intermittent high intensity exercise: preparation, stresses, and damage limitation. London: E & F N Spon, 1993. [RL1] Robergs, R.A. & Landwehr, R. The surprising history of the “HRmax=220-age” equation. Journal of Exercise Physiology, 5(2), 1-10, 2002. [S1] Saville, B.P. Physical testing of textiles. Woodhead Publishing Ltd, Cambridge, 1999.
Open Rotator Cuff Surgery in Swiss Elite Rock Climbers (P31) Hans-Peter Bircher1, Christoph Thür2, Andreas Schweizer3
Topics: Climbing; Extreme sport. Abstract: Rock and indoor climbing has become a very popular sport. For moving in a vertical or even overhanging wall an intact shoulder function is mandatory beside a good grasp function. Hand and particular finger injuries in climbing are well described in the literature. Almost no report is avaible about shoulder injuries in climbers. In our retrospective case study 21 shoulders in 20 climbers (18 male, 2 female, age 28 to 65) with open rotator cuff surgery between 1998 and 2006 were analysed. The ability of climbing before and after the operation is a main outcome control. The level of difficulty in climbing is given by the french difficulty scale ranging from three to nine whereas the top grade nine is performed by only few climbers worldwide. In 10 climbers a single shouldertrauma while in 7 a repetetive traumas and in 3 chronic overuse prevented a situation compatible with rock climbing. There are 7 complete rotator cuff lesions and 14 partial tear of the supraspinatus tendon. In all climbers an open rotator cuff surgery was done including 17 enlargement of the subacromial space and 15 tenodesis of the long head of biceps. 6 months after the operation in average the climbers started again. Average time to return in a similar level of difficulty was 16 months in all but one climber. From the point of biomechanics, pathologies of climbing and swimming shoulder are discussed. Keywords: sportclimbing, injury, rotator cuff, open surgery, returning in sportclimbing.
1- Introduction Rock and indoor climbing has become a very popular sport. Sportclimbing is moving with the use of both arms and legs in steep or even overhanging walls or rock. This activity goes back on a rebelish group of young mountaineer in the early seventies in the USA revolting the old fashioned climbing style using technical aid to overcome difficulties. They created the motivation to climb „by fair means“ which excluded technical aid. The climbing gear is used for potection only. The difficulties in climbing increased since, 1. Orthopaedic Clinic, Hospital of Zug, 6300 Zug, Switzerland - E-mail: [email protected] 2. Shouldersurgery and Traumatology, 8001 Zürich, Switzerland - E-mail: [email protected] 3. Universityclinic Balgrist, 8008 Zürich, Switzerland - E-mail: [email protected]
178 The Engineering of Sport 7 - Vol. 1 which is equivalent to a decrease of grips and steps. The stress on the upper extremity in this sport is high. To climb difficult routes, a power and endurance training particular for the upper extremities is mandatory. Several tries in a projected route are needed before climbing it sucessfull. Finger injuries, particular strain or rupture of flexor tendon pulley, are frequent and well described [B1][B2][MG1][LO1] [RM1][S1][SJ1]. It is been estimated that up to 20% of elite climbers suffer shoulder problems [JA1][R1]. Very few information was found about general shoulder problems among climbers in literature. No publication was found about a specific shoulder problem in climbing. The presented paper analyses rotator cuff pathologies in elite climbers and recreational climbers.
2- Material and Methods In a retrospective case study 21 operated shoulders with rotator cuff pathologies in 20 climbers were controlled in two orthopaedic clinics (Clinic of Shouldersurgery and Traumatology, Zürich and Orthopaedic Department, Hospital of Chur, Switzerland). 13 were professional (10 mountain guides, 2 competitive climbers, 1 gym instructor). 7 were recreational climbers on a high level. The avarage age was 44 years (28- 65). In this collective the comparison in climbing prae- and postoperativly is used as a quality control for the operative treated rotator cuff problem. The difficulties in climbing were rated in the world wide accepted french scale, ranging from three to nine. Three is a simple climbing, nnie is an extreme difficult climbing. Each routerating is set by the climber who climbed the wall or rock the first time. Grade 9 is reached by a few climbers world wide only. Each degree is subdivided in a, b and c, rising in difficulty. 10 climbers had a causual shoulder trauma with a tear/distorsion. 7 climbers had repeated shouldertraumas including contusion. 3 climbers showed an overuse of the shoulder before the onset of the dysfunction. Conservative treatment was performed in 16 patients (7 physiotherapy, 2 infiltration with steroids, 2 alternative treatment and 5 with regularly analgetic medication). None of the collective was able to climb even near his top level because of shoulder dysfunction. An arthro-MRI was obtained of all operated shoulders. The intervall between onset of shoulder pain and /or trauma and operation was 16 months (3-24). All patients were operated in general anaesthesia with a supplementary scalenus block. 7 complete and 14 partial tears were diagnosed (Fig. 2). The complete tears were in 6 cases limited on the supraspinatus tendon (1x2 cm to 3x4 cm) In one case the tendon rupture was 5x4 cm including the upper part of the infraspinatus- and subscapularis tendon. The 14 partial tears were subdivided in 2 PASTA, 4 bursal -, 4 articular side and 4 distinct inflammation in the rotator interval and supraspinatustendon. (Table 1)
The Engineering of Sport 7 - Vol. 1
179
Table 1 - Rotator cuff, biceps and SLAP pathologies.
tendon
complete 7
partial 14
supraspinatus
1x2, 2x3,2x3, 2x3, 1x2,3x4
PASTA articular bursal inflammation
2 4 4 4
rupture part inflammation Typ I Typ II
2 3 1 2 2
supraspinatus, infraspinatus subscapularis long head of biceps
5x4
rupture part. tear inflammation
3 4 2
SLAP
In 10 patients a diagnostic arthroscopy was performed. All operations on the rotator cuff were done in an open way. Reinsertion was done with braided non- absorbable sutures (Ethibond No. 2). Intertendinous sutures were fixed with absorbable monofil material (PDS No. 1). In 9 cases a transosseus supraspinatus reinsertion with a doubling of the tendon was done. In 5 cases a supraspinatus reinsertion and in 2 cases a debridement were performed. An acromioplasty was made in 17 shoulders. (13 acromion lift up osteotomies [TJ1] and 4 modified plasties according to Neer). 15 tenodesis of the long head of biceps were done (11 key houle and 4 suture fixation). Additionally two SLAP II lesions were refixated arthroscopicly with absorbable suture ancors (Panalok loaded with PDS 1). Two SLAP I lesions were debrided arthroscopicly. (Table 2) Table 2 - operativ procedure in rotator cuff.
tendon
open operation
arthroscopic operation
supraspinatus
reinsertion and doubling of supraspinatus reinsertion deltoid flap
9 5 1
diagnostic tenotomy biceps debridement
10 8 2
tenodesis - key hole - suture
tenotomy biceps
8
11 4 refixation debridement
2 2
supraspinatus infraspinatus subscapularis long head of biceps
SLAP
Postoperativly 16 patients were treated with an shoulder abduction splint for 6 weeks. In 5 cases the shoulder was immobilized for 6 weeks in a shoulderwest. Physical therapy started the first day after operation obtaining range of motion passivly: abduction/adduction 90-40-0 and internal/externalrotation 20-0-20. After the forth postoperative week
180 The Engineering of Sport 7 - Vol. 1 isometric abduction in 90° with flexed elbow was allowed. A clinical and radiological controll was done 6 week after operation. Active physiotherapy is started to build up a full range of motion. The patients were seen 12, 26 and 52 weeks ambulatory after operation. A telephonic questioning was done in average 49 months postoperativly (12-108). A systematic questionnaire about pain, range of motion, restart of climbing and top level was used.
3- Results In a healty shoulder situation praeoperativly the top level of 5 climbers was grade 8 (1x8c, 3x8b, 1x8a). 4 climber were familiar with grade 7 (1x7c, 1x7b, 2x7a). 9 climbers made grade 6 (7x6c, 2x6b) and 2 climbers grade 5 (2x5c). (Table 3) We saw no complication in the operated collective. The ability of working was given after 4 months postop (1-6). Restart in climbing was in average 8 months after operation (6-12). The top level after the operation was reached 16 months after the operation by the collective (8-24). The top level after the operation of grade 8 climbers was similar to the praeoperativ level in 3. Two grade 8 climber reached 8a instead of 8b praeoperativly. In grade 7 two climbers did the same top level as before the operation. One climber stopped his carrier out of other reason than the operated shoulder. One climber is performing grade 6c instead of 7a praeoperativly. In grade 6 four climbers had small loss, but remained on the 6. degree. In grade 5 there was no difference. (Table 3) Table 3 - Climber and his top level prae- and postoperativly (french scale).
climber A.R. G.H. M.A. A.L. A.L. M.W. R.B. S.A. L.S. N.E. D.M. T.U. L.C. D.Z. C.A. A.R. D.M. A.R. Z.B. F.J. H.D.
prae-op
post-op
8c 8b 8b 8b 8b 8a 7c 7b 7a 7a 6c 6c 6c 6c 6c 6c 6c 6b 6b 5c 5c
8c 8b 8a 8a 8a 8a 7c 7a 6c 6c 6c 6c 6c 6b 6b 6a 6b 6a 5c 5c
The Engineering of Sport 7 - Vol. 1
181
4- Discussion The epidemiology of shoulder injuries in climbing is given by 15-20% of elite and recreatinal cimber in the literature. Finger is the most commen injuiered part of the upper exremity in climbing. Finger injuries are supposed to be three to four times more frequent than shoulder. But shoulder dysfunction is a real handicap. The aethiology of shoulder dysfunction is either cuff tear by overload. Further repetitive overstrain may result in cuff tear and /or subacromial bursitis with a following outlet impingement. In our collective almost every climber mentioned several contusion or distorsion of shoulder while climbing in the past. The most commen reason of these accidents is a fall into the securing rope. This occurs usually during exploring a difficult climbing part. In our experience rotator cuff pathologies and subacromial impingement problems represent the main part of shoulder problems. Beside these we observed in climbers instabilities, omarthritis, AC- dislocation, AC- arthritis and distal tendon rupture of biceps. The cuff and biceps pathologies in the presented collective show a large spectrum. There seems not to be one typical pathology. It is difficult to compare the shoulder in climbers with other sport. Due to a distribution on both arms for moving and a similar load subdivision between arms and legs [PT1] the shoulder of swimming may be comparable with the climbing shoulder. Competitive swimmer perform nearly 500’000 stroke revolutions per arm and season [RJ1]. The swimmer’s shoulder has an element of general laxity due to repetitiv arm motion. It is thought that a loss of static components in stabilisation require a greater contribution from the rotator cuff [RB1][ZM1]. A certain laxity in climbing shoulder is found as well. This kind of laxity in the climber shoulder may be explained by the repetitive grasping toward the top. But the number of tractions and the dynamic load in the pull through or climb through part of movement is difficult to compare. On the side of the impingement, Swimmer’s shoulder show moments of subacromial but also moments of nonoutlet type of impingement [YH1][YH2]. Due to the pathology of cuff lesions in the climbing shoulder a certain nonoutlet form of impingement must be discussed. Open cuff surgery shows an excellent outcome in our collective. Postoperativly 19 of twenty climbers took up climbing in a similar level including the elite. The postoperativ rehabilitation is a key point in teh treatment as well. The climbers showed a progressiv strength in their operated shoulder up to 2 years. The top level in climbing reach the climbers in average 8 months after the restart.
5- References [B1] Bollen S., Soft tissue injury in extreme rock climbers. Br J Sports Med, 22(4): 145-7, 1988 [B2] Bollen S., Upper limb injuries in elite rock climbers. J R Coll Surg Edinb. 35(6 Suppl): S1820, 1990 [JA1] Jones G, Asghar A, Llewellyn DJ. The epidemiology of rock climbing injuries. Br J Sports Med, 2007
182 The Engineering of Sport 7 - Vol. 1 [LO1] Largiader U., Oelz O., [Analysis of overstrain injuries in rock climbing]. Schweiz Z Sportmed, 41(3): 107-14, 1993 [MG1] Moutet F., Guinard D., Gerard P., Mugnier C., [Subcutaneous rupture of long finger flexor pulleys in rock climbers. 12 case reports]. Ann Chir Main Memb Super, 12(3): 182-8, 1993 [PT1] Pink M., Tibone J.,The painful shoulder in the swimming athlete. Ortop Clin Noth Am, 31(2): 247-261, 2000. [R1] Rooks MD. Rock climbing injuries. Sports Med, 23(4): 261-70, 1997 [RB1] Rupp S., Berninger K.,Hopf T., Shoulder problems in high level swimmers – impingement, anterior instability, muscular imbalance? In Int J Sports Med, 16(8):557-562, 1995. [RJ1] Richardson A., Jobe F., Collins H., The Shoulder in competitive swimming, Am J Sports Med, 8(3):159-163, 1980. [RM1] Rohrbough J., Mudge M., Schilling R. Overuse injuries in the elite rock climber. Med Sci Sports Exerc, 32(8): 1369-72, 2000 [S1] Schweizer A., Lumbrical tears in rock climbers. J Hand Surg [Br], 28(2): 187-9, 2003 [SJ1] Schoffl VR, Jungert J. Closed flexor pulley injuries in nonclimbing activities. J Hand Surg [Am], 31(5): 806-10, 2006 [TJ1] Thür C., Jülke M., Bircher H., [Lifting osteotomy of the acromion as a new principle in treatment of impingement syndrome, especially in correlation with reconstruction of large rotator cuff lesions] Unfallchir 101(3):176-83, 1998 [YH1] Yanai T., Hay J., Miller G., Shoulder impingement in front crawl swimming: I. A method to indentify impingement. Med Sci Sports Exerc, 32(1):21-29, 2000. [YH2] Yanai T., Hay J., Shoulder impingement in front crawl swimming: II. Analysis of striking technique. Med Sci Sports Exerc, 32(1):30-40, 2000. [ZM1] Zemek M., Magee D., Comparison of glenohumeral joint laxity in elite and recreational swimmers. In Clin J Sport Med, 6(1): 40-47, 1996 Jones G, Asghar A, Llewellyn DJ. The epidemiology of rock climbing injuries. Br J Sports Med, 2007 Largiader U, Oelz O. [Analysis of overstrain injuries in rock climbing]. Schweiz Z Sportmed, 41(3): 107-14, 1993 Rohrbough JT, Mudge MK, Schilling RC. Overuse injuries in the elite rock climber. Med Sci Sports Exerc, 32(8): 1369-72, 2000
A Quantitative Analysis Of Beach Casting (P33) Benjamin Charles1, Darryl P Almond1, Aki I T Salo2 Presented by Alan N Bramley1
Topics: Sea fishing, video analysis. Abstract: Video analysis techniques have been used to analyse quantitatively the overhead casting method employed by sea anglers fishing from beaches. The techniques have been used to estimate launch velocities and launch angles achieved by this casting method. Cast distances achieved have been compared with results of projectile calculations, using observed launch angles and velocities. The measured cast distances have been found to be between two and three times shorter than predicted by basic projectile calculations. Further experiments indicate the drag of the running line to be the main cause of the reduction in cast distance. Keywords: video analysis, projectile dynamics, sea fishing, casting.
1- Introduction Beach casting is a very popular form of fishing which involves using a long rod to cast hooked bait or lures from a beach, pier, breakwater or off rocks into the sea. A lead weight is attached to the end of the line to ‘pull’ the hooked bait or lures out from the shoreline and across the water during a cast. The main objective of the cast is to project the hooked bait or lures as far as possible from the shoreline into the sea. Despite the enormous number of people participating in this and other forms of angling, there has been no quantitative scientific analysis of the casting method. There have been some publications on fly casting (Robson 1990, Spolek, 1986) and various forms of beach casting and the equipment have been described (Holden, 1982). We report here the results of a detailed study of beach casting in which the casting process has been subjected to video analysis and casting distances achieved are compared with the predictions of conventional projectile calculations. The method of casting studied here is the “overhead” or “ground” cast which is one of the most commonly used methods. At the start the lead weight at the end of the line 1. Department of Mechanical Engineering, University of Bath, Claverton Down, Bath, BA2 7AY, UK. E-mail: [email protected]; [email protected] 2. Sports and Exercise Science, University of Bath, Claverton Down, Bath, BA2 7AY, UK - E-mail: [email protected]
184 The Engineering of Sport 7 - Vol. 1 is suspended from the rod tip by between one and two metres of line (the “leader length”) and laid on the beach behind the angler. The angler then rotates the rod rapidly in an arc set in a vertical plane. The weight is towed around, following the rod tip, until the line is released when the angle between the rod and the sea is approximately 45. The mechanism is very similar to that employed by the medieval siege engine, the trebuchet, to project missiles over great distances. The momentum of the weight at the point of release determines the distance it travels across the sea. This distance is also affected by the aerodynamic drag of the weight and the bait or lures and by the drag of the line as it is drawn off the fishing reel and through the line guide rings along the length of the rod. There is a number of other casting methods, notably the pendulum cast, that involve rotating the weight in a near horizontal plane to build up momentum prior to release. These methods have not been investigated here because of the complexity that arises in video analysis of a 3D motion of this type. The advantage of the overhead or ground cast is that the motion occurs, to a good approximation, in the vertical plane alone, allowing it to be recorded by a single video camera.
2- Equipment The rod used was a 12 foot Shakespeare Targa Beachcaster. 12 feet (3.66 metres) is the most commonly used length for beach cast fishing. The rod is made of carbon fibre which is currently the standard material for fishing rod manufacture. Since the rod is manufactured from carbon fibre, it has a dark almost black appearance. White tape was therefore applied to the rod at intervals along its length to ensure it stood out in the video recordings. The reel used was a Targa Beachcaster Fixed Spool Reel, as supplied with the rod. The line, supplied with the reel, was a 15 pound (6.8 kilogram) test strength nylon “running line” that has a diameter of 0.5 mm. In the tests reported here, only a weight was cast. Three different weights were used, 2, 3 and 4 ounce (57, 85 and 113 grams). These were standard torpedo shaped lead weights. One of the weights was attached to a ~ 4 metre length of 50 pound (24 kg) test strength nylon line, known as the “shock leader”. This was in turn tied to the running line. The purpose of the shock leader is to withstand the high centrifugal forces exerted by the weight on the line during casting that could break the lighter running line. During casting, the angler holds the end of the shock leader line with one finger against the rod handle so that casting forces are exerted on the shock leader line and not on the running line. At the release point of the cast, this finger is raised, releasing the shock leader which is drawn through the line guide rings by the motion of the weight, which in turn draws the running line off the reel and through the same rings and away from the rod. The casts were recorded using a Sony DCR-TRV 900E Mini-DV Camcorder with an effective video resolution of 420 k pixels. This camera records at 25 frames per second, thus yielding 50 fields per second for the analysis.
The Engineering of Sport 7 - Vol. 1
185
3- Casting experiments A large number of casts were video recorded and the corresponding cast distances were measured. Different combinations of weight and shock leader lengths were tried. Casts of one particular combination of shock leader length and weight (4 foot (1.22m) and 3 ounce (85 grams) were repeated 14 times to study the variability in cast distance. The video analysis was completed using Vicon Peak Motus software [Vicon Motion Systems, Inc., Centennial, CO]. The digitizing process of the video sequences involves progressing through each field in the video (there are two fields per frame, and thus 25 frames per second (yielding 50 fields per second) and marking all the relevant locations the user wishes to analyse further. For this analysis, nine points were selected along the rod, and one for the weight. The points used were where the white tape had been added to the rod, mentioned above. The locations on the rod were not equal distances apart as fewer are required towards the base of the rod which has a thick section that bends little during casting.
4- Results
Figure 1 - A sequence of digitised stills from the video recording of a cast.
Figure 1 is a selection of images from the video recording of a cast. The magnitude of the velocities of the weight and the rod tip during the cast, given by the Peak Motus software, are shown in figure 2. Figure 2 shows the weight and rod tip to have similar velocities for the majority of the cast as the weight is towed up from the ground.
186 The Engineering of Sport 7 - Vol. 1
Figure 2 - Resultant velocities of the weight and the tip of the rod plotted against time. Data is obtained from the digitised images, figure 6, using Peak Modus software.
The period of acceleration, figure.2, corresponds with the bending of the rod evident in figure 1. This bending is a reaction to the force being used to accelerate the weight. Measurements of the stiffness of the rod confirm the bends in figure 1 to match the acceleration of the weight indicated in figure 2. It can be seen from figure 1 that towards the end of the cast, when the rod is approximately vertical, the weight begins to orbit the rod tip. It is during this phase of the cast that the line is released and the precise time of release determines the angle at which the weight moves away from the rod. The centrifugal forces in the shock leader line are at their maximum during this phase of the cast. The data at times after the line release were used to estimate the magnitude of the weight velocity and the angle at which it was moving at line release. An error analysis indicated an uncertainty of ± 1 ms-1 in estimations of weight launch velocity and ± 2o in launch angle. The results shown in figure 3 were obtained using three different weights, 2, 3 and 4 ounce (57, 85 and 113 grams) and shock leader lengths of 0, 2, 4, 6 and 8 ft (0, 0.61, 1.22, 1.83 and 2.44 m)
The Engineering of Sport 7 - Vol. 1
187
Figure 3 - Comparison of predictions of cast distance, calculated with and without weight drag, with actual experimentally measured distance.
5- Analysis of cast distances The first approximation is to assume the weight to be a freely moving projectile and to estimate cast distances from observed launch angles and launch velocities. However, it is clear from figure 1 that the weight is released from a point several metres above ground and this factor must be added to calculations at the outset. It is well known that the distance, s, travelled by a freely moving projectile launched from a point on the ground is given by: (1) In which u is the launch velocity, is the launch angle and g is acceleration due to gravity. It can be seen from this equation that the optimum launch angle is 45° as this leads to sin2 taking its maximum value of 1. It can also be seen that projectile distance is fairly insensitive to launch angle because of the slow variation of sin2 for values of around 45°. If the projectile is launched at a distance h above the ground, the projectile distance becomes: (2)
188 The Engineering of Sport 7 - Vol. 1 Where ~ 45° and h <<s, the additional distance travelled by the projectile is ~ h. These conditions are met here leading to the difference between cast distances predicted by equations 1 and 2 being ~ the launch height h, which can be seen from figure1to be ~5.5 m. The next approximation is to include in calculations the drag of the air acting on the weight. The drag force F acting on a projectile moving at a velocity v through air is: (3) Where: = Air Density, taken as 1.3 kg/m2, standard air density at sea level. CD = Drag Coefficient, taken at 0.5, an average figure for a sphere. A = Cross-sectional area of the weight. The 2, 3 and 4 ounce weights were torpedo shaped with maximum diameters of 14, 16 and 18 mm, respectively. The results of calculations of cast distance based on equation 2 and incorporating the effect of the drag force, equation 3, are shown in figure 3. The results indicate the drag force on the weight to reduce the cast distance by a relatively small amount ~ 10 m. However, there is a substantial difference between the cast distances predicted by eqns. 2 and 3 and the actual cast distances achieved in the videoed tests. These can be seen to be a factor of between two and three times smaller than the theoretical predictions. It must be concluded that there are additional drag forces on the weight that have not been included in the analysis. The potential sources of these additional forces are: the drag of the air on the running line as it is pulled through the air by the weight; the friction of the running line as it passes through the rod rings on leaving the reel and the force needed to pull this line off the reel. A further set of experiments were performed to eliminate the drag of the running line during casting. In these experiments, casts were made without the running line. The weight was attached to a length of shock leader line alone, sufficient to allow casting. The length of line was chosen to give a leader length of 4ft (1.22 m) and just sufficient extra line for its end to be held in the usual way during casting, a total length of ~ 13 ft (4 m). The line was attached to a 3 ounce (85 gram) weight but not to the running line, to avoid the drag of the running line during a cast. The cast distances achieved without the running line are shown in figure 3.
Figure 4 - Cast distances achieved without running line, 3 ounce (85 gram) weight and a leader length of 4ft (1.22 m).
The Engineering of Sport 7 - Vol. 1
189
These results are directly comparable with those for the 14 repeated casts, average distance 54.43 m, as they involve the same weight and shock leader length and they were obtained in the same experimental session under the same wind conditions. It is clear from the cast distances achieved without the running line that the drag force of the running line is substantial and the major reason for the discrepancy between the theoretical and experimental results in figure 3.
6- Conclusions The video analysis of the overhead form of beach casting shows it to be a fairly reliable method of launching the weight at an angle that is close to the optimum angle of 45°. Analysis, eqn.1, indicates the insensitivity of cast distance to launch angle θ. Application of this formula to the recorded data in figure 3 shows that the effect of deviations in θ from 45° accounts for, at most, a 5% reduction in cast distance. The variation in launch velocity is significant. Examination of figure 3 does not reveal any strong correlation between launch velocity and shock leader length or the mass of the weight. It is concluded that the variation is characteristic of the mechanical process of casting repeated by the angler. Equations 1 and 2 indicate cast distance to be strongly dependent on launch velocity (∝U2) but there is no clear evidence of this dependence in the data contained in figure 3. The most striking result of this study is the large difference between the actual measured cast distances and those predicted on the basis of the weight launch characteristics, shown in figure 3. This result lead to the conclusion that the running line exerts a very significant drag force on the weight; causing cast distances to be far shorter than expected. This conclusion was substantiated by the tests in which the weight was cast free of the running line (figure 4.). The main conclusion of this study is that the beach caster is hampered in his/her efforts to cast far from the shore by the drag of the running line. Further research may reveal the proportion of this drag caused by the line being towed through the air and the proportions cause by friction between line and rod rings and or reel. An understanding of the principle factors determining running line drag may lead to the design of equipment that will lead to its reduction.
7- References J M Robson, “The physics of fly casting”, Am.J.Phys. 58 (1990) 234-240. G A Spolek, “The mechanics of flycasting: the flyline”,Am.J.Phys. 54 (1986) 832-838. J Holden, “Long distance casting” The Crown Press (1982).
An Assessment of Sensing Technologies to Monitor The Collision of a Baseball and Bat (P34) Lawrence Fallon1, James Sherwood2, Michael Donaruma3
Topics: baseball, sensors. Abstract: Some of the key parameters that make up the collision of a baseball and a bat include the velocities of both just prior to impact, the horizontal and vertical location of impact, the angle of impact and the geometric and physical properties of the bat and ball. The objective of the batter is to optimize this collision to produce the maximum batted-ball speed at a desired trajectory angle. Through experience, a professional batter uses his senses of hearing, vision and touch to help him develop the swing and timing to create this optimum collision. The location of this optimum impact is commonly referred to as the sweetspot on the bat. A study was performed to evaluate different types of sensors, such as accelerometers, strain gages and microphones, and to quantify their ability to identify these collision parameters. High-speed motion analysis was used to evaluate the accuracy of the measurements as they relate to location and velocity. Data were collected and were statistically quantified to compare the effectiveness of each technology. Keywords: baseball, sensors, collision.
1- Introduction A wood baseball bat was instrumented with several different types of sensors. A series of tests was performed using this bat to make an initial assessment of the applicability of these sensors to identify impact location and the swing speed of the bat. The study was as much an investigation of baseball as it was an exercise in simultaneous recording of multiple technologies using a single data acquisition system. The objective of this initial assessment was to obtain test data, identify problems associated with using these sensors and evaluate methods to process the data. The actual values of strain, acceleration, sound-pressure level were not the main focus, but relative levels were used to draw comparisons of impact locations and swing speed. 1. University of Massachusetts Lowell, One University Ave Lowell, MA USA 01854 - E-mail: [email protected] 2. E-mail: [email protected] 3. E-mail: [email protected]
192 The Engineering of Sport 7 - Vol. 1 A total of five different types of sensors were mounted onto a bat, including four strain gages, four microphones, four piezoelectric accelerometers, a variable capacitance accelerometer and a gyro assembly. The series of tests included stationary bat impacts, noncontact swings and finally swings impacting a ball off a tee. For the stationary impact tests, the bat was supported by a single hand grip approximately 15 cm (6 in.) from the knob. Swinging the bat at the batter’s maximum potential would exceed the range of some of the sensors. An effort was made to swing the bat at approximately half-speed.
2- Methodology 2.1 - Instrumented Solid-Ash Bat A bat’s sweetspot is typically located about 15 cm (6 in.) from the end of the barrel of the bat. The solid ash bat used for these tests measured just under 89 cm (34.8 in.) and weighed 0.91 kg (32 oz.) prior to instrumentation. The gyro assembly was mounted on the upper part of the handle facing upward to measure the main axis of rotation. The variable capacitance accelerometer was mounted on the barrel end of the bat. The remaining sensors were grouped into four sets of three sensors and mounted along the bat per the locations specified in Table 1. All distances are referenced from the knob end of the bat. Table 1 - Sensor Locations.
The strain gages were installed on the bat using a flexible adhesive over copper tape. The accelerometers and gyro were also attached using a fast curing adhesive, and the variable capacitance accelerometer was further secured with a small wood screw. The microphones were installed along the impact side of the bat, whereas the PE accelerometers and strain gages were bonded to the opposite side. There was concern that the microphone might be affected by the wave propagating along the bat as opposed to the sound pressure transported through the air. To minimize this effect, a viscoelastic damping foam was placed between the microphones and the bat which were then secured using reinforced tape. The speed of sound in ash along the fibre is 4000 m/s (890 mph) and in theory, the impulse travels from the barrel to the handle in 0.21 ms (Cross 1998). However, tests demonstrated that the travel time was significantly longer because the impact on the side of the bat caused an initial propagation across the growth rings which has a speed of sound about 1/4 that of the fibre direction (Brinsmead 1889).
The Engineering of Sport 7 - Vol. 1
193
2.2 Sensors Pre-wired 350-ohm strain gages, Omega type KFG-3-350-C1-11L3M3R, were used for the testing. The strain gages had a gage factor of 2.1. For the stationary bat tests, the data acquisition settings included a quarter bridge setup, a 5K gain, an excitation of 5V, a 3.3 kHz filter and a sensitivity of 525 mV/Vex. The handle and mid-bat settings were further adjusted to account for increased strain during swinging conditions. The piezoelectric (PE) accelerometers were Endevco model 2250A-10 which has a frequency range of 2 to 8000 Hz and a maximum acceleration of 500 g. The variable capacitance (VC) accelerometer was Endevco model 7596-30 which has a frequency range of 0 to 800 Hz and a maximum acceleration of 30 g. The microphones were Panasonic model WM-63PR, Omnidirectional Electret Condenser Microphone Cartridges. Omnidirectional microphones pick up sounds from any direction, because the electronic pick-up is placed in the center of a mesh-covered dome. The microphones were isolated from the ash bat with the objective of only recording the sound wave travelling through the air at 343 m/s (770 mph). The gyro assembly module had been previously used on rail gun tests and was specifically designed by Draper Lab for high-shock environments. The gyro’s support electronics increased the commercial gyro’s capabilities to measure rates as fast as 300 deg/s about an axis perpendicular to the module.
2.3 Data Acquisition System A Liberty (part no. 269-945083, manufactured by LDS Test and Measurement LCC, Wisconsin, USA) ruggedized DAQ (Data Acquisition) System was used to acquire all accelerometer, gyro, strain-gage and microphone responses. The Liberty DAQ System consisted of the Liberty mainframe, signal conditioning modules, battery module, Perception software and a PC (personal computer) running on a network. The system was supported by a custom aluminium frame. The signal conditioning modules included an eight-channel Bridge Signal Conditioner (BR-8) to acquire the strain gage data, an eight-channel General Purpose Signal Conditioner (GP-8) to acquire the VC accelerometer and a 16-channel General Purpose Signal Conditioner (GP16) to acquire the gyro module data and an eight-channel ICP Signal Conditioner (ICP-8) to acquire the piezoelectric accelerometer and microphone response data. The gyro module required an external 5V power supply, and the VC accelerometer was powered with 12V. Depending on the internal settings, the sampling rate was limited from 20 kS/s to 50 kS/s. All of the data collected were stamped using an IRIG (InterRange Instrumentation Group) time. The Liberty uses two internal oscillator clocks and was synchronized to a motion analysis system using a video playback feature. A motion analysis system was used to validate the impact location and to measure the bat velocity at impact which could be correlated to the centripetal acceleration measured by the VC accelerometer. The motion analysis system recorded the impacts at 1000 frames/s. For the stationary bat tests, the bat was impacted with a ball lightly covered with chalk. The chalk left a removable mark on the bat identifying the location of impact.
194 The Engineering of Sport 7 - Vol. 1
2.4 General Description of Data Analysis Sensor response data were captured for the three series of tests and imported into Microsoft Excel files for processing. The data file included time and response values for each sensor. Time values were extracted which corresponded to each sensor’s first response to the impact of the ball on the bat. Additionally, the time values associated with the peaks of the first microphone responses were also extracted. The time values were plotted against the sensor’s location along the bat. A line was curve-fit through the data, and the location in which the slope of the line was zero was recorded as the predicted impact location by that type of sensor. This predicted impact location was compared against the actual as measured by the chalk mark and the motion analysis system.
3- Results 3.1 Stationary Test A total of eight impacts were recorded for the stationary tests. Two impacts were recorded at the sweetspot, above the sweetspot, below the sweetspot and near the midpoint of the bat. The variable capacitance accelerometer and gyro assembly data were not recorded during stationary tests as their function was more applicable to determining swing speed. Because the maximum sampling rate for this setup was 20 kS/s, which translates to 0.00005 s/S, the speed of data acquisition did not provide many timestamps between adjacent sensors. Some interpretation of data was required to obtain the initial response of the sensors. The microphone peaks were recorded for only the first pulse because of secondary waves resulting from reflections off the test-room walls. A second time value was also extracted for the microphones which corresponded to the peaks of their first response. Impacts near the handle or end of the bat resulted in some saturated strain-gage data during non-sweetspot impacts. The data from the stationary tests were manually extracted from the DAQ files and are listed in Table 2. These data show when the pulse reached each sensor. The accelerometers recorded the first response, and the impact was then received by the strain gage and microphones. The timing between impact locations (1-4) was somewhat inconsistent. For example, Run No. 2 was impacted at 72 cm. The pulse first reached accelerometer 2 and accelerometers 1, 3 and 4 with delays of 0.02, 0.06 and 0.13 ms, respectively. The microphone’s responses to the impact occurred from 0.57 to 0.93 ms following the initial measurement and similarly the strain gages responded with a lag of 0.76 to 1.21 ms.
The Engineering of Sport 7 - Vol. 1
195
Table 2 - Stationary Test Normalized Signal Start Times (ms).
For the stationary bat tests, the bat was impacted with a ball lightly covered with chalk, which left a removable mark on the bat identifying the location of impact. The sensor location was plotted vs. the acquisition time (first response or “start time” for all three sensor types and peak response for the microphone). A sixth-order polynomial fit (Figure 1) was used to determine an estimated impact location. Table 3 summarizes the results and present a first-order accuracy. The PE accelerometers resulted in the best predictor of impact location—being able to predict the impact location within 5 cm (2 in.) and within 2.5 cm (1 in.) 63% and 38% of the time, respectively. The microphones were slightly less accurate with the start time analysis providing better results than the peak-time analysis. The strain gages were the least accurate in estimating impact location partly due to the difficulty in identifying a signal near the end of the barrel. Perhaps a method that offers better accuracy is a comparison of the peak-value ratios. Table 4 presents the ratios for the PE accelerometers and the microphones. Because of saturation and unidentifiable data on the strain gages, this method was not applicable. A comparison of the peak response levels demonstrated very consistent results for the accelerometers. If a modal analysis were to be performed on the bat, these ratios may offer an excellent method to identify impact. For example, for the two endof-the-barrel impacts, the ratios of accelerometers 2, 3 and 4 with respect to accelerometer 1, were 2.0, 3.0 and 1.8 respectively for the 80-cm impact and 2.0, 2.8 and 1.4 respectively for the 81-cm impact.
Figure 1 - Stationary Run No. 5 Sensor Timestamp vs. Sensor Location.
196 The Engineering of Sport 7 - Vol. 1 Table 3 - Stationary Test Results Summary (Time Data).
Table 4 - Stationary Test Results Summary (Peak Data).
3.2 Swing and Tee-Hitting Tests With the bat fully instrumented, several practice swings were taken to set the gains on the transducers and to capture swing data without impact. Figure 2 shows a photo of the tee-hitting test set-up. The photograph shows the instrumented bat, the ball on a tee (construction cone), the power supply, DAQ system the locations 1 through 4 and sensor wiring harness.
The Engineering of Sport 7 - Vol. 1
197
Figure 2 - Tee-Hitting Test Set-Up.
Figure 3a plots non-impact swing data of the gyro module and VC accelerometer. The response levels were normalized and are dimensionless. The accelerometer provided essentially a noise-free response, while the gyro had considerable noise. A polynomial trend line through the gyro data aided in interpreting the results.
Figure 3 - a) Non-Impact Swing, b) Impact off Tee.
Several impacts were recorded during Tee-Hitting including two sweetspot and two non-sweetspot impacts, and those data are presented in Figure 3b, which plots each of the sensor types during a sweetspot impact. The data were rescaled to help in visualizing the results. The plots indicate some gyro and VC accelerometer offset after impact. The VC accelerometer also saturated at maximum swing speed. The strain on the handle indicates the gages in compression during the start of the swing with the gages 3 and 4 switching to tension prior to impact. If these gages are correlated for the length of the bat, it is possible to determine if the “whip effect” was maximized upon impact (i.e., the small strains at the location opposite the impact point switch from compression to tension precisely at impact).
198 The Engineering of Sport 7 - Vol. 1 Based on the accelerometer data for the four impacts, the average pulse speed along the bat was 2100 m/s (6900 ft/s) which falls between the published numbers for along the fibre and across the growth rings. To improve the curve-fitting method used to calculate impact location, an imaginary (extrapolated) accelerometer was assumed to be 18 cm (7 in.) beyond Accel 1, just off the end of the barrel. It was assumed that this accelerometer start time occurred at a time equivalent to that of Accel 1 plus 18 cm (7 in.) divided by 2100 m/s. Table 5 lists the accelerometer response data which were slightly less accurate than the stationary bat results. Impact number 5 was first recorded by Accel 2 followed by Accels 1, 3 and 4 with lag times of 0.03, 0.21 and 0.32 ms, respectively. Impact 5 predicted an impact location 2 cm different from that recorded by the motion analysis system. Additional testing with the bat rotated 90 degrees and an off-center impact demonstrated that the accelerometers and strain gages could be used to determine the severity of the eccentric impact. Table 5 - Tee-Hitting Test Results Summary (Normalized Accel Start Time Data).
4- Conclusions All five sensors provided useful data when analyzing the swing and impact. There were issues with limitations which caused voltage saturation in some cases. The VC accelerometer, gyro module and strain gages provided information on the swing although the strain gages and gyro had considerable noise. The accelerometers and microphones were the best at determining impact location with the most consistent and most promising method of data interpolation being accelerometer peak G ratios as compared to modal analysis results.
5- References [B1] Brinsmead, E., The History of the Piano, Simpkin Marshall & Co., 1889. [C1] Cross, R., The Sweetspot of a Baseball Bat, American Association of Physics Teachers, 1998.
6- Acknowledgement The authors would like to recognize the support of Chris O’Brien, Bob Wadland and Elliott White.
Correlation Between the Linear Impulse and Golf Ball Spin Rate (P35) Woo-Jin Roh1, Chong-Won Lee1
Topics: Golf. Abstract: Golf ball spin rate after impact with club is created by the contact force, which is greatly influenced by ball and club mass, material, impact speed, and club loft angle. Previous studies showed that the contact force is determined as the resultant force of the reaction forces normal and tangential to the club face at the contact point. The normal force causes the compression and restitution of the ball, and the tangential force creates the spin. Especially, the tangential force takes either positive or negative value as the ball rolls and slides along the club face during impact. Although the positive and negative tangential forces are known to create and reduce the back spin rate, respectively, the mechanism of ball spin creation has not yet been discussed in detail. It is shown in this work that the linear impulse of the tangential force is directly related to generation of back spin rate of golf ball. The linear impulse can be calculated from the tangential force, which depends upon many factors such as ball and club mass, material, impact speed, and club loft angle. In this research, the influence of the contact force between golf club and ball is investigated to analyze the mechanism of impact. For this purpose, the contact force and the contact time at impact between golf club head and ball are computed using finite element method (FEM). Keywords: back spin rate, impact, contact force, tangential force, linear impulse, FEM.
1- Introduction The impact duration between the golf club using a #1 wood and ball is known to be about 400~600 ms depending upon the club’s initial velocity and the composition of the ball. The ball gains a spin rate of approximately 2000~10000 rpm, as it launches off a club. The contact force (F) between a golf club head and a ball is determined as the resultant force of the reaction forces normal, and tangential to the club face at the contact point, Fn and Ft, respectively, as shown in Figure 1. Previous studies showed that Fn and Ft, respectively, attribute to flying and backward spinning of ball (Roh and Lee, 2007). 1. Center for Noise and Vibration Control (NOVIC), Department of Mechanical Engineering, KAIST, Daejeon, Korea E-mail: seungbusa,[email protected]
200 The Engineering of Sport 7 - Vol. 1
Figure 1 - Golf club head and ball models at impact.
Figure 2 is the typical plot of the tangential force acting on the contact surface of a golf ball during impact, which was calculated from the FE models of golf club head and ball in Figure 1. The tangential force is positive and creates back spin at the initial stage of impact between the club and the ball. Later in the impact, the tangential force changes its direction and reduces the back spin. Previous studies showed that the change in the direction of tangential force takes place, because the golf ball is twisted at the initial stage of impact and then this deformation is released during the expansion stage of impact (Moriyama et al. 2004). The tangential force directly affects on generating the spin rate of golf ball after impact. The tangential force varies depending upon many factors such as the impact velocity and loft angle of club, the golf ball and club material, and so on. It means that it becomes extremely difficult, if not impossible, to investigate the effect of so many factors on the tangential force profile in detail and the generation of ball spin rate. In this paper, we attempt to find the correlation between the linear impulse, which is the integration of the tangential force in time, and the spin rate, so that we can directly estimate the latter from the information of the former, irrespective of other factors.
Figure 2 - Tangential force acting on golf ball during impact.
The Engineering of Sport 7 - Vol. 1
201
2- Linear impulse 2.1 Theoretical analysis of impact
Figure 3 - Impact of club and ball.
The club approaches the golf ball with a velocity V at a loft angle , as shown in Figure 3. The golf ball is moving upward along the inclined surface of the club face at impact. If the ball deforms and slides without friction along the club face, it continues to travel along the club face at Vsin , but if the force of friction is large enough to prevent the ball from pure sliding, the ball starts to roll and, before the end of the impact, it may be rolling upward along the club face with a smaller velocity. We may calculate the final velocity of the ball (v2) by equating the impulse on the ball from the possibly varying frictional force F(t) opposing the motion of the ball to the decrease in the momentum of the ball along the club face. Therefore, the linear impulse-momentum relation for a ball to initially slide into the club face can be written as (1) The final velocity of the ball (v2) and final spin rate () was calculated by Jorgensen using the compression factor f (Jorgensen, 1994) as (2) (3) From Eqs. (1) to (3), we obtain the relations (4)
202 The Engineering of Sport 7 - Vol. 1
(5) and the expression for the spin rate (6) Equation (6) shows that we can simply calculate the spin rate from the linear impulse independent of other factors including the compression factor. However, Eq. (6) is an approximate relation for the actual spin rate, ignoring the detailed force-deformation analysis.
2.2 Comparison of theoretical and FEM results The numerical analysis using FEM has been carried out to calculate the tangential forces in detail for various cases. For example, the ball spin rate was calculated to be about 2000 rpm by using FEM result (Roh and Lee, 2007), for the case of V = 35 m/s, = 11.5°, m = 46 g, and R = 21.4 mm. On the other hand, we can calculate the ball spin rate from the approximate relation (6) as (7) where the linear impulse is calculated to be 0.0693 N-sec using the tangential force profile from the FEM. The estimated ball spin rate of 2441 rpm is larger by about 20 % than the FEM result of about 2000 rpm. The discrepancy is due to the simplistic assumption of v2 = R, implying that the ball is assumed to be not deforming during impact. In fact, the effective radius of deformed ball during impact varies drastically, because the real golf ball is greatly deformed during the whole process of impact. Figure 4 shows the FEM result that the ball is compressed at the initial stage of impact, and then swollen after about 300 ms, the radius of ball at the final stage of impact being even larger than the free ball radius of 21.4 mm temporarily.
Figure 4 - Effective radius of deformed ball at impact.
The Engineering of Sport 7 - Vol. 1
203
The ball spin rates that are calculated by Eq. (6) and FEM are compared, with the impact velocity and loft angle varied, in Figures 5 and 6. The larger the impact velocity and loft angle are, the larger the linear impulse and thus the ball spin rate. Note that there exists a consistent tendency of proportionality between the spin rates calculated from two different methods. The discrepancy in proportionality constant between the two results, as the linear impulse increases, is due to the decrease in the effective radius of deformed ball at impact.
Figure 5 - Spin rate vs. impact velocity (=11.5 °).
Figure 6 - Spin rate vs. loft angle (V=35 m/s).
3- Correlation between linear impulse and spin rate 3.1 Effect of the ball material stiffness on spin rate As previously mentioned, the larger the linear impulse is given, the larger the ball spin rate is. This means that the positive tangential force increases the ball spin rate and the negative tangential force decreases the ball spin rate. The tangential force is a function of club speed and loft angle at impact. However, Eq. (6) implies that, for given club speed and loft angle at impact, the tangential force may have an adverse effect on the ball spin rate. Although there is a discrepancy in effect of tangential force on the spin rate between Eq. (6) and the FEM results, the reason why the previous results show similar tendency, as shown in Figures 5 and 6, can be interpreted as: other factors have great effect on the ball spin rate. For example, when the impact speed or the loft angle is increased, the linear impulse is also increased. However, the product of the impact velocity and the loft angle increases much faster than the linear impulse, according to Eq. (6). The ball spin rate can also depend upon the ball material stiffness with the other factors unchanged such as the impact speed and the loft angle. Tanaka demonstrated that the ball spin rate tends to increase (decreases) with the Young’s modulus of the core (the cover) (Tanaka et al. 2006). Meaning that a softer cover will give a higher spin rate than a harder cover, as experienced by players. On the other hand, a softer core will give a smaller spin rate than a harder core, which has not been reported. We can calculate the linear impulse and the ball spin rate depending upon the ball core or cover material stiffness using the FEM, as shown in Figure 7. It shows that a
204 The Engineering of Sport 7 - Vol. 1 harder core gives the higher linear impulse than a softer core, whereas a harder cover results in a smaller linear impulse than a softer cover. Note that the ball spin rate computed from Eq. (6) differs in tendency from the FEM results and the results by Tanaka. For the normal golf ball core and cover Young’s modulus of 40 Mpa and 300 Mpa, respectively, the two results provide quite similar spin rate. Thus, Eq. (6) is a good approximation for calculation of spin rate for given linear impulse only for the most of commercial golf balls. Equation (6) inherently fails in reflecting the ball material properties in calculation of spin rate. This is the subject for a future study.
Figure 7 - Spin rate vs. ball material hardness.
3.2 Empirical formula for ball spin rate
Figure 8 - Spin rate vs. linear impulse: FEM.
The Engineering of Sport 7 - Vol. 1
205
Although the correlation between the linear impulse and golf ball spin rate is complicated as explained before, we attempt to induce an approximate linear relation based on the FEM results, so that we can estimate the ball spin rate for given linear impulse, irrespective of other factors. Figure 8 shows the ball spin rates computed by FEM with respect to the linear impulse when the impact factors are varied. From the figure, we can derive a simple empirical formula by linear curve fitting to the simulated results, i.e. (ball spin rate in rpm) = 2.4 104 (linear impulse in N – sec)
(8)
Note that the curve-fit error is about 5 %. The rational of the formula (8) can be checked by comparison with previous results. For example, Figures 9 shows the tangential forces calculated by Gobush (Gobush 1990) for impact speed of 29 m/s and loft angle of 20 °, and Moriyama (Moriyama et al. 2004) for impact speed of 34 m/s and loft angle of 12 °, respectively. The linear impulses computed from Figure 9 are 0.1315 and 0.095 N-sec, respectively, resulting in the spin rates of 3100 and 2200 rpm according to the formula (8). The estimated spin rates are quite compatible with other results.
Figure 9 - Tangential force by Gobush (1990).
4- Conclusions It is important to improve the initial launch conditions of golf ball at impact between golf club and ball to get longer flight distance or a fair green regulation. The ball flight is greatly influenced by the initial launch conditions such as ball speed, launch angle and back spin rate. However, the mechanism of ball spin generation has not yet been comprehended. Back spin rate is known to be created by the tangential force, which attributes to the linear impulse. In other word, the linear impulse can be calculated from the tangential force, which depends upon many factors such as ball and club mass, material,
206 The Engineering of Sport 7 - Vol. 1 impact speed, and club loft angle. Thus, if we know the linear impulse, we can estimate the ball spin rate without considering other factors with fair accuracy. Using FEM results, we can obtain the simple empirical formula between the linear impulse and the ball spin rate. It provides a good estimation of the spin rate from given linear impulse. The error of about 5 % in estimation is mostly due to the ignorance of other factors such as golf ball and club material properties.
5- References [G1] Gobush, W. Impact force measurements on golf balls, Science and Golf, Proceedings of the World Scientific Congress of Golf, 219-224, 1990. [J1] Jorgensen, T. P. The physics of golf, American Institute of Physics, 1994. [MY1] Moriyama, K., Yamaguchi, T., and Yabu, M. The influence of mechanical impedance of the golf club and the golf ball on ball spin, Proceedings of The engineering of sport, 5th International Conference, 337-343, 2004. [RL1] Roh, W. J., and Lee, C. W. Analysis of golf ball spin mechanism at impact by FEM, Proceedings of 14th International Congress on Sound and Vibration, Cairns-Australia, 2007. [T1] Tanaka, K. et al. Construction of the finite-element models of golf balls and simulations of their collisions, Proceedings of the Institution of Mechanical Engineers Vol. 220 Part L: J. Materials: Design and Applications, 13-22, 2006.
Dynamics-based Force Sensor Using Accelerometers-application of Hammer Throw Training Aid- (P37) Ken Ohta1, Koji Umegaki2, Koji Murofushi3, Ayako Komine1, Chikara Miyaji1
Topics: Athletics; Biomechanics; Measurement Systems. Abstract: Dynamics-based force sensor using accelerometers which measures forces and joint torques has been developed. In this study we have applied this method to hammer throw training aid integrating small sensors, signal processing, short-range wireless transmission, wearable data-logger and biofeedback training system. The purpose of this study was to establish methods for the measuring of rotational movement and the biofeedback training system for hammer throwers. Microelectromechanical systems accelerometers were chosen as the sensor platform capable of because they are noninvasive miniaturized devices and have wide bandwidth. In this system, a wireless data-logger was developed as a wearable device to replace cables and reduce constraint caused by wearing cables. The transmitted data were given as biofeedback information over a speaker through signal processing and voltage to frequency conversion. Keywords: accelerometer; biofeedback; force sensor; measurement system; dynamics model-based sensor.
1- Introduction Accelerometers are used in the study of a variety of human movement. One of the most typical applications of accelerometory for sports and human activity is the estimation of energy for multiple applications. A small portable accelerometer to estimate the energy expenditure of daily activities (Montoye et al. 1983) and a portable data processing unit for the assessment of daily physical activity (Bouten 1997) were developed. Mercer et al. investigated relationship between impact acceleration and running velocity using uniaxial accelerometers (Mercer et al. 2005). Since accelerometers are also useful for measuring rigid body movement, the methods for accelerometer-based rotational move1. Japan Institute of Sports Science, 3-15-1 Nishigaoka, Kita, Tokyo, 115-0056 JAPAN E-mail: ohta, komine.ayako, [email protected] 2. Maizuru National College of Technology, 234 Shiroya Maizuru, Kyoto 625-8511, JAPAN E-mail: [email protected] 3. Chukyo University, 101 Tokodachi, Kaizu, Toyota, Aichi, 470-0393 JAPAN - E-mail: [email protected]
208 The Engineering of Sport 7 - Vol. 1 ment measurement have been presented. Bogert developed a method to calculate total resultant force and moment on a body segment from accelerometer data, but incorrect assumption leads unreliable results (Borgert et al. 1996). Signals, which are captured by accelerometer, are composed of linear, gravitational, centrifugal and angular acceleration. To decompose each acceleration in three dimensions, at least nine accelerometers are required (Zappa et al. 2001). This accelerometry prepares complete set of kinematic data for inverse dynamic analysis. It allows to real-time calculation of forward dynamic analysis as well as inverse one. Using the method with haptic interface (Yano et al. 2003, Ando et al. 2004) it also realizes the force display devices capable of realizing virtual tennis racket, golf club, and other tools. In this study we applied this method to hammer throw training aid integrating small sensors, signal processing, short-range wireless transmission, wearable data-logger and biofeedback training system. The purpose of this study was to establish methods for measuring of rotational movement and biofeedback training system of hammer throwers during their training. Microelectromechanical systems (MEMS) accelerometers were chosen as the sensor platform capable of noninvasive, wearable, and real-time monitoring. In this system, a wireless data-logger was developed as a wearable device to replace cables and reduce constraint caused by wearing cables. The transmitted data were given as biofeedback information over a speaker through signal processing and voltage to frequency conversion.
2- Hammer movement
Figure 1 - Coordinate system.
The Engineering of Sport 7 - Vol. 1
209
Figure 2 - Hammer head angular velocity and angular acceleration.
Fig.1 shows a coordinate system to describe the hammer movement. The angular , velocity and the angular acceleration Yxy of hammer, which were measured by motion capture system, are shown in Fig.2 (Murofushi 2005). During double support phase, when the thrower has both feet on the ground, the most of angular acceleration of hammer is positive. The single support phase, contrarily, is when the thrower has one foot off the ground. Throughout the throw, the thrower repeated both phases. To accelerate the hammer the integration of the angular acceleration must be as greater as possible since the movement of instantaneous center for the hammer is small. Assuming that the rotational movement of the hammer head composes a plane, motion of equations for the hammer head are written in polar coordinates as (1) (2) where m is the mass of hammer, fr and f are the forces acting on the hammer for the radial coordinate r and the angular coordinate . In this model, corresponds to the angular velocity xy on the moving coordinate system in Fig.1. From the equation (6), Coriolis force (-2mr) and the angular component of wire force f have effect to accelerate the hammer. However, arising Coriolis force makes shorter radius r and greater angular velocity accordingly. Consequently, this makes the small angle between the wire force vector and the line pointing from the hammer head to the instantaneous center of rotation and f. Considering this compound effect of Coriolis force and the angular component of the wire force, resultant angular acceleration must be important information as the skill training for hammer throwing. However, since angular compo-
210 The Engineering of Sport 7 - Vol. 1 nent f is quite smaller than radial one fr, it is hard to sense the angular acceleration during throwing. We choose consequently resultant angular acceleration as key information which must be fed back to thrower during training.
3- Method 3.1 Accelerometry for Hammer throwing
Figure 3 - Coordinate system.
As a general case, two triaxial accelerometers are shown in Fig.3. When a triaxial accelerometer is attached to a rigid body segment at any point L of a segment, the acceleration at point L is: (3) where r is the position vector of the origin of the segment coordinate system S from the origin of world coordinate system O, g is the gravitational acceleration vector, =[x, y, z]T is the angular velocity vector of the segment, l is the position vector of a point L. These vectors are with respect to the segment coordinate system S. When triaxial accelerometers A1 and A2 are attached on a same axis of segment coordinate system S aligned parallel with each other, the difference of two accelerometers is (4) where l12 is a relative position vector pointing from A1 to A2. When the accelerometers are on axis z as shown in Fig.3, the equation (2) is rewritten as scalar equations (5)
The Engineering of Sport 7 - Vol. 1
211
where d is a distance between the accelerometer A1 and A2. Note that relative signals of two accelerometers a12 are centrifugal and angular accelerations which are linear with respect to the distance d. Since rotational movement z of hammer is small, relative signal a12 is rewritten as (6)
The method was applied for analysis of hammer movement and practical training aid system. As shown in Fig.4, two dual axis, ±50g MEMS analog devices accelerometers (ADXL 278) (Analog devices 2005) for axis z and two dual axis, ±2g MEMS STMicoroerectronics (LIS3L06AL) (STMicroelectronics 2006) for axis x and y were mounted to a board and aligned on a line creating two triaxial accelerometers. The distance of two triaxial accelerometers d is 100 mm.
3.2 – Wireless data-logger and biofeedback system
Figure 4 - Sensor board and wireless data-logger.
A wireless data-logger was developed for the training aid system providing minimally cable constraints and real-time monitoring (Fig.4). The logger transferred the data of accelerometers sampled with a frequency of 400 Hz using 12-bit analog-to-digital converters to a PC. This system comprised of an analog devices ADUC812 microcontroller, a Class1 Bluetooth wireless system, a 256MB flash memory, and a secondary battery. The sensor board was connected the data-logger through connecters set near a handle of hammer. The connectors were disconnected after releasing of the hammer, and consequently, no signals were transmitted. The transmitted signals of the angular acceleration were given as biofeedback information over a speaker through signal processing and voltage to frequency conversion on PC. The throwers used this signal as an additional virtual sensor detecting important skill for throwing (Fig.5).
212 The Engineering of Sport 7 - Vol. 1
Figure 5 - Biofeedback system for hammer throwing.
4- Results The data of the angular velocity and angular acceleration measured by accelerometers were compared with one measured by gyro sensors to evaluate the accuracy of accelerometers. The sensor board, which were mounted accelerometers, and two gyros (Murata GYROSTAR), which detected angular velocity x and y, are attached on a hammer wire. The hammer was turned several times without releasing of hammer and sensors data was sampled at 400 Hz and logged. The angular acceleration and the centrifugal acceleration were calculated by, equation (6) using the accelerometers. Fig.6 shows the data of angular acceleration Yx, which compose major rotation of this test movement, and of centrifugal acceleration 2xy. Numerical differentiations of the gyro signal were computed to evaluate the angular acceleration measured by accelerometers. This result shows the high accuracy of those sensors unless both sensors have same error.
Figure 6 - Sensor board and wireless data-logger.
5- Conclusion Dynamics-model-based force sensor using accelerometers which measures forces and joint torques has been developed. In this study we have applied this method to hammer throw training aid integrating small sensors, signal processing, short-range wireless transmission, wearable data-logger, and biofeedback training system. During throws the
The Engineering of Sport 7 - Vol. 1
213
throwers are hard to realize angular acceleration of hammer, which mainly affects the speed of the hammer head, in force level because the centrifugal acceleration composes most of the wire tension. Without the feedback of the information, throwers may exert effort without the effect. To realize efficient sports training, presenting the information, which athletes can’t detect by themselves, through virtual sensory organ might be useful for athlete. As a device which measures angular movement, gyro sensors are commonly used. However, measuring range for consumer gyro sensors are generally restricted up to ±300 deg/s. It is in general too small for measuring sports movement. Accelerometers provide enough range for measuring sports movement on the contrary and give a suitable data set, in particular, for inverse dynamical calculation online. Using this method it also realizes the force display devices, which are able to make virtual reality environment.
6- References [A1] Analog Devices (2005) ADXL278 Data Sheet [Online], Available: http://www.analog.com/en/prod/ 0,,764_800_ADXL278%2C00.html [AS1] H. Ando, M. Sugimoto and T. Maeda (2004) Wearable moment display devices for nonverbal communications, IEICE Transactions on Information and Systems, E87-D, 1354-1360. [BK1] C.V.C. Bouten, K.T.M. Koekkoek, M. Verduin, R. Kodde and J.D. Janssen (1997) A Triaxial accelerometer and portable data processing unit for theassessment of daily physical activity, IEEE Transactions on Biomedical Engineering , 136-147. [BR1] Anton J. van den Bogert, L. Read and B.M. Nigg (1996) A method for inverse dynamic analysis using accelerometry, J. Biomech. 29, 949-954 [MB1] J.A. Mercer, N.E. Bezodis, M. Russell, A. Purdy and D. DeLion (2005) Kinetic consequences of constraining running behavior, Journal of Sports Science and Medicine 4 144-152. [MS1] K. Murofushi, S. Sakurai, K. Umegaki and K. Kobayashi (2005) Development of a system to measure radius of curvature and speed of hammer head during turns in hammer throw, Int. J of Sport and Health Science 3, 116-128. [MW1] H. Montoye, R. Washburn, S. Servais, A. Ertl, J. Webster and F. Nagle. (1983) Estimation of energy expenditure by a portable accelerometer, Med Sci Sports Exerc., 15, 403-407. [S1] STMicroelectronics (2006) LIS3L06AL Data Sheet [Online], Available: http://www.st.com/stonline/ products/literature/ds/11669/lis3l06al.htm [YY1] H. Yano, M. Yoshie and H. Iwata (2003) Development of a non-grounded haptic interface using the gyro effect, Proc. the 11th Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems (HAPTICS 2003) 32-39. [ZL1] B. Zappa, G. Legnani, A. Bogert and R. Adamini (2001) On the number and placement of accelerometers for angular velocity and acceleration determination, Transaction of the ASME, 123, 552-554.
Influence of Pedal Foot Position on Muscular Activity during Ergometer Cycling (P39) Stefan Litzenberger1, Sandrina Illes1, Martin Hren1, Martin Reichel1, Anton Sabo1
Topics: cycling, biomechanics, muscular activity. Abstract: Among cyclists it is a matter of common knowledge that the axis of the pedal has to be placed right beneath first metatarsal head at the forefoot (metatarsophalangeal position). Thus plantarflexion of the foot provides forces that result in evenly distributed cranktorque. Recent research has shown that even top cyclists neither deliver constant crank-torque nor flex the ankle excessively. Considering these facts and making biomechanical considerations it seems possible that much of the muscular output of the lower limb is used exclusively to stabilize the ankle. Torque in the ankle joint is - due to the long lever arm - fairly high. Reducing this lever arm by placing the pedal beneath the metatarsus (tarsometatarsal position) reduces torque in the ankle-joint and thus reduces muscular activity in the lower limb. The present paper examines changes in muscular activity of eight selected muscles of the leg by changing the pedal-foot-position from metatarsophalangeal to tarsometatarsal. Measurements were conducted on a bicycle ergometer (Daum, GER) at different power-outputs (75W, 150W) at two different pedal-foot-positions (metatarsophalangeal, tarsometatarsal) (N=12, non-cyclists). Muscular activity was recorded using an eightchannel-EMG-System (Noraxon, USA). Changes in muscular activity between different power outputs and pedal-foot-positions are analysed. Results are that metatarsal pedalling reduces muscular effort for calf muscles up to more than 20%. Activation durations of thigh muscles do not channge significantly, frontal shank muscle activation duration rises remarkably. This and further research could ultimately result in the development of a new cyclingshoe construction. Keywords: cycling, pedal-foot-position, muscular activity, EMG, cycling-shoe.
1- Introduction Cycling is all about transmitting human power to the road and thus produce speed. In this process it is aimed to use the human power produced with the maximum possible efficiency. Thus bicycles are lightweight, tires have low rolling resistance and aerodyna1. Fachhochschule Technikum Wien University of Applied Sciences, Höchstädtplatz 5, A-1200 Vienna - E-mail: litzenberger, illes, hren, reichel, [email protected]
216 The Engineering of Sport 7 - Vol. 1 mics are optimized. But one of the major points in being more or less efficient is the application of force to the pedal. As only force that is applied normal to the crank produces torque - and therefore mechanical power (Gressmann 2002)- cyclists usually aim to concentrate on appplying torque relevant force during the whole crank-cycle – so called “circling”. They even try pulling on the pedal during the upstroke. This is widely considered the most efficient way of pedaling. Research has shown that even elite-cyclists do not show positive work during the upstroke (Wilson 2004, Henke et al. 2001, Hillebrecht et al. 1998) but apply negative forces on the pedal. Korff et al. (Korff et al. 2007) have shown that there are differences in mechanical and metabolic efficiency. While “pulling” provides high mechanical efficiency, metabolic efficiency is lowest. Gross efficiency improves for pushing (more force in downstroke) as well as “circling” and is best when subjects use their preferred style of cycling. Broker (Broker 2003) assumed that more muscular work would be needed to eliminate negative forces. To overcome the turning points on the bottom and top of the crank cycle plantarflexion of the ankle is widely used. Previous research (Litzenberger et al. 2005) has shown that plantarflexion is lowest in higher workloads and does not differ greatly between metatarsophalangeal and tarsometatarsal position. It is concluded that muscular activity in the lower limb does not contribute to crank torque but mainly stabilizes the anklejoint. Reducing the lever arm of the foot (pedal axis to ankle joint) by placing the pedal closer to the ankle joint will reduce joint-torque and thus reduce muscular activity in the calf-muscles (m. gastrocnemius, m. soleus). Assuming that every active muscle is consuming oxygen the reduction of the number and the reduction of duration of activity of the muscles used, could result in less metabolic effort for a given workload. The present paper concentrates on the question whether muscular activity in the lower limb is different for forefoot or metatarsal pedalling and its consequences.
2- Methods Twelve healthy male subjects (N = 12, age: 25.2 ± 4.2 years, weight: 76.2 ± 7.6kg, height: 180.1 ± 4.5 cm, non cyclists1) performed a test on a Daum Ergobike 8008i bike-ergometer (daum electronic, Fürth, GER). Each test consisted of four trials each. Therefore a total of 48 trials was analysed. Subjects did four runs alternately using forefoot pedalling and metatarsal pedalling. The first two runs were performed at a workload of 75 Watt, runs three and four were performed at 150 Watt. Subjects were advised to use their preferred pedaling style and a cadence of 70 rotations per minute.
1. non-cyclists: only recreational or commuting cycling of less than three hours per week
The Engineering of Sport 7 - Vol. 1
217
Figure 1 - adidas Voltage MTB shoe with cleat in forefoot and metatarsal position.
To provide rigid connection between cycling shoe and pedal Time Atac system pedals (Time Sport International, Vauzelles, FRA) were used in combination with adidas Voltage MTB-cycling-shoes (adidas, Herzogenaurach, GER). The shoes were equipped with two Time cleats, one in the forefoot and one in the metatarsal position (Figure 1). Eight muscle groups of the right leg (Table 1) were equipped with surface electromyography-electrodes according to the recommandations of the SENIAM-project (SENIAM 2007). The muscle groups observed are the most important muscle groups for cycling and most of them are commonly used in Surface Electromyography (SEMG) studies in cycling (Ericson et al. 1985, Strunz and Wolf 2003). Data was acquired with a Noraxon Telemyo IV unit (Noraxon, Scottsdale, USA) at a sampling frequency of 1000Hz. SEMG cables were fixed with medical adhesive tape to minimize movement artefacts. The cycle-ergometer was additionaly equipped with a reed-sensor on the top position of the crank-cycle and a magnet on the crank. Thus on every passing of the crank in the top-positon an analog signal was sent and recorded simultaneously with EMGdata. Later on this signal was used as a trigger-signal to separate single crank rotations.
2.1 Data processing Data was processed using a specially designed Matlab-application (TheMathWorks, Natick, USA). Raw EMG-Data was imported for each run, rectified and filtered using a forward and reverse moving average filter (window size = 40ms) with no time shift. Afterwards data was divided into single rotations by the use of the trigger signal mentioned above. Single rotations were then interpolated to a length of 360 datasets (0 to 359) each. Each data representing one degree of the crank-cycle (0 deg : top position, 90 deg: forward horizontal position, 180 deg : bottom position, 270 deg : backward horizontal position). Mean of the signals and standard deviation were calculated for each subject. Moreover a total of 32 correlation coefficient matrices (eight muscles, four trials) were calculated for every single of the twelve subjects representing the correlation of the
218 The Engineering of Sport 7 - Vol. 1 EMG-signals of every single rotation. A mean correlation coefficient was then calculated for every muscle and trial resulting in a total 32 correlation coefficients. Mean and standard deviation of the correlation coeefficients for each muscle over all four trials were calculated to identify intra-individual variability of the EMG-signal. Table 1 - Muscles measured with SEMG and position of the Motopoint region (used for electrode placement)
For calculating the activation pattern of every muscle maximum and minimum amplitude for each mean-amplitude was calculated representing a status of maximum and no activation during one crank-cycle respectively. EMG-amplitude was expressed in percent of maximum amplitude. Mean EMG-data of all subjects was then calculated for every trial. Threshold was calculated via percentage of local EMG maximum. Threshold percentage was set to 30 percent of maximum EMG-Amplitude, minimal subperiod duration to 5 degree of crank cycle (Konrad 2005). To obtain the activation length the same approach to threshold as in the calculation of the activation pattern was used. Activation length was calculated for every single subject by its mean amplitude. Data was then averaged for all twelve subjects and standard deviations were calculated.
3- Results Results show different activation lengths and partly different activation patterns for forefoot and metatarsal pedalling. EMG-amplitudes and activation patterns show high similarity to former research (Ericson et al. 1986, Stapelfeldt and Mornieux 2005). Data showed that there was high intra-individual correlation betweeen EMG-signals. However correlation coefficients for EMG-data of muscles of the lower leg were higher than for thigh muscles. Maximum correlation coefficient was reached for m. gastrocnemius medialis (0.896 ± 0.011), lowest for m. rectus femoris (0.737 ± 0.024). Activation pattern for muscles of the thigh do not differ greatly between workloads of 75 Watt and 150 Watt respectively. There are differences to be observed in activation patterns between forefoot and metatarsal pedalling (Figure 2). Especially m. tibialis anterior shows a difference. It is activated longer. Activation does not end until about 40 degrees in metatarsal pedalling, whereas in forefoot pedalling activation ends at about zero degrees (uppermost point of crank cycle). All other muscles show similar patterns as well for forefoot and metatarsal pedalling as for different workloads.
The Engineering of Sport 7 - Vol. 1
219
However changes are obvious for activation length of different pedal-foot-positions (Table 2). Muscles of the thigh show slightly less or nearly equal activation length for metatarsal pedalling compared with forefoot pedalling due to high standard deviations. These results are not statistically significant. A tendency for shorter activation is observed for m. gluteus maximus and m. semitendinosus. Mucles of the shank show different effects due to change of pedal-foot position. Whereas m. tibialis anterior shows greater activation length in metatarsal pedalling, muscles of the calf (m. gastrocnemius medialis, m. soleus) are activated shorter.
Figure 2 - Activation patterns for all four trials (upper left : forefoot 75 W, upper right : metatarsus 75 W,lower left : forefoot 150 W, lower right : metatarsus 150 W) of muscles (from bottom to top: 1 : m. gluteus max., 2 : m. biceps fem., 3 : m. semitendinosus, 4 : m. rectus fem., 5 : m. vastus med., 6 : m. tibialias ant., 7 : m. gastrocnemius med., 8 : m. soleus)
220 The Engineering of Sport 7 - Vol. 1 Table 2 - Mean activaton length of all eight muscles in percent of crank cycle in four trials and differences between forefoot and metatarsal pedalling percent of forefoot activation length (FF = forefoot pedalling, MT = metatarsal pedalling). Negative numbers indicate shorter activation length in metatarsal pedalling
Figure 3 - Activation length for all eight muscles during all four trials in percent of crank cycle. Bold lines represent mean value, grey boxes standard deviation
The Engineering of Sport 7 - Vol. 1
221
4- Discussion and Conclusion Results indicate a reduction of the activation length of the calf-muscles (m. gastrocnemius medialis, m. soleus). For most muscles of the thigh (m. gluteus maximus, m. biceps femoris, m. semitandinosus, m. rectus femoris) no definite statement can be made due to high standard deviations. It is especially noticeable that activation length of the frontal shank muscle (m. tibialis anterior) is considerably higher in metatarsal pedalling. Activation patterns do not change greatly between forefoot and metatarsal pedalling. Solely m. tibialis anterior changes its activation pattern. Instead of being activated only in the last fourth of the crank cycle in forefoot pedalling, it shows additional activation in the first phase of the downstroke (until about 40 degrees). As there are as well positive as negative effects on different muscles a conclusion whether one pedalling style is advantageous in terms of reduction of total muscular effort and thus metabolic effort can not be drawn. Data shows that there is no significant difference in the activation patterns of the muscles indicating either that no adaption by the subject to the new pedal-foot-position has occurred or that no adaption is necessary. It can be assumed that due to less muscular effort in the shank muscles a metatarsal pedalling position might be advantageous for triathlon where a cycling split is followed by a running split and less strained calf-muscles could result in a better running performance. However further research especially concentrating on metabolic changes due to pedal-foot-position seems inevitable.
5- References [B1] Broker, J. Cycling Biomechanics: Road and Mountain. in: Burke, E., (ed.): HighTech Cycling, Human Kinetics, p. 147-175, 2003. [EN1] Ericson M.O., Nisell R., Arborelius U.P., Ekholm J. Muscular Activity during Ergometer Cycling; Scandinavian Journal of Rehabilitation Medicine, 17, 53-61, 1985. [G1] Gressmann, M. Fahrradphysik und Biomechanik. Kiel, 7. Ausg., 2002. [HM1] Henke, T., Monfeld, C., Heck, H. Trettechnik – Einzelzyklusdarstellung im Radsport; BISp-Jahrbuch, p. 127-144, 2001. [HS1] Hillebrecht, M., Schwirtz, A., Stapelfeldt, B., Stockhausen, W.; Buehrle, M. Tritttechnik im Radsport: Der “runde Tritt” - Mythos oder Realität? Leistungssport, Deutscher Sportbund, Frankfurt, pp. 58-62, 1998. [K1] Konrad P. EMG-Fibel, Version 1, 2005, http://www.badsassendorf.de/ generator.aspx/property=Data/id=113064/EMG__Fibel.pdf, 10.12.2007 [KR1] Korff, T., Romer, L., Mayhew, I., Martin, J. Effect of Pedaling Technique on Mechanical Effectiveness and Efficiency in Cycling. Medicine & Science in Sport & Exercise, 39 (6), p. 991-995, 2007. [LF1] Litzenberger, S., Fechter, A., Balac, M., Pfeifer, M., Reichel, M., Sabo, A. Einfluss der Pedalfußposition auf ausgewählte Parameter beim Mountainbiken. in: K. Witte et al. (ed.) Sporttechnologie zwischen Theorie und Praxis IV. Shaker, Aachen, p. 363-375, 2006.
222 The Engineering of Sport 7 - Vol. 1 [S1] SENIAM, http://www.seniam.org/, visited 20.12.2007 [SM] Stapelfeldt, B., Mornieux, G. Biomechanik im Radsport. Sportorthopädie Sporttraumatologie, 21, p. 107-114, 2005. [SW1] Strunz, J., Wolf, R. (2003): Stationäre und mobile Untersuchungen zu Muskelaktivitäten und zur Kinetik der Tretbewegungen bei Hochleistungsradsportlern, BISp Jahrbuch, 299-306 [W1] Wilson, D.: Bicycling Science., MIT Press, London, 2004.
Accurate Trajectory and Orientation of a Motorcycle derived from low-cost Satellite and Inertial Measurement Systems (P42) Adrian Waegli1, Alain Schorderet2, Christophe Prongué2, Jan Skaloud1
Topics: Differential GPS/MEMS-IMU integration performance, trajectory analysis, lateral slipping of tires, determination of characteristics of tires. Abstract: Inertially aided satellite positioning can bring its benefits to all disciplines in which detailed knowledge of the trajectory is a prerequisite for improving performance. In motorcycling for instance, the determination of slips of tires requires the determination of the precise trajectory and the orientation of the motorcycle’s chassis. The correct exploitation of torque or force sensors as well as studies of the vibratory behavior of pneumatics necessitate the knowledge of the orientation of the sensors. Accurate position and orientation can be obtained by integrating inertial measurement units (IMU) with GPS (Global Positioning System). Unfortunately, the traditional, bulky and expensive high-quality GPS/IMU instrumentation is restricted to few disciplines with higher accuracy demands, while the ergonomic constraints of some sports (e.g. ski racing, motorcycling) urge to use devices based on mono-frequency differential GPS and Micro-Electro-Mechanical System (MEMS) inertial technology. Due to their small size, low cost and power consumption, MEMS sensors are suitable for trajectory analysis in sports where ergonomic aspects play an important role. In this article, an experimental low-cost differential GPS/MEMS-IMU system is applied in motorcycling. The system provides an absolute positional accuracy better than 0.5m, velocity estimates accurate to 0.2m/s and an orientation accuracy of 1-2°. Keywords: DGPS, MEMS-IMU, motor cycling, torque sensors, slip of tires.
1- Introduction Satellite-based positioning provides highly accurate position and velocity. It has already proven its effectiveness in car racing (How, Pohlman et al. 2002), rowing (Zhang, Grenfell et al. 2003) and Alpine Skiing (Skaloud and Limpach 2003). Combined with inertial navigation systems (INS), acceleration and orientation can be observed additio1. École Polytechnique Fédérale de Lausanne, Switzerland, Geodetic Engineering Laboratory E-mail: adrian.waegli,jan.skaloud}@epfl.ch 2. École Polytechnique Fédérale de Lausanne, Switzerland, Mechnical Systems Design Laboratory E-mail : alain.schorderet,christophe.prongue}@epfl.ch
224 The Engineering of Sport 7 - Vol. 1 nally. Traditional GPS/INS equipment consisting of dual-frequency GPS receivers and tactical-grade INS provides high accuracies even for large dynamics (cm for position, cm/s for velocity and 1/100° for orientation). However, it is bulky (a few kg) and expensive (> 40’000€) and is therefore unsuitable for many sports. Recently, smaller and lighter equipment consisting of low-cost MEMS triple axis accelerometer and gyroscopes together with inexpensive L1 GPS receivers were introduced by (Waegli and Skaloud 2007). As it has been shown, the combination of these sensors helps to overcome the lack of continuity in the reception of the GPS signals in obstructed environment and to determine accurately the orientation of the MEMS-IMU sensor. In this article, the design of the GPS/MEMS-IMU system is described. Then, its performance is illustrated based on an experiment where the system was installed on a motorcycle. In the second part of the paper, applications of motorcycling are highlighted where the use of such a system can bring considerable benefits. Firstly, an example of trajectory analysis is given. Secondly, the computation of the lateral slipping of a motorcycle based on the GPS/MEMS-IMU trajectory is explained. Lastly, it will be shown how GPS/INS derived parameters can be exploited in order to characterize the performance of tires.
2- GPS/MEMS-IMU Integration Performance MEMS-IMUs are subject to large random and systematic errors (biases, scale factors, misalignment, and noise) which need to be suppressed in order to provide accurate orientations. For instance, a typical bias of 0.5m/s2 of a MEMS accelerometer would result in positioning errors of 50m after 10s. Representative MEMS gyroscope biases are up to 1°/s and would lead to orientation errors of 10° after only 10s. A solution for calibrating these errors consists in the integration of MEMS-IMU with satellite positioning. However, the conventional GPS/INS integration strategies (Titterton and Weston 1997) need to be slightly adapted in order to take into consideration the error characteristics of the MEMS-IMU sensors (Waegli, Skaloud et al. 2007). A motorbike was equipped with MEMS-IMU (xsens MT-i) which is rigidly fixed to the GPS antenna. A low-cost mono-frequency receiver (u-blox AEK4) was used together with a dual-frequency GPS receiver from Javad for reference. The results presented below integrate differential L1 GPS solutions of low-cost receivers at 1Hz with the triple-axis accelerometer and gyroscope measurements of the MEMS-IMU which are at 100Hz.
Figure 1 - GPS/MEMS-IMU system setup.
The Engineering of Sport 7 - Vol. 1
225
The performance of the low-cost L1 GPS/MEMS-IMU system was investigated by (Waegli and Skaloud 2007) in Alpine skiing by comparing its performance to a higher order GPS/INS system. The mean accuracies are summarized in table 1. The system parameters converge rapidly which permits bridging of GPS data gaps of up to 10s with inertial navigation without loss of accuracy. Using dual-frequency receivers would increase the position accuracy to decimeter-level while velocity and orientation quality would remain at a similar level. Table 1 - System setup and mean accuracy of the presented differential GPS/MEMS-IMU system.
3- Applications to Motorcycling Accurate position, velocity, acceleration and orientation data is crucial in many domains in motorcycling. In the sequel, applications where the GPS/MEMS-IMU system provides innovative approaches are described.
3.1 Trajectory Analysis The precise trajectory allows visualizing and comparing any parameters related to the performance (Waegli and Skaloud 2007). This may be quantities derived from GPS/MEMS-IMU trajectory (e.g. velocity, acceleration, tire slips) or parameters related to the motorcycle (e.g. throttle, suspensions) or to the athlete (e.g. heart-rate). Figure 6 gives an example where the lateral slipping of the back wheel of the motorbike is visualized on two turns of the track.
3.2 Solution to the Reference Frame Problem When studying the motorcycle performance, various parameters are determined in different reference frames, e.g. in the reference frame fixed to the motorcycle (abbreviated with b for body) and in the reference frame fixed to the track (l for local-level frame, figure 2). Measurements (e.g. force or torque) can be converted between the two frames thanks to the rotation matrix Rlb which expresses the orientation of the motorcycle with respect to the track and which is estimated in the GPS/MEMS-IMU integration. xl = Rlb • xb where x stands for observations in either frame.
(1)
226 The Engineering of Sport 7 - Vol. 1
Figure 2 - Illustration of the reference frames.
3.3 Computation of the Lateral Slipping of Tires The lateral slipping of tires can be observed directly with the use of GPS/INS derived results. First, the direction of the trajectory can be derived from the velocity components vNorth and vEast (Figure 2) : v tan() = North vEast
(2)
Furthermore, the GPS/INS integration yields the heading hd of the motorcycle. Combining the two variable leads to the slip of tires :
= hd –
(3)
Figure 3 and figure 4 illustrate the slip angle with respect to the throttle, traction/braking torque at the rear wheel, the roll angle and the lateral acceleration respectively. The confidence level (1 1°) highlights the accuracy of the slips. The presented experiment was conducted in winter and the dynamics were correspondingly small. Nevertheless, statistically significant drifts were observed during the turns. The lateral acceleration is consistent with the roll angle. It can be noted that the motorcyclist was inclined even on a great portion of the straight lines (Figure 5). As seen from figure 4, this inclination was compensated by a lateral acceleration.
The Engineering of Sport 7 - Vol. 1
227
Figure 3 - Throttle with slip angle and accuracy indicator (1). The black vertical lines indicate the beginning of the turns.
Figure 4 - Torque, lateral accelerations and roll during the same period as figure 3. The black vertical lines indicate the beginning of the turns.
Figure 5 - Lateral slipping angles visualized on the GPS/MEMS-IMU derived trajectory.
228 The Engineering of Sport 7 - Vol. 1
3.4 Evaluation of the Characteristics and Performance of Tires The “magic formula tire model” defined by (Pacejka and Bakker 1992) provides a mathematical expression from which forces and moments acting longitudinally or laterally from the road on the tire can be related to its slip performance: y(x) = D • sin(C • arctan(B • x – E • (B • x – arctan(B • x)))) Y(X) = y(x) + y x = X + x
(4)
where – X represents the lateral slipping angle or the longitudinal slip . – Y stands for the lateral force Fy, the aligning torque Mz or longitudinal force Fx. – B, C, D, E, x and y are constant coefficients.
Figure 6 - Typical tire characteristics indicating the meaning of some coefficient of equation 4.
The constant coefficients are usually determined by laboratory experiments (on drum). The GPS/MEMS-IMU system permits the calibration of these deterministic parameters in situ. First, force and torque measurements need to be referenced with respect to the road which becomes possible due to the orientation determined by GPS/INS. The lateral slip angle is directly computed from GPS/INS and needs only to be corrected for the steering of the front wheel, the tire radius variation due to speed, load and roll angle as wells as the suspension pitch which can be measured directly by means of linear potentiometers. The longitudinal slip requires the knowledge of the longitudinal velocity (vsx, from GPS/INS) and of the longitudinal velocity of the tire at the contact patch (vx, from digital speed sensors, which must be compensated for the tire radius variation ): v k = – vsx x
(5)
Hence, combining all these measurements permits to characterize tires in the field. This calibration reflects the actual characteristics of the surface (temperature and road roughness) and therefore refines laboratory findings. The peak value of is often situated
The Engineering of Sport 7 - Vol. 1
229
at approximately 10° of slipping and slipping angles up to 30° can be modeled. In this range, the accuracy of the slipping angles accounts therefore for 3-10% of the error, whereas the torque and force measurements can be reached with a level of 5%.
3.5 Tire related vibrations Attitude determination also provides essential data to the motorcycle vibration analysis. Indeed, tire-related vibrations are today very important in motorcycle racing. The socalled “chattering” vibrations may appear at the curves’ entry, apex or exit and enforce the rider to reduce its speed. To solve this problem, the phenomenon is studied by means of numerical and experimental approaches (Duvernier, Fraysse et al. 2002). One approach consists in using hybrid eigenvector bases to perform a modal synthesis (Schorderet and Gmür 2004) (Schorderet 1997). The vibratory behavior of a given tire is depending on two external parameters: ground load (vertical) and roll angle. To build the modal basis, a laboratory experimental modal analysis provides the eigenfrequencies and eigenmodes for discrete values of these parameters. Then, vibration measurements are realized on the track and dedicated software is applied to identify the frequencies where chattering exists (rider data). The hybrid model is then used to reduce or cancel the chattering vibrations. The efficiency of this predictive model is depending on the experimental data quality. The use of the low-cost L1 GPS/MEMS-IMU system combined with a force measurement unit (body reference frame) allows determining accurate roll angles and evaluating the vertical force in the local reference frame. These two parameters are required for the definition of the precise dynamic conditions under which the vibrations appear.
4- Conclusion A low-cost GPS/MEMS-IMU system was presented which is suitable in terms of cost and ergonomy for deployment in motorcycling. The system continuously tracks the 3Dtrajectory of the motorcycle which allows monitoring and comparing a large number of parameters related to the performance. Any performance parameter can be represented with respect to the trajectory and compared to subsequent turns or laps. Its accuracy (50cm in position, < 0.2m/s for velocity and 1-2° for orientation) opens new possibilities in analyzing many factors related to motorcycle performance. The system provides orientations which enable the computation of the lateral slipping angles in relation to the track and measurements of the motorcycle (e.g. force, torque measurements). Combining all this data enables the calibration of tire models in situ.
5- Acknowledgement This research is financed by TracEdge, based at Grenoble, France. The experimental tests have been performed using torque sensors provided by NRCtech, based at EPFL Science Park, Lausanne, Switzerland.
230 The Engineering of Sport 7 - Vol. 1
6- References Duvernier, M., P. Fraysse, et al. (2002). Tyre Modelling for NVH Engineering in ADAMS. 1st MSC.ADAMS European User Conference, London. How, J., N. Pohlman, et al. (2002). GPS Estimation Algorithms for Precise Velocity, Slip and Racetrack Position Measurements. SAE Motorsports Engineering Conference & Exhibition. Pacejka, H. B. and E. Bakker (1992). “The magic formula tyre model.” Vehicle System Dynamics International Journal of Vehicle Mechanics and Mobility 21: 1-18. Schorderet, A. (1997). Synthèse modale et problème inverse en dynamique des structures. Lausanne, EPFL. PhD: 1698. Schorderet, A. and T. Gmür (2004). “Structural dynamics optimization based on a hybrid inverse synthesis method using a quadratic approximation.” ASME Transactions : Journal of Vibration and Acoustics 126(2): 253-259. Skaloud, J. and P. Limpach (2003). Synergy of CP-DGPS, Accelerometry and Magnetic Sensors for Precise Trajectography in Ski Racing. ION GPS/GNSS 2003, Portland. Titterton, D. H. and J. L. Weston (1997). Strapdown inertial navigation technology, Peter Peregrinus Ltd. Waegli, A. and J. Skaloud (2007). Assessment of GPS/MEMS-IMU Integration Performance in Ski Racing. ENC, Geneva, Switzerland. Waegli, A. and J. Skaloud (2007). “Turning Point – Trajectory Analysis for Skiers.” InsideGNSS(Spring 2007). Waegli, A., J. Skaloud, et al. (2007). Assessment of the Integration Strategy between GPS and Body-Worn MEMS Sensors with Application to Sports. ION GNSS, Fort Worth, Texas. Zhang, K., R. Grenfell, et al. (2003). Towards a Low-Cost, High Output Rate, Real-Time GPS Rowing Coaching and Training System. 16th International Technical Meeting of the Satellite Division of The Institute of Navigation, Portland.
Wireless Impact Measurement for Martial Arts (P43) J.I. Cowie1, J.A. Flint1, A.R. Harland1
Topics: Apparel; Measurement Systems Abstract: The integration of electronics into performance sports equipment has increased in recent years and with it the facility to provide more accurate judging, improved coaching, participant health monitoring and to enhance the entertainment of spectators. The focus in this paper is to respond to recent interest in various martial arts in quantifying and categorising impacts which occur during competition. An impact measurement technique has been developed which incorporates a non-invasive sensor system into a body protector worn in Taekwondo. The demonstrator system is integrated into the garment and utilises Bluetooth technology to transmit sensor readings back to a computer for analysis. The sensors considered for impact measurement include thin film piezoresistive force and pressure sensors, and also much newer technologies such as MEMS (Micro-Electro-Mechanical Systems) accelerometers. Impact analysis has been carried out in both the frequency and time domain in order to determine the suitability of the different methods and the bandwidth requirements. The key metrics of force, pressure, velocity and impact duration have been assessed. It has been shown by field experiments and by employing standardised test methods that impacts can be best characterised by making use of various types of sensor in combination. Bluetooth allows for 100 m range; however it was found that the limited bandwidth afforded makes real time data capture problematic unless the sampling rate, bit resolution or number of channels is kept relatively small. In conclusion, a highly compact measurement system has been realised which is capable of quantifying martial arts impacts. The Bluetooth standard is able to support the data transfer requirements, however development of a ‘smart’ on-garment sensor system would be the next logical step to manage a multi-sensor measurement system. This way the gathered data can be compressed or coded to reduce the transmission bandwidth. Keywords: Impact; Measurement; Piezoresistive; MEMS Accelerometers; Wireless; Bluetooth.
1. Loughborough University, Loughborough, England - Email: J.I.Cowie, J.A.Flint, [email protected]
232 The Engineering of Sport 7 - Vol. 1
1- Introduction In Taekwondo points are scored for a punch or kick to the thorax or head, these are recorded by four judges situated around the match arena, decisions are made purely on visual inspection and the officials judgement. The World Taekwondo Federation has identified the requirement for an electronic scoring system in the game to assist the match officials. As well as assisting officials, the integration of an electronic measurement system could be tailored to a number of other functions; to monitor the health of the players to determine if too many impacts have been received in the same location, and for the purposes of coaching to allow coaches to monitor technique, for example where players are dropping their guard or where they are not being the most effective during attacks. In order to achieve this instrumentation would be required in the protection equipment that is worn in matches and coaching sessions. This instrumented protective clothing would need to measure one or more of the key metrics of the impacts; force in the order of 2kN (Pierce et al. 2004), pressures of several hundred kPa, the velocity of the impact up to 10ms-1 (Gulledge, Dapena 2008) and displacements of tens of millimetres. This paper describes the measurement system for martial arts which incorporates piezoresistive sensors, MEMS accelerometers and a Bluetooth radio system. An overall system diagram is shown in Figure 1. The individual elements of the system will be discussed in the following sections. The problem can be broken down into a number of areas. A sensor technology is required to detect the impacts of kicks and punches. A microprocessor system needs to be developed to perform some signal processing on the sensor outputs and encode them for transmission. Radio frequency transceivers and antennas are needed to transmit the data from the protective equipment to a computer. Software also needs to be developed to take the data received, store them and present them for judges to make decisions.
Figure 1 - Block diagram of the system.
The Engineering of Sport 7 - Vol. 1
233
2- Sensor Technologies 2.1 Piezo Transducers Piezoresistive and piezoelectric force and pressure sensors have been used for some time (Millet et al. 1998, Bachus et al. 2006, Schmidt 2007) in sports applications (impacts, skiing and golf) to gather data generated by human interactions. Piezoresistive materials function by having a resistance which varies with the force which is applied to it. Piezoelectric materials generate a voltage across them when a force is applied to them, this voltage can be very large (much greater than the 10V maximum input to most analogue to digital converters). An instrumented punch bag with a piezoelectric strip to measure force has been used to demonstrate piezoelectric sensors. Figure 2 shows a voltage trace from a punch impact to the bag. This clearly demonstrates a known problem with the piezoelectric materials as the higher values have been clipped due to the capabilities of the data acquisition system.
Figure 2 - Voltage response from piezoelectric force sensor after a gentle punch.
2.2 MEMS Accelerometers One of the key metrics to be determined is velocity of the impact this can not be measured directly, and can not be determined from force pressure data. The use of accelerometers to measure soft tissue displacements has been shown (Boyer, Nigg, 2006). The use of Micro-Electromechanical Systems (MEMS) accelerometers has been studied to measure accelerations, velocities and displacements. These devices are becoming more accessible due to their mass marketing in commercial items such as games consoles and mobile phones. MEMS accelerometers operate using a mass spring system, as the device
234 The Engineering of Sport 7 - Vol. 1 is accelerated the mass’s inertia deforms the spring, this deformation is then measured. A punch bag has been instrumented with an accelerometer and various drop tests and punches have been carried out. An ADXL320 accelerometer from Analogue Devices has been used this is mounted with support circuitry on flexible PCB (Printed Circuit Board) substrate, Figure 3. A National Instruments USB data acquisition card is used to capture and digitise the data from the accelerometer. Figure 4 shows the resulting trace from an accelerometer during a punch impact.
Figure 3 - Accelerometer on flexible PCB attached to punch bag.
Figure 4 - Acceleration response of a gentle punch.
The Engineering of Sport 7 - Vol. 1
235 (1) (2)
From the acceleration an integration (Eq. 1) can be carried out to give a velocity profile and a double integration (Eq. 2) can be carried out to determine the displacement. Numerical integration methods were used to calculate velocity and displacement as can be seen in Figure 5. Errors have emerged in the velocity and displacement responses due to the constants of integration, the displacement has a falling response following c1t+c2. This problem is solved in post processing by the inclusion of a high pass filter with cut off frequency around one hertz.
Figure 5 - Velocity and displacement data from accelerometer.
3- Bluetooth Transmission The transmission from the garment is achieved using the Bluetooth transmission protocol operating in the ISM (Industrial Scientific and Medical) 2.4GHz license free band. The Bluetooth class 1 band offers a range of 100m utilising a low power requirement it also gives a bandwidth of 1Mb/s which can be enhanced to 3 Mb/s. For the purposes of this project a commercially available Bluetooth transceiver is used, a Roving Networks BlueSentry XP is used. The BlueSentry contains a RN-41 class 1 Bluetooth module for transmission, this is controlled by a Microchip PIC16F73 microcontroller. The BlueSentry also includes a Texas Instruments ADS8344 analogue to digital converter allowing the connection of sensors directly to the device. The analogue to digital converter allows for 8 channels to be used with a 16 bit resolution, this gives 4 sensors
236 The Engineering of Sport 7 - Vol. 1 differentially connected or 8 single ended sensors. Access to the device is given through the use of the Bluetooth Serial Port Profile, SPP. The SPP simulates an RS232 serial port giving the ability to send the device simple ASCII instructions to set up the analogue to digital converter and transmitter. Software has been written in visual basic to allow a Bluetooth enabled computer to interface with the BlueSentry, this software gathers data from the digital channel displays it graphically on screen and logs it to a comma separated values file with a time stamp for later offline analysis. The 8 channels available is a limitation of the Bluetooth protocol, the bandwidth offered will not allow for any more data to be transmitted continuously.
4- Integration In order to bring these systems together a method for integrating them with the personal protective equipment has been devised. This includes a study investigating where movement occurs in the human to determine the best location to place rigid electronics and to calculate the strains required to be withstood by the interconnections and transmission lines within the garment. The use of flexible printed circuit board substrates has been demonstrated with the accelerometers, further use of this material is required. Work has also been carried out on the use of conductive yarn to provide flexible interconnections in fabrics. A silver coated nylon thread is being used, this yarn can be sewn using a commercially available domestic sewing machine as well as by hand. It has a resistance of 3k per meter, this relatively high resistance can be lowered by stitching the same path multiple times, effectively increasing the number of strands in the interconnecting wires. The use of a very fine stitch pattern also helps in reducing the resistance, with these to methods a resistance of 200 per meter has been demonstrated. The integration of stitched wires into a fabric allows for the creation of a fully flexible circuit.
5- Conclusion A couple of methods of measuring impacts has been presented and a method for transmitting the data from the protective clothing to the judge has been outlined. Due to the limited bandwidth available from the Bluetooth protocol only a small number of sensors or an unacceptably low sampling frequency on the analogue signals is possible. To over come this problem a system of microprocessors with a certain level of intelligence will need to be developed to allow a large number of highly coupled sensors to operate simultaneously, whilst only relevant data is transmitted, this will reduce the amount of data which is needed to be transmitted while increasing the amount of data captured on the garment. With the use of multiple sensor types impacts in martial arts can be fully categorized and identified using the important metrics of force, pressure, acceleration, velocity and displacement. A highly capable wireless measurement system has been developed for use in Taekwondo or similar contact sports to assist in judging, coaching and health monitoring. It makes use of wireless technologies, micro-electromechanical systems, materials and microprocessor technologies. It was found that one limitation is the amount of information versus bandwidth and the need for a strategy to manage this information. There is also the possibility to use other sensor technologies not mentioned
The Engineering of Sport 7 - Vol. 1
237
here and the need to develop suitable antennas for use in a garment. Further work needs to be carried out to make the system robust enough to be used in match situations so that it can withstand all the movement and impacts that are associated with the sport.
6- References [BD1] Bachus, K. N., DeMarco, A. L., Judd, K. T., Horwitz, D. S. and Brodke, D. S. Measuring contact area, force, and pressure for bioengineering applications: Using Fuji Film and TekScan systems. Medical Engineering & Physics 28: 483–488, 2006 [BN1] Boyer, K. A. and Nigg, B. M. Soft tissue vibrations within one soft tissue compartment. Journal of Biomechanics 39: 645–651, 2006 [GD1] Gulledge, J. K. and Dapena, J. A comparison of the reverse and power punches in oriental martial arts. Journal of Sports Sciences 26(2): 189 – 196, 2008 [MH1] Millet, G. Y., Hoffman, M. D., Candau, R. B. and Clifford, P. S. Poling forces during roller skiing: effects of grade. Medicine and Science in Sports and Exercise 30 (11): 1637-1644, 1998 [PR1] Pierce, J., Reinbold, K., Lyngard B., Goldman, R., and Pastore1, C. Direct measurement of punch force during two professional boxing matches. Journal of Sport & Exercise Psychology 26(Suppl.); 150, 2004 [S1] Schmidt, E. R. Measurement of grip force and evaluation of its role in a golf shot. April 2007
7- Acknowledgments The authors would like to thank the co-investigators on this project Roy Jones and Richard Hague and also the IMCRC (Loughborough University Innovative Manufacturing and Construction Research Centre) who with EPSRC (UK Engineering and Physical Sciences Research Council) have funded it.
A Comparative Study of Ball Launch Measurement Systems; Soccer Case Study (P44) Jouni Ronkainen, Chris Holmes, Andy Harland, Roy Jones1
Topics: Measurement systems, Soccer. Abstract: The sports ball market is extremely competitive and in the US alone valued in excess of $1200 million [SG1]. In order to research and develop sport balls it is vital to quantitatively measure the launch and flight characteristics of the ball. Original equipment manufacturers (OEM’s) are currently using a wide range of systems to measure these parameters, allowing direct comparison between products. The purpose of this investigation is to compare some of the most methods currently available to measure soccer ball launch characteristics. The three measurement systems used were an optical system, a radar system and a high speed video (HSV). A detailed operational description of each method is provided which highlights system strengths and weaknesses. All systems were tested by assessing the launch characteristics of 30 kicks representing a maximal velocity strike, along with 30 impacts representing a curve kick. The kicks were carried out using a purpose built mechanical kicking simulator developed at Loughborough University (LU) in order to obviate inconsistencies achieved with player testing. The main findings from the work showed statistically that the optical system gave a higher soccer ball velocity, whereas the radar system gave a lower launch angle to the other systems. The measured spin rates for kicks highlighted the limitations of current measurement systems, due to discrepancies between all measured spin rate values. In summary this is the first comprehensive study to compare current soccer ball launch measurement systems using a highly repeatable kicking simulator. The study highlighted the uncertainties involved and particular attention was given to the fidelity of the spin measurement. Keywords: High speed video, launch characteristics, soccer, spin, measurement systems.
1- Introduction The sports ball market is very competitive; valued in excess of $1200 million in the US alone, comprising of £100 million in soccer ball sales [SG1]. In order to research and develop sport balls it is vital to quantitively measure the launch and flight characteristics 1. Sports Technology Institute, Loughborough Science & Enterprise Park, Loughborough, Oakwood Drive, Leicestershire, UK E-mail: J.A.Ronkainen, C.E.Holmes, A.R.Harland, [email protected]
240 The Engineering of Sport 7 - Vol. 1 of the ball. OEM’s are currently using a wide range of systems to measure these parameters, allowing direct comparison between products. It is fair to state that almost all soccer ball manufacturers have their own testing protocols in place to develop their latest range of soccer balls. Initially OEM’s are concerned with the ball achieving the FIFA ‘approved’ or ‘inspected’ insignia, however once they have achieved this status, they must design and develop their balls to outperform the competition. This study compares some of the most advanced methods currently available to measure soccer ball launch characteristics, these devices are essential in benchmarking balls so that future modifications can be objectively assessed against predecessors or competitor products.
2- Methodology The four main pieces of equipment used for this study are outlined in detail, highlighting the strengths and weaknesses of each device. The testing protocol is clearly defined.
2.1- Equipment 2.1.1- High Speed Video Camera The initial flight of the soccer ball, after impact with the mechanical kicking simulator, was captured using a Photron APX HSV camera. The camera was operating at a resolution of 1024 512 pixels, recording at 1000 fps with a shutter speed of 1/3000 seconds in order to improve the clarity of the image. A voltage output from the mechanical kicking simulator was used to trigger the camera at a known time interval, ensuring the capture of the entire ball flight. Figure 1 depicts a composite image of the soccer ball (20 ms apart) in flight during a straight kick (a) and curve kick (b). The strength of the system is that the HSV is a well established technique for launch measurement. Weaknesses include the manual nature of the digitisation process and using one camera only allows 2D analysis.
Figure 1 - Composite high speed images of soccer ball in flight, (a) straight kick, (b) curve kick.
2.1.2- Optical system This system, as shown in Figure 2, was purpose built for soccer ball launch characteristic measurement. The optical system which has previously been described in detail [NJ1 & NJ2], is an image based acquisition system. The device captures two images of the soccer ball; (1) an image of a static ball at approximately 1.25 metre stand off distance and (2)
The Engineering of Sport 7 - Vol. 1
241
an image of the ball, triggered by the impact sound, 6 ms after impact. The system calculates launch angle and velocity of the ball using elementary image processing techniques. The ball is marked with 50 dots; 25 blue and 25 yellow, the dots are arranged so that the relationship between each one is unique. It is inconsequential at what orientation the ball is viewed, avoiding the need to pre-align the ball. The spin rate is calculated by the system locating 7 dots on the ball, in each acquired image. In conjunction with the ball launch characteristics measurement, the system also captures the player’s kicking technique using a video acquisition device. The strengths of the system are that it can calculate the ball velocity, launch angle, spin rate and spin axis of the ball. The main weakness of the system is, in order to calculate the spin rate and spin axis, the fiducials must be present on the ball.
Figure 2 - The optical system.
2.1.3 Radar system The technology has descended from missile tracking applications, utilising Doppler radar technology, operating at 10.5 GHz bandwidth. The system is able to locate the ball in 3D space using the monopulse principle [R1], using one micro wave emitter and three receivers. A continuous measure is achieved, so that the position of the ball is located throughout the flight of the ball. The system is able to calculate the ball velocity, vertical and horizontal launch angle and spin rate. The lift and drag coefficients of the ball are plotted for the ball throughout flight. The system has been primarily designed for golf measurement, and this is the first time the radar based system is reported for soccer ball flight measurement. The prodigious strength of the system is that it is able to measure the ball throughout flight, the system is effortless to setup and once acquisition is initiated data is recorded indefinitely. A weakness in the system is due to the low spin rates achieved in soccer, the manufacturer claims that spin rates under 300 RPM are not demodulated accurately by the system. The system outputs total spin rate but spin axis is not defined.
242 The Engineering of Sport 7 - Vol. 1
2.1.4 Mechanical kicking simulator Athletes have been extensively used by manufacturers during the development and testing of sporting equipment. However athletes will tire during testing and therefore provide inconsistent performance. Mechanical simulators have been developed to obviate the issue arising from athlete fatigue and have been used since 1926. “Smart structures have been used increasingly over the last four decades to recreate consistent athlete/sports motions which offer increased repeatability” [H1]. Mechanical simulators are now used by manufacturers, governing bodies and academic institutes to provide accurate and repeatable simulations. In order to produce soccer ball impact conditions accurately and reliably, the simulator, as shown in Figure 3, consists of a simple leg rotation device, designed around a rigid A-frame, manufactured from 50 50mm steel box section. The adjustable ball teeing system is capable of manipulation about three axes, allowing the creation of different flight characteristics. The drive system uses a Lenze 6.9 kW asynchronous geared servo motor to rotate the kicking leg which is made from two aerospace grade aluminium plates. The leg impacter is capable of accelerating to a maximum velocity of 37 ms-1 in 270°. The machine’s software allows the length and rate of acceleration and deceleration to be specified, allowing different impact conditions to be achieved. The maximum ball velocity achieved during preliminary testing was 45ms-1, with the repeatability of the leg speed calculated to 0.032 ms-1 (2 SD).
Figure 3 - CAD images depicting the mechanical kicking simulator.
2.2- Protocol The purpose of this investigation was to compare three current measurement systems. All systems were tested simultaneously, using the mechanical kicking simulator impacting the ball at 16, 20 and 23 ms-1 leg velocity. A straight kick and curve kick were recreated by the mechanical kicking simulator by varying the position of the ball at impact, using the adjustable teeing mechanism, as shown in Figure 4. Five elite dot marked soccer balls, inflated to a pressure of 1 bar, were impacted a total of 10 times, for each leg velocity and kick type, with the initial launch characteristics measured using each system. Figure 5. shows the test and instrumentation arrangement. The radar and optical systems use a series of inbuilt algorithms within the analysis software to produce launch data, whilst the recording from the HSV was later digitised in order to allow comparison.
The Engineering of Sport 7 - Vol. 1
243
Figure 4 - Impact position for (a) straight kick, (b) curve kick.
Figure 5 - Plan view of testing protocol.
2.3 Analysis The HSV recording was manually digitised using Image Pro Plus, which determined the x and y pixel co-ordinate values for the centre of the ball at two positions a known time interval apart, by encompassing the circumference of the ball with a circle. The centre co-ordinate positions were used to calculate the ball velocity and elevation angle in the plane of the camera. The soccer balls were marked with a circumferential line which allowed the spin rate to be determined by analysing the number of frames taken to complete half a revolution. The uncertainty of the high speed video measurement Q Total can be defined using Equation 1. The uncertainty of the device Q Device was assumed to be very small by comparison to other uncertainties that exists within the measurement procedure. The
244 The Engineering of Sport 7 - Vol. 1 uncertainty of the setup Q Setup was attributed to the ball flight not being perpendicular to the camera placement. Q Setup was difficult to measure and was controlled by accurate alignment of the camera position, and mechanical kicking simulator. Uncertainty of the analysis Q Analysis can be described as a measure of repeatability. The repeatability of the analysis was calculated by analysing the same image a number of times and defined for velocity as ±1.01 ms-1 (95% confidence) and for launch angle ±0.14° (95% confidence). (1) Equation 1 - Uncertainty of the measurement procedure.
3- Results and Discussion The results were achieved successfully and the weather conditions were, dry and cool (10 to 13°C) throughout the testing protocol. Figure 6 - 7 display the mean ± 1 S.D for the velocity and elevation angle, measurements during straight kicks (a) and the curve kicks (b). Figure 8 displays each successful spin measurement taken during the entire testing protocol, for the straight kicks (a) and the curve kicks (b).
Figure 6 - Launch velocity measurements (average ± SD), (a) straight kick, (b) curve kick.
All the results in Figure 6 consistent indicated by low standard deviations, although a slight increase was observed for the curve kicks. The ball velocities are considerably lower for the curve kick than the straight kick, due to leg energy being transferred into rotational and translational energy. As the leg velocity of the robot increases, a consequent increase was observed in the ball velocity. A drastic reduction in measured velocity, was observed with the HSV measurement during the curve kick. This was accounted due to the HSV measuring 2D velocity, observed in Figure 1 (b), as the ball travels out of plane in the camera field of view. The optical system measured considerably higher velocities than the other two systems, which was due to the system acquiring two images; one static image and one dynamic image, therefore the impact point in time must be known precisely, otherwise errors are induced in velocity calculations.
The Engineering of Sport 7 - Vol. 1
245
Figure 7 - Launch angle measurement (average ± SD), (a) straight kick, (b) curve kick.
The HSV and optical systems recorded statistically similar results for the straight kick launch angle at the 95% confidence level, as shown in Figure 7. This was calculated using a one way ANOVA test and this was the only comparison to provide statistically similar results between the measurement systems. Similar results were achieved because both systems measured the launch angle in an identical way. It was noticeable that the curve kick results show a larger launch angle for the optical and HSV measurements, this was due to the alteration of the ball teeing position, in order to replicate a realistic curve kick. Due to the position of the radar system, it was not possible to measure the ball immediately after impact, which accounts for the low measured launch angle, as the ball was approaching its apex.
Figure 8 - Spin rate measurements, (a) straight kick, (b) curve kick.
Figure 8 highlights the fidelity of the soccer ball spin rate measurement and shows the results exhibit considerable variation. The radar system appears to be the worst performer regarding spin rate measurement, highlighted during the 20 ms-1 straight kick, where the spin rate measurement ranged from ~170 to 650 RPM. The HSV results show the most consistency, however spin rate was only measured around one axis, which was assumed to give a good representation for the straight kick. For the curve kick compound spin was placed on the ball and the HSV method will underestimate the spin rate. For this reason the optical system was expected to give the most accurate spin rate measurements for the curve kick. It must be noted that the systems all measured spin differently; the HSV measured spin around one axis only, the radar method measured total spin rate and the optical system measured total spin rate and spin axis.
246 The Engineering of Sport 7 - Vol. 1
4- Conclusions Uniform results were achieved for launch velocity and launch angle for all measurement systems. For comparative studies all the measurement systems give consistent results, therefore all systems could be used for ball dynamic measurements. However differences were observed between all systems, therefore users should be cautious when using the methods for benchmarking. If a new system is acquired for development purposes, it is recommended that a calibration test is carried out so that results from both systems can be compared. The test outlined here is recommended for this purpose. The only results between systems to show significant results were between the HSV and optical systems for straight kick launch angle measurement. The most difficult variable to measure was the spin rate, which was proven by the large variation in results. This work highlighted the need to develop more advanced spin rate measurement techniques with traceability. The optical system was proven to be a useful tool for soccer ball launch characteristics measurement, however this work recommends that both images captured for analysis are acquired post launch, as it would remove rotational and translational accelerations associated with the impact. This is the first comprehensive study to compare current soccer ball launch measurement systems using a highly repeatable kicking simulator. The results have shown the difficulty in measuring ball launch characteristics in a consistent and accurate manner, therefore an industry standard is required in order to calibrate new and existing measurement systems.
5- References [H1] Harper, T., 2007. Robotic simulation of golf swings. PhD thesis, Loughborough University, UK. [NJ1] Neilson, P., Jones, R., Kerr, D., and Sumter, C., 2004. An automated system for the measurement of soccer ball flight characteristics. The Engineering of Sport, 5, (2), pp. 180-6. [NJ2] Neilson, P., Jones, R., Kerr, D., and Sumter, C., 2004. An image recognition system for the measurement of soccer ball spin characteristics. Institute of physics publishing, Measurement Science and Technology, 15, pp. 2239-47. [R1] Rhodes, D., R., 1959. Introduction to monopoles. McGraw-Hill. [SG1] Sporting Goods Manufacturing Association, (SGMA)., 2007. Manufacturers sales by category report, US wholesale value of annual manufacturers’ shipments ($million).
Testing Protocol for Quantitative Comparison of Top of the Range Soccer Boots (P45) Jouni Ronkainen1, Dan Toon1, Joe Santry2, Tom Waller1
Topics: Soccer. Abstract: Soccer manufacturers are spending increased amounts of time and money developing their soccer boots. Claims such as more grip, more control and more speed have been used in advertising campaigns and no doubt will be used again in the future. The proof behind such claims is often not realistic to match play situations therefore this investigation set out to compare three top of the range branded boots and one pair of prototype boots using University level players to strike an instep swerving free kick. Subjective player feedback was given regards to the comfort and feel of each boot and how it felt to strike the ball. Objective results were obtained for each strike using a soccer launch monitor, used to measure the launch characteristics for each kick, allowing a direct measure of ball velocity, launch angle and spin rate. Player testing was used in order to achieve feedback on the boots, this presented an inherent problem of inconsistent strikes. Therefore ten repeats of each kick were carried out and no miss kicks were recorded. Since soccer is predominantly played outdoors, the weather conditions can drastically influence the players kicking ability. In wet conditions an aquaplaning effect is observed when the boot contacts the ball reducing the amount of grip between boot and ball. Therefore it was vital to test the boots in wet and dry conditions. The results obtained showed clear trends, in the wet conditions less spin and less velocity were imparted on the ball. No significant differences were observed between the launch characteristics measured between the different boots, suggesting that boots performed similarly, however distinguishing factors such as comfort, feel, aesthetics and branding were deemed very important by the players. Keywords: Soccer, soccer boots, launch characteristics, comparative study.
1. Sports Technology Institute, Loughborough Science & Enterprise Park, Loughborough, Oakwood Drive, Leicestershire, UK E-mail: J.A.Ronkainen, D.Toon, [email protected] 2. Progressive Sports Loughborough Science & Enterprise Park, Loughborough, Oakwood Drive, Leicestershire, UK E-mail: [email protected]
248 The Engineering of Sport 7 - Vol. 1
1- Introduction The governing body for soccer, the Fédération Internationale de Football Association (FIFA) has 208 national affiliated associations, consisting of more member states than the United Nations. FIFA contains confederations on 6 continents which act as umbrella organisations, thus the global presence of the game is undoubted. A key element of the game is the footwear that the player’s play in. In the US alone, in 2006, the wholesale value of soccer footwear was $296 million [SG1]. Soccer manufacturers are spending increased amount of time and money developing their soccer boots, in order to maintain competitive advantage. Claims such as more grip, more control and more power have been used in advertising campaigns and no doubt will be used again in the future. The proof behinf such claims is often not realistic to match play situations, therefore this investigation sets out to compare three top of the range branded soccer boots as well as a pair of novel prototype boots. It is not questioned that the sports industry is marketing orientated, therefore boot contracts with players such as Ronaldinho are lucrative for the players and the companies alike, however it is the performance aspect of boots that are investigated here.
2- Methodology Four boot types (A-D) from four different brands were selected for inclusion in the investigation. Three of the four boots were commercially available and considered market leading. Boot selection was based on retail price point and advice from players and coaches. An additional prototype boot was also included in the current investigation. The testing was split into two different categories: (1) a mechanical friction testing protocol was carried out in order to deduce if the grip of the material influences factors such as more control and more grip. (2) a player testing protocol was devised in order to achieve objective measurements of players striking the ball. The kick was designed to replicate a free kick scenario, considered a key set piece in soccer. The different testing protocols are outlined as follows.
2.1 Mechanical friction testing protocol Static and dynamic frictional resistance of the four soccer boot uppers (uppers A, B, C and D) were determined to assess the force necessary to initiate sample movement, and the force required to sustain sample movement respectively. A weighted sled with the attached boot sample was pulled a distance of 100 mm (BSISO: 15113-2005) across a surface constructed from a net form FIFA approved football (2007 Umbro England X ball). The ball was mounted upon the bed of a steel friction test rig. The force required to pull the sled and its displacement were recorded at 40 ms intervals using a Lloyd LRX constant rate extension machine, the resolution of measurement was stated at ± 0.02 Newtons. The friction testing setup is shown in Figure 1.
The Engineering of Sport 7 - Vol. 1
249
Figure 1 - Friction testing setup.
The boot samples were attached to the sled base using Cyanoacrylate adhesive, the front of the sample was rolled over the edge of the sled to reduce the amplified mechanical friction that would be expected from an edge interaction. A normal load placed upon the sled held the boot sample against the ball base during the test. Two displacement speeds were used during the testing, 50 mm·min-1 (BS-ISO: 15113-2005; speed 1) and 1000 mm·min-1 (speed 2). BS-ISO: 15113-2005 stipulates that a velocity of 1000 mm·min-1 (the maximum of the machine) may cause frictional heating consequently altering frictional properties. It must be noted that speeds of this magnitude and much higher are expected during matchplay. The friction tests were repeated in wet conditions; the surface of the ball was lightly sprayed with water from an atomised spray. Each individual test combination was repeated 5 times.
2.2 Player testing protocol Three players were selected to participate in the current investigation (age 23.7 ± 3.2 years, shoe size UK 9). The players selected had extensive competitive experience at University level and above. All testing was carried out on a long pile (65mm) third generation in-fill (sand and rubber) artificial pitch. Prior to testing a health screening questionnaire was completed and informed written consent was obtained in accordance with Loughborough University’s ethical advisory regulations. The test procedure was outlined to the players at the start of each test. The players were asked to perform a swerve kick from the instep of the foot. The kick was chosen since out of the 147 goals scored in the 2006 World Cup in Germany, 33% of the goals were scored from set pieces [AY1]. Set pieces are predominantly carried out using the instep kick. A typical instep kick consists of controlled components of spin and power,
250 The Engineering of Sport 7 - Vol. 1 but each athlete was encouraged to perform the kick in their own style to facilitate individual kicking reproducibility. The ball was positioned at a fixed point before each kick and targets were arranged in the field at fixed heights to promote consistent kicks with curved flight. The main marker was positioned so that when the ball travelled past it, the ball launch angle was circa 15°, in order to simulate a free kick going over the wall. The players wore their own boot on the supporting foot as a control boot, and the test boot on the kicking foot; this allowed for any differences in the outsoles of the boots to be eliminated. Soccer is predominantly played outdoors, the weather conditions can drastically influence the players kicking ability. In wet conditions an aqua planning effect is observed when the boot contacts the ball reducing the amount of grip between boot and ball. Therefore the test procedure was repeated for both wet and dry conditions. The wet conditions were simulated by lightly coating the kicking surface of the boot with a fixed amount of water using an atomising spray and dipping the ball in water prior to each kick. Boots were presented to the player in randomised order and each player performed ten successful kicks in each boot. The dynamics of ball flight were determined for every kick using a bespoke ball strike launch monitor system (Quinspin, Sports Dynamics Ltd.), positioned 1.25 metres from the soccer ball, perpendicular to the initial direction of travel. Spin rate and ball velocity were analysed for each kick and the data presented is a mean + 1 standard deviation (SD) of ten kicks. The player testing setup is shown in Figure 2. A one-way ANOVA was conducted to determine if any of the boot conditions generated significantly different ball velocity or spin. Accepted levels of significance were set at P0.05 and all statistical analysis was conducted using SPSS v.13.
Figure 2 - Plan view of player testing setup.
The Engineering of Sport 7 - Vol. 1
251
3- Results 3.1 Friction testing protocol Figure 3 illustrates the static and dynamic frictional force (Mean +1 SD) of the boot samples across the flat football surface. Both speeds and conditions are displayed. In Figure 3 (a) upper D shows a significant (P0.05) increase in frictional resistance for all ‘speed 1’ and ‘wet’ conditions against all other samples. However due to its large standard deviation, during tests under the higher speed, it does not show a significant increase over upper B in the condition ‘dry speed 2’. Upper B exhibits significantly (P0.05) higher frictional resistances than upper C and A in all conditions apart from ‘dry speed 1’ where it is only significantly (P0.05) higher than upper A. Upper C demonstrates significant (P0.05) increases in friction over upper A in all conditions excluding the condition ‘dry speed 2’ where no significant difference was observed. The dynamic frictional force between samples showed very similar trends and hierarchies to the static frictional results. The graphed data illustrates lower mean forces and decreased standard deviations throughout the data, as shown in Figure 3 (b). Uppers A, B, and C showed similar percentages of decrease in force and therefore remained in the same hierarchal order as in the static test. However, upper D showed a larger percentage decrease over the other three uppers. This is illustrated with upper B showing significantly (P0.05) higher dynamic frictional resistances in both ‘dry’ conditions over upper D, however no significant difference is seen between the ‘wet’ conditions with these two samples. Upper D still shows significantly (P0.05) higher dynamic frictional forces than uppers C and A.
Figure 3 - Mean frictional forces, (a) Static condition, (b) Dynamic condition.
3.2 Player testing protocol Figure 4 illustrates the mean velocity for all players and all kicks recorded for each of the boots in wet and dry conditions. No significant differences were exhibited between the boots although there is a consistently lower velocity achieved in the wet condition, the mean wet condition velocity was 1.46 ms-1 less than in the dry for all boots. On average, approximately a 6%
252 The Engineering of Sport 7 - Vol. 1 decrease in measured ball velocity was achieved in the wet. Boot B showed the greatest difference between wet and dry ball velocity.
Figure 4 - Mean velocity measurements (Mean + 1 SD).
Figure 5 illustrates the mean spin generation for all players and all kicks recorded for each of the boots in wet and dry conditions. No significant differences were exhibited between the boots although there is a consistently lower spin rate achieved in the wet condition, the mean wet condition spin rate was 82.3 RPM less than in the dry for all boots. Wet spin rates for boot B and boot C were noticeably lower than the equivalent dry spin rates.
Figure 5 - Mean spin rate measurements (Mean + 1 SD).
The Engineering of Sport 7 - Vol. 1
253
4- Discussion The testing protocol was carried out successfully and reports one of the first investigations to measure soccer ball launch characteristics using an automated measurement system. The measured velocities for the instep kick are comparable with previous results, which range from 18 to 31 ms-1 [AA1, N1, P1, RM1], however this is the first study to show the effects on velocity due to wet versus dry conditions of the boot and ball. The prototype boot tested against market leading boots showed that the acquired results were comparable. The large variability in results does not allow for statistically significant results. Player testing is accounted for the large variation in results. In order to obviate this, a mechanical kicking simulator could have been used, currently however they do not mimic foot kinematics satisfactorily and obviously would not allow for player feedback on the boots. It is thought that using professional players more consistent kicks could have been achieved, which could have led to statistical significance between boots. Anecdotal evidence proves that the comfort, feel, aesthetics and branding were deemed important by the players even if the boots performed similarly. Feel of the ball on the foot was important and it was stated that the prototype boot compromised this, even though the strike ‘felt good’. Players stated that when striking the ball with boot B and D they felt as if they ‘imparted more spin’ on the ball, the ‘material feels really rough, grippy’, which correlates with the friction testing findings, however strikes with boot C seemed to give the highest spin rate values. It was assumed that the large friction values would correlate to higher spin rates, since more grip would facilitate longer contact time between foot and ball on impact, allowing more spin to be imparted upon the ball. Ideally a larger number of boots would have been tested, and a bottom of the range boot could have been selected, it is thought that this would have performed statistically worse than the other boots. This would confirm the capability of the testing protocol to measure differences using player testing for velocity and spin measurement. Boot D appeared to be the most consistent performer in both wet and dry conditions regarding velocity and spin rate. This was the only material to elicit an increase in friction conditions when wet.
5- Conclusions The quantitative testing protocol allows the investigator to substantiate claims such as more power, more control and more swerve, since absolute values can be measured for successive strikes of the ball. The test shows interesting findings, clear trends are observed between wet and dry conditions when striking the ball. Noticeably lower launch velocities and spin rates are measured in the wet condition. This is the first study to report such behaviour. Since football is predominantly played outdoors, the testing suggests that footwear selection should perhaps be based on weather condition. Based on this, playing in the dry condition, from a performance point of view, boot B would be recommended, however if playing in wet conditions boot B seemed to be the worst performer. This is another
254 The Engineering of Sport 7 - Vol. 1 consideration for boot manufacturers in the design and development process for new boots. The testing showed that the prototype boot B, that was tested against market leading soccer boots was very comparable in relation to measured performance characteristics. It is therefore concluded that the boot is ready to be released for the mass market.
6- Acknowledgements The authors wish to thank the external collaborator for funding this work, the enthusiasm shown by the soccer players and Mr R. Waters and Mr A. Gray for assistance with the player testing protocol.
7- References [AA1] Asai, T., Akatsuka, T., and Haake, S., 1998. The physics of football. Physics World, 11, (6), pp. 25-7. [AY1] Acar, M., F., Yapicioglu, B., Arikan, N., Yalcin, S., Ates, N., and Ergun, M., 2007. Analysis of goals scored in 2006 World Cup, VIth World Congress on Science and Football, book abstracts, January 16-20th, Turkey. [N1] Neilson, P., J., 2003. The dynamic testing of soccer balls. PhD Thesis. Chapter 4 – Determination of soccer ball performance parameters, pp. 59-74. [P1] Plagenhoef, S., 1971. Patterns of human motion, a cinematographic analysis, Prentice Hall. [RM1] Roberts, E., M., and Metcalfe, A., 1968. Mechanical analysis of kicking. Biomechanics I, Baltimore University Press, pp. 513-9. [SG1] Sporting Goods Manufacturers Association (SGMA)., 2007. Manufacturers sales by category report. s
Development of a Measurement-Prosthesis for a Ski Boot Test Bench (P48) M. Reichel1, A. Haumer1, H. Schretter2, A. Sabo1
Topics: Ski & other Winter Sports; Biomechanics; Materials; Measurement Systems; Shoes; Testing, Prototyping, Benchmarking; Abstract: In cooperation between HTM Sport- und Freizeitgeräte AG (Tyrolia) and the University of Applied Sciences Technikum Wien a ski boot test bench for stiffness measurement was developed. The former implemented prosthesis of the lower leg was only equipped with a hinged ankle joint. To optimize the measurement system a new prosthesis of the lower human leg was designed. Due to significant differences between ski boots for men and for women, prosthesis for male and female has to be distinguished. The differences are the varying sizes of feet for men and women and the different shapes of the lower legs. The foot is a “Greissinger plus Fuß” made by Otto Bock Healthcare Products GmbH with mobility in the forefoot area. The ankle joint of the new prosthesis is based on the human ankle joint, so the axes conform to the axes of the upper and lower ankle joint with anatomically range of motion. The mobility of the joint also conforms to the anatomic range of motion and is limited by elastomers. Furthermore the ankle joint is linked to a tube by an adapter of Otto Bock, which is used to transmit force. A lower leg is positioned around the tube which guarantees a natural fit inside the ski boot. The developed prosthesis was compared to the former prosthesis, at the ski boot test bench. The analysis addresses both the reproducibility of the two prostheses and the interpretation of the differences in measurement results for ski boot stiffness. The results of the study demonstrate that measurements of the new prosthesis are more reproducible and that lateral movement of the ski boot, previously inhibited by the single hinged ankle joint, is now possible with the new prosthesis. Keywords: Ski Boot, Flex, Stiffness, Prosthesis, Ankle Joint, Test Bench.
1. University of Applied Science, Technikum Wien, Sports-Equipment Technology, Vienna, Austria E-mail: [email protected] 2. HTM Sport- und Freizeitgeräte AG, Schwechat, Austria - E-mail: [email protected]
256 The Engineering of Sport 7 - Vol. 1
1- Introduction For ski control the ski boot is the most important element beside ski and binding. Forces are needed to turn initiated by the rider are transmitted to the ski via boot and binding. Beside force transmission the ski boot should fulfil several other main criteria. For an average customer the wearing comfort is very important. The more comfortable a boot the softer all materials should be, like cushioning elements and the shell as well as the shaft. This means that a softer boot will deform more than a stiffer one. Manufacturers declare and often indicate this by a flex-index; the higher the index the stiffer the boot. These characteristics of ski boots can be measured in the lab, where boundary conditions can be held more or less constant. For this purpose a ski boot test bench has been developed at HTM Sport- und Freizeitgeräte AG, Schwechat, Austria (Tyrolia) in cooperation with the University of Applied Sciences Technikum Wien [A2]. This test bench is able to measure the applied force, which results in deformation in longitudinal (heel toe) and transversal direction, and the forces as well as the torques in all spatial directions. Furthermore boundary conditions can be changed like temperature, deformation speed, canting adjustments and boot angle in relation to direction of applied force. In the strict sense, the force is applied to a prosthesis of the lower leg, which is positioned into the boot. In principle this prosthesis consists of a foot and a shank connected by a hinge. Tyrolia uses a prosthesis where the foot and the shank are connected by a simple uniaxial hinge joint, which enables only longitudinal forward a backward movements of the shank. This means that no anatomical human movement is possible, which will result in measure errors especially in cases of nonlongitudinal force applications. A further goal is to consider differences between male and female, because female boots have got a softer shank top, mostly a fleece lining, a female last as well as heel wedges due to anatomical conditions. These features should give a higher wearing comfort, especially for the lower female calf onset to maintain good blood circulation. From this point of view, an improved model of a prosthesis should be developed to consider most of above mentioned factors and finally to compare to the old model.
2- Fundamentals The goal is to develop a new prosthesis, consisting again of jointed foot and shank, but the joint should be based on an anatomical human ankle joint. For female a size of “Mondopoint 25” and for male “Mondopoint 27” was chosen, but the male and female prosthesis should also differ in shape of the shank. The ankle joint should be a “standard” ankle joint and used for both male and female prosthesis.
2.1 Ski-boot test bench The ski boot test bench mainly consists of three elements (Figure 1): a base frame, a boot-fixing plate (“binding”) mounted on a 3D force plate (SHUNK) and a lever arm to apply the force to the prosthesis by a servo drive. The applied force is measured by a load cell and the resulting displacement is measured by a path measurement sensor. Figure 1 shows also the calibrated coordination system for torque and force measure-
The Engineering of Sport 7 - Vol. 1
257
ment. The boot is fixed in the x-y-plane, where an angle of +90° to -90° in steps of 5° can be adjusted to simulate more realistic skiing conditions. Force is always applied in xdirection.
Figure 1 - Ski-boot test bench with measurement coordination system (left) and classic “hysteresis” curve of ski boot flex (right)
The arrangement allows measurement of reaction forces and torques between ski and boot as well as applied force and the flex angle of the prosthesis or the boot respectively. For this purpose well known “hysteresis” curves (Figure 1b) can be achieved [S1]. A personal computer is connected to the test bench via an A/D-converter controlled by a “Labview” application to collect all measurement data and store them to a data base.
3- Methods Nowadays several measurement prostheses are available, like e.g. prosthesis with an uniaxial hinge joint (Tyrolia), human anatomic based ankle joint without elastomers [E1] (Sports-Equipment and Material Group at the Technical University of Munich) and the gait analysis ankle joint with elastomers but not for use in ski boots (Technology Development Group of “Fraunhofer Institut”). To develop a human anatomic bases ankle joint with implemented elastomers to simulate joint forces and also useable in ski boots, an anatomical model with axis and positions of upper and lower ankle joint is used (Figure 2, [I1] and [I2]). Considering the movements of upper and lower ankle joint all rotations are monoaxal, but can individually vary in a wide range [A1, D1, G1, F1]. The mean values are an inclination of 79° and deviation of 84° in the upper as well as an inclination of 23° and a deviation of 41° in the lower ankle joint. The distance of the axes is 5mm [I1, I2]. Based on these values an accordingly ankle joint can be constructed with the help of the software “ProEngineer Wildfire 2”. The joint consists of 3 parts, which are a base plate, a middle plate and a top plate. The parts are connected by fit screws which axes correspond to the upper and the lower axes of the ankle joint (Figure 3). The inserted elastomers in the upper ankle joint are of
258 The Engineering of Sport 7 - Vol. 1 diameter 20mm (70 Shore A) and in the lower of diameter 16mm (90 Shore A for outer and 70 Shore A for inner side).
Figure 2 - Superior (a) and lateral (b) view of the axis of the upper and lower ankle joint.
Figure 3 - 3D-view of the ankle-joint (left) and with inserted elastomers (right).
Figure 4 - Assembled measurement prosthesis (left) and its components (right); (1) “Greissinger plus Fuß”, (2) ankle joint with elastomers, (3) adapter to join tube and ankle joint, (4) tube with drilling to adjust shank, (5) fork joint to apply force, (6) adjust pin, (7) milled shank.
The Engineering of Sport 7 - Vol. 1
259
To compare the new developed anatomically based prosthesis (ABP) to the old one uniaxial hinged (UHP), several tests have to be performed: • Reproducibility (7 measurement cycles with Head Edge 9) • Comparison with boot fixation in x-y-plane at 0° • Comparison with boot fixation in x-y-plane at 20° The comparison tests are performed with 3 different boots corresponding to different flex indices: • Boot1 (B1): Head Ezon 8.5 (soft, index = 60) • Boot2 (B2): Head Edge 9 (middle, index = 70) • Boot3 (B3): Head Raptor Supershape (stiff, index = 100)
4- Results For analysing the results the pathway of the servo motor – moving the prosthesis – at 4 defined measurement points is determined; at +100Nm, +200Nm, -100Nm and 200Nm (Figure 5).
Figure 5: Hysteresis curve of ski boot flex measurement; defined points for analysing steps at 100Nm and 200Nm in forward as well as -100Nm and -200Nm in backward direction.
4.1 Reproducibility For the measurement series to determine reproducibility, each prosthesis is used within the boot of middle stiffness (Head Edge 9) in 7 separate measurement cycles. Old (UHP) and new (ABP) prosthesis are measured alternatively with a break of 10min between each cycle to enable complete restoring of boot deformation. Table 1 shows the determined values and deviations at the defined points.
260 The Engineering of Sport 7 - Vol. 1 Table 1 - Values of reproducibility measurement with hinged (UHP) and anatomical (ABP) prosthesis; values in bracket indicate the measurement torque My, M/SD…mean values and standard deviation of 7 cycles.
4.2 Comparison at 0° The comparison measurement between the 2 prostheses at 0° should give the possibility to analyse mainly the behaviour of the ski boot in forward and backward direction (Table 2). Table 2 - Values of comparison measurement at boot fixation of 0° in x-y-plane with hinged (UHP) and anatomical (ABP) prosthesis; values in bracket indicate the measurement torque My, maximum torque Mx and Mz are measured in (for)ward and (back)ward direction.
4.3 Comparison at 20° The comparison measurement between the 2 prostheses at 20° should give the possibility to analyse mainly the behaviour of the ski boot in partially transversal direction (Table 3), like it can occur in onset period of a turn. Table 3 - Values of comparison measurement at boot fixation of 20° in x-y-plane with hinged (UHP) and anatomical (ABP) prosthesis; values in bracket indicate the measurement torque My, maximum torque Mx and Mz are measured in (for)ward and (back)ward direction.
The Engineering of Sport 7 - Vol. 1
261
5- Discussion This work should point out, if there is a need for a new prosthesis based on a human anatomical ankle joint due to measurement errors of a prosthesis with just a uniaxial hinged ankle joint. The new prosthesis is able to fulfil anatomical correct movements which mean that for the ski boot also a freely transversal movement (in y-direction) is possible. This is for lab conditions more close to practical slope behaviour than it is possible with a uniaxial hinged ankle joint. The results of the reproducibility measurements show (Table 1), that the deviations in values of deformation x are lower with the anatomically based prosthesis and therefore the reproducibility is higher. Loading the prosthesis in forward direction at a boot fixation of 0° (Table 2) B3 is the stiffest, B2 is middle and B1 is the softest boot for both types of prosthesis. In backward direction is B2 the stiffest, B1 the middle and B3 the softest boot. In spite of the elastomers the range of motion is higher for ABP than for UHP, probably caused by blocking of UHP. If the torque Mx is high, the shaft of the prosthesis would move utwards but it is blocked by the boot or the joint of the prosthesis. Table 2 shows, that Mx is lower for ABP than for UHP which means, that UHP blocks the naturally movement of the shank and therefore the movement of the shaft of the boot. Also the values of torque Mz is a sign of blocking due to UHP. Measurements with all 3 boots have higher values with UHP than with ABP because ABP enables twisting in z-axis. Table 4 - Values of differences between boot fixation of 20° and 0° in x-y-plane with hinged (UHP) and anatomical (ABP) prosthesis; values in bracket indicate the measurement torque My, M/SD…mean values and standard deviation at each torque value.
Loading the prosthesis in forward direction at a boot fixation of 20° (Table 3) B3 is the stiffest, B2 is middle and B1 is the softest boot for both types of prosthesis. In backward direction is B2 the stiffest, B3 the middle and B1 the softest boot. In spite of the elastomers the range of motion is higher for ABP than for UHP, probably caused by blocking of UHP. At the boot fixation of 20° the difference between UHP and ABP should clearly be seen. Therefore mean differences between fixation of 20° and 0° in displacement of prosthesis at the defined measuring points (200/100/-100/-200Nm) have been worked out (Table 4). The values show, that both types of prosthesis have clear mean differences in forward direction, but the deviation is much higher in UHP which lowers reliability dramatically. In backward direction even the differences are not clear for UHP and also the deviation is very high, which approves the data from forward direction. Mx is similar in UHP and ABP either in forward or in backward direction and Mz is higher in UHP than in ABP, which underlines the block of twisting in UHP (Table 3).
262 The Engineering of Sport 7 - Vol. 1
6- Conclusion Summarising the results it can be stated, that the newly developed human anatomically based prosthesis (ABP) is able to perform measurements with higher reproducibility applied to the ski boot test bench. To analyse ski boots it is necessary to move the measuring prosthesis in all special directions concerning the geometry of ski boots. With the uniaxial hinged prosthesis (UHP) these kind of movements are not or hardly possible because this kind of joint blocks torques in x- and zdirection (Mx, Mz). With ABP shank and shaft movements in all relevant directions can be performed caused by its anatomically based ankle joint.
7- References [A1] Alt, W. W., 2001. Biomechanische Aspekte der Gelenkstabilisierung, 1. Auflage. Geislingen: C. Maurer [A2] Amara, T., 2007: Inbetriebnahme und Evaluierung einer Skischuhprüfmaschine und Entwicklung von Mess- und Auswertungsmethoden. Diplomarbeit. FH Technikum Wien [D1] Dettwyler, M. T., 2005: Biomechanische Untersuchungen und Modellierungen am menschlichen oberen Sprunggelenk im Hinblick auf Athroplastiken. Dissertation. Eidgenössische technische Hochschule Zürich [E1] Ebert, C., 2007: Die Interdisziplinarität in der Sportgeräteentwicklung am Beispiel des alpinen Skischuhs: Suche und Bewertung funktionsrelevanter Konstruktionsparameter. Dissertation. TU München [F1] Faller, A., 1999. Der Körper des Menschen: Einführung in Bau und Funktion, 13. Auflage. Stuttgart: Thieme Verlag [G1] Galik, K., 2002: The effect of design variations on stresses in total ankle arthroplasty. Dissertation. University of Pittsburgh [I1] Inman, V. T., 1976. The joints of the ankle. Baltimore: Williams & Wilkins [I2] Isman, R. E., 1969. Anthropometric studies of the human foot and ankle. Bulletin of Prosthetic Research [S1] Senner, V., 2001: Biomechanische Methoden am Beispiel der Sportgeräteentwicklung. Dissertation. TU München
Development of Multi-platform Instrumented Force Pedals for Track Cycling (P49) Jean-Marc Drouet1, Yvan Champoux1, Sylvain Dorel2
Topics: Bicycle, Measurement Systems. Abstract: The aim of this research was to develop instrumented force pedals that meet the specific requirements of track cycling. Both pedals are instrumented with eight strain gauges and provide the ability to measure normal and tangential pedalling forces. Rotary encoders are used to determine the angular position of the pedals relative to the crank arm and the angular position of the crank arm relative to the bicycle frame. One original feature is that the instrumented pedals can be fitted with interchangeable pedal platforms: the clipless LOOK CX7 and the Shimano 600 (PD-6400) with a toe-clip and strap for sprint and kilometre time trial events. It is therefore possible to measure pedal loads for all track cycling disciplines. The pedals have a very high mechanical resistance in order to withstand the pedal loads produced by track sprinters which are the highest encountered among all cycling disciplines. Their mechanical design allows the pedals to be installed on any crank arm model without requiring any crank modification. With the data acquisition system attached to a modified Camelbak pack carried by the cyclist, the pedals permit on-track pedal load measurements. Post-processing software was developed to calculate derived parameters, which include the effective power, the effectiveness index and two components of the total force: the effective force component that is the component normal to the crank arm and the force component in line with the crank arm. The derived parameter calculations and analysis can be done on site for each leg and allow specific qualities to be evaluated (peak power, peak force, etc.). Typical results for these parameters are presented in this paper. Keywords: instrumented force pedals, track cycling.
1- Introduction The measurement of pedal loads is essential in acquiring a better understanding of the pedalling process as well as providing load data for bicycle design. In designing the proposed instrumented force pedals for track cycling use, specific requirements must be 1. Mechanical Engineering Department, VélUS Group, Université de Sherbrooke, Sherbrooke, Canada E-mail: Jean-Marc.Drouet,[email protected] 2. Laboratoire de Biomécanique et Physiologie, INSEP, Paris, France - E-mail: [email protected]
264 The Engineering of Sport 7 - Vol. 1 addressed. One of these requirements is that the instrumented force pedals have a very high mechanical resistance to withstand the pedal loads produced by track sprinters. These loads are the highest encountered among all cycling disciplines. Even with a high load capacity, the pedals must accurately measure the pedal loads throughout the loading range. In order to be useful for all track cycling disciplines, another requirement is that the instrument force pedals be fitted with two different pedal platforms: clipless pedals and, toe-clip and strap pedals. The latter are used in sprint and kilometre time trial events. Also, considering the many different crank arm lengths used in track cycling, it would more practical and cost effective to have an instrumented force pedal design that allows for normal installation on the crank arm rather than to have a design that requires a dedicated crank arm (Alvarez and Vinyolas 1996, Rowe et al. 1998, Reiser et al. 2003). Because on-track situations such as full-power starts from a dead stop, drafting during a team event and riding on a banked track are difficult or impossible to replicate in a laboratory, on-track measurements are required to obtain realistic pedal load data in these situations. Many different pedal dynamometers have been described in the literature. Some of these dynamometers are restricted to laboratory use (Bolourchi and Hull 1985, Boyd et al. 1996, Davis and Hull 1981, Newmiller et al. 1988, Wheeler et al. 1992) while others permit load measurement outside of the laboratory. In the latter category, Alvarez and Vinyolas (1996) proposed a pedal dynamometer for road use based on a Time clipless pedal platform; Rowe et al. (1998) developed a pedal dynamometer for off-road bicycling; and Reiser et al. (2003) proposed instrumented pedals based on Shimano PD-6500 clipless pedals. These three pedal dynamometers require a dedicated crank arm with integrated bearings. Also, the maximum pedal load capacity reported by Rowe et al. (1998) and Reiser et al. (2003) is not high enough for track cycling use. Since none of these previously described pedal dynamometers satisfy the specific aforementioned requirements for track cycling use, the aim of this research was to develop instrumented force pedals that meet these requirements.
2- Methods The instrumented force pedals can be fitted with two interchangeable pedal platforms: the clipless LOOK CX7 platform (LOOK CYCLE International, Nevers, France) and the Shimano 600 (model PD-6400, Shimano Inc., Osaka, Japan) toe-clip and strap platform (Fig. 1a and 1b). In order to ensure that the instrumented force pedals are functionally equivalent to the original pedals, some parts of the original pedals have been used for the construction of the platform-specific bodies. These platforms are bolted to a pedal base assembly (Fig. 2), which consists of three principal elements: the ball bearing assembly, the instrumented spindle and the rotary encoder. The bearing allows the rotation of the instrumented spindle relative to the crank arm. A thin delrin rod is inserted into the hollow of the instrumented spindle and connects the encoder axle to the bearing assembly, thus allowing the encoder to measure the angular position of the instrumented spindle relative to the crank arm.
The Engineering of Sport 7 - Vol. 1
265
Figure 1 - Photographs of the instrumented force pedal with the LOOK CX7 and the Shimano 600 platforms.
Figure 2 - Photograph of the instrumented force pedal base assembly (exploded view).
The instrumented force pedals measure only the mutually orthogonal force components Fx and Fz (Fig. 3). These forces were measured using a total of eight strain gauges on each pedal (Fig. 2). The strain gauges were arranged in two full Wheatstone bridges, one in the x-y plane (Fx) and the other in the y-z plane (Fz). Theoretically, the position of the strain gauges and their interconnection give bridge signals that are independent of the location of forces Fx and Fz as well as insensitive to the unmeasured loads (Fy, Mx, My and Mz). The instrumented force pedals maximum load is 2500 N. In order to withstand the pedal load, a double row angular contact ball bearing with a dynamic load rating of 10600 N was used. High strength materials were also used for the construction
266 The Engineering of Sport 7 - Vol. 1 of the pedals. Heat-treated 17-4 PH stainless steel (yield strength (Sy) = 1250 MPa) was used for the spindle and the bearing assembly. For the two platforms, 7075-T651 aluminium (Sy = 500 MPa) was used. The mass of one instrumented force pedal is 422 g with the LOOK CX7 platform and 512 g with the Shimano 600 platform (original pedals typical mass: 200 g for LOOK and 282 g for Shimano).
Figure 3 - Pedals local coordinates system with three force components (Fx, Fy, Fz), three moment components (Mx, My, Mz), the pedal angle relative to the crank ( ) and the crank angle relative to the bicycle frame () (right pedal shown).
A supplemental rotary encoder located on the bicycle frame was used (Fig. 4) to measure the angular position of the crank arm relative to the bicycle frame (). Two pulleys (transmission ratio: 1:1) and a synchronous belt were used to link the left crank to the encoder.
Figure 4 - Crank angle () measurement system.
The Engineering of Sport 7 - Vol. 1
267
Measurement data from the instrumented pedals and from the encoder located on the bicycle frame are collected by a data acquisition system (model pro7, ISAAC Instruments inc., Chambly, Canada) attached to a modified Camelbak pack carried by the cyclist (Fig. 5). The pedals as well as the encoder located on the bicycle frame are wired to the data acquisition system. The electrical cables are attached to the cyclist’s legs with elastic bands and Velcro fasteners. The mass of the data acquisition system (including the Camelbak pack) is 1.2 kg. Post-processing software was developed using Matlab to calculate derived parameters, which include the effective power, the effectiveness index and two components of the total force: the effective force component that is the component normal to the crank arm and the force component in line with the crank arm. The derived parameter calculations and analysis can be done on site for each leg and allow specific qualities to be evaluated (peak power, peak force, etc.).
Figure 5 - Track cyclist and bicycle equipped with the instrumented force pedals (Shimano 600 toe-clip and strap platform) and data acquisition system attached to a modified Camelbak pack.
Calibration was performed by applying force and moment loads to measure the direct sensitivity of the in-plane loads (Fx and Fz) and both the calibratable and noncalibratable cross-sensitivities (Rowe et al. 1998). The calibratable sensitivity matrix (V/N, normalized for gain and input voltage ) and the non-calibratable sensitivity matrix (V/N or V/Nm, normalized for gain and input voltage) are given by equations (1) and (2) respectively. Moment My was not considered because it is produced by the pedal bearing friction and can be ignored. (1) (2) Using the extreme loading amplitudes indicated in Table 1, the maximum total root mean square error of the instrumented pedal was found to be 1.6% FS (Full Scale) for Fx and 1.5% FS for Fz. The hysteresis was also determined from the calibration data and it was
268 The Engineering of Sport 7 - Vol. 1 found that hysteresis introduced a maximum error of 0.9% FS. The instrumented pedal repeatability and reproducibility have been assessed by successive calibration. They were found to be less than 1% FS. The natural frequency along the x and z axis was determined using the pedal stiffness and assuming half the weight of a 75 kg cyclist clipped onto the pedal. The natural frequencies for both directions were about 125 Hz. The resolution of the rotary encoders mounted on the instrumented force pedals was 0.4º. A zeroing adjustment for both components of force (Fx and Fz) and pedal angle ( ) was carried out before each measurement session. All the signals were acquired at a sampling rate of 1 kHz (USB data acquisition, ISAAC Instruments inc., Chambly, Canada) and stored on a computer.
Table 1 - Pedal loads for the evaluation of the direct sensitivities and the cross-sensitivities.
3- Results Different experimental sessions in field conditions (i.e. on a track) were carried out, demonstrating the ability of these instrumented force pedals to provide a good quantity of relevant information. The present data describe an example of some measured and calculated derived variables obtained for an elite cyclist (i.e. world class sprinter) during a specific 125 m all-out effort. The athlete was asked to perform maximal effort and encouraged to produce the greatest possible acceleration. A starting block was used for the start and the cyclist naturally adopted a standing position throughout the exercise. Following the test session, raw data stored in the acquisition system were uploaded on a computer for subsequent analysis. The typical time course of the raw data (Fx, Fz, pedal angle of the left pedal and crank angle) measured by the device are presented on Fig. 6. Fz reached very high negative values during each downstroke phase and especially at the beginning of the sprint (>2200 N) while non-negligible positive values (almost 400 N) were observed during the upstroke phases. Oscillations between positive and negative values were also observed for Fx but the magnitude remained much lower (between -500 N and +600 N).
Figure 6 - Measured and calculated derived variables for a world class sprinter during a specific 125 m all-out effort.
The Engineering of Sport 7 - Vol. 1
269
Based on Fx, Fz, the pedal angle and the crank angle (Fig. 6), the total force was calculated by trigonometry and resolved into two components: one orthogonal to the crank (effective force) and another along the crank (ineffective force). A time derivative of the crank angle was used to obtain crank angular velocity and hence the power output (i.e. the product of effective torque and crank angular velocity). Evolution of total and effective forces on the left pedal, crank angular velocity and total power output (i.e. on both pedals) is depicted in Fig. 6. The effective force remained positive throughout the test (except during the last 2-3 seconds): although the effective force produced during the downstroke was very high (peak value > 2200 N), the value during the upstroke was also significant (>500 N), which highlights the importance of all zones of the crank cycle in carrying out the maximal acceleration of the system.
4- Discussion One of the design requirements for the instrumented pedals was that they be able to withstand the very high loads applied by track sprinters. As can be imagined, the severe consequences for the cyclist in the event of mechanical failure of the pedals during high force and high-speed runs are such that this design criterion is an important safety issue. Finite elements analysis (FEA) was used to evaluate stress levels in the pedal. The selected ball bearing was also experimentally tested to verify its mechanical resistance. Functional equivalence of the force pedals with the original LOOK CX7 and Shimano 600 pedals was also a design requirement. With the LOOK CX7 platform, the instrumented force pedals allow for normal engagement and disengagement of the LOOK Delta cleat and for full use of the cleat’s floating angle. For the Shimano 600 platform, the original toe-clips are used. The double straps go underneath the pedals and are held securely in place using tie-wrap fasteners. To avoid contact between the right pedal and the track when riding on the steep banking of a track, another consideration was that the pedals be compact. The use of high strength materials permitted us to reduce the pedals size. FEA was also used to ensure that no mechanical interference occurs between the pedals components under maximum load conditions. A consequence of reducing the pedal size was the reduction of pedal mass, which is of importance considering the high accelerations encountered in track cycling. During the design process, FEA was used for the optimization of the area moment of inertia of the instrumented area of the spindle in order to increase sensitivity. It was also used to determine the corner radius at each end of the instrumented area of the spindle. The stress concentration in the vicinity of these radii was taken into account to determine the location of the strain gauges and to numerically ensure low cross-sensitivities. The accuracy of the in-plane forces was established through calibration by evaluating the direct sensitivity and also by measuring the influence of the other load components. Extreme loadings were considered and it was established that the influence of non-measured loads is small. The direct cross-sensitivity between measured forces Fx and Fz is also small and does not contribute significantly to measurement error. For the measured forces, the linearity is very good and the hysteresis is small. The first natural
270 The Engineering of Sport 7 - Vol. 1 frequency of 125 Hz is high enough to assume a dynamic flat response of the instrumented pedals within the operational measured frequency band of interest of 0-30 Hz. The aerodynamic drag of the data acquisition system located on the back of the cyclist was a concern in high-speed runs (over 60 km h-1). Slower times were observed for 200-m sprints during which the cyclist is in a low crouch and the acquisition system is exposed to airflow. The electrical cables connecting the pedals to the acquisition system were attached to the external side of the cyclist’s legs. It was reported that when routed this way, the cables do not interfere with leg movement even at very high pedalling cadence (over 160 rev min-1). The data presented in this paper constitute a typical sample from among the different possibilities allowed by this new device. The device allows researchers and coaches to analyse the data using practical concerns while providing relevant and useful information that reflects not only the muscular capacity but also the technical abilities of the cyclist. Among the technical aspects that can be measured are the maximal effective force and power output, the index of asymmetry and the index of effectiveness.
5- Conclusion In this paper, instrumented force pedals that meet the specific requirements of track cycling were presented. They are functionally equivalent to the original LOOK CX7 and Shimano 600 pedals and provide accurate measurements of the mutually orthogonal force components Fx and Fz with very low cross-sensitivities. Considering that these instrumented force pedals are a research tool as well as a tool for use in the field, multiple perspectives are offered for the future in terms of scientific approaches to training.
6- References [AV1] Alvarez G. and Vinyolas J. A New Bicycle Pedal Design for On-Road Measurement of Cycling Forces. In Journal of Applied Biomechanics, 12(1):130-142, 1996. [BH1] Bolourchi F. and Hull M.L. Measurement of Rider Induced Loads During Simulated Bicycling. In International Journal of Sport Biomechanics, 1(4):308-329, 1985. [BH2] Boyd T., Hull M.L. and Wooten D. An improved accuracy six-load component pedal dynamometer for cycling. In Journal of Biomechanics, 29(8):1105-1110, 1996. [DH1] Davis R.R. and Hull M.L. Measurement of pedal loading in bicycling: II. Analysis and results. In Journal of Biomechanics, 14(12):857-872, 1981. [HD1] Hull M.L. and Davis R.R. Measurement of pedal loading in bicycling: I. Instrumentation. In Journal of Biomechanics, 14(12):843-856, 1981. [NH1] Newmiller J., Hull M.L. and Zajac F.E. A mechanically decoupled two force component bicyle pedal dynamometer. In Journal of Biomechanics, 21(5):375-386, 1988. [RH1] Rowe T., Hull M.L. and Wang E.L. A Pedal Dynamometer for Off-Road Bicycling. In Journal of Biomechanical Engineering, 120(1):160-164, 1998. [RP1] Reiser II R.F., Peterson M.L. and Broker J.P. Instrumented bicycle pedals for dynamic measurement of propulsive cycling loads. In Sport Engineering, 6(1):41-48, 2003.
The Engineering of Sport 7 - Vol. 1
271
[WG1] Wheeler J.B., Gregor R.J. and Broker J.P. A Dual Piezoelectric Bicycle Pedal With Multiple Shoe/Pedal Interface Compatibility. In International Journal of Sport Biomechanics, 8(3):251258, 1992.
In-Situ Measurement of Clipless Cycling Pedal Floating Angles (P51) Yvan Champoux1, Daniel Paré, Jean-Marc Drouet, Denis Rancourt
Topics: Bicycle, Measurement Systems. Abstract: In cycling sports, clipless pedals are used by athletes and dedicated cyclists to attach the shoe to the pedal because this allows efficient energy transfer to the bike. Most clipless pedals now offer a degree of freedom (float) to the shoe around an axis normal to the pedal surface. This feature was originally introduced in an attempt to reduce knee injuries due to overuse. Most studies reporting on the influence of the clipless pedal floating angle on knee injuries have been carried out under laboratory conditions and little is known about the use of the float in real road conditions. This paper is an evaluation of the design and accuracy of a new apparatus that can measure the in-situ clipless cycling pedal floating angle. A high-sensitivity sensor that can measure magnetic field orientation is embedded in a commercial pedal. Small magnets temporarily clipped onto the cleat create the required magnetic field around the sensor. This measurement technology eliminates the need for a physical connection between the sensor and the shoe, thus allowing the pedal to be used normally. Static calibration and a subsequent accuracy check revealed that the angle measurement uncertainty was found to be within a range of ±0.25º with an hysteresis of less than 1% Full Scale. Typical in-situ sample floating angle measurements are included to demonstrate the ability of the instrument to provide useful information. Keywords: Measurement, clipless pedals, floating angle, cycling.
1- Introduction In cycling sports, 41% of overuse injuries/complaints occur at the knee. Several studies have investigated potential knee injury mechanisms in cycling (Ericson et al. 1984) (Ruby, Hull and Hawkins, 1992) ( Ruby et al. 1992) (Bailey et al. 2003). A clipless pedal was first commercialized by LOOK Cycle in 1984. This type of pedal attaches the shoe to the pedal and allows efficient energy transfer to the bike. Modifications of the clipless pedal have included the introduction of an additional degree of freedom to the shoe (float) around an axis normal to the pedal surface in such a way that the foot’s internal 1. Mechanical Engineering Department, VélUS Group, Université de Sherbrooke, Sherbrooke (Quebec) Canada J1K 2R1 E-mail: Yvan.Champoux,Daniel.Pare,Jean-Marc.Drouet,[email protected]
274 The Engineering of Sport 7 - Vol. 1 and external rotation is allowed within a limited and set range of motion. The foot movement allowed by the float reduces pedal loads and varus/valgus knee moments (Ruby and Hull, 1993) (Boyd et al. 1997). To the authors’ knowledge, the studies reporting on the influence of the clipless pedal floating angle on knee injuries were carried out under laboratory conditions. Consequently, there is no information available on how cyclists actually use the floating angle in different road situations. The goal of this study was to develop and test an apparatus specially designed to provide accurate and reliable monitoring of the in-situ time variation of the floating angle of commercially available clipless pedals. The apparatus needed to be lightweight, small, easy to use and to not perturb the natural motion of the foot while allowing the cyclist to easily clip in and out. The floating allowed by the two commercial pedal models selected for use in this study (LOOK KéO and LOOK PP 336) was limited to an angular rotation range of 9° along an axis normal to the pedal surface, as shown in Fig. 1. The floating axis is located approximately 25 mm in front of the pedal axis. The 0º mark of the floating angle corresponded to the most counter-clockwise position of the floating range as seen from above. The positive floating angle corresponded to a clockwise rotation of either the left or right shoe. Most commercial clipless pedals are generally based on the same functioning principle and use similar components. A cleat is solidly attached to the underside of the shoe. The spring loaded pedal mechanism holds the cleat solidly in place and allows the cyclist to clip in and clip out of the pedal. Most clipless pedals offer a rotational floating along the Z axis but some of them also allocate a few centimetres of freedom along the Y axis.
Figure 1 - Clipless pedal floating representation. View from above the right pedal. Floating is identified by a rotation along the floating axis in –Z direction. Floating angle increases with a clockwise rotation of the shoe.
The Engineering of Sport 7 - Vol. 1
275
Within the floating 0º-9º range, the resistance moment of the pedal along the Z axis is small and is generated by dry friction between the cleat and the pedal. To disengage the foot from the pedal, the cyclist must apply a strong moment along the Z axis to rotate the shoe beyond the resistance-free 0°-9° range up to angles of approximately –16° and +25°. Consequently, at angles of less than 0° or greater than 9° the cyclist experiences external moment loads around the Z axis.
2- Measurement system The commercial clipless pedals were equipped with a Honeywell angular displacement sensor HMC 1501. This sensor is composed of a Wheatstone bridge and a high-resolution low-power magnetic resistance transducer that measures magnetic field angle direction with a resolution of 0.07º.
Figure 2 - Diagram of the apparatus
The bandwidth response is between 0-5 MHz. The transducer is very small and occupies a volume of only 5 mm x 4 mm x 1.2 mm. Unlike incremental encoding devices, the sensor detects absolute position and requires no indexing for proper positional output. The available full-scale output range is 120 mV for a bridge excitation of 5 V. The small transducer was held in place under the pedal with a plastic support glued to the pedal body as shown in Fig. 2. The additional weight of the device for one pedal is of the order of 80 g. A ceramic horseshoe magnet was anchored underneath the cleat with a small aluminium holder. Magnetic north and south poles created a strong magnetic field around the sensor. The rotation of the shoe/cleat modified the magnetic field which is detected by the sensor. The sensor was aligned along the axis of rotation of the shoe/cleat. There was a gap of 4 mm between the magnets and the top surface of the sensor. An ISAAC Instruments Dual action Wheatstone bridge conditioner model MODWBD-101
276 The Engineering of Sport 7 - Vol. 1 (3 mV/V Full Scale) was used to supply the proper signal conditioning to the sensor. An inductive proximity probe with sufficient sensitivity to detect the passage of the crank without requiring target or reflective tape was installed on the frame. This generated a one-pulse-per-revolution synchronisation signal which was used to segment the signal and to calculate cycle averages. A Model v7 Pro ISAAC Instruments Data Acquisition System was used to store the signals on three channels. With a sampling rate of 1 kHz and a memory capacity of 128 Mb, signals can be recorded for over 5 hours. The recorder and its battery pack were mounted on a modified Camelback® hydration pack. Small electric wires connecting the pedals to the recording system were routed along the cyclist’s lower limb. The recorded data was transferred to a computer with a USB connexion for post processing.
3- Calibration The Honeywell HMC 1501 sensor electrical voltage output V is V = VsS sin(2ø)
(1)
where Vs is the supply voltage, S is a constant determined by the material and ø is the angle between the orientation of the sensor and the magnetic field. For such a sinusoidal function at angles near 0°and for a small angle range (15º), a linear behaviour of the sensor can be assumed. For calibration, an accurate protractor was directly attached to a cleat in place of a shoe. Floating angles and sensor outputs were simultaneously recorded and used to determine the instrument’s direct sensitivity. A very good linearity was obtained (R2=0.999) and a nominal sensitivity of 0.45 mV/V/° was measured. Because small angle variations are measured, any unwanted relative displacement (besides the rotation related to the floating angle) between the magnet and the sensor provoked by pedal loadings would perturb the measurements. Static Loads (forces Fx, Fy, Fz; moments Mx) were individually applied to a cleat mounted on the instrument to verify the sensitivity of the measured floating angle to loads applied to the pedal. Moment My was not applied because it is produced by the pedal bearing friction and can thus be neglected. Moment Mz was also not considered because it is directly responsible for the floating angle variations. Using the load ranges shown in Table 1, maximum angle errors respectively for each load were measured. The total root mean square error related to pedal loads was found to be smaller than 0.25°. The hysteresis was also determined from the calibration data; it introduced a maximum error of less than 1% Full Scale. The calibration procedure was repeated at several occasions over a long period of time (more than 2 years) and the maximum variation of the sensitivity was less than 1% guarantying a good reproducibility in the measurements.
The Engineering of Sport 7 - Vol. 1
277
Table 1 - Pedal loads for the evaluation of the instrument’s cross-sensitivity.
4- Sample data for the in-situ floating angle To give examples of typical results and to demonstrate the ability of the floating angle measurement instrument, in-situ loading angles were measured in various conditions as shown in Figure 3. Each single floating cycle is described by its amplitude and angular position. As shown in Fig. 3c, the amplitude of the selected cycle (2.1°) corresponds to the peak-to-peak amplitude over the cycle. The angular position (3.5°) of the cycle corresponds to the central angular value of the cycle. A modified version of a polar plot provides an interesting global representation of a complete run, as shown in Fig. 4. Each cross denotes the amplitude (radial distance from the origin of the plot) and position (angular position) of a single cycle. The white dot in the centre of the grey square indicates the average angular position and amplitude for all the cycles under consideration. The two thin curved lines on both the left and right of the pedal polar plot indicates the zone limits within which the cyclist does not exceed the pedal floating range.
Figure 3 - Measured signals as a function of time. a) Synchronisation signals generated by the trigger signal b) Left pedal floating angle c) Right pedal floating angle.
278 The Engineering of Sport 7 - Vol. 1 The results in Fig 4a) show that for the cyclist being tested during flat road sitting, the right pedal floats near the centre of the available floating range. However, the results for the left pedal indicate that the cyclist consistently exceeds the limit of the available floating range and pushes the spring loaded pedal mechanism, increasing foot loads and creating additional stress to the knees (Ruby and Hull, 1993; Gregersen and Hull, 2003). Figures 4 b) and 4c) show the respective results for climbing in the sitting and standing position. The left foot shows large amplitudes that systematically exceed the range limit of the pedals (0-9°) for climbing.
5- Discussion The goal of this work was to design an instrumented clipless pedal that would provide an accurate measurement of the floating angle under real operating conditions. Using commercial pedals to develop the instrument ensures normal use of the pedal floating in order to obtain realistic measurements. The use of a magnetic sensor also guaranteed that the pedal could be used safely because the cyclist is able to dismount easily if needed. Upon calibration, the direct sensitivity is very linear and consistent and the hysteresis is small. A commercial clipless pedal is not an infinitely rigid structure and a slight deformation of the pedal body was observed when loads were applied. This was identified as the most important contributor to measurement error. Taking this fact into account, the influence of the pedal loading was measured and included in the measurement error. Tolerance variations of the shape of the cleats do not allow an exact repositioning of the magnet on each cleat. It was found that the instrument’s sensitivity is not significantly influenced by the variation in magnet position on the cleat. However, it may create a small angle measurement bias and zeroing is therefore required at each test to eliminate this bias. The measurement principle is based on the evaluation of the direction of a magnetic field. The magnet placed very close to the sensor succeeds in generating a sufficiently strong magnetic field to yield measurements that are not influenced by the surroundings. For example, it was verified that the use of a steel bike or the earth’s magnetic field did not influence the measurement. No significant drift was noted due to barometric pressure or temperature change. No special effort was taken to make the instrument waterproof and consequently, it is not recommended for use in wet conditions. The apparatus requires small wires to be attached along the cyclist’s legs. All of the cyclists tested stated they were not bothered by the wires.
The Engineering of Sport 7 - Vol. 1
279
Figure 4 - Polar plot indicating the amplitude and position of the cleat for each cycle and for the left and right pedals; MA = Mean Amplitude; MP = Mean Position. a) Flat road; Sitting; Number of cycles = 64; Left pedal: MA = 2.3º; MP = 1.4º; Right pedal: MA = 1.0º; MP = 5.2º b) Climbing and sitting; Number of cycles = 48; Left pedal: MA = 2.1º; MP = 5.5º; Right pedal: MA = 1.0º; MP = 6.1º c) Climbing and standing; Number of cycles = 20; Left pedal: MA = 3.4º; MP = 1.9º; Right pedal: MA = 0.9º; MP = 6.2º
280 The Engineering of Sport 7 - Vol. 1
6- Conclusion Although measurements of very small angle variations in the field is a difficult task, the authors were able to demonstrate that the instrumented commercial pedal suitable for in-situ measurement of the floating angle used in these tests provides reliable and accurate results. No other instrument enabling the in-situ quantification of the floating angle of a commercial pedal has yet been reported. This quantification is useful to investigate the influence of power and cadence on the in-situ use (hill climbing, sitting position) of the floating angle. It is also useful for the development of cleat fitting techniques to optimize the use of clipless pedal floating.
7- Acknowledgments The authors wish to thank ISAAC Instrument and LOOK Cycle International for their collaboration. This study received financial support from the National Sciences and Engineering Research Council of Canada (NSERC).
8- References [BM1] Bailey, M.P., Maillardet F.J. and Messenger, N. Kinematics of cycling in relation to anterior knee pain and patellar tendinitis. Journal of Sports Sciences, 21:649-657, 2003. [BN1] Boyd, T.F., Neptune, R.R. and Hull, M.L. Pedal and knee loads using a multi-degree-offreedom pedal platform in cycling. Journal of Biomechanics, 30:505-511, 1997. [EN1] Ericson, M.O., Nisell, R. and Ekholm, J. Varus and valgus loads on the knee joint during ergometer cycling. Scandinavian Journal of Sports Sciences, 6:39-45, 1984. [GH1] Gregersen, C.S. and Hull, M.L. Non-driving intersegmental knee moments in cycling computed using a model that includes three-dimensional kinematics of the shank/foot and the effect of simplifying assumptions. Journal of Biomechanics, 36:803-813, 2003. [RH1] Ruby, P. and Hull, M.L. Response of intersegmental knee loads to foot/pedal platform degrees of freedom in cycling. Journal of Biomechanics, 26:1327-1340, 1993. [RH2] Ruby, P., Hull, M.L. and Hawkins, D. Three-dimensional knee joint loading during seated cycling. Journal of Biomechanics, 25:41-53, 1992. [RH3] Ruby, P., Hull, M.L., Kirby, K.A. and Jenkins, D.W. The effect of lower-limb anatomy on knee loads during seated cycling. Journal of Biomechanics, 25:1195-1207, 1992. [WH1] Wilber, C.A., Holland, G.J., Madison, R.E. and Loy, S.F. An epidemiological analysis of overuse injuries among recreational cyclists. International Journal of Sports Medicine, 16:201206, 1995.
Correlation Between Treadmill Acceleration, Plantar Pressure, and Ground Reaction Force During Running (P52) Alex, J. Y. Lee1, Jia-Hao Chou1, Ying-Fang Liu2, Wei-Hsiu Lin3, Tzyy-Yuang Shiang4
Topics: Biomechanics, Innovation & Design Abstract: The purpose of this study was to investigate the correlation between peak treadmill acceleration (PTA), peak plantar pressure (PPP) and peak ground reaction force (PGRF) during running. Eight active college students (mean age: 21 ± 0.8 yrs; height: 169.9 ± 7.4 cm; weight: 62.5 ± 9.8 kg) wore standardized shoes and ran on a speed calibrated treadmill (95Te, Life Fitness, USA) at seven speeds (1.3, 1.8, 2.2, 2.7, 3.1, 3.6, and 4.0 m/s). PTA was measured by a dual axis accelerometer (MAX2312G/M, MEMSIC, Inc, USA) which plugged into the middle of the treadmill running board, with an MP100 data acquisition system (BIOPACK Systems, Inc, USA). In-shoe PPP and PGRF were measured by a wireless foot pressure measuring system (F Scan Mobile, Tekscan, USA). The PTA, PPP, and PGRF data were recorded for 10 seconds at a sampling rate of 500 Hz for each running speed. PPP and PTA data at different speeds were compared across the range of speeds by a repeated measures one-sample t-test. A standard linear least squares correlation was used to calculate the coefficients of determination (r2) between PTA, PPP, and PGRF. The running speed had a different effect on PTA, PPP, and PGRF at the seven speeds. The PTA, PPP, and PGRF during the fastest speeds (4.0 m/s) increased approximately 425% (2.2 g vs. 0.53 g), 216% (225.7 psi vs. 104.4 psi), and 228% (311.2 kgs vs. 136.6 kgs), respectively, when compared to the slowest speed (1.3 m/s). The coefficients of determination between PTA and PPP, PTA and PGRF, PPP and PGRF were 0.75, 0.78, and 0.91, respectively (p<.05). The PTA, PPP, and PGRF increased linearly with faster running speed and were strongly correlated with each other. This study demonstrates that an onboard treadmill accelerometer could be used to estimate the PPP and PGRF during running. Key words: accelerometer; foot pressure; foot and ankle; foot biomechanics; running impact.
1. Department of Physical Education, National HsinChu University of Education, Taiwan, ROC E-mail: [email protected] 2. General Education Center, Hsin Sheng College of Medical Care and Management, Taiwan, ROC 3. General Education Center, Tzu-Chi College of Technology, Taiwan, ROC E-mail: [email protected] 4. Institute of Exercise and Sports Science, National Taiwan Normal University, Taiwan, ROC E-mail: [email protected]
282 The Engineering of Sport 7 - Vol. 1
1- Introduction Ground reaction forces (GRFs) are defined as the forces that act on the body as a result of the body’s interaction with the support surface, and these forces are used to evaluate the function of normal and pathological gaits by providing objective measurements of changes in loading under static or dynamic conditions (Chen and Bates, 2000). The musculoskeletal structures of the lower extremities transmit force and experience shock as a result of the external GRF acting at the foot-ground interface during locomotion (Lafortune et al., 1995). It has been reported that increased GRFs during foot-ground contact have been associated with overuse running injuries (Hreljac, 2004) and increased incidents of anterior cruciate ligament (ACL) injuries (Chappell et al., 2002). Plantar pressure (PP) is a measure of the pressures or forces acting on the interface between the shoe and the foot while the foot is in contact with the supporting surface. Measurements of PP provide an indication of foot and ankle function during functional activities, and can be used for the evaluation and management of patients with a wide variety of foot impairments associated with neurological and musculoskeletal disorders (Orlin and McPoil, 2000, Veves et al., 1992). Although the force platform provides valuable information regarding both the vertical and shear components of the GRF (Orlin and McPoil, 2000), insole plantar pressure measurement systems can provide more in-field information for measuring the PP and GRF directly applied to the foot during less constrained tasks. In addition, insole plantar pressure measurement systems can be used to assess other parameters outside the laboratory with relative ease, such as the force distribution along the plantar region or between a bilateral foot (Santos-Rocha and Veloso, 2007). The forces on the internal muscles, bones, ligaments, and tendons during landing in physical activity have been related to various joint pathologies and are a substantial cause of lower extremity injuries in the general population (Hurwitz et al., 2000, Murphy et al., 2003, Richards et al., 1996). Although the foot facilitates the transmission of weight to the ground and protects the rear foot and lower extremity during the early stage of the gait cycle, it has been demonstrated that the passive impact portion of the GRF and tibial shock are linked with the incidence of chronic overuse injuries (Rodgers, 1995). Moreover, the GRF at the heel strike produces internal loading of the lower limb and causes transient stress waves that increase with a greater force (Liddle et al., 2000). Previous studies have been performed to measure and calculate the internal forces in the knee during jump landing (Decker et al., 2003, Pflum et al., 2004). A recent study also demonstrated that the tibial axial force (along the long axis of the tibia) can be correlated with peak vertical ground reaction forces during jumping (Derrick and Mercer, 2004). In addition, both the maximum force and peak pressure under the heel were highly associated with walking speed (Menz and Morris, 2006), and peak impact accelerations in jump landing have been correlated with peak GRFs (Elvin et al., 2007a). However, the actual relationships between the peak treadmill acceleration (PTA) of the running board, the peak plantar pressure (PPP) and the peak ground reaction force (PGRF) during walking and/or running exercise have not yet been studied; therefore, the purposes of this study was to investigate the relationship between PTA, PPP, and PGRF during different speeds of walking and/or running exercise.
The Engineering of Sport 7 - Vol. 1
283
2- Methods Eight young active college students (mean ± SD: age 21 ± 0.8 years; mass 62.5 ± 9.8 kg; height 169.9 ± 7.4 cm) participated in the study. All subjects were volunteers from the university student body and were right leg dominant, as established by their preferred kicking leg (Lin et al., 2008). None of the subjects reported a history of orthopedic injury, lower extremity trauma, deformities, or vascular disease, and all subjects signed an informed consent form approved by the University Human Subjects Internal Review Board. Evaluation included a thorough clinical examination to ensure the absence of hip or knee pathology that might affect gait. Foot examination included measurement of passive and active range of motion of the ankle joint, subtalar joint, metatarsophalangeal joints, and pronation/supination. The F-Scan Mobile in-shoe pressure measurement system (Tekscan, Boston, MA, USA) was used to collect foot PPP and PGRF data during walking and/or running exercise (Chen and Bates, 2000, Orlin and McPoil, 2000). The F-Scan insole system uses resistance-based technology, and the insole consists of two polyester sheets whose inner surfaces are printed with electrical circuits. Sandwiched between the circuits is a semiconductive ink which has an electrical resistance change inversely proportional to the applied pressure. The insole consists of 960 individual resistive sensing elements arranged in a rectangular grid with a spatial density of 4 elements/cm2. Calibration was performed prior to the foot scan of each subject according to the method recommended by the F-Scan manufacturer. The F-Scan system comes with the Research 5.72 software (Tekscan, Boston. MA), which converts the pressure information of each element into the corresponding color and presents all the data as a colored image of the foot. The reliability of the F-Scan pressure sensitive insole system has been well documented by other studies (Baumhauer et al., 1997, Randolph et al., 2000, Brown et al., 1996, Hsiao et al., 2002, Kernozek et al., 1996). Ahroni et al. (Ahroni et al., 1998) reported fair to good reliability for high-pressure levels under the foot, heel, metatarsal head and hallux. Another study also suggested that with a careful procedure, the F-Scan insole system could be a useful device to measure vertical GRF during gait (Chen and Bates, 2000). Furthermore, when the applied pressure was comparable with the calibration pressure, the measurement error was in the range of 1.3-5.8% (Hsiao et al., 2002). A custom-built dual axis accelerometer (MAX2312G/M, MEMSIC, Inc, USA) with an MP150 data acquisition system (BIOPACK Systems, Inc, USA) was attached to the middle of the treadmill (95T, Lifefitness, USA) running board (figure 1) to collect PTA data during walking and/or running exercise. The accelerometer was synchronously activated by connection to an external electronic pulse with the F-Scan Mobile in-shoe pressure measurement system. The accelerometer sensor node only measured acceleration in the vertical direction, and the speed of acceleration data acquisition was set to 500 Hz (matching the F-Scan Mobile system) to collect and record the changes in acceleration during each speed test.
284 The Engineering of Sport 7 - Vol. 1
Figure 1 - A side view of the location of the accelerometer for measuring PTA on a treadmill.
Before data collection, each subject’s age, height, and weight were recorded. Each subject wore standardized socks and shoes provided by the investigators to minimize possible confounding effects of different shoes. Each subject ran on a treadmill continuously at seven different speeds: 1.3, 1.8, 2.2, 2.7, 3.1, 3.6, and 4.0 m/s. These speeds were chosen because they included slow, normal, brisk and fast speeds experienced during walking and/or running exercise (Tillman et al., 2002, Lelas et al., 2003, Segal et al., 2004, Taylor et al., 2004). The subjects were allowed to acclimate to each speed through a minimum of 10 steps before data collection, which did not begin until the subject verified that he or she was comfortable at the given speed. Ten seconds of data were then collected at each speed, and eight complete steps in the middle of each trial were identified for future analysis. The F-Scan Mobile system and accelerometer sensor node were sampled at 500 Hz for 10 seconds at each speed. None of the subjects reported discomfort due to the pressure insoles, and their movement was not affected in any way by the use of the pressure insoles. The data from the first three and the last two stances were not included in the data analysis to ensure stabilization of the subject’s performance, because changes and/or differences of pressure are known to occur upon initiating and/or terminating a walk (Randolph et al., 2000). The study demonstrated that no statistical difference was found in plantar measurements between the right and left limbs (VanZant et al., 2001). Other studies have also demonstrated no significant differences in sensorimotor function between the dominant and non-dominant limbs in healthy young adults and children (Lee and Lin, 2007, Lin et al., 2008); therefore, the PTA, PPP, and PGRF data were only collected from the dominant limb over the 8 steps, and these steps were then averaged for data analysis. Statistical analysis was performed using the Statistical Program for Social Sciences (SPSS, version 11, USA) for Windows. Repeated measures one-way ANOVA was used to identify significant differences between each speed for PTA, PPP, and PGRF. Post hoc comparisons were performed using Bonferroni-adjusted t-tests. A standard linear least squares correlation was used to calculate the coefficients of determination (r2) between PTA, PPP, and PGRF. The level of significance for all tests was set at p < 0.05.
3- Results Table 1 presents mean with standard deviation (M ± SD) values for PTA (g), PPP (pound per square inch, psi) and GRF (kg) measurements for the eight healthy participants at each speed during walking and/or running exercise. The PTA varied from subject to subject, ranging from 0.31 g to 2.57 g. The PPP during running varied from
The Engineering of Sport 7 - Vol. 1
285
subject to subject, and ranged from 57 to 247 psi. The PGRF during running varied from subject to subject, and ranged from 80.5 to 326 kgs. Table 1 - The mean and standard deviation of peak treadmill accelerations (PTA), peak plantar pressures (PPP), and peak ground reaction force (PGRF) at seven different speeds during walking and/or running exercise.
Note. a. indicates that significant differences were found when compared at speeds of 2.2, 2.7, 3.1, 3.6, and 4.0 meters per second (p<0.05). b. indicates that significant differences were found when compared at speeds of 1.3, 1.8, 3.1, 3.6, and 4.0 meters per second (p<0.05). c. indicates that significant differences were found when compared at speeds of 1.3, 1.8, 2.2, and 2.7 meters per second (p<0.05).
A repeated measures one-way ANOVA showed significant differences between each speed for PTA, PPP, and PGRF (F=60.50, 70.39, and 47.47, p<0.05, respectively). The values for PTA, PPP, and PGRF were smallest for the running speeds of 1.3 and 1.8 meters per second, and greatest at running speeds over 3.1 meters per second. In addition, the correlations (r2) between PTA and PPP, PTA and PGRF, and PPP and PGRF were 0.75, 0.78, and 0.91, respectively, and were significantly (p<.05) correlated with each other during walking and/or running exercise (figure 2).
Figure 2 - Correlations between PTA and PPP (a), PTA and PGRF (b), and PPP and PGRF (c) during different running speeds.
4- Discussions GRF is traditionally measured by force plates and generally represents barefoot measurements; however, isolated steps and barefoot measurements have some disadvantages, as they cannot properly measure and predict natural gait during dynamic locomotion. On the contrary, pressure-sensitive insole systems allow the dynamic analysis and measurement of pressures and/or forces between the foot and the ground, which could be more suitable for use in clinical, rehabilitation, and sports fields. The purpose of this study was to examine the relationship between PTA, PPP and PGRF during walking and/or running exercise. The results of this study demonstrated that the PTA, PPP, and
286 The Engineering of Sport 7 - Vol. 1 PGRF increased with increasing walking and/or running speeds, and there were strong correlations between PPP and PTA, PTA and PGRF, and PPP and PGRF for each walking and/or running speed. Another interesting result is that relative to the linear increase in speed during walking and/or running exercise, the changes in the PTA, PPP, and PGRF have a non-linear response pattern. The results of this study are consistent with other studies which have demonstrated that the plantar pressure and ground reaction forces at the heel increased linearly as gait speed increased in walking (Burnfield et al., 2004, Kernozek and Zimmer, 2000, Menz and Morris, 2006, Segal et al., 2004, Warren et al., 2004). However, this study further examined the plantar pressures and forces during running and thus might provide more extensive information for further research and clinical application. In addition, the effect of cadence on plantar pressures was also examined, and showed that as walking cadence increases, pressure-time integrals and foot-to-floor contact durations decrease, and peak plantar pressure increases (Zhu et al., 1995). It is believed that the changes in plantar pressure parameters with increasing cadence or velocity can be explained by Newton’s Second Law: force = mass acceleration, where the acceleration is the derivative of velocity with respect to time (dV/dt). Furthermore, similar to the results of some recent studies (Segal et al., 2004, Taylor et al., 2004), this study also demonstrated that despite linear increases in speed and regardless of experimental settings, many of the changes in plantar parameters occurred between walking (1.1-1.8 m/s), normal running (2.2-3.1 m/s), and fast (3.6-4.0 m/s) running. This suggests that the motion of the foot has a non-linear response to increases in velocity (Figure 3 and Table 1), i.e., fast walking is not simply a faster version of normal walking and fast running is not simply a faster version of normal running. Increased GRFs and decreased knee flexion angles during contact have been associated with increased incidence of ACL injuries (Chappell et al., 2002), and specific training programs have been devised to help athletes reduce their GRFs at contact (Myer et al., 2006). Therefore, when walking or running at a fast speed, specific changes to the movement patterns of the lower limbs (such as hip and knee flexion, tibial rotation and foot pronation) may occur that are not evident in the slow-to-normal speed transition. In-shoe plantar pressure measurement systems are capable of measuring pressures at the interface between the shoe and the foot, and are useful in clinical, rehabilitation, and sports fields for assessing patterns of forces between the agent and the ground (Santos et al., 2001). Significant differences in plantar pressure patterns have been reported in people with diabetic neuropathy, leprosy, and rheumatoid arthritis compared to healthy subjects (Caselli et al., 2002, Frykberg et al., 1998), and numerous studies have employed plantar pressure technology to demonstrate significant changes in foot function when wearing orthoses. The F-Scan insole system utilizes an ultrathin (0.18 mm) pressure sensitive sensor. Each sensing trace is coated with pressure-sensitive, resistive ink such that a sensing cell is created at each grid crossing point. The resistance for each cell is inversely proportional to the applied surface pressure. By scanning the grid and measuring the resistance at each intersection, the pressure distribution on the sensor surface can be determined. The insole sensor consists of a Mylar substrate with 960 individual pressure cells. The pres-
The Engineering of Sport 7 - Vol. 1
287
sure-sensing points are evenly distributed at 5 mm intervals. The sensing range of each individual cell is 8-124 pounds per square inch (psi). Unlike the forceplate, which is factory calibrated, the F-Scan insole system requires a standard procedure to calibrate the sensors with the subject’s body weight by collecting data from a single leg stance on each insole sensor. The F-Scan system utilizes a calibration factor to convert the raw FScan unit to a selected pressure unit. During walking locomotion, foot pressure is generated as body weight (BW), and is transferred onto the stance limb. Initially, BW is loaded solely on the heel region, resulting in high peak pressures in this area (Eils et al., 2002, Rosenbaum et al., 1994). Afterward, peak heel pressure dissipates once the forefoot contacts the ground (foot flat), and BW is distributed over a larger surface area. In late stance, as the body progresses anterior to the ankle joint and the heel elevates from the ground, the plantar force is once again concentrated over a relatively small region, the forefoot (Eils et al., 2002). High forefoot pressures in late stance are a consequence of this posture (Rosenbaum et al., 1994). Even the loading of the heel is strongly influenced by events prior to heel-strike (Morag and Cavanagh, 1999), and walking speeds are strongly associated with higher pressures, primarily due to increased peak force values during rapid locomotion (Burnfield et al., 2004). GRFs are influenced by the entire BW rather than the impacted segments, and the impact conditions that can be studied are restricted because the foot must land on a force platform. The accelerometer, attached in the center/middle of the running board in this study, should be more versatile with regard to the impact conditions that can be evaluated, and it is possible to measure a more localized effect of the impact (Derrick and Mercer, 2004, Elvin et al., 2007a, Elvin et al., 2007b). During a typical heel-toe running cycle, the foot impacts the ground and causes a rapid increase in the vertical GRF that reaches a maximum after heel contact. This impact force accelerates a portion of the lower extremity, also resulting in peak impact acceleration. The results of this study also demonstrated that the correlations between PTA, PPP and PGRF were highest for PPP and PGRF (r2=0.91), then for PTA and PGRF (r2=0.78), and finally for PTA and PPP (r2=0.75). It is likely that impacts only contribute to injury when they are coupled with items such as abnormal anatomy or kinematics, excessive duration or inadequate rest between bouts (Derrick et al., 2002). In addition, the acceleration experienced by the body segments depends on the magnitude of GRF, and the damping effects of the body shock absorbers (Lafortune et al., 1995) provide a possible explanation of the correlations between PTA and PGRF (r2=0.78) (p<0.05). Although the location of the accelerometer was different from other studies, the results of this study are consistent with the findings of other studies; for example, the GRFs were correlated with leg accelerations during running (Derrick and Mercer, 2004, Lafortune et al., 1995) and vertical jumping (Elvin et al., 2007a, Elvin et al., 2007b). Therefore, the attachment of an accelerometer to the treadmill running board could provide a valuable in-field tool and technique for estimating and/or detecting GRF and/or PPP during treadmill running exercise, especially for measuring and detecting abnormal impacts for injury prevention and rehabilitation.
288 The Engineering of Sport 7 - Vol. 1 In conclusion, this study demonstrated that PTA, PPP, and PGRF increase with increased walking and/or running speeds, and there are strong correlations between PPP and PTA, PTA and PGRF, and PPP and PGRF. The change in PTA, PPP, and PGRF during running exercise has a non-linear response to a linear increase in velocity and a 2nd order polynomial would better fit the data. Therefore, onboard treadmill accelerometers could be used to estimate the PPP and PGRF during running exercise in healthy young adults. It is suggested that the experiment should be repeated for all types of socks/shoes-combinations in order to calculate the best regression line.
5- References [AB1] Ahroni JH, Boyko EJ, and Forsberg R. Reliability of F-scan in-shoe measurements of plantar pressure. Foot Ankle Int 19: 668-673, 1998. [BW1] Baumhauer JF, Wervey R, McWilliams J, Harris GF, and Shereff MJ. A comparison study of plantar foot pressure in a standardized shoe, total contact cast, and prefabricated pneumatic walking brace. Foot Ankle Int 18: 26-33, 1997. [BR1] Brown M, Rudicel S, and Esquenazi A. Measurement of dynamic pressures at the shoe-foot interface during normal walking with various foot orthoses using the FSCAN system. Foot Ankle Int 17: 152-156, 1996. [BF1] Burnfield JM, Few CD, Mohamed OS, and Perry J. The influence of walking speed and footwear on plantar pressures in older adults. Clin Biomech (Bristol, Avon) 19: 78-84, 2004. [CP1] Caselli A, Pham H, Giurini JM, Armstrong DG, and Veves A. The forefoot-to-rearfoot plantar pressure ratio is increased in severe diabetic neuropathy and can predict foot ulceration. Diabetes Care 25: 1066-1071, 2002. [CY1] Chappell JD, Yu B, Kirkendall DT, and Garrett WE. A comparison of knee kinetics between male and female recreational athletes in stop-jump tasks. Am J Sports Med 30: 261-267, 2002. [CB1] Chen B, and Bates BT. Comparing of F-Scan in-sole and AMTI forceplate system in measuring vertical ground reaction force during gait. Physiotherapy Theory and Practice 16: 43-53, 2000. [DT1] Decker MJ, Torry MR, Wyland DJ, Sterett WI, and Richard Steadman J. Gender differences in lower extremity kinematics, kinetics and energy absorption during landing. Clin Biomech (Bristol, Avon) 18: 662-669, 2003. [DD1] Derrick TR, Dereu D, and McLean SP. Impacts and kinematic adjustments during an exhaustive run. Med Sci Sports Exerc 34: 998-1002, 2002. [DM1] Derrick TR, and Mercer JA. Ground/foot impacts: measurement, attenuation, and consequences. Med Sci Sports Exerc 36: 830-831, 2004. [EN1] Eils E, Nolte S, Tewes M, Thorwesten L, Volker K, and Rosenbaum D. Modified pressure distribution patterns in walking following reduction of plantar sensation. J Biomech 35: 13071313, 2002. [EE1] Elvin NG, Elvin AA, and Arnoczky SP. Correlation between ground reaction force and tibial acceleration in vertical jumping. J Appl Biomech 23: 180-189, 2007. [EE2] Elvin NG, Elvin AA, Arnoczky SP, and Torry MR. The correlation of segment accelerations and impact forces with knee angle in jump landing. J Appl Biomech 23: 203-212, 2007. [FL1] Frykberg RG, Lavery LA, Pham H, Harvey C, Harkless L, and Veves A. Role of neuropathy and high foot pressures in diabetic foot ulceration. Diabetes Care 21: 1714-1719, 1998. [H1] Hreljac A. Impact and overuse injuries in runners. Med Sci Sports Exerc 36: 845-849, 2004.
The Engineering of Sport 7 - Vol. 1
289
[HG1] Hsiao H, Guan J, and Weatherly M. Accuracy and precision of two in-shoe pressure measurement systems. Ergonomics 45: 537-555, 2002. [HR1] Hurwitz DE, Ryals AR, Block JA, Sharma L, Schnitzer TJ, and Andriacchi TP. Knee pain and joint loading in subjects with osteoarthritis of the knee. J Orthop Res 18: 572-579, 2000. [KL1] Kernozek TW, LaMott EE, and Dancisak MJ. Reliability of an in-shoe pressure measurement system during treadmill walking. Foot Ankle Int 17: 204-209, 1996. [KZ1] Kernozek TW, and Zimmer KA. Reliability and running speed effects of in-shoe loading measurements during slow treadmill running. Foot Ankle Int 21: 749-752, 2000. [LL1] Lafortune MA, Lake MJ, and Hennig E. Transfer function between tibial acceleration and ground reaction force. J Biomech 28: 113-117, 1995. [LL1] Lee AJ, and Lin WH. The influence of gender and somatotype on single-leg upright standing postural stability in children. J Appl Biomech 23: 173-179, 2007. [LM1] Lelas JL, Merriman GJ, Riley PO, and Kerrigan DC. Predicting peak kinematic and kinetic parameters from gait speed. Gait Posture 17: 106-112, 2003. [LR1] Liddle D, Rome K, and Howe T. Vertical ground reaction forces in patients with unilateral plantar heel pain - a pilot study. Gait Posture 11: 62-66, 2000. [LL1] Lin WH, Liu YF, Hsieh CC, and Lee AJ. Ankle eversion to inversion strength ratio and static balance control in the dominant and non-dominant limbs of young adults. J Sci Med Sport 2008. [MM1] Menz HB, and Morris ME. Clinical determinants of plantar forces and pressures during walking in older people. Gait Posture 24: 229-236, 2006. [MC1] Morag E, and Cavanagh PR. Structural and functional predictors of regional peak pressures under the foot during walking. J Biomech 32: 359-370, 1999. [MC1] Murphy DF, Connolly DA, and Beynnon BD. Risk factors for lower extremity injury: a review of the literature. Br J Sports Med 37: 13-29, 2003. [MF1] Myer GD, Ford KR, McLean SG, and Hewett TE. The effects of plyometric versus dynamic stabilization and balance training on lower extremity biomechanics. Am J Sports Med 34: 445455, 2006. [OM1] Orlin MN, and McPoil TG. Plantar pressure assessment. Phys Ther 80: 399-409, 2000. [PS1] Pflum MA, Shelburne KB, Torry MR, Decker MJ, and Pandy MG. Model prediction of anterior cruciate ligament force during drop-landings. Med Sci Sports Exerc 36: 1949-1958, 2004. [RN1] Randolph AL, Nelson M, Akkapeddi S, Levin A, and Alexandrescu R. Reliability of measurements of pressures applied on the foot during walking by a computerized insole sensor system. Arch Phys Med Rehabil 81: 573-578, 2000. [RA1] Richards DP, Ajemian SV, Wiley JP, and Zernicke RF. Knee joint dynamics predict patellar tendinitis in elite volleyball players. Am J Sports Med 24: 676-683, 1996. [R1] Rodgers MM. Dynamic foot biomechanics. J Orthop Sports Phys Ther 21: 306-316, 1995. [RH1] Rosenbaum D, Hautmann S, Gold M, and Claes L. Effects of walking speed on plantar pressure patterns and hindfoot angular motion. Gait and Posture 2: 191-197, 1994. [SV1] Santos-Rocha R, and Veloso A. Comparative study of plantar pressure during step exercise in different floor conditions. J Appl Biomech 23: 162-168, 2007. [SC1] Santos D, Carline T, Flynn L, Pitman D, Feeney D, Patterson C, and Westland E. Distribution of in-sole dynamic plantar foot pressures in professional football players. Foot 11: 10-14, 2001. [SR1] Segal A, Rohr E, Orendurff M, Shofer J, O’Brien M, and Sangeorzan B. The effect of walking speed on peak plantar pressure. Foot Ankle Int 25: 926-933, 2004.
290 The Engineering of Sport 7 - Vol. 1 [TM1] Taylor AJ, Menz HB, and Keenan AM. Effects of experimentally induced plantar insensitivity on forces and pressures under the foot during normal walking. Gait Posture 20: 232-237, 2004. [TF1] Tillman MD, Fiolkowski P, Bauer JA, and Reisinger KD. In-shoe plantar measurements during running on different surfaces: changes in temporal and kinetic parameters. Sports Engineering 5: 121-128, 2002. [VM1] VanZant RS, McPoil TG, and Cornwall MW. Symmetry of plantar pressures and vertical forces in healthy subjects during walking. J Am Podiatr Med Assoc 91: 337-342, 2001. [VM1] Veves A, Murray HJ, Young MJ, and Boulton AJ. The risk of foot ulceration in diabetic patients with high foot pressure: a prospective study. Diabetologia 35: 660-663, 1992. [WM1] Warren GL, Maher RM, and Higbie EJ. Temporal patterns of plantar pressures and lowerleg muscle activity during walking: effect of speed. Gait Posture 19: 91-100, 2004. [ZW1] Zhu H, Wertsch JJ, Harris GF, and Alba HM. Walking cadence effect on plantar pressures. Arch Phys Med Rehabil 76: 1000-1005, 1995.
Development of Immediate Feedback Software for Optimising Glide Performance and Time of Initiating Post-Glide Actions (P56) Roozbeh Naemi1, Serdar Aritan2, Simon Goodwill3, Steve Haake4, Ross Sanders5
Topics: Performance Sports, Biomechanics, Measurement Systems. Abstract: Performance in starts and turns is a major contributor to success in swimming and is influenced greatly by the glide efficiency and the timing of commencing the post-glide action (including kick in all strokes and the underwater pull in breaststroke starts and turns). The main aim of this research is to develop and test ‘user friendly’ software for providing immediate feedback to swimmers and coaches to optimise glide performance and time of initiating post-glide actions in starts, turns, and the glide phase of the breaststroke. The developed software reads the video file of a glide followed by the post-glide action (PGA) gathered from a single underwater camera positioned perpendicular to the swimmer’s trajectory path. A set of body markers are digitised to provide the records of the displacement data. The mathematical model based on the ‘hydro-kinematic method’ (Naemi, 2007) is coded in MATLAB to calculate the glide efficiency parameters, reconstructed instantaneous velocity, and the actual and optimal times and distances of commencing post-glide action using the raw displacement. A Graphic User Interface with user friendly icon based system enables the results to appear in an aesthetic and effective screen display, which includes a video replay, displacement and velocity graphs, and tabulated results. Early tests revealed that the effectiveness of the ‘Hydro-kinematic’ method in ‘fine-tuning’ performance can potentially be improved by the developed software that enables rapid turnaround of results. The accompanying video replay enables qualitative assessment of postures and orientations enabling refinement in subsequent trials. The information from the software is also particularly important in predicting the exact timing for initiating the post glide action for a particular swimmer with a distinct glide efficiency and post-glide performance. Keywords: Glide, PGA Timing, Immediate feedback, Glide Efficiency, Streamlining. 1. Centre for Aquatics Research and Education, The University of Edinburgh, St Leonard’s Land, Holyrood Road, Edinburgh, UK EH8 8AQ - E-mail: [email protected] 2. Biomechanics Research Group, School of Sports Sciences and Technology, Beytepe, 06800,Ankara, Turkey E-mail : [email protected] 3. Sports Engineering @ Centre for Sport and Exercise Science Sheffield Hallam University, Collegiate Hall Sheffield, UK S10 2BP - E-mail: [email protected] 4. Sports Engineering @ Centre for Sport and Exercise Sciences Sheffield Hallam University, Collegiate Crescent, Sheffield, UK S10 2BP - E-mail: [email protected] 5. Centre for Aquatics Research and Education, The University of Edinburgh, St Leonard’s Land Holyrood Road, Edinburgh, UK EH8 8AQ - E-mail: [email protected]
292 The Engineering of Sport 7 - Vol. 1
1- Introduction Performance in starts and turns is a major contributor to success in swimming (e.g. Chatard et al. 1990) and is influenced greatly by the glide performance and the timing of commencing the post-glide actions (including kicking in all strokes and the underwater pull in breaststroke starts and turns).The average velocity over a glide as the indicator of performance is dependent on the initial velocity of glide (release velocity from the wall after push-off) as well as to the changes in velocity over the glide period. The characteristic of body that determines its ability to maintain its velocity and to avoid retardation over time (deceleration) is known as glide efficiency (Naemi 2007). While the initial velocity of a glide is related to the preceding action, such as pushing off the wall in turns, the glide efficiency is dependent on both the virtual mass of the swimmer, being the sum of body mass and the added mass of water, and the passive drag. Due to the differences in the swimmer’s posture during a real glide compared to towing at constant velocity and the inherent difference between the steady state and transient flow conditions, the drag forces found based on constant velocity towing methods can not represent the glide efficiency in an actual glide under real conditions in which velocity decreases . For example, based on the fact that larger and heavier swimmers experience higher values of passive drag, Benjanuvatra et al. (2001) concluded that this type of swimmer decelerates faster during glides. This suggestion did not take into account the fact that the heavier swimmers have higher inertia that reduces the rate of deceleration during the glide compared to lighter swimmers. As the added mass of water for human body during deceleration can not be determined with theoretical and experimental approach, the glide efficiency of body could not be calculated. Recently an innovative method known as ‘Hydro-Kinemtaic’ was proposed by Naemi (2007) that quantifies the glide efficiency in an accurate and reliable way using a ‘Hydro-Kinematic’ Method (HKM) (Naemi 2007). The HKM requires only a single underwater camera positioned perpendicular to the swimmer’s glide plane, any 2D digitising software, and the mathematical model (Naemi 2007). The glide efficiency of body depends on the drag force which is affected by the shape and alignment of a body. Thus to maximise glide efficiency one should adopt the most streamlined position, that can be achieved by adopting a body alignment with a minimum angle of attack and maintaining an appropriate posture with suitable joint angles. In addition to the need to increase glide performance, it is highly desirable to optimise timing of the post-glide phase. It has been established that maintaining a passive streamlined posture at higher velocities than a critical value is more beneficial than an active propulsive movement (Sanders and Byatt-Smith, 2001). This has important implications as a swimmer can lose speed unnecessarily if the passive glide is finished prior to or continued beyond an optimum speed. In these cases, additional energy and time is required to regain the optimal velocity. By knowing the initial velocity and glide efficiency, an accurate record of instantaneous velocity can be reconstructed for a glide trial. It is obvious that the instantaneous velocity of a body during the whole period of a glide can be directly calculated from the
The Engineering of Sport 7 - Vol. 1
293
raw position data. However, due to variations in the velocity as a result of instabilities in flow (Howe et al. 2001, Wang et al. 2003) that cannot be distinguished from noise, the determined instantaneous velocity data includes excess noise and may not be valid for comparison (Klauck and Daniel, 1976, Bilo and Nachtigall, 1980, Naemi and Sanders, 2004). The HKM overcomes that problem by enabling precise reconstruction of the velocity variations of the body during a glide. This information is particularly important in predicting the exact timing for initiating the post-glide action for a particular swimmer with a distinct glide efficiency and post-glide performance. For example for each subject the glide should end when the velocity reaches the velocity which can be sustained by the post-glide actions. Although the HKM developed by Naemi (2007) can readily determine the glide efficiency and the time of initiating the PGA, its use in the field was limited by the time required to conduct the analysis (two hours for each trial). It was realised that the effectiveness of the method in ‘fine-tuning’ performance could be increased greatly by developing software that enables rapid turnaround of results so that swimmers can respond to feedback and adjust technique immediately. The main aim of this project was to develop and test ‘user friendly’ software for providing immediate feedback to swimmers and coaches to optimise glide efficiency and time of initiating post-glide actions using ‘Hydro-kinematic’ method. This could have particular applications in improving the performance of starts, turns, and breaststroke mid-pool swim.
2- GlideCoach Software The software developed for this purpose ‘GlideCoach’, comprises MATLAB routines for digitising body landmarks and calculating the variables of interest via a Visual Basic user interface. The interface is designed in an aesthetic and user-friendly fashion resembling successful commercial software in the field of sports performance. The digitising module incorporates a tracking algorithm coded specifically for tracking markers underwater. It enables automatic or semi-automatic (allowing user intervention when the point being tracked is ‘lost’) tracking of markers. The calculation routines employ advanced curve fitting and optimisation techniques to calculate all the parameters related to glide performance and post-glide action timing based on the ‘Hydro-kinematic’ method (Naemi 2007). The parameters of interest are: The parameters related to glide performance including Glide Factor as a measure of glide efficiency as well as the Initial Velocity as a measure of push-off performance. Also an average velocity as a measure of glide performance was calculated based on the values of glide factor and initial velocity that incorporated both the push-off performance and glide efficiency. In addition to these the reconstructed instantaneous glide velocity as well as the sustainable PGA velocity were calculated. This information was used to calculate optimal and current velocity, displacement and time of initiating the PGA.
294 The Engineering of Sport 7 - Vol. 1
2.1 Glide Performance and PGA Timing modes (stages) The software comprises two modes - ‘Glide Performance’ and ‘PGA Timing’ to quantify parameters related to glide performance and timing of initiating PGAs respectively. Having two distinct modes in the software enables the user to focus on improving performance of each independently. The usual sequence is to use the ‘Glide Performance’ mode first as appropriate PGA timing varies according to the level of glide performance.
2.2 Preparation of subjects Prior to the pool test, each subject is marked with the permanent black markers at five anatomical landmarks including the styloid process of the ulna, head of the humerus, greater trochanter of the femur, level of the most lateral side of the patella (the lateral epicondyle of the femur), and the lateral malleolus of the fibula. These were marked on the left side of the body to represent the joint centres at the wrist, shoulder, hip, knee and the ankle. A marker on the left side of the head at the eye level on the side of the goggle was used to assess consistency in the head angle during the trials. The markings facilitated the process of automatic tracking as a faster and more desirable option than the manual digitizing. The subject was then instructed to push off the wall and to adopt a streamlined position during the glides. A submersion, preparation and push-off protocol was taught to the subject in order to facilitate glide followed by PGA in a horizontal rectilinear path so that to enable the HKM to be used to quantify parameters related to glide performance and PGA timing in each mode (Naemi 2007).
2.3 Filming Requirements AVI files of the swimmer performing passive glides only (for Glide Performance mode) or a passive glide followed by a PGA (for PGA Timing mode) are input to the software. For the calculations to be accurate the followings are required: 1- A camera sampling at either 25 frames per second, (or 50 fields in case of interlaced video) needs to be positioned underwater 10 m from the subject’s intended glide path with its axis perpendicular to the intended glide path. 2- In the ‘Glide Performance’ mode the camera needs to be zoomed to a field of view of approximately 6 meters in the direction of the glide to ensue that it covers adequate duration (1.5 s) of glide. Although at the beginning of each session the camera view should be adjusted by some trials to make sure that adequate glide duration of 1.5 second is covered. 3-In ‘PGA Timing’ the camera needs to be adjusted to a field of view that covers the swimmers glide and PGA of adequate cycles (3 kick cycles in front crawl, butterfly and backstroke, or a complete long pull in breaststroke). In this case the field of view needs to be approximately 9m wide. This needs to be specifically adjusted for a swimmer depending on the level of performance and height. 4-A second camera needs to be used on a boom at a 5 m height directly above the subject’s line of glide (or PGA) with its axis perpendicular to the intended glide path.
The Engineering of Sport 7 - Vol. 1
295
This needs to make sure that the subject does not deviate from the ‘T line’ over the period of glide or glide and PGA.
2.4 Capturing and importing The recorded video can be imported to the software. The details are filled in for each subject including name, session date, time, the trial number as well as the sampling frequency of the camera. This information is stored in a database that allows comparison of performance within subject, which allows the current performance of swimmer to be compared with the best and previous performance to assess the improvement and set targets. This feature will be further explained in the ‘Results’.
2.5 Playing and trimming In the player the video of an untrimmed glide including submersion, preparation, pushoff and glide (together with PGA in PGA Timing) can be played. The video control panel both allows scrolling and moving frame by frame that facilitates finding the exact frame to start the trim (i.e. from toe-off). The end of trim will be chosen automatically as the frame that comes in 1.5 seconds after the ‘start trim’ frame. As in order to be able to compare the glide performances across trials 1.5 s glide duration was chosen based on the average glide duration after turns for swimmers at elite level. In ‘PGA Timing’ the end trim will be determined as the end of three cycles of kick or the end of long pull in breaststroke. This is the end of the third cycle for kick that is at the start of the forth cycle just before when the legs start to come up for starting the forth cycle. In breaststroke the end of long pull determined as when the swimmers hand reaches the hip marker, is considered as the end trim point. Another point known as ‘switch point’ that is the starts of PGA is determined. The switch point is the start of the first cycle that is determined as the moment when the legs tend to come up (i.e. hyperextension, or knee flexion) in cases where kicking is used as a PGA. In breaststroke when long pull is used as a PGA, the switch point is when hands get apart. The playback feature allows playing both the video at different speeds including slow and fast forward that is particularly useful for visual feedback.
2.6 Analysing The trimmed video is digitized in the ‘Analyser’ that allows both manual and automatic tracking of markers. First the number of joints that needs to be digitised is chosen in the ‘Digitising Points’ panel. The zoom in feature provides a larger view of the skin markers that facilitates exact positioning of the digitising points on the body as well as monitoring the accuracy of tracking in all frames. The hip marker’s horizontal position is found to be more stable compared to other joint markers when the body alignment changes over a glide period. Thus the glide factor calculated based on the hip marker is more reliable in reflecting the changes to the body alignment (and angle of attack). For these reasons the hip’s horizontal position data was used for calculating all the parameters related to glide and PGA. Hip marker has also the advantage of being visible from camera view in PGA of all four strokes while
296 The Engineering of Sport 7 - Vol. 1 remains fairly stable in vertical direction. Therefore all calculations related PGA Timing were based on the hip marker position data which assures consistency of calculations. The hip marker is the minimum that needs to be digitised at all trials. Digitising other markers can be done for finding the joint angles, a feature that is going to be added to the software in future version. After the position of hip was pointed at in the first frame of the trimmed video, the auto analyse next frame button will send the initial positions of the markers in the initial frame as well as the current and the next frames to a tracking code.
2.7 Tracking Code The main problem is to determine the position of a given pattern in an image, so called region of interest (ROI), which in this case, ROI is the skin markers. One basic approach that can be used is template matching. This means that the position of the given pattern is determined by a pixel-wise comparison of the image with a given template that contains the desired pattern. We used normalised cross-correlation (NCC) in our tracking algorithm. In order to utilise the algorithm a code is written in MATLAB that provides the position of the marker in the next frame, knowing the initial marker position using images from the current and the next frame. The same procedure is followed up to the last frame. As the tracking code was specifically developed for underwater video, the code is more reliable tracking than the commercial software available. In early pilot works it was found that tracking frame by frame needs just some minor manual corrections in some cases. Further developments are underway to enable full automatic tracking that needs no intervention in the future version.
2.8 Calibration The tracked coordinates then are converted to the real position data using the twodimensional Direct Linear Transformation (DLT) techniques (Walton, 1981). The Calibration panel allows a numbers of reference points on the ‘T’ line at the bottom of the pool and on the push-off wall with known coordinates to be digitized and used for calibration Also a feature in the calibration panel allows the use of a previously made calibration file to make calibration easier in consequent testing when the camera is fixed between trials.
2.9 Calculation Core Code The digitized coordinates then will be sent to two Matlab codes developed separately for Glide Performance and PGA Timing.
2.9.1 Glide Performance Core Code The Glide performance core code receives the position data of glide along with the frequency at which the data was collected at. The code calculates the glide efficiency parameters based on the ‘Hydro-kinematic’ method (Naemi 2007). This is achieved by
The Engineering of Sport 7 - Vol. 1
297
fitting a logarithmic function that represents the position changes of a representative body to the raw position data of the swimmer’s body during a rectilinear glide. As the logarithmic function was derived from the equation of motion of the body during a horizontal rectilinear glide, the values of the curve parameter that provides the best fit for the data determine glide performance parameters. These include Glide Factor (m) and Initial Velocity (m/s) from which an average velocity can be calculated.
2.9.2 PGA Timing Core Code In PGA Timing module in addition to the displacement, and video frequency, the frame number for switching from glide to PGA is sent to the PGA Timing code. In PGA Timing code the raw position data corresponding to glide (the data sequence up to the switch point) is used to calculate the glide parameters including the Glide factor, initial velocity and average velocity using Hydro-kinematic method (Naemi 2007). The raw position sequence corresponding to PGA performance is used to calculate the sustainable velocity of PGA. Combining the reconstructed instantaneous velocity of glide that can be obtained based the initial velocity and glide factor, and the sustainable PGA velocity, the time at which the glide velocity reaches a velocity that can be sustained by kick is quantified. Based on these the parameters related to timing of PGA that includes current and optimal velocity, time and distance of initiating the PGA are calculated (Naemi 2007).
2.10 Results The results in each mode are presented separately.
2.10.1 Glide Performance. In the results panel in addition to the video and the parameters related to glide including Glide Factor, Initial Velocity and Average Velocity are displayed (Figure 1- left). The accompanying video replay enables qualitative assessment of postures and orientations enabling refinement in subsequent trials.
Figure 1- Screen shots of the results tab in the Glide Performance (left) and PGA Timing mode (right).
298 The Engineering of Sport 7 - Vol. 1 As the higher the glide efficiency is an indicator of a better streamlining position including body posture and alignment (segmental angle of attack), the best streamlined position of the swimmer is achieved using a trial and error approach. In this approach the swimmer is advised to adopt different joint angles and segmental angle of attack, from which the best streamlined position can be chosen based on having highest glide efficiency. In addition to glide factor, the initial glide velocity (wall release velocity) is displayed that represents the effectiveness of push-off both in terms of the push-off force and proper positioning of upper body and trunk during the push-off itself. As the overall glide performance is dependent both on the initial velocity as well as the glide efficiency, the average velocity is calculated and displayed as an indicator of the overall glide performance. The best glide performance as the one with highest average velocity, along with the previous performance appears in a table to allow monitoring progress and setting targets. In case of two glide trails with the same average velocity, the one which has higher glide factor is chosen as a better one, as the advantage was gained with less physiological cost (Naemi 2007) All the three parameters related to glide performance are displayed in an average velocity contour graph with initial velocity and glide factor displayed in the horizontal and vertical axes respectively (Figure 1- left). The contour graph is coloured in shades ranging from red (dark in grayscale) at lower left corner that represent poor performance, to green ( light in grayscale ) at upper right corner that represents high performances. This facilitates initial interpretation of results and enables the coach to determine areas that need to be worked on in order to increase the glide performance e.g. push-off or streamlining.
2.10.2 PGA Timing In PGA Timing in addition to the parameters related to glide performance of the glide part of the trial, the time difference between the current and optimal time and distance (from wall) of initiating PGA is to advise swimmer on either delaying the initiation of PGA to a later time, or initiating it earlier than the current time. The better trial is determined as the one which has close to zero time/distance difference. A criterion for the overall glide and PGA is under development that is intended to appear in overall performance bar in future version. In addition to these a velocity-time diagram of the performance is shown in the ‘Results’ section that allows the changes of velocity over the period of glide as well as the current and optimal velocities of initiating PGA to be displayed graphically (figure 1right). Two dashed lines show the current and optimal time of initiating PGA on the time axis. Two distance lines were also reproduced on the video in the positions corresponding to real distances to allow the swimmer to realize the optimal and current positions of initiating the PGA. The PGA Timing panel allows displaying both the optimal and the current time, velocity and distance in a table that facilitates more insight into the timing parameters. As it can be seen in this sample the PGA was initiated 0.2 sec earlier than an optimal time. The kicking over this 0.2 sec causes the velocity to drop faster as compared to if the swimmer continues the passive glide. This is due to the fact that the net decelerating
The Engineering of Sport 7 - Vol. 1
299
force (the difference between the active drag and the propulsive force) is greater than the passive drag for a gliding position (Lyttle et al. 2000).
3- Discussion and further developments The development of the ‘GlideCoach’ software provides for the first time the ability to apply the Hydro-kinematic method to improve the streamlining posture and optimising the time of initiating PGA on the poolside. Initial pilot works with the software revealed that feedback to the swimmer can be provided in less than 5 minutes from the start of the trial. This enables the swimmer to be benefited from the developed method without being affected by the prolonged procedures of trimming, tracking, digitising, importing and exporting files from and to different packages which was approximated to last more than two hours for each trial not using the software. The software is also one of the few in the field that provides immediate quantitative feedback on the kinematic parameters related to performance and indeed the only available software in the field of sports performance that provides immediate feedback on a complex parameter such as glide factor that has both a kinematic and a hydrodynamic entity. Further developments are underway specifically with regards to fine tuning the tracking code so that the tracking could be done with minimal manual intervention. Also for the future version the values of joint angles is going to be included in the result section of Glide Performance software. Further improvement in the software is planned after its use for a larger numbers of swimmers on the poolside. In addition providing overall performance in PGA Timing and adding other features like calculating the ‘Glide Coefficient’ (Naemi 2007) that allows comparison between subjects in terms of their glide efficiency can be added to the software.
4- Conclusion The ‘GlideCoach’ software was developed that provides immediate feedback on the glide efficiency and timing of PGA initiation using ‘Hydro-kinematic’ method. The software was developed as a ‘user friendly’ menu driven interface. The auto-tracking algorithm was developed in Matlab using advance image processing technique that enabled tracking markers to be fast and reliable. The calculations were based on the Hydro-kinematic method and done in two core codes for each mode in Matlab that provides the important parameters related to glide performance and timing of PGA initiation in each mode. Output the results in an aesthetic and effective screen display that include a video replay, graphs, and tabulated results can potentially improve the effectiveness of the software and its use in the field. By facilitating immediate feedback in applying the ‘Hydro-kinematic’ method to quantify the swimmer’s glide performance and timing of PGA initiation, the ‘GlideCoach’ software can provide an opportunity, for the first time, to ‘fine tune’ swimmers’ streamlining posture as well as to optimise timing of initiating postglide action in the field.
300 The Engineering of Sport 7 - Vol. 1
5- Acknowledgement We thank funding support provided through EPSRC and UK-Sport as well as the support of Scott Drawer, Fred Vergnoux, Ian Wright, Andrei Vorontsov, Tom Bruce, Brian Blanksby, Gareth Irwin, Bill Easson, Barry Wilson and Khelil Sefiane.
6- References [BN1] Bilo D. and Nachtigall W. . A simple method to determine drag coefficients in aquatic animals. In Journal of Experimental Biology, 87: 357-359, 1980 [BB1] Benjanuvatra N., Blanksby B. A., and Elliott B.C. Morphology and hydrodynamic resistance in young swimmers. In Paediatric Exercise Science, 13(3): 246-255, 2001 [CL1] Chatard J. C., Lavoie, J. M. Bourgoin B.,and Lacour J. R. The contribution of passive drag as a determinant of swimming performance. In International Journal of Sports Medicine, 11(5): 367-372, 1990 [HL1] Howe M.S., Lauchle G. C. and Wang J. Aerodynamic lift and drag fluctuations of a sphere. In Journal of Fluid Mechanics 436: 41–57, 2001 [KD1] Klauck J. and Daniel K. Determination of man’s drag coefficient and effective propelling forces in swimming by means of chronocyclography. In Biomechanics VB, Intentional Ser. on Biomechanics. Vol. 1 B: 250-257, 1976 [LB1] Lyttle A. D., Blanksby B. A., Elliott B. C., and Lloyd, D. G. Net forces during tethered simulation of underwater streamlined gliding and kicking techniques of the freestyle turn. In Journal of Sport Sciences, 18: 801-807, 2000. [N1] R. Naemi. A Hydro-kinematic method for quantifying glide efficiency of swimmers. Doctoral Thesis, The University of Edinburgh, 2007. [NS1] R. Naemi and R.H. Sanders. A comparison of two functions representing velocity of a human body subject to passive drag. International Symposium on Biomechanics in Sports 2004, Proceedings of conference, 2004. [SB1] R.H. Sanders and J. Byatt-Smith. Improving feedback on swimming turns and starts exponentially. International Symposium on Biomechanics in Sports 2001, Proceedings of Swim Sessions, 2001. [W1] J.S. Walton. Close-range Cine-photogrammetry: A generalized techniques for quantifying gross human motion. Masters dissertation, Pennsylvania State University, 1981. [WL1] Wang J., Lauchle G. C., and Howe M. S. Flow-induced force fluctuations on a sphere at high Strouhal number. In Journal of Fluid and Structure. 17: 365-380, 2003
Rod Response Analysis to Fish Bite Based on Multi-link Model Solved by Lower Triangularization of Sparse Symmetric Coefficient-matrix (57) Shigeyuki Yamabe1, Hiromitsu Kumamoto1, Shingo Nishioka1
Topics: Fishing. Abstract: There have been many studies on dynamic behavior of aerial lines for fly fishing, whereas few studies have been made for rod responses to fish bites for sea or pond fishing. A multi-link model is used to represent the rod, and dynamical responses are analyzed. A new solution method is proposed for a lower triangularization of a sparse symmetric coefficient matrix, and unknown variables such as link accelerations and internal forces are obtained by computation time only linearly increasing with the total number of links. This research not only facilitates understandability of rod fishing but also improves skills of rod anglers through visualizations of rod responses. The submerged line length is an important response parameter. Keywords: Rod response, Fish bite, Multi-link model, Kinetic analysis, Lower triangularization.
1- Introduction This paper considers a traditional fishing tackle consisting of a short rod of about 2 meters, a main line followed by a main sinker, and a hook line followed by a hook. It should be noted here that this type of tackle was originally used for pond fishing by Japanese samurais in the Edo period as a way of keeping skills and mentality for swordsmanship. Artistic bamboo-rods were used in those days, whereas carbon or glass-fiber rods are far more popular today. In this paper the rod is modeled by a multi-link system. The main and hook line portions are also modeled in a similar way. A total of several hundred links show up. Many numerical solution methods have been proposed for the multi-link models especially in robot-manipulator fields. These include Armstrong[W1], Featherstone[F1], Brandl[H1], Ascher[UDB1] where the computation time increases only linearly with the total number of links. These methods are less understandable because of complexity or a need for knowledge of 6-dimensional “spatial” vectors. 1. Kyoto University, Yoshida Honmachi, Sakyo-ku, Kyoto, 606-8501, JAPAN. - E-mail : yamabe, kumamoto, [email protected]
302 The Engineering of Sport 7 - Vol. 1 This paper first selects accelerations and internal forces as unknown variables, and arranges kinematic and kinetic equations suitably. A resultant sparse symmetric coefficient matrix is manipulated for a lower-triangularization, and the unknown variables are determined by a linear computation time.
2- Multiple link rod model Yamabe and Kumamoto[SHO1] considers a simple multi-link model where a link consists of a ball mass at the mass-less link tail. Each internal force acts only in a link direction. A bending moment between adjacent links is not considered. The present paper extends the simple model to handle the fishing rod: 1) Mass is distributed over the entire link and moment of inertia is considered accordingly. 2) Two types of internal forces are considered. One acts in a link direction and the other in a vertical direction to link. 3) Two types of motions are considered. One is a translation of the link centre of gravity (COG) and the other is a rotational motion around the COG. 4) Two types of bending moments are considered. One is determined from a joint angle between adjacent links, and the other is calculated from joint angle time derivative, i.e., a damping contribution.
2.1 Model schematic The rod model is shown in Fig.1. Each link is identified by an integer, h = 1,…, n. The unit vector in the direction of link h is denoted by ih. The another unit vector in the counter clockwise vertical direction is denoted by jh. Symbols and are accelerations in directions ih and jh, respectively, where suffix x denotes the direction of ih and suffix y the direction of jh. Symbols and are internal forces in directions ih and jh, respectively. Variable h is a relative angle between link h and its parent link. Variable 1 is exceptionally measured from a perpendicular direction. Symbols and are external forces in directions ih and jh, respectively, mh is the mass of ~ link h. lh is the link length, lh is length between the head of the link to the link COG, and l*h is the length between the COG and the tail of the link. The head of the first link is held by an angler hand. The first link head acceleration vector is assumed to be a zero vector. The corresponding velocity vector is also zero. Joint angle 1 can vary. The moment around COG of link is denoted by mh.
2.2 Kinematic and kinetic equation The head acceleration vector of link h can be represented in terms of its components in the local coordinate system: (1)
The Engineering of Sport 7 - Vol. 1
The head acceleration vector of , angular velocity of parent link h.
303
of the next link can be expressed in terms , and angular acceleration (2)
Express the acceleration vector be rewritten as:
in its local coordinate system. Then Eq. (2) can
(3) Where the sine and cosine of the joint angle are denoted respectively by Sh = sin h and Ch = cso h. This is the kinematic equations.
Figure 1 - A rod represented by a multi-link model.
Consider next kinetic equations. The translation and the rotation with respect to link h COG are subject to the following equations: (4)
304 The Engineering of Sport 7 - Vol. 1 Symbol Ih is the moment of inertia of link h. For example, a solid cylinder with radius rh has the moment of inertia of: (5) Table 1 summarizes the kinematic and kinetic equations.
2.3 Obtaining symmetric coefficient matrix ~ The coefficient matrix of Table 1 is not symmetric. Values lh h = 1,2 on rows 5 and 10 can be erased by~ using value 1 on rows 4 and 9. Value l*1 on columns 6 and 7 on row 5 changes to l*h + lh = lh. A symmetric coefficient matrix is obtained as Table2. Rows 5 and 10 of Table 2 are now kinetic equations of rotation around the heads of the links 1 and 2, respectively. The internal force to acting at the head of the link does not generate the moment for the rotation.
3- Lower-triangularization of coefficient matrix Introduce the following symbols to express block matrices in Table 2.
Table 3 shows a block representation of Table 2. Symbol T in Table 3 denotes a matrix transpose. A procedure of lower-triangularization is shown below. Total number of links is two for Table 3, i.e., n = 2. Table 1 - Difference approximation matrix for pattern of two links.
The Engineering of Sport 7 - Vol. 1
305
1) Matrix H2 for the last link is a positive scalar. Therefore, matrix KT2 can be erased by using row 10. Matrix G2 changes to . The mass matrix of link 2 is positive semi-definite, and hence matrix is also positive semi-definite, where the mass matrix is defined for link n by:
2) Denote by Tn the 4 4 matrix on the diagonal position on rows 6 to 9.
The inverse of Tn is obtained explicitly. Matrix RT1 and QT1 on rows 3 to 5 can eventually be deleted by matrix manipulation using rows 6 to 9. After this deletion, matrix K1 changes to , changes to . to , and G1 changes to . 3) Note that is again a positive scalar because is a positive scalar and is positive semi-definite. Thus, can be deleted and becomes . The final stage is shown by Table 4. Known an, bn, cn values at the last column can be transformed accordingly. Unknown accelerations and internal forces can be obtained from the block-lower-triangular coefficient matrix of Table 4 where diagonal blocks can be inverted easily. Table 2 - Sparse symmetric coefficient matrix.
306 The Engineering of Sport 7 - Vol. 1 Table 3 - Block matrix representation of coefficient matrix.
Table 4 - Result of lower triangularization.
Figure 2 - Rod vibration response by the multi-link model.
4- Rod parameter identification A typical rod is considered. The Young’s modulus of E = 3.89 1010 N/m2 is estimated from a bending experiment. The moment at link joint consists of the moment proportional to the joint angle and the moment proportional to the time derivative of the joint angle. (6) Here, 1/R is a curvature at the joint, lD,h is geometrical moment of inertia, and ch is a damping constant. The positive direction of the moment is a counterclockwise direc-
The Engineering of Sport 7 - Vol. 1
307
tion. The net moment acting on link h is a difference of the head and bottom joint moments. (7) The unknown damping constant ch is estimated in the following way. A sinker of 10.0g is connected to the rod tip. The tip drops about 40mm. The sinker is then detached and the rod is subject to a free vibration. The damping constant is estimated by letting the multi-link numerical solution conforms to the observed free vibration. The rod is divided into 20 links. The observed vibration period is about 0.1s. The damping ratio at each period is observed and damping constant of ch = 6.5 10-4 N • m • s/rad is obtained. The multi-link numerical solution with this damping constant can conform to the experimental result. Figure 2 shows a numeric solution. The vibration cycle is 0.11s, and the vibration of the rod is reproduced accurately.
5- Rod response to fish bite Figure 3 shows a tackle for rod fishing. The figure on the right hand side is a result of a multi-link numerical calculation under a leftward tidal flow. The diameter of the main line is 330μm and hook line is 235μm. Hook portion with bait is approximated by a ball with a diameter of 7mm. The tidal velocity is 0.1m/s. It turns out that the frictional force from the submerged line portion plays a dominant role. This force acts longitudinally along the line, i.e, in parallel direction to the unit . vector ih of link h. The force is proportional to relative longitudinal speed x between the . line and the fluid. A total frictional force is estimated as cx from a salt water vibration experiment without a tidal flow, where constant c is calculated as c = 0.0055 • l + 0.0281 and l is the total length of the submerged line. The longer the submerged portion, the . larger the frictional force, given the same relative longitudinal speed x. The rod is attenuated earlier for the longer submerged line portion. This is consistent with the actual rod behavior observed during the salt water fishing.
Figure 3 - A rod fishing tackle depicted by the model.
Figure 4 - Rod tips vibration by downward fish-bite.
308 The Engineering of Sport 7 - Vol. 1 The baseline value of 0.0281 is a contribution independent of the submerged line. The length coefficient of 0.0055 represents a contribution by unit length of the submerged line. These observations can be incorporated with the multi-link line model . where each link can have different x. The rod is divided into the 20 links and the line into the 110 links. There is no tidal flow. The depth of the main sinker is 8.5 m. Figure 4 shows the rod tip response to a downward fish-bite starting at 80s. The bite continues for 10s but the rod vibration vanishes rather quickly. As an extreme case, an over-damping response without any vibration may be observed.
6- Conclusions 1) A multi-link model is proposed to represent the rod and line. Kinematic and kinetic equations are arranged, and the coefficient matrix for the unknowns can be made symmetric by a simple matrix manipulation. 2) An efficient algorithm is given for lower triangularization of this sparse symmetric coefficient-matrix. 3) Rod parameters are identified, and the multi-link model can reproduce actual rod vibrations. 4) The frictional force from the submerged line portion turns out to play a dominant role for the rod response. The friction parameter is identified, and the rod response to the fish bite is reproduced by the multi-link model.
7- References [W1] W.W.Armstrong, Recursive solution to the equations of motion of an n-link manipulator In Proc. of the 5th World Congress on Theory of Machines and Mechanisims, 1343-, 1979. [F1] Featherstone.R, The calculation of robot dynamics using articulated-body inertias, International Journal of Robotics Research, Vol.2, No.1, 13-, 1983. [H1] H.Brandl, R.Johanni and M.Otter, A very efficient algorithm for the simulation of robots and similar multibody systems without inversion of the mass matrix, In Proc. of IFAC/IFCP/IMACS int. Symp. on Theory of Robots, Vienna Austria, 1986. [UDB1] Uri M.Ascher, Dinesh K.Pai and Benoit P.Cloutier, Forward dynamics, elimination methods, and formulation stiffness in robot simulation, International Journal of Robotics Research, Vol.16, No.6, 749-, 1997. [SHO1] S.YAMABE, H.KUMAMOTO, O.NISHIHARA, Analysis and Animation of Bar Float Response to Fish Bite for Sea Fishing, 3rd Asia-Pacific Congress on Sports Technology APCST2007, The Impact of Technology on Sport II, Taylor & Francis, 537-543, 2007.
Design and Manufacture of Customised Orthotics for Sporting Applications (P62) Paul Crabtree1, Vimal Dhokia, Martin Ansell, Stephen Newman
Abstract: The prescription and manufacture of personalised bespoke orthotics aim to correct and prevent symptoms related to a number of conditions concerning poor gait. Current fabrication methods involve modular designs built upon prefabricated components. This paper reviews classifications for orthotics and materials for their manufacture depending on whether they are rigid, semi rigid or soft. Manufacturing methods for the production of rigid customised carbon fibre orthoses are described. Epoxy pre-impregnated carbon fibre material was laid up onto a positive mould of a subject’s foot to achieve an accurate geometrical relationship between the plantar surface of the foot and the insole. Manufacturing techniques including vacuum bagging and autoclaving were used to produce a thin profile, allowing the device to be placed in footwear where space is at a premium. The final results were observed to meet the requirements of the subject, providing comfortable and supportive rigid orthoses. Keywords: Orthotics, CFRP, vacuum bagging, customised manufacture.
1- Introduction The use of foot orthoses can be crucial for the correction of foot deformities and biomechanical inefficiencies. Often an incorrect skeletal alignment can result in overuse injuries further up the kinetic chain. Foot orthotics can be prescribed for a number of conditions, for example a biomechanical correction is a common reason as well as comfort enhancement and for impact attenuation, but they are also used for medical applications such as the prevention of sores for diabetic patients and the treatment of rheumatoid arthritis. The manufacture of personalised orthotics has been an established technique for many years with podiatrists treating pathological stresses occurring in the foot. However, the production techniques currently used by many practitioners have lengthy lead times and it is very difficult to accurately replicate designs. With the rapid advances in scanning technologies and the routine application of vacuum bagging techniques, it is the 1. Department of Mechanical Engineering, University of Bath, UK - Email: [email protected]
310 The Engineering of Sport 7 - Vol. 1 intention of this research to demonstrate a methodology for the manufacture of bespoke insoles for sporting applications.
2- Orthotic Classification There have been many attempts to classify categories of foot orthoses and there are differing opinions with respects to applications, fabrication methods and rigidities. Orthotics classed according to their rigidities fall into one of three categories; rigid, semi-rigid or soft orthotics (Razeghi and Batt, 2000). Rigid orthotics are typically functional orthotics that is to say insoles that provide joint stability and correct function by controlling the motion of the foot. A rigid material is required so little deformity or flexion is experienced. These insoles are typically prescribed where a high level of stability and control is required and where a level of surface finish is important, e.g. cases such as diabetes and rheumatoid arthritis. Rheumatoid arthritis sufferers often suffer foot deformities which in turn can cause pain whilst weight bearing so a functional orthotic will prevent overloading of areas of the foot to avoid any further degeneration of the foot structure (Magalhães et al., 2006, Clark et al., 2006). Semi-rigid orthotics may also be used for functional insoles but will also incorporate an impact attenuation aspect to improve comfort. It has been demonstrated that wearing sports shoes with a foamed midsole can reduce impact forces on the foot, although an amount of support is required to prevent excessive inversion or eversion (De Wit et al., 1995). It could be assumed that an insole displaying properties similar to a midsole material would be advantageous. A semi-rigid orthotic provides a compromise between the control a rigid orthotic provides and the cushioning properties of a soft insole. There are three basic types of foot orthoses according to manufacture methods, – Prefabricated insoles which are bought ‘off-the-shelf’ and are typically mass produced, providing general arch support or cushioning to areas of the foot. – Customised orthotics which usually consist of a modified prefabricated component, often called a ‘shell’. This is then built up in a modular fashion to address certain problems, for example a heel lift to treat a leg length discrepancy. – Custom-moulded orthoses, manufactured from a cast or mould of the subject’s foot and produce the best results.
3- Materials Selection The choice of material when designing an orthotic is vital to its success; if the material is too hard then any incorrect aspects to the design could result in an uncomfortable and biomechanically detrimental insole. There is a growing belief among podiatrists that providing the orthosis is correctly prescribed and fits the subject well, then a rigid material should be used as the orthosis should be of a reasonable comfort due to correct skeletal alignment. The important characteristics in the manufacture of orthoses has been defined by Paton et al. (2007), these are the density, resilience, compressive stiffness, coefficient of friction, durability and compression set. These characteristics will vary depending on the requirements of the orthoses. Materials used for the fabrication of foot orthoses can be
The Engineering of Sport 7 - Vol. 1
311
divided into categories including plastics, composites and foams (Nicolopoulos et al., 2000). Plastic materials such as polypropylene can be used in the manufacture of rigid and semi-rigid orthoses, depending on the thickness of the design; the material exists above its glass transition temperature at ambient temperature. Foamed materials are used for soft and some semi-rigid orthoses. They can be of open or closed cell structure consisting of a continuous polymer phase containing pockets of gas achieved through blowing agents. Open cell foams allow the movement of the gas in between the cells whereas a closed cell foam encloses the gas within the cell walls allowing for a water-tight material, this is desirable for orthoses manufacture as sweat will not be able to penetrate into the material to cause premature degradation. Ethylene Vinyl Acetate (EVA) is a popular material in the manufacture of both orthotics and midsoles of sports shoes due to its versatility in terms of the variation of density and hardness. It has been suggested that a high density EVA possesses the properties most suitable for motion control (Paton et al., 2000). Typically these foams are heat mouldable and are commonly delivered in sheet form, manufactured through injection moulding or similar. Many orthotic applications require the insole to be functional whilst also providing the wearer with a degree of comfort. Another aspect to consider in the design of orthotics is the type of shoe in which the device will be placed. Some footwear used for sports would be unaccommodating for an insole due to the space inside. An example of this is the recent changes in design to football boots; recent developments in the way football is played and emphasis in aesthetic design have resulted in boots becoming lighter because a heavier boot will have an effect on the oxygen consumption and ultimately the fatigue of a player (Millet et al., 2006). Thus boots are becoming slimmer in design, with thinner sole plates and tight fitting uppers. Conventional orthotic dimensions would not be conducive with these designs and so a space saving insole is required. Carbon fibre reinforced plastics (CFRP) are an ideal solution to this problem; the material has a high stiffness to weight ratio and so can be made very light and, more importantly, thin to achieve a rigidity similar to that of a much thicker conventional design. These insoles can also be used for applications such as fashion shoes where similarly space is at a premium. CFRP insoles are the subject of the research reported here.
4- Foot Biomechanics The biomechanics of the foot are very complex and as such are outside the scope of this report. It is widely accepted that an orthotic device should orient the foot into a neutral subtalar joint position (Padhiar, 2001, Nigg et al., 1999). The neutral subtalar joint affects the motion of joints both distally and proximally to it and so establishing its neutral position is crucial for correct skeletal alignment. Biomechanical inefficiencies within the foot have the potential to cause injury to other areas, most commonly the knee. This is typically due to excessive amounts of inversion/eversion at the foot causing external/internal tibial rotation as a compensatory measure. Correcting this inversion/eversion aims to contribute to the prevention of overuse injuries to the knee. A study by MacLean et al. (2006) into the preventative measures of custom orthoses for healthy runners reported a significant reduction in maximum rearfoot eversion angle
312 The Engineering of Sport 7 - Vol. 1 and velocity, subsequently reducing the internal ankle inversion moment thus helping with the control of subtalar joint pronation. Biomechanical requirements with respect to the foot will change for different sporting activity, for example a tennis player may require a different prescription than a rugby or football player due to the factors such as playing surfaces and footwear. This area will be researched in order to define a prescriptive method for a number of different sports according to movements synonymous with the activity, allowing a bespoke insole to be manufactured according to not only the functional requirement of the athlete but also the sport and type of footwear worn.
5- Experimental Programme - cast fabrication A plaster cast mix was used to produce a positive mould of a subject’s foot. This was achieved by taking an impression of the plantar surface of the foot by pressing the foot into a phenolic foam box, manufactured by A. Algeo Ltd. With help from a podiatrist the foot was manipulated into a neutral subtalar joint position for correct biomechanical function. The plaster mix was then poured into the negative mould and left to set, resulting in a geometrically accurate representation of the foot, shown in figure 1. The cast was digitised using a 3D laser scanner to create a point cloud which was subsequently imported and edited within a CAD system to create a surface, shown in figure 2. Ribs created from Bezier curves were driven through planes along the x and y axis to create a wireframe on which the surface could be placed.
Figure 1 - Plaster cast mould of manipulated foot position
Figure 2 - CAD image of orthotic surface geometry
This surface maps accurately to the surface of the foot cast and so provides an accommodative insole. For a functional, corrective orthotic, specific control points could be manipulated in order to achieve the desired geometry, for example the arch of the device could be dragged up, providing a higher arch support.
6- Experimental Programme – Vacuum bagging of carbon fibre orthotics Carbon fibre orthoses were manufactured utilising vacuum bagging and autoclaving to produce high quality components that were free of voids. The material used for the manufacture of the insole was an epoxy pre-impregnated sheet (prepreg); a woven
The Engineering of Sport 7 - Vol. 1
313
prepreg material consist of two directional yarns of carbon fibre (weft and warp) impregnated with an epoxy resin matrix. There are many types of prepreg resultant of the variations in fibre directions and matrices. The material used for the orthotics was obtained from Blatchford; the prepreg is used for manufacturing prosthetics and so was suitable for orthotic applications. The material is called HEXPLY 920-793-38.5% (909220) and is an open 5 shaft satin weave, illustrated in figure 3, selected due to its low crimp and good drapeability, this is desirable due to the complex geometry of the plantar surface of the foot. A low crimp will result in the fibres being oriented straighter, allowing a greater load to be carried, thus giving improved mechanical performance. The three main weaves are plain, satin and twill weaves, each with different crimps and drapeabilities. The fibre volume fraction of the material was approximately 57%.
Figure 3 - Satin weave prepreg material.
In order to produce an accurate bespoke insole, a copy of the plaster cast was used as a positive mould onto which the prepreg material was laid up. Due to the porosity of the plaster, a peel ply material was required in order to prevent the resin from flowing away from the fibres and into the cast. Conventional peel plies such as films wrinkled on the surface due to the geometry of the foot resulting in a poor surface finish of the carbon fibre. A smooth finish was achieved by vacuum forming a layer of polythene onto the plantar surface of the foot cast on which to apply the prepreg, shown in figures 4 and 5.
Figure 4 - Vacuum forming of polythene onto cast
Figure 5 - Smooth polythene peel ply to prevent resin from flowing into cast
314 The Engineering of Sport 7 - Vol. 1 The prepreg was applied to the cast in layers, each bonded together to achieve a compact part. The prepreg material was first heated to increase its plasticity, thus making the draping process easier. Subsequent layers were heated and applied to the mould; these were rolled to ensure that there were no air pockets between the plies that could potentially create voids in the part. The insole designed used two layers of the woven prepreg material with a layer of unidirectional material in between to maintain symmetry. Two areas of the insole were also reinforced to ensure the required rigidity was attained at the arch of the insole i.e. the area where a large amount of control is required such as the arch and also where the orthotic will come into contact with the shoe and these are shown in figure 6.
Figure 6 - Laying up process highlighting the reinforcing areas.
When all of the plies had been laid up, a bleeder cloth was placed on top to absorb any excess resin, followed by a perforated film which prevented any further flow of the resin when curing; the film was slightly porous to allow the passage of air and any volatiles. A stockinette was then placed over the whole mould before a final layer of breather fabric was draped over the top to assist with the removal of air and volatiles and allow for vacuum bagging. This process is demonstrated in figure 7. A vacuum bag was then manufactured to fit over the mould and the air was evacuated, shown in figure 8.
Figure 7 - Typical vacuum bagging structure (Hexcel Corporation).
The Engineering of Sport 7 - Vol. 1
315
Figure 8 - Evacuation of air from vacuum bag.
The vacuum bag was placed inside the autoclave and held under high pressure and elevated temperature; a thermocouple was placed under the component to obtain an accurate localised temperature reading rather than just the temperature of the air within the autoclave. The cure cycle was programmed into the autoclave according to the properties of the prepreg; a temperature of 120°C was required for a minimum of one hour. First the pressure within the autoclave was increased at a rate of 10PSI per minute until the desired pressure of 100PSI was reached and subsequently held. The temperature was then ramped at 3°C per minute until 120°C was reached and this was then held for a duration of 75 minutes. Then, both the pressure and temperature were allowed to drop at the same rate as it was increased until it had returned to ambient temperature, and thus was ready for removal. Figure 9 shows the temperature and vacuum pressure profiles for the curing cycle, displaying the pressure and temperatures within the autoclave, both from the thermocouple placed on the part and from the air temperature. Finally the insoles were removed from the autoclave and the vacuum bag before they were ground to shape, resulting in a profile suitable for the subject’s feet. The final orthotics are shown in figure 10. Many orthotics need to be reshaped in order to fit correctly into the subject’s shoes; if the orthotic is too wide then it will not sit on the sole of the shoe, conversely if the orthotic is too narrow, it may shift medially within the shoe. An ideal orthosis width is such that it sits flat on the top of the insole of the shoe, just touching the medial and lateral edges of the upper.
Figure 9 - Vessel chart showing pressure and temperature throughout the cure cycle.
316 The Engineering of Sport 7 - Vol. 1
Figure 10 - Final orthotics ground to shape for placement within a shoe.
Due to the bespoke nature of the orthoses, any destructive testing would be inappropriate. However, the geometry and the feel of the insole can be tested by examining how well the geometry maps to both the casts taken and the subject’s feet. These observations proved to be successful and it is intended to further prove the effectiveness of the devices by carrying out a detailed motion analysis of the subject whilst (a) barefoot, (b) in shoes without the orthotics and (c) with the orthotics in place to confirm an improvement in gait.
7- Conclusions The manufacture of bespoke, functional insoles specific to the biomechanical needs of the subject aims to correct and improve gait, thus decreasing the risk of injury. The carbon fibre orthoses manufactured were of an appropriate stiffness to achieve the required control whilst also being thin enough to fit inside the desired shoe, saving space. It was observed that the orthoses were of the correct geometry, successfully mapping onto the original cast of the foot and providing the subject with a comfortable and stable solution. Future work will focus on manufacturing techniques for semi rigid and soft orthotics and CNC machining will significantly reduce fabrication time. Scans of the foot will be directly imported into a CAD system, manipulated and a polymer substrate will be subsequently machined. Cryogenic machining, a method of freezing a foamed material to below its glass transition temperature will allow soft and semi-rigid foams to also be machined.
8- References [CR1] Clark, H., Rome, K., Plant, M., O’Hare, K. and Gray, J. A critical review of foot orthoses in the rheumatoid arthritic foot. In Rheumatology, 45(2): 139-145, 2006
The Engineering of Sport 7 - Vol. 1
317
[DD1] De Wit, B., De Clercq, D. and Lenoir, M. The effect of varying midsole hardness on impact forces and foot motion during foot contact in running. In Journal of applied biomechanics, 11(4): 395-406, 1995 [MD1] Magalhães, E.P., Davitt, M., Filho, D.J., Battistella, L.R. and Bértolo, M.B. The effect of foot orthoses in rheumatoid arthritis. In Rheumatology, 45(4): 449-453, 2006 [MM1] MacLean, C., McClay Davis, I. and Hamill, J. Influence of a custom foot orthotic intervention on lower extremity dynamics in healthy runners. In Clinical biomechanics, 21(6): 623630, 2006 [MP1] Millet, G., Perrey, S., Divert, C. and Foissac, M. The role of engineering in fatigue reduction during human locomotion – a review. In Sports engineering, 9(4): 209-220, 2006 [NB1] Nicolopoulos, C.S., Black, J. and Anderson, E.G. Foot orthoses materials. In The foot, 10: 1-3. 2000 [NN1] Nigg, B.M., Nurse, M.A. and Stefanyshyn, D.J. Shoe inserts and orthotics for sport and physical activities. In Medicine and science in sports and exercise, 31(7): S421-S428, 1999 [P1] Padhiar, N. Foot orthoses and their role in injury management. In SportEX medicine, 8: 68, 2001 [PJ1] Paton, J., Jones., R.B., Stenhouse, E. and Bruce, G. The physical characteristics of materials used in the manufacture of orthoses for patients with diabetes. In Foot and ankle international, 28(10): 1057-1063, 2007 [RB1] Razeghi, M. and Batt, M.E. Biomechanical analysis of the effect of orthotic shoe inserts: a review of the literature. In Sports Medicine, 6(6): 425-438, 2000
Analysis of Snowboard Stiffness and Camber Properties for Different Riding Styles (P65) Aleksandar Subic1, Patrick Clifton, Jordi Beneyto-Ferre, Arnaud LeFlohic, Yoshiki Sato, Victor Pichon
Topics: Sports Technology and Design. Abstract: Research has indicated that the flex pattern and camber of a snowboard are crucial to its overall “feel” and response. These features are the primary cause of variation in snowboard performance for different riding styles. Consequently, this article deals with the identification of stiffness and camber characteristics for both freestyle and freeride boards, and their statistical correlation to a comprehensive list of qualitative “feel” based performance requirements. It has been determined that the test boards spanning the major styles all possessed similar bending and torsional profiles. Neglecting the assumed effect of the binding inserts, all of the profiles were highly representative of each snowboard’s respective thickness distribution. This leads to a conclusion that the thickness of a snowboard at any location along its chord will govern its resulting stiffness characteristics. The stability, edge grip and accuracy of any snowboard all appear to be highly positive linearly dependent on both the body bending and torsional stiffness. The camber on the other hand showed a particularly strong linear positive correlation to a snowboard’s manoeuvrability. Despite it being polar opposite by definition to the stability, the manoeuvrability also showed good positive linear correlation to the stiffness, implying that these subjective parameters are not mutually exclusive. However it appears that relatively soft boards with high camber are considered the most lively, although the mass of the rider will greatly affect how the stiffness of the board is perceived in terms of overall liveliness. Keywords: Snowboards; Stiffness; Static Testing; Correlation.
1- Introduction Modern snowboard design is dictated predominantly by the desired application or the style of the ride, with boards generally falling under one of two headers; Freestyle (park and trick based) or freeride (all-mountain). Certain boards are also considered ‘versatile’, and are designed to bridge the gap between the two major styles. A third less popular 1. School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University, Melbourne, Australia E-mail: [email protected]
320 The Engineering of Sport 7 - Vol. 1 race specific category also exists, that of freecarving or alpine. Manufacturers currently spend significant time and money trialling new designs, relying heavily on the feedback of professional riders in attempting to design-in the “feel” and optimise the performance of the board. The Snowboard Research Group at RMIT University in Melbourne set out to fully characterise the “feel” of snowboards for the two main riding styles. By correlating subjective evaluations to objective laboratory and field based data, the relevant matrices of parameters leading to the desired “feel” of the board can be determined. Preliminary research (Subic et al 2008) has indicated that the flex pattern (bending and torsional stiffness distributions) and camber (maximum height of the running surface) of the snowboard are crucial to its overall “feel” and response, and further drive the difference in performance between the major styles. This article deals with the identification of stiffness and camber characteristics for both freestyle and freeride boards, and their correlation to the following list of qualitative “feel” based performance requirements: Stability = How stable the rider feels on the board. Feedback = The amount of stress felt on the rider’s body including the effects of board chatter. Speed = The gliding speed of the board compared to other boards of similiar length. Accuracy = The precision of board movement in response to rider input. Forgiveness = The tolerance of the board to errors from the rider. Edge Grip = The level of grip exhibited during turns. Manoeuvrability = How easily the board responds to rider inputs. Transition smoothness = How easily the board flows from edge to edge. Board Liveliness = The level of ‘pop’ or spring in the board when performing a jump. The article presents a comprehensive approach to determining bending and torsional stiffness characteristics of snowboards for different riding styles with an in-depth discussion of relevant parameters governing the design of new generation versatile snowboards.
2- The Importance of Stiffness in Snowboard Design A thorough benchmarking process has been undertaken to identify the key snowboard design parameters and potential innovation opportunities (Subic et al 2008). The analysis used the Quality Function Deployment (QFD) method to evaluate relevant customer requirements and relate them to objective technical attributes of selected boards for both major riding styles. Additionally, key design drivers were used in conjunction with the Market Opportunity Mapping (MoM) to identify potential gaps in the snowboard equipment market. The qualitative data relating to customer requirements consisted of subjective ratings and importance levels for test boards spanning the major riding styles, and was obtained through a variety of online and on-snow surveys and interviews. The objective technical attributes on the other hand were based primarily on the ASTM Standard F1107-1995 – Standard Terminology Relating to Snowboarding, although several other material properties parameters were defined to cover all relevant aspects of the snowboard design. All of
The Engineering of Sport 7 - Vol. 1
321
the parameters were measured in the laboratory or obtained from published data sheets. Upon processing, the QFD methodology pinpointed the bending stiffness distribution as the overall key objective parameter in the design of a snowboard. A comprehensive gap analysis was also completed to identify possible design innovation and product development opportunities. Using the ratings from the first online survey of board models manufactured between 2004 and 2007, the snowboard’s cumulative performance under the prescribed qualitative headers was plotted against its published style, within a range between pure freeride and pure freestyle. The resulting Market Opportunity Map showed that there were practically no high performing versatile boards present, at odds with the desires of modern snowboarders from the various surveys and interviews. This confirmed that a gap in the current snowboard marketplace exists. In order to realise the identified design innovation opportunity, it was important to pinpoint the objective design parameters that effect the versatility of snowboards. A versatility value was formulated as a measure of the extent variation in a parameter will drive the feel and performance from freestyle to freeride or vice versa. The benchmarking indicated varying the bending and torsional stiffness distributions and the camber would appear to be the key approach to altering the feel and performance of a snowboard across the major riding styles.
3- Experimental Investigation of Stiffness Characteristics Given the determined importance of the bending and torsional stiffness distributions (plus camber) to the overall “feel” of snowboards, as well as driving the difference between the major riding styles, a thorough investigation into the bending and torsional characteristics of the three test boards was undertaken. The results should assist greatly in the generation of a versatile design characterised by optimal feel, as well as provide further insight into snowboard designs. The procedures for the tests were based primarily on ISO Standard 5902: Alpine skis - Determination of the elastic properties although various modifications were made to the procedures in order to apply them to snowboards. Most changes from the tests prescribed in the standards were a result of the differing geometry between snowboards and skis, requiring alternate testing dimensions as well as the application of greater forces and moments on the test boards to generate the requisite levels of bending and torsion. However, as each ski only possesses one binding as opposed to the dual binding system present on snowboards, the location of clamping during the tests also had to be modified. Furthermore, since only the basic testing methods were given by the standard, the testing rig had to be designed from scratch (see below for details). Also, the standard only gave procedures for calculating spring constants, with units of N/mm and N.m/º for bending and torsion respectively, so further transformation of the data was required to enable the calculation of stiffness values. This transformation was based around the work of Darques et al in 2004, which set out the calculation of bending and torsional stiffness from basic deflection and angular deformation measurements. The combina-
322 The Engineering of Sport 7 - Vol. 1 tion of both of these approaches allowed the determination of bending and torsional stiffness curves for all three test boards.
3.1 – Bending Stiffness The bending stiffness of the three test boards was calculated using the following basic formula (Darques et al., 2004): (1) Where EI is the bending stiffness (N.m2), M is the applied moment (N.m) and ƒ” is the curvature of the snowboard (m-1). The apparatus designed for the tests based on the descriptions given in the standard is pictorially shown in Figures 1-3 below. The test rig displayed in Figure 1 consists of a 1500mm long C-channel base, two adjustable supports with 20mm diameter rollers (supporting the entire width of the snowboard 220mm from the tip and tail respectively), and finally a load (F) application device consisting of two 20mm diameter rollers supporting two 16kg masses via hooks. Figure 2: Curvature Measurement Device shows the apparatus used to calculate the curvature of the snowboard, consisting of a 20mm comparator positioned centrally in a 200mm wide C-section. The device allowed accurate measurement of the localised relative deflection (‰) at 50mm intervals along the chord, and thus calculation of the curvature using simple geometry.
Figure 2 - Curvature Measurement Device.
Figure 1 - Bending Tests Rig.
Figure 3 - Alternate Setup.
To determine the bending stiffness at the heel and shovel of the snowboard, a different setup was required, as shown in Figure . For those particular tests the forebody or aftbody of the board was clamped using sets of 40mm wide metal plates 100mm from
The Engineering of Sport 7 - Vol. 1
323
the tip/tail, and the board was deflected using 22kg of total mass and measured at 50mm intervals as previously described. The combined results of the bending tests on all three test boards are shown graphically in Figure .
Figure 4 - Bending Stiffness Tests Results
3.1 Torsional Stiffness The torsional stiffness profiles were obtained using the following formula (Darques et al., 2004): (2) Where GJ represents the torsional stiffness (N.m2), T is the applied torque (N.m), Ø is the angular deformation (rad) and d is the length of the area under consideration (m). Given that the board materials were non-homogenous, the measurements were made upon four sections of the board approximately 200mm long. This length was chosen as although it should ideally have been as small as possible, due to angle measurement accuracy considerations it was considered appropriate. It was also a slight departure in procedure from the method of Darques et al., who used the total length from the clamping device, but resulted in higher levels of accuracy as the original formula was for a homogenous beam. Using the same rig as per the forebody and aftbody bending tests to clamp the board, the section under consideration was twisted using a dual system, comprising of a hanging mass on one side of the snowboard section, and a mass pulling the board upward via a pulley and flagstaff on the opposing side. Note that the masses applied to each test board varied between 23kg and 11kg, to ensure adequate angular displacement without straying into the plastic deformation zone. The board was clamped in four separate configurations for the tests. Firstly, the basic forebody/aftbody tests were conducted, whereby the board was clamped along the centreline, with the rollers positioned 170mm from the tip/tail. Secondly, to simulate the twist placed on the body section of each snowboard by the rider, the board was clamped at the
324 The Engineering of Sport 7 - Vol. 1 forward binding location, with the rollers positioned on the aft binding inserts. This test was repeated with the opposing setup, with both tests utilising a test section of the stance width plus 100mm. The second set of tests was again a departure from both the standard and the work of Darques et al., in order to highlight the particular response of the body section of each test board, determined to be crucial from the prior research (Subic et al., 2008). To calculate the resulting angular deformation along each test section, again a comparator (on a guided rail) was utilised to measure the vertical displacement of locations along the edge of the board (at 50mm intervals), which were then converted into angles using the board width distribution and simple geometry. The resulting torsional stiffness profiles for the three test boards are displayed in Figure below.
Figure 5 - Torsional Stiffness Tests Results.
4- Discussion of Results The bending stiffness profiles in Figure showed a number of interesting trends across the style spectrum. All three boards possessed similar bending profiles, comprising a steep rise in stiffness from the tip and tail towards the centre, yet all featuring a substantial trough in the centre of the board. The wavelike profiles for all three snowboards were likely a result of the steel binding inserts providing extra bending resistance at the locations of the peaks. Interestingly, the stiffness range between the peaks and trough varied quite drastically across the test snowboards, but this variance was most likely due to differing insert configurations, and possibly materials. Neglecting the assumed effect of the inserts, all three profiles are highly representative of each snowboard’s respective thickness distribution, and hence it can be postulated that the thickness of a snowboard at any location along its chord will drive its resulting bending stiffness. The freeride test board unsurprisingly showed the greatest level of overall stiffness, with the peaks and troughs significantly higher than the remaining two test boards. Surprisingly however, the versatile test board profile possessed higher peaks compared to the freestyle curve, yet a lower trough. Whether this was indicative of its versatility (being a combination of the other two curves) or was merely a result of a different binding insert pattern requires further analysis. It was also noted that the freestyle profile
The Engineering of Sport 7 - Vol. 1
325
was far more even throughout, with less drastic variation from tip to tail. This was not entirely unexpected given its intended trick riding application and the requirement of the board to have an even response. The torsional profiles shown in Figure possessed similar trends to the preceding bending stiffness profiles, with the characteristic wave profile (again presumably due to the inserts) apparent in all three test boards. Again, the freeride test snowboard exhibited the greatest torsional stiffness, but in this case by a significant margin (approximately 150%) unlike the previous bending curves. The torsional profiles for the versatile and freestyle curves were very similar, but interestingly, the freestyle test board possessed a slightly greater torsional stiffness throughout the body. This was most likely a result of the board’s relatively large waist width. Furthermore, unlike the preceding bending profiles, the peaks and the troughs for the versatile and freestyle boards had similar heights, implying the inserts had a similar effect despite the differing binding configurations. Finally, whilst the versatile and freestyle curves were relatively symmetric (within a reasonable margin of error), the freeride curve showed a significantly higher forebody peak. The only reasonable explanation is the extra resistance is for greater hold and accuracy in the initiation and maintenance of a turn for the leading foot, however this cannot be confirmed.
5- Correlation Analysis To provide further insight into the connection between the bending and torsional stiffness characteristics (as well as camber) of snowboards and their performance, a statistical correlation analysis was performed between the experimentally obtained values of these objective parameters and the subjective performance ratings for each test snowboard. Table below shows the respective Pearson correlation coefficients for each relationship, on scale between +1 (increasing linear) and -1 (decreasing linear). These values were obtained by applying the following standard formula: (3) Where Correl (X,Y) is the correlation coefficient, x is the objective parameter value, y is the subjective parameter value, whilst x– and y– are the respective sample means. Table 1 - Correlations.
326 The Engineering of Sport 7 - Vol. 1 The stability, edge grip and accuracy of any snowboard all appear to be highly positive linearly dependent on both the body bending and torsional stiffness (all coefficients over 0.88), which was unsurprising as of the three test boards, the highly stiff freeride board was very stable and accurate, yet noted as rather demanding to ride. Interestingly, for all of these subjective parameters the shovel and heel bending stiffness showed a medium linear negative correlation (all between -0.38 and -0.74), implying that relatively soft nose and tail sections are preferred for stable and accurate boards with good grip. The camber however showed a particularly strong linear positive correlation to a snowboard’s manoeuvrability. Despite it being polar opposite by definition to the stability, the manoeuvrability also showed good positive linear correlation to the stiffness, implying the parameters are not mutually exclusive. The results suggest that a stiff, highly cambered body section will allow the rider to swiftly generate turns by aggressively pushing the body of the snowboard into the snow (flattening the camber). These same properties would also allow a quick exit from a turn as the camber is returned to its natural position. On the other hand, considering that the freestyle board received the highest liveliness ratings yet possessed the lowest body stiffness levels, it appears that relatively soft boards with high camber are considered the most lively, although the mass of the rider will greatly affect how the stiffness of the board is perceived in terms of overall liveliness. The positive correlations for the heel and shovel were caused solely by the extra stiffness in these sections for the freestyle board, to ensure adequate spring for jumps and landings. The speed and transition smoothness of a board seem to increase with the overall body stiffness, yet decreasing camber. This is despite previous results suggesting that camber promotes manoeuvrability, and implies that if high levels of camber are present, the significant return spring on exiting the turn will make the transition to the following turn jerky and sudden. However the medium positive correlations of transition smoothness to both body bending and torsional stiffness parameters suggests that a certain level of body stiffness is required for a smooth transition between edges, as otherwise the board will feel limp and unresponsive, and the transition will be forced. It appears overall that a delicate balance between camber and stiffness parameters is required for a stable, manoeuvrable and accurate board that still is smooth in transition. The coefficients related to speed on the other hand were generally expected to be all strongly positive with the exception of camber, as increased camber would reduce the snow contact area of the board. Finally, further examination into the feedback and forgiveness of snowboards is required as the results do date are inconsistent with preconceived notions of these parameters. Specifically, it was expected that increased stiffness would generally increase the feedback to the rider, and thus the medium negative coefficients for the body stiffness parameters were very difficult to reconcile with the remaining medium to very strong coefficients in this section. Regarding forgiveness, the results also suggest that the stiffness of the heel and shovel sections are generally unrelated to the overall forgiveness of the board, although again this is highly questionable as they are undoubtedly used in the landing of jumps, where forgiveness is crucial.
The Engineering of Sport 7 - Vol. 1
327
5- Conclusion All of the test boards examined possessed similar bending and torsional profiles, comprising a steep rise in stiffness from the tip and tail towards the centre, yet all featured a substantial trough in the centre of the board. Neglecting the assumed effect of the binding inserts, all of the profiles were highly representative of each snowboard’s respective thickness distribution, and thus it can be concluded that the thickness of a snowboard at any location along its chord will drive its resulting stiffness characteristics. The stability, edge grip and accuracy (plus the speed and transition smoothness to a lesser extent) of any snowboard all appear to be highly positive linearly dependent on both the body bending and torsional stiffness. The camber however showed a particularly strong linear positive correlation to a snowboard’s manoeuvrability. Overall, the analysis has provided insight into the stiffness characteristics of snowboards and furthermore the relationships between bending, torsional stiffness and camber to subjective performance parameters. The information would help simplify tailoring the design of modern snowboards towards specific performance objectives.
6- References [DC1] Darques J., Carreira R.P., de la Mettrie A., Bruyant D. Quality Function Deployment – A means for developing adequate skis and snowboards. The Engineering of Sport 5 - Conference Proceedings of The Engineering of Sport, 13-17 September 2004, Davis, California, p 428-434. [I1] International Standards Organization. Alpine skis - Determination of the elastic properties. ISO 5902:1980. [SC1] Subic A., Clifton P., Beneyto-Ferre J. Identification of innovation opportunities for snowboard design through benchmarking. Sports Technology, Vol 1, No 1, 2008.
The Fluctuating Flight Trajectory of a Non-Spinning Punted Ball in Rugby (P67) Kazuya Seo1, Osamu Kobayashi2, Masahide Murakami3
Topics: Rugby. Abstract: This paper describes the aerodynamic characteristics of a non-spinning rugby ball and its flight trajectory on the basis of the unsteady aerodynamic data obtained from the non-spinning ball. It has been found that the side force oscillates with the time. The oscillation depends on the angle of attack and the position of the lace. As the angle of attack increases, the amplitude of the side force oscillation increases. At a certain angle of attack, which is between 60° and 70°, the amplitude of the side force oscillation becomes comparable to the ball’s weight. The simulated flight trajectory fluctuates. The amplitude is estimated at a few centimeters. Keywords: rugby, aerodynamic force, fluctuating flight trajectory, time variation data.
1- Introduction A punted ball kicked with no spin sometimes fluctuates during flight like a knuckle ball does in baseball. This might be caused by steady aerodynamc forces (Seo et al., 2004) as well as unsteady aerodynamic forces. We have measured the aerodynamic forces acting on a non-spinning ball in a low-speed wind tunnel, and we have simulated the fluctuating flight trajectory by integrating equations of motion of a rigid body on the basis of the unsteady aerodynamic forces.
2- Wind tunnel test A full-size rugby ball was employed to determine the aerodynamic forces acting on a ball in a low-speed wind tunnel with a 1.5 m 옰 1.0 m rectangular nozzle. We used a commercially available ball (Triple Crown, Gilbert, the official rugby World Cup ball). The wind speed was set at 30m/s (Holmes et al., 2006). Data was acquired with strain-gage load 1. Faculty of Education, Art and Science, Yamagata Univ., Yamagata, Japan - E-mail: [email protected] 2. School of Engineering, Tokai Univ., Hiratsuka, Japan - E-mail : [email protected] 3. Graduate School of Systems and Information Engineering, Univ. of Tsukuba, Tsukuba, Japan E-mail: [email protected]
330 The Engineering of Sport 7 - Vol. 1 cells for 16 seconds using a personal computer with the aid of an A/D converter board (PCI-3120, Interface). The sampling rate was 1000 per second.
Figure 1 - Definition of aerodynamic forces and characteristic parameters.
The aerodynamic force data were taken as functions of the angle of attack and the lace angle as shown in Figure 1. The angle is the angle between the longitudinal axis of the ball and the direction of the flight path. The lace angle is 0° if the lace is situated on the top against the wind, and =90° if the lace is situated on the right against the wind. Since the rolling moment and the yawing moment were negligibly small, the following forces and a moment were measured. The drag, the lift and the side force are denoted by D, L et Y, and the pitching moment is denoted by M, respectively as shown in Figure 1. The time averaged aerodynamic coefficients multiplied by the projectile area (& the length) at =180° are shown in Figure 2. Figure 2-a is the case of CD·A· and CL·A, while Figure 2-b is the case of Cm·A·l. Since the time variations of the drag, the lift and the pitching moment are relatively stable, the time averaged data is used for the silmulation. In the case of Figure 2-b, the pitching moment should be 0 when = 0 and 90°, and it also sholud be symmetrical at =45°. Therefore, the quadratic regression curve was applied for the simulation, though there were only 4 data points. Figure 2 shows that CD·A increases with increasing. CL·A increases up to about = 60°, when the effect of aerodynamic stall appears. Cm·A·l has positive values (nose-up) except when = 0 or 90°.
Figure 2 - Time averaged aerodynamic coefficients, CD, CL et Cm, multiplied by the projectile area, A (and the length, l), as a function of the angle of attack, at =180°.
The Engineering of Sport 7 - Vol. 1
331
Figure 3 - Time variations of the side force coefficient multiplied by the projectile area.
The time variations of CY·A at |V➝|=30 m/s are shown in Figure 3. CY·A is the product of the side force coefficient CY and the projectile area of the ball A. Figure 3-a shows the time variations at =180° ( is taken as a parameter), while Figure 3-b shows time variations at =90° ( is taken as a parameter). It can be seen that the time variation depends on and . From Figure 3-a, the time variation becomes more unstable as is increased up to 62°, where the amplitude is extremely large. The amplitude corresponds to the ball’s weight. This large variation also depends on . If the lace is not situated on the back against the wind (=180°), the amplitude becomes much smaller. Though the amplitude becomes smaller above 62°, the time variation is still unstable at 90°. From Figure 3-b, the mean value of the time variation depends on the lace angle. This is because the side force is influenced by the four corners (seams) of the ball (Seo et al., 2004). It is necessary to know the aerodynamic forces ➝F (t)=(D, L, Y) to simulate the flight trajectory. This was obtained using the Fourier series of the 50th harmonics in equation (1). Tough the Fourier series of the 50th harmonics was applied in order to reconstruct the time variation data exactly, It seems that the first 5 harmonics was enough, and the smaller set didn’t influence on the flight trajectory. Since the time variations depend on and , as shown in Figure 3, Fourier coefficients, ➝ a 0 through ➝ a 50 and b➝0 through b➝50, are defined as functions of and . Since the experiment was carried out using 30° intervals for both of and , the linear interpolation is applied for estimating the Fourier coefficients. (1)
3- Basic equations The inertial coordinate system in the right-handed coordinate system is shown in Figure 4. The origin is defined as the point of intersection of the player’s own goal line and the left touch line, while the XE-axis is in the horizontal forward direction, the YE-axis is in the horizontal right direction and the ZE-axis is in the vertical downward direction,
332 The Engineering of Sport 7 - Vol. 1 respectively. Figure 5 shows the Euler angles. The coordinate system in the frame of the ball is denoted by xb,yb, and zb with the origin at the center of gravity of the ball. The xb axis is along the longitudinal axis, while the yb and zb axes are along transverse axes with the zb-axis in the direction of the valve. The sequence of rotations conventionally used to describe the instantaneous attitude with respect to an inertial coordinate system is shown in Figure 5. Starting with the inertial coordinate system the following sequence is followed; 1. Rotate about the ZE axis, nose right (positive yaw ), 2. Rotate about the y1 axis, nose up (positive pitch ), 3. Rotate about the xb axis, right wing down (positive roll ).
Figure 4 - Inertial coordinate system.
Figure 5 - Euler angles.
In terms of coordinate transformations (Kato, 1999) we then have (2)
(3)
The Engineering of Sport 7 - Vol. 1
333
Here, (U, V, W) are the (xb, yb, zb) components of the velocity vector. Since there is a mathematical singularity (Gimbal lock) at = 90°, quaternion parameters (, , , ) should be used instead of Euler angles (Stevens, 2003). (4)
The equations of motion and moment equations are (5) (6) (7) (8) (9) (10) Here, (Xa, Ya, Za) are the (xb, yb, zb) components of the aerodynamic force, (P, Q, R) are the (xb, yb, zb) components of the angular velocity vector, mb is the mass of the ball, g is the gravitational acceleration, (La, Ma, Na) are the (xb, yb, zb) components of the aerodynamic moment, and IL and IT are the moments of inertia of the ball on its longitudinal axis and on its transverse axis, respectively. Assuming that the rugby ball as a hollow ellipsoid, IL = 0.0026 kgm2 and IT = 0.0033 kgm2 (Brancazio, 1987), respectively. It is necessary to convert the experimental aerodynamic data D, L, Y and M into Xa, Ya, Za and La, Ma, Na. The plane which contains the velocity vector V➝ and the unit vector xb is considered. Since the direction of the side force is perpendicular of xb component ^ to this plane, the unit vector of the direction of the side force is expressed by the vector product in equation (11). (11)
334 The Engineering of Sport 7 - Vol. 1 ➝
Here, (^ xb,^ yb,^ zb) are the unit vector of (xb, yb, zb). The angle wt between V and ^ xb is defined by equationn (12). (12) Having derived the unit vectors in the same manner, D, L and Y can be converted into Xa, Ya and Za using equation (13).
(13)
The moments L, M and N are converted into La, Ma and Na in the same manner.
(14)
Assuming L=N=0, equation (14) can be simplified. The derivatives of quaternion parameters (Stevens, 1992) are expressed by (15) (16) (17) (18) By integrating equations (2), (5) through (10), and (15) through (18), the flight trajectory can be obtained.
4- Flight trajectory A punted kick from the center of the field was simulated. The initial position is assumed to be (XE, YE, ZE) = (50, 35, -0.5). Figure 6 shows the velocity vector as well as the angular velocity vector. For a punted kick, there should be nine initial conditions; these are the
The Engineering of Sport 7 - Vol. 1
335
➝
➝ magnitude of V 0, the flight path angle, 0, the azimuth angle, 0, the magnitude of 0, ➝ ➝ the elevation angle of 0, 0, the azimuth angle of 0, 0 and the three Euler angles (Figure 5): The flight trajectory from the catcher’s view is shown in Figure 7. The initial ➝ ➝ conditions was |V 0| =25m/s, 0 =62°, 0=3.0°, | 0| =0rps, 0 =4.4°, 0 =-126°, 0 =-80°, 0 =32°, & 0 =249°. There was no special reason to use these initial conditions. After several simulations, the initialconditions which made the trajectory fluctuated was used. These nine initial conditions were converted into 13 initial conditions to solve 13 differencial equations (2), (5) through (10), and (15) through (18). The hang time is 3.5 seconds. The amplitude of the fluctuations is estimated at a few centimeters. It seems that the the estimated side force by the Fourier coefficients varies so frequently that the reacting time is not enough because of the inertia of the ball.
Figure 6 - Velocity vector (Angular velocity vector).
Figure 7 - The flight trajectory from the catcher’s view.
5- Summary We have carried out wind tunnel tests on a non-spinning rugby ball and simulated the flight trajectory on the basis of the time variations of aerodynamic forces. The results are: 1) The time variation of the side force depends on the angle of attack and the position of the lace. The amplitude of the side force oscillation increases as the angle of attack increases up to 62°, where the amplitude is comparable to the ball’s weight. 2) The simulated flight trajectory fluctuates.
6- Acknowledgement This work is supported by Inamori Foundation and Grant-in-Aid for Young Scientists (A).
336 The Engineering of Sport 7 - Vol. 1
7- References [SK1] Seo, K., Kobayashi, O. and Murakami, M. Regular and irregular motion of a rugby football during flight. The Engineering of Sport 5, 1, 567-573, 2004. [HJ1] Holmes, C., Jones, R., Harland, A.R. and Petzing, J.N. Ball Launch Characteristics for Elite Rugby Union Players, The Engineering of Sport 6, 1, 211-216, 2006. [KO1] Kato K., Ohya A. and Karasawa K. Stability and Control of Airplanes, 11, University of Tokyo Press, 1999. (in Japanese) [SL1] Stevens, B.L. and Lewis, F.L. Aircraft Control and Simulation, 2rd Ed., 25-34, Wiley, 2003. [B1] Brancazio, P.J. Rigid-body dynamics of a football. Am. J. Phys., 55, 415-420, 1987.
Aerodynamics of Bicycle Helmets (P68) Firoz Alam, Aleksandar Subic, Aliakbar Akbarzadeh1
Topics: Aerodynamics. Abstract: The aerodynamic efficiency is an important design criterion for bicycle helmets. The characteristics of venting geometry, venting orientation and venting location play a vital role in aerodynamic drag. In order to design an aerodynamically efficient bicycle helmets, a comprehensive study is required. Therefore, the primary objectives of this work were to study the aerodynamic efficiency as a function of venting geometry, venting location and orientation for a series of recreational production helmets currently available in Australian market. Keywords: Bicycle helmets, aerodynamic drag, wind tunnel, yaw and pitch angles.
1- Introduction Bicycle helmets are mandatory for recreational or professional bicycle riders in many countries including Australia. Over the past decade the designers and manufacturers have been engaged in helmet developments, which can comply with the safety standards. The regulations set the standards of impact protection to reduce head injury during serious accidents. Currently, there are no such standards concerning helmet ventilation or the effects associated with thermal stress and aerodynamic efficiency. It has only been recently that attention of such areas in helmet design has been implemented mainly pursued in the professional arena. In the pursuit for greater thermal comfort there is a trend to increase the number of ventilation openings. Unfortunately increasing the ventilation of a helmet not only degrades the structural integrity of the helmet but also increases the aerodynamic drag. At around 30 km/h speed, the aerodynamic resistance (drag) constitutes almost 80 percent of total resistance (remaining 20% is rolling resistance). Out of total aerodynamic drag, the rider position counts approximately 65 to 80 percent depending on body position, helmet and clothing. The remaining drag is coming from bicycle frames, wheels (mainly front wheels) and other components and adds-on (see Kyle and Weaver 2004) for details). Although, the amount of aerodynamic drag from the helmet is approximately 2 to 8 percent depending on the aerodynamic shape 1. School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University, 264 Plenty Road, Bundoora, Melbourne, VIC 3083, Australia - E-mail: [email protected]
338 The Engineering of Sport 7 - Vol. 1 of the helmets (at around 30 km/h cruising speeds), the use of an aerodynamically efficient helmet can play a vital role by providing an advantage both at recreational and racing bicycle riding. Although several studies (by Alam et al. 2007, 2006 & 2005, Bruhwiler 2003 and Reid and Wang 2000 & 1999) were conducted to measure the aerodynamic drag and temperature measurement techniques for bicycle helmets, these studies were not comprehensive and most studies except Alam et al. did not consider the aerodynamic efficiency at all. Hence, the primary objective of this work as a part of a larger project is to study the aerodynamic and thermal efficiency (comfort) of a series of current production bicycle helmets available in Australia for recreational uses as a function of air speeds, pitch and yaw angles. However, in this paper, only the aerodynamic property such as drag will be considered.
2- Experimental Procedure 2.1 Facilities and Equipment The study was conducted in RMIT Industrial Wind Tunnel. It is a closed return circuit wind tunnel with a turntable to simulate the cross wind effects. The maximum speed of the tunnel is approximately 150 km/h. The dimension of the tunnel’s test section is 3 m wide, 2 m high and 9 m long and the tunnel’s cross sectional area is 6 square meter. The experimental set up in the test section of RMIT Industrial Wind Tunnel is shown in Figure 2. The tunnel was calibrated before conducting the experiments and tunnel’s air speeds were measured via a modified NPL ellipsoidal head Pitot-static tube (located at the entry of the test section) connected to a MKS Baratron pressure sensor through flexible tubing. A mounting stud was manufactured to hold the dummy head and the helmet and was mounted on a six component force sensor (type JR-3). Purpose made computer software was used to compute all 6 forces and moments (drag, side, lift forces, and yaw, pitch and roll moments) and their non-dimensional coefficients.
2.2 Description of Helmets Five production helmets were used for this study. All helmets are different in terms of venting holes, venting location and venting geometry. All helmets were new and manufactured by Rosebank Australia, a subsidiary of Dunlop Pacific. These helmets are: Blast, Mamba, Nitro, Summit and Vert (see Figure 1). The aerodynamic drag were measured under a range of wind speeds (20, 30, 40, 50 and 60 km/h), yaw angles (0, ±30º, ±60º and ±90º) and pitch angles (90º, 60º, 30º and 0º from horizontal axis) using a six component force sensor. A dummy head was used to measure the aerodynamics forces of all helmets in the wind tunnel. In order to understand the effects of no venting, the Vert helmet was modified and tested twice as standard configuration with the venting and modified configuration without the venting (venting blanked off), see Figure 1.
The Engineering of Sport 7 - Vol. 1
339
2.2.1 Description of Venting Geometry- Effects of Venturi In order to understand the effects of geometry on aerodynamic and thermal performance, the geometry of venting (cross sectional area), venting orientation and venting heights were considered and compared results for each helmet against these three parameters. The venting is generally designed to promote the transfer of heat from the head through forced convection. This forced convection is achieved by varying the inlet and outlet areas of the vents. Figures 3 & 4 show the channelled airflow in a flute configuration to accelerate the flow over the head and achieve optimal quantities of heat removal. In this study, the mechanism that involves in increasing the airflow through the helmet is called the Venturi Effect. Unfortunately, the Venturi effect introduces airflow resistance over and through the helmet. The air velocity at the exit of the vents can be defined as (1) Where, A1 and A2 are the cross sectional areas of entry and exit of venting, and V
is the velocity of air at the entry of venting. The relationship describes the potential for the vents to increase the inlet velocities through the venting. It is an important parameter to assess a helmet’s potential heat removal and aerodynamic drag generated by the venting obstructions.
2.2.2 Description of Venting Geometry- Effects of Venting Orientation The orientation of venting can have significant effects on aerodynamic and thermal performance of the helmet. The orientation of venting effectiveness can be described as a ratio of venting longitudinal length and venting lateral width times the cosine angle with the wind direction as shown in Figure 3. The venting is generally more effective when the longitudinal length (L) is parallel the incoming airflow. This is to ensure that the vents channel the flow with minimal flow disruption. Therefore the venting system can be defined in terms of its overall effectiveness as the culmination of the venturi effect and the venturi potential to channel incoming flow which is adversely affected by the orientation to the flow. This can be defined as (2)
Where Figure 5.
for in line airflow and
for oblique airflow, see
2.2.3 – Description of Venting Geometry- Effects of Venting Height (Roughness) Variation The aerodynamic performance of a helmet can also be affected by the helmet surface roughness. This is generally a result of how the vents protrude from the surface of the helmet. Figure 5 shows the venting angle, which is adequate for capturing the airflow but
340 The Engineering of Sport 7 - Vol. 1 can impede the flow over the helmet that can in turn affect the overall drag of the helmet. Figure 3 shows the definitions for the geometric relationship for the roughness factor. This concept is an important issue when designing a helmet for maximum drag reduction and can significantly impede airflow if the height deviation for each vent is large. The roughness factor may be defined as (3) The venting height deviation (roughness factor) specifies the roughness magnification of the helmet. Generally, the higher the roughness factor, the higher the aerodynamic resistance (drag). The roughness of the helmet is mostly due to the vent design and ideally needs to keep minimum in order to minimize the aerodynamic drag.
Figure 1 - A plan view of helmets and their dimensions.
The Engineering of Sport 7 - Vol. 1
Figure 2 - Experimental setup in the test section with a dummy head and helmet
Figure 3 - Venting height deviation (Roughness factor)
Figure 4 - Venting geometry (Venturi)
Figure 5 - Orientation of venting
341
3- Results 3.1 – Effects of Reynolds numbers on Aerodynamic Drag Coefficients The drag forces were converted to non-dimensional parameter, drag coefficient (CD) and are plotted against the Reynolds number (
) varied by air speeds only. All
results are shown for zero yaw angles only. The CD as a function of Reynolds numbers under a range of pitch angles (90º, 60º, 30º and 0º) are shown in Figures 6 to 10. There was no significant variation in CD with Reynolds numbers for 90º and 60º pitch angles for all helmets. However, a slight variation was noted for the Vert helmet. However, a significant variation was noted for 30º pitch angle and a lesser extend at 0º pitch angle (see Figures 6 to 10). Flow visualisation by smoke (photographs not shown here) indicated that the air flow around the helmets at 30º and 0º pitch angles was significantly complex and turbulent compared to 90º and 60º pitch angles. It may be noted that pitch angles (30º to 0º) are not practical for the recreational bicycle riders and the results for these angles are not significantly important.
342 The Engineering of Sport 7 - Vol. 1
Figure 6 - CD variation with Re (Mamba helmet).
Figure 7 - CD variation with Re (Blast helmet).
Figure 8 - CD variation with Re (Nitro helmet).
Figure 9 - CD variation with Re (Summit helmet)
Figure 10 - CD variation with Re (Vert helmet)
Figure 11 - Effects of Venturi on CD
3.2 – Effects of Venturi on Aerodynamic Drag Coefficients The venting effects on the aerodynamic drag coefficients for all helmets are shown in Figure 11. With an increase of the venturi effectiveness, the aerodynamic drag increases for all helmets. The figure also indicates that the mamba helmet which has the highest venturi effectiveness, generates highest aerodynamic drag. On the other hand the vert exhibits the opposite trend. The vert helmet possesses the lowest of venturi effectiveness and the lowest drag coefficient. It indicates that the existence of not just vents, but larger
The Engineering of Sport 7 - Vol. 1
343
channelled vents such as the mamba, increase the aerodynamic drag coefficient significantly. The venting system is generally complex and needs more study to understand venturi effectiveness of bicycle helmets. The similar results were also obtained for the blast helmet which faired much the same as the mamba.
3.3 Effects of Venturi Roughness on Aerodynamic Drag Coefficients The relationship between the helmet surface roughness factor and the aerodynamic drag coefficient for all helmets is shown in Figure 12. The summit helmet has the highest value of surface roughness factor and also it demonstrates the highest value of aerodynamic drag coefficient. Generally higher the roughness factor, greater the venting obstruction of airflow over the helmet. Figure 12 clearly indicates this trend. However, it is believed that the aerodynamic drag can be minimized with the smoother and less obtrusive vents on the helmets as there is a direct correlation between the vent roughness factor and the pressure induced drag. The Vert helmet has minimal venting holes and less obtrusive (minimum roughness factor), produces minimum aerodynamic drag. The study also shows that having a large amount of vents is not so critical if the roughness of the vents is minimal as higher roughness can trigger the airflow separation over the helmet.
3.4 Effects of Venting Roughness on Orientation Sensitivity Aerodynamic Drag Coefficients The sensitivity of the helmet is based on the ratio of drag coefficients in two different orientations. The greater the ratio; higher the possibility of venting roughness will play a major role in reducing the aerodynamic performance. The statement is evident in the Vert helmet which has no change in the orientation factor. The aerodynamic drag (due to induced pressure) will be almost independent of venting orientation if the venting roughness can be kept minimal. However, it is not the case when vent’s numbers with larger dimensions are significant. This observation is evident in Summit and Blast helmets (see Figure 13). As venting orientation is fixed, the helmet will generate induced drag under crosswind condition (under large yaw angles).
Figure 12 - Effects of venting roughness on CD. Figure 13 - Effects of venting roughness on orientation.
344 The Engineering of Sport 7 - Vol. 1
4- Concluding Remarks The three main areas of the bicycle helmet have been discussed at length and provide a better insight to the results obtained experimentally and also demonstrate the influence of geometry upon the performance of a helmet. This relationship between experimental data and the physicality of the helmets can also provide designers with a better appreciation into the impacts of the helmet design on performance, and thus can be adopted as an approach to the design of new helmets.
5- Acknowledgements The authors would like to express their sincere thanks to Mr Anthony Resta and Mr Rhys Solomons for their assistance with the testing and data analysis. We are also indebted to Rosebank Australia for providing the helmets.
6- References [1] Alam, F., Subic, S., Akbarzadeh, A. and Watkins, S., “Effects of Venting Geometry on Thermal Comfort and Aerodynamic Efficiency of Bicycle Helmets”, in The Impact of Technology on Sport II (edited by F. K. Fuss, A. Subic and S. Ujihashi), pp 773-780, ISBN 978-0-415-45695-1, Taylor & Francis, London, 2007. [2] Alam, F., Subic, A. and Watkins, S., “A Study of Aerodynamic Drag and Thermal Efficiency of a Series of Bicycle Helmets”, Proc. of the 6th International Conference on Engineering of Sports, ISEA, 11-14 July, Munich, Germany, 2006. [3] Alam, F., Watkins, S. and Subic, A., “Aerodynamic efficiency and thermal comfort of bicycle helmets”, Proc. of the 6th International Conference on Mechanical Engineering (ICME2005), TH-32 (1-6), ISBN 984-32-2846-4, 28-30 December, Dhaka, Bangladesh, 2005. [4] Bruhwiler, P. A., “Heated, perspiring manikin headform for the measurement of headgear ventilation characteristics”, Meas. Sci. Technol. 14, pp 217-227, 2003. [5] Kyle, C. R. and Weaver, M. D., “Aerodynamics of human-powered vehicles”, Proceeding of the Inst. Mech Engs, Vol 218, Part A: J. Power and Energy, pp 141-154, 2004. [6] Reid, J. and Wang, E. L., “A system for quantifying the cooling effectiveness of bicycle helmets”, J Biomech Eng. 122 (4), pp 475-460, 2000. [7] Wang, E. and Reid, J., “Designing Bicycle Helmets for Thermal Comfort”, ASME Summer Bioengineering Conference, June 16-20, Big Sky, Montana, USA, 1999.
Aerodynamics of Cricket Ball-an Effect of Seams (P70) Firoz Alam, Roger La Brooy, Aleksandar Subic, Simon Watkins1
Topics: Aerodynamics. Abstract: A cricket ball possesses six rows of stitches with approximately 60 to 80 stretches in each row and a prominent seam at the joining of two halves. Asymmetric airflow over the ball due to seam orientation and surface roughness of the ball generally causes the flight deviation (swing) and the unpredictable flight. Swing makes difficult for the batsman to hit the ball with his bat and guard the stamps. The primary objectives of this work were to understand the aerodynamic properties of cricket balls, thus the mechanism of swing of a cricket ball. The aerodynamic forces and moments were measured under a range of speeds at various seam orientations. The airflow around the cricket ball was visualised and documented with video and still images Keywords: Swing, drag, wind tunnel, seam, flow visualization, cricket ball.
1- Introduction The cricket game is 2nd most popular and widely watched games following various forms of football in the world with a potential viewing audience of over 1.5 billion people across all continents. One of the centre pieces of a cricket game is the cricket ball. A cricket ball is constructed of a several layers of cork tightly wound with string. The ball is covered with a leather skin comprising 4 quarters stitched together to form a major seam in an “equatorial” plane. Moreover the quarter seams on both halves of the ball are internally stitched and juxtaposed by 90 degrees. The seam comprises six rows of stitches with approximately 60 to 80 stretches in each row. A cricket ball with a mass of 156gm and approximate diameter of 70 mm (diameter of a tennis ball is approximately 54 mm) is much heavier than a tennis ball. However, the prominence of the seam and mass can vary from one manufacturer to another. At present, over a dozen firms manufacture cricket balls, under the auspices of the International Cricket Council, overseeing the game at the highest level (Test cricket). 1. School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University, 264 Plenty Road, Bundoora, Melbourne, VIC 3083, Australia - E-mail: [email protected]
346 The Engineering of Sport 7 - Vol. 1 The aerodynamic properties of a cricket ball are affected by the prominence of the seam, the surface roughness of the ball in play and the launch attitude of the ball by the bowler. Asymmetric airflow over the ball then causes the flight deviation (swing). One of the objectives of this research is to understand the aerodynamic properties of the ball and to explain the mechanism of swing and de-mystify the unpredictability of the ball’s trajectory. The sideway deviation of the ball during the flight towards the batsman is called swing. There are various types of swing: conventional swing and reverse swing. Conventional swing results in the ball experiencing a sideways force directed away from the shiny half of the ball. Such a force is achieved by maintaining laminar boundary layer of air flowing over the shinny or smooth half with a turbulent boundary layer of air flowing over the rough half. Roughness over one half of the ball is a result of its natural deterioration during play whilst a shiny side is maintained by polishing the ball when the opportunity presents itself to the fielding team. Conventional swing can be achieved in at least two ways: a) By angling the seam to the batsman and with the mean direction of the flight so that one side experiences laminar (smooth) airflow and other side experiences turbulent airflow caused by the angulation of the seam itself. The points on each side of the ball half where the flow separates is asymmetrical, generating aerodynamic pressure variations with a component transverse to the ball’s motion causing eventual trajectory deviations. Generally, the ball is pushed towards the half where the airflow is remains attached. b) By bowling a deteriorated ball possessing shiny and rough halves to a batsman. However, by aligning the ball’s seam under some angles, the bowler can generate different type of swings which will be discussed later. Generally, if the ball moves in a direction away from the bat, the deviation is called out swing. Conversely when the ball moves towards the batsman the resulting sideway deviation is called inswing. Figure 1 illustrates a typical swing of a cricket ball to a righthanded batsman. A typical cricket ball used in One Day and Test game is shown in Figures 2 & 3 respectively.
Figure 1 - A schematic of outswing & inswing [ref 9].
Figure 2 - A typical Test cricket ball.
Figure 3 - A typical One Day cricket ball.
The Engineering of Sport 7 - Vol. 1
347
Reverse swing is generally achieved when the airflow becomes turbulent on both sides of the ball. Here the turbulent airflow at sides separates earlier on one side than the other. The phenomenon may occur with a ball that is bowled fast. It is not clear where the limits of velocity exist for this type of swing. A comprehensive study is required to answer this question. In reverse swing, unlike conventional swing, the ball deviates toward the rough side of the ball. Usually, any swing makes difficult for the batsman to hit the ball with his bat and guard the stamps. Traditionally reverse swing occurs when one half of the ball is has been naturally worn significantly. In most cricket matches, the phenomenon of reverse swing believed to be occurred after 40 or more overs (one over consists of a set of six bowled balls). The mechanism for a reverse swing is complex and still not fully understood due to the degree of ball’s surface roughness and required seam alignment angles with the mean direction of the flight. Although, some studies by Alam et al. [1, 2], Barrett et al. [3], Mehta [4, 5, 6], Sayers and Hill [7], Wilkins [8] were conducted to understand the aerodynamics of cricket ball, a comprehensive study to understand the gamut of complex aerodynamic behaviour resulting a wide range of swing under a wide range of wind conditions, relative roughness, seam orientations and seam prominence is yet to be conducted. Therefore, a large research project has been undertaken in the School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University to understand the overall aerodynamic behaviour of the cricket ball both experimentally as well computationally. The work presented here is a part of this large research project.
2- Experimental Procedure 2.1 Facilities and Equipment The study was conducted in RMIT Industrial Wind Tunnel. It is a closed return circuit wind tunnel with a turntable to simulate the cross wind effects. The maximum speed of the tunnel is approximately 150 km/h. The dimension of the tunnel’s test section is 3 m wide, 2 m high and 9 m long and the tunnel’s cross sectional area is 6 square meter. The experimental set up in the test section of RMIT Industrial Wind Tunnel is shown in Figures 1 & 2. The tunnel was calibrated before conducting the experiments and tunnel’s air speeds were measured via a modified NPL ellipsoidal head Pitot-static tube (located at the entry of the test section) connected to a MKS Baratron pressure sensor through flexible tubing. A mounting stud was manufactured to hold the ball and was mounted on a six component force sensor (type JR-3). Purpose made computer software was used to compute all 6 forces and moments (drag, side, lift forces, and yaw, pitch and roll moments) and their non-dimensional coefficients.
2.2 Description of Balls In order to understand the general behaviour of the airflow around a cricket ball, a large cricket ball (450 mm in diameter) was constructed to visualise airflow using smoke trails. Additionally, three new and two old practice cricket balls including a One-Day
348 The Engineering of Sport 7 - Vol. 1 International ball were used. These balls (made by Kookaburra, Australia) are: Turf white (One Day game ball), Regulation (Test game ball), Club Match (First class game ball), used ball (one side smooth and other side rough), and practice ball (one side white and other side rough & red). These balls except the club match are shown in Figure 7. The aerodynamic properties (drag, lift and side force and their corresponding moments) under a range of wind speeds (40 km/h to 140 km/h in increments of 20 km/h) at seam angles of 0°, 10°, 20°, 30° and 45°with the mean direction of winds were measured. It is believed that a practical seam angle during bowling is approximately 20 to 30 degree with the mean direction of flight. The larger diameter (450 mm) ball with six-rows of artificial seams (shown in Figure 6) was used to visualise the airflow around it. The larger diameter ball was photographed to check flow characteristics and quantify the flow effects with seam orientation angles with wind directions. Smoke was used to see the airflow trail around the ball. Six rows of seam were replicated using a siliconerubber compound and the surface roughness at one side of the ball was also replicated using the same material. Airspeeds ranged from 10 km/h to 40 km/h for the smoke flow visualisation. The forces acting on the real balls were determined by testing ball with the supporting gear and then subtracted from the forces acting on the supporting gear only. A cricket ball with the mounting device on a six component force sensor is shown in Figures 4 & 5.
Figure 4 - Experimental Set-up in RMIT Industrial Wind Tunnel.
Figure 5 - A Motorised Supporting Device for Cricket Ball mounted on a Six-Component Force Sensor.
Figure 6 - A simplified 450mm diameter cricket ball in the RMIT Industrial Wind Tunnel.
The Engineering of Sport 7 - Vol. 1
349
Figure 7 - Various Kookaburra Cricket Balls used in this study.
3- Results 3.1 Flow Visualisation Airflow characteristics under various seam orientation are shown in Figures 8 to 12. Figure 8 shows the flow separation occurring at the ball’s apex (90 degrees) whilst the seam remains parallel to the wind direction (zero seam angle with the horizontal axis) reaffirming the classical flow separation point from a sphere. Here, the seam does not play any role in triggering flow separation at all. However, when seam is angled to the flow direction (shown here at approximately 30 degrees with surface roughness elements located upstream of the seam, the flow no longer separates at the apex (as it was in Figure 8), but accelerates and separates at around 30 degrees past the apex. The surface roughness and the seam then enable turbulence and enhance the delayed flow separation. When seams are artificially placed at approximately 70 degrees to the flow direction (see Figure 10), the airflow separation is still delayed but not as much as was the case in Figure 9. Here the seams and surface roughness locations are close to the natural trigger of flow separation (see Figure 8). In Figure 11, the seams and surface roughness location trigger the natural flow separation as they are located at the critical zone (apex) and the airflow separates earlier than in the case of Figure 8. The airflow characteristics in Figure 12 show the similar pattern as in Figure 8. In this case, both surface roughness and seams do not play any role in the flow separation at all.
3.2 Effects of Seams All five balls were tested in the wind tunnel at a range of speeds under a range of yaw angles as mentioned earlier. The ball was yawed relative to the force balance (which was fixed with its resolving axis along the mean flow direction whilst a ball was yawed above
350 The Engineering of Sport 7 - Vol. 1 it) thus the wind axis system was employed. The Regulation and Turf (white) balls are new with no wear and tear. However, the red-red and white-red balls are used balls and one side is of the ball is rougher than other (see Figure 7). The forces and moments were converted to nonOriginal dimensional parameters (CD, CS, CL and their moment coefficients) and tare forces were removed by measuring the forces on the sting (mounting device) in isolation and them removing them from the force of the ball and sting. As mentioned previously, the side way deviation of a cricket ball in the flight is generally due to the side force variation. However, the variation of side force of a cricket is very small and it is relatively difficult to measure. The side forces measured in this study are difficult to interpret as they are within experimental error. A more sensitive force sensor will be used to accurately quantify these side forces. Therefore, in this work, only the drag forces for all five balls are presented and they are plotted against Reynolds numbers (varied by the tunnel wind speeds) as a function of yaw angles. These plots are shown in Figures 13 to 16.
Figure 8 - Seam orientation Figure 9 - Seam orientation parallel to flow direction approximately 30 degrees (0 degrees). with.
Figure 11 - Seam orientation approximately 90 degrees with flow direction.
Figure 10 - Seam orientation approximately 70 degrees with.
Figure 12 - Seam orientation parallel to flow direction (30 degrees).
The Engineering of Sport 7 - Vol. 1
Figure 13 - Drag Coefficient variation with Reynolds numbers as a function of seam angle (Turf white).
351
Figure 14 - Drag Coefficient variation with Reynolds numbers as a function of seam angle (Regulation).
A comparison of drag coefficients at all speeds and yaw angles for all five cricket balls indicates that there is a slight variation of drag coefficients between new and used balls. The new ball’s drag coefficient (CD) at high Reynolds numbers (above 100 km/h) is approximately 0.5 regardless of seam orientation. The findings agreed well with the published data (see Sayers and Hill [8]). The seam angles have negligible effect on drag coefficients at high speeds on Kookaburra made balls. However, a small variation in drag coefficients was noted for two used balls (see Figures 15 & 16). The used practice ball (half white and half red) shown in Figure 16, possesses slightly higher drag coefficient in all speeds tested. However, the used ‘Red-Red’ (one side smooth &other side rough) ball has the lowest aerodynamic drag coefficient at higher speeds (0.45). It is not clear why the old ball displays laser drag coefficient compared to new balls.
Figure 15 - Drag Coefficient variation with Reynolds numbers as a function of seam angle (One side smooth & other side rough).
Figure 16 - Drag Coefficient variation with Reynolds numbers as a function of seam angle (Half white & half red with roughness)
3- Concluding Remarks The following conclusions were made from the work presented here: • The airflow around a cricket ball is complex due to the surface roughness, seams and spins involved. • The drag coefficient of a cricket ball is approximately 0.5 at speeds over 100 km/h. There is minimum variation in aerodynamic drag coefficient between new and used balls made by Kookaburra.
352 The Engineering of Sport 7 - Vol. 1 • The seam location close to the apex of the ball triggers early flow separation (evidenced by flow visualisation). The seam location close to the mean direction of the airflow (horizontal axis) enhances the delayed flow separation (evidenced by flow visualisation). • Further flow visualisation is required to quantify the exact location using a real cricket ball.
4- Acknowledgements The authors would like to express their sincere thanks to the Australian Cricket Control Board (Cricket Australia) for their financial support for this work. Special gratitude is also due to Mr Peter Thompson, Kookaburra Australia for providing the balls and advice.
5- References [1]. Alam, F., La Brooy, R. and Subic, S., “Aerodynamics of Cricket ball- An Understanding of Swing”, in The Impact of Technology on Sport II (edited by F. K. Fuss, A. Subic and S. Ujihashi), pp 311-316, ISBN 978- 0-415-45695-1, Taylor & Francis, London, 2007. [2]. Alam, F., La Brooy, R. Watkins, S. and Subic, S., “An Experimental Study of Cricket Ball Aerodynamics”, Proceedings of the 7th International Conference on Mechanical Engineering (ICME2007), pp ICME07-FL- 17 (1-6), 29-31 December, Dhaka, Bangladesh, 2007 [3]. Barrett, R. S. and Wood, D. H., “The theory and practice of reverse swing”, Sports Coach, Vol 18, pp 28- 30, 1996 [4]. Mehta, R. D., “Cricket Ball Aerodynamics: Myth vs Science” in Engineering of Sports, pp 153167, Blackwell Science, Oxford, 2000 [5]. Mehta, R. D. and Wood, D. H., “Aerodynamics of the Cricket Ball”, New Scientist, Vol. 87, pp 442-447, 2000 [6]. Mehta, R. D., “Aerodynamics of Sports Balls”, Annual Review of Fluid Mechanics, Vol. 17, pp 151-189, 1985 [7]. Sayers, A. T and Hill, A., Aerodynamics of a cricket ball”, Journal of Wind Engineering and Industrial Aerodynamics, Vol. 79, pp 169-182, 1999 [8]. Wilkinson, B.,“Cricket: The Bowler’s Art”, Kangaroo Press, Australia, ISBN 0-86417-899-9, pp 1-191, 1997 [9]. www.cricketfundas.com/cricketcoachingjan0907swing
Numerical Modelling of the Flow Around Rowing Oar Blades (P71) Anna Coppel1, Trevor Gardner1, Nicholas Caplan2, David Hargreaves3
Topics: Sailing/Water Sports; Modelling; Virtual Reality & Computer application in Sports Abstract: Recent experimental studies have presented the lift and drag coefficients for a range of oar blades. However, these studies were not able to provide insight into the mechanisms of lift and drag generation. The aims of this study, therefore, were to model the flow around an oar blade using computational fluid dynamics (CFD), and to validate this model against experimental data. A CFD model of the flow around oar blades was constructed using Fluent (ANSYS Inc., USA). To evaluate the performance of the model the geometries of the blade, domain and fluid free stream velocity of the CFD model simulated published quarter scale experimental measurements. The geometries were determined directly from the quarter scale blades used previously, and the blades were tested at a range of static angles of attack, representative of the motion of the blade from catch to finish. In a second series of simulations, the blade geometries and the fluid velocity were scaled to full size. A k- SST RANS turbulence model was used as the solver, as it is capable of resolving the turbulence at the blade surface (in the near wall region) and in the bulk of the flow, at all angles of attack. A close agreement was achieved between experimental and CFD lift and drag coefficients, demonstrating the validity of the model. When CFD simulations of the blade geometry and fluid velocity were scaled to full size, lift was similar to quarter scale, but drag was reduced at all angles of attack. This reduction in drag at full size can be explained by the shape of the blade. For a flat plate with similar Reynolds number to those seen here, drag should be constant at increasing velocity. However, the curvature of the oar simulated here influenced fluid flow over and around the blade. Keywords: computational fluid dynamics; validation; rowing; oar blades.
1. The University of Birmingham, School of Sport and Exercise Sciences, Birmingham, B15 2TT E-mail: alc640*,[email protected] 2. Northumbria University, School of Psychology and Sport Sciences, Newcastle Upon-Tyne, NE1 8ST E-mail: [email protected] 3. University of Nottingham, School of Civil Engineering, Nottingham, NG7 2RD E-mail: [email protected]
354 The Engineering of Sport 7 - Vol. 1
1- Introduction Despite advances in oar blade design through the use of lightweight composite materials, there has been little research into the fluid dynamic properties of the blades and the work that has been undertaken to determine the forces exerted by the oar blade through the stroke have been largely experimental. Much of the research has been carried out by the oar blade manufacturers Concept II (Vermont, USA) and also by Kleshnev (1998) with on-water tests. More controlled laboratory experiments were carried out by Barré and Kobus (1998) who adopted a dynamic approach and by Caplan and Gardner (2007) who employed a quasi-static setup. For the quasi-static models, the oar is held fixed at discrete angles of attack and the fluid flows over it. Information about a physical process is ideally provided by experimental measurement on actual full-size equipment to predict how identical copies of the equipment would perform under the same conditions. However, such full size tests are, in most cases, prohibitively expensive and often unfeasible. The alternative is to perform experiments on small scale models with the resulting problem that measurements must then be extrapolated to full-scale using dimensional analysis. (Patankar, 1980). Both the laboratory experiments mentioned above had to use small scale oar blades, Barré and Kobus used a scale of 0.7 while Caplan and Gardner used quarter scale blades. The alternative is to predict the fluid flow characteristics by using a theoretical model that represents a physical phenomenon through a series of non-linear equations that are then solved by a computer. The main concern with using a computational fluid dynamic (CFD) model is proving the validity of the mathematical model and the numerical methods. Where a new phenomenon is investigated there is no doubt that experiment leads and computation follows (Patankar, 1980). However, with sufficient validation of the computed results by comparison with experimental data, computational methods can then be used with confidence. This validation is addressed in the first part of this study where oar blade lift and drag coefficients derived by the CFD model are compared with the experimental data of Caplan and Gardner (2007). Once a computational result has been validated against experimental data it is possible to test situations that may not easily be replicated through experiment. The use of CFD to simulate the lift and drag of full size oar blades which was unfeasible experimentally is addressed in the second part of this study.
2- Experimental Data The CFD predicted results were compared with experiments undertaken by Caplan and Gardner (2007). The flume had a free stream width and depth of 0.64m and 0.15m respectively. Due to the inherent edge resistance effects on the free stream velocity and the restrictions on flume width and height it was necessary to use quarter scale models, so the length of the blades were less than a quarter of the flume width, corresponding to a blockage factor (ratio of projected area of model to test section) of 0.052. With any fluid dynamics test requiring the use of scaled models, a dimensional analysis must be performed to ensure actual full size conditions are being modelled (Taylor, 1974). As the model blades were scaled exactly from the full size blades, geome-
The Engineering of Sport 7 - Vol. 1
355
tric similarity was met. A ratio of 1:4 was chosen for the length scale, which in terms of the characteristic length, L, is: (1) where, the subscripts m and p represent model blade and prototype blade respectively. Since in rowing the blade is close to the water/air interface it is necessary to match the non-dimensional parameter Froude number, Fr, for the model and prototype blades (Hartwanger & Bunt, 1996, Barré and Kobus, 1998). If the model and prototype Froude numbers are equated this will give us a ratio of the model to prototype blade free stream velocities, v, of: (2) If the prototype relative blade velocity, vP, is 5 ms-1, then it follows from equation (2) that the required model relative blade velocity in the flume, vm, is 2.5 ms-1. Unfortunately the maximal velocity in the experimental flume was restricted to 0.85 ms-1, which prohibited the correct model velocity being achieved, resulting in dynamic similarity not being met. To circumvent this problem, Reynolds number dependence on the lift and drag coefficients was studied by Caplan & Gardner (2007) for a flat plate and it was found that with a flume velocity of above 0.7 ms-1 the coefficients were independent of Reynolds number. This idea of Reynolds number independence is corroborated in literature by Munson et al. (2002) and Hartwanger & Bunt (1996) who state that for a fully immersed bluff body, such as a flat plate, in the range of Reynolds numbers considered, that a CD versus Re curve (drag coefficient, and Reynolds number, respectively) will level off to a flat line, when 105Re106. Thus, Reynolds number is not expected to be critical for objects that are fully submerged, provided the flows are turbulent. Caplan & Gardner (2007) made the assumption that if this was true for a flat plate, similar comparison could be made for rowing oar blades which only have a shallow curve. Therefore, a fluid velocity of 0.75 ms-1 was used by Caplan and Gardner which they believed was high enough to overcome any influence of Reynolds number up to a velocity of 2.5 ms-1, but also low enough to ensure that any effects arising from the walls of the flume would not adversely influence the validity of the results. They used a variety of different oar blade designs in their experiments at a range of angles of attack, , to the free stream. In this investigation only the Macon blade will be used for comparison with the CFD data.
3- Methods 3.1 Validation The first step is to ensure that the CFD predicted results are providing an accurate model of the fluid dynamic behaviour. The numerical modelling undertaken involves the solution of the Navier Stokes equations which are based on the assumptions of the conservation of mass and momentum within a moving fluid. The conservation of mass in tensor form is described by the differential equation:
356 The Engineering of Sport 7 - Vol. 1
(3) where, is the density and u is the component of the velocity of the fluid in the xdirection. The conservation of momentum is similarly described by the equation: (4) where, is the pressure, g is gravitational acceleration, and is the stress tensor. A detailed derivation of the solutions of the Navier Stokes equations may be found in Ferziger and Peric (1997). These equations are non-linear and coupled, and are subsequently very difficult to solve directly so numerical methods are used to obtain an approximation within the desired accuracy. The CFD software Fluent v6.3 (ANSYS Inc., USA) was used which relies on the finite volume method. In order to simulate the effects of turbulence on the flow, additional transport equations are solved. The robust k- model (Launder and Sharma, 1974) was employed as the preliminary choice of turbulence model. This is one of a range of models classified as the Reynolds Averaged Navier-Stokes (RANS) models. The k- model is known to be of limited accuracy when used to model the flow over objects with a curved boundary layer such as an oar blade. However, it is good practice to start simulations with a robust model like the k- before switching to a more complex and accurate turbulence model. Therefore, the k- Shear Stress Transport (SST) model (Mentor, 1993) was subsequently investigated as it would provide more accuracy in the near wall region due to it solving the accurate k- model in the near wall region. The k- SST model also has the advantage that it gives accurate results for a wide range of grid densities and has shown better potential to predict the key features of separated flows, which can occur with oar blades, than other models (Rhee and Koutsavdis, 2005). The sensitivity of the results to these turbulence models and to a number of discretization schemes was tested. Figure 1 shows the layout of the computational domain as seen from above and through its cross-section. As a number of blade angles were investigated it was necessary to create an inner mesh containing the blade (see circular region in overhead view), while a second mesh was also produced containing the rest of the flume. This allowed the inner mesh to be rotated to the desired blade angle using the non-conformal interfaces within Fluent. Also shown are the boundary conditions imposed and their position. A uniform velocity at inflow to the flume of 0.75 ms-1 was specified, and a symmetry condition on the upper boundary (the water/air interface) of the flume was set. A no-slip condition was applied at the flume walls and blade surfaces which were both assumed to be smooth since the experimental flume was constructed of PVC and glass and the blade was aluminium plate. The inner mesh consisted of unstructured tetrahedral cells with size functions and boundary layers attached to the faces of the blade to improve accuracy in the near wall region. In the outer mesh, structured hexahedral cells were generated by using the Cooper algorithm from top to bottom. A mesh size of 350,000 resulted. The simulations were run until lift and drag coefficients stopped varying by more than 0.01% and all the
The Engineering of Sport 7 - Vol. 1
357
other typical CFD residuals were monitored too to ensure they had fallen below acceptable convergence levels.
Figure 1 - Schematic showing the layout and boundary conditions of the domain.
3.2 Comparison of full size to scaled models Once validated the CFD model was used to simulate a full size Macon blade to test the assumption of Caplan and Gardner’s that force coefficients will remain constant above a relative blade velocity of 0.75 ms-1. For this to be true the CFD computed drag coefficients of a quarter scale blade with a relative blade velocity of 0.75 ms-1 should match the quarter scale blade at a velocity of 2.5 ms-1 and a full size blade at 5.0 ms-1. The layout and boundary conditions are as before with the dimensions of the computational domain being four times larger than the water flume used in the experiments of Caplan and Gardner and the blade corresponded to the dimensions of a typical full size Macon blade. The cross-sectional shape of the blade is shown in Figure 1.
4- Results and Discussion 4.1- Validation results The primary measure of accuracy between experimental (measured) and CFD (predicted) results was the ability of the CFD models to predict the lift and drag coefficients at a number of angels of attack. Figure 2 shows the drag coefficients for the Macon blade plotted against angle of attack (in radians). Comparison is made between the measured and experimental data and the CFD predictions using the different turbulence models and discretization schemes. It can be seen that the k- SST model shows the best correlation at all angles of attack for the drag coefficient and it was most successful when second order differencing was used. Similarly sucessful results were found for the lift coefficient.
358 The Engineering of Sport 7 - Vol. 1
Figure 2 - Effect of turbulence model on the variation of drag coefficient with angle of attack for the Macon blade.
Since the CFD model has been validated through accurately predicting the force coefficients around the Macon blade, it is now valid for us to use the same model to predict the flow conditions at full scale and with different velocities.
4.2 Comparison of full size to scaled results Figures 3 and 4 present the lift and drag coefficients, respectively, at the range of angles of attack tested for the full size Macon blade, using a relative blade velocity of 5 ms-1. The quarter scale force coefficient profiles for a relative blade velocity of 2.5 ms-1 and 0.75 ms-1 were also compared to test the validity of Caplan and Gardner’s assumption. For their assumption to be correct force coefficients should remain constant above a relative blade velocity of 0.75 ms-1. Looking initially at Figure 3, which presents the variation of lift coefficient with changing velocity and angle of attack, it would seem that this assumption is correct as there is little variation between the quarter and full size values at the different velocities. However, when we look at Figure 4, the variation of drag coefficient, the hypothesis appears invalid. Comparing the drag coefficients of the full size blade with a velocity of 5 ms1 and the quarter scale blade at 0.75 ms-1 it can be seen that at all angles the drag is overestimated at quarter scale.
The Engineering of Sport 7 - Vol. 1
359
Figure 3 - Lift coefficient against angle of attack comparing CFD simulations of full size Macon at 5ms-1, a quarter scale Macon at 2.5ms-1 and a quarter scale Macon at 0.75 ms-1.
Figure 4 - Drag coefficient against angle of attack comparing CFD simulations of full size Macon at 5ms-1, a quarter scale Macon at 2.5ms-1 and a quarter scale Macon at 0.75 ms-1.
However, comparing the drag coefficients of the dynamically matched quarter and full size blades, at free stream velocities of 2.5 ms-1 and 5 ms-1 respectively (Figure 4), it can be seen that at all angles of attack the values for drag coefficient correlate well. Therefore, it is reasonable to suggest that there is a scaling disparity between the quarter and full scale results for drag coefficient when dynamic similarity is not matched. This discrepancy is due to the assumption that drag coefficient is independent of velocity which is based upon the premise that an oar blade has similar fluid dynamic behaviour to a flat plate. By making this assumption the curvature of the blade is neglected, which may play an important role in the way drag is produced. Curvature, along with vorticity
360 The Engineering of Sport 7 - Vol. 1 and pressure gradients are important factors that govern the formation of turbulence and can be significant in oar blades by modifying aspects of the flow in a nontrivial fashion through the way it influences turbulence (Mathieu & Scott, 2000). Further, by increasing velocity to 2.5 ms-1 the Reynolds number will also increase, this will result in a change in the nature of the flow so that it becomes increasingly turbulent. When Reynolds number is large the viscous forces, and hence the energy dissipation, are small and unsteady vortices will appear and interact with each other (Davidson, 2004). It is due to these reducing viscous forces that the overall drag on the surface of the blade is also seen to decrease with increasing velocity. Therefore, it can be argued that the magnitude and exact nature of the flow around an oar blade is dependent both on the shape of the blade, as well as the free-stream velocity of the flow.
5- Conclusions The use of computational fluid dynamics has been validated against published experimental data for a quasi- static simulation of a Macon oar blade at a range of angles of attack to the free stream. The main flow measures were the lift and drag coefficients both of which were found to be sensitive to the choice of turbulence model. The k- SST model with second order differencing was shown to be the most accurate turbulence model. When compared with CFD solutions at full size for a free stream velocity of 5 ms-1, the quarter scale solutions at 0.75 ms-1 overestimate drag coefficients but estimate lift coefficients much more accurately. However, when dynamic similarity was matched between quarter and full scale, requiring a relative blade velocity of 2.5 ms-1 at quarter scale, both lift and drag coefficients were unchanged. This leads to the conclusion that the assumption of Caplan and Gardner concerning Reynolds number independence above 0.75 ms-1 was incorrect. The magnitude of the drag coefficient, rather than the manner in which it changed with angle of attack, was found to be a function of the shape of the blade.
6- References [A1] Ansys Inc. (2005). Fluent (Version 6.3) [Computer software]. [BK1] Barré, S. & Kobus, J. (1998). New facilities for measurement and modelling of hydrodynamic loads on oar blades. In Engineering of Sport (pp. 251-260). Cambridge: Blackwell Science. [C1] Concept II. www.concept2.co.uk. 2008. Accessed: 29-01-08. [CG1] Caplan, N. & Gardner, T. (2007). A fluid dynamic investigation of the Big Blade and Macon oar blade designs in rowing propulsion. Journal of Sports Sciences, 25, 643-650. [D1] Davidson, P. A. (2004). Turbulence: an introduction for scientists and engineers. Oxford: Oxford University Press. [FP1] Ferziger, J. & Periç, M. (1997). Computational methods for fluid dynamics. Verlag, Berlin: Springer. [HB1] Hartwanger, D. & Bunt, E. A. (1996). Modelling of scull drag at the orthogonal point of a rowing stroke. MSc University of Witwatersrand.
The Engineering of Sport 7 - Vol. 1
361
[K1] Kleshnev. V. (1998). Estimation of biomechanical parameters and propulsive efficiency of rowing. Australian Institute of Sport. [LS1] Launder, B. E. & Sharma, A. (1974). Application of the Energy-Dissipation Model of Turbulence to the Calculation of Flow Near a Spinning Disk. Letters in Heat and Mass Transfer, 1, 131-138. [M1] Mentor, F. (1993). Zonal two equation k-? turbulence models in aerodynamic flows. AIAA Paper. [MS1] Mathieu, J. & Scott, J. (2000). An introduction to turbulent flow. (1st ed.) Cambridge, UK: Cambridge University Press. [MY1] Munson, B. R., Young, D. F., & Okiishi, T. H. (2002). Fundamentals of fluid mechanics. (4th ed.) New York: John Wiley and Sons, Inc. [P1] Patankar, S. V. (1980). Numerical heat transfer and fluid flow. New York: Hemisphere Publishing Corporation. [RK1] Rhee, S.H., Koutsavdis, E. (2005). Two-dimensional simulation of unsteady marine propulsor blade flow using dynamic meshing techniques. Computers and Fluids, 34, 1152-1172.
The Acute Response to a Garment-based Elastic Thoracic Load, Applied During Exercise on Inspiratory Muscle Strength and Pulmonary Function (P72) Ashley R. Gray1, Dr Tom M. Waller2, Prof Mike P. Caine2
Topics: Apparel and respiratory muscle training. Abstract: Respiratory muscle training (RMT) is becoming increasingly accepted as an ergogenic and therapeutic aid to performance athletes and the general population, respectively. However, the acute benefits of RMT are less frequently reported. A new proprietary technology has made it possible to load the thorax through a garment (RespiVest, Canterbury of New Zealand Ltd & Progressive Sports Technologies Ltd) that can be worn during exercise, providing an IMT stimulus. Methods: Twelve recreationally active males (24 ± 3 yrs, 81.3 ± 7.1 kg, 1.80 ± 0.1 m; mean ± SD) took part in a randomised cross over study. Wearing RespiVest (RV-IMT) or control garment, participants exercised at 100 W for 10 min on a rowing ergometer (Concept II, USA). For the final 5 min, 6 x 5 sec high-intensity sprints were performed separated by 1 min 100 W recovery bouts. Pre and post exercise, an indirect measure of inspiratory muscle strength; maximal inspiratory pressure (MIP) at residual volume, was obtained using an MIP device (Chest Scientific, UK). Lung function (flowvolume loops) values were obtained using static spirometry (Super spiro, Micro Medical, UK). Heart rate was monitored constantly throughout the exercise protocol. Results: A RVIMT warm-up significantly increased maximal inspiratory pressure (8.1 ± 2.0%; mean ± SE; P < 0.01) post-exercise. After using control, maximal inspiratory pressure was reduced (-2.0 ± 2.6%). No other differences in lung function parameters pre and post exercise were seen between RV-IMT and control. Conclusion: These data suggest that inspiratory muscle strength can be enhanced using RV-IMT during short duration, low intensity whole body exercise. In addition this was more effective than using a control garment. RespiVest may therefore be useful in facilitating optimal warm-up preparations by providing a respiratory warm-up effect similar to the one present in locomotive muscles. Keywords: Inspiratory muscle training, maximal inspiratory pressure, warm-up, RespiVest, rowing.
1. Progressive Sports Technologies Ltd, Sports Technology Institute, Loughborough University Science & Enterprise Park, 1 Oakwood Drive, Leicestershire, LE11 3QF, UK - E-mail: [email protected] 2. Sports Technology Research Group, Sports Technology Institute, Loughborough University Science & Enterprise Park, 1 Oakwood Drive, Leicestershire, LE11 3QF, UK - E-mail: t.m.waller, [email protected]
364 The Engineering of Sport 7 - Vol. 1
1 - Introduction Increasing levels of evidence are now becoming available supporting the chronic use, during a training intervention, of the ergogenic benefits of respiratory muscle training (RMT) (Romer et al. 2002, Stuessi et al. 2001, Volianitis et al. 2001, Griffiths and McConnell 2007, Johnson et al. 2007). RMT can now be practised with respective specific devices, in three distinct modes; inspiratory muscle training (IMT), voluntary isocapnic hyperpnoea (VIH) and flow resistive loading. Literature investigating an acute response to RMT is somewhat more limited. Initial work by Volianitis and colleagues in 1999 demonstrated a significant enhancement in MIP (8.5 ± 1.8%) after participants executed 60 breaths on an IMT device (POWERbreathe®, IMT Technologies Ltd) when compared to a general cycling, and rowing warm-up. A later study by Volianitis et al. (2001) incorporated a respiratory muscle warm-up with a rowing warm-up. Two bouts of 30 breaths at 40% of maximum inspiratory pressure (MIP) provided a significant enhancement in MIP by 7.0 ± 1.0% from baseline values. After a 6 min all out rowing effort, an IMT combined rowing warm-up was found to be significantly beneficial in attenuating inspiratory muscle fatigue. This was evidenced in a reduced drop off in MIP values when compared to the same rowing warm-up without IMT, and a sub-maximal rowing warm-up. Indeed of greater interest was that this translated into performance enhancement; with power output during the 6 min test was found to be significantly elevated by 1.2% when compared to the rowing warm-up without IMT. In a study by Furain et al. (1998) attention was focused on an alternative method of IMT. Rather than targeting the buccal cavity, Furian and colleagues used a latex band to load the thorax. Participants completed maximal inspirations for 30 min, 5 times a week over a 5 week period. The results demonstrated that similar responses to traditional RMT could be achieved using this technique, with significant improvements seen in maximal voluntary ventilation (MVV), respiratory muscle endurance (RME), work time at anaerobic threshold and attenuate lactate response. A new propriety technology, called the RespiVest, has now made it possible to load the thorax through a garment based elastic load that can be worn during exercise, and thus provide an inspiratory muscle training stimulus ‘on-the-go.’ The aim of the present study was to determine the influence of a whole-body RespiVest inspiratory muscle training (RV-IMT) warm-up intervention on inspiratory muscle strength and pulmonary function.
2- Methods 2.1 Experimental design Twelve recreationally active males (24 ± 3 yrs, 81.3 ± 7.1 kg, 1.80 ± 0.1 m) provided written informed consent to take part in a randomised cross over study. Procedures were administered in accordance with ethical advisory committee guidelines. Prior data collection participants were familiarised with mouth pressure measurements and flow
The Engineering of Sport 7 - Vol. 1
365
volume manoeuvres. Participants were requested to refrain from alcohol and vigorous exercise 24 hr before entering the laboratory. Maximum mouth pressures and pulmonary function were assessed before and after a rowing warm-up protocol.
Figure 1 - IMT garment (RespiVest, Canterbury of New Zealand & Progressive Sports Technologies Ltd, UK)
2.2 Maximum inspiratory pressure (MIP) MIP is commonly used to measure inspiratory muscle strength. It reflects the forcegenerating capacity of the combined inspiratory muscles during a brief, quasi-static contraction (Meuller manoeuvre). A hand-held mouth pressure meter (Chest Scientific, UK) was used to measure a minimum of five and a maximum of nine technically satisfactory measurements. This involved the initiation of inspirations from residual volume for 2-3 sec while verbal encouragement was provided. Trials which did not represent the participant’s maximum effort, according to their subjective feeling, were discarded. Repeat measurements were taken until the highest of three measurements with 5% variability, or within 5 cmH2O difference, was defined as maximum (Whitelaw and Feroah, 1989).
2.3 Pulmonary Function Pulmonary function was assessed with a Super Spiro spirometer (Micro Medical, UK). Following familiarisation, each test was repeated three times and the highest recorded value was used for subsequent analysis (Cotes 1993). Forced vital capacity (FVC), forced expiratory volume in one second (FEV1), percentage expired (FEV1/FVC), peak expiratory flow (PEF) and peak inspiratory flow (PIF). Maximum voluntary ventilation (MVV) was measured while seated with approximately 5 min separating two repeated measures.
2.4 Rowing warm-up Wearing either a respiratory muscle training garment (RV-IMT; RespiVest, Canterbury of New Zeland & Progressive Sports Technologies Ltd, UK), or normal upper body garment (control) participants performed 10 min of rowing on an ergometer (Concept
366 The Engineering of Sport 7 - Vol. 1 II, USA) at approximately 100 W. In the final 5 min, 6 x 5 s all-out sprints were performed, separated by 1 min of 100 W low intensity rowing, as used by Mandic et al. (2004). Heart rate was assessed using short-range telemetry (Polar Sport, Finland).
2.5 Statistical Analyses Student’s t-test for paired samples was used to compare differences between lung function parameters before and after the rowing warm-up, and between the IMT and control trial. Values of P < 0.05 were considered statistically significant. Data points were means (± SE) unless otherwise stated.
3- Results 3.1 Maximum inspiratory pressure and pulmonary function Baseline pulmonary function and MIP were all within normal limits and were not different between RV-IMT and control. Post rowing warm-up, RV-IMT induced a significant increase (8.1 ± 2.0%) in MIP from a mean baseline of 155.0 ± 6.1 cmH2O to 167.7 ± 7.4 cmH2O (P < 0.01). MIP was found to be slightly reduced (-2.0 ± 2.6%) in the control with mean MIP at baseline measured 158.25 ± 4.8 cmH2O and 154.5 ± 5.1 cmH2O after the rowing warm-up, this was not significant (P < 0.05). The difference between treatments therefore equated to 10.2 ± 3.1%, (P < 0.01). Additionally, it is of interest to note, elevated, but non-significant, MVV values found after post rowing warm-up for RVIMT. No significant differences were found in other lung function parameters from baseline after each treatment or between treatments, as summarised in Table 1.
Figure 2 - Maximum inspiratory pressure (MIP) in RespiVest inspiratory muscle training (RVIMT) (filled bars) and control (open bars) pre and post rowing warm-up (mean ± SD). **Significantly different from Pre-IMT (P < 0.01)
The Engineering of Sport 7 - Vol. 1
367
Figure 3 - Maximum voluntary ventilation (MVV) in RespiVest inspiratory muscle training (RV-IMT) (filled bars) and control (open bars) pre and post rowing warm-up (mean ± SE). Table 1 - Mean ± SE and percentage changes between baseline and after rowing warm-up, and percentage difference between trials, *(P < 0.05), **(P < 0.01)
3.2 Heart rate response Mean heart rate was found to be elevated by approximately 23 bpm after participants increased rowing workload from 100 W to a sprint. Indeed after each of the subsequent 6 sprints, heart rate progressively increased. It is of interest to note, a small tendency for heart rate to be slightly elevated by 1-3 bpm, after the third sprint at the 7 min timepoint, however this difference was not significant.
3.3 - Test-retest reproducibility of MIP The two baseline measurements of MIP before the warm-up protocol permitted a test-retest assessment of MIP. The mean baseline MIP values between the IMT and control were not significantly different, the mean difference being less than 5 cmH2O.
368 The Engineering of Sport 7 - Vol. 1 The mean coefficient of variation (CV = 100% x SD/mean) for the baseline MIP measured on two occasions was 4.2%. During the warm-up average power, stroke rate and participant’s heart rate were not found to be significantly different between trials.
4- Discussion The main finding of the present study was that MIP increased significantly (8.1 ± 2.0%; P < 0.01) following the rowing warm-up using RV-IMT, but not during a warm-up without RV-IMT (-2.1 ± 2.6%). This finding is in corroboration with Volianitis et al. (1999) where a significant increase (8.5 ± 1.8%; P < 0.0001) in MIP was found after a 2 x 30 breath respiratory warm-up protocol using the POWERbreathe® (IMT Technologies Ltd., UK) device at 40% MIP load. It is important to note that the present study involves IMT during exercise, in contrast to Volianitis et al. (1999) where IMT was performed with no exercise component. In addition the present study’s design did not measure MIP load initiated by the RespiVest, although it is likely to be significantly less than 40% MIP used by Volianitis et al. (1999). When RV-IMT was compared to control, MIP was enhanced by 10.2 ± 3.1%. If such MIP gains are functionally relevant to a specific sport, the RV-IMT could provide an effective passive IMT warm-up, while other warm-up regimens (such as stretching, light jog, skills) are followed. In performance sport, where time is so often highly restricted, this factor alone could significantly enhance adherence, and eliminate the need to dedicate a specific time period for the sole use of a traditional IMT device such as POWERbreathe®. Furthermore, heart rate data during this short warm-up showed no significant alterations in response, providing an indication that RespiVest does not induce excessive cardiovascular stress, which could potentially negate the ability to perform optimal warm-up strategies. The precise mechanism(s) responsible for the increase in MIP following the respiratory warm-up cannot be easily identified. A skeletal muscle warm-up has been reported to have an effect on maximum isometric force when the change in the muscle temperature is substantial (Ranatunga et al. 1987). Under current test conditions it was not possible to measure respiratory muscle temperature, however it is important to consider that such muscles are likely to be highly thermoregulated due to their proximity to the body core. Therefore as suggested by Volianitis et al. (1999), an altered motor control hypothesis should be considered more closely. Indeed it is possible that the intermuscular coordination between inspiratory and expiratory muscles is improved in a manner similar to the one identified by other skeletal muscles (Komi 1992). Repeated performance of the specific recruitment pattern might decrease the degree of co-contraction known to exist between inspiratory and expiratory muscles at residual volume and consequently improve force generation. With reference to Furian and colleagues’ initial work with the latex band, it is of interest to note MVV data collected in the present study. Indeed, no significant enhancement in MVV was found, in contrast to Furian et al.’s 5 week training study. However, a visual inspection of results demonstrated that the RV-IMT may be having some effect in attenuating an MVV drop-off (Figure 3).
The Engineering of Sport 7 - Vol. 1
369
To the best of the author’s knowledge this is only the third study to assess the acute effects of IMT after Volianitis et al. (1999) and Volianitis et al. (2001), and indeed the first where IMT was performed during exercise with a garment. Therefore at present no literature has explored the length of time such ergogenic effects remain after the initial IMT bout. In addition, optimal prescription of use indicating resistance and time used is as yet unknown. The current literature base has shown that in well controlled and designed studies that respiratory muscle training interventions have an ergogenic effect on performance (McConnell and Romer, 2003). Hitherto, in such an early stage of research, neither the present study design or Volianitis et al. (1999) have attempted to measure the short term effects of IMT on subsequent performance. In corroboration with Volianitis et al. (1999), the present study found that a general warm-up without IMT did not significantly affect MIP. A possible explanation may be that the modest ventilatory response induced by the low intensity warm-up protocol was not sufficient for the respiratory muscles to be ‘warmed-up’. Consequently in the sporting context it may be that the traditional low intensity warm-up is not sufficient to optimally prepare athletes prior to performance. An improved functional capacity of the inspiratory muscles induced by an IMT warm-up, may allow a decrease in fibre recruitment requirements during performance and as a result reduce the sensation of breathlessness (dyspnea). Indeed a strong relationship between the recruitment of inspiratory muscles and the perception of dyspnea has been suggested (Killian and Jones, 1982). The avoidance of dyspnea is now believed to be crucial in delaying fatigue in sub-maximal exercise through blood flow redistribution to exercising limbs (McConnell and Romer 2003). The importance of attenuating dyspnea was further highlighted during the performance trial by Volianitis et al. (2001). Post an IMT warm-up, during a 6 min all-out effort, the perception of dyspnea was significantly reduced when compared to a submaximal warm-up without IMT. Indeed in this same trial the IMT warm-up led to a significant 1.2% enhancement in power output, with distance completed elevated by 18 ± 13 m. Pearson’s correlation coefficient revealed that the changes in dyspnea accounted for 22.5% of this variance in power output. In the present study, anecdotally there were indications that the perception of dyspnea in some individuals had been affected, with comments made by participants, such as “my lungs feel more open” and “I feel more confident breathing after a RespiVest warm-up”. Unfortunately the present study failed to log this quantitatively using a visual analogue scale, and therefore should be considered in future work. With a closer inspection of individual results interest, two of the highest responding participants to the RVIMT warm-up elicited enhanced MIP values to the order of 15% and 21%. Both of whom interestingly had been previously diagnosed with asthma. In elite sport the process of diagnosing borderline asthma or exercise induced asthma to enable a performer to be legally prescribed asthma medication under WADA regulations is sometimes problematic. Therefore future studies should consider the efficacy for the use of RV-IMT as a non-pharmacological relief from bronchi spasm.
370 The Engineering of Sport 7 - Vol. 1
5- Conclusions Performing a rowing warm-up when wearing RespiVest enhanced inspiratory muscle strength more effectively than without, thus supporting the notion that a warm-up phenomenon, similar to the one present in locomotive musculature exists in inspiratory muscles. Indeed, such results provide sensible rationale for the use of RespiVest to facilitate whole body optimal warm-up preparations.
6- References [C1] J.E. Cotes. Lung function: assessment and application in medicine, 5th edn. Blackwell Scientific Publications, London, 496-497, 1991 [FH1] Furian T.C., Hirschberg F. and Ritthaler F. Improvement of the aerobic endurance after training the muscles of respiration with a latex band around the chest. International Journal of Sports Medicine. 19:S15, 1998 [GM1] Griffiths L.A. and McConnell A.K. The influence of inspiratory and expiratory muscle training upon rowing performance. European Journal of Applied Physiology. 99:457-466, 2007 [KJ1] Killian K.J. and Jones N.L. Respiratory muscles and dyspnea. Clinical Chest Medicine. 9:237-247, 1988 [K1] P.V. Komi. Strength and power in sport. Oxford: International Olympic Committee, Blackwell Publications, 1992 [MQ1] Mandic S. Quinney H.A. and Bell G.J. Modification of the Wingate anaerobic power test for rowing: Optimization of the resistance setting. International Journal of Sports Medicine. 25:409-414, 2004 [MR1] McConnell A.K. and Romer L.M. Respiratory muscle training in healthy humans: Resolving the controversy. International Journal of Sports Medicine, 2003 [RS1] Ranatunga K.W., Sharpe B. and Turnbull B. Contractions of a human skeletal muscle in different temperatures. Journal of Physiology. 390:383-395, 1987 [RM1] Romer L.M., McConnell A.K. and Jones D.A. Effects of inspiratory muscle training upon time trial performance in trained cyclists. Journal of Sports Science. 20:547-562, 2002 [SS1] Stuessi C., Spengler C.M., Knopfli-Lenzin C., Markov G. and Boutellier U. Respiratory muscle endurance training in humans increases cycling endurance without affecting blood gas concentrations. European Journal of Applied Physiology. 84:582-586, 2001 [VM1] Volianitis S., McConnell A.K., Koutedakis Y. and Jones D.A. The influence of prior activity upon inspiratory muscle strength in rowers and non-rowers. International Journal of Sports Medicine. 20:542-547, 1999 [VM2] Volianitis S., McConnell A.K., Koutedakis Y., McNaughton L., Backx K. and Jones D.A. Inspiratory muscle training improves rowing performance. Medicine and Science and Sport and Exercise Science. 33:803-809, 2001 [WF1] Whitelaw W.A. and Feroah T. Patterns of intercostals muscle activity in humans. Journal of Applied Physiology. 67:2087-2094, 1989
Aerodynamic Performance of Cycling Time Trial Helmets (P76) Kim B. Blair, Ph.D.1, Stephanie Sidelko2
Topics: bicycle. Abstract: The aerodynamic performance of equipment is critical for cycling’s “race of truth”, the time trial, where time differences of seconds can separate the top finishers. Recently several manufacturers have developed aerodynamic helmets specifically for use during time trials. To date, a comprehensive study of the aerodynamic performance of these time-trial helmets has not been performed. In this study, the aerodynamic helmets were tested on a mannequin in the time trial positions over a range of yaw angles, from 0° to 15°, in increments of 5°. The helmets were tested at three head angle positions at each yaw angle in order to best mimic actual riding conditions. A control road helmet was used to serve as a comparative tool. The testing results showed that all of the aerodynamic helmets offer drag reduction over a standard road helmet. However, the ranking of helmets in order of performance varied depending on yaw angle and head angle. Keywords: bicycling, helmets, aerodynamics.
1- Introduction The disciplines of the time trial in cycling and the cycling leg of a triathlon (for nondrafting events) have the common goal of getting from the start of the event to the finish in the shortest time possible. In this event, nearly 90% of a cyclist’s power output is used to overcome the resistance of aerodynamics (Blair, 2007). Even small reductions in the aerodynamic drag on the cyclist can result in the savings of seconds during the event. Consequently, a large market of commercial products exists to aid the cyclist in reducing aerodynamic drag, including aerodynamic bike frames, wheels, handlebars, and now helmets. Lukes, Chin and Haake (2005) have provided an overview of the pertinent literature in this domain. While aerodynamic head fairings have been used for many years, the Union Cycliste Internationale recently enacted a requirement that all professional cyclists must wear protective helmets. Manufacturers were quick to respond to this requirement, resulting 1. Xenith, LLC, 672 Suffolk Street, Lowell, MA, USA - E-mail: [email protected] 2. Massachusetts Institute of Technology, Sports Innovation @ MIT, Cambridge, MA, USA E-mail: [email protected]
372 The Engineering of Sport 7 - Vol. 1 in a wide variety of protective aerodynamic helmets becoming available for recreational cyclists and triathletes. Blair (2007) recently demonstrated that the reduction in drag afforded by an aerodynamic helmet is significant, with a larger impact than aerodynamic wheel sets. Thus, the aerodynamic helmet is a very effective tool for improving a cyclist’s or triathlete’s performance. In this work we have compared the performance of several aerodynamic cycling helmets. Testing was carried out in a wind tunnel at varied yaw angles using a mannequin in an aerodynamic position. The wind speed was held constant and the yaw angle of the mannequin, as well as the angle of the helmet relative to the mannequin’s back, was varied for each helmet tested.
2- Methods Manufacturers donated ten aerodynamic cycling helmets for this project under the conditions of anonymity. Helmets that were supplied with face shields were tested both with and without the face shield. It was agreed that the results could be published, but that the specific identity of the helmets not be identified to protect proprietary data. A standard road cycling helmet was also tested and used for a comparison. As the cyclist/triathlete is required to wear some form of helmet during competition, a comparison with a typical road helmet provides an indication of the improvement possible for competition. Testing was completed in the Massachusetts Institute of Technology Wright Brothers Memorial Wind Tunnel. The wind tunnel is a closed return tunnel with a 2.1 m X 3 m oval test section. The tunnel data acquisition system records tunnel parameters of static and dynamic pressure, temperature and humidity at 1000 Hz. These values are used to calculate a humidity corrected tunnel wind speed (MIT, 2002). The wind tunnel’s pyramidal balance was used to collect the aerodynamic force data. This balance remains in a fixed position and thus measures the aerodynamic forces in the wind tunnel axes. The drag force, DT is measured along the axis of the tunnel, with a positive direction to the rear of the tunnel, and the side force ST is measured perpendicular to the flow positive direction to the right of the tunnel when viewed from the top. However, the aerodynamic forces of critical importance to the cyclist are the force acting against the rider’s direction of motion DR, and the force perpendicular to that force, SR. The forces measured by the wind tunnel balance were converted to the rider’s drag and side force values by the equations DR = DT cos – ST sin
SR = DT cos – ST sin
(1)
where T and R represent the tunnel axes and rider axes respectively and is the yaw angle of the rider, measured in a clockwise direction when viewed from the top of the rider. A 0° yaw angle means that the tunnel and rider axes are aligned, and the rider is facing directly into the wind. For the remainder of this paper, the terms drag force and side force will refer to the forces in the rider coordinate system.
The Engineering of Sport 7 - Vol. 1
373
Wind tunnel air speed was set to 13.4 m/s for all testing. This wind-speed has become somewhat of a de-facto industry standard. Yaw angles were varied from 0° to 15°, in increments of 5°. In order to eliminate the test-to-test error inherent in using a cyclist, an upper body mannequin was used for the test program. The mannequin is in the time-trial cycling position as shown in Figure 1.
Figure 1 - Upper body mannequin in the wind tunnel. The lower half of the skin-suit was taped up behind the mannequin durng testing.
The effect of helmet position relative to the rider’s back was also investigated in this study. As the mannequin did not have an articulated neck, the helmet was tipped in three different angles on the mannequin’s head. As the helmets had a variety of tail shapes and lengths, it was determined that the most consistent method to set the helmet angle was to use lines on the mannequin’s forehead to align the front edge of the helmet as shown in Figure 2. Position 1 results in the tail of the helmet being very close to the back, position 3 has the tail pointing nearly straight up in the air, as if the rider were looking straight down at the ground.
Figure 2 - Reference positions for helmet alignment.
A data set was recorded for each helmet at three different helmet positions at four different yaw angles yielding twelve data sets for each helmet. The helmets that have visors were tested both with and without the visor. Data was collected for 30 seconds at
374 The Engineering of Sport 7 - Vol. 1 1000 Hz for each test point. For each test point recorded, the drag value of the mannequin alone is subtracted from the result for the mannequin wearing the helmet. Thus, if the drag of the mannequin wearing the helmet is less than the mannequin alone, the result with be negative, indicating that the combination of the mannequin and helmet is “faster” (i.e. less drag) than the mannequin alone. The potential performance benefit to a cyclist resulting from a reduction in aerodynamic drag afforded by an aero-helmet can be estimated in terms of the power savings available to the rider. While the benefit to a specific rider cannot be predicted without detailed knowledge for that rider, an estimate of the power savings for a professional and amateur rider can be estimated for illustrative purposes. For the professional cyclist, we assume a total drag of 22 N while wearing a road-race helmet, with the rider capable of sustaining 450 watts of power output over a long (40 km) time-trial. For an amateur cyclist, corresponding values of 27 N of drag and 225 W of power output can be assumed (Martin et al. 1998). Determining the percent reduction in drag and multiplying it by the rider’s power output can calculate the power savings, Ps, (2) where DRH and DAH represent the drag with a road helmet and aero-helmet and PAve is the rider’s power output. Here the assumption is made that the rider’s speed remains the same, so there are no other differential energy changes in the system.
3- Results Figure 3(a) shows the results for each helmet in the three positions for the case of 0° yaw. Recall that the drag value of the mannequin alone is subtracted from the result for the mannequin wearing the helmet. Thus in Figure 3, the case of the mannequin without a helmet results in a drag value of 0. The goal is to reduce drag, so lower values indicate better performance. Helmet A is the control helmet, a typical road-racing helmet. For 0° yaw with the helmet in position 1, with the tail laying nearly flat on the back, ten of the aero-helmets result in a lower drag than the case of the mannequin not wearing a helmet (drag < 0). For position 2, only eight of the aero-helmets show drag that is less than the mannequin without a helmet. For both positions 1 and 2, all aero-helmets show a lower drag than the control helmet. For position 3, the tail of the helmet pointing nearly straight up, none of the aero-helmets show a drag value less than the no helmet condition. Further, only nine of the aero-helmets result in a lower drag than the control helmet. For each test condition, the helmets rank in a different order for drag, with none of the helmets showing a consistently high performance across all head positions. Figure 3(b) shows the results for each helmet in the three positions for the case of 5° yaw. Note that helmet A, the standard road helmet, shows substatially the same drag at all yaw angles. For the helmet in position 1, seven of the aero-helmets result in a lower drag than the case of the mannequin not wearing a helmet. None of the aero-helmets show less drag than the mannequin alone for positions 2 and 3. The control helmet shows the highest drag for positions 1 and 2. For position 3, only eight of the aero-
The Engineering of Sport 7 - Vol. 1
375
helmets show a lower drag than the control helmet. For each position the helmets rank in a different order of performance, with helmet E showing good performance across all positions, with a rank of fourth, first and first. Figure 3(c) shows the results for each helmet in the three positions for the case of 10° yaw. For this yaw angle, none of helmets show a lower drag than the mannequin without a helmet. For positions 1 and 2, all aero-helmets show a lower drag than the control helmet, with only eight aero-helmets performing better than the control for position 3. For this yaw condition, helmet L results in the lowest drag for each head position. Figure 3(d) shows the results for each helmet in the three positions for the case of 15° yaw. As with the 10° yaw condition, none of helmets show a lower drag than the mannequin without a helmet. For positions 1 and 2, all aero-helmets show a lower drag than the control helmet, with only six aero-helmets performing better than the control for position 3. For this yaw condition, helmet L results in the lowest drag for head positions 1 and 3, and ranks fourth for position 2.
Figure 3 - Drag results for a) 0° yaw angle, b) 5° yaw angle, c) 10° yaw angle, and d) 15° yaw angle. Helmet position indicated as Position 1 Position 2 Position 3.
The benefit of the reduction in drag by the use of an aero-helmet can be further illustrated by estimating the rider power savings resulting from the lower drag. Table 1 shows this power savings estimate calculated for the helmet that performed in the median for each test condition and for the case of both a hypothetical professional and amateur cyclist.
376 The Engineering of Sport 7 - Vol. 1 Table 1 - Power savings resulting from median aerodynamic helmet.
4- Discussion The results clearly show that there is a significant reduction in aerodynamic drag resulting from the use of an aero-helmet as compared to a road-helmet. For some helmets, at low yaw angles and with the helmet in Positions 1 or 2, using an aero-helmet will result in a lower drag than the no helmet condition. The results show that there is no clear choice among aerodynamic helmets for all riding conditions represented by the range of yaw angles investigated herein. However, helmet L performed generally better at high yaw angles across all helmet positions. Due to the differing performance of the helmets for each of the test conditions, a cyclist may want to select the helmet appropriate to their racing style and conditions. Factors to consider include rider speed, prevailing wind on typical racecourses, and a preferred head angle. Higher effective yaw angles can be expected from either a slower rider, or strong cross winds. If either of these conditions applies, the cyclist should consider helmets that perform well at high yaw angles. Further, if the rider cannot ride comfortably with the head looking up to maintain the helmet in Position 1, the cyclist should select a helmet that performs well in Position 2. The results clearly show that the cyclist should avoid looking down at the ground while racing, as the drag in Position 3 is higher across all helmets at all yaw angles. Estimating the power savings resulting from the use of an aero-helmet clearly illustrates the benefit of these helmets, even for a median performing helmet. The potential to gain between 10 and 30 watts of power savings will allow a cyclist to significantly increase their racing speed while maintaining their optimal power output.
5- Conclusions The benefit of aerodynamic helmets has clearly been demonstrated, although no helmet showed a clear performance advantage across all test conditions. However, the study is somewhat limited, in that a mannequin with a single body shape and riding position was used in this study. Unpublished proprietary results of working with cyclists in the wind tunnel has shown that there can be a significant interaction between the rider’s body shape, the arch of the back, and riding position and the helmet that results in the lowest aerodynamic drag. To truly optimize the helmet selection, the cyclist will need to try each helmet of interest during a wind tunnel test.
The Engineering of Sport 7 - Vol. 1
377
The results of this study show that there is a potential to optimize the shape of aerodynamic helmets to work across a wider variety of conditions. In order to do so, a more detailed study of the flow characteristics of the flow field around the helmet will be required.
6- References [B1] Blair K. B. Cycling Aerodynamics. Presented at Serotta International Cycling Institute, SICI Cycling Science Symposium and Expo. Hotel Boulderado, Boulder, CO. 23 January, 2007. [LC1] Lukes R. A., Chin S. B., and Haake, S. J. The understanding and development of cycling aerodynamics. In Sports Engineering, 8(2): 59-74, 2005. [MM1] Martin, J. C., Milliken D. L., Cobb, J. E., McFadden K. L., and Coggan A. R., Validation of a Mathematical Model for Road Cycling Power. In Journal of Applied Biomechanics, 14: 276-291, 1998. [MIT1] MIT, Massachusetts Institute of Technology - Wright Brothers Wind Tunnel Information for Use by Industry,” http://web.mit.edu/aeroastro/labs/wbwt/index.html , 2002.
Physical Motion Analysis of Nordic Walking (P77) Takayuki Koizumi1, Nobutaka Tsujiuchi2, Masaki Takeda3, Yusuke Murodate4
Topics: Biomechanics; Fitness. Abstract: Recent years have seen a worldwide increase in people participating in Nordic Walking with a heavy concentration in Northern Europe. This trend has led to abundant research in Nordic Walking and to reports that this type of exercise is effective in reducing load on the lower limbs. At the same time, there has been no comprehensive experimental study to our knowledge on what change in muscular activity brings about this load-reduction effect. To clarify the exercise structure of Nordic Walking and particularly the mechanism behind this load-reduction effect on lower limb joints, this study simultaneously measured the load on lower limb joints during Nordic Walking on a level surface and the amount of muscular activity at 16 locations on the upper and lower limbs and compared the results with ordinary walking. Results revealed a decrease of 8% or greater in the downward force perpendicular to the floor at each lower limb joint suggesting a reduction in load on lower limb joints. They also showed that differences in lower-limb muscular activity were great at the knee extensor muscles of gastrocnemius and quadriceps femoris. In the upper limbs, all measured locations exhibited an increase in muscular activity with difference in activity of flexor carpi radialis being particular large compared to that of other muscles. These results indicate that bodily movement in Nordic Walking has the potential of reducing load on lower limb joints. Keywords: Nordic Walking; Physical motion analysis; Muscular activity; Load of joint.
1- Introduction Nordic Walking is a form of exercise born in Finland in which a walker uses a pole in each hand much like cross country skiing. In recent years, the number of people participating in Nordic Walking has been increasing in Northern Europe and other regions around the world, and research on Nordic Walking has been quite active as a result. In 1. 1-3 Tatara Miyakodani, Kyototanabe-City, Kyoto, JAPAN - E-mail: [email protected] 2. 1-3 Tatara Miyakodani, Kyototanabe-City, Kyoto, JAPAN - E-mail: [email protected] 3. 1-3 Tatara Miyakodani, Kyototanabe-City, Kyoto, JAPAN - E-mail: [email protected] 4. 1-3 Tatara Miyakodani, Kyototanabe-City, Kyoto, JAPAN - E-mail: [email protected]
380 The Engineering of Sport 7 - Vol. 1 Nordic Walking, the use of poles results in no subjective difference in exercise intensity despite a rise in oxygen consumption, heart rate, and calorie consumption compared to ordinary walking [RV1]. It has been pointed out that pushing poles into the ground as in Nordic Walking may reduce the load on lower limb joints compared to ordinary walking [WT1] [BA1]. Nordic Walking can consequently be treated as a new means of promoting health with the capability of providing safer and more effective exercise than ordinary walking. However, though reports have been made on the load-reduction effect of Nordic Walking on lower limb joints, there has been no comprehensive experimental study to the our knowledge on what change in muscular activity brings about this loadreduction effect. While there have been experimental reports on muscular activity [SE1], those reports have focused only on muscular activity leaving out a simultaneous analysis of motion. In addition, muscle sites that have been measured are few resulting in an insufficient amount of data. Obtaining a comprehensive understanding of the loadreduction effect of Nordic Walking on lower limb joints requires more than just examining the load on the lower limb joints of a walker. That load must be examined simultaneously with the functioning of lower-limb muscular activity and the load on poles. In this study, with the aim of clarifying the exercise structure of Nordic Walking and particularly the mechanism behind this load-reduction effect on lower limb joints, we measure the load on lower limb joints and poles during Nordic Walking on a level surface simultaneously with the amount of muscular activity at 16 locations on the upper and lower limbs. We then compare the data so obtained with that of ordinary walking to clarify the mechanism behind the load-reduction effect on lower limb joints during Nordic Walking from the viewpoint of muscular activity.
2- Methods 2.1 Subjects To perform the experiment of this study, we selected nine healthy males (age: 22.9±1.6 [yrs], height: 169.0±5.8 [cm], weight: 59.0±8.4 [kg]) exhibiting no ailments as subjects. Each subject was fully briefed on the purpose and method of the study, and each consented to participating in the experiment. All subjects had no experience in Nordic Walking. They received instruction from a Nordic Walking specialist before the experiment and learned how to comfortably engage in Nordic Walking.
2.2 Experimental conditions Each subject performed Nordic Walking and ordinary walking one time each. Two force platforms were arranged on the floor in the path of Nordic Walking and ordinary walking, and subjects were asked to pass over the force platforms starting about five meters before them and continuing for about 5 meters after. Figure 1 shows the experimental setup for Nordic Walking. No particular conditions were established for the speed of walking and subjects were simply instructed to perform Nordic Walking and ordinary walking at whatever speed was comfortable. For Nordic Walking, subjects were
The Engineering of Sport 7 - Vol. 1
381
instructed to step with their left foot on the left force platform and to push their righthand pole into the right force platform as shown in Fig. 2. For ordinary walking, subjects were only asked to step with their left foot on the left force platform. Before commencing measurements, subjects practiced to adjust their stride accordingly and satisfy these conditions for Nordic Walking and ordinary walking. For each subject, the amount of muscular activity and lower limb load was measured for both Nordic Walking and ordinary walking and the pole’s reaction force with respect to the floor was measured for Nordic Walking.
Figure 1 - Experimental condition of measurements.
Figure 2 - Walking condition during Nordic Walking.
2.3 Measurements of muscular activities The surface muscle potential generated by various muscles (Table 1) was measured during walking to determine the amount of muscular activity in subjects during both Nordic Walking and ordinary walking. This measurement was performed using disposable electrodes (from MEDICOTEST) and an electromyograph (from Mega Electronics). An electromyogram (EMG) was derived from surface electrodes by placing an electrode over the belly of each muscle in line with muscle movement and another electrode nearby as a ground electrode. Since it is said that the frequency component of surface muscle potential is generally in the range of 20-500 [Hz], measurements were performed at a sampling frequency of 1000 [Hz]. To remove noise, the raw data obtained from the experiment was cut at 15 [Hz] and below using a Butterworth filter. Table 1 - Measured Muscles.
382 The Engineering of Sport 7 - Vol. 1
2.4 Assessment of load of lower limb joints A three-dimensional motion analysis system (from Motion Analysis) was used to measure the load applied to a subject’s lower limb during Nordic Walking and ordinary walking. Here, spherical markers conforming to the Helen Hayes Marker Set were attached to a subject’s body at 19 locations to measure the subject’s joint positions while walking. Using ten cameras, a subject’s joint coordinates were measured at a sampling frequency of 60 Hz during Nordic Walking and ordinary walking, and inverse dynamics was used in conjunction with floor reaction force on the force platform to calculate force in the direction perpendicular to the floor and extension/flexion moments of each joint. Calculations were performed as shown below. Dx, Dy, Dz are distances between distal endpoint and segment center of gravity. Px, Py, Pz are distances between proximal endpoint and segment center of gravity. Fdx, Fdy, Fdz are force platforms or calculate forces on distal end of segment. Mdx, Mdy, Mdz are distal moments. x, y, z are angular accelerations. x, y, z are angular velocities. Ix, Iy, Iz are segment’s inertia tensor from Dapena’s (1978) correction factors. Forces on proximal end of segment with equations: (1) (2) (3) Moments about segment center of gravity with equations: (4) (5) (6) Joint moments with equations: (7) (8) (9)
2.5 Statistical analysis Significant difference between the data obtained by this experiment for Nordic Walking and ordinary walking was analyzed using a paired t-test. However, for data not
The Engineering of Sport 7 - Vol. 1
383
exhibiting a normal distribution, analysis was performed using a Wilcoxon signed-rank test.
3- Results and Discussion Table 2 compares the average integrated EMG per cycle time of each muscle site between Nordic Walking and ordinary walking, and Table 3 compares average integrated downward force on the floor acting on lower limb joints and extension/flexion moments per stance phase between Nordic Walking and ordinary walking. Table 2 - Comparison of Average Integrated EMG between Nordic Walking and Ordinary Walking (μV).
Table 3 - Comparison of Average Integrated Forces and Moments per Stance Phase between Nordic Walking and Ordinary Walking.
3.1 Muscular Activity The results in Table 2 reveal that a significant difference (P<0.005) exists in muscular activity in the upper limb between Nordic Walking and ordinary walking and that the amount of muscular activity is higher in Nordic Walking than ordinary walking. In particular, musculus flexor carpi radialis is 46.5 times more active, musculus biceps brachii 7.1 times, musculus deltoideus 5.8 times, musculus triceps brachii 10.4 times, musculus pectoralis major 3.8 times, musculus rectus abdominis 3.2 times, musculus
384 The Engineering of Sport 7 - Vol. 1 latissimus dors 4.4 times, upper range of musculus trapezius 2.2 times, and Musculus erector spinae 1.6 times. The muscle that exhibits the greatest difference in muscular activity among the muscles targeted in this study is musculus flexor carpi radialis. The reason for this difference is thought to be that the subject grips poles in Nordic Walking. We can also say that the difference in muscular activity for musculus triceps brachii is caused by the elbow’s extension force for pushing the poles in the backward direction, and that for musculus biceps brachii is caused by the weight of the poles acting on the arms when the subject swings the poles forward applying a load to elbow bending. As for the lower limb, a remarkable significant difference (P<0.005) was observed for musculus vastus medialis as well as a significant difference (P<0.05) for musculus rectus femoris. These two muscle sites had higher values of muscular activity for Nordic Walking of about 1.4 and 1.5 times, respectively, compared to that of ordinary walking. An insignificant, but a tendency of difference (P<0.1) was also observed for musculus gastrocnemius indicating that the amount of muscular activity increases in Nordic Walking compared to ordinary walking. These muscles are put to work at the time of knee extension, and it is considered that muscular activity increases because of kneejoint extension.
3.2 Load of Joint Referring to the data in Table 3 for lower limb joints, we see that downward force decreased by 8% or more (p<0.05) in Nordic Walking compared to ordinary walking. This is thought to occur because the weight of the walker becomes distributed as the walker pushes the poles into the ground thereby reducing the reaction force from the floor on the legs. Here, to investigate how much load is applied to a pole during Nordic Walking, we calculated the maximum reaction force from the force platform on the floor to the pole in the axial direction. This force was found to be 68.5±16.0 [N] on average for the subjects in this experiment, which shows that the subjects were actually applying force to the poles when engaging in Nordic Walking. From the above, we can say that Nordic Walking has a load reduction effect on lower limb joints compared to ordinary walking. In terms of moments, a significant difference between Nordic Walking and ordinary walking was observed only for knee extension with that moment being about 87% greater in Nordic Walking. This increase is thought to reflect the longer stride that is generally said to be a feature of bodily movement in Nordic Walking. To underscore this point, the muscular activity of musculus quadriceps femoris and musculus gastrocnemius, which generally function at the time of knee extension, was found to be greater in Nordic Walking compared to ordinary walking (Table 2). This is why the knee extension moment is thought to increase in Nordic Walking.
4- Conclusion This study measured the load and muscular activity in lower limb joints during ordinary walking and Nordic Walking to determine the exercise structure of Nordic Walking and, in particular, the mechanism behind load reduction in lower limb joints. The results of this study revealed the following characteristics.
The Engineering of Sport 7 - Vol. 1
385
Compared to ordinary walking, Nordic Walking 1) Lowers the downward force on the floor at each lower limb joint by 8% or more thereby decreasing the load on lower limb joints, and generates a floor reaction force on a pole of 68.5±16.0 [N]; 2) Generates a knee-joint extension moment that is about 87% larger; and 3) Generates greater muscular activity in the lower-limb knee-extension muscle sites of musculus gastrocnemius and musculus quadriceps femoris. Muscular activity at all measured locations in the upper limb also increases during Nordic Walking, with difference in muscular activity being especially noticeable in musculus flexor carpi radialis. Bodily movement in Nordic Walking therefore has the potential of bringing about a load reduction in lower limb joints. The above results show that Nordic Walking actively works both upper-limb and lower-limb muscle groups more so than ordinary walking and distributes body weight to the poles. In short, Nordic Walking has the advantage of reducing load on lower limb joints while being a safe and effective form of exercise.
5- Acknowledgment This work was partially supported by Grant-in-Aid for Scientific Research (c)(2)(19560266), Japan Society for the Promotion of Science.
6- References [RV1] Rodgers, C. D., J. L. Vanheest, and C. L. Schachter. Energy expenditure during submaximal walking with Exerstriders. Med. Sci. Sports Exerc., Vol. 27, No. 4, pp. 607–611, 1995. [WT1] Willson, J., M. R. Torry, M. J. Decker, T. Kernozek, and J. R. Steadman. Effects of walking poles on lower extremity, gait mechanics. Med. Sci. Sports Exerc., Vol. 33, No. 1, 2001, pp. 142–147. [BA1] Bohne, M., and J. Abendroth-Smith. Effects of Hiking Downhill Using Trekking Poles while Carrying External Loads Med. Sci. Sports Exerc., Vol. 39, No. 1, pp. 177–183. 2007. [SE1] A. Sabo, M. Eckelt and M. Reichel. Development of a Novel Nordic-Walking Equipment Due to a New Sporting Technique. The Impact of Technology on Sport II, pp. 163-168, 2007.
Driving Performance Variability Among Elite Golfers (P79) Ian C. Kenny1, Eric S. Wallace2, Steve R. Otto3
Abstract: In recent years studies have been carried out on single subjects, for example by Kinugasa et al., 2004. It has been reported that it is unlikely any two golfers will have an identical swing, and even that an individual golfer is unlikely to produce two identical swings. The present study aimed to address performance variability among elite level golfers to ascertain whether single-subject (SS) analysis is merited for golf studies. Six elite golfers (0.1 ± 2.2 handicap, 22.1 ± 2.3 yrs) performed eight trials each using three randomly assigned drivers specifically constructed with matched physical properties for the current study. Testing was carried out on a purpose-built outdoor practice hole. A stereoscopic high-speed camera was used to record club head and ball launch conditions prior to and immediately after impact. Two laser range finders were positioned approximately 250 yards (229 m) from the tee providing measures of carry and accuracy. There existed significant differences in overall performance between subjects. Club head velocity, spin axis tilt, launch angle and dispersion all exhibited inter-subject differences (p<0.05). In addition, club head velocity exhibited significant intra-subject variability (p<0.01) among all subjects. However, whilst statistically significant variations in carry and dispersion were observed for shots performed with matched drivers, absolute variation was actually very small (<0.5 %). Results suggest that golf research merits SS analysis although intra-subject variability was also noted among even elite level golfers. Keywords: golf; variability; performance.
1- Introduction Biomechanics researchers including Hatze (2005) and Farrally et al. (2003) have discussed the need for subject-specific investigation into human motion by means of computer models and movement simulation. They called for development of models that are anthropometrically tailored for individual subjects, therefore providing a better correlation between experimental and theoretical results. Indeed, Kenny et al. (2006) and 1. University of Limerick - E-mail: [email protected] 2. University of Ulster - E-mail: [email protected] 3. R&A Rules Ltd - E-mail: [email protected]
388 The Engineering of Sport 7 - Vol. 1 Nesbit and Ribadeneira (2003) have developed such models, aiding investigation into the golf swing and driver parameters. Statisticians including Bates (1996), Bates et al. (2004) and Kinugasa et al. (2004) have expressed confidence in conclusions drawn using appropriate statistical techniques which may be applied to perform analyses on data collected during single-subject investigations. In recent years biomechanical studies have been carried out on single subjects, for example by Bates, 1996; Bates et al., 2004; Kinugasa et al., 2004 and Reboussin and Morgan, 1996. However, despite a number of large and small-scale musculoskeletal models and computer simulations emerging in recent years, it remains to be ascertained whether golf drive shot performance analyses, using single-subject elite golfer kinematic data, is valid. Naturally, intra-subject trial data will usually correlate better than inter-subject data. The large number of degrees of freedom associated with whole body movements, and the larger number of motor control units and muscles involved in multi-joint movements mean that the method by which a golfer moves the driver club head from the address position to make appropriate impact with the ball is highly likely to differ in three dimensional space for each shot. According to Bernstein (1967), the musculoskeletal system with this large number of degrees of freedom, allows goal-directed tasks to be accomplished in a variety of ways. Task and mechanical constraints help to reduce the large number of degrees of freedom to a clearly recognisable and relatively invariant movement pattern. When learning a novel motor problem, the subject can resolve the problem by rigidly fixing (freezing) certain components and/or strongly coupling their displacements, thus reducing the number of initial degrees-of-freedom. Elite golfers are deemed to be highly efficient in creating this reduction. In the course of practice, these couplings could then be relaxed to permit more economical coordination through the use of the internal and external forces acting on the system. These hypotheses formulated by Bernstein (1967) have been confirmed by one study in which the subjects had to acquire a novel cycle or discrete coordination pattern (Temprado et al., 1997). Latash et al.’s (2002) study reported that “an essential feature” of a coordinative structure is that if one of the component parts introduces an error into the common output, the other components automatically vary their contribution to movement organisation and minimise the original error. It is this reorganisation that the present study addressed. During a round of golf, a golfer can make shots with a number of different clubs and in some cases club physical properties often differ by very small amounts. Using drivers that actually did not differ in physical properties, the authors sought to investigate, for a highly skilled cohort of players, did initial club head and ball launch conditions or shot outcome vary significantly between players, and between trials for individual players. As part of a larger investigation into the use of three dimensional motion analysis data to drive musculoskeletal models of the golf drive by the authors, intra-subject variability, or lack of, is important in the assessment of validity of such computer models.
The Engineering of Sport 7 - Vol. 1
389
2- Methods 2.1 Subjects & Equipment Six male subjects (0.1 ± 2.2 handicap, 22.1 ± 2.3yrs, 76.93 ± 9.45kg, 1.80 ± 0.04m) signed an informed consent form and completed an activity and medical history questionnaire. Ethical approval was granted by the University of Ulster Research Ethics Committee. Driver component brands were selected based on perceived quality and their popular use by elite golfers. Fifteen shafts and fifteen club heads were purchased, all of which underwent static testing in a laboratory to determine their key properties prior to club assembly. Club heads were tested for mass, volume, loft, lie and face area to identify three club heads best matched for these properties. Similarly, the fifteen shafts were statically tested for shaft mass, torque and frequency so that three closely matched shafts could be selected. A 350 cc, 200.05 g club head was temporarily fitted to each shaft in turn for the purposes of the test. The butt end of the shaft was clamped in a Golfsmith’s™ frequency analyser, a support placed at a distance of 0.15 m from the hosel and a weight clamp positioned 0.05 m from the hosel. A protractor fixed to the weight clamp was used to determine angular displacement of the clamp when a mass of 50 g was placed on the distal end of the clamp. A Golfsmith’sTM frequency analyser clamped the butt end of the shaft in place for assessment of shaft frequency. Masking tape was placed around the shaft and the tape marked around its circumference at 15 degree intervals. Downward pressure, resulting in 0.10 m deflection, was placed on the club head which, when released, allowed the club head to oscillate naturally. For each 15 degrees angular displacement, the test was performed 3 times and frequency values obtained from the analyser. Drivers were assembled by a skilled club assembly qualified PGA professional. Measures of loft, lie and overall mass were repeated during the assembly process to minimise clubs differences. Table 1 shows the physical properties for the main components measured for ‘matched clubs’. Some industry standard units are also shown (inches), as is also the case later in the current paper for drive length (carry in yards). Table 1 - Test clubs physical properties.
Data from a previous study (Egret et al., 2003) determined that normal club head velocity for the skill level of the subjects recruited was in excess of 44.7 ms-1 (100mph) and less than 51.4 ms-1 (115mph). As such, the speed of a swing affects the amount of bending experienced by the shaft, therefore club head/ball impact characteristics, the
390 The Engineering of Sport 7 - Vol. 1 magnitude of shaft deflection increasing as swing speed increases. Swing speed for the subjects recruited for the current study suited a stiff shaft. Club length for the subjects’ own drivers ranged from 1.13m (44.5”) to 1.17m (46”) and matched drivers were constructed 46° in length.
2.2 Experimental Procedures Testing was carried out on a purpose-built practice hole with a straight fairway cut 40 yards (36.58 m) wide, 330 yards (301.75 m) from tee to pin, with a raised tee box and visible flag on the green. Figure 1 illustrates the set-up used for testing. A stereoscopic high-speed camera positioned perpendicular to the intended direction of ball flight was used to record club head and ball launch conditions prior to and immediately after impact which included club head velocity, club head orientation, initial ball velocity, ball backspin, sidespin (component of spin axis tilt), and ball launch angle, both elevation and side angle. Two laser range finders were positioned approximately 250 yards (228.60 m) from the tee such that using calibration coordinates and known distance from one laser to the other, and the second laser to the tee, ball carry position as identified by two ball spotters could be determined within a coordinate frame, giving both carry and dispersion from a fairway centre line. Personnel were in place so that for each shot, data were recorded for club head and ball launch conditions using the launch monitor, for anecdotal information at the tee relating to quality and direction of the shot, and from each of the laser range finders for ball carry and dispersion. Premium balls were used for the present study. The test conditions, such as with the noted wind (average 1 to 5 km.hr-1, “Light air” on the Beaufort scale, right to left for shots being played), actual fairway and target factors, provided an ecologically valid test environment.
Figure 1 - Schematic testing set-up.
To evaluate golfer skill and determine the level of variability, as indicated by the range or standard deviation of measures recorded, subjects were required to hit shots with the test drivers. The three constructed ‘matched’ drivers, were randomly assigned the identification numbers 1, 2 and 3. Each golfer performed their usual warm-up routine which involved stretching followed by 10 practice shots with their own driver. Three sets of 8
The Engineering of Sport 7 - Vol. 1
391
trials were performed by each golfer using the randomly assigned matched drivers. Subjects were not informed that drivers were physically matched.
2.3 Data analysis Launch monitor and ball carry and dispersion data were amalgamated in tabular form using MSTM Excel v9.0.3821 SR-1 and included anecdotal information obtained from the subject and the experimenter at the tee. Anecdotal information identified any of the 8 shots which were mis-hit or which subjects reported as being markedly inferior. Additional shots were performed in such cases. Descriptive statistics were calculated relating to the central tendency of the measures recorded, namely mean, standard deviation, and the standard error in the mean (/n). Inter-subject variance was statistically analysed using a one-way ANOVA with a post-hoc LSD test applied to any measures that showed significant variance. ANOVA assumes that data has been sampled from populations that follow a Gaussian bell-shaped or normal distribution. Biological data never follow a Gaussian distribution exactly, because a Gaussian distribution extends infinitely in both directions, including both extremely low negative numbers and high positive numbers. But many kinds of biological data, such as that collected in the present study, follow a bell-shaped distribution that is approximately Gaussian. Because ANOVA works well even if the distribution is only approximately Gaussian these tests are used in many fields. Graphical display in SPSS of the data collected in the present study confirmed that the distribution was normal. The post-hoc test that was selected, LSD, provided the simplest and most powerful means by which to clearly identify where any differences rested, in this case signifying inter-subject variability.
3- RESULTS 3.1 Results Table 2 illustrates the means and standard deviations for all subjects for club head velocity, ball carry and dispersion. Also shown is the standard deviation of the mean for dispersion. It can be seen that there existed significant difference in overall performance between subjects. Club head velocity at impact and ball carry showed significant differences between subjects for all clubs, whilst dispersion from the fairway centre was statistically significant among subjects only for Club 3. No significant differences were found across the three clubs for any variables indicating no club effect. Table 2 - Club head velocity and shot performance means (± s.d.) for matched drivers for all subjects.
392 The Engineering of Sport 7 - Vol. 1 Table 3 shows further descriptive statistics across subjects for launch characteristics recorded by the stereoscopic launch monitor for all trials using each matched driver. Significant difference was demonstrated for measures of side angle, launch angle and backspin. The dispersion chart Figure 2 shows all shots performed by all subjects using the three drivers. Again, industry standard yards are the units used for graphical representation. The shaded area represents the fairway which was cut 40 yards (63.58 m) wide. Table 3 - Launch angles and spin rate means (± s.d.) for shots performed using matched drivers.
Figure 2 - Scatterplot for all shots by all subjects using three ‘matched’ drivers.
In addition, Table 4 presents descriptive data for club head velocity immediately prior to impact for each subject for each club. It can be seen that, in terms of mean club head velocity and standard deviation, there existed considerable difference in performance within the small group of elite golfers studied. Table 4 notes also present test scores for the one-way ANOVA and post-hoc LSD performed, showing significant intersubject variability (p<0.01) between numbered subjects. Furthermore, data showed considerable range in standard deviation of club head velocity for individual subjects from 0.24 ms-1 up to 0.96 ms-1 for sets of trials, indicating significant intra-subject variability, as the post-hoc LSD confirmed (p<0.01).
The Engineering of Sport 7 - Vol. 1
393
Table 4 - Club head velocity at impact means (± s.d.) for matched drivers for individual subjects.
*significant difference among subjects (F=177.73, p0.01); Post-Hoc LSD (subject v subject)1v3,4,5,6; 2v3,4,5; 3v4,5,6; 4v6; 5v6
4- Discussion Highly skilled acts are characterised more by the consistency of their output or results than by the consistency of the muscular contractions needed to achieve them. The aim of the current study was to investigate the launch characteristics and driving performance of low-handicap golfers using identical drivers for assessment of inter-subject variability. Inter- and intra-subject variability for club head velocity immediately prior to impact were both found to be significant, even with the elite group of golfers (<3 handicap) studied here. However, whilst statistically significant variations in carry and dispersion were observed for shots performed with matched drivers, absolute variations were actually very small. It may be concluded that such small variations are unlikely to have any affect on overall shot performance. The variable nature of performance may be debated as to whether or not it is a detractor to shot outcome. Traditionally, increased variability has been viewed as associated with decreased stability (Heiderscheit, 2000). However, variability in joint coordination may provide the flexibility required for superior movement execution. Whilst the subjects tested were ‘good’ amateurs with a high level of skill, there will nonetheless be a period of time needed during which the golfer will use feedback, afferent, auditory, tactile in nature, to become accustomed to new drivers. It may be the case that a subject will constantly perform poorly with a particular club no matter how long a period they have to become accustomed to it. It may require days or weeks of practice with a club in order to familiarise oneself with it, in which case the small number of trials used in the present study (n=8) is only an indication of shot variability. The use of drivers with which the golfer was unfamiliar would have introduced some error into the normal
394 The Engineering of Sport 7 - Vol. 1 swing path, but for the highly skilled golfers studied, this error is considered to be minimal. Significant inter-subject performance differences were noted for each of the test drivers. Differences, though, were mainly in terms of ball spin and launch angle, neither of which appeared to have any significant effect on carry nor dispersion from the fairway centre. Average differences in club head velocity and ball carry, measures which the majority of club manufacturers and golf coaches will highlight as being important, were less than 0.5%. These findings lead the authors to support the selection of any given subject from the cohort examined for future single-subject analyses, postulating that they would provide data representative of this skill level. Further development of computer simulations using a single highly-skilled (<3 handicap) golfer would provide a valid representation of the kinematics and kinetics of the elite golf swing.
5- Conclusion The present study aimed to address performance variability among elite level golfers to ascertain whether single-subject (SS) analysis is merited for golf swing studies. Whilst there existed significant differences in overall performance between subjects with matched drivers, absolute variation was actually very small (<0.5 %). Intra-subject variability was also noted among these elite level golfers. However, it is concluded that golf swing research merits SS analysis where the focus on an individual representative of an elite skill level may provide key swing characteristics that permit further in-depth investigation, such as that offered by computer modelling.
6- Acknowledgements We would like to express our thanks to the R & A Rules Limited and the College of Agriculture, Food & Rural Enterprise, Northern Ireland, Greenmount Campus for their valuable contributions to the study.
7- References [B1] Bates, B. T. (1996) Single-subject methodology: an alternative approach. In Medicine & Science in Sports & Exercise, 28(5), 631-638. [BR2] Bates, B. T., Rodger, C.J. and Dufek, J.S. (2004) Single-subject analysis, In Innovative analyses of human movement, (Ed. Stergiou, N.), Illinois, Human Kinetics, 3-28. [B3] Bernstein, N. (1967) Coordination and regulation of movements, Pergamon Publications, New York. [EV1] Egret, C.I., Vincent, O., Weber, J., Dujardin, F.H. and Chollet, D. (2003) Analysis of 3D kinematics concerning three different clubs in golf swing. In International Journal of Sports Medicine, 24, 465-469. [FC1] Farrally, M. R., Cochran, A.J., Crews, D.J., Hurdzan, M.J., Price, R.J., Snow, J. T. and P. R. Thomas (2003) Golf science research at the beginning of the twenty-first century. In Journal of Sports Sciences, 21, 753-765.
The Engineering of Sport 7 - Vol. 1
395
[H1] Hatze, H. (2005) Towards a comprehensive large-scale computer model of the human neuromusculoskeletal system. In Theoretical Issues in Ergonomics Science, 6(3-4), 239-250. [H2] Heiderscheit, B.C. (2000) Movement variability as a clinical measure for locomotion. In Journal of Applied Biomechanics, 16, 419-427. [KW1] Kenny, I.C., Wallace, E.S., Brown, D. and Otto, S.R., 2006, Validation of a full-body computer simulation model for the golf drive for clubs of differing length. In The Engineering of Sport 6, Munich, edited by Moritz, E.F. and Haake, S.J. (New York: Springer), pp.11-16. [K2] Kinugasa, T., Cerin, E. and Hooper, S. (2004) Single-subject research designs and data analyses for assessing elite athletes’ conditioning. In Sports Medicine, 34(15), 1035-1050. [LS1] Latash, M.L., Scholz, J.P. and Schoner, G. (2002) Motor control strategies revealed in the structure of motor variability. In Exercise and Sports Science Review, 30(1), 26-31. [NR1] Nesbit, S.M. and Ribadeneira, M.X. (2003) Sport biomechanical analysis using full-body computer models. In IASTED: Proceedings of the 14th IASTED International Conference on Modelling and Simulation, Palm Springs, USA, 24th-26th February. [RM1] Reboussin, D.M. and Morgan, T.M. (1996) Statistical considerations in the use and analysis of single-subject designs. In Medicine & Science in Sports & Exercise, 28(5), 639-644. [TD1] Temprado, J.J., Della-Grast, M., Farrell, M. and Laurent, M. (1997) A novice-expert comparison of (intra-limb) coordination sub-serving the volleyball serve. In Human Movement Science, 16, 653-676.
Power Measurement in Cycling using inductive Coupling of Energy and Data (P80) Reinhardt Tielert†, Norbert Wehn1, Thomas Jaitner2, Roland Volk1
Topics: Bicycle; Innovation & Design; Measurement Systems; Performance Sports; Testing, Prototyping, Benchmarking. Abstract: The power exerted on the pedal is the most reliable parameter to determine the training load in cycling biomechanically and hence a crucial factor to optimize performance. Commercial power meters are meanwhile part of the standard equipment of professional cyclists, but also used by an increasing number of non professional cyclists. In this paper we present a system to measure the torque, the cadence and power in cycling using inductive coupling of energy and data. Sensors and signal pre-processing electronics work without any battery on the turning parts. Under dynamic conditions, an overall accuracy of ±10% can be determined. The newly developed power meter can be characterized by low maintenance and energy consumption as well as by an increased number of detected physical values (e.g. 30 degree sectors, individual measurements of left and right pedals). This allows long-term measures as well as more detailed analyses of the pedalling techniques (coordination between left and right leg, e.g.). The first prototype has been successfully integrated into a bicycle and was tested under conditions of training. A second prototype that allows more detailed measures (e.g. section-wise detection of torque and power) runs under laboratory settings. Keywords: power meter; inductively air gap coupling of data and energy; sub-milliwatt range.
1- Introduction The power output exerted on the pedal is the most direct parameter to determine the training load in cycling biomechanically and hence it is a crucial factor to optimize performance (Jeukendrup & van Diemen, 1998, Stapelfeldt et al. 2007). Due to technological innovation during the recent years, power meters are meanwhile part of the standard equipment of professional cyclists, but also an increasing number of non professional utilizes technological support to improve their training. There are three systems that are widely used: The SRM power meter, the Power Tap system (PT) and the 1. Department of Electrical and Computer Engineering, Technical University, Kaiserslautern, Germany E-mail: wehn, [email protected]. 2. Department of Social Sciences, Technical University, Kaiserslautern, Germany - E-mail: [email protected]
398 The Engineering of Sport 7 - Vol. 1 Ergomo Pro system (EP). The SRM system consists of a crank set that continuously measures the power output from torque and angular velocity. The torque is determined by up to 8 strain gauge sensors that are located between the crank axis and the chainring. Due to its high validity and reliability the SRM system often serves as a reference system for the measurement of power output (Gardener et al. 2004, Bertucci et al. 2005, Duc et al. 2007). The PT system is also based on strain gauge measurement, but the sensors are located in the hub of the rear wheel. Both devices, the SRM and the PT, calculate the overall power output generated at the left and right pedal. The EP uses two optoelectronic sensors located in the bottom bracket. The torque is calculated by the torsion of the bottom bracket against the left crank. Hence, only the power output generated by the left limb can be determined. Assuming that the power output on both pedals is almost identical, the overall power output is displayed by multiplying the measured values by 2. This is considered as a major disadvantage of the EP because some cyclists have shown asymmetries in pedalling technique (Duc et al. 2007). On the other hand, the EP sensors are not sensitive to temperature and only a low additional mass is added compared to the standard equipment (0,074 kg /EP vs. 0,152 kg /PT or 0,280 kg /SRM, respectively). In this paper we present a new power meter system to measure the torque and power output using inductive coupling of energy and data. The main features of this system are the sector-wise detection of the torque, the differentiated measurement of the torques generated at the right and left pedal, its low weight, low power consumption and energy concept for turning parts.
2- Power meter The system for measuring the torques consists of two sensor units. One strain gauge sensor is located in the chain-ring and measures the torsion against the axle. This torsion is proportional to the torque and hence the total torque generated by the cyclist can be calculated. A second strain gauge sensor is integrated in the bottom bracket. By the torsion of the axle, the torque of the left pedal can be determined (fig 1). By subtraction of the torque of the left pedal from the total torque, we can distinguish the torque of the left pedal and the torque of the right pedal. For cadence measurement, 12 reed relays are arranged in a way that one turn of the crank-set is divided in sectors of 30º (fig. 2). Additionally, the torque and the power output can be detected in these 12 sectors for the left and right pedal, respectively. The overall weight of the power meter is 58g. Sensors and signal pre-processing electronics work without any battery on the turning parts. The power input is 6,4 mA. Not only the sensor values, but also the energy is inductively air gap coupled transmitted. The energy consumption of the total electronics is in the sub-milliwatt range.
The Engineering of Sport 7 - Vol. 1
399
Figure 1 - Bottom Brackets with Sensors for measuring torque of the left pedal.
Figure 2 - Crank-set with the turning part of the sensor-system.
3- Principle of transport for energy and data The measure system consists of fixed and turning parts. In figure 2, the circuit board that is turned by the chain-ring can be seen in front. Behind, there are the fixed parts of the measure circuits. The transmission of energy and data is realized by over-air magnetic induction via two coils that are fixed on each circuit board. A low power microcontroller is located on the circuit board which belongs to the fixed part of the measure system. The clock signal produced by the microcontroller is used to generate the signal for the energy transmission. In fig. 3, this signal is displayed in a dark blue colour.
400 The Engineering of Sport 7 - Vol. 1 For over-air transmission of the data, a long and a short pulse are modulated on the 8 MHz frequency by the turning part of the circuit. The rising and falling edge of the long signal can be detected quite clearly on the 8 MHz signal (dark blue) at the bottom of figure 3. On top, the demodulated signal (red) with its short and long pulse is shown. A change of the torque exerted at the pedal results in a variation of the electric resistivity. This variation is used to modulate the time shift between the short and the long pulse. Hence, the change of the torque is quantified by the time shift between the rising edges of the modulated signals. The amplitude of the power supply signal should not affect the detection of the position of the rising edge as well as the duration of the pulses to ensure a reliable data acquisition. Therefore, we use an analog comparator. So the rising and falling edge of the data signal (red) can be determined by a comparison with an additional modulated signal with a lower gradient (green). The light blue signal in figure 3 shows the positions of the rising and falling edges resulting from the analog comparator. In the following the calculation of the torque is described more detailed. Figure 4 shows exemplarily two pulses, a short one and a long one. The duration of short pulse is about 11μs, the long pulse lasts for about 45 μs. The length of these pulses is only necessary for the microcontroller to distinguish the two pulses by counting. t1 refers to the time interval between the rising edge of the short pulse and the rising edge of the long pulse, t2 indicates the interval between the long and the short pulse. The Sum of t1and t2 is fixed to about 3.36ms (figure 4).
Figure 3 - Signal for over-air energy- and data-transmission.
The Engineering of Sport 7 - Vol. 1
401
Figure 4 - Sequence of the short and the long pulse in the modulated signal.
A change of the torque does not affect the sum of t1+t 2, but varies the length of t1 and t2. So the torque can be calculated as follows
with calibration factor k and offset factor d. Due to tolerance range induced by the hardware components, t1 and t2 will not have the same length if the torque equal to zero. Therefore, an offset factor d is determined during the first milliseconds at the beginning. This factor is calculated as the mean value of t1-t2 under no-load-condition. The calibration factor k is a constant factor that can be derived by an initial calibration under static conditions. This calibration has to be carried out once for each power meter. Figure 5 shows exemplarily the results of a static calibration procedure with loads from 0 to 8.8 kg. A linear relationship between the difference of the time intervals t1-t2 and the load can be observed (correlation r=.99997). Similar results were obtained if the static calibration was repeated with loads up 45 kg.
Figure 5 - Determination of factor k by static calibration.
402 The Engineering of Sport 7 - Vol. 1
4- Accuracy under dynamic conditions Figure 6 shows the test set-up for the accuracy measures under dynamic conditions. A swinging-stator-dynamometer was connected to the axle of a bicycle. The bicycle was fixed on a TACX™ 1680 FLOW ergometer. Tests were run with a power output between 50 and 250 W. A high correlation (r = .995) was observed between the power exerted by the swinging-stator dynamometer and the power output measured by the power meter. Error values were in the range of 1.38% to 10.86% by a mean percent error of 7.22%.
Figure 6 - Bicycle coupled to the swinging-stator-dynamometer.
5- Discussion and Conclusions The newly developed power meter can be characterized by low maintenance overhead and energy consumption (no battery necessary on turning parts) as well as by an increased number of detected physical values (e.g. 30 degree sectors, individual measurement of left and right pedals). Due to the low energy consumption, it can be applied during long lasting training sessions or competitions such as cycling marathons. In addition, the power meter offers a wide range of application for more detailed analysis of the pedalling techniques. As example, the coordination between the right and left leg can be analyzed by a comparison of the power output exerted by the single legs. Further, the power output in the different sectors of the pedalling cycle provides an insight in the muscle force production at different joint angles or might even be used to optimize the cycling position. A first prototype has been successfully integrated into an Assisted Bicycle Trainer, an Ambient Intelligence system that was developed at the University of Kaiserslautern to support team training in cycling (Fliege et al. 2006). Up to now, several test runs were performed under training conditions. A second prototype that allows more detailed
The Engineering of Sport 7 - Vol. 1
403
measures (e.g. section-wise detection of torque and power) runs under laboratory settings. Further work emphasis the improvement of the accuracy of the current system which is not yet satisfactory.
6- Acknowledgement This work was supported by the research centre Ambient Intelligence at the University of Kaiserslautern.
7- References [BD1] Bertucci W, Duc S, Villerius V, Pernin JN, Grappe F. Validity and reliability of the PowerTap mobile cycling power meter when compared with the SRM Device. Int J Sports Med. 10:868-73, 2005. [DV1] Duc, S., Villerius, V., Bertucci, W. and Grappe, F. Validity and Reproducibility of theErgomo®Pro Power Meter Compared with the SRM and Powertap Power Meters. Int. J. Sports Phys and Perf, 2:270-281, 2007 [FG1] Fliege, I., Geraldy, A., Gotzhein, R., Jaitner, T., Kuhn, T., Webel, C. An ambient intelligence system to assist team training and competition in cycling. In: Moritz, E.F., Haake, S. (Eds.) The Engineering of Sport 6. Volume I: Developments for Sports, 97-102, 2006 [GS1] Gardner AS, Stephens S, Martin DT, Lawton E, Lee H, Jenkins D. Accuracy of SRM and Powertap power monitoring systems for bicycling. Med Sci Sports Exerc. 36:1252-1258, 2004. [JA1] Jeukendrup, A. E., van Diemen, A. Heart rate monitoring during training and competition in cyclists. J. Sports Science, 16: S91-S99, 1998 [SM1] Stapelfeldt, B., Mornieux, G., Oberheim, R., Belli, A. and Gollhofer, A. Development and evaluation of a new bicycle instrument for measurements of pedal forces and power output in cycling. Int J Sports Med, 28: 326-332, 2007
Online-Monitoring of Multiple Track Cyclists During Training and Competition (P81) Thomas Kuhn1, Thomas Jaitner2, Reinhard Gotzhein3
Topics: Bicycle; Measurement Systems; Virtual Reality & Computer application in Sports. Abstract: In cycling, the regulation of the training load is of particular importance to improve performance and to prevent overtraining. Meanwhile, biomechanical and physiological parameters such as power, cadence and heart rate can be monitored online by commercial power meters, and experienced athletes can use these data to adapt the load accurately to the actual physical disposition during training or competition. However, young and less experienced athletes do not possess the appropriate ability and knowledge for such deliberate control, and therefore external feedback or advice by the trainer is needed. An Assisted Bicycle Trainer (ABT) has been developed as an ambient intelligence system for the training of a group of cyclists (Fliege et al. 2006). The objective of the ABT is to improve training such that each cyclist is as close to his individual target heart rate as possible. The focus of this paper is on the tailored communication system of the Assisted Bicycle Trainer and on its outdoor evaluation. A prototype cyclist system consists of an Ergomo™ power meter including a heart rate receiver and a MicaZ mote (Crossbow 2007). MicaZ motes are low power and low weight nodes, consisting of micro controller, serial communication, and wireless transceiver, supporting the dynamic creation of a wireless network. We have devised a new communication stack, supporting robust multi-hop communication in high-mobility environments. The ABT network consists of three types of nodes: mobile cyclists, stationary or mobile trainer nodes and stationary forwarders. Forwarders can be used to extend the network coverage, i.e. to cover a round track. Sensor data from all cyclists is propagated to all trainer nodes and is visualized immediately. The trainer system is installed on a laptop with a sophisticated graphical interface to monitor and direct the training. Trainers can send commands to specific, or to all cyclists, enabling quick reactions, i.e. to change pulse rate or power output. Sensor data is stored on the trainer notebook and locally on every bicycle for future evaluation. Keywords: Ambient intelligence, Cycling, Embedded Computing, Wireless Ad-hoc Networks, Measurement Systems.
1. Department of Computer Science, University of Kaiserslautern, Germany - E-mail: [email protected] 2. Department of Social Sciences, University of Kaiserslautern, Germany - E-mail: [email protected] 3. Department of Computer Science, University of Kaiserslautern, Germany - E-mail: [email protected]
406 The Engineering of Sport 7 - Vol. 1
1- Introduction Preventing overtraining by regulating the training load is of particular importance in sports in general. During the last years, technique evolved and yielded affordable, commercial power meters that are able to monitor biomechanical and physiological parameters. Especially for bicycle training, the relevant parameters power, cadence, speed, and heart rate can be monitored by power meters. Typical examples are the SRM™ system (Schoberer Rad Messtechnik, Germany) and the Ergomo™ power meter (SG-Sensortechnik, Germany). While experienced athletes can use this monitored data to adapt their training to their physical disposition (Jeukendrup and van Diemen, 1998), younger, less experienced athletes lack the required knowledge and experience. These athletes require external guidance by their trainers, which require therefore access to live sensor values. This is especially true for bicycle training, and is of even greater importance, if cyclists train in groups (Gregor and Conconi, 2000). In a typical training situation, a trainer will either accompany several athletes by bicycle or car or, when cycling on a circuit or track, observe them from a convenient position. In both cases, the trainer normally will not be able to monitor status data of the cyclists using power meters, because existing budget-priced solutions are not able to transmit sensor values from cyclists to trainers over a wireless connection. This significantly affects the trainer’s possibilities to detect and immediately correct suboptimal training. Rather, they have to analyze data collected during training sessions and change the training plan afterwards. On the other hand, wireless transmission of sensor data and visualization to trainers in real-time support immediate intervention if a non-optimal training is perceived. We have developed the Assisted Bicycle Trainer (ABT) as a low-cost ambient intelligence system for the training of a group of cyclists (Fliege et al. 2006). The objective of the ABT is to improve training such that each cyclist is as close to his individual target heart rate as possible, while reducing the generated size and weight overhead to a minimum. This is achieved by facilitating wireless transmissions to trainer stations, which can send advice and commands back to cyclists. Interfaces to existing power meters or tailored hardware were especially developed for the ABT, including a communication system that supports wireless communications between cyclists and trainers. Trainers are equipped with notebooks that periodically receive sensor data of all cyclists. Bicycles may be equipped with different sets of sensors – it is neither necessary to equip every bicycle with the full set of sensors, nor is it necessary to equip all bicycles with the same set of sensors. We believe that this approach reduces costs and also the weight overhead for cyclists. The remaining parts of this paper are structured as follows: Section 2 describes the ABT and our reference scenarios. Section 3 describes the communication system of the ABT, which is one of its core components. Section 4 presents results from outdoor training and discusses results. Section 5 draws conclusions and lays out future work.
The Engineering of Sport 7 - Vol. 1
407
2- The Assisted Bicycle Trainer Our ABT consists of an Ergomo™ power meter (SG-Sensortechnik, Germany) including a heart rate receiver and a MicaZ mote (Crossbow 2007). Figure 1 shows the core unit of the Assisted Bicycle Trainer mounted on a bicycle. The MicaZ motes used for the ABT are low power and low weight nodes, consisting of a micro controller and a low bandwidth, ultra-low power wireless transceiver that enables the creation of a wireless network. All bicycles are additionally equipped with a small input and output device that provides interaction between the assisted bicycle trainer and the cyclist. The bicycle utilizes wireless communications. Trainers use pc-style hardware, i.e. notebooks that are connected to MicaZ motes. The communication system of the ABT supports multiple trainer nodes, and currently up to 30 cyclists in one network. The transceiver hardware is able to utilize different radio channels; therefore, multiple networks may be active at the same location in the same time interval.
Figure 1 - The core unit of the Assisted Bicycle Trainer mounted on a bicycle.
Three types of nodes form the Assisted Bicycle Trainer: mobile cyclists, trainer nodes, which can be stationary or mobile, and stationary forwarder nodes. Forwarder nodes are necessary, because the transmission range of the used low-power transceivers is limited. To extend network coverage to a larger area, i.e. the whole round track, forwarders that extend the transmission range of the mobile nodes are necessary. Extension of cyclist transmission ranges is supported by the tailored communication system of the ABT. The communication system periodically propagates sensor data from all cyclists to all trainer nodes, where it is visualized immediately by the trainer software system. The trainer system is installed on a laptop, with a sophisticated graphical interface to monitor and direct the training. Trainers have the ability to send commands to specific cyclists, or to all cyclists, enabling quick reactions i.e. to changing pulse rate or power output. Therefore, a variety of different training scenarios is supported by the ABT: – One possible training scenario is track cycling, where the cyclists train in a defined environment. In this environment, the trainer node will be stationary. According to the
408 The Engineering of Sport 7 - Vol. 1 training objectives, all athletes could cycle in a group, or every athlete could cycle at its individual speed. Depending on the size of the track, a direct wireless connection between cyclists and trainer cannot be guaranteed. The ABT handles this situation using stationary repeater nodes. With 8 repeater nodes, we were able to cover the track around a regular soccer field. For our evaluation, we mounted the forwarders on mobile tripods; alternatively, they might be located at fixed positions along the track. – In another possible scenario, the trainer follows a group of cyclist performing group training on a bicycle or in a car within transmission range of at least one cyclist. Although we cannot use stationary forwarders in this scenario, our hardware achieves more than 50 meters of transmission range under realistic outdoor conditions. More power consuming transmission technologies achieve transmission range from 500 to 1000 meters (MeshNetics 2008). – A third scenario of our assisted bicycle trainer is the individual training, in which an athlete cycles by oneself. The assisted bicycle trainer records all sensor data with timestamps and offers the facility to the trainer to analyze the data offline. The focus of this paper is on the online-monitoring of performance-related parameters and the communication between trainer and athlete. Our experiments with real athletes were performed on a round track; therefore, we will focus on the first scenario for the remaining part of this paper.
3- The communication system To provide sophisticated training functionalities, all nodes must be able to communicate with each other. For the ABT, the following communication requirements are identified: – Transmission of sensor values to all trainer stations. Therefore, efficient one-to-many communication must be provided by the network. – Communication between one cyclist and one trainer. This feature requires one-to-one communication to be available through the wireless network. –Communication from the trainer to all cyclists: This communication is used to broadcast one identical message to all cyclists. For this requirement, also one-to-many communication is necessary. – The ultra low power transmitters do not provide sufficient transmission range to transmit directly to all network nodes. Therefore, multi-hop communication must be provided by the communication system. Current wireless networks are not able to handle these requirements. Networking technologies like wireless LAN and Ultra Wideband require too much power, the required batteries with an adequate capacity would be too large to be effectively operated in that domain. Additionally, Wireless LAN in ad-hoc mode does not support multi-hop communication. In the domain of sensor networks, Zigbee is a standardized communication methodology (ZigBee 2003). Zigbee networks are able to provide multi-hop communication. However, ZigBee is not able to handle mobile nodes satisfactorily. Therefore, we have decided to create a tailored communication stack. The software system on every node consists of three layers: The hardware interface, the communication system and the application. Nodes of the ABT only differ in the
The Engineering of Sport 7 - Vol. 1
409
hardware interface and in the application. The communication system on all nodes is identical.
Figure 2 - Software structure of the Assisted Bicycle Trainer.
Figure 2 shows the main software layers of the cyclist nodes. The basic structure of the other nodes is similar. The application layer provides application specific services, such as processing input from the user. The middleware layer provides communication services, such as encoding of messages, sequence numbers, detection of duplicate messages and multi-hop communication. The lower layers are hardware specific and provide functionalities that are tailored to the hardware platform, i.e. the user interface component is tailored to the needs of the restricted user interface available to the cyclists. When multi-hop communication is used, messages may be transmitted over multiple paths to the receiver. Therefore, a duplicate detection system is necessary. The assisted bicycle trainer stamps every transmitted frame with the ID of the creator, and a unique sequence number for this creator. Forwarders forwarding the frame will not alter this information. Thus, the receiver can perform duplicate detection, eliminating duplicates of frames that arrive over multiple routing paths (Figure 3).
Figure 3 - Message forwarding over multiple routing paths.
410 The Engineering of Sport 7 - Vol. 1 The communication system is self-configuring. New trainers and cyclists are detected automatically by the communication system. Cyclists that left the training are removed after a defined amount of time. Forwarders can also be added and removed while the system is running. Our routing mechanism guarantees that a transmitted frame reaches the destination node as long as a path exists to it. The communication system uses two communication paradigms: – One-to-one communication is mapped to regular unicast communication. This is used in the case that a trainer communicates directly with exactly one cyclist, i.e. to issue new orders. – One-to-many communication is mapped to global broadcasts (Fliege et al. 2006) and realized using publisher/subscriber communication. Data is published on channels, i.e. cyclists publish their sensor data on their individual channels, and trainers publish their commands that are directed to all cyclists on a trainer channel. Every channel allows multiple publishers and multiple subscribers. This way, every node listens to those channels that are of interest to it, and filters out messages from unrelated channels. The publisher/subscriber communication is a powerful approach because it enables also peer-to-peer applications between cyclists. For example, a cyclist could subscribe to the sensor data channel of another cyclist, and therefore compare its training performance to the one of its competitor in real-time. Although these facilities are not used in our current system, future extensions may benefit from them.
4- Results and discussion So far, 4 bicycles have been equipped with the ABT. Several training sessions were run successfully in an outdoor environment. We placed the trainer and forwarder nodes as indicated in Figure 3. The cyclists moved around the course at their individual speed. The wireless communication was reliable and especially the fast movements of the cyclists did not affect the performance of the multi-hop communication system in a measurable manner. The performances of all cyclists were monitored on the trainer’s notebook (Figure 4). Heart rate, pedal power output, cadence and speed were shown for each cyclist separately. Additionally, the setpoint values for each parameter according to the individual training plan are displayed. Therefore, the trainer could easily identify deviations from the predefined training load and interfere immediately by sending a feedback to the athlete or changing the setpoint values. Directly after the training, a graphical presentation of the performance data was available for a first post-training analysis in short time. All data were stored for further analysis.
The Engineering of Sport 7 - Vol. 1
411
Figure 4 - Screenshot of the trainer application of the Assisted Bicycle Trainer.
Our current configuration supports training sessions up to 20 hours per bicycle and for all forwarders before recharging is required. In comparison to the operating times of commercial power meters such as SRM or Ergomo (12 to 30 hours), this is quite satisfactorily. Of course, the battery capacity of the notebook might be a limiting factor if a power plug cannot be used.
5- Conclusions In this paper, we have presented the communication system of our Assisted Bicycle Trainer, an ambient intelligence system that supports individual training, as well as team training in cycling. Our communication system for the Assisted Bicycle Trainer offers wireless communications between cyclists and trainers. Thus, online monitoring and intervention by trainers during a running training session is facilitated. Additionally, all sensor data is recorded for offline analysis after training. Several outdoor training sessions have shown the applicability and reliability of this system under real training conditions. According to youth cycling trainers, the Assisted Bicycle Trainer will make a considerable contribution to high-quality training, and therefore shall be used regularly for training and testing elite youth cyclists. Future work will be on the collection of accurate real-time sensor data to create tailored training plans for each cyclist. Based on this data, automated training analysis routines shall support the trainers in the development of performance over shorter as well as longer periods or time.
412 The Engineering of Sport 7 - Vol. 1
6- Acknowledgements The development of the Assisted Bicycle Trainer is a joint research activity between research groups from the computer science department, the electrical engineering department, and the social sciences department of the University of Kaiserslautern. It is conducted in the context of the research center “Ambient Intelligence” at the University of Kaiserslautern.
7- References [CT1] Crossbow Technology Inc. (2007) MICAz Wireless Measurement System. Document Part Number 6020-0060-03 Rev B. Retrieved October 15th, 2007 from www.xbow.com/Products/Product_pdf_files/Wireless_pdf/ MICAz_Datasheet.pdf [FG1] Fliege, I., Geraldy, A., Gotzhein, R., Jaitner, T., Kuhn, T., Webel, C. An ambient intelligence system to assist team training and competition in cycling. In: Moritz, E.F., Haake, S. (Ed.) - The Engineering of Sport 6. Volume I: Developments for Sports, 97-102, 2006 [GC1] Gregor, R.J., Conconi, F. Road Cycling. Blackwell Science, 2000 [JD1] Jeukendrup, A. E., van Diemen, A. Heart rate monitoring during training and competition in cyclists. J. Springer Science, 16, S91-S99, 1998 [IE1] IEEE Std. 802.15.4: Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal Area Networks (LRWPANs). IEEE Computer Society, New York, NY, USA, Oct. 2003. [ME1] MeshNetics: ZigBit Module. http://www.meshnetics.de/zigbee-modules. Valid in 02/2008.
A Model Predictive Controller for Sensor-based Training Optimization of a Cyclist Group (P82) Ankang Le1, Lothar Litz2, Thomas Jaitner3
Topics: Bicycle, Virtual Reality & Computer application in Sports. Abstract: Determining the optimal exercise intensity is a crucial factor in cycling to improve performance and avoid overtraining. Novel sensor technologies allow to optimize the training not only for an individual cyclist but also for an entire group. A sensor-based Team Cycling Training System (TCTS) has been developed to optimize the group training in cycling. This system consists of three major parts: a hardware platform with basic sensors for training data acquisition, a wireless ad-hoc network that establishes the communication among multiple bicycles, and a control algorithm for the optimization of the training. The focus of this paper lies on the development of the control algorithm, a Model Predictive Controller (MPC). The MPC uses a cycling performance model to predict the physical work loads of the cyclists according to various conditions such as road profile, headwind, speed and position of cyclists within the group. Based on the predicted physical exercise loads, the MPC uses a cyclist individualized dynamic heart rate prediction model to determine the physiological load of each cyclist and regulates the group training by advising the cyclists to change the position in group, to adjust the group speed, or to split the group in such a way that each cyclist can meet his training plan as exactly as possible. Training sessions with two or four group members have been conducted under different conditions. The results of the trainings indicate that the TCTS with the MPC is an effective aid for the group training in cycling. Keywords: model predictive control, training optimization, cycling group training.
1- Introduction Determining the optimal exercise intensity is a crucial factor in cycling to improve performance. Low intensities will not result in the desired training effect, but too high intensities may cause overtraining or illness (Kuipers and Keizer, 1988). Therefore, it is important to monitor exercise intensities during training and competition. Typically, biomechanical parameters such as power, cadence and speed are used to quantify the 1, 2. Institute of Automatic Control, University of Kaiserslautern, Germany - E-mail: le, [email protected] 3. Department of Social Sciences, University of Kaiserslautern, Germany - E-mail: [email protected]
414 The Engineering of Sport 7 - Vol. 1 external load. Among these parameters, the power exerted on the pedal can be considered as a direct and objective indicator of the external load (Coyle et al. 1991, MacIntosh et al. 2000, Stapelfeldt et al. 2006). To estimate the internal load or physical stress that results from an external load, the heart rate (HR) is a widely chosen parameter. The HR may change with the blood lactate concentration, hand (torso) position, temperature of the environment, altitude, training duration and so on. (Achten and Jeukendrup, 2003, Jeukendrup and van Diemen, 1998, Too 1990). However the study of Lucía et al. (2000) has confirmed that the values of the target HR generally remain stable for professional cyclists during the course of the season. Even though cycling is primarily known as an individual sport, teams play an important role in training and competition (Gregor and Conconi, 2000). In particular, team time trial (TTT) is a standard event in track and road races such as team pursuit and UCI Pro Tour TTT. Moreover, road cycling in groups is common in training. In typical team training, a group of cyclists covers a distance of up to 200 km with a varying road profile. The speed for all cyclists in the team is the same, but the power output depends on the position within the group. Due to the head wind the power output of the leading cyclist is up to 36% higher than the power output of subsequent cyclists (Neumann, 2000). To achieve the best training effects, each cyclist should ride with an individual exercise intensity that depends not only on factors such as cyclist’s physical capabilities and skills, bike aerodynamics, road surface and incline, head wind and temperature (Atkinson et al. 2003, Too 1990), but also on the position within the group. A sensor-based Team Cycling Training System (TCTS) has been developed to support the training of a group of cyclists (Litz et al. 2004, Jaitner et al. 2006, Le et al. 2007, Jaitner and Trapp, in print 2008). The objective of the TCTS is to optimize the group training such that each cyclist is as close to his individual predetermined exercise intensity as possible by changing the positions of the cyclists within the group or adjusting the group speed. The system consists of three major parts: a hardware platform with basic sensor technologies for training data acquisition, a communication middleware that establishes the communication among multiple bicycles, and a control algorithm for the optimization of the training. The focus of this paper lies on the control algorithm, a Model Predictive Controller (MPC). Based on a cycling performance model and a cyclist individualized HR prediction model (Le et al. in print 2008), the MPC predicts the physical and physiological loads of the cyclists online during training or competition. It optimizes the group training by minimizing the difference between actual and predicted HR values of the whole group. In section 2, we describe the TCTS system. The development of the MPC is presented in section 3. It is followed by the evaluation of the training results in section 4. We conclude this study in section 5.
2- The Team Cycling Training System (TCTS) The TCTS with four bicycles is shown in Figure 1. Each bicycle is equipped with an Ergomo™ Pro power meter (SG Sensortechnik GmbH & Co. KG, Germany) and an Ultra Mobile Personal Computer (UMPC). The communication between the Ergomo™
The Engineering of Sport 7 - Vol. 1
415
System and the UMPC is established via serial port (RS232). All UMPCs are connected to each other via an ad-hoc network formed by Wi-Fi technology (Jaitner and Trapp, in print 2008). The TCTS collects training status data dynamically and delivers them to a self-organizing automatic optimization algorithm which is running on the UMPC. Actual values as well as target values of heart rate, power, speed and cadence are displayed for each cyclist separately on his UMPC. The training instructions are also displayed on the panel guided by an audio signal simultaneously. These instructions may advise the cyclists to adjust the group speed or to change the position within the group. Additionally, all training data, instructions, information of group formation and training environment are stored on the UMPC for post-training analysis.
Figure 1 - The team cycling training system with 4 bicycles.
3- The Model Predictive Controller (MPC) In model predictive control, a dynamic process model is used for online prediction on a moving horizon of the future actions of the manipulated variables on the plant output. The future moves of the manipulated variables are determined by optimization with the objective of minimizing a defined cost function which comprises the predicted errors subject to operating constraints. The optimization is repeated at each sampling time based on updated information from the plant (García et al. 1989, Rawlings 2000). For the optimization of the cycling group training, the entire cyclist group is the control plant. The training intensities are the manipulated variables which are represented by the HR or power out of the cyclists. In order to optimize the training intensities of the cyclists by the MPC, a cycling performance model and a HR model are required to predict the physical and physiological loads of the cyclists during training. These models will be described in the following subsections.
2.1 The cycling performance model In cycling, the power of a cyclist is required to overcome a complex interaction of resistive forces presented by the cycling environment. These forces include air resistance, rolling resistance, gravity, inertia and frictional losses from the drive chain and wheel bearings (Atkinson et al. 2003, Faria et al. 2005).
416 The Engineering of Sport 7 - Vol. 1 Air resistance (FW) is the major resistive force at normal cycling speed. It increases with the square of speed and is given by (1) where cd represents the drag coefficient, Ap is the projected frontal area of cyclist and bike, is air density, vb is steady-state bike speed and vW is head (positive) or tail (negative) wind speed (Olds et al. 1993). Cyclists may have the drafting effect by riding in the slipstream of the lead rider. The effect of drafting diminishes for the following cyclists on subsequent positions as the wheel spacing increases. A correction factor was derived by Olds (1998) as (2) where CFdraft is the ratio of air resistance under drafting conditions to that without drafting, and dw is the wheel-to-wheel distance between the preceding and the following cyclists. It is assumed that there is no benefit of drafting by cycling more than 3 m behind another rider (Olds 1998). The second major resistive force that must be overcome is rolling resistance (FR). It is expressed as (3) where GR is the road gradient, Mc and Mb are the masses of the cyclist and the bike respectively, cR is the coefficient of FR and g is the gravity acceleration (Martin et al. 1998). If the course is not flat, work is performed against or with the grade. The force (FG) due to the road gradient is related to the mass of rider and bike and is represented by (4) Besides the physical resistances described above, work is also performed against or with the cycling direction while varying the bike speed. The force (FA) due to the speed variation is shown by equation 5. (5) with acceleration a. Thus the power to overcome these resistances for the lead cyclist without drafting is calculated as (6) Additionally, when the wheels rotate, the spokes slice through the air like the blades of a fan. That causes also resistance (Martin et al. 1998). Finally, the frictional losses in wheel bearings and drive chain must also be considered. However, the wheel rotating resistance and frictional losses are relatively small and can be expressed together as a mechanical efficiency factor cm. Therefore the total power output PLT of the lead cyclist is modelled by (7)
The Engineering of Sport 7 - Vol. 1
417
Similarly, the total power output PDT of cyclist riding with drafting is calculated as shown by equation 8. (8) If all of the model parameters can be accurately determined, these equations should provide a precise prediction of the necessary power to overcome all of the resistances in cycling. However the power output represents only the external physical load of cycling. A well-trained cyclist can perform a high power output, but the same power value might be too high for other cyclists with lower training conditions. Therefore, to prescribe the optimal exercise intensity for an individual cyclist, we need a HR model to predict the internal physiological load of the cyclist under certain power output.
2.2- The heart rate prediction model Le and others have developed a dynamic HR prediction model and confirmed its validity under laboratory conditions (Le et al. in print 2008). This model predicts the HR of a cyclist based on the physical response of the HR to the exercise load online during training or competition. It considers the diverse HR responses to different exercise intensities according to the cyclist’s individualized blood lactate threshold. Eight parameters are used to calculate the future heart rate values from current values, training duration and training load by taking consideration of other effects such as exhaustion and recovery. Mathematically, the HR kinetics during training is modelled by the equations 9 and 10 (Le et al. in print 2008). (9) (10) where k is sampling step, HR(k) is HR value of k, HRS is HR value before the exercise start, and HR(k) is the change of the HR due to the cycling workload. P(k) is the power output of the cyclist, TA is sampling time, and is a time constant related to training duration. HRiAT is the HR at cyclist’s individual anaerobic threshold. Kon is the time point, at which HR(k) is greater than HRiAT for the first time. () is the unit step function with (11) K1, K2, K3, K4 and K5 are model parameters which represent the ratio of the HR kinetics to training work load under different intensity levels. They were identified using the least-squares method by minimizing the loss function that is the sum of the squared differences between the measured data and the fitted functions. Based on the cycling performance model and HR model, the design process of the MPC algorithm can be started.
418 The Engineering of Sport 7 - Vol. 1
2.3 The control algorithm The purpose of the MPC is to find the future values of the control signals by minimizing a predetermined cost function which comprises the predicted errors subject to operating constraints. For the group training optimization, each cyclist in the group should meet his training plan as exactly as possible. It means that the difference between the training HR and the predetermined reference value of each cyclist must be minimized. Therefore, the errors between the predicted HR and reference values from the training plan are comprised in the cost function. In group training, the group speed must be the same for all the cyclists. Due to the drafting effect, the training load of each cyclist can be optimized based on his individual performance profile and training plan by riding in different positions within the group. Thus, we take these three control signals for the training optimization: changing positions within the group, adjusting the group speed, and splitting the group. These control signals have different influences on the heart rates of the cyclists while regulating the training loads. The MPC must select the best way to regulate the load at each sampling step so that each cyclist can meet his training plan as exactly as possible. To achieve this objective, these control signals are taken into consideration while minimizing the cost function. Therefore, the cost function is defined to consist of the following two parts: 1. The errors between the predicted HR and reference values weighted by the position number within in the group. 2. Control signals: changing positions within group, adjusting group speed and splitting the group. Mathematically, the cost function J is expressed by equation 12. (12) where L is the number of prediction steps, M is the minimal formation time. During the period of M*TA, the group formation can not be changed. N is the number of the group members, n is the position number in the group, Wn is the normalized weighting factor of the cyclist’s position in the group with . (13) (HRn) is the error between the predicted HR and the reference value of the cyclist in position n. WP, WS and WG are weighting factors for the control signals Pos, Spd, and Grp respectively. The control signal of splitting the group Grp is not considered for a homogenous small group and is not considered in our training tests either. If the cyclist position and the group speed are not necessary to be changed, both Pos and Spd have the value 0. Pos = 1 when the cyclist position is to be changed. Spd = 1 for the cases group speed will be increased or decreased. Thus we have four possibilities at each prediction step in total: changing position, speed increment, speed decrement and no changing. For the L prediction steps, we have 4L possible combinations of the control signals within the prediction horizon L*M*TA. Each combination of the control signals
The Engineering of Sport 7 - Vol. 1
419
yields one value of J. There is one minimum among these 4L of J values. Then, the combination of the control signals that corresponds to the minimal J gives the optimized control sequence for the period of L*M*TA. However, only the first control signal of the optimized sequence is applied to the cyclist group. At the next sampling step, new sensor information is gained. Based on the new information, the entire control trajectories are recalculated and the optimization procedure is repeated.
4- Evaluation of training results Several training sessions of two- or four-cyclist group were conducted on a circular stadium track to evaluate the MPC. As an example, we will show the training results of a two-cyclist group. The physiological characteristics and the HR model parameters of the cyclists participated in the test are listed in Table 1. The values of the HR model parameters were identified by least squares method of the interval training data (Le et al. in print 2008). Table 1 - Physiological characteristics and HR model parameters of the cyclists participated in the training tests
The environment temperature was between 14.5 °C and 15.3 °C during the training test with a sea level atmospheric pressure. Because the test was conducted on a circular track and the wind speed was very low, the influence of wind speed was neglected. The training was started after a warm up with the start HR of 108 and 100 bpm for cyclist AC and DS, respectively. The standard dropped racing posture was taken by the cyclists. The total projected frontal area of the cyclist and bicycle was calculated corresponding to Heil (2002) by the equation (14) According to our test bicycle type and track surface, we took the parameters listed in Table 2 to calculate the resistances and cost function. The parameters for the calculation of the resistance were selected by comparing with the data from Di Prampero (2000) and Olds (1998). The parameters for the cost function were determined by our training simulations according to the cyclist’s performance and training plan. From these parameters, we can see that the prediction horizon L*M*TA is 30 seconds. The minimal leading time M*TA is 6 seconds. Table 2 - Parameters for the calculation of cycling resistances and cost function
420 The Engineering of Sport 7 - Vol. 1 During the training, the cyclists were required to ride with a wheel-to-wheel distance of 0.5 m between the leading and subsequent cyclists. While changing the position of cyclists, the lead cyclist was required to stop pedalling shortly and then drop to the end of the group. The cyclist behind him took the lead position automatically without varying the group speed. When the controller gave out the instruction of adjusting the group speed, the lead cyclist was instructed to adjust the group speed such that his power output might be changed by 20 watt. The training result is shown in Figure 2. The target HR of cyclist AC and DS are 135 and 130 bpm, respectively. The mean values of their training results are 135.06 and 130.13 bpm, respectively. The standard deviations are 3.38 and 4.66 bpm.
Figure 2 - Training result of
a two-cyclist group.
The training result of cyclist AC is shown in Figure 3. The grey line is the distribution of his position during the training. Position 1 means the lead position and 2 the end of the group with drafting. We can see that the training is started when cyclist AC is in lead position. His HR rises when he is in lead position and drops when he is in non-lead position. During the 28 minutes training test, the cyclist AC is about 12 minutes in the leading position.
Figure 3 - Training result of
cyclist AC.
The Engineering of Sport 7 - Vol. 1
421
The distribution of the control signals are shown in Figure 4. In order to present the control signals more obviously, they are shown in the figure with different values. Value 2 means position change, value 1 means group speed increment, value 0 means no change (keep going) and value -1 means group speed decrement. We can see that after the beginning of the training the controller requires the cyclists to increase the group speed. When the group speed has reached the desired value, the controller instructs the cyclist to hold the speed and the group formation. After about 1 minute, the cyclists are instructed to change the position. There are totally 29 times of position changes during a training period of 28 minutes. The position is changed nearly once per minute on average. Because of the cardiac drift effect, the cyclists are also instructed to decrease the group speed 6 times during the training. The frequency of changing the position and speed can be adjusted by the weighting factor WP and WS in the cost function according to the training plan.
Figure 4 - Distribution of control signals during training period
5- Conclusion In this study, we have presented a model predictive controller integrated in a sensorbased TCTS system for the optimization of the cycling group training. Based on a cycling performance model and a dynamic heart rate prediction model, the model predictive controller predicts the exercise intensity of each cyclist in the group and optimizes the group training by minimizing a predetermined cost function. Training test results of a two-cyclist group on a circular stadium track indicate that the TCTS with the model predictive controller is an effective aid for the group training in cycling
6- Acknowledgement This work was supported by the research centre Ambient Intelligence at the University of Kaiserslautern. The authors gratefully acknowledge Marcus Trapp for his assistance in the completion of this study, as well as Andreas Christmann and Daniel Schmidt for their enthusiastic participation in the training tests.
7- References [AD1] Atkinson G., Davison R., Jeukendrup A. and Passfield L., Science and cycling: current knowledge and future directions for research, Journal of Sports Sciences, 21: 767-787, 2003. [AJ1] Achten J. and Jeukendrup A., Heart rate monitoring: Applications and Limitations. Sports medicine, 33(7): 517-538, 2003.
422 The Engineering of Sport 7 - Vol. 1 [CF1] Coyle E.F., Feltner M.E., Kautz S.A., Hamilton M.T., Montain S.J., Baylor A.M., Abraham L.D. and Petrek G.W., Physiological and biomechanical factors associated with elite endurance cycling performance. Medicine and Science in Sports and Exercise, 23: 93-107, 1991. [D1] Di Prampero P.E., Cycling on Earth, in space, on the Moon. European Journal of Applied Physiology, 82: 345-360, 2000. [FP1] Faria E.W., Parker D.L. and Faria I.E., The science of cycling, physiology and training – part1. Sports medicine, 35(4): 285-312, 2005. [FP2] Faria E.W., Parker D.L. and Faria I.E:, The science of cycling, factors affecting performance – part2, Sports medicine, 35(4): 313-337, 2005. [GC1] Gregor, R.J., and Conconi F., Road Cycling. Blackwell Science, Oxford, 2002. [GP1] García C.E., Prett D.M. and Morari M., Model predictive control: theory and practice_a survey, Automatica, 25(3): 335-348, 1989. [H1] Heil D.P., Body mass scaling of frontal area in competitive cyclists not using aero-handlebars. European Journal of Applied Physiology, 87: 520-528, 2002. [JT1] Jaitner T., Trapp M., Niebuhr D. and Koch J., Indoor-simulation of team training in cycling. In: Moritz, E.F., Haake, S. (Eds.) The Engineering of Sport 6. Volume I: Developments for Sports, 103-108, 2006. [JT2] Jaitner T. and Trapp M., An ambient intelligence system to support team training in cycling. E-Journal Movement and Training, in print, 2008. [JV1] Jeukendrup A. and Van Diemen A., Heart rate monitoring during training and competition in cyclists. Journal of Sports Sciences, 16(3): 91-99, May 1998. [KK1] Kuipers H. and Keizer H.A., Overtraining in elite athletes. Sports Medicine, 6: 79-92, 1988. [LH1] Lucía A., Hoyos J., Pérez M. and Chicharro J.L., Heart rate and performance parameters in elite cyclists: a longitudinal study. Medicine and Science in Sports and Exercise, 33(10): 1777-1782, 2000. [LJ1] Le A.; Jaitner T. and Litz L., Sensor-based Training Optimization of a Cyclist Group. Proceedings of 7th International Conference on Hybrid Intelligent Systems, pp. 265-270, September 2007. [LJ2] Le A., Jaitner T., Tobias F. and Litz L., A Dynamic Heart Rate Prediction Model for Training Optimization in Cycling. The 7th conference of the international sport engineering association. in print, 2008. [LW1] Litz L., Wehn N. and Schuermann B., Research Centre "Ambient Intelligence" at the University of Kaiserslautern. VDE Kongress, 1: 19-24, 2004. [MM1] Martin J.C, Milliken D.L., Cobb J.E., McFadden K.L: and Coggan A.R., Validation of a mathematical model for road cycling power, Journal of Applied Biomechanics, 14: 276-291, 1998. [MN1] MacIntosh B.R., Neptune R.R. and van den Bogert A.J., Intensity of cycling and cycle ergometry: Power output and energy cost. In: B.M. Nigg, B.R. MacIntosh, J. Mester. (Eds.) Biomechanics and biology of movement, 129-148, 2000. [N1] Neumann G., Physiologische Grundlagen des Radsports (Physiological Bases of Road cycling). Deutsche Zeitschrift für Sportmedizin, 5: 169-175, 2000. [O1] Olds T.S., The mathematics of breaking away and chasing in cycling, European Journal of Applied Physiology, 77: 492-497, 1998. [ON1] Olds T.S., Norton K.I. and Craig N.P., Mathematical model of cycling performance, Journal of Applied Physiology, 75(2): 730-737, 1993. [R1] Rawlings J.B., Tutorial overview of model predictive control, IEEE Control Systems Magazine, pp. 38-52, 2000.
The Engineering of Sport 7 - Vol. 1
423
[SM1] Stapelfeldt B., Mornieux G., Oberheim R., Belli A. and Gollhofer A., Development and evaluation of a new bicycle instrument for measurements of pedal forces and power output in cycling. International Journal of Sports Medicine, 28: 326-332, 2006. [T1] Too D., Biomechanics of cycling and factors effecting performance. Sports
A Dynamic Heart Rate Prediction Model for Training Optimization in Cycling (P83) Ankang Le1, Thomas Jaitner2, Frank Tobias3, Lothar Litz4
Topics: Bicycle, Modelling. Abstract: Heart rate can be considered as a reliable indicator of the physiological load both for immediate training monitoring and for post-training analysis in cycling. The aim of this paper is to present a dynamic heart rate prediction model which will be used by a model predictive controller to optimize the cycling training. This model predicts the heart rate of a cyclist online during training or competition based on the physical dynamics of the heart rate to exercise work load. It uses eight parameters to calculate the future heart rate from current values, exercise duration and exercise work load by taking consideration of other effects such as fatigue, exhaustion and recovery. These parameters are identified from training data of nine well-trained cyclists by the least squares method. Each cyclist performed first a stepwise incremental test on a bicycle ergometer to determine his individual anaerobic threshold. Afterwards, they executed two interval tests on the same bicycle ergometer according to their individual anaerobic thresholds for the identification and evaluation of the model parameters. For all subjects, the mean absolute error and standard deviation between the measured and modelled heart rate values without updating are 3.06 and 3.95 bpm respectively. The mean correlation coefficient is 0.9686. If the model output is updated with the measured values every 20 seconds, then the mean absolute error is 1.31 bpm, the standard deviation is 1.92 bpm and the mean correlation coefficient is 0.9907. The result indicates that this model is able to predict the heart rate of cyclists accurately and can be used by a model predictive controller for training optimization. Keywords: heart rate prediction model, training optimization, cycling, exercise intensity.
1- Introduction The exercise intensity is the most crucial factor to improve a cyclist’s performance. If the exercise intensity is too low, performance will not be improved or even reach a lower level. Excessive training at high intensities may lead to illness or overtraining (Kuipers and Keizer, 1988). Therefore, it is important to monitor the exercise intensity during trai1, 4. Institute of Automatic Control, University of Kaiserslautern, Germany - E-mail: le, [email protected] 2, 3. Department of Social Sciences, University of Kaiserslautern, Germany - E-mail: [email protected]
426 The Engineering of Sport 7 - Vol. 1 ning. When determining the best way to monitor exercise intensity, a balance has to be found between the validity of the parameter and practicality of using that parameter for intensity monitoring (Achten and Jeukendrup, 2003, Jeukendrup and Van Diemen, 1998). Recent developments of sensor technologies allow to monitor the exercise intensity by measuring the heart rate (HR) as well as the power output of the cyclist during the training or competition for immediate feedback to the athlete and for later analysis. Compared with other indicators of exercise intensity such as speed, cadence or percentage of maximal oxygen uptake, HR and power output are reliable and easy to monitor, and can be used in most situations (Achten and Jeukendrup, 2003, Faria et al. 2005, Gilman 1996). Furthermore, HR plays an important role in the detection and prevention of overtraining (Jeukendrup and Van Diemen, 1998). Some previous studies have modelled the HR with monoexponential or biexponential equations. HR was expressed as the sum of a baseline value plus one or two first order e-function with time delay (Bearden and Moffatt, 2001, Mavrommataki et al. 2006, Linnarsson 1974). However, in those models the work load values were not directly considered by the calculation of the heart rates. Those studies were conducted with constant work loads, moderate or heavy exercise intensities. They did not describe the dynamic responses of HR to the varying workloads quantitatively. Furthermore, the cardiac drift effect was not taken into consideration. Heart rates have been shown to drift upwards up to 20 beats per minute (bpm) during exercises lasting 20-60 minutes, despite unchanged work loads and steady or decreasing plasma lactate concentrations (Kindermann et al. 1979, Mognoni et al. 1990). Another study has shown that while exercising at a HR which is 5% below the HR at the ventilatory threshold, the work load has to be reduced by approximately 17% (from about 220 to 183 watt) over time, although the HR was kept relatively stable (176-180 bpm) (Boulay et al. 1997). Cardiac drift is accentuated by numerous factors such as dehydration and head stress and therefore an important factor for the HR modelling (Achten and Jeukendrup, 2003). The HR prediction model used in the study of Le et al. (2007) has considered the cardiac drift effect, but the influence of the aerobic and anaerobic mechanisms of the energy metabolism on the HR kinetic was not taken into account. Depending on the individual’s physical state, the aerobic and anaerobic mechanisms of the energy metabolism will interact differently to provide the energy needed to perform at a given intensity (Bangsbo et al. 1990). The study of Grazzi et al. (1999) confirmed by 500 tests during a 2-year period that the deflection point of the power output/heart rate relationship at the anaerobic threshold is a physiological phenomenon. Thus, the individual blood lactate threshold and the dynamics between power output and blood lactate should also be considered while modelling the HR kinetics to prescribe the adequate exercise intensity in training. The purpose of this investigation is to present a dynamic heart rate prediction model which will be used by a model predictive controller (MPC) to optimize the training of a cyclist group (Le et al. in print 2008). In the group training optimization, a cycling performance model and a cyclist individualized HR model are used by the MPC to predict the future HR values for each cyclist online during training or competition. Based on the predicted information, the MPC regulates the training load of each cyclist
The Engineering of Sport 7 - Vol. 1
427
by advising them to adjust the group speed or to change the positions within the group in such a way that each cyclist can reach his predetermined exercise intensity as far as possible. This model predicts the HR of a cyclist based on the physical response of the HR to the exercise load. It considers the diverse HR responses to different exercise intensity levels according to the cyclist’s individualized blood lactate threshold. Eight parameters are used to calculate the future heart rate values from current values, training duration and training load by taking consideration of other effects such as fatigue, exhaustion and recovery. In section 2, the subjects, equipments for the test, test protocol and procedure will be introduced. Moreover, the mathematical model will be derived in this section. In next section, the result of the model identification and evaluation will be presented. It is followed by a short discussion about the results in section 4. We will conclude this study in section 5.
2- Methods 2.1 Subjects and equipments Nine healthy well-trained male cyclists participated in the study. Their levels of performance ranged from regional to national. The possible risks and benefits of the tests were fully explained to the subjects before they gave their consent. All cycling tests were carried out on an electromagnetically braked cycle ergotrainer (Tacx™ T1680 Flow, Netherlands) with a built-in power meter (Ergomo™ Pro, SG Sensortechnik, Germany). The Ergomo™ power meter determines the current power output of the cyclist right in the bottom bracket. During pedalling, the axle twists slightly. This torsion is measured by comparing two signals and converted into power by using the angular velocity. The manufacture confirms that this system has a power measurement precision of ±0.5%. Additionally, it calculates the cycling speed by means of a magnet placed on the rear wheel and monitors the HR with the help of a pulse transmitter (e.g., Polar™ WearLink, Polar Electro GmbH, Germany).
2.2 Test protocol and procedure Each subject performed three test sessions in our laboratory on different days. They were requested not to participate in any competition seven days before the tests started. Additionally, they did only light-recovery training 24 hours before each test. The tests for each cyclist were performed over a 10-days period with at least one day recovery between two tests. During the tests, the cyclist was allowed to drink water. Our laboratory had a temperature of about 21-22 °C with a sea level atmospheric pressure during the test period. First, the power output (PiAT) and HR (HRiAT) at the individual anaerobic threshold (iAT) of each cyclist were assessed via a stepwise incremental cycling test on the ergometer. Each cyclist was required to pedal at a constant power output with his favourable cadence, beginning at 50 W with a stepwise increase of 50 W every three minutes. The
428 The Engineering of Sport 7 - Vol. 1 test was completed when the cyclist could not maintain the required power output. HR and workload were measured and stored every second. Arterialized blood samples were taken from the hyperemic earlobe (Stegmann et al. 1981) at rest, during the last 30 s of each exercise step and at 1, 3, 5, and 10 minutes post-test for the determination of the lactate concentration. The cyclist remained seated during the post-test recovery period. The iAT was determined using a lactate analyzer (Super GL ambulance, Dr. Mueller Geraetebau GmbH, Germany) and verified by the method described by Stegmann et al. (1981). After PiAT and HRiAT had been determined, each cyclist was requested to perform two interval tests on the same ergometer. The first test had a low work load Plow (60% of the PiAT) and a high work load Phigh (110% of the PiAT). It started with a 15 minutes warm-up, followed by two minutes Plow and three minutes Phigh in alternating sequence. The cyclist pedalled with his favourable cadence and tried to keep the cadence stable. The test was ended when the cyclist reached his subjective exhaustion. During the test the values of HR, power output, cadence and speed were recorded every second and stored in a computer for the HR model identification. The protocol of the second interval test was similar with the first one, but had a Phigh = PiAT. The result of the second test was applied to evaluate the parameters of the HR model identified by the first interval test.
2.3 Derivation of the mathematical model According to the fundamental physiological principles, the HR response to the training work load during exercise is generally expressed by the equation (1) where k is sampling step, HR(k) is HR value in bpm, HRS is HR value before the exercise start, and HR is the change of HR according to the cycling workload at sampling step k. Due to the fact that the dynamic of the HR responses to the work loads may differ under different exercise intensity levels, HR is modelled separately according to diverse exercise intensity levels. Under moderate exercise intensity, the energy production mainly relies on the aerobic pathways. It is well known that HR increases linearly with increasing exercise intensity over a wide range up to sub maximal intensities (Achten and Jeukendrup, 2003, Grazzi et al. 1999, Lucía et al. 2000). Therefore, under moderate exercise intensity HRmod is modelled as (2) with P(k) the work load in watt. Because of the cardiac drift effect, HR can drift upwards by up to 20 bpm during exercise over time, despite unchanged work loads and steady or decreasing plasma lactate concentrations. The exact amount of the HR drift depends on the training duration and actual work load. Thus the cardiac drift is modelled by equation 3. (3)
The Engineering of Sport 7 - Vol. 1
429
where TA is the sampling time and Ù is a time constant related to the training duration. The individual anaerobic threshold represents the maximal lactate steady state, i.e. the balance point of oxygen supply and demand (Kindermann 2004, Stegmann et al. 1981). Up to this point, the oxygen delivered by the cardiovascular system is sufficient to meet the energy needs of the body. However, when the HR is over the HRiAT, the demand for oxygen exceeds the rate of supply and the muscles have to rely on the stored reserves in the body. The energy production becomes more dependent on anaerobic glycolysis. That results in an oxygen deficit in the body and a rapid increase of blood lactate. Training in this phase can not be maintained for long periods and will lead to exhaustion. The HR response to the training work load in this phase is different from that under moderate exercise intensity. It is modelled by equation 4. (4) where () is the unit step function with (5) The term HRexhaust represents the exhaustion effect on the HR. It integrates the total duration while the cyclist is exercising over his individual anaerobic threshold. The longer the cyclist exercises over his iAT, the greater becomes the effect of the exhaustion. However, the effect of the exhaustion can be compensated by recovery with exercise intensity under the PiAT. During the recovery the cardiovascular system is still stimulated and tries to pay back the oxygen deficit by supplying oxygen to help break down the lactates. When the blood lactate is decomposed, the muscles are regenerated and have again the ability to undertake intensive exercise. The effect of the recovery is expressed by the equation (6) where Kon is the time point, at which HR(k) is greater than HRiAT for the first time. This parameter Kon gives the condition that the recovery effect can occur only after exhaustion. Summing equations 1, 2, 3, 4 and 6 yields the following equation for the HR model: (7) (8) Here K1, K2, K3, K4 and K5 are model parameters which represent the ratio of the HR kinetics to training work load under different intensity levels. They are identified using the least-squares method by minimizing the loss function that is the sum of the
430 The Engineering of Sport 7 - Vol. 1 squared differences between the measured data and the fitted functions (Noble and Daniel, 1988).
3- Results The physical and physiological characteristics of the subjects used in the study are presented in Table 1. The mean and standard deviation of PiAT and HRiAT for all cyclists are 223.9±26.2 W and 149.8±7.3 bpm, respectively. They were determined by the stepwise incremental cycling tests. Table 1 - Physical and physiological characteristics of all subjects participated in this study.
Table 2 shows the parameters of the heart rate models of all subjects identified by the least squares method using the data of the interval tests. Among the parameters from K1 to K5, the parameter K2 is the dominant one with the mean value of 0.9856. Table 2 - Parameters of the heart rate prediction models of all subjects.
Fig. 1 - The trajectories of the modelled (without updating) and measured heart rate.
The Engineering of Sport 7 - Vol. 1
431
Table 3 summarizes the descriptive statistics of the mean absolute error, the standard deviation and the correlation coefficient between the measured HR and the HR calculated by the mathematical model. For the calculation of the model output with equation 8, the power output of the subject P(k) was the same as that of the interval test measured by the Ergomo™ power meter. The mean absolute error and standard deviation between the measured and modelled heart rate values without updating are 3.06 and 3.95 bpm, respectively. The mean correlation coefficient is 0.9686. When the model output is updated with the measured values every 20 seconds, the mean absolute error is 1.31 bpm, the standard deviation is 1.92 bpm and the mean correlation coefficient is 0.9907. As an example, the trajectories of the modelled (without updating) and measured heart rate of cyclist FT is shown in Fig. 1.
Table 3 - The absolute error (ae), standard deviation (sd) and correlation coefficient (r) between the measured and modelled HR values for all subjects.
4- Discussion This study has demonstrated that the actual heart rate of a cyclist during the exercise can be predicted with high accuracy and reliability by a mathematical model based on the physical and physiological principles. For all subjects, the absolute error between modelled and measured HR is about 3 bpm with a mean standard deviation of 3.95 bpm. The mean correlation coefficient is 0.9686. Due to the fact that the change of the HR is caused mostly by the change of the actual work load, the parameter K1 describes the direct response of the cyclist’s HR to the work load and is considered as an important parameter besides K2. Therefore, the level of a cyclist’s performance can roughly be presented by the parameters K1 and K2. Among all subjects, the subject AK is the best of all with a PiAT of 268 W and HRiAT of 156 bpm. From the model parameters in Table 2, his model parameters confirm his strength with the largest K1 of 0.005 bpm/W and the smallest K2 of 0.9708 among all subjects. Our model predicts the HR values also based on the individual anaerobic threshold. Many studies have shown that the lactate thresholds may vary throughout the sports season (Denis et al. 1982, Keith et al. 1992, Tanaka et al. 1986). However the study of Lucía et al. (2000) has confirmed that the values of the target HR generally remain stable in professional cyclists during the course of the season. Therefore, in order to ensure the
432 The Engineering of Sport 7 - Vol. 1 precision of the prediction model, it may be meaningful to update the HRiAT and PiAT for each cyclist in the middle of the season. For prescribing the exercise intensity, power output may be the most direct indicator. However, it is variable during a whole training session or a competition. HR is more constant than power output, whereas it may change with the blood lactate concentration, hand (torso) position, temperature of the environment, training duration, altitude and so on (Jeukendrup and Van Diemen, 1998). In our model, most factors have been considered directly by the calculation of the HR value. For the special application in the model predictive control, the influences of these parameters are so small that they may not affect the precision of the prediction, because the HR will be updated by the measured values at each prediction step.
5- Conclusion In this study a dynamic heart rate prediction model was derived and the values of each model parameter were determined. This model predicts the HR of the cyclist during training or competition online based on physical and physiological principles. Comparison of the heart rate values predicted by the model with the values that were measured directly confirmed that the model was a valid and accurate representation of cyclist’s heart rate. This model facilitates the coach or cyclists to predict the HR tolerance range of the athletes before training or competition. It also enables the cyclist to plan his optimal training workload or to distribute his energy at a best way in competition. Furthermore, it establishes the basis model for the model predictive control to optimize the cycling training by a computer.
6- Acknowledgement This work was supported by the research centre Ambient Intelligence at the University of Kaiserslautern. The authors also express their gratitude to the cyclists of the triathlon club of 1. FC Kaiserslautern for their collaboration and support.
7- References [AJ1] Achten J. and Jeukendrup A., Heart rate monitoring: Applications and Limitations. Sports Medicine, 33(7): 517-538, 2003. [BG1] Bangsbo J., Gollnick P.D., Graham T.E., Juel C., Kiens B., Mizuno M. and Saltin B., Anaerobic energy production and O2 deficit-debt relationship during exhaustive exercise in humans. Journal of Physiology, 442: 539-559, 1990. [BM1] Bearden S.E. and Moffatt R.J., and heart rate kinetics in cycling: transitions from elevated baseline. Journal of Applied Physiology, 90(6): 2081-2087, June 2001. [BS1] Boulay M.R., Simoneau J.-A., Lortie G. and Bouchard C., Monitoring high-intensity endurance exercise with heart rate and thresholds. Medicine and Science in Sports and Exercise, 70: 125-132, 1997. [DF1] Denis C., Fouguet R., Poty P., Geyssant A. and Lacour J.R., Effect of 40 weeks of endurance training on the anaerobic threshold. International Journal of Sports Medicine, 3: 208-214, 1982.
The Engineering of Sport 7 - Vol. 1
433
[FP1] Faria E.W., Parker D.L. and Faria I.E., The Science of Cycling, Physiology and Training – Part1. Sports medicine, 35(4): 285-312, 2005. [G1] Gilman M.B., The use of heart rate to monitor the intensity of endurance training. International Journal of Sports Medicine, 21(2): 73-82, Feb 1996. [GA1] Grazzi G., Alfieri N., Borsetto C., Casoni I., Manfredini F., Mazzoni G. and Conconi F., The power output/heart rate relationship in cycling: test standardization and repeatability. Medicine and Science in Sports and Exercise, 31(10): 1478-1483, 1999. [JV1] Jeukendrup A. and Van Diemen A., Heart rate monitoring during training and competition in cyclists. Journal of Sports Sciences, 16(3): 91-99, May 1998. [K1] Kindermann W., Anaerobe Schwelle. Standards der Sportmedizin. Deutsche Zeitschrift für Sportmedizin, 55(6): 161-162, 2004. [KJ1] Keith S.P., Jacobs I. and Mclellan T.M., Adaptations to training at the individual anaerobic threshold. European Journal of Applied Physiology, 65: 316-323, 1992. [KK1] Kuipers, H. and Keizer, H.A. Overtraining in elite athletes. Sports Medicine, 6: 79-92, 1988. [KS1] Kinderman W., Simon G. and Keul J., The significance of the aerobic-anaerobic transition for the detection of work load intensities during endurance training. European Journal of Applied Physiology, 52: 25-34, 1979. [L1] Linnarsson D., Dynamics of pulmonary gas exchange and heart rate changes at start and end of exercise. ACTA Physiologica Scandinavica, 415:1-68, 1974. [LH1] Lucía A., Hoyos J., Pérez M. and Chicharro J.L., Heart rate and performance parameters in elite cyclists: a longitudinal study. Medicine and Science in Sports and Exercise, 33(10): 1777-1782, 2000. [LJ1] Le A.; Jaitner T. and Litz L., Sensor-based Training Optimization of a Cyclist Group. Proceedings of 7th International Conference on Hybrid Intelligent Systems, pp. 265-270, September 2007. [LL1] Le A., Litz L. and Jaitner T. A model predictive controller for sensor-based training optimization of a cyclist group. The 7th conference of the international sport engineering association. in print, 2008. [MB1] Mavrommataki E., Bogdanis G.C., Kaloupsis S. and Maridaki M., Recovery of power output and heart rate kinetics during repeated bouts of rowing exercise with different rest intervals. Journal of Sports Science and Medicine, (5): 115-122, 2006. [MS1] Mognoni P., Sirtori M.D., Lorenzelli F. and Ceretelli P., Physiological responses during prolonged exercise at the power output corresponding to the blood lactate threshold. European Journal of Applied Physiology, 60: 239-243, 1990. [ND1] Noble B. and Daniel J.W., Applied Linear Algebra, 3. Edition, Prentice Hall, London 1988. [SK1] Stegmann H., Kindermann W. and Schmabel A., Lactate kinetics and individual anaerobic threshold. International Journal of Sports Medicine, 2(3): 160-165, Aug 1981. [TW1] Tanaka K., Watanabe N. and Konishi Y., Longitudinal associations between anaerobic threshold and distance running performance. European Journal of Applied Physiology, 55: 248-252, 1986.
Stability Training and Measurement System for Sportsperson (P84) S.N. Omkar1*, D.K. Ganesh2
Topics: Measurement Systems, Prevention and Health. Abstract: Balance and stability are very important for everybody and especially for sportsperson who undergo extreme physical activities. Balance and stability exercises not only have a great impact on the performance of the sportsperson but also play a pivotal role in their rehabilitation. Therefore, it is very essential to have knowledge about a sportsperson’s balance and also to quantify the same. In this work, we propose a system consisting of a wobble board, with a gyro enhanced orientation sensor and a motion display for visual feedback to help the sportsperson improve their stability. The display unit gives in real time the orientation of the wobble board, which can help the sportsperson to apply necessary corrective forces to maintain neutral position. The system is compact and portable. We also quantify balance and stability using power spectral density. The sportsperson is made stand on the wobble board and the angular orientation of the wobble board is recorded for each 0.1 second interval. The signal is analized using discrete Fourier transforms. The power of this signal is related to the stability of the subject. This procedure is used to measure the balance and stability of an elite cricket team. Representative results are shown below: Table 1 represents power comparison of two subjects and Table 2 represents power comparison of left leg and right leg of one subject. This procedure can also be used in clinical practice to monitor improvement in stability dysfunction of sportsperson with injuries or other related problems undergoing rehabilitation. Subject Subject 1
Power (x104 ) 2.0759
Subject Subject 8
Power (x104 ) 3.5990
Table 1 - Both leg stance. Subject Subject 8
Power-Left leg (x 104) 7.3961
Power-Right leg (x 104) 1.7495
Table 2 - Single leg stance. 1, 2. Department of Aerospace Engineering, Indian Institute of Science, Bangalore-560 012, India; * Yoga consultant for Indian cricket team and National Cricket Academy - E-mail: omkar,[email protected]
436 The Engineering of Sport 7 - Vol. 1 Keywords: Balance, Stability, Vision, Power spectral density, Display unit, Inertial Measurement Unit.
1- Introduction To self sufficiently perform acts of living; avoid falls causing injuries, good stability and balance is mandatory. The tendency to excessively stiffen our muscles during imbalance interferes with the working of our whole body and the efficiency of our delicate balancing mechanism. Subjects who demonstrate poor balance have nearly seven times as many ankle sprains as subjects who have good balance (McGuine et al. 2000). Maintenance of standing balance requires inputs from visual, somatosensory, and vestibular sensory systems. Different sensory inputs provide the central nervous system with different frames of reference about an individual’s position relative to the base of support. The central nervous system needs to compare and integrate all the incoming inputs in order to maintain balance. The visual system has been shown to be important in the balance by means of various tests (Edwards 1946, Henriksson et al. 1966, Travis 1945). Visual impairments in adolescents lead to inadequate static balance and incommodious postural control (Bouchard D and Tetreault S 2000). In the current work, we are proposing a compact and simple system to estimate the standing balance using Inertial Measurement Unit and to assess the importance of visual sensory input for balance and stability of each subject. After extensive literature survey it can be safely said that this is a unique and simple methodology in which Inertial Measurement Unit, Wobble board and a real time visual display unit are used together, to estimate the stability of a subject and the importance of visual sensory input for balance. The data recorded is treated as a signal and its power/ stability value is estimated using Power spectral Density. Based on this test data, suitable exercise program or physical therapy can be recommended to improve the subject’s overall stability and balance.
2- Equipments The equipments required for this test are: Inertial Measurement Unit (IMU), a wobble board and a personal computer for Real-time visual feed back and data recording. The wobble board used is a wooden rectangular plate with semi-cylindrical bottom as shown in the fig.1 and 2. The individual stands on the upper portion which is flat, non-slippery surface. The bottom side, which goes on the ground, has a semi-cylinder along the board length. This allows the board a total of 60 degrees of movement, ± 30 degrees of movement along longitudinal axis. Uses: Wobble board is mainly used for exercise to enhance balance stability (Victoria Clark, Adrian Burden 2005).
The Engineering of Sport 7 - Vol. 1
Figure 1 - Wobble Board Top Surface. The IMU is attached to the top surface of the wobble board.
437
Figure 2 - Wobble Board Bottom Side, with semi-cylindrical surface.
Inertial Measurement Unit 3DM-GX1 combines three angular rate gyros with three orthogonal DC accelerometers, three orthogonal magnetometers, multiplexer, 16 bit A/D converter, and embedded microcontroller, to output its orientation in dynamic and static environments, operating over the full 360 degrees of angular motion on all three axes. This is mounted on the wobble board to measure the change in the wobble board angle with equal time intervals. Since the wobble board has only one degree of freedom – angle change in this direction alone is captured by Gyro Enhanced Orientation Sensor. It is connected to the computer and the readings are directly recorded into it. The fig.3 below shows the orientation sensor used. Uses: Unmanned Aerial/ Underwater Vehicles navigation, artificial horizon, etc.
Figure 3 - Inertial Measurement Unit (IMU).
3- Methods A Total of 14 male subjects participated for the test. The test was conducted to measure the balance and stability of an elite cricket team. The mean age of the participated subjects was 28±8 years. Detailed instructions of the test were given to all the subjects and each subject was given one practice run.This is a very unique methodology as this test combines the use of a wobble board with IMU and a visual display unit to estimate
438 The Engineering of Sport 7 - Vol. 1 the stability of a person. A personal computer was placed at a distance of one meter from the experimental area, exactly aligned to the line of sight. This is used to provide visual feed back to the subject. The orientation sensor is initially set to read zero degree. After the instructions, the subject is made to stand on the wobble board with hands straight down. The wobble board is first placed on flat ground and each subject is made to stand straight on the wobble board with feet/ foot parallel to smaller edge of the board. The neutral position for the wobble board is when the board’s flat surface is perfectly parallel with respect to the ground. The subject tries to balance the wobble board at a neutral position for two different experimental conditions, 1. With Real-time visual feed back 2. Without visual feed back Each subject is made to repeat the two experimental conditions for three different types of balance standpoint 1) With both the feet on the wobble board (fig.4). 2) With only left foot on the wobble board (fig.5). 3) With only right foot on the wobble board (fig.6).
Figure 4 - Subject standing with both the feet on the wobble board.
Figure 5 - Subject standing with only left foot on the wobble board.
Figure 6 - Subject standing with only right foot on the wobble board.
The change in the angle of the wobble board is measured at equal intervals of time by the IMU and directly recorded into the computer.
4- Measurement and data collection The IMU measured the angle of the wobble board along longitudinal axis with a sampling rate of 0.1 seconds. The sensor was connected to the computer through USB port. The data was recorded directly into the computer with the help of the software provided with the orientation sensor. The total time for which the data was recorded was 50 seconds for each subject in each experimental condition. Initial 10 seconds and the last 10 seconds are not considered during analysis because of the subject getting on and off the wobble board respectively. The fig.7 below shows the orientation of the wobble board from neutral axis with respect to time for a typical run.
The Engineering of Sport 7 - Vol. 1
439
Figure 7 - Typical run of a subject showing the orientation of the wobble board, with and without Real time Visual feedback.
5- Data Analysis The data obtained, i.e. the angular movement of the wobble board with respect to time can be treated as a signal. For any given random signal, power spectrum describes how the power of a signal or time series is distributed with frequency (Steven w. Smith 1997). Power spectrum is used for varied application from identifying noise in a given signal to estimating the systolic blood pressure to analysing the color characteristics of a particular light source. To analyse the data, discrete fourier transform is used. Discrete Fourier transforms are extremely useful because they reveal periodicities in input data as well as the relative strengths of any periodic components. The results of Fourier transform of the data is a complex vector output. The magnitude of the Fourier transform output squared is called the estimated power spectrum (Steven w. Smith 1997). The power spectrum versus frequency is plotted for each trial of each subject. The fig.8 below shows a typical power spectral density versus frequency for one subject with and without Realtime Visual feed back.
Figure 8 - shows a typical PSD versus frequency for one subject with and without Real-time Visual feedback.
440 The Engineering of Sport 7 - Vol. 1
6- Comparative analysis The power is calculated for each experimental condition to estimate the balance stability. Two different types of comparisons are done for the data collected. 1) The stability value of each subject for both the feet on the wobble board is compared between Real-time visual feedback and without visual feedback experimental conditions. Relatively the experimental condition with the lower power/stability value is said to be better balanced. 2) The stability value of each subject for single leg stance is compared between Realtime visual feedback and without visual feedback experimental conditions. The power/ stability value of left foot and right foot are compared with one another for each subject. This is done to identify the unstable limb of the two and to suggest suitable exercise to the subject in order to improve the subject’s over balance.
7- Results The three different types of balance standpoint tests were conducted for each subject and the respective Power/stability values were estimated. The Fig.9 below shows the comparison of the stability values for all the standpoints with and without visual feedback. From the Fig.9 it is apparent that the subjects have better stability with Real-time visual feed back as compared to stability without feedback. During the single leg stance the stability with Real-time visual feed back to a great degree is better than stability without visual feedback. The difference in the standing balance measurement of the left leg and right leg is also compared. The Table 1, below, gives typical power/ stability value of two subjects (subject 1 and 8) in both legs stance, with and without visual feedback. The Table 2, below, gives typical power/ stability value of one subject (subject 8) in single legs stance, with and without visual feedback.
Figure 9 - Typical power/ stability value of two subjects in both legs stance, with and without Real time visual feedback.
The Engineering of Sport 7 - Vol. 1
441
Table 1 - Typical power/ Stability values for two subjects in both legs stance, with visual feedback and without Real time visual feedback
Table 2 - Typical power/ Stability values for one subject in single leg stance, with visual feedback and without Real time visual feedback
8- Discussions and Conclusion This unique methodology can be used to measure the standing balance of a subject in lateral and longitudinal axes simultaneously, with the help of a wobble board having a semi spherical bottom. It can be clearly seen from the test that majority of the subjects have shown better balance with Real-time visual feedback in all the three balance standpoints. This helps in estimating the importance of visual sensory input for stability and balance of each subject. These test results can be used as a reference to help the subjects improve their balance by suggesting suitable exercises. This methodology is extremely useful in the world of sports, in order to prevent injuries to the sports person during the course of the game, by providing suitable balance training (McGuine et al. 2000) based on these test results. Subjects who demonstrate poor balance have nearly seven times as many ankle sprains as subjects who have good balance (McGuine et al. 2000). The results obtained from this methodology can be used for various applications, a few of them worth mentioning here are: to compare the stability of one subject with another, for example, the stability of case with leg injuries or related problems can be compared to a healthy subject’s stability value. From the results we can estimate the difference in stability of each limb; this procedure can be used in clinical practice to monitor improvement in stability dysfunction of patients with injuries or other related problems undergoing rehabilitation. Balance and walk impairments in elderly citizens has increased the risk of falls, which constitutes to a majority of accidental casualty and injury-related visits to hospitals (Tinetti et al. 1988). As a result, it is crucial to look into balance instability in order to identify elderly citizens at risk of a fall, and to reduce balance impairment. Standing balance measurement will help in clinical practice to estimate a patient’s current stability and suggest suitable physical therapy to improve the patient’s stability. This methodology can also be used as a reference to compare the progress made by a patient during the course of exercise from time to time.
9-References [BT1] Bouchard D, Tetreault S. The motor development of sighted children and children with moderate low vision aged 8–13. J Vis Impair Blin 2000;94:564–73.
442 The Engineering of Sport 7 - Vol. 1 [E1] Edwards AS. Body-sway and vision. J Exp Psychol 1946;36:526– 35 [HJ1]Henriksson NG, Johansson G, Olsson LG, Ostlund H. Electric analysis of the Romberg test. Acta Otolaryngol (Stockh)1966;224(Suppl):272–9 [MG1] McGuine, TA, Greene JJ, Best T, Leverson G. (2000) Balance as a Predictor of Ankle Injuries in High School Basketball Players. Clinical J Sports Med, 10:239-244. [S1]Steven w. Smith, (1997) The scientist and engineer’s guide to digital signal processing. Newnes: Demystifying Technology Series [T1] Travis RC. An experimental analysis of dynamic and static equilibrium. J Exp Psychol 1945;35:216–34 [TS1]Tinetti me, speechley m, Ginter sf. Risk factors for falls among elderly persons living in the community. N engl j med 1988; 319: 1701–7. [VA1] Victoria m. Clark, Adrian m. Burden. A 4-week wobble board exercise program improved muscle onset latency and perceived stability in individuals with a functionally unstable ankle. Physical therapy in sport 6 (2005) 181–187.
SRM Torque Analysis of Standing Starts in Track Cycling (P85) Paul Barratt1
Topics: Performance Sports, Measurement Systems, Bicycle. Abstract: The SRM (Schoberer Rad Messtechnik, Welldorf, Germany) power monitoring system has been used extensively in applied field based studies to provide an accurate measurement of cycling power (Gardner et al. 2005). The SRM system consists of a PowerMeter (instrumented crank), a PowerControl (data logger and onboard data display), and a sensor cable (linking data transfer from crank to the onboard powercontrol). One limitation of the SRM system is an inability to attain power outputs until one entire crank revolution is achieved. A crucial element of track cycling time trial events is the standing start in which the athlete is generating maximum torque at low cadence. In this performance setting the cyclist accelerates the bicycle from rest and consequently initial power data is not recorded by the standard SRM system. A modification to the SRM system is described which allows for collection of instantaneous crank torque. The modification consists of an electronic de-modulator developed by SRM and a multi-channel datalogger (DL16CAN, Tellert, Germany) connected in series between the PowerMeter and the PowerControl. The pulse width modulated signal from the PowerMeter is de-modulated into a cadence voltage signal and a torque frequency signal. The signals are synchronised and recorded separately on a time base. The modification allows crank torque to be acquired at 200Hz. This equipment has opened up a new avenue of analysis in a previously under-researched and ultimately performance impacting area of track cycling. Keywords: Cycling, SRM, Standing Start.
1- Introduction The standing start it is a key feature of track cycling. It occurs in individual and team time-trial events where the aim is to complete a set distance in the shortest time possible (Craig and Norton, 2001). During the standing start the bike is released from a stationary position and the cyclist attempts to accelerate the bike up to full speed as quickly 1. English Institute of Sport, Sportcity, Gate 13, Rowsley Street, Manchester, M20 4NZ, United Kingdom E-mail: [email protected]
444 The Engineering of Sport 7 - Vol. 1 as possible. The acceleration phase will typically last up to one lap, with maximum torque generated by the cyclist in the first few metres of the effort. It is a feature of eight of the seventeen World Championship disciplines and four of the twelve Olympic disciplines (UCI Regulations 1.3.025). The SRM power measuring system is used extensively in applied research where an accurate measure of cycling power is required (Gardner et al. 2005). Unfortunately, the standard SRM system has limitations as a method of analysing standing starts. A key limitation is the relatively low sample rate. Due to the SRM requiring a cadence measurement to calculate power, data is not attained until one full crank revolution is completed, and even then power is only acquired as an average per revolution. This makes it suitable for long road stage races, however this is a very different performance environment to track sprinting. The ability to sample continuously at a higher rate could prove advantageous when attempting to understand the components of a fast track start.
2- SRM Power Measuring System The standard SRM system comprises of an instrumented crank (“PowerMeter”), a data logger and display unit (“PowerControl”) and a sensor cable to transmit the data between the two. Depending upon the version of the PowerMeter, it will contain 2, 4, or 8 (amateur, professional or science respectively) strain gauges in a wheatstone bridge configuration. The strain circuit produces a voltage when torque is applied about the bottom bracket, the voltage is converted into a frequency by a voltage-to-frequency converter within the PowerMeter. The angular velocity of the crank is termed the “cadence” or “pedalling rate” and is normally reported in revolutions per minute. This is measured by the PowerMeter by way of a reed switch. Once per revolution the reed switch is tripped by a magnet located on the bike frame. Torque is acquired at a sampling rate of 200Hz, cadence is acquired once per revolution. The torque and cadence signals are frequency-modulated and transmitted to a sensor located on the bike frame via a radio signal. This signal is then transmitted to the PowerControl by the sensor cable where is it de-modulated back into 2 signals. A processor within the PowerControl multiplies the average torque and cadence per revolution together to produce a power value. Power is calculated as an average per revolution. Figure 1 shows a functional diagram of the SRM system from the SRM manual.
Figure 1 - SRM Operating Principle (image from SRM Manual, www.srm.de).
The Engineering of Sport 7 - Vol. 1
445
3- SRM Torque Analysis The standard SRM system generates an average value of power for every pedal revolution. This limitation is due to the system only acquiring a cadence value once per revolution. The raw torque data is acquired at 200Hz by the PowerMeter, however this data is discarded by the PowerControl once an average torque for a pedal revolution has been calculated. SRM produce a “torque box” which can be used in conjunction with the PowerMeter to retain the instantaneous torque signal normally discarded by the PowerControl. The torque box is designed for ergometer use, but it can be placed on a bike, in conjunction with a separate data logger, for on-bike instantaneous torque measurement. Both the torque box and data logger can be located under the saddle of a standard track bicycle.
Figure 2 - Location of the torque box and data logger under the saddle.
The torque box receives the signal from the PowerMeter via the sensor cable, and demodulates the combined torque and cadence signal (Hz) to generate separate torque (Hz) and cadence signals (V). A multi-channel data logger (DL16CAN, Tellert, Germany) receives the output from the torque box, acquires the torque data at 200Hz and cadence once per revolution. This enables instantaneous torque to be acquired at 200Hz, as well as cadence and power to be acquired as average values per revolution. From the instantaneous torque data, it is possible to view the torque profile of any given pedal revolution and to acquire values of peak torque, rate of torque production, and mechanical impulse. Figure 2 shows the torque box and data logger located under the saddle of a track bicycle.
446 The Engineering of Sport 7 - Vol. 1 On-bike crank torque has previously been collected during road cycling (Bertucci et al. 2005; Bertucci et al. 2007). However, for these investigations the PowerControl on the bike has been connected via a cable to a computer in a following car, which is obviously not a suitable option for track cycling.
4- Data Processing The data can be easily downloaded from the data logger to a laptop computer via a USB cable. The raw data is collected in the form of a torque frequency signal (Hz) and a cadence voltage signal (V), both synchronised on a time base. TEMES software (supplied with the data logger) is used to trim the signals to an individual effort, and then export the data as a .txt file. Figure 3 shows the raw data for a sample 60 metre standing start effort in the TEMES software.
Figure 3 - Sample crank torque data from a 60 metre standing start effort (TEMES software).
The .txt file is imported into a Microsoft Excel spreadsheet. The zero offset (the frequency signal at zero load) is removed from the raw signal and, using the calibration slope of the powermeter, converted into a torque value (Nm). A Microsoft Excel MACRO written in Visual Basic is used to automatically calculate variables in the data. The MACRO identifies individual pedal revolutions within the effort and generates values of peak torque (Nm), rate of torque production (Nm/s) and mechanical impulse (Nm.s) for each revolution. The MACRO also uses the average torque per revolution to calculate average power per revolution (W) as in a conventional SRM system.
5- Calibration considerations for SRM Torque Analysis The PowerMeter converts the raw frequency value from the strain gauges into a torque value (Nm) by way of a calibration slope. The slope is generated during a calibration procedure in which the PowerMeter is removed from the bike, fixed in a vice and a mass is hung from the crank to produce a known torque. A frequency value is collected from
The Engineering of Sport 7 - Vol. 1
447
the PowerControl, and this is used to calculate a calibration slope. (For a full description of the static calibration procedure, see Wooles et al. 2005). Figure 4 gives a graphical description of the calibration slope and zero offset, taken from the SRM manual.
Figure 4 - Description of the calibration slope and zero offset of the PowerMeter (image from SRM Manual, www.srm.de).
Using instantaneous torque analysis creates the need to be aware of any possible variation in calibration slope at different positions of the crank. This is perhaps not necessary for the standard use of the SRM system, as the data is recorded as an average per revolution. However, torque analysis allows for instantaneous torque values to be collected in different positions around the crank, and as such it is necessary to ensure the accuracy of these values. To assess the variation in slope around the crank, eight PowerMeters were randomly selected and statically calibrated in 4 positions around the crank (0deg., 90deg., 180deg., 270deg.). Calibration slopes were generated for the 4 positions and a mean and standard deviation produced for each PowerMeter. For a standard calibration the mean value would be used to generate the actual value of the slope. The experimental data is shown in figure 5.
Figure 5 - Variation of slope around the crank for 9 randomly selected PowerMeter’s.
The mean standard deviation in slope for the group was 0.065Hz/Nm. For a given PowerMeter slope of 16.00 Hz/Nm this represents a variation of less than 0.5% around
448 The Engineering of Sport 7 - Vol. 1 the crank, which is highly satisfactory for a measuring instrument. For a torque input of 100Nm this is a variation of between 99.60Nm and 100.41Nm around the PowerMeter. The other known method of calibration used for an SRM PowerMeter is a dynamic calibration. For this method the PowerMeter is spun at a known cadence whilst a dynamic brake is applied to the PowerMeter. This method produces a rolling average of power, and it is not possible to attain a calibration slope at specific positions around the crank. For this reason it may be argued that it is an unsuitable method of calibration for a PowerMeter on which torque analysis will be used.
6- Discussion / Conclusion The major benefit of this technology is the size and weight of the measuring system. Although currently too large to be considered for competition use, it is very suitable for the training environment. It allows crank torque to be routinely monitored during practice standing start efforts, and to assess the change in key performance indicators over time. The innovation is based on an independently validated measuring system. Although the manner in which data is acquired is modified, the results of the calibration trial would confirm that confidence can be placed in the crank torque data. One area of future development may be to include instantaneous angular displacement, and therefore instantaneous angular velocity, into the data collection. This would allow an angular velocity profile to be generated, similar to the torque profile, and ultimately allow measurement of instantaneous power. In the 2007 UCI World Track Cycling Championships, the margin between first and second in the Men’s Team Sprint was 0.002 seconds. Technological innovations such as the modification described in this paper, allow performance to be analysed at a greater depth than previously possible. With such fine margins between winning and losing in modern sport, this ability could be the key to success at world level.
7- References [GS1] Gardner AS, Stephens S, Martin DT, Lawton E, Lee H, Jenkins D. (2004) Accuracy of srm and power tap power monitoring systems for bicycling. Med Sci Sports Exerc 36:1252–1258 [CN1] Craig NP, Norton KI (2001) Characteristics of Track Cycling. Sports Medicine 31(7): 45768 [WR1] A.L. Wooles, A.J. Robinson, P.S. Keen. (2005) A static method for obtaining a calibration factor for SRM bicycle power cranks. Sports Engineering 8;137-144. [BG1] Bertucci W, Grappe F, Girard A, Betik A, Rouillon JD. (2005) Effects on the crank torque profile when changing pedalling cadence in level ground and uphill road cycling. J Biomech. May;38(5):1003-10. [BG2] Bertucci W, Grappe F, Groslambert A. (2007) Laboratory versus outdoor cycling conditions: differences in pedaling biomechanics. J Appl Biomech. May;23(2):87-92.
Aerodynamic Study of Ski Jumping Flight Based on High-Speed Video Image (P86) Masahide Murakami1, Nobuyuki Hirai1, Kazuya Seo2, Yuji Ohgi3
Topics: Ski & other Winter Sports. Abstract: Aerodynamic forces were derived from the data analysis of high-speed video image of ski jumping flight, which was taken at Hakuba Ski Jumping Stadium in Japan. Initial 40 m parts of flights were recorded by a fixed high-speed video camera at a frame rate of 250 frames/s. The primary target of the present study is to extract the aerodynamic force data during real jumping flights, in particular, in the initial phase of a quasi-steady flight between 1 and 2 sec. after take-off. It is concluded the present aerodynamic force data derived from the image analysis are found to be in fair agreement with existing wind tunnel data for the cases of jumping flights in the quasi-steady flight phase. The aerodynamic data in the cases of relatively large forward leaning angle, which is typical in the initial phase of a jumping, could also be obtained. In addition, data of an Olympic team jumper and a high school jumper are compared to discuss the detail of their flight skills. Keywords: Ski Jumping Flight, Video Image Analysis, Lift, Drag.
1- Introduction The database for the aerodynamic forces of ski jumping is now available (Seo et al. 20041), which was constructed on the basis of the wind tunnel test data measured using a real sized model. The data was used in the jumping flight optimization study (Seo et al. 2004-2). However in the database, data are lacked in the range of large forward leaning angle, that is the ski-body angle, due to some technical problem that the blockage factor in the test section of wind tunnel becomes too large as well as a difficulty in the wire support method of the model in the test section. The jumper posture with a large forward leaning angle is typical in the initial phase immediately after a take-off. Consequently, in the above optimization study the analysis was restricted to flights after about 0.5 sec. from take-off. Therefore, the measurement of the aerodynamic force 1. Graduate School of Systems and Information Engineering, University of Tsukuba, Tsukuba Japan E-mail : [email protected] 2. Faculty of Education, Art and Science, Yamagata University, Yamagata Japan - E-mail: [email protected] 3. Graduate School of Media and Governance, Keio University, Fujisawa Japan - E-mail: [email protected]
450 The Engineering of Sport 7 - Vol. 1 during initial 1 sec. of flight is the very target of this study. The real time derivation of the aerodynamic force data would be used in the computer-aided ski jumping coach system based on visual image. In this system, a coach of ski jumping can give an advise to a jumper immediately after his flight based on the aerodynamic force data as well as on the video image. In this study as a part of a series of researches, we first focused on the improvement of the accuracy in the aerodynamic force computation. Numerical filtering and regression analysis are introduced in the computation (Murakami et al. 2007). Second, the number of analyzed sample data was increased so that ensemble average can be taken to improve the accuracy. And some discussion was developed with the comparison of the data between an expert Olympic team and a high school jumpers.
2- Acquisition of video image In the field measurements in Hakuba, high-speed video images of jumping flights were taken. And an accelerometer and a gyro-sensor were installed on the back of a jumper to measure the acceleration and the angular velocity of the trunk around the horizontal axis. The data measured with these sensors are made public at this conference (Ohgi et al. 2008). The flights of large-hill class jumping were selected as examination objects. The camera was set on the top of the coach tower of the normal-hill jumping that stands at the immediate side of the take-off point as shown in Figure 1. The location is about 15 m downward along the large-hill landing slope from the take-off point and 60 m away from the large-hill landing slope. The field of view of the camera with a wide-angle lens with a focal length of 28mm covers upper 40m range of the landing slope. This roughly corresponds to a flight for initial 2 sec. The high-speed digital video camera (Photron Fastcam-X 1280PCI) was mostly operated in the mode of 1024512 pixels at 250 frames/sec. The subjects in these measurements were members of Japanese Olympic team jumpers and a number of high school jumpers. The effect of wind was ignored, because no wind data were provided because the measurements were done during usual practices. As a matter of fact, wind was weak to moderate during the experiments The
Figure 1 - Sketch of the video installation place in Hakuba Ski Jumping Stadium.
Figure 2 - Coordinate system of the jumping hill and the definition of angles characterizing the posture of a jumper.
The Engineering of Sport 7 - Vol. 1
451
position of the centre of gravity of a jumper is expressed in the coordinate system indicated in Figure 2, where the coordinate origin is the take-off point, and the x-axis is taken horizontally and the y-axis vertically downward.
3- Results and discussion 3.1 Image data analysis The centre of mass of a flying jumper must be pursued in images in order to derive the velocity and acceleration data. In this study, the centre of mass was calculated as a weighted average of the positions of the waist, head and toe. The specific weights of body segments of a jumper were calculated by referring to the paper (Chandler 1975). In addition to these data, the mass of ski is also taken into account, and the mass of a jumper is assumed to be 70 kg. The image tracking for these selected positions was mostly accomplished automatically with the aid of image analysis software, but partly with some
Figure 3 - Trajectory data in a x-y plane of the Olympic team jumper and the high school jumper.
Figure 4 - Time variations of the x- and y-velocities. The simple numerical differentiation results by the marks, the low-pass filtered result by the dotted lines and the regression curves by solid lines are shown for comparison.
Figure 5 - The regression curves for the time variations of the x- and y-accelerations.
452 The Engineering of Sport 7 - Vol. 1 manual assist for cases where the positions were hard to be distinguished because they seemed as if they were blended into in the background. In these measurements, the single fixed video camera with a wide-angle lens was used instead of adopting a three-dimensional camera system. Some distortion being not avoided in the images taken with a camera with a wide-angle lens must be corrected in the image digitizing stage. A simple correction by a quadratic formula was applied to the original time series position data. Shown in Fig. 3 are the trajectories of two jumpers, an Olympic team jumper and a high school jumper drawn in an x-y plane, which are the loci of the centre of mass of jumpers. In this paper, the data of the Olympic team jumper are mostly treated, but those of the high school jumper are also presented in some figures for comparison. Both of the trajectories, in the appearance, look almost same parabolic curves. However we see below the two flights are much different. The flight distances were 124m for the Olympic team jumper and 117m for the high school jumper for the almost same initial speed, 25 m/s. For further data analysis, the first and the second order time derivatives of the trajectory data should be computed for the velocity and the acceleration from which the aerodynamic forces are derived. Rather higher order numerical differentiation scheme as below is adopted for the velocity and the acceleration computations. (1) It is, however, seen that some data smoothing procedures are further required for the significant derivation of the velocity and the acceleration from the time series data of the centre of mass. The low pass filtering with a cut-off frequency of 3 Hz and the weighted least square regression with a higher order polynomial are introduced for this purpose. The x- and y-velocity data, u and v, for the Olympic team jumper are shown in Figure 4. In this figure, the simple numerical differentiation data, the low-pass filtered results and the regression curve are shown, respectively by marks, by broken lines and by solid lines. It should be noted that data for initial 0.2 sec. and later than 1.7 sec. couldn’t be obtained because of the prohibited application region of data filtering. Gradual deceleration in u due to the air drag and rather rapid acceleration in v due to gravity are clearly seen.
Figure 6 - Time variations of the absolute jumper velocity U. Two results of the Olympic team jumper and the high school jumper are compared.
The Engineering of Sport 7 - Vol. 1
453
Some data smoothing is also inevitable for the derivation of the acceleration. The results are shown in Figure 5 in which the data are processed by the low-pass filtering and then regression curve fitting. The x-acceleration, ax, primarily results from the air drag. In the initial phase, ax slightly decreases due to the forward leaning action of jumper. The later small increase is the result of aerodynamic stall due to large angle of attack. It is seen that ay is smaller than the acceleration due to gravity from the initial phase of flight. Even immediately after the take-off the lift affects the flight. Slight decreases after 1 sec. is caused by the increase in the lift due to the increase of the angle of attack and deeper forward leaning of the body. The time variation of the absolute velocity U defined below is shown in Fig. 6 for two jumpers. (2) In the case of the Olympic team jumper, it decreases for the initial 0.5 sec., and then increases. In the initial phase, the contribution of u that decreases with time due to the drag dominates in U. And then, U increases due to the increase of v as the result of the falling motion. This type of time variation in U also found in the simulation result (Seo et al. 2004-1) is typical for such expert jumpers as the Olympic team jumpers. However, in the case of the high school jumper, it never increased through the observation period. It is seen this is caused by larger drag as a result of too large upward angle of ski resulting in too large angle of attack as discussed later. The time variations of several angles characterizing a jumper posture are shown in Figures 7 and 8. The plots of the angle of attack of ski, , and the flight angle, , are shown in Figure 7, and those of the forward leaning angle, , and the ski angle with respect to the horizontal line, - , are shown in Figure 8 where the data of the high school jumper are also added for comparison. The flight angle that is measured downward from the horizontal line is about 10 deg. immediately after the take-off, as the angle of the take-off point (Kante) is 11 degrees downward. The angle increases gradually with the lapse of time, which results from the increase in the falling component, v due to gravity. The angle of attack, rapidly increases until about 0.7 sec. as a result of jumper’s operation to lift the tip of ski in order to gain the lift. After this period, slowly increases responding to the increase in to reach the stall angle. The jumper proceeds into a quasi-steady posture in 0.7 sec. We see from the comparison between the Olympic team jumper and the high school jumper in the quasi-steady phase presented in Figure 8 that the time variations of the forward leaning angle are almost similar between the two jumpers. However, ski angle, - , of the high school jumper is always larger than that of the Olympic team jumper by about 10 degrees. This is evident from the original video images. This is the very reason why the drag of the high school jumper is larger than that of the Olympic team jumper, in particular both in the very initial phase and after 1.1 sec. when aerodynamic stall occurred, to result in smaller flight velocity leading to smaller flight distance as discussed above.
454 The Engineering of Sport 7 - Vol. 1
3.2 Aerodynamic force data The aerodynamic forces, that is the lift L that is the normal component of the aerodynamic force with respect to the jumping trajectory and the drag D that is the tangential component, are computed in terms of the acceleration data, ax and ay and the flight angle . It should be noted that the effect of gravity g is subtracted in the calculation.
Figure 7 - The time variations of the angle of attack of ski, , and the flight angle, .
Figure 8 - The time variations of the forward leaning angle, , and the ski angle with respect to the horizontal line, - , for the Olympic team jumper and the high school jumper.
Figure 9 - Variations of the lift area SL and the drag area SD during a flight. 9-a: The variations of SL and SD. as a function of time. The SD data of the high school jumper is also presented for comparison. 9-b: Variations of SL and SD as a function of angle of attack .
(3) (4) where m is the mass of a jumper. For the sake of comparison of the results with the existing wind tunnel data (Seo et al. 2004-1), the lift area SL and the drag area SD that were introduced by Tani, et al. (Tani and Mitsuishi 1951, Tani and Iuchi 1971) are used instead of the lift coefficient CL and the drag coefficient CD
The Engineering of Sport 7 - Vol. 1
455
(6) where is the air density. The time variations of SL and SD are shown Fig. 9-a, where the SD data of the high school jumper is also drawn. Both SL and SD increase with time for initial 1 sec., and then SL decreases while SD still increases. The result of SD in the later period is considered to result from the aerodynamic stall. The time when the stall appeared is in fair agreement with the time for to exceed 30 degrees. This fact becomes clear from the SL and SD data plotted against as shown in Fig. 9-b. It may be generally concluded that the aerodynamic stall occurs when the angle of attack reaches about 30 to 40 deg. This result is consistent with the wind tunnel experimental data of Seo et al. (2004-1). It is evident SD of the high school jumper is larger than that of the Olympic team jumper through the jump flight as suggested from the result of large ski angle. Shown in Figure 10 may be considered as a summary of the measurement results, in Figure 10-a for SL and in Figure 10-b for SD. The ensemble average of the SL and SD data over 17 jumping samples are plotted by taking and as parameters, of which values are given by dividing into some ranges. The solid lines are the regression curves for the present data, and the broken lines are the data of the previous wind tunnel test (Seo et al. 2004-1) that lack in the region of large . In the range of around 20 to 40 deg. of which postures typically appear in early quasi-steady flight phase, the present SL data are in good agreement with the wind tunnel data for larger than 20 deg. in which region flight data are available. On the other hand, it seems the present SL data are slightly larger than the extrapolated values to the wind tunnel data in the larger range. This is because the jumpers fly in the posture in which their upper trunk are bent nearly horizontally in the waist while in the wind tunnel experiment the whole body of the model was kept straight. However, the discussion on the data for near 90 does not seem conclusive as the wind tunnel experimental data do not exist in this region.
Figure 10 - The ensemble average data of the SL and SD over 17 jumping samples plotted against the angle of attack by taking the forward leaning angle as a parameter. The solid lines are the regression curves for the present data, and the broken lines are the data of the previous wind tunnel test (Seo et al. 2004-1). 10-a : SL. 10-b : SD.
456 The Engineering of Sport 7 - Vol. 1 The present SD data is in fair agreement with the wind tunnel experimental data in quasi-steady flight phase when is around 20 deg. However, in the cases of large , the present SD does not increase to the wind tunnel data, and does not increase so that
may increase. The reason for this is also attributed to the difference in the posture of jumpers as in the case of SL. The jumper posture in which his upper trunk is bent leads to smaller air drag in particular in the case of near 90 degrees corresponding to the very initial phase of flight.
4- Conclusions The conclusions of this study are summarized as follows: 1. It is demonstrated that the aerodynamic force data for ski jumping flight can be extracted from high-speed video images. The data are found to satisfactorily agree with the existing wind tunnel experiment data for the jumping flight in quasi-steady flight phase. In the very initial phase of flight, the aerodynamic data is not conclusive and further data analysis is still required. 2. The absolute jumper velocity decreases in response to the reduction in u due to the drag in the initial 0.5 sec, and then increases due to the increase in v mostly caused by gravity effect. 3. The skill of a jumper can be evaluated on the basis of the aerodynamic force data derived from high-speed image.
5- References [SW1] Seo K., Watanabe I. and Murakami M. Aerodynamic force data for a V-style ski jumping flight. Sports Engineering 7-1: 31-39, 2004. [SM1] Seo K., Murakami M. and Yoshida K. Optimal flight technique for V-style ski jumping. Sport Engineering 7-2: 97-104, 2004. [MH1] Murakami M., Hirai N., Seo K. and Ohgi Y. Aerodynamic forces computation from highspeed video image of ski jumping flight. Presented at Asia-Pacific Congress on Sports Technology 2007 Sept. 2007 [OS1] Ohgi Y., Seo K., Hirai N. and Murakami M. Aerodynamic study of ski jumping flight based on inertia sensors. Presented at this Conference, 2008. [C1] Chandler, R. F. Investigation of inertial properties of the human body. Technical Report AMRL-74-137, Wright Patterson Air Force Base, 1975. [TM1] Tani I. and Mitsuishi, T. Aerodynamics of ski jumping. Science (Japanese edition), 117-52, 1951. (in Japanese) [TI1] Tani I. and Iuchi M. Flight-mechanical investigation of ski jumping. Scientific study of skiing in Japan. Hitachi, 35-52, 1971.
The Role of Materials and Construction on Hockey Ball Performance (P88) Dan Ranga, James Cornish, Martin Strangwood1
Topics: Hockey, materials, testing. Abstract: Three commercial hockey balls, two with mixed cork, wool and polymer cores and one which was hollow, were tested dynamically using ball rebound and ‘hardness’ tests. The load / deformation behaviour of selected ball types was determined at low deformation rates (0.02 – 2 mm s-1) using a Zwick Z100 universal tester and at high deformation rates (16 – 24 m s-1) using a gas cannon and steel target. In addition, cylindrical samples taken from the core were compression tested (using the same slow deformation rates, also on the Zwick Z100) to determine their tangential modulus variation with strain and deformation rate as well as their viscoelastic energy losses. The trends shown during impact tests were consistent with those in the low strain rate compression tests and these correlated with the materials properties and dimensions. Lower impact speed behaviour was associated with deformation being dominated by the cover, whilst higher deformation rates gave significant differences in ball behaviour due to the composition and dimensions of their cores. These differences were, however, demonstrated in the core compression testing indicating that the core materials play a greater role as the speed of the shots increases and deformation spreads beyond the cover. The qualitative relationships between ball construction at deformation behaviour at varying speeds is also summarised. Keywords: Hockey, sports balls, impact behaviour, stiffness, energy balance.
1- Introduction Most hockey balls are an example of solid ball construction being composed of mixtures of cork, wool and elastomer with a polyurethane (PU) or polyvinyl chloride (PVC) cover (McHutchon, 2006) and so, as for other solid construction balls, would be expected to demonstrate strain and strain rate dependent properties. The strain and strain rate will both be determined by the speed of a hockey shot, which will have a number of effects. As the shot speed increases then deformation will extend through the PU / PVC cover to involve larger volumes of the core. In addition, the strength and modulus of the 1. Sports Materials Research Group, The University of Birmingham, Department of Metallurgy and Materials, Elms Road, Edgbaston, Birmingham, B15 2TT, UK - Email: [email protected]
458 The Engineering of Sport 7 - Vol. 1 viscoelastic material will increase (Gibson et al. 1981), which, for the same stress level would reduce the strain induced in the core material. However, increasing shot speed also increases the stress applied, which has a greater effect than the increased modulus so that the overall strain increases. The result of increasing speed would be to alter the degree of energy loss during the impact and so can affect the speed of the ball of the stick or the bounce of the ball on the pitch. Balls used in competitive international hockey matches need to conform to a number of requirements, from durability and consistency to playing performance. These are contained within the 10 standard tests stipulated by the FIH (Federation International d’Hockey). Of these 10 criteria (FIH, 1999), only two of the tests measure the dynamic response of the ball, but are conducted at low impact speeds (4.66 - 6.26 m s-1) relative to those experienced by the ball during play, which have been recorded at up to approximately 49 m s-1 (Rai et al., 2002). Thus, the behaviour of the ball during play may not be accurately reflected in the standard tests. The possibility also exists that balls, which satisfy the standard criteria might exhibit significantly different behaviour at the higher speeds experienced during play. This paper reports a number of tests carried out on three commercial hockey ball types over a range of deformation rates representative of standard testing and of hockey play in order to determine the significance of strain rate effects encountered during hockey.
2- Experimental Four samples each of three commercial hockey balls were used for this study. Ball compositions, materials, and dimensions are shown in Table 1. Table 1 - Ball types investigated in this study.
Low deformation rate tests were carried out on two samples of each ball type using a Zwick Z100 universal tester at crosshead speeds of 0.02, 0.2 and 2 mm/s to deformations of 3.6, 7.2 and 10.8 mm. Samples 5 ± 1.09 mm diameter and 10 ± 1.75 mm length extracted from the cores of each ball (cover for ball G) were uniaxially compressed at the same crosshead speeds to strains of 5, 10 and 15 %. The resulting load / deflection or stress / strain curves were analysed to give stiffness (ball) or modulus (cylinder) values as a function of deformation / strain as well as hysteresis (energy loss on each loading and unloading cycle). The static properties were also determined from Shore D hardness values taken across a radius from two examples of each ball type, sectioned along a diameter, using a Mitutoyo Shore D Durometer with a 30º steel cone, and a tip radius of 0.1 mm. The sectioned balls were also used to characterise ball diameter, composition, cover dimensions, and layer dimensions if they were present; distance measurements were made using a digital calliper (Mitutoyo Series 500, accuracy ± 0.02 mm) and
The Engineering of Sport 7 - Vol. 1
459
macroscopic image analysis software (Zeiss Axiovision 4) on images captured using a Zeiss MRc5 camera. The FIH standard ball rebound tests were carried out on 4 samples of each ball type and measured the rebound height of the ball when dropped from 2 m onto a steel plate, therefore giving a theoretical impact speed of 6.26 m s-1. To conform with FIH criteria, the rebound height must be 500 - 650 mm, corresponding to a CoR (Coefficient of Restitution) of 0.50 – 0.61 at 6.26 m s-1. The other dynamic FIH standard test (ball ‘hardness’) was also carried out by impacting each sample of the ball with a 5 kg mass at 4.66 m s-1 and measuring the peak deceleration of the mass in g; to conform to FIH criteria, a peak deceleration of 130 – 220 g should be observed. High deformation rate dynamic impact testing was conducted on two balls of each type by firing them horizontally at a steel target using an ADC Supercannon 2000, at speeds of 16 – 24 m s-1. Ball impacts were recorded with a high-speed digital imaging system (Kodak Ektapro 4540MX) operating at 18,000 fps with 256 x 64 pixel resolution. Images were subsequently analysed to determine ball impact speed, rebound speed, deformation and deformation rate.
3- Results and Discussion Images taken from sectioned samples of the three ball types are shown in Figure 1 and confirm the internal constructions noted in Table 1. From these images the likely viscoelastic response of the balls can be predicted based on the amount of elastomeric polymers present in their construction. Elastomeric polymers have been shown to demonstrate a greater viscoelastic response than natural cork (Gibson et al., 1981; Qi & Boyce, 2005), causing their behaviour to be more dependent on strain and strain rate, therefore the viscoelastic response of the balls is expected to decrease in the order ball G > ball I > ball H.
Figure 1 - Diametric sections of (a) ball G; (b) ball H; and (c) ball I.
The hollow nature of ball G means that the thicker cover has a slightly higher Shore D hardness, Figure 2, in order to provide sufficient strength and rigidity to the ball; Shore D hardness can be linked to polymer modulus (Qi et al., 2003). The Shore D values recorded through the cover of ball G were constant within experimental error, Figure 2. Both of the filled balls (H and I) had a softer, thinner cover than G, which lay over a soft cork outer core and then a harder inner core. This layered structure would tend to concentrate strain in the outer core layer, which had similar hardness and width in both
460 The Engineering of Sport 7 - Vol. 1 balls. The inner core of ball I was softer than that in ball H, due probably to the elastomer incorporated into the cork, Figure 1 (c). The stiffness – deflection plots for these balls, Figure 3, show that ball G has a lower stiffness than balls H and I and that it also gives a lower increase in stiffness with increasing deformation. Balls H and I showed similar responses, both in terms of initial stiffness and the increase in stiffness as deformation increased. Within experimental error balls H and I showed the same stiffening behaviour with deformation, which would be consistent with strain concentration in the similar stiffness outer core region.
Figure 2 - Radial variation in Shore D hardness for balls G, H and I.
Figure 3 - Variation in whole ball stiffness with deformation for balls G, H and I at a crosshead speed of 0.02 mm/s.
Stiffness is a property of both material modulus and component shape, whereas the component material cylinder testing can identify the modulus of the material. Compression testing results for the core materials are summarised in Table 2.
The Engineering of Sport 7 - Vol. 1
461
Modulus and hysteresis values for all component materials could not be obtained due to limits of minimum specimen size, but comparison between compressive modulus and Shore D hardness of the component materials tested shows a general agreement. The central core of ball H demonstrated a high modulus compared to published data on the mechanical properties of cork (Gibson et al. 1981), whereas the other samples all were within published values (Qi et al. 2005). The relationship between modulus and Shore D was, however, not linear. Hysteresis in the polymeric materials was also significantly higher, indicating that balls containing significant amounts of polymer will lose more energy in compression and also show a greater strain rate dependence. Table 2 - Component material properties.
The variation in whole ball hysteresis values as a function of deformation rate for the three deformations are shown in Figure 4. The deformation and deformation rate dependence of hysteresis for ball G, Figure 4 (a) shows a steady increase as both parameters increase consistent with the fully polymeric nature of this ball. The hysteresis of balls H and I shows very similar values and trends with deformation and deformation rate. As the amount of deformation increases more of the core is strained and, due to the viscoelastic nature of cork, the loss in energy increases, which should correspond to a decrease in CoR. As the deformation rate increases for ball H, then the hysteresis at the lowest deformation shows a steady increase. For the lowest amount of deformation, the contribution to total deformation from the cover is greatest and so the greater strain rate dependence of hysteresis shown by PU and PVC is still seen in the whole ball deformation behaviour for both balls H and I. At the higher two levels of deformation a greater amount of deformation is taken up by the cork in the core with its lower strain rate sensitivity (Gibson et al., 1981; Qi & Boyce, 2005). Thus, for ball H, Figure 4 (b), the rate of increase in hysteresis with deformation is reduced for the middle level of deformation and hysteresis is practically constant for all three deformation rates at the highest level of deformation. Similar trends are seen for ball I, Figure 4 (c), but there is still some small strain rate sensitivity at the highest level of deformation, consistent with the presence of an elastomer mixed with the cork in this ball, Figure 1 (c).
462 The Engineering of Sport 7 - Vol. 1
Figure 4 - Energy losses (U) with increasing deformation (3.5 mm ■; 7.2 mm ▲; and 10.5 mm ●) and deformation rate (d’) for balls G (a), H (b), and I (c).
The variations in hysteresis shown in Figure 4 for the three ball types are determined at very low deformation rates compared with those experienced during either standard FIH dynamic testing or during play. Using rebound heights (FIH testing) or the ratio between inbound and outbound ball speeds (air cannon testing) allows the coefficient of restitution to be determined for all three ball types over a wide range of impact speeds. Higher CoR values relate to lower energy losses during impact and so would correspond to lower hysteresis values during whole ball compression testing. The relationship between impact speed and CoR for three ball types is shown in Figure 5. It can be seen that the balls show decreasing CoR with increasing impact speeds, but a variation in behaviour between ball types. All three ball types fulfil the FIH dynamic test requirements, but as impact speed increases towards higher speeds seen during play then ball G shows a very rapid decrease in CoR. This would mean that the speed of this ball off the stick will be much less than the other two ball types and that the bounce of this ball for a higher speed shot will be lower. Ball G was hollow and so all the solid material in this ball is viscoelastic PU. The CoR results follow the same trends as shown in the whole ball compression tests. Ball I has the highest CoR (rebound height) during dynamic FIH testing; this shows a reduced variation in CoR with impact speed compared with ball G, which also correlates with the low deformation rate whole ball compression trends. The presence of an elastomer filler in the cork core of ball I causes some viscoelasticity to be seen as larger deformations cause more of the core to be strained. In contrast, ball H starts with the lowest CoR in the dynamic FIH tests, but shows the lowest strain rate sensitivity, maintaining its CoR value more as impact speed increases, which correlates with the lower strain rate sensitivity of cork and the results of the low deformation rate whole ball compression tests. The CoR values for balls H and I converge at higher impact speeds, but ball H would demonstrate greater consistency in response. Therefore, low deformation rate testing of whole balls in compression can be used to qualitatively predict the strain rate sensitivity of the balls provided that the balls are compressed to a large enough deformation; this would suggest that the hockey balls are more strain than strain rate dependent. The trends seen in both low deformation rate and impact testing are consistent with the materials used to construct the core and cover of the ball suggesting that the ball performance could be predicted from materials and
The Engineering of Sport 7 - Vol. 1
463
construction in a similar manner to that established for solid golf balls (Strangwood et al., 2006). In terms of controlling ball speed off the stick, this work would indicate that low hysteresis (high CoR) is needed, which, for high speed (large ball deformation) shots is favoured by a low hysteresis core. At lower shot speeds a high CoR would be favoured by low hysteresis cover and outer core layers. For the ball types studied this behaviour can be controlled through the mix of cork and elastomer used in the core as well as the cover material and thickness. This work is for impact against a rigid target and so reveals the ball behaviour. In play impacts will be the result of the interaction between ball and stick and so both components need to be considered. The standard tests measuring the dynamic response of the ball are conducted at impact speeds of 4.66 m s-1 and 6.26 m s-1; Figure 5 shows the range of conformity at these relatively low impact speeds. It can be seen though that the behaviour of the ball tested at these speeds are not necessarily representative of ball behaviour at the higher impact speeds experienced during play. The combination of whole ball compression testing and structural analysis would allow the impact speed behaviour to be assessed.
Figure 5 - Coefficient of restitution as a function of impact speed for ball types G, H, I. Impact speeds of 16 - 24 m s-1 were obtained through impact testing, and those at 6.26 m s-1 were obtained through ball rebound test. The standard conformity range of 0.5 - 0.61 at 6.26 m s-1 is indicated.
4 - Conclusions and Further Work It has been shown that as strain increases, ball behaviour shows a greater dependence on core properties, whereas at lower strains ball behaviour is primarily dominated by the cover and outer core layers. The influence of cover and core properties has been demonstrated in both low deformation rate compression and high deformation rate impact tests. The behaviour of balls in both these testing conditions has been seen to depend on materials in the ball, and also on the ball’s construction.
464 The Engineering of Sport 7 - Vol. 1 The trends shown in low deformation whole ball compression testing are a qualitative guide to the ball’s impact behaviour and behaviour across both deformation rate ranges can be related to the properties of the materials used in ball construction and their dimensions, provided the amount of deformation applied to the ball during low deformation rate testing is sufficient to represent the impact deformation. The balls studied demonstrated greater strain than strain rate dependence. Further analysis of the ball construction and deformation behaviour, e.g. through more quantified component properties and modelling of impact deformations, is needed to quantitatively predict impact behaviour from low deformation rate whole ball compression. The impact behaviour also needs to be characterised for ball interaction with hockey sticks as well as with rigid bodies.
5- Acknowledgements The authors would like to thank EPSRC for financial support during this research and Chingford Manufacturing for supplying the samples tested.
6- References [F1] FIH, Regulations on performance requirements for hockey balls, 1999. [GE1] Gibson, L.J., Easterling, K.E., Ashby, M.F., The structure and mechanics of cork. Proceedings of The Royal Society, London, A. 377, 99-117, 1981. [M1] McHutchon, M. A., Design methodologies for sports equipment: A case study in hockey sticks. Ph.D thesis, The University of Sheffield, 2006. [QJ1] Qi, H. J., Joyce, K., Boyce, M. C., Relationship between durometer hardness and the stressstrain behaviour of elastomeric materials. Rubber Chemical Technology. 76, 419-435, 2003. [QB1] Qi, H.J., Boyce, M.C., Stress-strain behaviour of thermoplastic polyurethanes. Mechanics of Materials. 37, 817-319, 2005. [RB1] Rai, R., Bhangu, G.S., Mohanty, S.K., Goel, A., Kinematic and Temporal Evaluation of Swings, Stick Length and their Interaction in Field Hockey. Medicine & Science in Sports & Exercise 34(5), 18, 2002. [SJ1] Strangwood M. Johnson A.D.G. and Otto S.R., Energy losses in viscoelastic golf balls. In Proc. I. Mech. E.,Part L: Journal of Materials, Design and Applications, 220(1): 23-30, 2006.
Shape Optimization of Golf Clubface using Finite Element Impact Models (P90) Willem Petersen1, John McPhee2
Topics: golf, finite element impact modelling, clubface shape optimization, impedance matching. Abstract: To model the impact dynamics of a golf drive, high-fidelity finite element models of the ball and the clubhead are created and combined to simulate the collision of the two bodies. A three-piece golf ball is modelled using only solid elements, while the clubhead is modelled using solid elements for the crucial area of the impact, i.e. the clubface, and using shell elements for the rest of the clubhead to improve the computational efficiency of the simulation. The correct transfer of forces and moments in the transition area between the shell and solid elements is assured by introducing kinematic nodal constraints using rigid elements. The finite element model of the clubhead is parameterized with three shape variables that are varied during an optimization of the launch velocity for central impacts. The optimization process is performed in three stages. After analyzing the modal behaviour of each body involved in the impact, the first stage minimizes the natural frequency of the clubhead so that it better matches that of the ball, according to the theory of impedance matching. This first stage only requires structural analyses of the clubface. In the second stage, the ball is included and impacted on the stationary clubhead. The rebound velocity is maximized and the final shape is used as the initial shape in the last optimization stage, to save time during optimization iterations. In the final stage, the entire clubhead model is used and driven into the ball, to determine the clubface geometry that maximizes the launch velocity of the golf ball for central impacts. A final clubface shape is reached and a total improvement of 4.8 m/s over the initial design is obtained. This is a 7% gain in launch velocity, which results in a driving length advantage of approximately 20 meters until the first contact with the ground. The optimization results are compared against those obtained using the principle of impedance matching. Keywords: Golf impact model, finite element method, coefficient of restitution, shape optimization, impedance matching.
1. Systems Design Engineering, University of Waterloo, Waterloo, ON, Canada N2L 3G1 E-mail: [email protected] 2. Systems Design Engineering, University of Waterloo, Waterloo, ON, Canada N2L 3G1 - E-mail: [email protected]
466 The Engineering of Sport 7 - Vol. 1
1- Introduction Over the last decade, golf drivers have become larger, hit the fall farther, and are more forgiving for off-centre ball strikes. In response, the USGA has placed an upper limit of 0.83 on the coefficient of restitution (COR) between the ball and the clubhead, and an upper limit of 5900 g•cm2 on the moment of inertia (MOI) around the vertical axis. To satisfy these new restrictions while designing even better drivers, golf engineers are studying MOIs about horizontal axes [Beach 2007], “hot” faces that have a COR of 0.83 for off-centre strikes, and the effects of center of mass (COM) location on hooks and slices. Insight into driver design and performance can be obtained by studying the impact between the clubhead and ball. One can neglect the dynamics of the golfer and the club shaft because the duration of impact is relatively small (approximately 0.5 ms). In this small amount of time, the golfer and the shaft can not affect the dynamics of the impact [Cochran and Stobbs 1968]. To model the energy losses in the impact, we have modelled the three pieces of a golf ball as spheres of solid finite elements with hyper-elastic properties [Tanaka et al. 2006]. A new high-fidelity model of the clubhead was created using a fine mesh of solid finite elements for the clubface, where the impact occurs, and shell elements for the remainder [Petersen and McPhee 2008]. The use of solid elements provides greater fidelity for the face model. It also allows for a 3D parameterization of the shape in a shape optimization of the face geometry that maximizes the launch velocity for central impacts. This is in contrast to the approach of Nakai et al. 2004, who optimized the thickness over a flat face geometry using shell elements. Our computational time was decreased by the use of appropriate initial conditions in three separate stages of the optimization process; the initial design for each stage was obtained from a prior stage using lower-fidelity (faster) finite element models. During all stages, the flexibility of the clubface from a standpoint of mechanical impedance matching in the frequency domain was taken into account [Yamaguchi et al. 1998].
2- Finite Element Model of Clubhead/Golf Ball Impact To study the golf ball/clubhead collision in detail, a finite element (FE) model of a golf ball and clubhead is developed and simulated using the Altair HyperWorks software and the explicit crash solver LS-DYNA. A finite element model was chosen because of the relatively large deformations in the golf ball during the period of impact. A detailed model is required to simulate those deformations and the resulting energy losses during the collision.
The Engineering of Sport 7 - Vol. 1
467
Figure 1 - Finite element model of the three-piece golf ball.
The FE model of the ball (see Figure 1) was created using solid elements and hyperelastic material properties, in the form of a Mooney-Rivlin model [Holzapfel 2000], to represent a commonly-used three-piece golf ball [Tanaka et al. 2006]; by sharing boundary nodes, the layers were constrained to move together at their interfaces. For the construction of the FE model of the clubhead, a CAD surface model of a new design for a 460cc driver head was created and prepared for meshing. The clubface was modelled using solid elements and the remaining geometry is broken down into seven components and modelled using shell elements [Petersen and McPhee 2008] as illustrated in Figure 2. The clubhead and face are made from the TiAl6V4 titanium alloy.
Figure 2 - Views of the finite element model of the golf clubhead.
468 The Engineering of Sport 7 - Vol. 1
3- Optimization To create a golf club design that maximizes the launch velocity for an impact on the “sweet spot”, a shape optimization of the finite element model of the clubhead/golf ball impact is performed. The clubface was parameterized by three basic shapes, which represent the optimization design variables. The optimization process is performed in three stages for which the existing finite element (FE) model was modified and used in three different ways: 1. Shape optimization of the clubface using the structural FE solver OptiStruct • clubface only model • minimizing the natural frequency of the first mode of the clubface 2. Shape optimization of the clubface using the FE crash solver LS-DYNA • clubface only model plus golf ball • maximizing the rebound velocity of the ball driven into the clubface 3. Shape optimization of the clubface using the FE crash solver LS-DYNA • entire clubhead model plus golf ball • maximizing the launch velocity of the ball This three-stage process is used to save iteration steps, and thus computational time, for the final optimization step in which the entire clubhead model is simulated. To save time-consuming iteration steps, the results of the previous optimizations are used as initial conditions for the design variables, which makes the optimization routine converge earlier.
3.1 Optimization Process in Three Stages Essentially, the shape of the clubface is optimized by making it more flexible. To perform a shape optimization on the clubface, its 3D geometry is parameterized by three design variables that define the basis vector of the optimization. The linear combination of these basis shapes creates the actual shape of the clubhead geometry. The three design variables, along with the initial clubface shape, can be seen in Figure 3. The three different shape variables were created by simply morphing the inside surface of the clubface. The outside surface, where the impact occurs, was kept constant. The first basis shape (Shape 0) is introduced to allow thinner clubfaces; it is created by simply moving the nodes on the back face in the normal direction to its surface. The second and third design variables, Shape 1 and Shape 2, are created by concentrating the masses along an elliptical ring around the centroid of the clubface and as an ellipsoidal dome in the centroid of the clubface, respectively, as proposed by Nakai et al. (2004). The optimizations in every stage are performed using the adaptive response surface optimization routine.
The Engineering of Sport 7 - Vol. 1
469
Figure 3 - Nominal clubface shape and shape design variables.
3.1.1 Optimization Step 1: Minimizing the natural frequency of the clubface The first optimization step is performed in order to get an understanding of the modal behaviour of the clubface, and to better match the impedances of the golf ball and clubhead. First, modal analyses of both the ball and the clubface are performed to get the natural frequency of the initial face geometry and the reference frequency of the ball. The modal analysis is performed using the structural solver OptiStruct, which uses an eigensolver based on the Lanczos algorithm. Modal analyses of the three-piece golf ball and clubface: Separate analyses were performed on the previously described golf ball model and the initial design of the clubface. Whereas no nodal constraints or forces are applied to the ball, all the nodes around the edge of the clubface were constrained for this analysis. The first natural frequencies of the ball and the clubface were calculated as 2727Hz and 6703Hz, respectively. The calculations also show that the clubface is most compliant in the direction of the major loads applied during the impact of the clubhead and the ball. The ball is equally compliant in all directions, and its first mode shape also describes a motion in the direction of the major loads. The first optimization stage is performed using only the solid element model of the clubface. It is parametrized by the earlier defined design variables. The clubface edge nodes are again fixed in all directions, but free in their rotational degrees of freedom. For
470 The Engineering of Sport 7 - Vol. 1 this stage, the mass of the clubface is constrained by a lower limit of 20g. The objective in this step is to minimize the natural frequency related to the first mode of vibration, so that it better matches the ball frequency according to the principle of impedance matching. Since the natural frequency of the initial clubface design is rather high, the thinning design variable Shape 0 is initially weighted with 1, and the other two design variable start from zero. The initial settings and the optimization results are listed in Table 1. Table 1 - Initial conditions and final outcome of the first optimization stage.
After 20 iterations, a natural frequency of 2971Hz for the clubface was reached, while the mass approached its lower limit of 20g. The curves of the objective function, mass constraint, and design variable weights are shown in Figures 4 - 5.
Figure 4 - Constraint and objective over optimization steps
Figure 5 - Design variable weights over optimization steps.
3.1.2 Optimization Stage 2: Maximizing the rebound velocity of the ball (clubface model) For the second stage in this optimization process, the clubface model is combined with the golf ball model used earlier. Whereas a purely structural analysis is performed in the first stage, Stage 2 optimizes the clubface geometry by maximizing the rebound velocity of a ball driven into the stationary clubface. This intermediate step is performed to obtain appropriate initial values, shown in Table 2, for the design variables in the final optimization stage.
The Engineering of Sport 7 - Vol. 1
471
3.1.3 Optimization Stage 3: Maximizing the launch velocity of the ball (full clubhead model) For this third and last stage in the optimization process, the full finite element model of the ball/clubhead impact is simulated. The shape optimization of the face geometry is then performed using the final shape parameters from the second stage. This time, the sweet spot of the clubhead is driven into the ball at a velocity of 44.7 m/s. The mass is constrained to lie in the range of 20g to 35g, and the objective is to maximize the launch velocity of the golf ball. The initial values are provided in Table 2, along with the values that were obtained during the optimization. Table 2 - Initial conditions and final outcome of the third optimization step
The optimization routine is run for eight iterations, after which the design variables are not changing significantly and a final launch velocity of 73.5 m/s is obtained. Only 0.5 m/s is gained by this last stage, so it can be concluded that the initial values of the design variable weights were close to the final optimal values. Compared with the driver design initially used in the optimization stage 1, which launches the ball with a velocity of 68.7 m/s under the same impact conditions, a total gain of 4.8 m/s (or 7%) is achieved. The curves of the launch velocity, mass, and design variable weights at each optimization step can be seen in Figures 6 - 7.
Figure 6 - Constraint and objective over optimization steps
Figure 7 - Design variable weights over optimization steps
The final clubface shape is illustrated in Figure 8. The first natural frequency of this optimized shape is 4040 Hz, which is quite a bit higher than that obtained after the first optimization stage. In the following section, the optimization results are compared with those obtained from the theory of impedance matching.
472 The Engineering of Sport 7 - Vol. 1
Figure 8 - Final optimized clubface shape
Figure 9 - Impedance matching theory compared to optimization results
4- Comparison to Impedance Matching Theory To compare our results against those obtained using mechanical impedance matching, a lumped-mass model of the impact was developed. The ball and clubhead were each represented by 2-DOF models, including spring and damper, and impacted against each other through a Hertzian contact. The model is simulated for different stiffnesses (i.e. natural frequencies) of the clubhead. The results show that the best energy transfer between the two bodies occurs when the ratio of their natural frequencies is unity [Petersen 2007], consistent with the theory of impedance matching [Yamaguchi et al. 1998]. The clubhead obtained by optimizing the ball launch velocity has a natural frequency of 4040 Hz, which is approximately 1.5 times the natural frequency of the ball. To investigate this further, six different clubface geometries with different natural frequencies of 3000-5500 Hz are simulated and tested on their release velocity. The ratio of the relative velocity of the clubface to the golf ball after the impact (v0) over the pre-impact relative velocity (v1) is plotted against the natural frequency ratio in Figure 9. The finite element models of the six clubface designs are represented by the red asterisks, while the values obtained from the simplified lumped-mass model are represented by the blue curve. The cubic green curve shows an approximate fit to the finite element results. The plot shows that the best ratio of natural frequencies is indeed close to 1.5, which confirms the results gained from the optimization. This also shows that impedance matching does not necessarily lead to the fastest launch velocity of a ball, possibly due to the overly-simplified linearized models that are typically used in impedance matching approaches.
5- Conclusions A detailed finite element model of the impact between a three-piece golf ball and a driver was developed, so that the physics of local deformations and wave propagation were correctly included in the impact simulation. This high-fidelity model was used to optimize the clubface geometry for a better energy transfer between the two colliding bodies
The Engineering of Sport 7 - Vol. 1
473
during the impact, thereby maximizing the launch velocity of the ball for central impacts. The clubface geometry was represented by a linear combination of three different basic shapes, which comprised the design variables for the three-stage optimization process. Using modal analyses results for the ball and clubhead, the first stage minimized the natural frequency of the clubface so that it better matched that of the ball, in accordance with the principle of impedance matching. Using the Stage 1 results as the initial design for the second stage, the rebound velocity of the ball was maximized when it was launched into a clubface whose edge nodes were fixed in translation. By using the results from simpler models as the initial designs for subsequent stages, computational time was saved since fewer simulations were required in the subsequent higher-fidelity models. Thus, the results of the second stage were used as the initial design configuration for the third stage, in which the full finite element model was used to simulate a central impact between moving clubhead and ball. The result of this last optimization stage was a clubhead design that launched the ball with a 7% higher velocity than the initial clubhead design. A comparison of the optimized finite element results against those from the theory of impedance matching shows that the latter theory does not necessarily lead to maximal ball launch velocities.
6- References [BA1] Beach, T., Anderson, D., and Vincent, B. (2007), “Golf Clubhead”, U.S. Patent number 7,198,575. [CS1] Cochran, A., and Stobbs, J. (1968), Search for the Perfect Swing. Triumph Books: Chicago, Illinois. [H1] Holzapfel, G.A. (2000), Nonlinear Solid Mechanics: A continuum approach for engineering. John Wiley and Sons: New York, USA. [NW1] Nakai, K.; Wu, Z.; Sogabe, Y.; Arimitsu, Y.: A study of thickness optimization of golf clubheads to maximize release velocity of balls. In: Communications in numerical methods in engineering 20 (2004), pp. 747–755. [P1] Petersen, W.: Dynamics of Impact between Golf Clubhead and Ball, Diplomathesis (2007), Hamburg University of Technology, Germany. [PM1] Petersen, W., and McPhee, J: Comparison of Impulse-Momentum and Finite Element Models for Impact between Golf Ball and Clubhead, Proceedings of the World Scientific Congress of Golf (2008), Phoenix, USA. [TS1] Tanaka, K.; Sato, F.; Oodaira, H.; Teranishi, Y.; Sato, F.; Ujihashi, S.: Construction of finiteelement models of golf balls and simulations of their collisions. In: Proceedings of the Institution of Mechanical Engineers, Part L: Journal of Materials: Design and Applications 220 (2006), pp. 13–22. [YI1] Yamaguchi, T.; Iwatsubo, T.: Optimum Design of Golf Club Considering the Mechanical Impedance Matching. In: Science and Golf III: Proceeding of the World Scientific Congress of Golf (1998), pp. 500–510.
An Examination of Cricket Bat Performance (P92) Lloyd Smith1, Harsimranjeet Singh1
Topics: Cricket, performance. Abstract: The aims of this study were to experimentally measure and numerically describe the performance of cricket bats and balls. A dynamic finite element model was employed to simulate the bat-ball impact. The ball was modeled as a linear viscoelastic material which provided the mechanism of energy loss during impact. An experimental test apparatus was developed to measure the performance of cricket bats and balls under dynamic impact conditions representative of play. Experiments were conducted to measure the elasticity and hardness of the cricket balls as a function of incoming speed. A bat-performance measure was derived in terms of an ideal batted-ball speed based on play conditions. The model found good agreement with experimental data for a number of impact conditions. A composite skin, applied to the back of some bats, was observed to increase performance experimentally and in the numerical model. While different treatments and designs typical of cricket bats had a measurable effect on performance, they were much smaller than the 10% difference observed between some solid-wood and hollow baseball and softball bats. Keywords: Cricket bat, Cricket ball, COR, Dynamic stiffness, BBS, Composite skin.
1- Introduction Although, the sport of Cricket is 500 years old [B1] there has been little research studying the bat or the ball. Since the 17th century, the cricket bat has been changed various times, but remains of solid wood [T1]. The aim of the bat is to maximize batted-ball speed, and minimize vibration to the batsman’s hands. The blade is made of willow which is strong, lightweight and has good shock resistance. The handle is made of cane which has good shock absorbing properties. The length of the bat cannot exceed 38 inches (96.5 cm), and the width of the blade must be less than 4.25 inches (10.8 cm) [T1]. The performance of cricket bats has been compared using their coefficient of restitution (COR); defined as the ratio of relative speed of the objects after and before the collision. One study found that the COR of the bat decreased as the bat stiffness 1. School of Mechanical and Materials Engineering, Washington State University, 201 Sloan, Spokane Street, Pullman, WA 99164-2920 USA - E-mail: [email protected]; [email protected]
476 The Engineering of Sport 7 - Vol. 1 increased [F1]. Recent advances in technology and materials have motivated a number of changes in cricket bat design. While some studies suggest these advances have not affected performance, more work is needed to quantify their contribution [S4]. Cricket balls are made from a cork nucleus with layers of wound wool and cork and a leather cover. The leather exterior is usually constructed from four sewn pieces. A cricket ball weighs between 5.5 to 5.75 ounces (155.9 to 163 g) and can be no more than 9 in. (22.9 cm) in circumference [T1]. Cricket balls are made in a number of different ways, with varying core design. Some have shown that greater deformation was found for impacts landing on the seam, compared to those landing perpendicular to the seam [C1]. There is little information on the effect of cricket ball properties on bat performance.
2- Numerical model 2.1 Ball To understand the impact between a bat and ball a finite element model was developed. A viscoelastic model was selected for the cricket ball, defined by the time dependent shear modulus [S1, M1, S3] as,
(1)
and a constant bulk modulus, k where Go is the instantaneous shear modulus G is the long term shear modulus, and determines the time sensitivity of the model. An apparatus, Figure 1, was used to characterize the ball at strain rates representative of play. The apparatus and test involved pitching a ball toward a rigidly mounted load cell. Ball speed was measured before and after impact using light gates. The apparatus and ball were modeled using finite elements. By tuning the properties of the ball good, agreement was attained with the load-time curve as shown in Figure 2 for a 60-mph (26.8-m/s) ball speed. The viscoelastic material properties are defined in Table 1.
The Engineering of Sport 7 - Vol. 1
477
2.2 Bat The geometry of the cricket bat was created using CATIA V5 R17. The geometry was imported into LS-DYNA (Livermore Software Technology, Livermore, CA) for finite element analysis. The mesh of the cricket bat consisted of 44544, 8-noded solid elements as shown in Figure 3. The density of the willow and cane were tailored to match the measured weight and mass moment of inertia (I) of the bat. Properties of the willow and cane are summarized in Table. 2.
Figure 2 - Comparison of finite element and experiment of cricket ball impacting a load cell at vp = 60 mph (26.8 m/s). Table 1 - Viscoelastic Material Properties of Ball.
Table 2 - Elastic Properties of Bat.
478 The Engineering of Sport 7 - Vol. 1
3- Experiment 3.1 Ball Testing Cricket balls were compared by their elasticity and hardness. Elasticity was quantified through their rigid-wall COR. Ball hardness was quantified through a so-called dynamic stiffness [S3], which was defined by equating the ball’s initial kinetic energy with its stored energy upon impact with the load cell. The unknown ball displacement was described by the measured force, assuming the ball acted as a linear spring. Accordingly, an expression for the ball’s dynamic stiffness, kd, was found as (2) where mb is the ball mass, Fp is the peak impact force, and vi is the inbound ball speed. All balls were conditioned at 72±2°F (22±1°C) and 50±5% relative humidity for at least 14 days (or saturation) prior to testing. Figure 4 shows the ball dynamic stiffness as a function of incoming ball speed. The FEA results are also included in Figure 4, which show good agreement with the experimental data over a range of speeds. Figure 5 compares the force-displacement curve of two representative ball brands with the FEA model. Displacement from the experiment was obtained by dividing the force by the ball mass and integrating twice with respect to time. It was observed that Ball A had 17% more deformation than Ball B. The different ball constructions and materials likely contribute to the characteristic responses of Balls A and B observed in Figure 5. The FEA model was tuned to Ball B, and the FEA model and Ball B show good force-displacement agreement.
Figure 3 - Cricket bat mesh.
The Engineering of Sport 7 - Vol. 1
Figure 4 - Dynamic stiffness and COR as a function of incoming ball speed.
479
Figure 5 - Representative force-displacement curves for two ball models.
3.2 Bat Testing Apparatus The bats were tested using a fixture similar to the ball test as shown in Figure 6. The rigid wall was replaced by a fixed pivot that allowed the bat to recoil after impact and controlled the impact location. In the bat tests, an incoming ball speed of 60 mph (26.8 m/s) was used to prevent accumulated bat damage from influencing the results. Light gates measured the ball speed before and after the impact. The ratio of the rebound to inbound ball speed is the so-called collision efficiency, ea [N1]. If the bat, vb, and ball, vp, speed in play conditions are known, the collision efficiency may be used to find the batted-ball speed, BBS, according to (3) The bowled-ball speed is relatively easy to measure in play and is usually taken as a constant when comparing bat performance. The release speed of fast bowlers approaches 100 mph (44.7 m/s) at the hand, while the speed at the bat after impacting the pitch is near 80 mph (35.8 m/s) [P1]. In this work, it was taken at 85 mph (38 m/s). The bat speed is more difficult to measure and has a greater effect on the BBS than bowled-ball speed. For this work, a bat speed of 70 mph (31.3 m/s), 21 in. (533 mm) from the knob end was used. This bat speed was found by considering a ball-flight trajectory of 450 yards or 411 m. Bat speed is not constant, but will vary with I and impact location, Q, [S2], [C2] according to (4) where vr (70 mph or 31.3 m/s) is the reference speed, Qr (21 in or 0.533 m) is the reference location, and Ir (10,000 oz in2 or 183 g m2) is the reference bat mass moment of inertia.
480 The Engineering of Sport 7 - Vol. 1 Each bat was impacted six times at multiple locations along its length until a maximum BBS location was found within 0.50 in. (13 mm). Representative performance curves for the experimental data and the finite element model are shown in Figure 7. Note the relatively small region which produces a maximum BBS. The algorithms used to manage contact between colliding bodies in explicit dynamic simulations are sensitive to node alignment and mesh density. In spite of the effort used to characterize the ball, its viscoelastic properties required tailoring to achieve the correlation observed in Figure 7. These properties are also summarized in Table 1.
Figure 6 - Schematic of test fixture used to test cricket bats.
3.3 Bat Testing Procedure While cricket rules require the bat to be made of wood, some manufacturers have added a thin composite skin to the back surface. The mass properties of a bat with and without a skin are compared in Table 3. The skin stiffens the blade and is purported to improve durability. The performance of a bat was compared with and without a composite skin. The results are included in Figure 8, which shows the skin increased the BBS by 1.4%. It should be noted that removing the skin reduced the bat’s I by 4.6%. The effect of I on bat performance, independent of the composite skin, may be considered in Eq. (3) and (4) using the bat-ball COR from the bat without the composite and increasing the bat I. Accordingly, a 4.5% change in I was found to increase bat performance by 0.85%. Thus, roughly half of the performance advantage attributed to a composite reinforced blade is due to its increased I. The effect of the 0.005-in. (0.13-mm) thick composite skin was considered in the numeric model using the properties of Table 2. The results are included in Figure 8, where the skin increased the BBS by 1.3% The FEA results of Figure 8 considered a bat model with a different weight distribution than was tested experimentally. This different weight distribution is likely the cause of the 1% lower BBS obtained from the FEA.
4- Summary This study considered the performance of cricket bats and balls. A finite element model has been used to investigate the performance of a bat numerically. A test apparatus to measure bat and ball properties at impact speeds representative of play condi-
The Engineering of Sport 7 - Vol. 1
481
tions was shown to have utility in comparing bat and ball response. The FE models showed good agreement with the experimental data for bat and ball performance. Mass distribution and composite reinforcement were shown to have a measurable effect on bat performance. In comparison with hollow baseball and softball bats, however, the effect was relatively small. Table 3 - Properties of Cricket bat and FEA used in the performance comparison.
Figure 7 - Representative performance curves Figure 8 - Comparison of finite element for bat and model. (Impact location measured model and cricket bat with and without from the pivot, 6 in or 152 mm from the knob.) composite skin.
5- References [B1] Bowen, R., “Cricket: A History of its growth and development throughout the world”. Eyre and Spottiswoode. London. 1970. [C1] Carre, M. J., James, D. M. and Haake, S. J.“Impact of a Non-homogeneous Sphere on a Rigid Surface” Proceedings of the Institution of Mechanical Engineers; ProQuest Science Journals; 218, 3; p273, March 2004. [C2] Cross, R. and Bower, R. “Effect of Swing-Weight on Swing Speed and Racket Power”. Journal of Sports Sciences;24(1): p23-30, January 2006. [F1] Fisher, S., Vogwell, J., Bramley, A. N. “The Effect of Structural Design on the Coefficient of Restitution for Some First Class Cricket Bats”. Sports Engineering, Vol. 7 Issue 4, p31-37, 2004. [M1] Mustone, T.J., Sherwood, J.A.“Using LS-DYNA to characterize the performance of the baseball bats”. 5th International LS-DYNA Users Conference, September 21-22, Southfield, MI, 1998. [N1] Nathan, A. M. “Characterizing the Performance of Baseball Bats” Am. J. Phys, 71 (2). February 2003.
482 The Engineering of Sport 7 - Vol. 1 [P1] Penrose, T., Foster, D. and Blanksby, B. “Release Velocities of Fast Bowlers During a Cricket Test Match” Supplement to the Australian Journal for Health, Physical Education and Recreation, p2-5, March 1976. [S1] Sandmeyer, B.J. “Simulation of bat/ball impacts using finite element analysis”. Master’s thesis, Oregon State University. 1994. [S2] Smith, L. V., Cruz, C. M., Nathan, A. M., Russell, D. A. “How Bat Modifications Can Affect Their Response,” APCST, Tokyo, Japan, The Impact of Technology on Sport, (Subic & Ujihashi, Eds) p. 33-38, 2005. [S3] Duris, J., Smith, L. V., “Evaluating Test Methods Used to Characterize Softballs,” The Engineering of Sport 5th International Conference, Vol. 2, pp. 80-86, Davis, CA, 2004. [S4] Stretch, R.A., Brink, A. and Hugo, J. “A Comparison of the Ball Rebound Characteristics of Wooden and Composite Cricket Bats at Three Approach Speeds”. Sports Biomechanics Vol. 4(1) pages 37-46, 2006 [T1] The laws of Cricket – 2000 Code, Marylebone Cricket Club, London, 2000.
Forces Applied on Rowing Ergometer Concept2®: a Kinetic Approach for Development (P94) Nicolas Découfour1, Franck Barbier1, Philippe Pudlo1, Philippe Gorce2
Topics: Sport materials. Abstract: Conception of rowing ergometer is constrained by many limits. One of them is that it must resist to forces applied on it, whatever is the frequency movement, and take few spaces in an apartment or a room. A good sport material must respect these two major constraints. In this study, we focus on resistance of material using forces quantification on each contact point on the rowing ergometer concept2® at different frequencies of practice. Our goal is to reveal where structure modifications could be effected in order to optimize conception. There are four contact points on this ergometer: “the handle, the seat and the two foot-stretchers”. On each point, forces are measured with a specific material. On the handle, a mono-directional sensor is used. Under the seat and the two foot-stretchers, 6 axis force plates were used. These measurements are realised with an expert rower who practice rowing on the concept2® ergometer more than twice in a week and is an international level rower on boat. One rowing cycle is selected for each produced stroke rate: 18 to 40 strokes per minute. First results show that forces are not really modified when stroke rate increases. Secondly, forces produced under the seat are not constant. An inertial parameter explains this inconstancy: the accelerated rower masses of his trunk and his upper limbs. Finally, on stretcher, forces are different too because of inertial effect of whole body at the end of recovery. In resume, the structure of concept2® ergometer can be optimised, changed and probably costs less if we take into account this type measurements before design a rowing ergometer. Keywords: Rowing ergometer, applied forces, kinetics and performance.
1- Introduction The design of sport ergometers requires thorough knowledge of the studied sport. The design is characterized by the dilemma space and weight equipment versus resistance to the activity. Mechanically, it has to resist to the efforts developed by costumers. This resistance has to be sustainable over time. Moreover, the finished structure must allow a 1. L.A.M.I.H. – UMR CNRS 8530, Université de Valenciennes, France - E-mail: [email protected]; [email protected]; [email protected] 2. HandiBio, Université du Sud Toulon-Var, France - E-mail: [email protected]
484 The Engineering of Sport 7 - Vol. 1 practice in all of the practicing movement amplitude [1], and must take a minimum living space too. Moreover, the weight of the finished structure becomes predominant when the population becomes more and more old [2]. In this paper, the studied activity is rowing. It imposes several constraints to manufacturers because it is a sport that can be done with a development of small forces (elders) to very high (expert rowers) at each contact point between costumer and ergometer. At the moment, some rowing ergometers coexist and each have their specificities, but to optimize the structure of the ergometer, engineers need to know precisely which constraints the whole structure will have to withstand. Currently in the literature few data are clearly described. Pudlo et al. [3] have instrumented a Concept2 rowing ergometer ® Type C and have tested it with a French regional level rower. Nevertheless, these data are not significant enough in order to characterize atypical or extreme practice. The purpose of this paper is to provide all the kinetic or kinematic data needed to design a rowing ergometer considering an extreme use. The presented data will come from a movement realised by a very high level rower at different stroke rates: training and racing. This study is based on the same existing ergometer as Pudlo et al. [3] but the expert rower is a French champion and member of the French team.
2- Protocol and Method 2.1 Protocol To define rowing ergometer characteristics, conception teams must have a review of kinematics and kinetics data of extreme rowing activity. Consequently, the followed data are measured or calculated with respect of experimental instructions.
2.1.1 Measured and calculated data
The Engineering of Sport 7 - Vol. 1
485
2.1.2 Experiments A French rower gold medallist at French championship in double scull lightweight is tested on the instrumented ergometer. His characteristics are: 28 y.o., weight: 72.5 kg, height: 1.73 m. The experiments consist to row on Concept2® ergometer at stroke rates between18 to 40(strokes.min-1). One cycle for each stroke rate is recorded.
2.2 Experimental Apparatus The experimental apparatus comprised a 3D motion analysis system, a Concept2® ergometer Type C (Morrisville, VT, USA), 3 six-axis force-plates, and 1 mono-dimensional force transducer (Figure 1).
2.1.1 The motion capture system The motion capture system VICON 612 (Oxford Metrics, Oxford, UK) is composed by 8 cameras placed around the ergometer and coupled with an analogue acquisition of the force transducers. The kinematics and kinetics data acquisition frequency is 60Hz. The reference frame is defined by the M/L axis (pointing to the right), A/P axis (pointing forwards) and V axis (pointing upward).
2.1.2 Stretchers Instrumentation Two three-axis forceplates (PX2000, LOGABEX, Giat-Industrie) are placed under two stretchers independent of the ergometer. The force measurement range is 5500 N for M/L and A/P components, 21000 N for the V component. These forceplates are considered to be sufficient since they allow recordings at levels higher than the force peak recorded by MacFarlane et al. (1997) which was equal to 900 N.
2.1.3 Seat instrumentation A third three-axis miniature forceplate (EX114.45-200, LOGABEX, Giat-Industrie) is fixed under the ergometer seat. Its measurement range is 2000 N for the V component, 500 N for the M/L and A/P components. The error given by the manufacturer on each component is 1% of the measurement range.
2.1.4 Handle instrumentation One mono-dimensional force transducers (ELHM-T3M-10KN, ENTRAN), with a measurement range of 10000 N, allowable overcharge of 5000 N and error is less than 0.01%, operates in traction-compression mode. This transducer was considered sufficient since tests performed on 81 national level heavy-weight rowers showed that cumulated force from both hands was maximal at the start of the race, reaching 1352 +/-109 N (Hartmann et al., 1993). This transducer allows dynamic and static forces measurements. This transducer is located at handle / chain interface. The 3 components of F are calculated in the global reference frame.
486 The Engineering of Sport 7 - Vol. 1
3- Results The results are exposed in relation with the design procedures and as a function of stroke rate.
3.1 Kinematics 3.1.1 Handle and seat displacements To determine seat rail and chain length needed for this rower practice, seat and handle movements have been measured and results are expressed in table 1. Table 1 - Seat and handle amplitudes with stroke rate increase.
This table shows that the biggest amplitude displacement of the seat is obtained at 18 strokes.min-1 with 531mm. It shows too that the biggest anteroposterior amplitude (1480mm) and vertical amplitude (215mm) are respectively obtained at 28 and 40 strokes.min-1 (see also, Figure 6). Consequently, a maximal seat rail and chain lengths is not necessary obtained with an extreme practice and confirm that stroke rate is not a necessary factor to take into account for these lengths.
3.1.2 Handle velocity and acceleration on each rowing phase To give maximal values of handle when the stroke rate increases, figure has to be analysed. The figure 1 (a) shows the handle velocity during propulsive phase. At 40 stroke.min-1, the handle velocity is at its maximum with 2.4m.s-1. More the costumer increases the stroke, more the handle velocity increases. Then, it is necessary to have a handle which can move at this velocity without adding passive friction. Indeed, the customers adjust the air-braked flywheel to limit the lumbar injury that can be induced by regular and intense practice. The injury mechanism can be explained by the Figure 1 (b). The acceleration is intense on start of propulsive phase and grows up to “0” at 1820% of the propulsive phase. When the acceleration is around 0 the force applied to the handle is maximal. At this moment, the joint moment applied at the level of L5 is maximal because the posture of the rower induced a maximal lever arm. On recovery phase, handle velocity and acceleration is quasi symmetrical to the propulsive phase. Likewise during the recovery phase, the passive friction of the handle has to be minimised because it does not penalise the chain coming in. During this phase, the flywheel does not produce; of course, any reaction force, nevertheless the recovery phase is important and can’t be ignored because it’s a rest time during the effort.
The Engineering of Sport 7 - Vol. 1
487
Kinematics data leads to determine the ergometer geometry. Kinetics data provides more information about the constraints exerted on the ergometer.
Figure 1 - Handle velocity (a) and acceleration (b) during propulsive phase at different stroke rates.
Figure 2 - Handle velocity (a) and acceleration (b) during recovery phase at different stroke rates.
3.2 External forces 3.2.1 Handle force Maximal handle force developed by the tested rower is around 1050 N (figure 3, a). This maximal value of handle force during propulsion is constant what the maintained stroke rate is. The figure 3 (b) shows that force developed at the handle by the rower during the recovery stays positive and quasi constant at 20N whatever the stoke rate. The fact that the handle force stays positive could be explained by the coming in of the chain by elastic. The recovery phase using an ergometer is different than the practice of rowing on boat [4, 5, 6]. The engineer can use this gap on the actual ergometer to propose technological solution to optimise the simulation of boat rowing on ergometer.
Figure 3 - Force measured at the chain / handle interface during propulsion (a) and recovery (b) rowing phase.
488 The Engineering of Sport 7 - Vol. 1
3.2.2 Seat forces The seat ergometer has been a revolution in rowing sport because it permits rowers to increase the distance of their blade on water activity. That’s why, on ergometer, seat is an important parameter because it permits rowers to have a greatest handle amplitude without any fright of disequilibrium. So the seat must have to support the weight of the rower but not only. The figure 4 represents forces at stroke rate 18 and 40.
Figure 4 - Force measured under the seat during propulsion (a) and recovery (b) rowing phase.
This figure shows a lightening of the rower during the first middle part of propulsive phase. Effectively, he pushes on his legs (Figure 4, a) backward and downward. But at the end of propulsive phase, the vertical forces measured grow up until 711 N (the weight of the rower). The same observation can be done at the beginning of the recovery phase (Figure 4, b). During the first 60% of the recovery phase, a vertical force is measured by the instrumented seat. It seems to have no compensation on this part of movement. Finally, on the other axis, there is no important force applied. Consequently, this observation can be used to propose a seat lightening in order to win weight on the ergometer.
3.2.3 Foot stretcher forces Foot stretchers have to support the pushing force produced by legs during propulsive phase but not also, they must be able to accept the pulling forces developed during inversion movement of rower during recovery phase. On propulsive phase, figure 5 (a) shows that on the first 50% of this phase, A/P force increases up to 500N and V force stay nearly constant at 300N. A/P force is the most comprehensive force applied on stretchers because the rower pushes on them. Such as V force is constant during those 50% we can suppose that the rower is “standing up” on the foot stretchers. Then, this result can explain the fact that on the seat, there is less weight measured than the body mass value. The end of propulsion is characterised by negative A/P and V forces. This data shows that rower realised a pulling phase on stretcher to compensate the trunk inclination at this moment. On recovery phase, V and A/P forces grow up during all the recovery phase. Indeed, the rower places his body segments near foot stretcher.
The Engineering of Sport 7 - Vol. 1
489
Figure 5 - Force measured under the right foot stretcher during propulsion (a) and recovery (b) rowing phase.
3.2.4 CoM positions during cycle Rower CoM during the whole cycle can give information about the multi-compensation of forces realised by the rower measured on ergometer. On figure 6, CoM altitude decreases about 80mm at the end of propulsion. This decreasing informs on the reaction force measured at the seat because trunk acceleration influence forces on the seat. This information is confirmed by upper results. Displacement of CoM during cycle is unchanged with stroke rate and so the coverage zone of CoM displacement on the rail is always the same for this rower. So, it can be determined where the rail must be the most robust.
Figure 6 - Rower CoM on a rowing ergometer cycle : recovery is represented by the lower part and propulsion by the upper part of curves.
4- Discussion Our results provide information needed by designers on handle, seat and structure when a high level rower uses them at different stroke rates. Handle and seat displacements leads to determine the geometry of the ergometer. Handle velocity and acceleration provide to designers characteristics of chain come in and in comparison with the real activity of on-water rowing. We can note that there is no attractive force on the row handle while on this ergometer there is one during recovery where the rower are not pulling the handle but pushing it. It informs designers who continue to consider the
490 The Engineering of Sport 7 - Vol. 1 centred pulling handle as the best solution while it does not represent the real practice of on-water rowing. External forces measured at each contact point shows that kinetics of rowing movement is not so easy. The handle pulling force analysis shows that it does not change with stroke rate level but its maximum is near the 1050N and so components must resist at this pulling force. The seat pressure analysis shows that the seat must support heavy charge especially at the end of propulsive phase when trunk is on back. Stretcher analysis shows that rower produces a leg pushing during the first 50% of propulsive phase and just after he adjusts his body segment posture in order to stop his body segment inertia. On recovery phase, the rower just returns on the catch position in order to engage a new cycle. Finally, CoM displacement of the rower shows that he uses his trunk as an accelerated inert mass in order to increase handle velocity until the 90% of propulsive phase that confirm past results on ergometer or on boat [5]. Designers must take into account this information to develop an intelligent seat which permits a better shock absorption at the end of propulsion. This study concerns one lightweight high level rower and so can not be sufficient for an efficient ergometer conception but gives a good starting database in order to design a new rowing ergometer on which tests will be reprocessed.
5- References [1] Découfour N. and Pudlo P. Effect of stroke rates on hand-curve on a rowing ergometer. In Computer Methods in Biomechanics and Biomedical Engineering, 8(1), pp 67-68, 2005. [2] Brutel C. La population de la France métropolitaine en 2050 : un vieillissement inéluctable. In Economie et Statistique, 355, pp 57-71, 2002. [3] Pudlo P., Pinti A. and Lepoutre F-X. Experimental laboratory apparatus to analyze kinematics and 3D kinetics in rowing. In Sports Engineering, 8, pp 39-46, 2005. [4] Martindale W.O. and Robertson D.G.E. Mechanical Energy in Sculling and in Rowing an Ergometer, Canadian Journal of applied Sport Sciences, pp 153-163, 1984. [5] Lamb D.H. A kinematic comparison of ergometer and on-water rowing. In the American Journal of Sports Medicine, Vol. 17(3), 367-373, 1989. [6] Dal Monte A. and Komor A. Rowing and sculling mechanics. In Biomechanics of sport, pp 53119, 1989.
JUMPICUS – Computer Simulation in Ski Jumping (P95) Heike Hermsdorf1, Falk Hildebrand2, Norman Hofmann1, Sören Müller2
Topics: Biomechanics, Modelling, Ski & other Winter Sports. Abstract: From the biomechanics point of view, ski jumping is a complex motion. An excellent tuning of the individual technique of the athlete and the equipment is necessary to achieve good placements in international competitions. For that reason computer simulation gains a growing importance. JUMPICUS is a highly detailed biomechanical multibody model for simulation in ski jumping and offers different methods of biomechanical analysis. Firstly, real jumps can be analysed. The real motion is captured by digital cameras. A video analysis provides 3D-coordinates of significant points on the athlete. By the method of inverse kinematics JUMPICUS evaluates the time history of joint angles as well as the time history of global position and orientation of the athlete. Based on these results, several ski jumping specific parameters (e. g. take-off velocity, angular momentum, angles between body segments and the flight trajectory) can be evaluated and compared with parameters from other jumps or with a given technical ideal. Secondly, using the forward dynamics simulation effects resulting from changing the athletes motion as well as the influence of equipment parameters can be analysed. JUMPICUS has been already applied to study the effect of modification of the take-off motion on the vertical take-off velocity and the angular momentum at the lip of take-off. Keywords: Ski Jumping, Dynamics Simulation, Inverse Kinematics.
1- Introduction In the field of biomechanics and sports science modern simulation methods get a great importance. Computer simulation in sports is useful in many respects. On the one hand it provides insight into the biomechanical behaviour of the system consisting of athlete 1. Institute of Mechatronics, Reichenhainer Straße 188, 09126 Chemnitz, Germany E-mail: hermsdorf, [email protected] 2. Institute for Applied Training Science, Marschnerstraße 29, 04109 Leipzig, Germany E-mail: hildebrand, [email protected]
492 The Engineering of Sport 7 - Vol. 1 and sports equipment, on the other hand it enables a detailed, both qualitative and quantitative assessment of athletic motions (Härtel and Hermsdorf, 2006). JUMPICUS is a simulation toolbox for ski jumping. It is intended to be used for analyses of ski jumps by sports and training scientists. A crucial part of a ski jump is the transitional flight phase that corresponds to approximately 17 m of the initial flight after the take-off and takes about 0.6 s. The transitional phase is unstable. It is decisive that the ski jumper quickly attains a pose corresponding to the stable flight phase following the transitional phase (Hildebrand et al., 2007). The transitional phase is characterized by a high angular momentum (nose down), whereas the flight phase is characterized by a low angular momentum (see figure1). High vertical take-off velocities admit long flight distances, whereas large angular momenta enable to rapidly reach a proper flight posture (Hildebrand et al., 2007). To this end the athletes must find a trade-off between a high vertical take-off velocity and a high angular momentum. In order to better understand the mechanisms behind the interplay of the vertical velocity and the angular momentum the JUMPICUS software toolbox was used as described in this paper.
Figure 1 - Angular momentum after take-off versus horizontal flight distance of 14 international top jumpers
2- The JUMPICUS-Toolbox The JUMPICUS toolbox is an add-on for the commercial multibody system simulation package alaska (alaska, 2007). The toolbox comprises detailed models for the athlete, ski, jumping hill, boots, and their mutual interactions. In particular, the athlete is represented by a DYNAMICUS® model – the biomechanical human model available in alaska (DYNAMICUS, 2007). JUMPICUS can be used in two different variants (see figure 2): JUMPICUS/Basic and JUMPICUS/Expert.
The Engineering of Sport 7 - Vol. 1
493
2.1 JUMPICUS/Basic JUMPICUS/Basic is meant for the simulation of a completely prescribed motion of the athlete and the skis. That is, the motion of the ski-jumper and the skis must be provided by the user. This can be done in two possible ways as follows. One option is to use the time history of a set of body marker coordinates (ankle, knee, shoulder, etc.) with respect to an inertial frame. This approach essentially corresponds to the widely accepted method for inverse kinematics via motion tracking of human models. The second option consists in providing the time history of joint angles and the coordinates of the right ankle marker. Since motion capturing, using video sequences, yields the time history of body markers the first option is the standard way of use. JUMPICUS/Basic performs a biomechanical analysis and computes general mechanical and biomechanical results characterizing the motion, such as joint angles, velocity of centre of mass, and angular momentum. The simulation allows for a detailed analysis of the motion of the athletes and the skis. As such ski jumping specific parameters the take-off velocity, flight trajectory, and V-angle can be determined.
2.2 JUMPICUS/Expert JUMPICUS/Expert can be used to compute the global motion of the athlete as a result from a prescribed motion of the joints and applied external forces acting upon the system as well as interaction forces acting within the system. That is, the user must provide the joint angles dependent on the position. JUMPICUS/Expert computes the motion of the overall system with respect to the inertial frame as result of the specified joint motion. JUMPICUS/Expert yields the same simulation results as JUMPICUS/Basic.
Figure 2 - JUMPICUS-Toolbox.
494 The Engineering of Sport 7 - Vol. 1 The simulation results are presented in different ways. A 3D animation of the jump enables a visual evaluation. Different camera settings allow views that can’t be realized by video cameras. Moreover, the user can visualize and easily compare different jumps in one animation. Additionally, results are presented as graphs, which can be exported to be processed in other tools. The features of JUMPICUS are integrated in a graphical user interface as shown in the figure 3.
Figure 3 - Graphical User Interface of JUMPICUS.
3- Model Components JUMPICUS represents a complex model of a multi body system. It comprises several components (see figure 4).
3.1 Athlete The model of the athlete is created using DYNAMICUS® (DYNAMICUS, 2007). DYNAMICUS® is a library of model components that enables the user to create a spatial multibody model of a human body in a very efficient way. The library offers several model components for parts of the body which differ in joint kinematics. Due to this library structure, it is possible to create a model of the athlete customized to the specific needs of the ski jumping simulation. The joint kinematics is defined in such a way that the motion of the athlete can be adequately modelled. The model of the athlete has a degree of freedom equal to 17. Anthropometric data can be given by input of individual body dimensions as well as by access to different data pools.
The Engineering of Sport 7 - Vol. 1
495
3.2 Ski The ski model consists of 13 rigid bodies which are connected among each other by revolute joints. Spring-damper-forces within the joints model the elastic properties of the ski. Extensive investigations with jumping skis were performed. Both static and dynamic properties were determined. Methods of parameter identification were developed to obtain model parameters, which realistically describe the elastic and dynamic properties of the ski.
3.3 Jumping Hill The construction data of a jumping hill is described by a FIS certificate. Based on these data a mathematical description of the jumping hill surface is created. JUMPICUS computes contact forces between the jumping hill surface and the skis.
3.4 Aerodynamic Forces To compute a realistic motion aerodynamic forces have to be considered carefully. Extensive wind tunnel studies have been performed by the Institute of Applied Training Science. The results of these studies are used to find functions which give the drag and lift coefficients dependent on the posture of the athlete. The focus lies on the aerodynamic forces acting on the athlete during the in-run phase and the take-off phase. A function was created, which gives the drag coefficient depending on the height of the head over the jumping hill surface. Analogously, a second function gives the lift coefficient depending on the angle of attack with respect to the upper part of the body. Using these coefficients as well as the current velocity of the airflow JUMPICUS/Expert computes a drag force and a lift force acting on the athlete.
Figure 4 - Model Components.
496 The Engineering of Sport 7 - Vol. 1
4- Application JUMPICUS/Basic has been used to investigate the take-off motion in ski jumping. The take-off motion of the athlete is of significant importance, because the vertical velocity and the angular momentum at take-off represent initial conditions for the flight phase. The athlete has to find a compromise between these two quantities (Schwameder and Müller, 1995), (Virmavirta and Isolehto, 2005). JUMPICUS/Basic has been applied to investigate how a modification of the take-off motion affects the take-off velocity and the angular momentum. Firstly, the motion of a real jump of a top athlete has been analysed. The jump was recorded with two pan, tilt and zoom video cameras. A video analysis provided the coordinates of body markers on the athlete with respect to the inertial frame. These marker coordinates are the input for JUMPICUS/Basic. JUMPICUS/Basic provides the time history of the joint angles. The joint angles are used to modify the motion of the athlete. Table 1 shows simulation results when the angle of the hip joint at take-off is modified. The time history of the modified joint angle is approximated by a quadratic function according to the specified conditions. The in-run position of the athlete, the angle of the ankle joint as well as the knee angle remained unchanged. The take-off velocity and the angular momentum computed by JUMPICUS/Basic are shown in the table 1. Table 1 - Modification of the hip joint angle.
The second modification causes an increase of the take-off velocity but a decrease of the angular momentum. The next modification step starts with jump 2 and tries to modify the angle of the ankle joint in order to achieve an angular momentum of 12 kgm2/s as for jump 1. Further, a sequence of modifications was carried out taking into account further boundary conditions in order to create take-off motions close to reality. Consequently, JUMPICUS can be used to optimize the take-off motion. Figure 5 shows a comparison of two different take-off motions using the 3D visualisation capability of JUMPICUS.
The Engineering of Sport 7 - Vol. 1
497
Figure 5 - Comparison of different take-off motions.
Since the modifications were made in JUMPICUS/Basic the influence of aerodynamic forces as well as the changes of friction between skis and the jumping hill are not taken into account. These effects will be investigated in future studies.
6- Conclusions The JUMPICUS-Toolbox enables the user to perform a detailed biomechanical analysis of ski jumps. Moreover, the influence of motion modifications can be analysed. Further development will be focused on aerodynamic forces acting in the transitional and the flight phase.
7- Acknowledgments This project is supported by the Bundesinstitut für Sportwissenschaft (BISp), Bonn, Germany.
8- References [A1] alaska 5.1. Reference Manual, Institute of Mechatronics, Chemnitz, 2007. (www.ifm-chemnitz.de) [D1] DYNAMICUS. Reference Manual, Institute of Mechatronics, Chemnitz, 2007. [HDM1] Hildebrand, F., Drenk, V., Müller, S. Stability during ski jumping flight phase. E. Müller (Ed.) 4th International Congress on Science and Skiing. Book of Abstracts, Salzburg p.106, 2007. [HH1] Härtel, Th., Hermsdorf, H. Simulation von Bewegungsabläufen in Kraft- und Techniksportarten. In Edelmann-Nusser, Witte (Hrsg.), Sport und Informatik IX, Bericht zum 6. Workshop Sportinformatik der DVS-Sektion Sportinformatik, Shaker-Verlag Aachen, 2006, 307313, 2006. [SM1] Schwameder, H., Müller, E. Biomechanische Beschreibung und Analyse der V-Technik im Skispringen. Spectrum der Sportwissenschaften 7, 1, 5-36, 1995. [VI1] Virmavirta, M., Isolehto, P., Komi, P., Brüggemann, G.-P., Müller, E., Schwameder, H. Characteristics of the early flight phase in the Olympic ski jumping competition. Journal of Biomechanics 38, Issue 11, 2157-2163, 2005.
Kinematic Response to Variations in Natural Turf During Running (P96) Stiles, V. H.1, Dixon, S.D.1, Guisasola, I.N.2, James, I.T2
Topics: Biomechanics Abstract: Important health and social benefits can be gained from participation in sports and exercise. Appropriate surface provision that aids sports participation, cannot be met by artificial surfaces alone – it requires natural turf surfaces to be utilised. Considerable improvement in the durability of natural turf surfaces and thus, a greater understanding of the human-natural sports surface interaction is required. Ground reaction force data have been used to help quantify how human participants respond to changes in natural turf properties during running and turning. A kinematic analysis would further this understanding. This EPSRC/UK funded study analyses kinematic response to variations in natural turf during running. Three different rootzone conditions (clay, sandy and rootzone) were constructed in portable plastic trays (0.60 m x 0.40 m x 0.08 m) and turfed with ryegrass. Trays were positioned in the laboratory on non-slip matting (6 mm thick) to form a continuous runway. Three-dimensional kinematic data (Vicon Peak, automatic, opto-electronic system 120 Hz) were collected for nine subjects wearing football boots (studded natural turf design) during running (3.83m.s-1). Group mean data for initial and peak ankle and knee angles and peak joint angular velocities were statistically compared using an analysis of variance with repeated measures (ANOVA R.M, p<0.05). Mechanical measures of surface hardness (Clegg Hammer) and shear were taken before and after subject testing and assessed using a paired t-test (p<0.05). Moisture content was also assessed. Kinematic data were found to be representative of typical running values presented in the literature. While mechanical measures revealed that natural turf conditions were not identical, changes in surface did not yield any significant kinematic differences. The consistent production of ankle and knee joint kinematics with changes in mechanical surface properties could suggest that humans prefer to maintain similar geometries when running on a variety of natural turf surfaces. Alternatively, the mechanical properties of the natural turf conditions may not have been sufficiently different to elicit changes in human response during running. Keywords: natural turf, kinematics, soccer, running.
1. School of Sport and Health Sciences, University of Exeter, Exeter, EX1 2LU, UK - E-mail: v.h.stiles; [email protected] 2. Cranfield Centre for Sports Surfaces, Cranfield University, Bedford, MK43 0AL, UK E-mail: i.t.james; [email protected]
500 The Engineering of Sport 7 - Vol. 1
1- Introduction The World Health Organisation (WHO) and Department of Health (DoH) in the United Kingdom recognizes that participation in sport and exercise activities yields important health and social benefits for the individual and reduced dependence on community primary care health provisions [W1 ; D1]. Promotion and attainment of a healthy nation can be aided by appropriate sports facilities that are affordable and safe. Benefits to health and society can be gained via participation in traditional, competitive sports such as hockey, football, tennis, rugby, cricket and lacrosse at school, club and elite level. However, it is important that in order for these activities to operate with relative ease, thus reaping the health advantages, the condition and provision of sports surfaces are appropriate at a variety of sporting levels. Sports such as tennis, hockey and to some extent soccer have benefited from incorporating artificial surfaces into the game as they provide year round playing opportunity. The influence of adverse weather conditions on a surface’s playing ability is also less affected when playing on artificial surfaces compared to natural turf. An artificial turf surface also requires a lower level of maintenance, can be constructed within a relatively small space and can tolerate regular multi-sport use compared to a natural turf surface [K1; NY1]. Artificial sports surfaces have made an important contribution to the provision of functional sports surfaces and increased sport participation. By comparison, a natural turf surface requires a larger area of ground providing the scope to rotate pitch usage as a method of surface regeneration, does not withstand the rigors of frequent multi-sport use and playing properties are considerably influenced by changes in the weather. There is a need however to continue to develop natural turf surfaces, the reasons for which are two-fold; the protection of green spaces and playing fields in the built environment is crucial and the preservation of fundamental playing characteristics for sports such as soccer, rugby, golf, cricket and lacrosse is paramount. It may be considered that the game of hockey has benefited from its move away from natural turf onto artificial pitches that started in the 1970’s. Modification of the hockey pitch changed the traditional playing characteristics of the game, resulting in certain surface-related skills being lost, but replaced with heightened reaction and agility skills encountered as a result of playing on a faster surface [SL1]. The move from natural turf surfaces to artificial surfaces for sports such as soccer appears to be somewhat more reluctant due to the risk of modifying playing characteristics of the game [U1]. Thus, in order for sports to maintain these playing characteristics, they must remain on natural turf surfaces, placing a greater demand on natural turf surface provision. Surface provision for training and competitive use for sports such as soccer and rugby can only be met if advancement is made in the construction and sustainability of natural turf surfaces. Despite natural turf being a common playing surface for popular sports such as soccer, rugby and cricket, few biomechanical studies have been performed using natural turf conditions. It is suggested that logistical complications of incorporating a natural soil media in the biomechanics laboratory have inhibited progress on understanding how wear and tear of a natural turf surface is achieved during sporting activity and how humans respond to changes in natural turf properties from the biomechanical perspec-
The Engineering of Sport 7 - Vol. 1
501
tives of injury and performance. Some analysis of natural turf properties has been achieved in the field for example during the assessment of traction performance during cutting maneuvers [CL1] and plantar pressures underfoot during sports specific movements [ES1]. However, incorporation of natural turf in the laboratory is important in order to be able to use biomechanical equipment that is either too sensitive or practically inappropriate to be used in an outdoor environment. Typical running, turning and accelerating ground reaction force profiles have previously been derived from subjects performing sports specific movements on a natural turf surface in the laboratory [SD2]. This work forms an initial benchmark for biomechanical analysis of natural turf however one that requires further support and validation from more sophisticated studies. Recently, biomechanical response to variations in natural turf surface quantified using ground reaction force data derived from multiple participants during running and turning in the laboratory has been presented [SD1]. Analysis of ground reaction force data, whilst useful, only quantifies one aspect of the human’s complex ability to respond to changes in turf condition. Numerous studies have used a kinematic analysis to understand how humans respond and alter their lower limb geometry when running with different mechanical properties of shoes and surfaces underfoot [BY1; DD1; DC1]. For example, an increase in initial knee flexion (cushioning flexion) during running has been yielded as a compensatory adjustment when contacting a surface with increased stiffness [BY1]. In addition, a flatter foot (reduced dorsi-flexion) has been observed for barefoot compared with shod running. Limited research is available however that reports kinematic data for participants running on natural turf surfaces. The present study proposes analysis of kinematic response (initial and peak ankle and knee angles and peak angular velocities) to variations in natural turf during running in order to characterise patterns of human movement when running on natural turf. Initial indications of how humans adjust their geometry when running on a variety of natural turf surfaces can also be studied. Variations in natural turf can be constructed via modifying the material properties of the rootzone. Variations in turf construction in-situ can be quantified using mechanical tests. Assessment of mechanical properties of a surface has frequently used devices such as the Clegg Hammer yielding a measure of ‘peak deceleration (g)’ to characterize and monitor natural surface hardness [C1; HB1]. The bulk shear strength of turf can be quantified using a cruciform Shear Vane, which is used in situ to measure un-drained shear strength by the rotation of a cruciform vane to soil failure (, in kN m-2; BS13779, 1990). Soil moisture content is a key factor in soil strength and can be measured using a dielectric probe (e.g. a Theta Probe, Delta-T, Cambridge) that determines the volume of water per unit volume of soil as a percentage (vol%), [GM1]. In keeping with literature evidence regarding changes in the shoe-surface interface, it was hypothesized that the turf condition with the highest mechanical hardness, lowest moisture content and highest shear strength would yield increased magnitudes of initial knee flexion (cushioning flexion) and reduced ankle dorsi-flexion compared to a turf surface with the lowest mechanical measure of hardness, moisture content and shear strength. The additional kinematic variables of peak knee flexion, peak ankle angle and peak ankle and knee angular velocities are assessed to monitor whether humans move consistently when running on a variety of natural turf surfaces in the laboratory.
502 The Engineering of Sport 7 - Vol. 1
2- Methods Sixteen portable plastic trays (0.60 m x 0.40 m x 0.08 m) were turfed with ryegrass in three different soils of a 0.07 m depth (Table 1). The ‘clay’ condition was typical of heavy clay football pitches, the high sand ‘rootzone’ condition was typical of modern, elite natural surfaces and the ‘sandy’ condition provided an intermediate sand-content condition. Trays were positioned sideways in the biomechanics laboratory on non-slip matting (6 mm thick) to form a continuous runway of one surface condition (Figure 1). Within the continuous runway, the target tray upon which participants were required to contact during running was positioned lengthways with additional trays on either side providing a safe area for subjects to pass over. Trays containing the other two turf conditions were positioned inside the laboratory at the same time as the condition constructing the runway and thus being tested. Surface conditions were rotated during a subject testing session as required to construct the testing runway. Before testing, the trays of turf located in the laboratory were mowed to a length of 29 mm. Table 1 - Turf Conditions.
Figure 1 - Laboratory lay-out (top view).
Nine male volunteers were recruited and consented to be participants in the study. Approval for the collection of data from human participants was obtained from the School of Sport and Health Sciences, University of Exeter Ethics Committee. Each participant was either a soccer or rugby player of a university or club standard and regularly participated in training and match playing sessions on a natural turf surface whilst wearing studded footwear. Participants were assigned standard metal studded soccer boots in their size (UK sizes, 10, 11 & 12). All participants wore the same model of boot. Participants were required to visit the laboratory on one occasion to complete running trials on all three test surfaces. During a test session, boot-shod participants were familiarized with the running movement on the runway whilst a spare tray of turf was positioned in the target area of the runway. A constant running speed of 3.83 m.s-1 (± 5%)
The Engineering of Sport 7 - Vol. 1
503
was required between two sets of photocells set 1 m away from the center line of the target tray (2 m distance between photocell sets). Participants were required to make a right-footed contact with the target tray during each trial without adjusting their running stride and rhythm. Starting positions were marked out in the laboratory during the familiarisation trials to assist participants in making contact with the tray without adjusting their running action. Where subjects failed to make contact with the target tray in a natural style or run within the specified speed range, data were discarded and the trial recollected. Kinematic data were collected (Vicon Peak, automatic, opto-electronic system 120 Hz) from 10 successful running trials for each subject on each type of turf condition (total of 30 successful trials per session). After 10 trials on one condition, the target tray was removed and preserved for shear strength analysis. Each target tray was therefore unique to a subject. If not considerably damaged, all other trays of turf were kept for future test sessions in order to create another runway for another subject on another day for one of the three conditions. All trays of turf were relocated outside overnight and on occasion rested with appropriate water maintenance for one or two days depending on laboratory test requirements and obvious turf damage. Three dimensional initial (frame immediately prior to ground contact) and peak ankle and knee joint angles (during stance) were assessed together with peak joint angular velocities (during stance) and respective times of occurrence relative to the start of ground contact. Kinematic data were filtered using a quintic spline, (Peak Performance default optimal smoothing technique using 5th degree quintic polynomials; [W2]). A combined and adapted version of joint coordinate systems presented by Soutas-Little, Beavis, Verstraete and Markus, [SB1] and Vaughan, Davis and O’Connor, [VD1] was employed to monitor joint movement at the ankle and knee. Joint angles were referenced to a relaxed standing position. A positive ankle angle represents dorsi-flexion. Anatomical marker locations are presented in figure 2. Statistical differences between surface conditions were assessed using a repeated measures analysis of variance (ANOVA R.M, p<0.05). Measures of surface hardness (Peak ‘g’) using a Clegg Hammer were performed immediately before and after the subject testing session on the target tray. Three Clegg Hammer test procedures were performed on the test tray in a diagonal formation; corner one (bottom left), center, corner two (top right). The mean of the three tests was taken to represent surface hardness for each tray and was calculated for each surface condition under each movement. Standard deviations are used to indicate the reliability of tray hardness within each surface condition. Volumetric soil moisture content was measured immediately prior to the test session for all trays in the runway using a Theta Probe. Surface mean soil moisture content was calculated and was presented for each movement. From these values, the degree of saturation was calculated as a percentage of the saturation moisture content (maximum volume of voids in the soil). Shear strength of the target tray was quantified prior to and after the subject testing session using a shear vane inserted to a depth of 33 mm. Shear strength is presented in kPa. Surface mean soil shear strength is presented for each movement.
504 The Engineering of Sport 7 - Vol. 1
Figure 2 - Anatomical marker locations and diagram of a marker.
3- Results Group mean data for nine subjects running on three different natural turf surfaces are presented in table 2 together with mechanical test results. Typical knee and ankle angle time histories are presented in figure 3 together with angular conventions. Initial ankle and knee joint angles remain at similar magnitudes with changes in surface. Peak ankle and knee joint angles are also similar across conditions. Ankle and knee ranges of movement (ROM) appear to be lower on the rootzone surface compared to the clay condition however these differences were not significant (p>0.05). Peak ankle and knee angular velocities were similar across turf conditions. Mechanical measures of hardness reveal that the rootzone condition possessed similar magnitudes of hardness (58.34 peak g, ±5.75) compared to the clay condition (59.47 peak g, ±15.91) prior to biomechanical testing. However, the rootzone condition compared to both clay and sandy conditions possessed significantly lower levels of hardness (64.30 peak g, ±9.67 compared to 70.47 peak g ±13.76 and 72.62 peak g, ± 11.83 respectively) after participant test sessions (p<0.05). Measures of shear demonstrate that the rootzone condition consistently possessed lower resistance to shear failure before and after (significant, p<0.05) participant test sessions compared to clay and sandy conditions.
Figure 3 - Typical ankle and knee angle time histories together with angular conventions.
The Engineering of Sport 7 - Vol. 1
505
Table 2 - Kinematic and mechanical mean data (* indicates significant differences at p<0.05 compared to the clay condition).
4- Discussion The present study collected kinematics data from nine participants performing running on three natural turf surfaces in the biomechanics laboratory. Kinematic running data from the present study demonstrate that typical magnitudes of knee and ankle variables have been yielded from participants running on a variety of natural turf surfaces in the biomechanics laboratory. These values compare well to those presented in the running literature [BY1] and therefore the dynamics of running have been satisfactorily reproduced. Initial indications of how humans adjust their geometry when running on a variety of natural turf surfaces have also been studied. Turf wear and soil deformation have been measured using standard techniques for natural turf sports surfaces (BS EN 12231:2003 & BS EN 14954:2005). During the running trials, compared to the other two surfaces, the sandy condition was hardest prior to and after participant testing. The shear data indicate that the rootzone condition had lower values than the clay and sandy conditions. Thus there appear to be distinct differences in the mechanical properties of the three turf surfaces.
506 The Engineering of Sport 7 - Vol. 1 Despite the differences in mechanical properties, kinematic results indicate similar running patterns on the three turf conditions. Based on the study hypotheses, a greater initial knee flexion and lower ankle dorsi-flexion would be expected on the harder sandy surface than on the clay or rootzone. However, similar values were observed for these kinematic variables across all three surfaces. The consistent production of ankle and knee joint kinematics with changes in mechanical surface properties could suggest that humans prefer to maintain similar geometries when running on a variety of natural turf surfaces. Alternatively, the mechanical properties of the natural turf conditions may not have been sufficiently different to elicit changes in human response during running. The lack of difference in running kinematics with changes in turf surface properties does not necessarily indicate that players will move similarly on different surface types for all movements. For example, turning or accelerating on the surfaces may require changes in movement patterns according to the hardness or shear properties of the surfaces. Turning and accelerating movement tasks that result in a change of direction of the performer have been found to yield approximately 4.2 times the magnitude of peak braking force and 3.8 times the magnitude of horizontal (braking) loading rate compared to when running on natural turf [SD1]. This finding therefore supports the notion that a turning manoeuvre if performed at subject self selected sub-maximal speeds result in the participant experiencing higher magnitudes of horizontal loading and rates of loading compared to running at 3.83 m.s-1. Compared to running, turning imparts greater horizontal forces and rates of loading on the turf thus placing greater reliance on the shear strength properties of the surface in order for the participant to successfully and consistently perform the movement in a stable manner. Given the increased need for the participant to utilise mechanical properties of the turf surface when performing a turning manoeuvre, it is suggested that this movement would provide more scope to study kinematic measures of human response with changes in natural turf condition. Future work will assess kinematic response to different turf surfaces when performing a turning movement.
5- Conclusions A kinematic analysis of running on three mechanically distinct natural turf surfaces has revealed that participants maintain similar ankle and knee joint geometries across all surfaces. Future work will assess kinematic response when performing a turning manoeuvre with changes in natural turf condition.
6- Acknowledgements The authors gratefully acknowledge the funding of this research by the Engineering and Physical Sciences Research Council, UK under project EP/C512243/1(P).
7- References [B1] BS1377-9:1990. Soils for civil engineering puposes – Part 9: in-situ tests. Brtish Standards Institute, London. ISBN 0 580 18242 8.
The Engineering of Sport 7 - Vol. 1
507
[BY1] Bobbert, M.F., Yeadon, M.R. and Nigg, B.M. Mechanical analysis of the landing phase in heel-toe running. Journal of Biomechanics, 25(3): 223-234, 1992. [C1] Clegg, B. An impact testing device for in situ base course evaluation. Australian Road Research Bureau Proceedings, 8: 1-6, 1976. [CL1] Coyles, V.R., Lake, M.J., & Patritti, B.L. Comparative evaluation of soccer boot traction during cutting manoeuvres – methodological considerations for field testing. In S.J. Haake (Ed.), The Engineering of Sport (p. 183-190). Blackwell Science: Cambridge, 1998. [DD1] De Wit, B., De Clercq, D. and Aerts, P. Biomechanical analysis of the stance phase during barefoot and shod running. Journal of Biomechanics, 33: 269-278, 2000. [D1] Department of Health At least five a week – evidence on the impact of physical activity and its relationship to health. Department of Health, Physical Activity, Health Improvement and Prevention, 2004. [DC1] Dixon, S.J., Collop, A.C. and Batt, M.E. Surface effects on ground reaction forces and lower extremity kinematics in running. Medicine and Science in Sports and Exercise, 32(11): 1919-1926, 2000. [ES1] Eils, E., Streyl, M., Linnenbecker, S., Thorwesten, L., Volker, K., & Rosenbaum, D. Characteristic plantar pressure distribution patterns during soccer-specific movements. The American Journal of Sports Medicine, 32(1): 140-145, 2004. [GM1] Gaskin, G. J., Miller, J. D. Measurement of soil water content using a simplified impedance measuring technique. Journal of Agricultural Engineering Research. 63: 153-160, 1996. [HB1] Holmes, G., Bell, M. J. A pilot study of the playing quality of football pitches. Journal of the Sports Turf Research Institute. 63: 74-91, 1986. [K1] Kolitzus, H.J. (1984). Functional standards for playing surfaces. In E.C. Frederick (Ed.), Sport shoes and playing surfaces: Biomechanical properties (p. 98-118). Human Kinetics Publishers Inc: Champaign, Illinois. [NY1] Nigg, B.M. and Yeadon, M.R. Biomechanical aspects of playing surfaces. Journal of Sports Sciences, 5: 117-145, 1987. [SB1] Soutas-Little, R.W., Beavis, G.C., Verstraete, M.C. and Markus, T.L. Analysis of foot motion during running using a joint co-ordinate system. Medicine and Science in Sports and Exercise, 19(3): 285-293. 1987. [SL1] Spencer, M., Lawrence, S., Rechichi, S., Bishop, D., Dawson, B. and Goodman, C. Time motion analysis of elite field hockey, with special reference to repeated-sprint activity. Journal of Sports Sciences, 22, 843-850, 2004. [SD1] Stiles, V.H., Dixon, S.J., Guisasola, I.N. and James, I.T. Biomechanical response to variations in natural turf surfaces during running and turning. In: ‘STARSS 2007’, Proceedings of the First International Conference of the SportSURF Network, September 2007, Eds P. Fleming, C. Young, S. Dixon, I. James, M. Carre, C.Walker, ISBN 978-1-897911-30-3, Loughborough University, UK, 2007. [SD2] Stiles, V.H., Dixon, S.J., & James, I.T. An Initial Investigation of Human-Natural Turf Interaction in the Laboratory. In E.F Moritz & S. Haake (Eds.), The Engineering of Sport 6, Vol 3, (pp. 255 - 260). New York: Springer, 2006. [U1] UEFA, Turf first in the Netherlands, (2005) Retrieved from www.uefa.com 06/10/05. [VD1] Vaughan, C.L., Davis, B.L. and O’Connor, J.C. Dynamics of Human Gait. Human Kinetics: Champaign, IL, 1992. [W1] WHO (2003). Health and Development Through Physical Activity and Sport. WHO/NMH/NPH/PAH/03.2, World Health Organization, Geneva, Switzerland.
508 The Engineering of Sport 7 - Vol. 1 [W2] Woltring, H.J. On optimal smoothing and derivative estimation from noisy displacement data in biomechanics. Human Movement Science, 4(3): 229-245, 1985.
Finite Element Simulation of Ice Pick Torquing (P97) Rae S. Gordon1, Kathryn L. Franklin
Topics: Modelling, Climbing. Abstract: The ice axe is an essential tool for any mountaineer climbing in icy conditions. Ice axes are often used for climbing frozen waterfalls and icy rock faces. One of the techniques used in such climbs is known as torquing. This technique involves placing the pick of the ice axe into a crack and pulling on the ice axe handle to wedge the pick in the crack to give purchase for the climber. This technique was initially frowned upon by ice axe manufacturers as the pick was not designed for such loading, but it has become an accepted climbing technique and the manufacturers have modified their product accordingly. There have been a number of ice pick failures due to fatigue, where the crack was initiated at the root of the ice pick tooth. It has been suggested that torquing is responsible for these failures. In this study the ice axe pick was modelled using three dimensional elements. Two cases were analysed: the first a static finite element analysis assuming the tip of the pick is clamped and the second a transient analysis modelling the tip in a crack in the rock. The results showed that in the static case the highest stresses were not at the root of a tooth, but there were tensile stress concentrations in the root of the teeth nearest the handle. The transient analysis revealed a high shear stress at the root of the two teeth nearest the axe handle. In both cases these may result in the fatigue failure of an ice pick with the crack initiating at the root of a tooth. Keywords: Mountaineering, Ice Tools, Stress, Crack Initiation, Fatigue.
1- Introduction Ice axes are a common tool used by climbers and mountaineers in wintry conditions or at altitude where there is permanent snow and ice. Ice axes can be categorised into either walking or technical axes. Walking ice axes are used by mountaineers for fall arrest in the event of a climber slipping or as a “third leg” for extra stability. Technical ice axes are used to climb frozen waterfalls and icy rock faces. An ice axe consists of a handle, a pick and either an adze or hammer. The handle generally has a spike at is base, a rubber grip 1. University of Glamorgan, Pontypridd, Wales - E-mail: rgordon,[email protected]
510 The Engineering of Sport 7 - Vol. 1 and a leash which attaches to the climbers wrist so that the ice axe stays attached to the body. A technical ice axe handle is generally curved in some way and tends to be shorter than a walking ice axe. The pick consists of a sharp point, which is used to strike the ice and a series of serrated teeth with a backwards sweep along the bottom edge which grip the ice as the climber pulls themselves up the ice face. The top surface of the pick is a ridge to aid cleaving the ice to remove the ice axe. The adze, which is the opposite side of the handle to the pick, is used to chip away loose ice. A climber climbing an ice wall uses two ice axes and on the second ice axe the adze is replaced by a hammer, which is used to hammer ice pegs into the ice so that the climber can attach the safety ropes. In many cases the pick, adze and hammer are detachable from the ice axe so that they can easily be replaced if damaged. A typical ice axe is shown in Figure 1. Technical ice axes are primarily used for climbing frozen waterfalls, but as climbers seek new challenges and routes up a mountain often these ice axes are used on icy rock faces this is referred to as mixed climbing. On icy rock faces the ice is not always suitable for using the ice axe in its traditional manner. Climbers have developed alternative techniques for climbing this type of rock face. One of the popular techniques employed by climbers is called torquing. This involves placing the pick of the ice axe in a crack and applying a torque to turn the pick and wedge it at an angle in the crack. This provides an anchor point for the climber, who then pulls themselves up. This technique was initially frowned upon by ice axe manufacturers as the pick was never designed to be subjected to this type of loading. Mountaineering handbooks [FP1] state on the subject of torquing “This is not generally the use intended for the product by the manufacturer, so equipment damage is to be expected if a large amount of mixed climbing is done”. However, as the technique has gained acceptance the manufacturers now consider torquing loads when designing ice axe picks. Technical ice axes suffer a great deal of abuse in use and are subject to a wide range of loading conditions. The two main forms of loading are due to the impact between the pick and the ice/rock and torquing. The failure of ice axe picks tends to be due to either fatigue or buckling [H1]. The failure due to buckling occurs when the pick is not very thick and this subjected to a high impact load. The cause of the fatigue failures is often not so clear, with cracks initiating at the base of the serrations on the bottom of pick. The aim of this paper is to investigate the effect of the torquing load and to determine if it could play a role in crack initiation and subsequent fatigue failure.
2- Literature Review As ice axes are designed to prevent falls from a height they are classed as personal protective equipment and as such are subjected to a British and European standard. The standard that applies to ice axes is BS EN 13089:1999 and this was developed in conjunction with the Union International des Associations d’Alpinisme (UIAA) and associated bodies such as the British Mountaineering Council (BMC). The standard provides a series of static tests for the handle and a static and fatigue test for the pick. The fatigue test consists of placing the free end of a pick in a vice like clamp up to a depth of
The Engineering of Sport 7 - Vol. 1
511
25±0.2mm. The handle is subjected to a cyclic force of between 0 and 80-85N at a distance of 250±1mm from the mid-point of the clamped thickness of the pick. The cyclic load is applied at a frequency between 0.5-2Hz for 50,000 cycles. The axe is assumed to have passed if the pick does not break or any components become loose. This load case is similar to the action of a climber moving the axe up and down to loosen it from the ice. The stress analysis of the pick due to an impact load has been investigated using photoelasticity [HB1]. An experienced climber was photographed using an ice axe to determine the orientation and speed of the ice axe pick on impact with the ice. This study found that the pick was at an angle of 20±3º to the normal and that it was travelling at 9.3±1.3 ms-1 just prior to impact. A test rig was developed to mimic the action of the climber and to produce repeatable strikes. Initially the pick struck a block of ice, but the ice cleaved upon impact and this was replaced by balsa wood. The ice axe pick was manufactured from EN24 steel and coated in 2mm photoelastic material. The study found that the pick acted like a beam in bending. Unfortunately no fringe orders could be obtained in the area around the root of the serrated teeth and any stress concentrations in this area would be missed. It was concluded that the impact loading alone was unlikely to cause failure of the pick and other load cases such as torsion should be considered. The BMC encourages its members to report incidents of equipment failure to its technical committee. These failures are then investigated and recommendations made to the manufacturer and to other members if necessary. One such equipment failure [A1] was investigated for a fatigue failure of an ice pick. The pick had failed during on a mixed climbing route. It was determined that the pick had been used repeatedly on such routes. The crack had initiated at the root of one of the teeth. Examination of the ice showed that it was free from defect and that it was considered that the failure of the pick by fatigue was due to repeated flexing (torqing). A finite element simulation of the impact of an ice pick with a semi-rigid surface has been investigated [GF1]. The premise of this study was that with a sufficient impact load the compressive stress concentrations at the root of the serrated teeth would exceed yield and result in a tensile residual stress. The repeated use of the pick would then result in a tensile-compressive stress pattern and hence conditions for crack initiation and growth would exist. The finite element model consisted of a two-dimensional representation of the pick which was fixed at the handle. A block of material of equivalent to the mass of the ice axe, but with the stiffness of granite was impacted against the pick using the speed and orientation defined by Haake et al. 1997. This study showed that if yield occurred in compression at the root of a tooth, then conditions for fatigue could exist. The issue of the effects of a torquing were not addressed. The British and European standard does not seem to include any tests which examine the case when the pick is subjected to a torsional load. The BMC [G1] suggested in a report to the UIAA that torsional tests should be carried out on ice axe picks. Initial work carried out involved clamping the free end of the pick to a depth of varying between 25 and 50mm and then applying a torque of between 100 and 150Nm. It was finally recommended that the free end of the ice axe pick be placed in a vice to a depth
512 The Engineering of Sport 7 - Vol. 1 of 40mm and a torque of 50Nm be applied to the shaft without significant permanent deformation and that a 150Nm torque be applied to the shaft without fracture. It would seem that this recommendation was not acted upon.
3- Finite Element Analysis In this study two finite element analyses were carried out to examine the effect of a torquing load on the ice pick and to determine if this could cause a crack to initiate at the root of a tooth. The first analysis carried out was a static analysis model with the free end of the pick being clamped and then subjected to a torsional load. The second analysis was a dynamic analysis where the pick was placed in a gap between two surfaces and rotated to simulate the actual torqing process.
3.1 Static Analysis Model The first case to be examined using finite element analysis was based on the torsional test recommended by [G1]. The free end of the pick was clamped to a depth of 25mm. This was chosen over the recommended 40mm as it was thought to be a more extreme load case. A three-dimensional model of the ice pick was created and meshed using 10 noded tetrahedral elements in ANSYS V11. It was assumed that the handle would be much stiffer than the pick and was therefore modelled using multi-point constraint (MPC) (i.e. rigid) elements. The MPC elements were defined with beam like behaviour. A MPC element was created from each node on the end of the pick and linked to a single node at the centre of the ice axe head. The handle length (330mm) was modelled using a MPC single element from this point. A point load of 151.5N was applied at the free end of the handle so that it would produce a torque of 50Nm. The material properties used were those of steel with Young’s modulus of elasticity of 210GN/m2, Possion’s ratio of 0.3, yield stress of 750MN/m2 and a tangential modulus of 151MN/m2.
3.2 Static Analysis Results The von-Mises stresses for the fully clamped model can be seen in Figure 1. This shows that the point of highest stress (the point where yield will occur first) is at the top of the pick at the position of clamping and has a value of 263MN/m2. This does not indicate high stresses at the root of the teeth, the area where fatigue cracks tend to form. As the von-Mises stress is always positive it does not distinguish between high tensile or compressive stress. Therefore, the 1st principle stress was examined as this shows the most tensile stresses. Although, the highest principle stress is at the top of the pick where it attaches to the handle (this is possibly due to use of MPC’s), there is another area of high stress at the top of the pick near the clamped end there is evidence of a high tensile stress at the root of the last two teeth nearest to the ice axe handle. This can be seen in Figure 2.
The Engineering of Sport 7 - Vol. 1
513
Figure 1 - The von-Mises stress (MN/m2) distribution due to a torsional load on a fully clamped pick.
Figure 2 - Tensile stress (MN/m2) (1st Principal) concentration at root of teeth.
3.3 Dynamic Analysis Model The second case to be examined was based on a slightly more realistic scenario, in that in reality the pick will have some movement within the crack in the rock. It has to be borne in mind that cracks come in all shapes and sizes, but in this case a vertical crack
514 The Engineering of Sport 7 - Vol. 1 is being modelled. This is possibly an extreme condition as the pick will be wedged in this type of crack using the torsional load exerted by the climber alone [FP1]. On applying a torsional load the pick will turn in the crack and contact will be with the tips of the teeth on one side and the top of the pick on the other. This may result in an alternative stress distribution, which will indicate possible fatigue failure nearer to the tip of the pick. The pick geometry used in this analysis was similar to that of the static analysis. The difference being that the sharp point at the front of the pick and the fillet radius at the top of the pick were removed. Half of the head and handle were modelled, these were assumed to be stiffer than the pick. The geometry was then meshed using 10 noded tetrahedral elements and the surfaces were meshed using 8 noded brick elements. The pick, which was taken to be 4mm thick, was placed midway between two flat vertical surfaces separated with an overall distance of 9mm. The distance between the surfaces was arbitrarily chosen. The free end of the pick was placed between the two surfaces to a depth of 25mm, so comparisons could be drawn with the first analysis. As the ice axe would initially be moving with a rigid body motion LS-Dyna was used for this analysis. The surfaces were defined as being rigid, but with a stiffness equivalent to that of the pick. A node to surface contact model was used. It was assumed that there was no friction between the pick and the rigid surfaces, as would be the case if there were a thin layer of ice in the crack. Prior to the finite element analysis, an ice axe was used to investigate how the pick would contact with the surfaces of the gap. This investigation resulted in two possible scenarios, the first was that as the climber torques the ice axe the pick turns in the crack and both the top and bottom of the pick wedges simultaneously at which point the load induces bending. The second scenario was that the as the climber moves the ice axe one face contacts one surface of the crack and the pick then turns in the crack pivoting about a line at the front of the crack. This continues until the point of the pick comes into contact with the opposite wall, whereupon the pick rotates about the front point until it wedges. It was decided that the first scenario be modelled. From observation it was determined during torquing the axe rotates about an axis at the top of the handle. To simulate this, the model of the axe was given an initial rotational velocity of 20rad.ss about that axis, this is an assumed value. The results obtained from this analysis will only provide an indication of the positions of the maximum stresses. The back face of the handle was restrained so that it remained parallel with the front face of the wall.
3.3 Dynamic Analysis Results Figure 3 shows the displaced shape of the pick as it is wedged between the two rigid surfaces. It can be seen that contact is made between the first two teeth on one side and by a line of material at the crack edge at the other. This shows the difference between the assumed condition of restraint of the first case and what actually happens in practice. The von-Mises stress was examined and showed that the area of highest stress was at the root of the two largest teeth nearest the ice axe handle in a similar position to those of the first analysis (Figure 2). However, as this does not indicate the nature of maximum stress the principle stresses were examined. In all cases the maximum stress occurred where the pick joint the handle. The individual stress components were then investigated
The Engineering of Sport 7 - Vol. 1
515
and it was found that the stress concentration was due to shear. The shear stress distribution is shown in Figure 4.
Figure 3 - Displaced shape as pick wedges against the two rigid surfaces.
Figure 4 - Shear stress (N/m2) distribution for the pick.
4- Discussion of Results It can easily be seen from the two analyses carried out the behaviour of the ice axe is quite different. The fully clamped model shows that the highest stress occurs in the area where the pick is restrained. This is as expected, however this is not repeated when the
516 The Engineering of Sport 7 - Vol. 1 pick is subject to a torsional load in a physical gap. In both models a stress concentration occurs at the root of the two large teeth nearest the handle, however in the fully clamped model this is masked to a large extent by the stresses around the clamped area. In the model where the end of the pick is clamped it is subject to a bending moment as well as the torsional load. This bending load may not be present in practice depending upon the technique of the climber. The dynamic analysis shows that the stress concentration is mainly due to the shear, once again this would seem reasonable. In both these cases there are stress concentrations present at the root of a tooth which may result in a fatigue crack initiating. In the study of impact loading on the pick [GF1] it was found that if compressive yield was exceeded a residual tensile stress occurred on the teeth close to the free end of the pick, hence giving the condition for a fatigue failure. This compares with this study which shows that a fatigue crack is more likely to initiate at the root of teeth nearer the pick handle. This would suggest that in the investigation of a fatigue failure the type of load that has caused failure may be determined using finite element analysis. As different ice pick profiles were used for these two studies, further work should be carried out to compare these two models on the same profile. Further work should be carried out on the dynamic model to provide more realistic boundary conditions, i.e. an accurate determination of the rotational velocity of the ice axe when used by a climber.
5- Conclusions The modelling of the stresses on an ice axe pick due to torquing using finite element analysis is more realistic when the tip of the pick is placed in a gap between two surfaces rather than clamping the end. This type of clamping restraint could never be experienced on a mountain. The most likely location for a fatigue crack to be initiated on the base of the pick is at the teeth nearest to the handle of the axe, but this may vary if there are changes in section along the pick. The finite element method could be used in a failure investigation to simulate the two main types of loading experienced by an ice axe pick to determine which contributed most to the failure.
6- References [A1] R. Allen. Technical Committee Memorandum TCM98/8, The British Mountaineering Council, Manchester. 1998 [B1] BS-EN 13089:1999 Mountaineering Equipment – Ice tools – Safety requirements and test methods. British Standards Institute. 19999 [FP1] A. Fyffe and I. Peter. The handbook of climbing, Pelham Books, London. 1995 [GF1] R. Gordon and K. Franklin. Finite element simulation of ice axe impacts on a semi rigid surface. In The Engineering of Sport 6, Vol. 2 Development for disciplines. Ed E.K. Moritz and S. Haake. Springer, New York. 2006
The Engineering of Sport 7 - Vol. 1
517
[G1] N Grandison. Technical proposals for testing ice axes for discussion by UIAA ice axe working group. British Mountaineering Council, Manchester, Technical Note 85/9. 1985 [HB1] Haake, S.J., Blackwood, W.D. and Yoxall A. Photoelastic analysis of ice axe impacts. Proceedings of SPIE International Society of Optical Engineering, 3111: 482-489, 1997 [H1] T. Helen. The engineering stability of metallic climbing equipment. In The science of climbing and mountaineering, Ed N. Messanger, W. Patterson and D. Brooks. Human Kinetics. 1999
A Sociological Analysis of a Controversy in French Sport Science Field: How to Manage Teams Specialising in Technological Innovation (P99) Philippe Terral1, Cécile Collinet2
Topics: Social Sciences. Abstract: A Sociological Analysis of a Controversy in French Sport Science Field. This paper aims at presenting a scientific controversy in the world of French sport sciences. It focuses upon electrical stimulation, a technique used to increase muscle with an electrical device sending electric impulses into the muscle. This technique was the subject of many scientific research studies, but the results stemming from them are contradictory, and thus create a typical controversy. This controversy involves several categories of actors, which this paper identifies while studying the type of arguments developed to impose one’s point of view. Through the analysis of 50 scientific papers and 15 interviews of the main researchers involved in the controversy, this paper studies the social processes at work in the construction and resolution of the controversy. The latter engages various conceptions of scientific research, and particularly enhances the conflicts between fundamental and applied science. The controversy is also grounded upon axiological positions and values, notably various conceptions of competitive sport, or the relationship between research and the sport industry. In addition, this paper shows how the conflicts can be better understood if one considers the researchers’ social stances in the French sport science field and the interests associated with these stances. This article is relevant to the field of sports engineering as it delineates the origins of the tensions between the different actors in the world of sports and, in doing so, it offers perspectives to optimise these actors’ relationships. In this respect, sociological studies provide useful knowledge on how to maximize the management of teams specializing in technological innovation. Keywords: France – researchers – controversy – electrical stimulation – muscle strength.
For about thirty years, a close analysis of technical objects has stressed the appearance of a precise technique devoted to the training of high-level athletes – electrical stimulation (ES). It was soon used as part of reeducation programs either in medicine or in physiotherapy. ES “consists in bypassing the nervous system by sending an electrical stimulation directly into the core of muscles” (Le Chevalier, Pradet 2003, p.55) thanks to an electric apparatus – the stimulator. It is said to work efficiently for enhancing muscular strength 1. SOI, Université Toulouse III, 11 rue Froideterre 31200 Toulouse, France, - E-mail: [email protected] 2. GREHSS, Université Marne La Vallée (Paris Est) - E-mail: [email protected]
520 The Engineering of Sport 7 - Vol. 1 and endurance but also for a certain number of other applications (recovery, muscular warm-up and so on).
1- Problematic and Theoretical Stances The electrical stimulator as an electrical object and, above all, ES as a technique for improving athletic performance, prove to be highly stimulating research items for anybody concerned with scientific and technical controversies. As ES was transplanted from the medical area to the world of sports, various research works were carried out to assess the effect of ES on athletic improvements and more particularly on the development of muscular strength. In the 1980s and 1990s, ES was definitely given a sporting orientation by coaches and athletes, as well as physiotherapists and other sports training “technicians”, which strongly aroused the interest of scientists studying high-level athletes or, more generally, the development of strength. From then on, ES became a scientific focus leading to contradictory results and creating a real controversy between different scientific and technical protagonists, all related to sports training. In keeping with Chateauraynaud and Torny (1999), we want to show that this controversy, strictly scientific at the beginning, gradually encompassed non scientific considerations, especially as more and more actors took part in it. The purpose of this sociological work is to delineate the origins of the tensions between the different actors in the world of sports in order to optimise their relationship and more generally, to maximise the management of teams specializing in the production and commercialisation of technological innovations. Several social groups got involved in the issue connecting ES to high-level sports. The scientific protagonists were at the heart of the controversy that started in the 1980s and lasted for about ten years. This group was not homogeneous and was composed of very diverse actors – professors from UFRSTAPS and the INSEP1, but also therapists – physicians and physiotherapists – specialized in sports. Moreover, another category of people took great interest in this issue – employees from the firms that sold electrical stimulators. Although they belonged to different institutions, these two groups were not independent but interacted a lot (for example, the firms ordered scientific researches and used them to promote their product). Furthermore, we must also take into account the group of sport technicians. Coaches, as well as athletes were, to some extent, involved in the controversy too. In this article, we would like to emphasize the overlapping of three forms of justifications that fuelled the debates among the sport scientists interested in ES. Thus, we will see that this controversy tackles the notions of scientific truth, of technical efficiency but also of moral fairness, and this, through the different representations of the nature of training and sport practice but also through the relationships that must be established between science and industry. We will first introduce the controversy by considering the categories (notably institutional) of the actors involved and the content of the discourses. We will next focus on the epistemic grounds of the controversy (i.e. 1. The UFRSTAPS (Training and Research into Physical and Sport Activity Units) are an integral part of the French University. The INSEP (National Institute of Sport and Physical Training), dedicated to the training of high-level athletes, is supervised by the Ministry of Sports.
The Engineering of Sport 7 - Vol. 1
521
the question of the scientific validity and the technical efficiency of the declarations), then on its more axiological grounds through a study of the representations of the nature of training and sport practice but also of the relationship that must be established between science and industry.
2- Method We chose to start our study with the scientific field since it is where the controversy stems from. We first apprehended the debates through a literature review on the subject. We consulted the articles written by French researchers in books and periodicals - both French and foreign. We referred to about fifty articles. We were able to detect different types of “forum” (resources and authorities through which the contradictors can make their point of view stand out). Indeed, we could identify on the one hand a formal forum with theorizations, experimentations, publications in reviews as well as congress communications and on the other hand, an informal forum with articles for the layman, or promotional materials. After reading the whole range of articles, we contacted the researchers to interview them. We singled out fifteen researchers according to their publications. The exploratory interviews together with the readings enabled us to determine a certain number of questions that finally resulted in a questionnaire. The latter permitted us to survey the researchers’ involvement in ES research (its origin, scientific working modes, and so forth), the acknowledgement by the researcher himself of the controversy and his own positioning, the impact of ES on a certain number of physiological modifications, the researcher’s opinion on the role and use of ES in training and, finally, the collaborations among researchers. Eventually, some interviews were held after receiving the questionnaires in order to shed light on certain answers.
3- Results and Discussion The exploration of the data linked to sports enabled us to pick out the first articles written by French authors in the middle of the 1980s and to notice a boom in production between 1989 and 1993. The first foreign scientific works on the subject go back to 1970 (Krcka and Zrubak 1970, Kots and Chwilon 1971). The controversy developed over the discrepancies between the results of the foreign works and those of new experiments.
3.1 The Protagonists of the Controversy The interviews and the questionnaires revealed two modes of implication from the part of the scientists. It was the increasing use of ES in training that aroused the interest of a first category of scientists. Their implication was generally built around contacts with sportsmen - coaches and athletes - or with the firms that sold ES products. A second category of scientists came through their own research works (on strength for instance) to study particular processes to develop muscles (and thus ES). These researchers either set forth researches highly connected to the social demand of the world of sports or, on the contrary, devoted themselves to more academic research perspectives, even called “fundamental research” in certain circles. The first results produced in the 1980s were contradictory (Miller and Thépaut-Mathieu 1985, Cometti 1988, Joly 1989) and gene-
522 The Engineering of Sport 7 - Vol. 1 rated oppositions. In the interviews or questionnaires, the researchers themselves alluded to the constitution of groups with differing opinions. It is noteworthy that these groups were related to very specific institutions. We can single out three main groups of scientists taking part in the debates. The first one principally dealt with performance enhancement thanks to ES. These researchers were mainly from the Sport University of Dijon and belonged to the CEP (Performance Evaluation Centre), from the Jean Monnet University in Saint-Etienne and from a few provincial universities (Nice or Clermont-Ferrand). The most important researchers were G. Cometti, N. Maffiuletti, G. Pousson, J.C. Chatard, L; Martin, S. Colson. Another group, on the contrary, questioned the interest of such a technique to enhance performance. This second group was composed of researchers from the INSEP: C. ThépautMathieu and C. Miller (the work by S. Morth on the role of ES regarding recovery is to be linked to this group). Then a third group can be delineated – that of therapists, with mainly two physiotherapists (M. Pujo and B. Joly) and some physicians (Questel, Stéphan). The members of this group were not in touch with one another and worked in different geographical areas: the INSEP for Pujo, a physical therapy centre in Alsace for Joly. Nevertheless, they all agreed that ES had some importance for training either to increase muscle strength or to recover after efforts. It is worthy of note that Joly and Cometti did work together (Joly notably took part in the writing of the 1988 book by Cometti). This division into three groups was relevant up to 1997, when the book about the increase of strength by Thépaut-Mathieu was published. The following period seems different with a huge series of publications by the first group demonstrating the positive effects of ES and this, up to the early 21st century (notably Maffiuletti et al. 2002). By then, the number of articles by the second group questioning the interest of ES had already decreased.
3.2 The Content of the Controversy When studying a scientific controversy on a sociological level, one cannot but start with paying a closer attention to the theoretical problems. That’s why we are going to sum them up briefly now. ES aroused a whole range a scientific questions. As we surveyed the literature on the topic and the questionnaires, we determined the central issue arising from the research: the effects of ES on the improvement of muscular strength. Contradictory positions were held. Generally speaking, the scientific works on the subject pointed out that ES could develop the athletes’ maximum strength: “Training with ES brings about a significant development of muscular strength” (Miller 1990, p.5). However, the impact of this development was also apprehended differently: “diverse” for the INSEP researcher, or “impressive” for Cometti, for example: “Thus, we can assert that working with ES is efficient and dramatically improves strength” (Cometti 1988, p.292). More precisely, the question evolved around the comparison between on the one hand, the effects of training with ES, and on the other hand, the effects of bodybuilding exercises, or “in voluntary contraction (VC)” exercises. Answers were various. To some, the results were identical: “When we compare the increase in strength through VC or through ES, we can observe that the results can be identical” (Le Chevalier 2003, p.52 – an INSEP researcher). To others, the results with ES were inferior: “We must first acknowledge that
The Engineering of Sport 7 - Vol. 1
523
from a quantitative point of view, training with ES is less profitable than training with VC. Moreover, it presents a greater individual variability” (Thépaut-Mathieu 1997, p.341). Finally, a certain number of experiments tended to show a greater gain of strength when athletes are electrically stimulated (notably Maffiuletti – professor in Dijon – et al. 1999). It was thus difficult to get a clear-cut assessment. If it is important to understand the nature of the scientific debates on ES, it is even more important in this sociological approach, to determine how the diverse conceptions in this controversy are grounded upon a larger array of arguments.
3.3 The Controversy Grounds In a scientific controversy, the protagonists confront their arguments and methods to determine, assess, and interpret phenomena (Chateauraynaud, Torny, 1999: 80) and to impose their point of view. Two groups of arguments can be detected here, in keeping with two central grounds of controversy, namely epistemic and axiological grounds.
3.3.1 Epistemic Grounds While a single norm of truth or scientific validation could have enabled people to decide for one group or another, the truth of the different results is still to be clearly defined. This blurring impression is due to three factors revealing the problematic establishment of stable scientific results: the literature used to back the thesis researchers want to support (in favour or against ES), the lack of experiments, and their diversity, making research comparisons difficult. Then, we will also see that the epistemic grounds of the controversy stem from two different ontological conceptions of scientific truth, enhancing a conflict between “fundamental” or more “applied” research. Controversy and Modality of Presentation of the Scientific Results It was the foreign literature that offered us a first outlook of points of view. Concerning the detractors of ES, the literature revealed the uncertainty of the experimental results. Thépaut-Mathieu thus put the emphasis on the diversity of results concerning the effects of ES on muscular hypertrophy (Calbric et al. 1987, Currier et al. 1983, Gobelt et al. 1989, Halbach et al. 1980, Romero et al. 1982). Those in favour of ES also used foreign works but chose to put the stress on the positive results. This is particularly what Cometti did in his book in 1988 (p. 225) when he focused on the works by Kotz in the USSR, Anzil, Modotto, Zanon in Italy, Portman in Canada, Kopanski, Klepacki and Jaszczuck in Poland and Moreno-Aranda and Sereig. Moreover, the difficult definition of stable scientific results constitutes the second category of arguments developed in the controversy. Thus, detractors often criticised the too hasty conclusions of some researches due to a lack of experiments. When dealing with the supposed effects of ES, a majority of answers in the questionnaires and interviews underlined the lack of work on each of these effects. On average, four to six objectives assigned to ES were left without attested research. The third type of epistemic argument centres on the one hand on the denunciation of the heterogeneity of the research methodologies at work, and on the other hand on
524 The Engineering of Sport 7 - Vol. 1 the differences of opinion concerning the modes of interpretation of the results. Indeed, numerous researchers reckoned the heterogeneity of the methods prevented comparisons and impaired their knowledge of the effects of ES: “There are a lot of contradictions, principally because of the heterogeneity of the protocols, of the subjects, of the stimulators used and of the muscles stimulated” (extracted from a questionnaire). This statement gave birth to a whole range of rhetorical figures depending on the chosen points of view: on the whole, this heterogeneity was used to contest the efficiency of ES or on the contrary, to question the different criticisms of this technique. These different arguments are developed in order to support and impose a particular stance. They also refer to modalities of construction of truth and proof, notably through the different nature of the experiments at work. Ontological Dimensions of the Research Modalities Beyond the problematic establishment of stable results – after all, an intrinsic problem of scientific activity – our investigations lead us to consider more ontological dimensions within the epistemic grounds of the controversy. Among others, we think of the different conceptions of scientific truth (Terral 2003). Thus, we distinguished two visions enhancing two modes of validation of statements. Thépaut-Matthieu and Miller (1987, 1990, 1992, 1993) discussed the validation modes of the pro ES research. They tried to prove the scientific dimension of their works; a dimension they easily questioned concerning some scientific works developing the positive effects of ES on training. Their experimentation modes turned out to be very “academic” and “fundamental” in the sense that they isolated the effect of a small number of variables to study their mutual influences in a laboratory. The researchers in favour of ES also proposed such experiments in a laboratory. Yet, even if we can notice more “fundamental” works about the influence of ES on the muscle (more precisely on the sural triceps: Martin, 1993), we can also find numerous other works of a different nature, we could qualify as more “applied”. These works did not simply assess strength improvement or analyse the different muscular adaptations thanks to ES but were also applied to sport practices such as basket-ball (Maffiuletti, Cometti, Martin, 1999; Maffiuletti, Cometti, Amiridis, Pousson, Chatard, 2000), or swimming (Pichon, Cometti, 1994, Pichon, Chatard, Martin, Cometti, 1995). The specificity of these studies was to consider the effects of ES on athletic performance as it contributes to gain strength: “We wanted to see if the modifications of strength obtained thanks to ES resulted in modifications of performance in sports practices” (interview of a CEP member). Such research modes were criticised by the first group of researchers regarding the empirical procedures used to measure the impact of ES on performance enhancement: “I asked for the procedures but I have never got them”, “How do we pass from one (fundamental research) to the other (applied research)? (Extracted from the interviews). They considered such procedures as non scientific since the diverse variables are not isolated – in the way they are in laboratory research – and since as a consequence their respective effects cannot be detected. We believe what is at stake here are two ontological conceptions of scientific truth. The scientists engaged in “applied” research have developed validation modes of know-
The Engineering of Sport 7 - Vol. 1
525
ledge close to those of sports actors and cannot be compared to those of the scientists dedicated to more “fundamental” forms of research. Indeed, these forms are in line with a “realist” conception of truth (to quote Engel 1998), considering a statement is true as long as it is faithful to the real and is based on a description. On the contrary, the researchers engaged in “applied” science are consistent with a pragmatic conception of truth in the sense that they implicitly think the validity of a declaration doesn’t depend on its connection to an exterior reality but is determined according to practical consequences for training. This is in keeping with the theses of James (1975) who stated that truth is an event that depends on an idea. An idea becomes truth through facts. This verification through facts must be apprehended as a practical consequence that suits and satisfies us (for example, a performance enhancement that would be ascribed to the effects of ES without any thorough experimental verification and isolation of the effects of the diverse variables). These ontological conceptions of truth generate two opposite visions of the relationship between science and technique. Some people think the efficiency of a technique depends on its connections to scientific works. Others reckon that technique can do without science, and that the practice of training is sufficient to justify the use of ES.
3.3.2 Axiological Grounds The controversy we are dealing with develops scientific debates involving epistemic positions but also evolves around other types of arguments, more axiological ones. Chateauraynaud and Torny (1999) consider the passage from controversies (more purely scientific) to polemics when controversies start involving more actors and comparing extra-scientific arguments corresponding to different “visions of the world”. We can precisely observe a diversification of arguments and actors engaged in the controversy over ES. Coaches and athletes, therapists, as well as businessmen actively take part in the controversy. The arguments propose different “visions of the world”. On a precise scale, they offer representations of training and on a larger scale, of sports practice, and also of scientific activity in its relationship with what we will generally call industry. These two types of representations generate oppositions partly grounded upon ethical referents that seem to us at the core of the controversy. “Natural” or More “Artificial” Representations of Training and Sport Practice If we pay attention to the thesis of Thépaut-Mathieu, we notice that ethical problems arise several times (1997, 2000). Enhancing muscular endurance entails heavy workouts (several hours a day every day and for several weeks) which raises “ethical issues (…) in the case of such a routine in the physical preparation of an athlete” (Thépaut-Mathieu 1997, p.361). Beyond these ethical issues lies a representation of sports and training advocating the rejection of artificial processes. ES is indeed often dealt with in relation to the question of doping. The very opinion of athletes on the subject confirms this vision of sports as the fruit of both effort and work, a perspective very close to Pierre de Coubertin’s values (Duret and Trabal 2003, p.68). Some researchers interviewed underlined the incompatibility between the demand for stimulation (in time and intensity) and the representation of sports as the fruit of an active work, of efforts and of perseverance. Thus, the answer of some researchers to the questionnaire is very clear: “If we take
526 The Engineering of Sport 7 - Vol. 1 literally the definition of doping that establishes any artificial practice as a doping practice, I’d be tempted to consider ES as doping. Now, there are contradictory arguments. Indeed, if ES is considered artificial, what about bodybuilding? And here we are confronted to the problematic definition of an artificial practice. As far as I am concerned, I would nevertheless consider this practice as doping, but how can we test and control such a practice? There is no detectable trace of ES ”. On the other hand, the researchers in favour of training with ES, strongly reject the possibility that such a practice might be associated to doping. Cometti, in 1988, devoted his first chapter to the question: “Obviously, we do think that electrical stimulation has nothing to do with doping” (p.194). Moreover, the artificiality of training is much debated: “ We often use the word “artificial” to qualify work with ES. It wouldn’t be a natural means. If we refer to actual modern training, we notice that athletes work with 120 to 150% (of their maximum strength), now, is it really natural? The very notion of “naturality” has no relevance in the world of high-level sports”. Training is no more a natural practice fuelled by effort and talent to get the best performance but it now uses the most sophisticated devices to push back the limits of human possibilities. Cometti and the CEP clearly support such a vision and their research aims at increasing the possibilities of ES in order to “exceed the maximal voluntary strength” (we quote Cometti 1988). Autonomy of Science or Close Connection to Industry Private companies have been associated to the research on ES from the start. Like Collingridge and Reeve (1986), we tend to think that the relationship between research and lobbies is a crucial factor in the development of scientific controversies. The questionnaires and interviews collected revealed a need for industrial subvention from the part of certain researchers. In this light, a researcher from Dijon told us that because he lacked financial means for his research, he had no choice but to “contact stimulator manufacturers in the hope of making them give us equipment, or even finance our research with the argument that they could use their product in the sports environment”. We also noticed that businessmen ordered research to laboratories; indeed, an INSEP researcher confessed that Compex (the market leader in the high-level sport industry) had offered them a contract (which was turned down). Moreover, some actors involved in the polemic sell their own electrical stimulator themselves: this is the case of M. Pujo from the INSEP for example or a researcher interviewed who took part in a television shopping program on the subject. Eventually, some researchers (notably INSEP ones) deny such a link. Thus, the results of the CEP regarding the strength enhancement of certain athletes highlight the efficiency of the “Compex” stimulators (Ratton and Cometti 1990, Gillet and Cometti 1990, Champion and Pousson 1991, Martin and Cometti 1991). At the same time, these data are used by firms for their advertising campaigns. This phenomenon is of great importance as far as the controversy is concerned since firms have taken over real and registered scientific works and have made them public without stressing the scientific controversy. Therefore, they contributed to undermine it in the larger sphere of ES users by making its distribution easier. After developing, with a whole network of researchers, the question of the impact of ES on muscular development and
The Engineering of Sport 7 - Vol. 1
527
its efficiency regarding performance enhancement, the CEP team wove another network of alliances with the industrial world that spread and promoted their achievements. This network was only possible around a profit-sharing scheme, which was in fact two-fold: on the one hand, scientists got financial support and on the other hand, the useful aspects of the technique were highlighted, which eased its distribution. This process fuelled the controversy as the vision of the world conveyed by Compex and supported by some works (but that went far beyond it) conflicted with the one proposed by other scientists. They denounced the link between pro-ES researchers and the industry field as well as the types of relationship we have already alluded to. Indeed, if some researchers reckon that being linked to a commercial environment may not question the credibility of the research – as it merely enables them to finance this research -, others argue that it reinforces the impression of a lack of scientific rigor and of distortion of the works carried out in the process of extension-translation already quoted. Thus, the controversy also entails an evaluation of the scientific and independent dimensions of the research which is heightened when some actors sell their own products themselves. One of the INSEP researchers we interviewed underlined that their will to remain independent had for instance led them to have an apparatus built to suit their own experiments (this apparatus was never intended to be sold ).
4- Conclusion: Identifying the origins of the controversies in order to better manage teams specialising in technological innovation This sociological article pointed out that, due to various causes, tensions were generated around the use of a technical object (ES). We demonstrated that these tensions originated from different scientific results but also from non scientific considerations. This last statement is particularly interesting if one wishes to optimize the management of teams specialized in the production of technological innovations. Indeed, our research showed that it is necessary, as much as possible, to form teams which are culturally homogeneous when dealing with scientific and technological work conception or the representation of sport training. This cultural homogeneity seems to be the adequate basis for an efficient coordination of teams. This research also emphasized a useful principle for the management of diverse groups of actors: when faced with conflict, it is wise to locate the origin of the tensions in each actor’s broader conception. Indeed, each individual has a personal vision of the world which is often unconscious and which, as a consequence, cannot be changed easily. Therefore, the main task of the manager consists in making the actors’ broader conceptions evolve gradually so as to make them compatible. In doing so, he can achieve a compromise between the team members and guarantee an efficient development of the project.
528 The Engineering of Sport 7 - Vol. 1
5- Sociological References (the scientific articles on ES by the researchers quoted in our article are not included here) [CT1] Chateauraynaud F. and Torny D. Les sombres précurseurs : une sociologie pragmatique de l’alerte et du risque. Paris, Editions de l’EHESS, 1999. [CR1] Collingridge D. and Reeve C. Science speaks to power : the role of experts in policy making. New York, St Martin’s Press, 1986. [DT1] Duret P. and Trabal P. Le dopage dans le cyclisme professionnel : accusations, confessions et dénégations. In STAPS, 60 : 59-73. 2003. [E1] P. Engel. La vérité. Paris, Hatier « optiques », 1998. [J1] W. James. The Meaning of Truth. Harvard University Press (1ère édition en 1909), 1975. [T1] P. Terral. La construction sociale des savoirs du monde sportif : Sociologie des conceptions épistémiques. Thèse de doctorat en sociologie, Université Paris IV – Sorbonne, 2003.
Biomechanical Ingredients Measurement: A New Vision-Based Approach (P102) Mohammad Reza Mohammadi1, Hadi Sadoghi Yazdi2
Abstract: In this paper, gait and race is analyzed in various velocities. Biomechanical ingredients as position, velocity, acceleration, torque, force and power are measured using a novel approach in human motion tracking. Six velocity are considered in this study include 1.5, 3.5, 6, 8, 10, and 12 Km/h. The proposed method is developed to capture markers positions associated with human motion obtained from video data. A data-driven predictor is used for tracking of wearing reflective markers. The uncertainty and occlusion of detected markers increase the noise in captured positions. For solving these problems, we proposed a suitable approach which includes a memory based data-driven system. The result of tracking algorithm is tested over 12 sport men in 6 velocities over more than 30000 captured frames. Simulation results validate the analysis and ensuing method. Keywords: Biomechanics, motion analysis, video, tracking, occlusion.
1- Introduction Human gait is an important subject in various researches. Some of them attend to gait as a biometric feature in human identification problem [1]. Human gait recognition has attracted growing attention in video-based applications [2]. Recent research has shown that individuals have distinctive and special ways of walking and that human gait recognition has many advantages as human gait is a biometric feature that may be captured from a great distance and gait has the advantage of being unobtrusive. Nowadays, biomechanical studies use camera-based capture systems for determination of human motion in a room-based system. In this paper, a new tracking algorithm is presented for reducing of occlusion effects in human motion capturing. Object tracking in video streams has been a popular topic in the field of computer vision. Tracking is a particularly important issue in human motion analysis since it serves as a means to prepare data for pose estimation and action recognition. In contrast to human detection, human tracking belongs to a higher-level computer vision problem. 1. Department of physical education and sport science, Tarbiat Moallem University of Sabzevar, Sabzevar, Iran E-mail: [email protected] 2. Department of Engineering, Tarbiat Moallem University of Sabzevar, Sabzevar, Iran - E-mail: [email protected]
530 The Engineering of Sport 7 - Vol. 1 However, the tracking algorithms within human motion analysis usually have considerable intersection with motion segmentation during processing. Tracking over time typically involves matching objects in consecutive frames using features such as points, lines or blobs. That is to say, tracking may be considered to be equivalent to establishing coherent relations of image features between frames with respect to position, velocity, shape, texture, color, etc. Our focus is on discussing various methods used in the tracking process, so different tracking methods studied extensively in past work are summarized as model-based tracking, region-based tracking, active-contour-based tracking, and feature-based tracking. Because of can marker tracking categorized in feature-based tracking group, so we survey existing features tracking techniques. This tracking method uses sub-features such as distinguishable points or lines on the object to realize the tracking task. This type tracking is useful in the presence of partial occlusion; some of the sub-features of the tracked objects remain visible. Feature-based tracking includes feature extraction, feature matching and prediction algorithm.
2- Motivation We divide tracking problem to two problems: filtering and prediction. If filtering is perform using suitable model-driven as maximum a posteriori (MAP) probabilistic method upon captured real-time data and previous stored predicted data, we expect this new filtering process have good operation. Predicted data are collected when the predictor has been converged. So with selecting data-driven predictor and model-driven filter, a new mixed version of tracking algorithm is proposed. In the present work suitable combination of RLS and MAP algorithm, is obtained and it can be proved that power of this error is the less than the RLS algorithm. It is also applied in human motion analysis in the field of sport. Finding optimum solutions of performances in the field of sport have been developed using computer. Among, vision-based measurement can be done precisely using machine vision techniques. In the next sub-section we survey literature in the vision-based human motion analysis as we will find power of the proposed algorithm. A new tracking algorithm which has properties of model and data driven tracking algorithm is presented in this paper. The remainder of this paper is organized into three sections. Section 2 includes the proposed algorithm. In Section 3, the proposed tracking algorithm is applied in human motion analysis and results are discussed and final section proceeds to conclusion.
3- Methods 3.1 The proposed memory-based filtered input RLS algorithm In this section, a new method for reducing the noise effect in tracking algorithm is presented and then the noise statistics is computed by a recursive method. We combine a model-based smoother and recursive least square, RLS, algorithm as predictor. The proposed algorithm, namely, memory-based filter input RLS, MFX-RLS algorithm will
The Engineering of Sport 7 - Vol. 1
531
be applied to human motion analysis. In MFX-RLS algorithm, maximum a posteriori estimator, MAP, and RLS algorithm are used which is appeared in [7, and 8].
3.2 The MFX-RLS method The RLS algorithm in the prediction configuration, tracks the noise similar to signal. This is because of the filter is data-oriented and its output is a linear combination of the input signal. For canceling the noise algorithms such as Kalman filtering consider a model for the generation process and observing the signal. The Kalman filter in the made of tracking moving objects, demand a model of the object motion in linear form. If there is not a motion model or a process model, the application of this algorithm will be difficult. On the other hand, this causes the vulnerability of the algorithm against the process y(n)(the estimaand observation noise. In this paper, the purpose is to track the signal ^ tion of the original signal) instead of the signal x(n). For this purpose, we use an estimator which does not need to process model. The MAP estimator use mean and variance of original signal and noise. Initially we assume mean and variance of noise to be constant and known and in the next section, with a recursive method, will be calculated. The MAP estimator requires the mean and variance of the original signal and noise, but at tracking because of the existence of only one sample of the data can not calculate it. For this purpose, we predict M next state by the RLS predictor and save them. Then in each time step, we apply the input signal (with states that so far were predicted by the RLS filter) to the estimator, Figure 1. As the block diagram of the system of Figure 1 shows, the input signal x(n) is merged with the information taken from a 2Dimmensional table (Figure 2) and is applied to the MAP estimator as a data vector.
Figure 1 - The MFX-RLS algorithm
The structure of two-dimensional buffer is illustrated in Figure 2. As shown in Fig 2, in the first time 2-D buffer is empty and the MAP algorithm deliver input data directly to the RLS algorithm. But, after receiving 6 samples after convergence of the RLS algorithm, 2-D buffer approximately is full and the MAP estimator can estimate received data using stored predicted buffered data.
3.3 Applying of MFX-RLS in tracking In this section, we will show first the performance of the MFX-RLS algorithm and its superiority in marker prediction in human motion analysis.
532 The Engineering of Sport 7 - Vol. 1
Figure 2 - The structure of two-dimensional buffered data.
3.4 Application of the MFX-RLS in the prediction of the position in biomechanics Marker Tracker is one of the biomechanical motion analysis methods. In this method, markers are installed on specific points of human body. Many problems are exist in tracking of markers as follows, a) Occlusion conditions are caused losing of markers in some of duration time. We used the MFX-RLS algorithm which predicts marker position in a few frames. b) Reflection of light from peripheral equipments and no uniformity of gray value of markers. We encountered to this problem after data collecting. In our experiments, 18 markers are used which some of markers have different light. Of course, light of shirt of samples were another problem. The above problems are different types of noise which are observed in marker tracking that is an incentive for using adaptive filtering for noise reduction. These problems are caused marker tracking is converted to an outdoor tracking problem. The tracking algorithm has following section, a) Capturing of marker position in the first frame using suitable GUI1. b) Matching 7*7 buffered sub-image in consecutive frames using correlation method. Best search area is 7*7 windows. c) After convergence of the MFX-RLS for each marker, the predictor/smoother trajectory is used for prediction of the position and smoothing of trajectory for finding correct velocity and acceleration information’s. The MFX-RLS algorithm help for prediction of loosed markers in 5-20 frames depend on the loosed position relative to turning point. d) The GUI software provides for user which guides tracking or it provides HCI2 for user in the markers tracking. As it is seen eighteen markers are shown in Fig 3. We have used negative image for suitable showing of marker position. Results of tracking of marker no.13 have been shown in Fig 4 and Fig 5 shows 71 tracked frames. Effect of smoothing is seen in Fig 6. Smoothed trajectory is important for finding position without shaking and correct velocity and higher order of derivation of position. For finding the performance of the 1. Graphic User Interface 2. Human Computer Interaction
The Engineering of Sport 7 - Vol. 1
533
smoother, we collect a data set of manual main positions and mean square error is obtained for calculating of effect of smoother. The performance is obtained from (1), (1) Where, (2)
(3) Where, index of I is Ideal, index of S is smooth version and index of nS is no smoothed trajectories. N is number of captured samples. The performance is obtained over collected data set. The specifications of data set are, • 12 sport men moved over captured tool with conditions which have been mentioned in Table 1, 2. • The camera with 350 FPS3 with gray images. • Each man moved in 6 different velocities. Table.1 - Individual features of sport men
Table.2 - Anthropometry specifications in sport men.
1 Standard Deviation
For measuring robustness of the proposed tracker we added different level uniform noise to x,y position. As shown in Fig 7, the proposed MFX-RLS algorithm obtained better result for SNR less than 30dB. One of result of designed software is film of motion skeleton and combination of negative film and skeleton which in Fig 8, 9 and 10 numbers of constructed images have been shown.
3. Frame Per Second
534 The Engineering of Sport 7 - Vol. 1
Figure 3 - Eighteen markers in negative image.
Figure 4 - The tracked position of 13th marker.
Figure 5 - Trajectory of 13th marker in 71 frames.
Figure 6 - Smoothing of a) X position using the MFX-RLS algorithm, b) Y position
The Engineering of Sport 7 - Vol. 1
535
Figure 7 - The performance per SNR
Figure 8 - Motion skeleton in 5 frames from 50 tracked frames.
Figure 9 - Motion skeleton in 10 frames from 500 tracked frames
Figure 10 - Motion combined skeleton and negative frames in 10 frames from 500 tracked frames
Biomechanical ingredient measurements After tracking of markers as shown in Fig 3, we extract biomechanical ingredient from these positions and anthropometric specifications (Table 2). Biomechanical ingredients include position, velocity, acceleration, force, torque, power, work, and energy are measured. Biomechanical ingredients in 1.5 Km/h velocities for each part of body are
536 The Engineering of Sport 7 - Vol. 1 captured and for example we show one of them. In Fig 11, biomechanical ingredients about hip joint during gait have been shown.
Figure 11 - Biomechanical ingredients about hip joint.
4- Conclusion In this paper, biomechanical features are extracted using tracking system. New tracking algorithm helps for obtaining position in noisy environment and occlusion problem. The convergence and tracking behavior of the proposed method are proved. Obtained results showed the proposed algorithm gave better performance in tracking of motion body in biomechanical application.
The Engineering of Sport 7 - Vol. 1
537
5- References [1] A. K. Jain, A. Ross,and S. Prabhakar, ,”An Introduction to Biometric Recognition,” IEEE Trans. on Circuits and Systems for Video Technology, vol. 14, no.1, pp.4-20, Jan. 2004. [2] L. Wang, H. Ning, T. Tan, and W. Hu, “Fusion of Static and Dynamic Body Biometrics for Gait Recognition,” IEEE Trans. On Circuits and Systems for Video Technology, vol.14, no.2, pp.149158, Feb 2004. [3] S.Haykin, Adaptive Filter Theory, 3rd-ed, Printice Hall,1996. [4] S.Vaseghi, Advanced Signal Processing and Digital Noise Reduction, John Wiley & Sons Ltd, 1996. [5] J. Badenas, J. Sanchiz, F. Pla, Motion-based segmentation and region tracking in image sequences, Pattern Recognition 34 (2001) 661–670. [6] N. Peterfreund, Robust tracking of position and velocity with Kalman snakes, IEEE Trans. Pattern Anal. Mach. Intell. 22 (6) (2000) 564–569. [7] H. Sadoghi Yazdi, M.Lotfizad, M. Fathy, “ Car Tracking by Quantized Input LMS, QX-LMS algorithm in Traffic Scenes,” IEE Signal Processing, Feb. 2006. [8] H. Sadoghi Yazdi, M.Lotfizad, E. Kabir, M. Fathy, “Clipped Input RLS Applied to vehicle Tracking,” Eurasip Journal on Applied Signal Processing 2005, vol.8, pp.1221-1228. [9] S.Haykin, A.H.Sayed, J.Zeidler, P.Yee, P.Wei, “Tracking of linear Time-Variant Systems,” Proc. MILCOM, pp.602-606, San Diego, Nov. 1995. [10] J.W.Lee, I.Kweon, “MAP-Based Probabilistic Reasoning to Vehicle Segmentation,” Pattern Recognition, Vol.31, No.12, pp.2017-2026, 1998.
How Optimal Baseball Swings Change for Three Levels of Play (P103) Ann Chase1, Mont Hubbard1, Chris Ray2
Topics: Baseball Abstract: Hitting homeruns is a key factor in the sport of baseball. This paper presents an adaptation of a previous baseball impact and flight model to understand how optimal swing parameters (resulting in maximum range) vary for three levels of baseball play: Little League, high school, and college. The two parameters optimized describe the bat’s path to the ball: E, the undercut distance between the center of the bat and ball; and y, the bat swing angle. The optimization determines what values of E and y result in maximum range for each play level. The model consists of two main components: the analysis of the bat-ball impact and a flight simulation of the ball. The ball properties after the bat-ball collision were the initial conditions used in the flight analysis of the batted-ball range. The optimal values of E and y were found by iterating the above calculations until the largest range was found. This method was repeated using parameters consistent with each play level. Optimal strategies differ significantly for different levels of play. Generally, at higher levels of play, the optimal values of E and ψ decrease. Furthermore, the college-level optimal range was slightly larger than the optimal range results of Major League level players reported in the original model. This result suggests that the difference between aluminum (hollow) and wooden (solid) bats plays a significant role in batted-ball range. Keywords: bat-ball collision model; dynamic flight simulation; optimal control; play level.
1 - Introduction Over 30 million people play baseball, some starting as young as five years old. These players are diverse, ranging in age, size, and skill level. They experience vastly different conditions: bat mass and moment of inertia, swing speed, and pitch speed and spin. Yet, these players have a common goal: to hit home runs. Thus, it is natural to ask how the optimal home run swing strategies change, if at all, through the levels of the participating baseball population. 1. University of California, Davis, USA - E-mail: achase, [email protected] 2. Saint Mary’s College of California, Moraga, USA - E-mail: [email protected]
540 The Engineering of Sport 7 - Vol. 1 Sawicki et al. [SH1] investigated the maximum range (home run) strategy of Major League baseball players by quantifying the bat’s path to the ball using two parameters: E, the undercut distance between the center of the bat and ball and , the undercut angle (Figure 1). They found that there is an optimum swing strategy which is dependent on pitch type. An E of 2.64 cm (1.04 04 in.) and a of 9.13 deg., will result in the maximum range for a typical Major League fastball when swinging with a bat speed of 67 mph. Furthermore, an optimally hit curveball will result in a larger range than an optimally hit fastball, demonstrating the importance of spin-induced lift. This investigation adapts the model of Sawicki et al. [SH1] to determine how the parameters dependent on play level affect the optimum swing strategy. Little League level players use smaller and lighter bats made out of aluminum rather than wood. Should they swing differently than a Major League player does? We determine what equipment parameters and kinematic variables change among Little League, high school, and college; measure and estimate these parameters and variables; and finally determine the optimal swing strategies for each of three play levels.
Figure 1 - The kinematic properties of the bat-ball collision.
2- Methods The optimal swing strategies were found by using a mathematical model, scripted in MATLAB, which calculated the range of the ball by analyzing the bat-ball collision and the batted-ball flight as two sequential problems. The collision calculation used a rigidbody impact analysis to return the batted-ball velocity vbf, launch angle , and angular velocity bf given the pitched-ball velocity b0, arrival angle , and angular velocity b0 (Figure 1). These results were used as the initial conditions for the flight analysis, which consisted of a numerical integration of the flight equations of motion describing the batted ball. The forces acting on the ball in flight are gravity, lift, and drag. While gravity is constant, both lift and drag are proportional to velocity squared. Thus, the flight analysis model calculated these forces dynamically. The collision analysis depends significantly on play level as it requires the equipment parameters, or physical properties, and kinematic variables, that describe the bat and ball motion. The bat parameters were determined by measuring the physical properties of three typical bats, one for each play level (Table 1). Because the ball does not change between amateur and professional levels of play, the same ball parameters were used as
The Engineering of Sport 7 - Vol. 1
541
in Sawicki et al. [SH1] . The kinematic variables were estimated by using velocities and scaled spins typical to fastballs for each level of play (Table 2). Table 1 - Bat Properties.
Table 2 - Ball Properties.
As summarized by Sawicki et al. [SH1], it is assumed that the ball makes contact with the center of percussion of the bat (Nathan, 2000). This point of impact ensures that the energy lost due to vibration is minimized and that there is no impulsive reaction with the batter’s hands. Because the impact does not occur at the center of mass of the bat, the effective inertia of the bat at impact must be used by calculating the effective mass of the bat from (1) where MB is the effective mass of the bat, M’B is the actual bat mass, z is the axial distance between the impact point and center of mass, and kT is the centroidal transverse radius of gyration, determined by indirectly measuring the transverse moment of inertia for each bat. The period of small oscillations for each bat was found by suspending the bat as a compound physical pendulum at the knob of the handle. The moment of inertia about the bat’s center of mass is related to its period of oscillation by (2) where g is gravitational acceleration and rcm is the distance of the center of mass from the knob of the bat. The bat was assumed to be a cylindrical shell, which has an axial moment of inertia (3) where rB2 is the barrel radius. Although the coefficient of restitution, e, is higher for an aluminum versus wooden bat-ball collision, this model assumes that e is only a function of the relative bat-ball [HG1]. Finally, it was assumed that the bat had no initial axial angular velocity during impact.
542 The Engineering of Sport 7 - Vol. 1 The flight analysis returned the range of the batted ball by dynamically calculating the lift and drag forces and integrating the equations of motion of the ball. The lift force, L, is given by (4) where is air density, A= rb2 is the frontal area of the ball, Vr is the relative wind speed, the difference between the wind and ball velocity, and b is the ball angular velocity. The dimensionless lift coefficient, CL, is usually determined experimentally and found to be a strong function of the spin parameter, S ([WF1], [AH1]) (5) The drag force, D, acting on a baseball is given by (6) where CD, the dimensionless drag coefficient, is also found experimentally but is mostly a function of Reynolds number, Re ([AH1]), [B1]). The Re is the ratio of inertial forces to viscous forces acting on the ball and is given by (7) where rb is the ball radius and m is the dynamic viscosity of air. From Eqns. (5) and (6), it is clear that CL and CD are not constant. Rather, they are dependent on variables that change throughout flight. In fact, [F1] noticed that there is a sudden decrease in the magnitude of CD for a given Re known as the drag crisis and that this decrease occurs at speeds observed in baseball. Because CL=f1(S) and CD=f2(Re), fits of the CL and CD data available from the literature are used within the lift and drag calculations, respectively [SH1]. The optimal swing strategy was found by calculating the maximum range and returning the E and used in that calculation. A script was written to call the pre-defined MATLAB function ‘fminsearch’ for swing speeds ranging from 20 mph to 120 mph. The same set of bat swing speeds was used for all levels for comparison purposes. We emphasize that such swing speeds are not within the capabilities of all players. The entire process was completed six times using two typical linear pitch velocities for each of the three amateur play levels (Table 2). The pitched angular velocities were assumed to be proportional to the linear velocity and are all negative, corresponding to fastballs.
3- Results Range is an increasing function of bat velocity (Figure 2a). This agrees with intuition; the faster one swings the bat, the greater the batted-ball velocity and the farther the ball will travel. Bat speed has more influence on range than any other single variable. While
The Engineering of Sport 7 - Vol. 1
543
there is only a small increase in range over the two pitch speeds within a given play level, the optimal range increases noticeably between play levels. For example, the Little League (40 and 50 mph) curves are nearly overlapping while there is a significant range difference between the Little League and high school (50 and 60 mph) curves. This range difference must be due to the sizable difference in bat parameters between Little League and high school. As stated earlier, range depends on spin-induced lift, which explains the increase of hit ball angular velocity with both bat velocity and play level. (Figure 2b). The magnitude of both variables comprising the optimal swing strategy are dependent on play level (Figure 3a and 3b). Both E and ψ decrease in magnitude as players move up in play level, increase their swing speed, and use a bat with greater mass and inertia.
Figure 2a - Optimal range as a function of bat velocity.
Figure 2b - Optimal hit ball angular velocity as a function of bat velocity.
Figure 3a - Optimal undercut distance, E, as a function of bat velocity.
Figure 3b - Optimal bat swing angle, ψ, as a function of bat velocity.
544 The Engineering of Sport 7 - Vol. 1
4- Discussion Figure 3 clearly shows that the optimal swing strategy does in fact change with play level. The trends of the curves suggest that a Little League player should swing with a larger E and than a college level player for a given swing speed. Given that achievable bat swing speeds are so much lower for smaller, less experienced Little League players, their optimal E and are actually much larger than college level players. Note that the differences between the two extreme play levels of E and is approximately 0.3 in and 7 deg., respectively. Thus, the deviations in optimal swing strategies are small and require a player of significant skill level. Because of the complexity of the problem, it is difficult to attribute local maximums and minimums to a single independent variable. The algorithm receives four input variables (ball linear and angular velocity, ball arrival angle, and bat velocity) and then optimizes two additional parameters (E and ). It is impossible to attribute particular characteristics of the curves to one variable or another. However, the jaggedness of the curves is likely due to the optimization process. It is difficult to determine the exact location of a maximum during the iteration process. Careful observation of Figure 2a reveals that college-level players are able to produce slightly larger optimal ranges than Major League players at the same bat speed. This difference is initially surprising considering that the pitch speed used in the Major League study [SH1] was 4 mph faster than the larger collegiate one used here (90 mph). This result is explained by Figure 2b, however, which shows college-level players achieving larger optimal batted-ball angular velocities and thus greater spin-induced lift. The difference in bat composition between Major League and college play levels explains the discrepancy between the batted-ball angular velocities. College players use aluminum bats which have roughly twice the axial moment of inertia as the wooden bats used in the Major Leagues. This difference between the axial moments of inertia allows the aluminum bats to impart a larger angular velocity to the batted ball during impact. The effect of axial moment of inertia on the optimal swing strategy is currently under further study. While the axial moment of inertia is not the only difference between aluminum and wooden bats, it is the only one that results in range increase. The larger actual coefficient of restitution of an aluminum bat cannot account for the increased range because this difference was neglected; it was assumed that the coefficient of restitution was only a function of the relative bat-ball velocity. Furthermore, the collegiate aluminum bat’s smaller mass and larger transverse moment of inertia (compared to the wooden bat used in [SH1] both decrease the effective mass of the bat, thus decreasing the resulting range. This suggests that the axial moment of inertia plays a larger role in dictating the range than the other differences between aluminum and wooden bats.
5- Conclusions In summary, this investigation has determined what bat swing parameters change with play level, estimated magnitudes for these parameters using experimentation and common knowledge of the game, and modified the model of Sawicki et al. (2003) to use these values in models for bat-ball impact and subsequent ball flight. It was found that
The Engineering of Sport 7 - Vol. 1
545
the optimal swing strategy is dependent on play level. The model’s flight analysis is realistic because it dynamically calculates the lift and drag forces, thus accounting for the dependence of the lift and drag coefficients on the spin parameter and Reynolds number, respectively. However, the impact analysis assumed that the bat was a hollow cylinder, which does not account for the smaller handle radius and tapered shape of the bat. Furthermore, the optimal range results suggest that further investigation into the effect of bat axial moment of inertia and increased coefficient of restitution of aluminum bats on the optimal swing strategy is warranted.
6- References [AH1] Alaways L. and Hubbard M. Experimental determination of baseball spin and lift. In Journal of Sports Sciences, 19:349-358, 2001 [B1] Briggs L. Effect of spin and speed on the lateral deflection (curve) of a baseball; and the Magnus effect for smooth spheres. In The American Journal of Physics, 27:589-596, 1959. [F1] Frolich C. Aerodynamic drag crisis and its possible effect on the flight of baseballs. In The American Journal of Physics, 52:325-223, 1985. [HG1] Hendee S.P., Greenwald R.M., and Crisco J.J. Static and dynamic properties of various baseballs. In The Journal of Applied Biomechanics, 14:390-400, (1998). [NA1] Nathan A.M. Dynamics of the baseball-bat collision. In The American Journal of Physics, 68:979-990, 2000. [SH1] Sawicki G., Hubbard M., and Stronge W. How to hit home runs: optimum baseball swing parameters for maximum range trajectories. In The American Journal of Physics, 71:1152-1162, 2003. [WF1] Watts R. and Ferrer R. The lateral force on a spinning sphere: aerodynamics of a curve ball. In The American Journal of Physics, 55:40-44, 1987.
Graduated Compression Stockings and Delayed Onset Muscle Soreness (P105) Stéphane Perrey1, Aurélien Bringard1, Sébastien Racinais1, Kostia Puchaux2, Nicolas Belluye2
Topics: Exercise physiology, muscular function, stockings, engineering processes. Abstract: Delayed onset muscle soreness (DOMS) is a common experience following unaccustomed eccentric exercise. DOMS and associated force deficits may limit optimal performance in subsequent days. The cause of DOMS remains poorly understood, thus there is no effective treatment. Graduated compression stockings (GCS) are a commonly used intervention believed to diminish DOMS. The purpose of this study was to determine if GCS after eccentric walking exercise minimizes DOMS and associated deficits (e.g. muscle force capacity). Eight healthy subjects (age 26±4 yrs, height 175±8 cm, weight 70±5 kg) volunteered to perform a single bout of backward downhill walking exercise (duration 30 min, velocity 1 m.s-1, negative grade -25%, load 12% of body weight). Following walking exercise, subjects were required to wear 5 hours per day for 3 consecutive days GSC (SupportivTM) on one leg while the second was used as control. Muscle soreness and neuromuscular measures (Mwave, peak twitch, maximal voluntary torque or MVT) were taken pre and postwalk, then 2, 24, 48 and 72 hours post-walking exercise for the two legs. There was a 28% reduction in DOMS 72 h after exercise when wearing GCS (P<0.05) than in the control leg. Immediately after exercise there was a 15% decrease in MVT of the plantar flexors in both legs partly attributable to an alteration in contractile properties (-22% in electrically evoked mechanical twitch). In leg wearing GCS, MVT starts to recover while the contractile properties had significantly recovered within 24 h but not in the control leg. In the current study, GCS might have had the effect of compressing the muscle tissue to such an extent that less structural damage occurred relative to a control condition. GCS accelerated the recovery of the muscle force capacity at 24 hours beyond that achieved by the control condition. Keywords: stockings, tissue management, muscle damage, soreness, textile.
1. Faculty of Sports Science (UFR STAPS), EA 2991 Motor Efficiency and Deficiency Lab, 700 avenue du pic Saint Loup, 34090 Montpellier - E-mail: stephane.perrey, [email protected]; [email protected] 2. Decathlon Test and Research Center, 4 Bd de Mons 59650 Villeneuve d’Ascq, France - E-mail: kostia.puchaux, nicolas.belluye}@decathlon.com
548 The Engineering of Sport 7 - Vol. 1
1- Introduction Both recreational and serious athletes have experienced delayed onset muscle soreness (DOMS), pain, and stiffness following unaccustomed exercise or increased training workload. Although research on exerciseinduced muscle soreness dates back to 1902 (Hough), the exact mechanisms of soreness and pain, as well as the accompanying responses remain unclear. These responses include a prolonged loss in range of motion and strength, increases in muscle enzymes in the blood, swelling and structural damage. Several studies have shown that eccentric muscle actions result in greater soreness, fatigue and damage than isometric or concentric contractions. This explains why some activities such as hiking that incorporates a large degree of eccentric contractions result in considerable soreness while others, like cycling, that are not biased towards eccentric contractions produce little soreness. Exercise models developed to study muscle damage and soreness are those where eccentric contractions predominate such as downhill running and backward downhill walking (Nottle and Nosaka 2005). The latter allows the study of the effects of DOMS on alteration of muscle function of the triceps surae region. Compression stockings and tights are considered by many athletes to be beneficial for recovery treatment and related exercise symptom relief and are now very popular. One of the most recognized actions of compression tights during recovery relates to DOMS prevention. The increased microcirculation provided by compressive garments may prevent post-exercise damage and pain by reducing oedema and helping compensate for impaired venous return (Jonker et al. 2001). Two studies have shown that compression garments maintained muscle function and reduced perceived muscle soreness following eccentric exercise (Kraemer et al. 2001a, Kraemer et al. 2001b). These studies also showed that compression garments attenuate CK release from skeletal muscle into the circulation following eccentric exercise. Although compression is advocated in the recovery from exercise induced muscle damage (Noonan and Garrett 1999), there is little information on the effect of compression on intracellular metabolic function. The most convincing evidence comes from magnetic resonance spectroscopy that showed elevated skeletal muscle metabolites phosphodiester in the thigh 1-h following eccentric injury caused by 30 minutes of downhill walking compared to the control thigh (Trenell et al. 2006) but without changing perceived muscle soreness. Overall, these data suggest that wearing compression garments in the recovery from eccentric exercise may alter the inflammatory response to damage and accelerate muscle repair processes. Bringard et al. (2006) recently showed in sportsmen a significantly lower venous pooling (smaller blood volume changes measured with near infrared spectroscopy) and a higher calf oxygen saturation level with compression tight in comparison to elastic tight and shorts when subjects were lying supine and during upright standing. It appears that in comparison to elastic tights, compression tights have positive effects on muscle oxygenation and venous function at rest, and could be useful to oxygenate damaged muscles after exercise. In this context, wearing elastic compression stockings in 63-year-old sportspeople during an 80 min recovery period between two maximal 5 min exercises led to reduction in hematocrit and lactate concentrations. In this study, a 2.1% performance enhancement was found during the second bout of exercise (Chatard et al. 2004).
The Engineering of Sport 7 - Vol. 1
549
Therefore, the goal of this study was to determine whether the impaired muscle function and DOMS following a single 30-min bout of backward downhill walking exercise is attenuated by wearing compression stockings 24 h, 48h and 72h after.
2- Materials and methods 2.1 Subjects Eight healthy young men participated in this study. Their average age, height, and weight (means ± SD) were 26 ± 4 years, 175 ± 8 cm, and 70 ± 5 kg, respectively. All subjects were physically active and exercised 1 – 2 times a week, but they had not used an intense workload for their own resistance-type exercise for the last 3 years. Subjects were not allowed all forms of exercise throughout the period of the experiment. There is a wellknown repeated bout effect that performance of first bout of eccentric exercise will cause an adaptation such that there is less damage in the second bout (for a review, see McHugh 2003). Subjects gave informed, written consent to participate in this study. The procedures complied with the Helsinki declaration for human experimentation and were approved by the local Ethics Committee. None of the subjects suffered from muscle soreness or ankle injuries. Subjects were asked to avoid caffeine intake within the 8-h preceding the test and to avoid all vigorous activity during the 24-h preceding the test. Subjects were also asked to refrain from analgesic intake all along the protocol, which could have disturbed DOMS perception.
2.2 Experimental procedures Subjects visited our laboratory on four consecutive days. The first day consisted of a neuromuscular test session (described subsequently) on both legs followed by a backward walking exercise (description below) followed by further neuromuscular testing. On the second, third and fourth days, subjects returned to the laboratory at the same hour of day that they had finished the walking exercise and performed the neuromuscular testing for both legs. Immediately following exercise and until the end of the experiment, graduated compression stockings were worn 5 hours per day at 2 h, 24 h, 48h and 72 h following backward walking exercise. Graduated compression stockings (GCS) covered the calf on one leg (SupportivTM, France). The leg wearing GCS was randomly assigned (dominant versus non-dominant as determined by the non-preferred leg to kick a ball). Subjects exercised by walking on a motorized treadmill (S2500, HEF Techmachine, France) for 30-min at a constant velocity of 1 m.s-1 with a negative grade of -25%. To increase the eccentric loading on the plantar flexor muscles, the walk was performed backwards (Nottle and Nosaka 2005) whilst wearing a vest loaded with an additional weight equivalent to 12% of body weight. All the neuromuscular tests began with the determination of the stimulation intensity required to elicit a maximal M-wave amplitude (Mmax). Afterwards, three Mmax interspaced by 8 s were elicited from the relaxed muscle. The amplitude of the three twitches evoked at Mmax intensities were averaged for subsequent analysis.
550 The Engineering of Sport 7 - Vol. 1 Thereafter, subjects were instructed to perform three maximal voluntary torque (MVT) contractions of the plantar flexor muscles, each for 5-s. Subjects were verbally encouraged to perform maximally. Finally, a doublet (two electrically evoked twitches, 10 ms apart, Mmax intensity) was evoked during each plateau (superimposed twitch) and another doublet was evoked 4-s after each MVT (potentiated twitch). The ratio of the amplitude of the superimposed twitch torque over the amplitude of a twitch evoked at rest 4 s after the MVT was used to assess the level of voluntary activation (VA) where VA (%) = (1- Superimposed Twitch/Potentiated Twitch) x 100. A subjective evaluation of the extent of DOMS in the plantar flexor muscles was performed before each neuromuscular test by completing two subjective scales for evaluation of DOMS. The first was a visual scale of 9 cm without any graduation (horizontal line ranging from no pain at the left to extreme pain at the right).
2.3 Measurement and calculations The MVT of the plantar flexor muscles was recorded by a dynamometric pedal (Captels, St Mathieu de Treviers, France). Subject position was standardized with hip, knee and ankle angulations of 90°, and foot securely strapped on the pedal. The tibial nerve was stimulated with a cathode electrode with a diameter of 9 mm placed in the popliteal cavity (Contrôle Graphique Medical, Brie-Comte-Robert, France). Subjects were in a standardized position with motionless head and a standardized environment (i.e., same time-of-day, silent room, constant lighting). Furthermore, a constant pressure was applied to the electrode with the use of a strap. This was controlled by an air pressurerecorder (Kikuhime, TT MediTrade, Soro, Denmark). The anode (10 x 5 cm, Medicompex, Ecublens, Switzerland) was positioned distal to the patella. Electrical stimulations (400V, rectangular pulse of 0.2 ms) were delivered by a high-voltage stimulator (Digitimer DS7AH, Digitimer, Hertfordshire, England). The amperage was adjusted for each subject during the familiarization session. During this first session, the amperage was increased progressively (10 mA increment) until a plateau in twitch mechanical response (peak twitch, Pt) and Mmax were observed. Electrophysiological responses to the evoked action potentials were recorded on the soleus and gastrocnemius medialis muscles with bipolar Ag/AgCl electrodes (Contrôle Graphique Medical, BrieComte-Robert, France) with a diameter of 9 mm and an interelectrode distance of 25 mm. The reference electrode was placed on the wrist. Low impedance between the two electrodes (< 5 k) was obtained by abrading and washing the skin with emery paper and cleaning with alcohol. Signals were amplified and filtered (band pass 30 Hz – 500 Hz, gain = 1000), and recorded at high frequency (2000 Hz). The action potentials were recorded using MP30 hardware (Biopac Systems Inc., Santa Barbara, California, USA) and dedicated software (BSL Pro Version 3.6.7, Biopac Systems Inc., Santa Barbara, California, USA). The same equipment was also used to drive the stimulator. From the mechanical response obtained, we calculated Pt considered as an index of the contractile properties while Mmax amplitude represents an index of sarcolemmal excitability. Statistical analyses were performed with Systat software (Systat, Evanston, IL, USA). Data are reported as mean ± SEM and the level of statistical significance was set at P < 0.05.
The Engineering of Sport 7 - Vol. 1
551
3- Results The MVT significantly changed across the 3 days following the walking test in both legs (P < 0.02). Post hoc analysis showed a significant decrease in MVT after the walking exercise and which persisted during the next two days (P < 0.01). A significant recovery in MVT was observed on the third day (48-h versus 72-h after, P < 0.02). An enhanced recovery of 6 % in MVT for the leg with GCS was observed early during the recovery but failed to achieve statistical significance. In line with the evolution observed in MVT, post hoc analysis showed a significant decrease in VA level after the walking exercise (pre versus post-exercise: P < 0.02), which failed to recover by 48-h (post-exercise versus 24h and 48-h after exercise, NS). However there was a significant recovery by 72-h (postexercise versus after 72-h, P < 0.005). There was no difference between the two legs for the VA level. The electrically evoked Pt also displayed a significant variation following walking exercise (P < 0.001) and legs. Post hoc analysis revealed a significant decrease in Pt after the exercise (P < 0.001) followed by a significant recovery thereafter (P < 0.001). The recovery was better at 24 h for the leg wearing GCS. The walking exercise failed to induce significant changes in the evoked potentials at rest (Mmax) for any experimental conditions. The subjects feeling of DOMS increased significantly in the days following the walking exercise (P < 0.001). Post hoc analysis displayed significantly higher subjective DOMS for the three days following exercise compared to the termination of exercise (P < 0.001). Muscle soreness reached a maximum 48-h after exercise and began to recover by 72-h (48-h versus 72-h after exercise, P < 0.01). At 72 h, DOMS were significantly lower with the leg wearing GCS (P < 0.05, Figure 1).
Figure 1 - Mean ± SEM rating of perceived delayed onset muscle soreness following the exercise protocol for graduated compression stockings (GCS) and control conditions. * P < 0.05 significantly different from control.
552 The Engineering of Sport 7 - Vol. 1
4- Discussion The present experiment was designed to test the hypothesis that the triceps surae, which acts both concentrically and eccentrically during everyday activities, can become damaged after an eccentric exercise and that subsequently the muscle is able to show an adaptation response differently while wearing GCS. Muscles exert an eccentric action when they are lengthened while generating active tension. This occurred in our experimental setup when plantar flexor muscles acted as brakes to slow the motion of the body during backward walking (Nottle and Nosaka 2005). As expected the downhill walking exercise induced a significant decrease in the maximal voluntary torque of the plantar flexor muscles and of similar magnitude for both legs. Immediately after the walking exercise, the torque decrement of -15% appeared to be caused partly by an alteration in muscle contractile properties (i.e., -22% for Pt). Furthermore, this alteration was also associated to a decrease in voluntary activation (i.e., -5% for both legs) suggesting the concomitant existence of a “central modulation” (for a review, see Gandevia 2001). Drop in torque due to damage to muscle fibres may be confounded by the effects of fatigue. In the present study, muscle damage was produced by what subjects considered to be low-intensity exercise. Subjects did not tire during the exercise and by implication fatigue remained at very low levels. Therefore, the post-eccentric changes could not simply be assigned to the effects of fatigue. The first relevant finding of this study is that the MVT failed to recover before the third day, whereas the measure of the (peripheral) contractile properties had recovered significantly within the first 24 hours after exercise (P < 0.01). This delayed recovery in maximal voluntary torque appeared to be mainly associated to a decrease in VA. The time course of change in VA presents similarities with the time course of torque changes. Second, the previous findings on recovery patterns for MVT and Pt amplitudes were “speeded-up” while wearing GCS. Although not significant, it is worthy to note that Pt recovered fully with GCS while a 12% decline in Pt was still observed for the control leg at 24 h. Meanwhile, MVT tended to be higher (+ 6%) for the leg wearing GCS. As in the present study, several studies have reported a prolonged loss in the ability of eccentrically exercised muscle to generate force (Nottle and Nosaka, 2005, Newham et al. 1983). This change may be due in part to alterations in excitation-contraction coupling and crossbridge formation/function. In the present study, the contractile properties had significantly recovered during the first day but not MVT (within the third day). Further, ultrastructural damage worsens in the 2 to 3 days post-exercise as MVT is being restored, thus ultrastructural damage is not the only factor explaining force loss. Thus, the delayed recovery in MVT appeared to be mainly associated with a decrease in voluntary muscle activation. The latter is presented in the literature as a central protection of the muscle from further peripheral fatigue and damage (Gandevia 2001). Overall, these results suggest that contractile properties recovered early during the recovery but with temporal dissociation in the ability to generate force, and that GCS was able to modulate positively contractile properties within 24 h. Post-exercise muscle soreness is usually said to follow an inverted U-shaped curve over time with soreness low immediately following exercise, highest at 24 and 48 h and
The Engineering of Sport 7 - Vol. 1
553
falling at 72 h (see Figure 1). Subjective reports of muscle soreness showed an increase from baseline 2-h, 24-h and 48-h following eccentric exercise in both leg conditions (i.e., with and without GCS). Interestingly, there was a significant difference between both legs at 72 h (Figure 1). It should be noted that the perceived muscle soreness was significantly lower while wearing GCS, which is in line to previous observations (Kraemer et al. 2001a, 2001b). Soreness itself is unlikely result in a reduced ability to generate force, since MVT is already markedly reduced immediately following eccentric exercise before DOMS is perceived. Bringard et al. (2006) demonstrated that compression tights were effective in mediating a reduction in the amount of venous pooling and an increase in muscle oxygenation in the lower body. Different constructions of garments may mediate these overall effects via different physiological mechanisms related to fluid shifts and muscle tissue damage (Jonker et al. 2001). In the present study wearing GCS during 5 h per day may minimize oedema and muscle tissue disruption, thereby increasing comfort in the leg. Use of lightweight (low compression) gradient compression stockings is very effective in improving symptoms of discomfort, swelling, aching, as well as leg tightness when worn regularly (Weiss and Dufy 1999). Wearing graduated compression stockings after eccentric exercise appears to reduce delayed-onset muscle soreness but considerably later after exercise. This could have important implications not only for cross-country runners and trailers, but also for individuals who wish to embark on exercise regimes with associated pain following exercise. The ease of use, length of time and perceptual benefits of such a recovery aid may in itself be of assistance to athletes; however, given minimal differences in neurophysiological data, the likelihood of a placebo effect may be apparent. Whatever the explanation for the significant effect reported here for DOMS with GCS 72 h post-exercise, the current study has provided a small benefit for GCS in the management of the muscle function impairment associated with DOMS.
5- Conclusion In conclusion, GCS did affect differently some temporal measures of muscle function and perceived muscle soreness during 3-days of passive recovery after an exerciseinduced muscle fatigue on calf. These results suggest that GCS may attenuate soreness associated with delayed onset muscle soreness. However it may not be beneficial in the treatment of strength and functional declines. Therefore, as significant benefits were found when subjects wore GCS, it is recommended that active subjects use this additional recovery tool provided by a correctly fitted GCS after physical activity integrating severe muscle damage. However, further research is recommended to investigate the effects of GCS on other measures of muscle function, such as motor performance during dynamic exercise, and to know the necessary application time of GCS after some specific exercise. Such research is warranted to ensure that GCS provide an ergogenic aid so that subjects can exercise more often and with less pain, without impeding performance during successive training bouts.
554 The Engineering of Sport 7 - Vol. 1
6- References [BD1] Bringard A., Denis R., Belluye N. and Perrey S. Effects of compression tights on calf muscle oxygenation and venous pooling during quiet resting in supine and standing positions. In Journal of Sports in Medicine Physical and Fitness, 46(4): 548-54, 2006 [CA1] Chatard J.C., Atlaoui D., Farjanel J., Louisy F., Rastel D. and Guézennec C.Y. Elastic stockings, performance and leg pain recovery in 63-year-old sportsmen. In European Journal of Applied Physiology 93(3): 347-352, 2004 [G1] Gandevia SC. Spinal and supraspinal factors in human muscle fatigue. In Physiological Review, 81(4) : 1725-1789, 2001 [H1] Hough T. Ergographic studies in muscular soreness. In American Journal of Physiology, 7: 76-92, 1902 [JB1] Jonker M., de Boer E., Ader H.J. and Bezemer P.D. The oedema-protective effect of lycra support stockings. In Dermatology, 203: 294-298, 2001 [KB1] Kraemer W.J., Bush J.A., Wickham R.B., Denegar C.R., Gómez A.L., Gotshalk L.A., Duncan N.D., Volek J.S., Putukian M. and Sebastianelli W.J. Influence of compression therapy on symptoms following soft tissue injury from maximal eccentric exercise. In Journal of Orthopaedic and Sports Physical Therapy, 31(6): 282-290, 2001 [KB2] Kraemer W.J., Bush J.A., Wickham R.B., Denegar C.R., Gómez A.L., Gotshalk L.A., Duncan N.D., Volek J.S., Newton R.U., Putukian M. and Sebastianelli W.J. Continuous compression as an effective therapeutic intervention in treating eccentric- exercise-induced muscle soreness. In Journal of Sport Rehabilitation, 10: 11-23, 2001 [M1] McHugh MP. Recent advances in the understanding of the repeated bout effect: the protective effect against muscle damage from a single bout of eccentric exercise. In Scandinavian Journal of Medicine and Science in Sports 13: 88-97, 2003 [NG1] Noonan T.J. and Garrett W.E., Jr., Muscle strain Injury: diagnosis and treatment. In Journal of the American Academy of Orthopaedic Surgeons, 7: 262-269, 1999 [NM1] Newham D.J., Mills K.R., Quigley B.M. and Edwards R.H. Pain and fatigue after concentric and eccentric muscle contractions. In Clinical Science (London), 64(1): 55-62, 1983 [NN1] Nottle C. and Nosaka K. The magnitude of muscle damage induced by downhill backward walking. In Journal of Science and Medicine in Sport, 8(3): 264-273, 2005 [TR1] Trenell M.I., Rooney K.B., Sue C.M. and Thompson C.H. Compression garments and recovery from eccentric exercise: a 31 P-mrs study. In Journal of Sports Science and Medicine, 5: 106114, 2006 [WD1] Weiss R.A. and Duffy D. Clinical benefits of lightweight compression: reduction of venous-related symptoms by ready-to-wear lightweight gradient compression hosiery. In Dermatologic Surgery, 25(9): 701-704, 1999
A Study of Knuckling Effect of Soccer Ball (P106) Takeshi Asai1, Kazuya Seo2, Yousuke Sakurai3, Shinichiro Ito4, Sekiya Koike1, Masahide Murakami3
Topics: Soccer. Abstract: The aerodynamic properties and boundary-layer dynamics of a non-spinning or slowly-spinning soccer ball are not well understood. The purpose of this study is to discuss the magnitude and the frequency of the side force of non-spinning or slowly-spinning flight soccer ball, which called “knuckling effect ball”, using a high speed VTR image of a real place kick. The direct liner transformation method was used to obtain three dimensional coordinates of ball position. The magnitude and the frequency of the side force were measured by a digitizing software system in PC. The magnitude of the side force in real flight was measured to range from about 1 N to 8 N. Additionally, the frequency of the side force in real flight was estimated to range from 1.0 Hz to 3 Hz. Keywords: field testing, video analysis, aerodynamic force, knuckle.
1- Introduction Understanding the flight path of a soccer ball as it curves and drops is essential in the modern game [AT1, AT2]. Hence, there is a need to investigate the aerodynamic properties of soccer balls [M1, CM1, CM2, AT3]. The flight trajectory of a non-spinning or slowly spinning soccer ball may fluctuate in unpredictable ways. Such anomalous horizontal shaking or rapid falling is ascribed to a phenomenon called the ‘knuckle effect’ or ‘knuckling effect’. With the goal of explaining this, in recent years, some research on nonspinning and slowly spinning soccer ball flight has been conducted [BS1, AT4]. However, the aerodynamic properties and boundary-layer dynamics of a non-spinning or slowly spinning soccer ball are not well-understood. The purpose of this study is to analyse high-speed video images of ball flight in three dimensions by considering the knuckle effect to be a fundamental aerodynamic property. Two high-speed video cameras were used to analyse the three-dimensional flight trajectory of a ball in flight; one set behind 1. Inst. of Health and Sports Science, Tsukuba Univ., Tsukuba, Japan. - E-mail: [email protected] 2. Faculty of Education, Art and Science, Yamagata Univ., Yamagata, Japan - E-mail: [email protected] 3. Graduate School of Systems and Information, n Engineering, Univ. of Tsukuba, Tsukuba, Japan E-mail: [email protected] 4. Department of Mechanical Engineering, National Defence, Academy, Yokosuka, Japan - E-mail: [email protected]
556 The Engineering of Sport 7 - Vol. 1 the goal to record an image of the oncoming ball and the other placed to the side to record the lateral view.
2- Methods The high-speed videos were taken at the 1st soccer field at the University of Tsukuba. Fig. 1 is the plan view of the soccer field, showing the positions of the two cameras (Photron FASTCAM-1024PCI) and the point where the ball was placed. As shown in the figure, the ball was placed on the end-line and the soccer goal stood at a distance of 30 m from the ball. The No.1 camera (head-on view camera), with a focal length of 55 mm, was operated at 500 frames/s and a resolution of 1024 512 pixels. The No.2 (side view) camera was placed at a point 60 m from, and perpendicularly to, the ball flight plane, and 16.5 m from the end-line. The No.2 camera used a wide-angle lens with a focal length of 28 mm and was operated at 500 frames/s and a resolution of 1024 256 pixels. The X-axis is in the horizontal forward direction from set ball, the Y-axis is in the horizontal left direction and the Z-axis is in the vertical upward direction, respectively. The direct liner transformation method [AA1] was used to obtain three dimensional coordinates of ball position. The subjects (10 kickers) were elite soccer players from the University of Tsukuba who possessed a relatively high degree of skill. The subjects kicked a ball with little or no spin and at the same speed as they would in real games, aiming at the centre of the goal. Each subjects performed 20 knuckle ball kicks, we chose a typical knuckling effect ball for this analysis, respectively. Therefore, we analyzed 10 cases in these experiments. In these experiments, it is natural that the ball would not rotate in the same manner as in wind tunnel experiments. Moreover, it is impossible that the ball could be in a state of absolutely zero rotation in a realistic free-kick situation.
Figure 1 - Experiment set-up for ball trajectory.
As was stated previously, we assume that the flight trajectory of the non-spinning or slowly spinning soccer ball is the subject of the investigation. Fig. 2(a) shows the headon view of the ball flight trajectory, taken by the No.1 camera, and Fig. 2(b) shows the side view of the ball flight trajectory, taken by the No.2 camera.
The Engineering of Sport 7 - Vol. 1
557
Figure 2 - Examples of high speed VTR image using stroboscopic technique of front view (a) and side view (b) in knuckle ball (The X-axis is in the horizontal forward direction, the Y-axis is in the horizontal left direction and the Z-axis is in the vertical upward direction from set ball).
By the way the distortion of the image seen through a wide-angle lens must be corrected. The correction of the depth perception which is arisen from the image taken by the camera No.1 in front of the goal was attempted in the digitizing stage. In order to make clear the fundamental aerodynamic properties of knuckle effect, the time series data of the ball flight trajectory, the velocity and the acceleration must be obtained. For the calculation of the trajectory of a ball, the top, bottom, left and right points of a flying soccer ball image are first digitized, and then the center of the ball is calculated as an average of the four points. The tracking of the ball image was mostly accomplished with the aid of image analysis software that has a capability of tracking a target object automatically. However, since in some particular positions in an image it was hard to distinguish the envelop of a ball because the ball seemed as if it was blended into the background, the tracking of the ball position was made by hand-operation. In order to correct this distortion, the quadric correction expression is also applied to the original video image. This quadric correction expression is derived in the same procedure for the correction of the depth perception. After the correction by the above method, the z-coordinate values of the corrected image taken by the camera No.1 and that by No.2 must coincide with each other. In order to transform the pixel number of the point of the ball center into the ycoordinate in the ground-fixed space, the quadratic correction expression is applied to the pixel of the original video image. This quadratic expression is empirically derived referring to the images on which the points of the marker-cones placed in the soccer field at the fixed distance. The video image taken by the No.2 camera is somehow distorted because the image was seen through a wide-angle lens.
558 The Engineering of Sport 7 - Vol. 1 Each of the following coefficients were calculated: ball (wind) velocity (V), force acting in the direction of the wind (drag (D)), force acting perpendicular to the wind direction (lift (L)), and force acting sideways based on a frontal view (S). The aerodynamic forces acquired in the experiment were converted to the drag coefficient (CD), lift coefficient (CL), and side force coefficient (CS), as shown in equations 1 to 3. (1) (2) (3) where is the density of air (=1.2 kg/m3), V is the flow velocity, and A is the projected area of the soccer ball (A = 0.112 = 0.038 m2).
3- Trajectory analysis Typical examples of the digitized displacement (trajectory) data of the ball center point on knuckle ball and curve ball are shown in Fig. 3. Figures 3(a), (b), (c) show that the knuckle ball flight in the X-Z plane, the X-Y plane and the Y-Z plane, respectively. Similarly, Figures 3(d), (e), (f) show that of the curving ball flight. All of them are shown after the coordinate of the trajectory is converted into an image so that the ball flies right toward the center of the No.1 camera. Fig. 3(b) indicates that the ball fluctuation occurs due to the knuckle effect around X=10~15m with a fluctuation amplitude of about 0.03m. We can observe the fluctuation around the top of the trajectory in the front view (Fig.3(c)). On the other hand, Fig.3(a) seemingly does not show any fluctuating motion clearly but it just seems to be a parabola that is a normal ball flight trajectory in the vacuum. For the purpose of comparison, Fig. 3(e) and Fig. 3(f) are shown as an example of the trajectory of conventional the curving ball flight being influence by the Magnus effect, which does not show any major fluctuation as seen in Fig. 3(b) and Fig. 3(c). In this study, it is demonstrated that the trajectories under the influence of the knuckle effect can be visualized by the 3-dimmnessioal video image analysis for the images taken by two high-speed cameras. However the trajectories are not sufficiently smooth accompanied with considerable data scattering. It is clear that this data scattering results from primarily manual ball tracking in video image analysis. The data scattering is also noticeable in the velocity data and the acceleration data. Therefore the methods of calibration and the ball tracking should be improved in order to carry out more accurate analysis. The side force on a ball experiencing the knuckle effect and its trajectory are shown in Fig. 4. The thick line shows data for the side force, which is certainly one of the factors contributing to the knuckle effect, and the thin line shows data for the corresponding trajectory under the influence of the knuckle effect. The peak side force is about 2.5 N,
The Engineering of Sport 7 - Vol. 1
559
and the frequency of the variation is about 2.0 Hz. The magnitude of the side force in real flight was measured to range from about 1 N to 8 N (Table 1). It was seemed that the detect of upward acceleration was more difficult than that of sideward according to effect of gravity, therefore the side force and side force frequency were analyzed in this study. The frequency of the side force in real flight was estimated to range from 1.0 Hz to 3 Hz. The knuckle effect seems to cause random fluctuations in both the horizontal and vertical directions. In this study, the coefficients of the side force and lift force were investigated after removing the effect of gravitational acceleration.
Figure 3 - Examples of the digitized displacement (trajectory) data of the ball center point on knuckle ball (a, b, c) and curve ball (d, e, f).
560 The Engineering of Sport 7 - Vol. 1
Figure 4 - An example of ball trajectory and side force in knuckle ball. Table 1 - Side force and frequency of each subject on knuckle ball.
Fig. 5 shows the front view of the variations in the side-force and lift coefficients during flight, where the abscissa indicates the side force coefficient and the ordinate indicates the lift coefficient. In this figure, the thick line shows the variations in the side-force and lift coefficients under the influence of the knuckle effect, and the thin line, those of a curving ball (due to back spin and side spin) experiencing the Magnus force. From the figure, it is clear that the side-force and lift coefficients exhibit a wider range of variation under the influence of the knuckle effect than in the case of the curving ball. It is interesting to note that, on average, the positive and negative variations in the side-force and lift coefficients were equal. On the other hand, the variations in the side-force and lift coefficients of the curving ball are of a smaller scale and are distributed unevenly in the first quadrant. Thus, it seems that, in the case of the knuckle ball, the side and lift forces act on the ball almost to the same degree. It has not escaped our notice that the frequency of this fluctuating force we have calculated immediately suggests a possible depending on the large scale undulation of the vortex trail [AT3] in knuckle ball.
The Engineering of Sport 7 - Vol. 1
561
Figure 5 - An example of unsteady side force coefficient versus unsteady lift coefficient during flight on a knuckle ball and a curve ball.
4- Summary The purpose of this study is to discuss the magnitude and the frequency of the side force on a non-spinning or slowly spinning soccer ball, which is called a knuckle ball, using a high speed VTR image of a real spot kick. The results are summarized as follows. 1) Three-dimensional ball flight trajectories could be calculated by the video image analysis with two high speed cameras. 2) It was observed that the large scale fluctuation around the top of the trajectory in knuckling ball. 3) An example of the trajectory of conventional the curving ball flight being influence by the Magnus effect, which did not show any major fluctuation clearly. 4) The magnitude of the side force in real flight was measured to range from about 1 N to 8 N. 5) The frequency of the side force in real flight was estimated to range from 1.0 Hz to 3 Hz.
5- References [AT1] Asai, T., Akatsuka, T. & Haake, S.J. (1998) The physics of football, Physics World, vol.11-6, 25-27. [AT2] Asai, T., Carré, M.J., Akatsuka, T. & Haake, S.J. (2002) The curve kick of a football, I: impact with the ball. Sports Engineering, 5, 183–192. [AT3] Asai, T., Seo, K., Kobayashi, O. and Sakashita, R. (2006) Flow visualization on a real flight non-spinning and spinning soccer all. IN: The Engineering of Sport 6, Vol. 1, (eds. E.F. Moritz & S. J. Haake), pp. 327–332. International Sports Engineering Association, Sheffield. [AT4] Asai, T., Seo, K., Kobayashi, O. and Sakashita, R. (2007) Fundamental aerodynamics of the soccer ball, Sports Engineering, ISEA, 10(2), 101-109.
562 The Engineering of Sport 7 - Vol. 1 [BS1] Barber, S., Haake, S.J. and Carré, M.J. (2006) Using CFD to understand the effects of seam geometry on soccer ball aerodynamics. In: The Engineering of Sport 6, Vol. 2, pp. 127–132. The International Sports Engineering Association. [CM1] Carré, M.J., Goodwill, S.R., Haake, S.J., Hanna, R.K. and Wilms, J. (2004) Understanding the aerodynamics of a spinning soccer ball. The Engineering of Sport 5 (Eds. M. Hubbard, R.D. Mehta and J.M. Pallis). Pub. The International Sports Engineering Association, Sheffield, UK, 7076. [CM2] Carré, M.J., Goodwill, S.R. & Haake, S.J. (2005) Understanding the effect of seams on the aerodynamics of an association football. Proc. IMechE Vol.219 Part C: J. Mechanical Engineering Science, 657-666. [M1] Mehta, R.D. (1985) Aerodynamics of Sports Balls, Annual Review of Fluid Mechanics, 17, 151-189. [AA1]Abdel-Azis, Y. I. and Karara, H. M. (1971) Direct liner transformation from comparator coodinates into object apace coordinates in close-range photogrammetry. In: Proceedings of the ASP/UI Symposium on Close-Range Photogrammetry. Falls Church. VA: American Society of Photogrammetry, 1-18.
Ball and Racket Movements Recorded at the 2006 Wimbledon Qualifying Tournament (P109) Simon B Choppin1, Simon Goodwill1, Steve Haake1, Stuart Miller2
Topics: Tennis & other Rackets Sports; Testing, Prototyping, Benchmarking. Abstract: This paper contains the recorded shot movements of 13 players in practice conditions at the Wimbledon 2006 Qualifying Tournament. A 2-camera 3D system was used to track the racket and ball for a period of 0.02 seconds for each recorded shot. Custom-written analysis software was used to extract the required co-ordinates from the ball and racket positions and transform them into 3D. From this information the following things were obtained; ball velocity before and after impact; racket linear and angular velocity before impact; ball spin and impact position. It was found that although ball velocity was very similar for all players before impact, male players were able to generate higher ball velocities after impact. This was found to be due to a higher racket COM velocity. Impact position and angular velocities were very similar for both sexes.
1- Introduction Player testing is an important tool, and has its place in sports science, engineering and biomechanics. To date, and with tennis analysis firmly in mind, photogrammetric player testing has generally been performed in 2D at low (<200 fps) frame rates. The aim of such testing has varied from establishing a definition of player accuracy Blievernicht (1968), or more recently, studying advanced player kinematics Knudson (2005). Many camera and sensor systems now exist which are capable of recording multiple points in 3D co-ordinates. Visual systems have trouble operating in daylight, sensor systems are capable of tracking 3D points in real time, but require large, heavy sensors to be attached to the tracked object, and currently cannot process data at high enough levels to be of use in this case. A method capable of 3D tracking has been developed which was designed to be portable, have no redundancy and be of minimal intrusion to the player. It focuses solely on the racket and ball movement immediately prior to, and post impact. Analysis is 1. Centre for Sport and Exercise Science, Sheffield Hallam University Collegiate Campus, Sheffield S10 2BP, United Kingdom E-mail: s.choppin, s.r.goodwill, [email protected] 2. International Tennis Federation, Bank Lane, Roehampton, London SW15 5XZ, United Kingdom
564 The Engineering of Sport 7 - Vol. 1 concentrated on the movements of the racket and ball only, biomechanical movements are not considered. To obtain full 3D movement, specifically marked points on the racket were used to define a plane, the ball was tracked as a single point in space. With this information it was possible to accurately track racquet velocity (any point on racquet), ball velocity, impact instant, impact position, and all associated angular velocities. The advantage of this method is that unlike previous methods, velocities are not limited to a single tracked point, or singularly considered axis of rotation. This method has been used to record a total of 106 shots from 13 internationally ranked players in practice conditions at the 2006 Wimbledon Qualifying Tournament, full ball and racket movement was achieved in each case. This paper outlines essential details of the aforementioned methodology and describes the recorded racket and ball movement for the male and female players.
2- Methodology 2.1 Camera and Court Set-up A 2-camera stereo videogrammetric method was used to track ball and racket position over the duration of the testing. High speed phantom V 4.2 cameras were used running at 1000 frames per second. The duration of testing was such that 20 image frames were captured for each recorded shot. Ten frames were recorded before and after impact, in this way a total of 0.02 seconds of footage was used for each impact. A practice court at the Lawn Tennis Association’s grounds in Roehampton, London was used to record every shot. The player was made aware of the testing, but was not asked to perform in any particular way. Each of the two cameras were placed at either side of the court, in-line with the net. Positioning the cameras in this location ensured that clear impact images were obtained whilst minimising visual interference with the player. Figure 1 (taken from (Choppin, Goodwill et al. 2007)) shows the position of the two cameras relative to the court.
Figure 1 - The position of the two cameras at either side of the net.
The Engineering of Sport 7 - Vol. 1
565
A checkerboard was used to calibrate the court system for 3D analysis, it was moved into a variety of positions within a prescribed 2 ҂ 2 ҂ 2 m volume. The checkerboard intersections were tracked using a Matlab program in order to generate the required set of calibration parameters. An orthogonal frame with attached reflective markers was used to specify a global axes set, as shown in figure 2.
Global Reference Frame
Local Reference Frame Figure 2 - Reflective markers set on an orthogonal frame shown on the right were used to generate a global axes set located as shown on the right.
2.2 Ball and Racket Set-up Ball position was tracked from the footage by locating its centre at every instant. The ball was easily tracked from the recorded image due to the high reflectivity of the ball’s surface. The ball was marked discretely using a fabric marker in order to be able to count full revolutions and hence ball spin. The racket was marked using five high-contrast markers on the racket frame. The markers were used to track racket location and orientation using a local axes set as shown in figure 2.
2.3 Recording and Analysis If permitted by the player, the racket was marked and measured in order to accurately recreate any recorded movement. The player was asked to commence a normal practice session. A period of 0.3 seconds was recorded via a manual trigger which was depressed as close to the instant of ball/racket impact as possible. Each shot was analysed using custom-written software. The position of three markers and the ball was extracted from both images at every instant. The 2D co-ordinate sets for each shot were transformed into 3D co-ordinates automatically using this software. The software used the set of 3D co-ordinates to calculate racket and ball velocities, impact positions and spins.
566 The Engineering of Sport 7 - Vol. 1
3- Results The analysis method was able to calculate: • Ball velocity in three directions prior to and post impact. • Racket linear and angular velocity prior to impact. •Ball spin and angular velocity prior to impact. • All results are split according to the sex of the player. All the results are displayed as a frequency histogram in order to clearly display the spread of the obtained values.
3.1 Ball Velocity The measured ball velocities before and after impact are shown in figure 3. The velocities are shown according to the global axes set shown in figure 2.
Figure 3 - The pre and post impact ball velocities according to the global axes set shown in figure 2. The results are divided according to the sex of the player.
3.2 Racket Velocity The linear and angular velocity of the racket is shown in figure 4. In this case all results are prior to impact. The mass of the player’s arm affected the velocity of the racket after impact, these results were therefore omitted. The linear velocity values are taken from the racket COM and according to the global axes set. Angular velocity values are about the three local axes described in figure 2.
The Engineering of Sport 7 - Vol. 1
567
Figure 4 - The pre-impact linear and angular velocities of the racket.
3.3 Ball Spin and Impact Position The spin of the ball after impact is shown below in figure 5 along with impact positions. Impact position is described as distance from the racket butt (along the racket length) and distance from the racket’s central axis (off set impact distance).
Figure 5 - The post-impact ball spin and impact location described as distance from the butt and central axis.
568 The Engineering of Sport 7 - Vol. 1
3.4 Results Summary A summary of all the results shown above is included in table 6. In this case resultant values have been given with the mean and standard deviations in each case.
Table 6 - A summary of the results shown in section 3.
4- Discussion The results show that the ball has a very similar resultant velocity and direction before impact for both the male and female players. All of these results were obtained in practice conditions; the ball is hit to the player in a very repeatable way, usually by the coach. Given the very similar ball velocities before impact, male players hit the ball with a greater velocity, on average, after impact. The results show that the male and female players hit the ball at a very similar point on the racket and exhibit very similar racket angular velocities. The higher post-impact ball velocity is therefore due to the higher racket COM velocities generated by the male players. An 11% higher average resultant racket velocity generates around 9% higher resultant ball velocities. Figure 4 shows that vertical ball velocity after impact (global Y direction) is positive after impact and correspondingly higher for the male players. This is as one would expect, positive vertical velocity gives the ball the necessary trajectory in order to arc over the net to reach the opposite side of the court. Higher vertical ball velocity generated by the male players is reflected by higher ball spin values. The higher spin generates more down force, allowing the ball to be hit harder and still remain within the boundaries of the court. Both male and female players hit the ball, on average, on a very similar point on the stringbed. Longitudinal impact position was referenced from the racket butt so it can be easily compared between each racket. The mean impact position is located roughly around the racket’s stringbed centre, which has been found to be the location on the stringbed which players ‘aim’ for in previous studies Hatze (1994) and Coe (2000). The vast majority of the shots were played with top-spin; observation of the footage reveals that the majority of shots were also played on the forehand. This similarity in shot type makes the data useful for comparison, within the range of recorded shots there still exists a variety of spins and velocities. The lack of backhand and volley shots means that further testing is required in order to assess the particularities of this type of shot.
The Engineering of Sport 7 - Vol. 1
569
Generally, this work has provided a range of values for the baseline top spin forehand which can be used in future laboratory testing. The speeds and spins are generally as expected and impact positions tally with previous research. The values of racket and ball linear and angular velocity at impact, and the ball’s impact position on the racket give further insight into how the top players in the game of tennis execute their shots. A development of this methodology or collaboration with a racket manufacturer would allow testing to take place during true competition conditions. This is currently not possible due to markers placed on the racket face. Continued testing of this type and at different levels would constitute the most in-depth player shot analysis to date.
5- References [B1] Blievernicht, J. G. (1968). “Accuracy in the Tennis Forehand Drive: Cinematographic Analysis.” Res. Q. Ex. Sport 39(3): 776-779. [C1] Coe, A. O. (2000). The Balance Between Technology and Tradition in Tennis. Tennis Science & Technology, The International Tennis Federation, London, UK, Blackwell Publishing. [C2] Choppin, S. B., S. R. Goodwill, et al. (2007). 3D Player Testing at the Wimbledon Qualifying Tournament. Tennis Science and Technology 3, The International Tennis Federation, London, UK. [H1] Hatze, H. (1994). “Impact Probability Distribution, Sweet Spot, and the Concept of an Effective Power Region in Tennis Rackets.” Journal of Applied Biomechanics 10: 43-50. [KB1] Knudson, D. and J. R. Blackwell (2005). “Variability of impact kinematics and margin for error in the tennis forehand of advanced players.” Sports Engineering 8(2): 75-80.
Ball Spin Generation at the 2007 Wimbledon Qualifying Tournament (P110) John Kelley1, Simon Goodwill1, Jamie Capel-Davies2, Steve Haake1
Topics: Tennis & other Racket Sports. Abstract: The International Tennis Federation (ITF) monitors all aspects of tennis for their role of protecting the nature of the game. One area in which there is limited documented data is ball spin in match play. In previous work, the ball spin generated in the 2007 World Group Davis Cup tie between Switzerland and Spain, played indoors on Taraflex carpet, was studied. This paper uses footage captured from the Wimbledon Qualifying Tournament, an outdoor grass tournament that provides many new issues that need to be dealt with in order to capture high speed video. This paper describes the methods used and highlights the difficulties encountered and how they were overcome. The spin rate was measured prior and post impact to enable analysis of the spin rate generated both from the bounce and by the player. Some of the results are presented with a brief comparison with the previous Davis Cup study. It was intended that the majority of the shots should be of male tennis players however weather conditions limited the number of male matches played on the courts that could be filmed. This meant that the number of usable shots recorded for men was 54 from 4 players, whereas 152 shots for women from 10 players were recorded. All of the players filmed, both male and female, were in the top 300 in the world rankings. The maximum spin rate values found for women’s serve and forehand are 3529 rpm and 2727 rpm respectively. This study shows that spin rate analysis of outdoor player testing in match conditions using high speed video is feasible and as accurate as previous indoor work. Keywords: Tennis; Ball Spin; Player Testing; Competition; High Speed Video.
1- Introduction The International Tennis Federation (ITF) is the world governing body of tennis and has a number of objectives designed to protect and develop the sport. Included in these is the objective to make, amend and uphold the rules of the game and the objective to preserve the integrity and independence of tennis as a sport (ITF, 2008). The combina1. Sports Engineering Research Group, Sheffield Hallam University, UK. - E-mail: [email protected], s.r.goodwill, [email protected] 2. International Tennis Federation, Roehampton, UK. - E-mail: [email protected]
572 The Engineering of Sport 7 - Vol. 1 tion of these objectives implies that the rules of the game need to be chosen in order to protect the nature of the game. The Rules of Tennis (International Tennis Federation, 2007) requires several tests that the ball, racket and court surface are required to pass to become ITF Approved and also several other tests they must undergo to be ITF Classified (e.g. the Surface Pace rating of a court). To protect the nature of the game these rules and the associated test procedures need to be repeatedly reviewed. This review process will need to include the accurate monitoring of all aspects of tennis so the ITF understand the mechanics of the game as fully as possible (Goodwill et al., 2007). Also continuous innovation in tennis equipment by manufacturers means that constant monitoring is required to find any changes to the game caused by these innovations. Ball spin is one of the aspects of tennis that needs to be constantly measured so that trends can be monitored to help the ITF make the correct decisions. Accurate ball spin values for up to date match play tennis also ensures that any relevant ITF equipment tests are carried out at realistic ball spin ranges. Player analysis in sport is very well documented and tennis is no exception. There is a great deal of literature on areas such as injury prevention (Elliott, B., 2006) and also the movement of the ball, racket and/or player for specific shots (Blievernicht, 1968; Elliot et al., 1986; Choppin et al., 2005). However, accurate measurement of ball spin is not easily done especially in the field and there has been very little quantitative data collected on ball spin values in competitive matches. The first major study on the magnitude of ball spin was conducted at the 1997 US open (Pallis, 1997), which may not be representative of ball spin in the modern game. More recently, the ball spin generated in the 2007 World Group Davis Cup tie between Switzerland and Spain, played on Taraflex carpet, was studied by Goodwill, et al. (2007) on behalf of the ITF. The ITF requires continuous monitoring of ball spin generation in match situations and the aim of this study is to continue the work started by Goodwill, et al. (2007) at the 2007 Wimbledon Qualifying Tournament. This paper will highlight the problems encountered, how they were overcome and show that spin rate analysis from outdoor player testing in match situations without interfering with players or equipment is possible.
2- Method The Wimbledon Qualifying Tournament takes place at the Bank of England Sports Club in Roehampton and in 2007 was held from June 18th-21st. This tournament, played on outdoor grass courts, has been held before each Wimbledon Championships in various formats since 1919. Players who win all 3 rounds receive the prize of a place in the main Wimbledon draw. Each match used Slazenger Wimbledon tennis balls which each have a single sponsorship logo. The testing was carried out on the first 3 days of the competition, with the intention that the majority of shots recorded were to be from male matches that were scheduled on the first and third days. The filming was done using a Vision Research phantom v4.3 high-speed video camera operating at 1000 frames per second and with exposure time 200 μs, the same settings used by Goodwill, et al. (2007). However the phantom v4.3 camera allowed a higher image resolution of 800 x 600 to be used. The short exposure time of 200 μs is required to keep motion blur to a minimum
The Engineering of Sport 7 - Vol. 1
573
to aid analysis. Unlike Goodwill, et al. (2007), the matches were played outdoors so there were good levels of light. This meant the aperture size could be relatively small thus increasing the usable depth of field (the depth of the picture in which objects are in focus). This was useful to keep the ball in focus as long as possible. The aperture was not fixed at one setting since the light levels changed repeatedly with the change in cloud cover throughout the testing. The phantom v4.3 camera requires the use of a laptop, onto which the shots are saved. Footage was not continuously recorded from the camera; the recording was manually triggered every time a stroke was hit inside the field of view using a post trigger. Initially 500 frames (0.5 seconds) were recorded for each shot but this was reduced to 300 frames (0.3 seconds) since this was still more than long enough to capture the whole shot. Reducing the number of frames enabled more shots to be taken since they required less time to save to the laptop from the camera and the camera was ready for the next shot sooner. Mains power was needed for the camera and the laptop and was available from between the two areas of public courts, close to the practice area (see figure 1). It was routed between the unused courts and the practice area to the fencing along the line of 12 courts and then ran along the bottom of the fencing to the camera position. The cabling had to be raised over a walkway to ensure that it was out of the way of spectators. Any links between cables such as plugs were waterproofed with plastic coverings attached tightly around the cables completely covering the connections.
Figure 1 - Overhead of Roehampton site.
Filming was allowed from a raised bank next to the line of 12 courts, 10 of which were designated match courts and the other 2 were subsequently used to aid schedule congestion (figure 1). The camera had to be positioned on a bank using a high reaching tripod extended to 2m above the ground so that the fencing around the courts did not obstruct the view (figure 2). The camera was positioned on the secondary spectator bank to ensure that the camera did not obstruct the view of spectators and so that spectators
574 The Engineering of Sport 7 - Vol. 1 did not interfere with the equipment. This bank was sparsely populated because the courtside fencing partially obscures the spectator’s view. A small, high tent, 1m x 1m x 2m high, was erected near the camera so that the equipment, especially the camera and laptop, could be stored inside it whenever it rained. A high tent was chosen so the tripod and camera could be quickly stored without dismantling the set up.
Figure 2 - Photo of the camera and tripod set up set-up.
Measuring ball spin requires a clear, sharp image of the ball long enough to track the spin, so the ball must stay in focus for at least one revolution. This means that the ball cannot be moving directly towards the camera because this would provide a very limited period of time in which the ball is in focus, it would only be the length of the usable field depth (figure 3(a)). Ball movement across the field of view provides a much greater time in which the ball is in focus so spin can be measured and the ideal angle for this is close to right angles (figure 3 (b)). Therefore the camera had to be aimed at least partially side on to the court. The court of interest could not be the one directly in front of the set up area. Also the zoom restricted the choice to courts within 50m of the camera and the camera could only be adjusted while the immediately adjacent courts were not in play so that players were not disturbed. Hence there was only a limited choice of which matches could be filmed.
Figure 3 - Camera aim compared to ball movement.
The Engineering of Sport 7 - Vol. 1
575
Goodwill, et al. (2007) carried out preliminary testing to ascertain that professional players typically stand 1 metre behind the baseline when hitting groundstrokes. This was used to determine that the cameras should be aimed at the centre of the baseline. Similarly the camera was aimed at the baseline again to record groundstrokes in this study. However, the camera was not aimed at the centre on the baseline since the angle sometimes prevented a view across the whole baseline. Instead it was aimed at a point so that the serves and returns from the deuce court could be recorded and then so that as much of the rest of the baseline as possible could be seen (figure 4).
Figure 4 - Camera aimed at the baseline.
This study aims to measure the ball spin rates generated both off the racket after the ball has been hit and off the ground, after the ball has bounced but before it has been hit. The spin generated off the ground is found by measuring the spin rate before the ball is hit by the player and the spin generated off the racket is found by recording the spin after (figure 5).
Figure 5 - Diagram showing where at which point the spin is measured.
Ball spin rates were measured both before and after the ball was hit by manually tracking the orientation of the logo on the ball. As long as the ball rotation brought the logo into view it was easy to see since as the resolution of the cameras were high (see figure 6). The number of frames for the ball logo to return to its starting position, chosen to
576 The Engineering of Sport 7 - Vol. 1 be where the logo is closest to the centre, was counted and converted into a spin rate. No axis of revolution was determined since the logo size was too large to get accurate enough coordinates to find the axis. In a large number of the shots the ball went out of the field of view before more than one revolution had been completed so for consistency only one revolution was used for the frame count.
Figure 6 - Series of photos showing the logo move across the ball, the 5th picture shows a typical start frame.
In a repeatability study the spin rate off the racket of 6 shots were reanalysed 4 times. So there were 5 frame counts, and thus 5 spin rate values, for each of these 6 shots. The maximum variation in the frame count for any of the 6 shots was 2 frames, which corresponded to 134rpm for a spin rate in the region of 2000rpm. However the maximum variation in the spin rate value was 141rpm for a different shot with a spin rate in the region of 3000rpm. The variations in spin rate value are all less than 7% of the recorded spin rate value. These variations are comparable to the repeatability found by Goodwill, et al. (2007).
3- Results From the 3 days of filming 206 usable shots, all from match play, were collected and analysed, summarised below. Summary of Male results 1. A total of 54 usable shots were recorded. 2. Four different players were filmed, from 2 matches. All the players had ATP single rankings of 290 or better. 3. The range of shots were as follows 1- 27 were 1st serves. 2- 12 were forehand topspin strokes. 3- 4 were backhand topspin strokes. 4- 10 were backhand slice strokes. Summary of Female results 1. A total of 152 usable shots were recorded. 2. The majority of the shots were evenly distributed between 8 players from 4 matches and the remaining 3 were of 2 other players from another match. All the players had WTA rankings of 287 or better. 3. The range of shots were as follows 1- 50 were 1st serves. 2- 66 were forehand topspin strokes. 3- 4 were forehand slice strokes. 4- 30 were backhand topspin strokes. 5- 2 were backhand slice strokes.
The Engineering of Sport 7 - Vol. 1
577
Table 1 - Brief results of the spin rate analysis.
Weather conditions on the first and second day and subsequent schedule changes meant that more female matches were played on the courts that could be filmed and so the majority of the shots recoded were of female players. A brief outline of the values recorded is shown in table 1, NA is shown where there were not enough shots recorded for reliable values to be shown. The mean value of the spin rate off the ground for men is very close that found by Goodwill, et al. (2007), 3104 rpm compared to 3344 rpm and the maximum value is also very similar to that found by Goodwill, et al. (2007), 5000 rpm compared to 5200 rpm.
Figure 7 - The distribution of ball spin rates generated by female shots.
Figure 6 (a) shows that the women hit 1st serves with a range of spin rate values spread out quit evenly between above 1000 and below 3500 rpm. This wide range of spin values perhaps indicates a large variety of tactics used on serve. Figure 6 (b) shows that the majority of female forehand strokes were hit with spin rates between 1000 and 2000 rpm.
4- Discussion The problems encountered when filming in a difficult environment, outdoors with changing weather conditions, were overcome and 206 usable shots were recorded. The courtside fencing and the layout of the courts provided difficulties that required careful positioning of the camera. Continuous rain for most of the first day did not cause any damage to equipment even though it was kept in the field inside the tent and as soon as play started the equipment could be quickly and easily made ready to record. Similarly the short showers of the second day did not cause any problems. The 60m of power cabling required to access mains power was protected against the rain and set up so that there was no obstruction to the movement of spectators, players and officials. This was done on the morning of the event and did not present and both during and after set up there were no problems with power. High winds, especially on the third day, did not
578 The Engineering of Sport 7 - Vol. 1 interfere with the filming at all. The camera remained stable even on the high tripod and all the equipment remained reliable. The only problem the high winds caused was when erecting the tent. The ever changing cloud cover resulted in quite changeable light levels and this was dealt with by not using a fixed aperture setting. The aperture was adjusted between shots if the light levels changed significantly. Both grass and Taraflex are considered to be fast surfaces and the similarity of the spin rate values between this study and Goodwill, et al. (2007) is what would be expected. Since the values found in this study are close to those found by Goodwill, et al. (2007) on a similar surface indicates that the methods used in both studies are comparably accurate.
5- Conclusions It is possible to carry out accurate spin rate player testing on outdoor matches in a competitive environment without interfering with either players or spectators. The weather conditions highlighted the need to prepare the equipment carefully and provide protection against the elements and the precautions taken proved effective against rain and high winds. The analysis was successful which shows that the camera does not need to be set up precisely court side and that it is not necessary to modify equipment (adding markers to the ball for example) which would not be allowed in match play. The major problem that could not be solved was the schedule changes caused by the weather conditions which restricted the data that was recorded. The obvious solution to this would be to anticipate this complication and prepare to film for extra days if required. The ITF can use the mean and maximum spin rates for men found both here and by Goodwill, et al. (2007) to ensure any equipment tests and future experiments use realistic match play spin rate values. Also the spin rate values found here for women can be used to ensure that the ITF can protect the women’s game by ensuring tests and experiments involving spin rates are carried out at spin rate values relevant for women as well as men.
6- References Blievernicht, J.G. (1968). Accuracy in the Tennis Forehand Drive: Cinematographic Analysis. Research Quarterly for Exercise and Sport, 39, 776-779. Choppin, S.B., Whyld, N.M., Goodwill, S.R. and Haake, S.J. (2005). 3D Impact Analysis in Tennis. The Impact of Technology on Sport, Tokyo Institute of Technology, 1, 373-378. Elliott, B., Marsh, A. and Blanksby, B. (1986). A Three-Dimensional Cinematographic Analysis of the Tennis Serve. International Journal of Sport Biomechanics. 2. 260-271. Elliott, B. (2006). Biomechanics and Tennis. British Journal of Sports Medicine. 40. 392-396. Goodwill, S.R., Capel-Davies, J., Haake, S.J. and Miller, S.R. (2007). Tennis Science & Technology 3, 349-356. International Tennis Federation. (2007). Rules of Tennis. London: ITF ITF (2008). http://www.itftennis.com/abouttheitf/abouttheitf/index.asp Pallis, J. (1997). http://wings.avkids.com/Tennis/Project/usspin-05.html.
Analysis and Optimization of the Sliding Properties of Luge Steel Blades on Ice (P111) Mathieu Fauve, Hansueli Rhyner1
Topics: winter sports, measurement systems, performance sports. Abstract: The performance of a luge competitor is highly influenced by the interaction between the luge blades and the ice. This interaction depends on the environmental and loading conditions onto which the equipment must be adapted in order to get the lowest friction and still allow good steering of the luge. Until now, the evolution went mainly through the athlete’s perception and by trial and error during field testing. The objective of the present study was to optimise the sliding of steel blades on ice by an analysis of penetration and friction acting on the blades during sliding. In order to achieve this goal, a specific laboratory measuring campaign was conducted. The friction force and the penetration into the ice of four different steel blade cross-sections for various experimental settings were measured. It was observed that friction and penetration are highly correlated and mainly influenced by the outside angle of the blade. It was verified that the laboratory experiments are very accurate and reproduce well the field conditions. Keywords: luge, blade, ice, friction, penetration.
1- Introduction The performance of a luge competitor is influenced by their technical skill, aerodynamic behaviour and by the interaction between the luge runners and the ice. This interaction involves friction and penetration into the ice which determines the steering behaviour of the equipment. In order to minimize friction and still allow a good control of the luge, athletes try to adapt their equipment to the ice conditions by modifying the longitudinal curvature and the cross-section of the steel blades. A difficulty in finding the optimal geometry is that the force applied onto the ice varies from around 1200 N on straights up to 6000 N during curves. One of the main problems is that the optimisation of the equipment is only the results of subjective field testing during which numerous parameters can influence the results such as a heterogeneous ice surface or steering mistakes made by the athletes. 1. WSL Institute for Snow and Avalanche Research SLF, Davos Dorf, Switzerland - E-mail: fauve, [email protected]
580 The Engineering of Sport 7 - Vol. 1 The sliding of runners on ice has been studied by many authors (Bowden [B1], Ericksson [E1], Barnes et al. [BT1], Colbeck et al. [CN1], Hainzlmaier [H1]). The different physical processes involved during sliding on ice have been described in many studies (Evans [EN1], Oksanen et al. [OK1], Akkok et al. [AE1], Petrenko [P1], Petrenko et al. [PC1] [PR1], Hainzlmaier [H1]). It could be shown that frictional heating, pressure melting, ploughing, existence of a liquid-like layer, electrostatic charging and adhesion are involved in the sliding of runners on ice. The influence of many parameters such as the sliding material, the ice temperature or the load on friction has also been quantified mainly by conducting laboratory measurements (Bowden [B1], Ericksson [E1], Evans et al. [EN1], Tusima [T1], Forland et al. [FT1], Akkok et al. [AE1], Itakaki et al. [IH1], Hainzlmaier [H1]). The steering ability of runners has never been subjected to scientific studies. However, one can speculate that it should be highly related to the penetration of the blades into the ice. In 2005 Hainzlmaier [H1] has analysed with casting material the trace left by bobsleigh runners on the ice surface and compared his measurements with numerical simulations. The mechanical properties of ice have been studied by many authors (Bowden [B1], Petrenko et al. [PW1], Schulson [S1], Petrovic [P2]). It was clearly shown that the hardness and the compressive strength of ice are highly influenced by temperature. Since it has never been studied in the past and because of its importance concerning the performance of a luge competitor, the interaction between penetration and friction of luge blades sliding on ice was analysed in the present study. In order to achieve this goal, laboratory measurements of the sliding behaviour of steel blades on ice under controlled and stable conditions representatives of field conditions were performed. From a technical point of view, the objective was to optimise the sliders cross-section by adapting it to the ice and loading conditions.
2- Materials and methods 2.1 Sample The luge blade cross-section is defined by the two angles Wo and Wi and by the length L as presented on figure 1. For the present study, the aforementioned parameters were given two levels: high or low. For confidential reasons, the values of these parameters are not mentioned. In order to minimize the number of samples, the Taguchi statistic method (Roy [R1]) was applied and allowed to reduce it from 8 to 4. The four different geometries tested are summarized in the table 1. The samples had a contact length of 15 mm and were rounded at both ends to avoid a cutting effect of the ice. The samples were polished with very fine sand paper in order to obtain the same surface roughness as competition runners.
The Engineering of Sport 7 - Vol. 1
581
Figure 1 - Cross-section of a luge steel blade. Table 1 - Cross-sections of the samples used for the laboratory experiments.
2.2 Measuring devices and methods 2.1.1 Impact friction tester The impact friction tester measures the friction force and the penetration of different materials on ice or snow. The device is composed of a Kistler force plate and a movement controller onto which the sample is fixed (see Figure 2) and which can be moved horizontally and vertically. The force plate measures the normal, the friction and the lateral force when the sample is in contact with the ice. The movement controller records the vertical and the horizontal displacement. The experimental data was recorded at a sample rate of 25 kHz. For the present study, the normal force was set to 100 N and 200 N, which represent the forces applied on a length of 0.1 m of each blade of a luge on straights and high radius curves. The constant sliding speed was 1 m/s. The room and the ice temperatures were similar and set to –2°C and –10°C. The air humidity oscillated between 20 and 50%. Ice blocks were made of common tap water in aluminium boxes. The surface of the ice was flattened and smoothed by a cutting procedure with a sharp blade.
582 The Engineering of Sport 7 - Vol. 1
Figure 2 - Impact friction tester used to measure the friction and penetration of luge steel blades on ice.
2.1.2 Ice track analysis In order to obtain a precise description of the track left by the blade into the ice in depth and in width, imprints were taken. The imprints were also used to control the roughness of the ice surface and the reproducibility of the ice preparation procedure. The tracks’ imprints were made with the Mikrosil two components silicon paste. The accuracy of the paste is within 1 micrometer. The imprints were analysed with a MicroProf® optical device developed by Fries Research & Technology GmbH. The MicroProf® is a noncontact device for high precision surface measuring procedures as contour, roughness and topography. The sample is illuminated by a focused white light. An internal passive optics, using chromatic aberration, splits the white light into different colours corresponding to different wavelengths. A miniaturized spectrometer detects the colour of the light reflected by the sample and determines the position of the focus point, and by means of an internal calibration table, the vertical position measured on the sample surface. The Mark III® software was used to analyze the measurements.
2.1.3 Ice hardness measurements Because of the significant creep of ice (Bowden 1955, Barnes et al. 1971), the hardness of ice has to be measured with a dynamic test corresponding to a very short loading time as expected during luge. The Equotip® measuring device from Proceq S.A. was used to measure the ice hardness. The device is composed of an impact body with a hard metal tip (Ø 3mm), which is propelled by a spring against the surface of the ice. Surface deformation takes place when the impact body hits the test surface, which results in loss of kinetic energy. This energy loss is calculated by velocity measurements. The voltage of the signal is proportional to the velocity of the impact body, and signal processing by the electronics provides the hardness reading for display and storage. The measured hardness “L” is equal to the bounce velocity divided by the impact velocity. Conversion tables
The Engineering of Sport 7 - Vol. 1
583
can be used to compare the results with hardness values of other measuring methods (Brinell, Vickers, Rockwell) with a precision of 3%. The device can be used at temperatures between –20°C and +120°C. The average values result from 10 to 15 measurements. The measurement uncertainty is ±3%. The EQUOTIP is sensitive to the surface roughness, which must not exceed 2μm.
3- Results and discussion 3.1 Results 64 measurements were conducted with the impact friction tester. The measured friction coefficients ranged from 0.01 to 0.18. Penetration depths were measured between 0.005 mm and 0.35 mm depending on the ice temperature, normal force and sample tested. Results clearly showed that the samples G1 and G2, with the low outside angle, enabled the lowest penetration depths and friction coefficients on ice for both temperatures (see Figure 3).
Figure 3 - Boxplots of the coefficient of friction (left) and the penetration depth (right) into the ice for the four samples measured with the impact friction tester.
It is showed on Figure 4 that an increased penetration depths leads to a higher friction coefficient. Further, for the same penetration depth, the coefficient of friction is significantly higher at –10°C then at –2°C.
Figure 4 - Influence of penetration of the friction coefficient. Results were obtained with the impact friction tester.
584 The Engineering of Sport 7 - Vol. 1
3.2 Verifications 3.2.1 Accuracy of the impact friction measurements The penetration depths into the ice recorded by the impact friction tester were compared to the ones determined by optical analysis of imprints of the tracks. Figure 5 shows an optical analysis of the traces left by the four blades into an ice block. Penetration depths correlated up to 95 % between the two analyses methods indicating the good accuracy of the impact friction tester.
Figure 5 - Optical analysis of an imprint of the traces left by the blades on the ice during impact friction testing. The penetration depth measured with the impact friction tester is written under each trace.
3.2.2 Comparison of laboratory and field conditions Ice hardness measurements and imprints of runners traces were made on the artificial luge track of Oberhof on January 12th, 2007. Ice temperature was –2°C. The traces left by the blades had a depth between 0.06 and 0.08 mm on the flat parts which corresponds well to the values that were measured in the laboratory for the same loading situation. The mean ice hardness, L, had a value of 130 +/- 15 which corresponds well with the mean laboratory value of 121 +/- 25 measured at the same temperature. It was concluded that the laboratory experiments reproduce well the field conditions.
4- Conclusion and future work The influence of three main parameters, which determine the blades’ cross-sections, on the friction and penetration depth into the ice was analysed in a cold laboratory under controlled conditions with a new testing device. The results showed the high impact of the blades’ outside angle on penetration and friction. Further, it could be shown that this device is very accurate and reproduces the field conditions well. New cross-sections regarding low friction and specific penetration into the ice for good steering ability for
The Engineering of Sport 7 - Vol. 1
585
various types of ice and loading situations were proposed according to the results of the laboratory measurements. The new geometries have been tested on the field. Primary results seem to confirm the findings of this study. Analytical calculations of the real contact area and the penetration into the ice for different loading situations, ice temperatures and blade cross-sections are being developed. The aim of these calculations is to propose further blade geometries. Finally, further laboratory measurements are planed to analyse the effect of surface roughness, material type and surface treatment on the friction coefficient of the blades as no restrictions from the international luge federations exist concerning the aforementioned properties of the blades.
5- Acknowledgment The authors would like to thank the financial support of Swiss Olympic and the good collaboration with the Swiss national luge team.
6- References [AE1] Akkok M., Ettles C., Calabrese S.J., “Parameters affecting the kinetic friction of ice”, Journal of Tribology, 109 552-561, 1987. [BT1] Barnes P., Tabor D., Walker J.C., “The friction and creep of polycrystalline ice”, Proc. Roy. Soc. Lond. A., 324 127-155, 1971. [B1] Bowden F.P., “Friction on Snow and Ice”, Proceeding of the Royal Society of London. Series A, Mathematical and Physical Sciences, 217 (1131) 462-478, 1953. [CN1] Colbeck S.C., Najarian L., Smith H.B., “Sliding temperatures of ice skates”, Am. J. Phys., 65 (6) 488-492, 1997. [E1] Ericksson R., “Friction of runners on snow and ice”, Snow, Ice and Permafrost Research Establishment, Corps of Engineers, US Army, 1955. [EN1] Evans D.C.B, Nye J.F., Cheeseman K.J.,“The kinetic friction of ice”, Proceeding of the Royal Society of London A, 347, 493-512, 1976. [FT1] Forland K.A., Tatinclaux J.-C. P., “Kinetic friction coefficient of ice”, CRREL Report 85-6 ,1985. [H1] Hainzlmaier C. , A tribologically optimized bobsleigh runner, Zentralinstitut für Medizintechnik, Technische Universität München, 2005. [IH1] Itagaki K., Huber N.P., Lemieux G., “Dynamic friction of a metal runner on ice”, CRREL Report 89-14, 1989. [OK1] Oksanen P., Keinonen J., “The mechanism of friction of ice”, Wear, 78 315-324, 1982. [P1] Petrenko V.F., “ The effect of static electric fields on ice friction”, Journal of applied physics, Vol. 76 (2), 1216-1219, 1999. [PC1] Petrenko V.F., Colbeck S.C., “Generation of electric fields by ice and snow friction”, Journal of applied physics, Vol 77(9), 4518-4521, 1995. [PR1] Petrenko V.F., Ryzhkin I.A.., “Surface state of charge carriers and electrical properties of the surface layer of ice”, Journal of physical chemistry B, Vol 101, 1997. [PW1] Petrenko V.F., Whitworth R.W., “Physics of ice”, Oxford University Press, 1999. [P2] Petrovic J.J., “Review - Mechanical properties of ice and snow”, Journal of Material Science, 38 1-6, 2003.
586 The Engineering of Sport 7 - Vol. 1 [R1] Roy R.K, “Design of experiments using the Taguchi approach”, John Wiley & Sons, Inc., New York, 2001. [S1] Schulson E.M. , “The structure and mechanical behaviour of ice”, JOM, 51 (2) 21-27, 1999. [T1] Tusima K., “Friction of a steel ball on a single crystal of ice”, Journal of Glaciology, 19 (81) 225-235, 1977.
Brake Induced Vibration in Mountain Bikes (P112) Robin C. Redfield1
Topics: Cycling, vibration, braking. Abstract: Mountain biking design continues to evolve towards more sophisticated systems including frame structures, suspension strategies, and braking components. Frames become lighter with directionally designed stiffness. Suspensions are designed and tuned to respond to diverse riding scenarios including large and small bump compliance, high and low speed compression and rebound damping, and pedal feedback/suspension bob. Braking components are getting more powerful, with better modulation, while maintaining stiffness and weight character. As structures become lighter and brakes become stronger, interactions between the stick-slip friction of brake pads and rotors and the stiffness of rear frame structures sometimes induce limit-cycle vibration that can be transmitted throughout much of the bicycle frame. This annoying vibration can be felt and heard and is more often seen in relatively compliant frame structures with powerful, state-of-the-art disk brakes. This paper examines the dynamic causes of these vibrations and relates them to design and operational parameters. Kinetic versus static friction, structural stiffness and damping, brake power and modulation rate, and bike speed and brake geometry are all key contributors to the phenomenon. Specific relationships between these parameters that preclude the onset or continuance of this vibration are determined. Key words: Mountain biking, braking, vibration.
1- Introduction As mountain bike design increasingly answers to higher performance demands by the customer, engineering requirements become more constrained, there is less room for “over-design,” and unintended consequences emerge. Bikes must become lighter, stiffer, faster, and more. Examples of adverse consequences include suspension components overheating as size and weight diminish but energy absorption requirements increase; structures flex and vibrate as frame weight decreases, stiffness is directionally designed and inputs from the ground and braking increase. 1. Department of Engineering Mechanics, United States Air Force Academy, Colorado Springs, CO USA E-mail: [email protected]
588 The Engineering of Sport 7 - Vol. 1 It is the second example that is the topic of this paper. The combination of more powerful and quickly initiated braking coupled with insufficient frame stiffness in the “rear triangle” of a full suspension bike has resulted in an increasing occurrence of braking induced, limit-cycle vibration. The stick-slip nature of sliding friction can continually excite rear frame vibration that is often transmitted throughout much of the bicycle. The rider can feel the vibration in the seat and sometimes in the handlebars. This is a relatively low frequency event, not to be confused with brake squeal where the sound issue is paramount. It is the vibration of the rear structure (rear triangle or swing arm) of a full suspension mountain bike due to the input of braking forces through a pad and caliper (disk braking). Basic stick-slip friction in introduced in [R1]. More sophisticated analyses, including non-linear dynamics, chaos, and bifurcation effects are in [M1], [HS1], and [PS1]. The balance of the paper investigates single degree-of-freedom (DOF) vibration models excited by stick-slip friction. The first sections look at undamped, forced, and then damped phase plane responses. Next, stick-slip excitation is considered. The effects of system stiffness and damping, and of brake force and modulation, on vibration onset and continuation are examined.
2- Structural Vibration Model Braking induced structural vibration involves structural stiffness and damping, and applied forces from sliding friction between the pads and rotor. The next sections look at these effects separately and then together. Consider the forced, one DOF vibrating system in Figure 1 with structural mass m, damping coefficient b, and stiffness k. A braked bicycle is just such a system. The frame serves as a distributed mass with structural stiffness and damping and the braking acts as an applied force. The standard equation of motion is: (1) With no forcing, F, or damping, b, the time response for position comes from solving Eq. (1): (2) is the undamped natural frequency and A and B are constants determined by initial conditions of position and velocity. This would model a bike frame with negligible damping and no applied force subject to an initial condition (IC).
The Engineering of Sport 7 - Vol. 1
589
Figure 1 - Vibration structure with applied force. .
For the system starting at rest with an initial compression (x(0) > 0, x(0) = v(0) = 0), the time response of position x and normalized velocity (v/n), and a phase-plane plot of velocity versus position are graphed in Figure 2. Velocity is graphical horizontally and position vertically to demonstrate their contributions to the phase-plane result. Note that dividing the velocity by the natural frequency results in equal position and velocity amplitudes. A measure of distance traveled in the phase plane will be referred to as s.
Figure 2 - Phase plane representation for free, undamped vibration
The position and velocity are 90 degrees out of phase (due to the lack of damping) which results in a circular phase-plane trajectory that starts at the initial position on the positive horizontal axis. The radius is determined by the initial conditions. Note that if the initial condition were at the origin (at rest with zero spring displacement), the circle would have zero radius and the system zero response.
590 The Engineering of Sport 7 - Vol. 1
3- Damped, forced vibration If we add damping and a constant applied force to our system, the time response is the total solution of Eq. (1) and becomes (3) where is the damping ration and d is the damped natural frequency. The oscillations decay with the exponential and the position is offset in the direction of the applied force. Figure 3 shows the phase plane of a forced and damped system spiraling into its non-zero equilibrium. The same initial condition is used here as before.
Figure 3 - Vibration with damping and forcing.
If the initial conditions were at the origin (zero velocity and position), the trajectory would start with initial radius of F/k and still spiral and decay to the same equilibrium. The instantly applied force would initiate the vibration. The next vibration model examines the effect of Coulomb (stick-slip) friction, so we look at those characteristics first.
4- Coulomb (stick-slip) friction During braking, equal and opposite forces are developed on the rotor and caliper (through the brake pads). These are stick-slip friction forces that have distinct static and kinetic behavior (Figure 4).
Figure 4 - Stick-slip friction behavior.
The Engineering of Sport 7 - Vol. 1
591
When there is no slip velocity (locked, static condition), the friction force can be anywhere between the maximum and minimum static friction values, ±Fs. The maximum static force is proportional to the normal force between the surfaces through the friction coefficient μs. When the surfaces slip, the friction force drops to a lower, kinetic value of ±Fk. Some experimental evidence shows that this drop may not happen instantly and a more gradual drop is shown in the alternative model. Kinetic friction also depends on the normal force which, in the case of disk brakes, is applied by lateral pistons forced by hydraulic pressure. This paper employs the discontinuous model shown in bold.
5- Vibration with stick-slip friction Figure 5 shows an undamped system sliding under the influence of stick-slip where sliding friction is less than static, and the direction of the slip velocity controls the sign of the friction force. When the mass is moving positively (to the left), the friction force is constant and to the right. When the mass is moving to the right, the friction force is to the left.
Figure 5 - Structure with sliding friction.
The phase-plane response of the undamped system with stick-slip friction is in Figure 6 which shows the results for three different initial conditions. If the system is given an initial displacement greater than ±Fs/k, the spring overcomes static friction and the system slides, otherwise it stays at rest. During sliding, the system is oscillatory with a constant applied force (Fk, the kinetic friction force) so the trajectory is circular (as in Figure 2) and clockwise about the center, ±Fk/k (as in Figure 3). When the velocity is positive (trajectory A), the friction force is negative and the trajectory orbits about - Fk/k; when the velocity is negative (trajectory B), the orbit is about + Fk/k. The radius of the orbit depends on the initial condition as previously.
592 The Engineering of Sport 7 - Vol. 1
Figure 6 - Phase plane with stick-slip friction.
Once the velocity is zero, the system sticks (stops) if the static friction limit is greater than the spring force, Fs/k < x (trajectories A and B). Otherwise the system reverses and slips in the opposite direction while orbiting around a new center as in trajectory C. The change in center is due to the sign change of the friction force. If viscous damping is included for any of these cases, the trajectory spirals towards a center as in Figure 3.
6- Complete vibration model Figure 7 shows a schematic of a braking structure (effective mass, stiffness, and damping of m, k, b) forced by a moving rotor. During brake actuation, the friction between the brake pad and rotor causes a leftwards force on the braking structure.
Figure 7 - Brake and bike structure schematic.
The Engineering of Sport 7 - Vol. 1
593
As long as the structure is moving slower than the rotor, kinetic friction acts, and the friction force causes the structure to deform. If the structure does not attain rotor velocity, the response is of an oscillatory system with a constant applied force. If the structure accelerates up to the rotor velocity, the brake pad will stick to the rotor and continue to deform. Slip will again occur when the restoring force from the deforming structure overcomes limiting static friction.
6.1 “Low” instantaneous brake force Figure 8 shows the phase-plane response in the former case for an undamped system with an instantaneously applied friction force of Fk. The friction force is not great enough to get the structural velocity up to the rotor velocity, i.e. Fk/k < vr/n. The trajectory starts at the origin (zero ICs) and orbits the center of Fk/k with radius r = Fk/k. If damping were present, the trajectory would spiral toward the center and stop. Lower brake forces with more structural stiffness and higher bike velocities with lower frame natural frequencies preclude the structure from being accelerated to the rotor velocity and hence sticking.
Figure 8 - Phase plane response with no sticking; Fk/k < vr/n
6.2 Undamped case, “high” brake force If the braking force is greater, Fk/k < vr/n, the trajectory first orbits its center with r1 = Fk/k (the a-slip portion of the trajectory) in Fig. Figure 9. When v = vr stick occurs and the structure/rotor move together; the structures deforms until the elastic restoring force reaches the limits of static friction (b-stick portion). Slip then occurs again (c-slip) and the trajectory orbits the same center at a new radius of (4)
594 The Engineering of Sport 7 - Vol. 1
Figure 9 - Phase plane response with stick-slip; Fk/k > vr/n.
At the end of this circular phase of the trajectory, structural velocity again reaches rotor velocity and stick re-occurs. This stick-slip cycle continues indefinitely. Note that it is the difference between static and kinetic friction forces that determines the new (cslip) trajectory radius. The larger this radius, the more likely limit-cycle stick-slip will occur.
6.3 Damped case If sufficient damping is added to the system, the trajectory radius during the c-slip phase can be reduced so sticking does not re-occur as in Fig.Figure 10. The vibration then spirals into an equilibrium displacement of Fk/k. The ratio of F/k and damping controls the whether the system decays or continues in a limit-cycle.
Figure 10 - Damped system with stick-slip; Fk/k > vr/n.
The Engineering of Sport 7 - Vol. 1
595
6.4 - Non-instantaneous braking If the applied braking force is assumed to be exponential (or ramped) in nature instead of instantaneous, the trajectory flattens out substantially during brake application. dr/ds (s was defined in Figure 2) is greater than with instantaneous braking. The center of the trajectory, Fk(t)/k, moves with time from the origin to its steady state center, Fk( )/k (and the trajectory always spirals about this center). This results in reducing the maximum attained velocity of the structure. This reduction may avoid any stick in the first place resulting in just damped oscillation until equilibrium is reached as in Figure 11. Obviously, it is not generally advantageous to moderate how quickly brakes come to full power. Compliance in the mechanisms and hydraulics, and dead bands are the chief factors in non-instantaneous brake force application.
Figure 11 - Phase plane with ramped braking; sticking can be avoided.
7- Discussion The simple model of this paper modeled rear, bike frame vibration including structural mass, stiffness, and damping excited by disk brake induced stick-slip friction. The nature of this type of friction, with static values always above kinetic, is the cause of the limitcycle behavior. If Fs were equal to Fk, stick might still occur, but with any amount of damping re-stick would not. Given some F (Eq. (4)), bike and brake designers have some of options. Since fast acting and powerful brakes are not too “negotiable” (at least for most medium to high end bikes), material selection to minimize the difference in static and kinetic coefficients at the rotor and pad is very desirable. In terms of structures, stiffer rear frames in the direction of brake forces (usually in the wheel plane) and stiffer brake mounting hardware would be beneficial. Suspensions are often laterally stiff for bike handling but more compliant in other directions due to suspension linkages and material cross section. Also, adding a relatively small amount of structural (or other) damping can mitigate the re-stick regardless of stiffness. Brake hangers, for example, could be designed with intentional damping. Frame elements could also use laminates with damping character.
596 The Engineering of Sport 7 - Vol. 1
8- References [HS1] Hetzler, H., Schwarzer, D. and Seemann, W., 2007, “Analytical Investigation of Steady-State Stability and Hopf-Bifurcations Occurring in Sliding Friction Oscillators with Application to Low-Frequency Disk Brake Noise,” Communications in Nonlinear Science and Numerical Simulation 12, pp. 83-99. [M1] McMillan, A. J., 1997, “A Non-Linear Friction Model for Self Excited Vibrations,” Journal of Sound and Vibration, Vol. 205, No. 3, pp. 323-335. [PS1] Popp, K. and Stelter, P., 1990,“Stick-Slip Vibrations and Chaos,” Philosophical Transactions: Physical Sciences and Engineering, Vol. 332, No. 1624, pp. 89-105. [R1] Rao, S., 1995, Mechanical Vibrations, 3rd edition, Addison-Wesley, New York.
Aerodynamic Optimization and Energy Saving of Cycling Postures for International Elite Level Cyclists (P114) Luca Oggiano1, Stig Leirdal2, Lars Sætran3, Gertjan Ettema4
Topics: Sport aerodynamics. Abstract: Introduction: Drag in cycling counts for as much as 90% of total resistance opposing motion in a normal time-trial course. A small gain in term of drag reduction can, over a longer time-span (30 - 60 minutes) give a large advantage to the cyclists in terms of power output saved or velocity gained. The aim of present study was to aerodynamically optimize the cycling posture for each cyclist and thereby improve the athletes’ performances. We also wanted to quantify the power output saving, velocity gains and energy savings of this optimization. Methods: 11 elite cyclists with a maximal aerobic power output of 481 W were tested at 6 different positions on their time-trial bicycle in a wind tunnel with an air flow at 14.5 m/s. All positions were adjusted from their regular position and included both adjustment of seat and handlebar. All cyclists also went through an extensive physiological test, including lactate threshold and VO2max tests, allowing for individual efficiency calculations at several power outputs. Results: From the wind tunnel test individual power-output – velocity curves were plotted, showing the effect of the different positions in terms of saved power generation. Showing an average 21.9 W saving in power output and an average of 0.75 km/h gain in velocity at 500 W for the most aerodynamically position. Using each cyclist’s efficiency we calculated the theoretical effect of oxygen consumption, Kcal/h and heart rate. Average results show a 0.34 l/min, 101.5 Kcal/h and 14 BPM for the heart, in saving, for the most aerodynamically position. Conclusions: The effect of small adjustments on elite cyclists can have large effect on performance and energy saving. However, care should be taken as the new position can negatively affect power generation and pedaling technique, which might become more energy consuming. Data on this is also collected in present study but needs further analysis. Keywords: Aerodynamics, drag, cycling. 1. Norwegian University of Science and Technology, Faculty of Engineering Science and Technology, N-7491 Trondheim, Norway - E-mail: [email protected] 2. Norwegian University of Science and Technology, Faculty of Social Sciences and Technology Management, Human Movement Science programme, N-7491 Trondheim, Norway - E-mail: [email protected] 3. Norwegian University of Science and Technology, Faculty of Engineering Science and Technology, N-7491 Trondheim, Norway - E-mail: [email protected] 4. Norwegian University of Science and Technology, Faculty of Social Sciences and Technology Management, Human Movement Science programme, N-7491 Trondheim, Norway - E-mail: [email protected]
598 The Engineering of Sport 7 - Vol. 1
1- Introduction Performances in cycling are affected by a number of factors. Together with the important evaluations about physiology, biomechanics and all the possible research about materials, an accurate analysis of the forces acting against a cyclist during his motion is highly important. Considering a flat track, several forces act on a cyclist during his race (fig 1). Using the second Newton’s law, the force balance along the x-axis gives F = max + FR + C + D
(1)
and along the y-axis: mg = 2T
(2)
Where F is the total force produced by the cyclist during his cycling action, FR is the rolling resistance is the inertia force, C is the transmission losses (which can be neglected), D is the aerodynamic resistance or drag, and T is the reaction force.
Figure1 - Forces acting on a cyclist
1.1 Rolling resistance The rolling friction on the front wheel is different from the rolling friction on the rear wheel. This is due to the weight distribution on the bike. A simplified formula permits to consider the total rolling resistance as the sum of the front wheel rolling resistance (RTF) and the rear wheel rolling resistance (RTR). FR = RTF + RTR
(3)
Many studies have been carried out in order to determine FR. According and [KB1], can be written as follow:
The Engineering of Sport 7 - Vol. 1
FR = cR1mg + cR2V
599 (4)
Where cR1 and cR2 are constants. In some studies the dependency of FR on the speed has been neglected ([GCBR1], [CRBFD1], [D1], [DCMS1], [P1]) and thus FR has been considered proportional to the rolling coefficient cR1. The non dependency of FR on the speed gain importance when it comes to analyze the total forces resisting the motion of a cyclist. While D increase with the square of the speed FR is constant. Thus, when the speed increases, the importance of rolling resistance compared to the drag decreases.
1.2 Drag Aerodynamics has a central role. Drag resistance is in fact the major part of the total forces acting against a cyclist and it counts up to 90% at 14m/s (fig. 2a). A gain in terms of drag reduction permits a sensible improvement in performances. (fig. 2b)
Figure 2 - (a) Drag percentage on total force. Due to the quadratic relation between drag and speed, the drag gain more importance at high speeds (constant rolling friction coefficient cr of 0,003 and a standard mass of 80kg) [GKM1], [H1]). (b) Maximum speed reachable with different power values.
Considering the total system (cyclist + bike), seventy percent of the total drag is due to the cyclist and 30% is due to the bike drag [GKM1]. Under these assumptions, the cyclist position has a great importance in cycling performances. An aerodynamic efficient posture can give a huge advantage during competitions. The drag can be written as: (5) where A is the frontal area, Cd is the non dimensional drag coefficient and depends from the shape, r is the density of the air and V is the speed. The square dependency of the drag from the speed is significant because, a small reduction in Cd, at high speed, means a high reduction of the drag. This effect gain even more importance when it
600 The Engineering of Sport 7 - Vol. 1 comes to analyze the power spent in order to overcome the drag (fig 3). The power is in fact proportional to the cube of the speed (6) However, cyclist posture is rigorously related to his anthropometric characteristics and, often, an aerodynamic optimized position do not give good results in terms of force production. Optimization between biomechanics and aerodynamics is then highly important in order to find postures which can gives some advantages. The cyclist posture has been studied and analyzed for many years. Di Prampero [DCMS1], Kyle [KB1], and Hennekam [H1] showed some postures are more efficient than others while Capelli [CRBFD1] showed the importance of the bike itself in order to reduce the drag.
2- Methods 2.1 Subjects After approval form the regional ethical committee 9 male and 2 female elite cyclists with an average age of 22.2 (5.9) years, signed an informed consent to participate in the study. Average height and weight was 181.6 (5.8) cm and 73.8 (7.4) kg. Subjects had a maximal aerobic power of 481 (75) W and an average VO2max of 4.66 (0.59) l/min. Average maximal heart rate was 198 (5.2) BPM. Their lactate threshold power (lactate = 4 mmol/l hemolysed blood) was 297 (47) W at an average VO2 consumption of 3.82 (0.53) l/min and a heart rate of 175 BPM (5.6). Average energetical efficiency was 21.8 (1.02) %.
2.2 Wind Tunnel For the experiments, the wind tunnel of NTNU (Norwegian University of Science and Technology) in Trondheim has been used. The contraction ratio is 1:4,23, and the test section of the wind tunnel is 12,5 meters long, 1,8 m high, and 2,7 m wide. The wind tunnel is equipped with a 220KW fan that can produce a variation of speed between 0,5 - 30 m/s. The balance (Carl Schenck AG) used is a six components balance capable to measure the three forces and the three momentums around the three axes. Variations of forces and momentum are measured using strain gauges glued to the balance body. The voltage outputs are measured by a LABVIEW based PC program.
2.3 Aerodynamic test All cyclists were tested and 6 different positions (table 1), starting for their natural position were tested. The test has been run at 14.5m/s and the drag acquired for each posture has been acquired 5 times and averaged. All positions were adjusted from their regular position and included both adjustment of seat and handlebar and were within the UCI rules for bicycle geometry in relation to body size.
The Engineering of Sport 7 - Vol. 1
601
From the drag measurements at 14.5 m/s, the power output that each cyclist had to generate was calculated (F • v) and a power – velocity curve was fitted down to 0 km/h for each cyclist. Table 1 - The six positions tested in the wind tunnel. All positions are adjusted form each cyclists normal position.
Position
Seat
Handlebar
1 2 3 4 5 6
Normal Up 15 mm Down 15 mm Normal Up 15 mm Down 15 mm
Normal Normal Normal Down 20 mm + forward 20 mm Down 20 mm + forward 20 mm Down 20 mm + forward 20 mm
2.4 – Physiological methods To define the individual performance level all subjects performed a lactate threshold protocol and a VO2max test. The lactate threshold protocol started at 200 W for men and 150 W for women. Each 4 min workload was increased by 25 W. Gas exchange (Jaeger Oxycon Pro – mixing chamber) was measured the last minute of each workload and blood lactate (Lactate Pro) and heart rate (Polar s610i) at the end of each workload. The lactate threshold protocol was terminated when lactate was above 4 mmol/l haemolysed blood (lactate threshold) From the gas exchange measurements on the sub-maximal workloads (lactate < 4 mmol/l blood) gross energetical efficiency was calculated using the correct energy equivalent for oxygen based on the respiratory exchange ratio (RER = VCO2/VO2). The energetical efficiency was used for calculating the individual energy saving for the different positions in the wind tunnel. VO2max was tested starting at 300 W for men and 200 W for women with of 25 W increase in workload pr min. VO2max was defined as the highest average oxygen consumption over 1 min. Both physiological tests were performed at freely chosen pedal rate (FCPR) at constant power outputs using a bicycle ergometer (Velotron) with an electro-magnetic brake mechanism creating resistance. Subjects wore cycling shoes and the seat and handle bar position on the ergometer was adjusted to the preferred sitting position for each subject.
3- Results and discussion Position 4 (table 2) is resulted, in average, the position which gave the best performances in terms of drag reduction. This position has been obtained simply adjusting the handlebar. In order to evaluate the drag reduction per each athlete, a Drag reduction factor has been defined as:
602 The Engineering of Sport 7 - Vol. 1
(7) Position 1 2 3 4 5 6
%Gain
St.Dev
———— 0.8970 -0.864 -5.791 -4.078 -3.126
———3.050 2.060 5.326 3.995 4.806
However, not all the cyclists present in the test got the same improvement using the position 4 but results were quite scattered as confirmed by the standard deviation values. This is due to the fact that the starting position for each cyclist is his normal cycling posture and some of them have already empirically optimized their position with the experience acquired during trainings and competitions. Thus, an individual analysis in order to achieve the best results is strongly recommended. The physiological effects and benefits of each position have been estimated and calculated for each of the cyclist tested and then averaged. However, all cyclists told that the position tested did not affect pedalling technique. In fact, most of the subjects change their position permanently after these measurements. The physiological effects of each position have been estimated from the energetical efficiency for each of the cyclist tested and then averaged. Results in figure 3, showing position 4 to give 21.9 W saving in power output at 50 km/h (this speed can be maintained for approximately 30mins). Or from a different perspective, position 4 gives a 0.75 km/h gain in velocity at 500 W. Calculated from the energetical efficiency of each cyclist, position 4 gives an average oxygen consumption saving of 0.34 L/min which is a 7.3 % reduction compared to the cyclists maximal aerobic capacity (VO2max). For elite cyclist an equal increase of their maximal oxygen consumption would be a whole year of training to give the same effect. If we assume that heart rate and oxygen consumption are linear and that maximal heart rate and maximal oxygen consumption is synchronous, we calculate that position 4 in 50 km/h gives a heart frequency that is 15 BPM lower than at their normal position. This also gives an estimated 101.5 Kcal/h of energy saving, which is a substantial amount of energy. In conclusion, we have shown that adjusting the handlebar down 20 mm and forward 20 mm have a substantial effect on performance (gain in km/h at a given power output) or in energy saving (reduction in effort at a given velocity.
The Engineering of Sport 7 - Vol. 1
603
Figure 3 - (a) Change in power output per velocity per position. (b) Estimated change in oxygen consumption per position per velocity. (c) Estimated change in % of maximal aerobic capacity per position per velocity. (d) Estimated change in heart rate per position per velocity. (e) Estimated energy saving per position pr velocity. (f) Estimated change in velocity pr position at 400 W power output.
4- Conclusions The experiments carried out permits to affirm that a posture optimization is possible for each cyclist. However, even if a mean trend has been found, each athlete showed different results per each position tested. A specific individual test is then mandatory in order to obtain advantages from the aerodynamic point of view. Furthermore, each athlete has different anthropometric values which should be taken into consideration in order to maintain high biomechanics efficiency with the postures tested.
604 The Engineering of Sport 7 - Vol. 1 In average, posture 4 is the one who gave the best results in terms of drag reduction with consequents advantages from the physiological point of view. At the same time, posture 2 resulted to be ineffective for most of the athletes tested.
5- References [CRBFD1] - Capelli, C., Rosa, G., Butti, F., Ferreti, G., Veicsteinas, A. and Di Prampero, P. E. ”Energy cost and efficiency of riding aerodynamic bicycles”. European Journal of Applied Physiology. (1993) 67, 144-149. [D1] - Davies, C. T. M. 1980. ”Effect of air resistance on the metabolic cost and performance of cycling”. European Journal of Applied Physiology, 45, 245-254. [DCMS1] - Di Prampero PE, Cortili G, Mognoni P, and Saibene F. ”Equation of motion of a cyclist”. J Appl Physiol 47: (1979) 201-206, 1979. [GCBR1]-Grappe F, Candau R, Belli A, and Rouillon JD,. “Aerodynamic drag in field cycling with special reference to the Obree’s position”. Ergonomics 40,12. (1997): 1299-1311. [GKM1] - Gross, A. C., Kyle, C. R. and Malewicki, D. J. ”The aerodynamics of human-powered land vehicles”. Scientic American, 249, (1983) 126- 134. [H1]-Hennekam W.. “The speed of a cyclist”. Physical education 25,12. (1990): 1299-1311. [KB1] - Kyle CR, Burke E.R. ”Improving the racing bicycle”. Mech Eng. (1984) 34-45 [P1] - Pugh, L. G. C. E.”The relation of oxygen intake and speed in competition cycling and comparative observations on bicycle ergometer”. Journal of Physiology, 241, (1974) 795 - 808
A Comparison of Test Methodologies to Enable the Improved Understanding of Soccer Boot Traction (P115) J.D. Clarke1, M.J. Carré1, R.F. Kirk1
Topics: Footwear traction. Abstract: The difficulties of mechanically representing actual human movements have long been recognised. However, repeatable mechanical tests are necessary to quantify the characteristics of footwear-surface interactions. Improving understanding of the interactions between footwear and surfaces can advise athletes on the optimum choice of footwear in order to balance their athletic performance with injury risk. Many mechanical test devices have been developed to measure traction, generally simulating the movement of a shoe over a sports surface. Current velocity controlled devices measure the peak traction that occurs during a movement. However, high-speed video analysis suggests the peak traction does not represent an athlete losing performance when slipping. A force controlled traction rig has recently been developed to examine a different approach. This rig measures the initial resistance to movement, when the surface first fails. Load cells in the horizontal and vertical directions measure the resistance to movement of a loaded studded plate allowing different interaction mechanisms to be studied. A comparison between the methodologies used in the traction testing of sport surfaces is discussed. Results from velocitycontrolled and force-controlled rigs are presented and the virtues of each method compared. Keywords: Soccer, Sport Surface, Traction, Sport Injury, Studs.
1- Introduction Many manoeuvres carried out by football players require sufficiently high traction between the shoe and surface interface. However excessive traction can increase the risk of injury (Shorten 2007). A large number of test methodologies have been developed in order to investigate sport surface traction. The motivation behind such research is to optimise footwear properties for particular surface conditions, to reduce the risk of injury while maintaining or increasing athletic performance. 1. Department of Mechanical Engineering, Universit, Sheffield, Mappin Street, Sheffield, S1 3JD - E-mail: j.d.clarke m.j.carre, [email protected]
606 The Engineering of Sport 7 - Vol. 1 Numerous test methods involve the analysis of human movement. Anderson et al discuss the difficulties of understanding the player interactions leading to a football injury. They argue video based analysis can help determine real injury mechanisms (Andersen et al. 2003). Video observations of actual body movements can help establish the boundary conditions required to be accurately replicated by mechanical test devises (Carré et al. 2007). Mechanical test devices must therefore try to replicate the forces and velocities that appear in human movement (Shorten 2007). Dura compared friction tests with three dimensional player observations. He observed athletes adapting their movements according to the friction they experienced. Therefore the friction coefficient measured with a DIN 18032 Friction Machine differed from those measured with human subjects as it did not replicate the athlete movements performed (Durá 1999). González et al highlighted the difficulty for an athlete to naturally recreate sport movements in lab conditions, finding “significant statistical differences” between five subjects performing the same five “typical” movements (González et al. 2004). Mechanical test devices have the advantage of creating objective loading conditions which provides a repeatable measure of shoe-surface traction.
2- Velocity Controlled Test Devices Many mechanical traction test devices involve dragging a loaded sled along a surface with a shoe attached. The traction force is sampled throughout the motion and the maximum traction during the movement is usually taken as the traction parameter (Carré et al. 2007). Figure 1 shows how, with velocity controlled devices, the peak traction force measured can occur after movements in excess of 50 mm.
Figure 1 - Typical traction data from SERG velocity-controlled traction rig for two different stud types on third generation artificial turf (Carré et al. 2007).
Following high speed video observations of players performing football movements it is argued that the initial movement of a shoe through a surface is a more appropriate factor to examine. It is during the initial movements a player is likely to experience a lose
The Engineering of Sport 7 - Vol. 1
607
of performance and slip (Kirk et al. 2007). The peak traction measured by velocitycontrolled traction test devices may not be suitable to classify the traction during initial slipping of a player, as the peak recoreded form the velocitycontrolled device occurs after large displacements. It is therefore suggested a traction device which measures an increasing force until slipping occurs would be a more appropriate mechanical test (Carré et al. 2007).
3- SERG Force Controlled Traction Device In response to the high speed video analysis, a prototype force controlled traction rig has been developed, as shown in Figure 2.
Figure 2 - The SERG prototype traction rig.
A hydraulic ram provides a controlled constant vertical load applied to a stud plate. The plate was designed to allow the attachment of different stud types and configurations. A high pressure pneumatic ram provides a dynamic, increasing force in the horizontal direction. The horizontal load is controlled by a solenoid valve which is gradually opened to increase the force in the cylinder/ram. Load cells in the horizontal and vertical direction and a horizontal linear displacement voltage transducer (LDVT) provide the necessary data to measure an initial resistance to movement. Voltage signals from the load cells and LDVT are sampled simultaneously, via strain indicator boxes and a data acquisition device (National Instruments model number NI USB-6008), in real time and displayed in LabView (version 7.1 National Instruments). The signals were sampled and transformed into force and displacement measurements. An example of the variation of horizontal and vertical force with displacement is shown in Figure 3.
608 The Engineering of Sport 7 - Vol. 1
Figure 3 - Typical traction data from SERG force-controlled traction rig.
4- Method The stud plate allows studs of differing shape and configuration to be tested and compared. Initial tests compared the traction forces of four different stud shapes (see Table 1). Two stud types were from the forefront of popular adidas football boots, the Copa Mundial (Stud D) and the World Cup (Stud C) boots. The remaining two (A and B) were unique designs manufactured at The University of Sheffield. Stud Ref Picture Vertical Length (mm2) Cross Sectional Area to motion (mm2) Table 1 - The four studs analysed for initial testing.
The traction rig simulates a forward front push off between jogging and going into a full sprint (Kirk et al. 2007). For this manoeuvre an athlete requires sufficiently high traction to avoid slipping. It was therefore decided to replicate a typical five stud confi-
The Engineering of Sport 7 - Vol. 1
609
guration found on the forefront of an adidas Copa Mundial football boot. The same stud configuration was used for each stud type. During testing a 350 N vertical load was applied by the hydraulic ram to replicate the weight of a player while in contact with a surface. This was found to be the representative normal force when a player is most at risk of loosing performance when carrying out a forefoot push off movement (Kirk 2008). All tests were carried out on a laboratory based third generation artificial surface, with a pile length of 50 mm and rubber particle in fill. Tests were repeated 10 times for each stud with the surface brushed and reconditioned for every test.
5- Results The data was analysed to give the horizontal force after a displacement of 10 mm. The comparative value for each stud was taken as the median horizontal force of the ten tests after a displacement of 10 mm (see Figure 4). A displacement of 10 mm was chosen as high speed video analysis showed that the surface ‘gives’ by approximately 10 mm during a movement when the player does not slip. It can be argued that a horizontal displacement greater than 10 mm would result in the perception of slipping (Kirk 2008).
Figure 4 - Box-plot showing distribution of data (range, upper and lower quartile and median) for the four studs after a 10 mm displacement.
6- Discussion and Conclusions The results reveal stud geometry will affect the surface traction given to a soccer player when performing a forward front sprint movement. Interestingly the stud with longest length, stud B, shows the lowest median traction force. Although shorter in length, Stud A may require a higher traction force than Stud B, despite their similar ellipse shape. This suggests the cross sectional area of football studs in the direction of motion is an important factor to traction force. However, it is possible Stud A is so shallow that the force plate was in contact with the surface and resisting motion, and needs to be investigated
610 The Engineering of Sport 7 - Vol. 1 in any further work. This may be useful to shoe manufacturers; the plate or sole of the shoe needs to be contact with the surface to obtain comparable values of traction. The same trend was not found between the two round studs. The longer stud, Stud C, gave a higher median traction value than stud D but has a lower cross sectional area. This is expected as stud C is designed for softer surfaces, and hence should offer more traction. The ellipse shape of Studs A and B clearly offer less resistance force to initial movement than the round studs C and D.
7- Further Work The initial testing described demonstrated the potential the SERG force controlled traction rig can have in the understanding of soccer boot traction during an athlete’s initial movement. However a number of potential developments have been identified during testing to improve the usefulness of the prototype rig in the future. Replacing the stud plate with an actual boot for example, would allow comparisons of existing and concept designs. Also, Kirk et al, describe the importance the angle of stud penetration into a surface has during football manoeuvres; changing the angle the studs on the rig penetrate the surface may replicate a more realistic movement (Kirk et al. 2007). All the tests carried out were laboratory based, it is hoped the rig will become portable to further investigate the relationship between surface conditions and stud design.
8- References [AL1] Andersen T.E, Larsen O., Tenga A., Engebretsen L. and R Bahr. Football incident analysis: a new video based method to describe injury mechanisms in professional football. In British Journal Of Sports Medicine, 37: 226-232, 2003. [CK1] Carre M.J., Kirk R.F. and Haake S.J. Developing Relevant Tests for Traction of Studded Footwear on Surfaces. In STARRS 2007, Proceedings of the Science, Technology and Research into Sport Surfaces Conference, Loughborough (Eds. P. Fleming, C. Young, S. Dixon, I. James, M. Carré and C. Walker), Loughborough University, UK, September 2007 [D1] Durá, J.V. The influence of friction on sports surfaces in turning movements. Proceedings of the International Assossiation for Sports Surface Sciences Technical Forum conference, 1999 [GM1] González J.C., Martínez A., Montero J., Alemany S. and Gámez J. Analysis of the Horizontal Forces in Soccer Boot Studs for Specific Movements, Proceedings of the 6th symposium of the ISB Technical Group on Footwear Biomechanics conference, July 2003 [K1] Kirk R.F. Traction of Association Football Shoes, PhD Thesis, The University of Sheffield, 2008 [KN1] Kirk R.F., Noble I.S.G., Mitchell T., Rolf C., Haake S.J. and Carré M.J. High-Speed Observations of Football Boot –
The Engineering of Sport 7 - Vol. 1
611
Surface Interactions of Players in Their Natural Environment. In Sports Engineering, 10(3): 129144, 2007 [S1] Shorten M.R. Sport Surfaces and Injury : The Missing Links. In STARRS 2007, Proceedings of the Science, Technology and Research into Sport Surfaces Conference, Loughborough (Eds. P. Fleming, C. Young, S. Dixon, I. James, M. Carré and C. Walker), Loughborough University, UK, September 2007
How to Build an Optimized Movement Analysis Laboratory for High Performance Athletes of Various Sport Disciplines (P116) Lars Janshen1
Topics: Constructing an optimized movement analysis laboratory. Abstract: Movement analysis plays an important role for both, enhancing athletes’ performance and movement techniques in various sport disciplines as well as optimising sport equipment. Therefore an optimal movement analysis laboratory should provide most realistic conditions for a variety of sport disciplines, such as track and field, gymnastics and team sports as well as for testing sport equipment ranging from sport shoes, floor constructions, landing mats, apparatus for gymnastics and throwing objects like balls, discs or javelins. The purpose of this paper is to demonstrate the functional and technical requirements for an optimized movement analysis laboratory. Special emphasis is given to the constructive properties and the spatial restrictions imposed by highly accurate measurement devices for both, motion capture and ground reaction forces without interfering with the performance of the athletes. As an example for a state-of-the-art movement analysis laboratory the new constructed lab of the Centre for Sport Science and Sport Medicine is presented. The requirement analysis included location factors, demands of actual and future research projects in sport science and medicine, physical education, sport associations and industrial partners. In addition important information for the planning process including key positions and a general timeline are presented. From these data, recommendations how to plan and build a movement analysis laboratory are derived. Keywords: biomechanics, movement analysis, laboratory, construction, project management.
1- Introduction Over the last decades the instrumented movement analysis became more and more common to investigate motion of either humans or animals. Typical scientific fields for movement analysis in humans are clinical biomechanics, sport biomechanics, neuromuscular research and motor control as well as psychological and behavioural studies. 1. Humboldt-University Berlin, Germany, Centre for Sport Science and Sport Medicine Berlin (CSSB), Institute for Sport Science - E-mail: [email protected]
614 The Engineering of Sport 7 - Vol. 1 The clinical movement analysis ranges from gait analysis (Devita et al. 2007, Simon, 2004; Perry, 1992), the analysis of motor control in stroke patients (Caimmi et al., 2008) or in persons with cerebral palsy (Desloovere et al., 2006; Romkes and Brunner, 2002) up to the evaluation of movement ability and movement quality in the elderly (Benedetti et al., 2007) or in obese persons (Nantel et al., 2006). In psychological studies typically research areas for movement analysis in humans are investigating the interaction between humans (Wong, and Rogers, 2007) and to their environment (Kersting et al., 2005). In sport science movement analysis is performed with similar perspectives, as listed above. It is very common to investigate the biomechanical aspects in particular the movement kinematics and kinetics together with motor control in an interdisciplinary approach (Biewener et al, 2007; Delay et al, 2007). There are two major goals using this approach. One is to improve the performance of highly competitive athletes by optimizing their movement technique (Dun et al., 2008; Hay, 1993). The second goal is to avoiding critical overloading of the biological structures at the same time (Bahr and Krosshaug, 2005). Further, optimised sport equipment may help to increase performance and to reduce critical loads on the human body. Thus it is important to improve sport equipment used by high performance athletes and in recreational sports. This includes technological research and development (R&D) for example to optimise sport shoes for the specific demands in different sports such as running (Nigg et al, 2003), tennis (Hreljac, 1998) or soccer (Gehring et al., 2007) as well as the improvement of landing mats in gymnastics (Arampatzis et al, 2003; Janshen, 2001), sport flooring for indoor sports (Streepay et al., 2000; Stiles et al., 2006) and outdoor sports (Ford et al., 2006). For both, the research in human performance and for the R&D of sport equipment a highly accurate quantitative measurement of human movement is necessary. These methods need to be scientifically sound and tested. Today this is achieved by using various measurement devices to analyse three-dimensional kinematics and kinetics combined with the neuromuscular control during human movement. Therefore a modern movement analysis laboratory has to enable the simultaneous and synchronised use of movement analysis systems to capture kinematics, to measure acting forces on the body and to evaluate motor control during the movement. Most sport movements and especially their critical events, such as the foot contact in long jump, are highly dynamic. These fast movements with short duration require high recording frequencies of the kinematic and kinetic measurement. In addition the measurement systems should minimally interfere with the movements. Kinematic movement analysis is mainly based on digital video and infrared camera systems. Digital video cameras range from common DV camcorders up to high-speed video systems operating at up to 10,000 Hz. The accuracy of the video analysis increases with lens quality camera resolution, operating frequency and the number of cameras used at the same time to capture the movement from different perspectives. In addition reflective markers skin mounted on body landmarks of the subjects are commonly used to increase the accuracy of the measurements. Especially the high-speed-video systems require a high and homogeneous luminosity in the facility used. In addition infrared
The Engineering of Sport 7 - Vol. 1
615
(IR) camera systems with measurement frequencies up to 2,000 Hz are used to analyse movement kinematics. The major advantage of these systems is the improved identification and automatic tracking of the reflective markers. In contrast to the video systems these IR-camera systems are quite sensitive to light and would be impaired even at small sun radiation. For both kinematic measurements special restrictions in the construction of the movement laboratory such as lighting, darkening and the colours used in the room have to be considered. The kinetic analysis of human movement is performed using various force transducers implemented in devices such as strength training machines, force plates, pressure distribution plates, as well as mounted in or on machines, apparatus and objects where the acting external or internal forces should be measured. The use of force plates demands for a detailed and accurate planning and construction of the floor. In any case vibrations of the force plates have to be avoided or eliminated. The analysis of the neuromuscular coordination during movements is commonly done by surface electromyographical measurements (EMG). The EMG electrodes placed on the skin over specific muscles transmit the electrical signals after pre-amplification to an AD-board on a computer. Today telemetric systems based on radio transmission or WLAN are used. In addition to the technical devices of the movement analysis, the design of a movement analysis laboratory should allow for a multifunctional use. This is especially a mayor factor for movement analysis in sport science. A number of different sports techniques should be executable in the laboratory under most realistic conditions to keep the influence of the laboratory environment as little as possible. Therefore the spatial requirements for the different sports have to be considered. In addition the climatic conditions in the laboratory have to be controllable not only for repeated measurements. Therefore constant as well as adjustable climatic conditions with sufficient fresh air ventilation are necessary even at varying outside temperature and humidity. To the authors knowledge no paper about the construction of an optimized biomechanical laboratory was available. The purpose of this paper is to demonstrate the functional and technical requirements for an optimized movement analysis laboratory. Special emphasis is given to the constructive properties and the spatial restrictions to integrate highly accurate measurement devices for both, motion capture and ground reaction forces without interfering the sport specific movement patterns of the athletes.
2- General issues for conception and planning of a movement analysis laboratory The adequate design of a new or converted movement analysis laboratory should be based on the research requirements of the one or more institutions that will use the facility. According to the actual and future research programs involving movement analysis, each participating group has special requirements to the laboratory. A list of items was developed to standardise the different demands the facility may face. This list included the spatial requirements for the movements (e.g., run-up length, movement pathways, preferred flooring, room height), the measurement requirements (e. g. devices for kine-
616 The Engineering of Sport 7 - Vol. 1 matics, kinetics, EMG, X-ray and medical devices), the technical requirements (e. g. power supply, data networks, ventilation, temperature and humidity control) and special safety regulations. A mathematical model of the spatial requirements of the laboratory was developed based on various pathways of the movements that should be analysed (figure 1a). For most sport movements, the spatial requirements were taken from the regulation of the respective disciplines or from recommendations for sports facilities. The most important areas for the movement were determined relative to the respective overall movement pathway. From this, essential spots in the laboratory were defined using an optimisation technique by overlaying all movement pathways. One mayor spot for the movement analysis was defined. The positions of three additional subordinated spots (sub-spots) were described relative to this spot. Again, the distances between the spots were based on the movements. This also included the positioning of sport apparatus, mounted on the floor, the walls or the ceiling. At all spots a multiple instrumented movement analysis using various measurement devices should be enabled. Therefore additional technical requirements such as the minimum and maximum distance of cameras to the movement planes, the lighting at the spot and special demands for the floor construction to implement force plates had to be considered. In the second step the technical profile of the laboratory was developed (figure 1b). This included the types of power supplies (e.g., non-medical, medical standards with and without uninterrupted power supply (UPS)), the position of power and data sockets relative to the spots for the analysis, the type and positions of the lighting and heating system and the layout of the air-conditioning system. To ensure highly accurate measurements of the ground reaction forces a detailed concept to construct the floor of the laboratory was developed. This included the grounding, the floor stiffness and the material of the floor surface. Finally the colours used for the floor and the walls needed to provide maximum contrast for marker based as well as for markerless kinematic analysis.
Figure 1a - Model of the requirements and interactions to calculate the positions of the major and subordinated spots for the kinetic and kinematic movement analysis systems
Figure 1b - Schematic model of the technical requirements and their interactions that are important for the essential spots for the movement analysis.
The Engineering of Sport 7 - Vol. 1
617
In the third step key positions within the time of the planning process were defined. Based on the estimated duration of the individual construction phases a general timetable was developed. This timeline is used to control the individual steps during the construction process.
3- Realisation of constructing a movement analysis laboratory The definition of research areas revealed four major interests of the participating institutions. These were (1) the analysis of elite sports and high performance athletes, (2) the education of young researchers in sport science and physical education as well as in clinical sciences, (3) the clinical movement analysis especially for orthopaedic patients and (4) the research and development together with external partners form the sport equipment industry and the clinical implant industry. Typical sports that may be analysed in the laboratory are displayed in table 1. Based on the mathematical optimization of the movement pathways an optimal room size of 30 m in length, 15 m in width and 6 m in height is recommended. These measures result form typical movements, such as the run-up lengths in gymnastics horse jumping (max. 25 m (FIG 2001)), the last five to seven steps in jewelling (10 to 15 m) (Tidow, 1995), in high-jump (15 x 15 m), (IAAF, 2008) and from the estimated spatial requirements of typical cutting movements in soccer, handball and basketball (about 10 x 10 m). The room height mainly results from the demands in gymnastics using uneven bars and high bar (FIG, 2001). Table 1 - Movements and types of equipment to be analysed in the laboratory.
The laboratory in the present study was implemented in a renovated building. The optimal room with a height of 5.5 m was on ground floor and had no basement. To realise a room size as shown in figure 2 some internal walls were removed. As described in the methods, one major and three subordinated spots for the movement analysis were defined. The position of the major spots within the room resulted from the calculated overlay of the pathways and the most important areas of the different sport movements. At all spots bays are designed to mount force plates of different sizes in various positions (figure 2). The major spot consists of two platform bays (A and B). Their transversal position in the room is based on the positioning of the motion analysis cameras related to the longitudinal middle axis of the bays. The maximum distance is about 4 m if
618 The Engineering of Sport 7 - Vol. 1 mounted on the nearest parallel wall an about 7.5 m if mounted on the opposite wall. In addition the main internal door to the laboratory (ID1) is aligned to the longitudinal axes of bay A, B and C. Thus it is possible to extend the run-up length of about 6 m to the position of the force plates. The longitudinal position in the room of bay A is determined by a 6 m distance to the end of the room, given by the requirements of the landing area for horse jumps in gymnastics. Bay B is placed under a load-bearing steel girder. This provides both, an extended maximum distance between force plates that need to be mounted in line with bay A and the mounting of hanging constructions (e.g., gymnastics rings) directly over the force plates. The longitudinal position of bay C is aligned to the middle axis of the second internal door (ID2) to analyse cutting movements as they occur in ball games and team sports. The door is connecting the laboratory with a second one and could also be used to extend run-up lengths.
Figure 2 - Floor plan of the movement analysis laboratory including the major pathways for run-ups and gait analysis, the positions of force plate bays (A to E), different nets mounted on the wall and possible positions of the cameras systems. X indicate positions of ceiling mounted weight bearings, dots indicate positions of floor anchor points for apparatus for gymnastics. The security beam is mounted on the ceiling.
Two additional bays (D and E) are positioned parallel to the axis of bay A, B and C and in about 4 m distance to the nearest parallel wall. The opposite wall again is about 7.5 m away. One end of bay D is oriented to the middle of the longitudinal distance of the laboratory. Bay E is positioned in line with the longitudinal axes of bay D. The distance between bay D and E is defined by the alignment of bay E to the middle axis of the exterior door of the laboratory. This door enables the analysis of movements like sprint starts or throwing movements were a second wall would interfere with the natural movement pattern of the athlete. For all platform bays a layer of 0.5 m made from high density concrete (7 g mm-3) is used as a fundament. From this a fundament mass of about 5 tons (bay C, E), 10 tons
The Engineering of Sport 7 - Vol. 1
619
(bay B, D) and 16 tons (bay A). These masses are directly attached to solid ground by 1.0 m long poles (diameter = 0.2 m) from statically reinforced concrete. This massive ground work is necessary to eliminate vibrations induced by impacts to the mounted force plates. The fundament is covered by a levelling layer of a combined epoxy-cement material. On top of this layer a 20 mm strong steel plate is adhered. The whole plinth block of each platform bay is fully detached from the rest of the floor fundament build of 0.15 m of statically reinforced concrete. The platform bays are covered with wooden filling pieces in different sizes to allow for various mounting positions of the force plates. Both the fillings and the force plates were screwed to tapped holes in the steel plate. The floor surrounding the platform bay is covered by a wooden area-elastic sport floor with a linoleum surface to account for the demands of the sport movements. An orange colour of the surface provides sufficient optical contrast for black and white. This is necessary to work with both, marker based video systems based on white reflecting markers and markerless tracking systems that only work with dark clothing of the subjects. Around all platform bays circular husks are inserted into the floor fundament to hold steel pipes of 0.1 m in diameter. Theses pipes function as optional mounting points for the motion capture camera systems. The husks are physically detached from the elastic sport floor. This ensures that no vibrations from the sport floor are transferred to the cameras. In addition to the husks anchor points for the gymnastics apparatus are attached to the floor foundation. The position of the anchor points for the apparatus are related to bay A and D and allows for a kinetic analysis of landing movements measured by the force plates. The cameras for motion capturing are mounted on the walls. They are attached to a specially designed frame system of aluminium pipes. A vertical pipe is attached to an upper and lower horizontal pipe. The vertical pipe carries a 2.0 m long pivoting lever with the camera fixed at its end. This system enables a continuously horizontal, vertical and transversal positioning of the cameras in the room. Furthermore, wall mounted anchor points for the volleyball net are related to bay A and D to enable the kinetic analysis of athletes’ offensive or defensive movements near to the net. In addition to the frame system on the wall four additional fixings for cameras are mounted on the ceiling. Again the positions are defined relative to the major spot (bay A and B) to enable optimised view angles of the cameras, especially for markerless tracking. According to the floor the wall colours must provide a high contrast to dark colours but also allow for a sufficient contrast to white markers as used in the high-speed video analysis. This is best achieved with a yellow colour tending towards a light orange. To minimise objects and surfaces that may distract especially the automatic tracking of the marker based video analysis for motion capture, no reflecting materials or striated surfaces are allowed at the floor or walls. Therefore radiator plates were mounted under the ceiling instead of attaching them to the walls. The overall layout of the ceiling is based on the technical requirements for the heating, cooling and lighting system. Special attention is given to the area above the gait track (figure 2). In order to analyse orthopaedic patients or subjects with balance problems special safety features to secure the subjects are necessary. For this reason a metal beam was mounted exactly above the longitudinal
620 The Engineering of Sport 7 - Vol. 1 axis of bay D and E. A safety harness is attached to a carriage running on the beam. The construction is able to carry loads of up to 5,000 N which is about six times bodyweight of a 85 kg person. The metal beam must not interfere with the other technical systems on the ceiling, especially the lighting. Depending on the used luminaires the positioning of the lights are calculated to generate a sufficient and homogeneous luminosity in the laboratory. The luminosity is set to 1000 lx in a height of 1.1 m above the floor. To avoid interferences between the frequency of the luminaires (e.g., when using high intensity discharge (HID) headlamps) and the frame rate of high-speed-video cameras, incandescent lamps are used. Additional spotlights are installed to increase the luminosity for side view video. The spotlights are mounted on continuously height adjustable pantographs that are attached to drays running on horizontal aluminium rails close to the ceiling. The power supply in the laboratory consists of three systems. An uninterrupted power supply (UPS) network with and without medical standard and a power supply network without UPS. Together with the data network all cables are integrated in a wall mounted cable duct. The locations of the different sockets (power supply systems and data network) again is related to the movement analysis spots. In addition the maximum distances between sockets of the same system do not exceed 2.0 m. Furthermore, additional floor tanks with data and power supply sockets are implemented a eight positions in the floor.
4- Technical recommendation to concept, plan and realise a movement analysis laboratory As described above, it is important to involve all participating institutions that are using the movement analysis laboratory. From this, the used measurement systems should be defined. Commonly these are kinematic (camera based) and kinetic (force platform based) systems that are combined with electromyography, radiological systems and additional analogue measurement devices. As shown in this paper, the pathways of the movements that are going to be analysed and the measurements systems used should define the spatial and technical requirements of the laboratory. To allow for highly accurate measurements and a high functionality of the laboratory special constructional conditions are necessary. Close collaboration of researchers, architects and engineers is necessary to transfer the needs of the researchers into technical requirements to construct or renovate the building. To ensure an optimal communication between all technical groups there has to be an understanding of different technical terminology. In an ideal case, one person with strong backgrounds of scientific movement analysis research, engineering and housing technology is coordinating the planning and realisation process of the laboratory. The most important phase to build a new or renovated movement analysis laboratory is the preliminary planning phase. Here any general decisions regarding the laboratory such as location, included research fields, size, analysis capabilities, etc. are determined. In the following construction documentation phase the main constructive demands such as ground work, fundaments and superstructure work, used materials, type of floor have to be determined. In addition the housing technology in power supply
The Engineering of Sport 7 - Vol. 1
621
systems, lighting, heating, cooling and ventilation up to data networks etc. are defined. It is very important to precisely define all requirements to the laboratory.. Since the plans are the legal documents for all companies involved in the construction the construction work itself should not begin until the pre-planning is fully completed. Presuming a detailed and comprehensive planning phase the construction of an interdisciplinary use of a movement analysis laboratory may bring high benefits to all participating institutions. This may include the shared use of the in many cases highly specialised measurement equipment. Therefore it is possible for each of the institutions to save money. In addition the costs to run the laboratory can be shared.
5- References Arampatzis A., Morey-Klapsing G. and Brüggemann G.P. The effect of falling height on muscle activity and foot motion during landings. In Journal of Electromyography & Kinesiology, 13(6): 533-544, 2003. Bahr R., and Krosshaug T. Understanding injury mechanisms: a key component of preventing injuries in sport. In British Journal of Sports Medicine, 39(6): 324-329, 2005. Benedetti M. G., Berti L., Maselli S., Mariani G., and Giannini S. How do the elderly negotiate a step? a biomechanical assessment. In Clinical Biomechanics, 22(5): 567-573, 2007. Biewener A.A., and Daley M. A. Unsteady locomotion: integrating muscle function with whole body dynamics and neuromuscular control. In Journal of Experimental Biology, 210(17): 29492960, 2007. Caimmi M., Carda S., Giovanzana C., Maini E.S., Sabatini A.M., Smania N. and Molteni F. Using kinematic analysis to evaluate constraint-induced movement therapy in chronic stroke patients. In Neurorehabilitation and Neural Repair, 22(1): 31-39, 2008. Desloovere K., Molenaers G., Feys H., Huenaerts C., Callewaert B. and de Walle P.V. Do dynamic and static clinical measurements correlate with gait analysis parameters in children with cerebral palsy? In Gait and Posture, 24(3): 302-313, 2006. Daley M.A., Felix G. and Biewener A.A. Running stability is enhanced by a proximo-distal gradient in joint neuromechanical control. In Journal of Experimental Biology, 210(3): 383-394, 2007. DeVita P., Helseth J. and Hortobagyi T. Muscles do more positive than negative work in human locomotion. In Journal of Experimental Biology, 210(19): 3361-3373, 2007. Dun S., Kingsley D., Fleisig G.S., Loftice J. and Andrews J.R. Biomechanical comparison of the fastball from wind-up and the fastball from stretch in professional baseball pitchers. In American Journal of Sports Medicine, 31(1): 137-141, 2008. FIG - Federation International de Gymnastique. Apparatus Norms. Moutier, Switzerland, 2001. Ford K.R., Manson N.A., Evans B.J., Myer G.D., Gwin R.C., Heidt R.S. and Hewitt T.E. Comparison of in-shoe foot loading patterns on natural grass and synthetic turf. In Journal of Science and Medicine in Sport, 9(6): 433-440, 2006. Gehring D., Rott F., Stapelfeldt B. and Gollhofer, A. Effect of soccer shoe cleats on knee joint loads. In International Journal of Sports Medicine, 28(12): 1030-1034, 2007. Hay J.G. The biomechanics of sports techniques. Prentice-Hall, 1993. Hreljac A. Individual effects on biomechanical variables during landing in tennis shoes with varying midsole density. In Journal of Sports Sciences, 16(6): 531-537, 1998. IAAF – International Association of Athletics Federations. Competition Rules. Monaco, 2008.
622 The Engineering of Sport 7 - Vol. 1 Janshen L. Neuromuscular control during gymnastic landings II. In Hong Y. and Johns D.P. (Ed.) Proceedings of XVIII International Symposium on Biomechanics in Sports: 154-157, Hong Kong – China, 2000. Kersting U.G., Janshen L., Bohm H., Morey-Klapsing G.M. and Bruggemann G.P. Modulation of mechanical and muscular load by footwear during catering. In Ergonomics, 48(4): 380-398, 2005. Romkes J. and Brunner R. Comparison of a dynamic and a hinged ankle-foot orthosis by gait analysis in patients with hemiplegic cerebral palsy. In Gait and Posture, 15(1): 18-24, 2002. Nantel J., Brochu M. and Prince F. Locomotor strategies in obese and non-obese children. In Obesity (Silver Spring, Md.) 14(10): 1789-1794, 2006. Nigg B.M., Stefanyshyn D., Cole G., Stergiou P. and Miller J. The effect of material characteristics of shoe soles on muscle activation and energy aspects during running. In Journal of Biomechanics, 36(4): 569-75, 2003. Simon S.R. Quantification of human motion: gait analysis-benefits and limitations to its application to clinical problems, In Journal of Biomechanics, 37(12): 1869-1880, 2004. Streepey J.W., Gross M.M., Martin B.J., Sudarsan S. and Schiller C.M. Floor composition affects performance and muscle fatigue following a basketball task. In Journal of Applied Biomechanics, 16(2): 157-168, 2000. Stiles V.H. and Dixon S.J. The influence of different playing surfaces on the biomechanics of a tennis running forehand foot plant. In Journal of Applied Biomechanics, 22(1): 14-24, 2006. Tidow, G. Model technique analysis sheet - The Javelin throw. New Studies in Athletics 11(1): 4562, 1996 Wong W. and Rogers E.S. Recognition of temporal patterns: from engineering to psychology and back again. In Journal of Experimental Psychology, 61(2): 159-167, 2007.
Analysis of the Wobble of a Spinning Disc at Launch (P117) William Rae1, Mont Hubbard2
Abstract: Field observations (Hubbard and Hummel, 2000) and numerical simulations (Crowther and Potts,2007) of spinning discs in flight often show a wobble of the disc immediately after launch. This component of the motion causes an undesirable drag penalty, and its source is not well understood, occurring for a wide range of initial conditions. The analysis presented in this paper succeeds in identifying, for a flat disc, the frequencies and amplitudes of the wobble and their connection to the launch conditions. The approach used to make this identification is to express the equations of motion in a two-Euler-angle reference frame, leading for small wind angles and Euler angles to a pair of eigensolutions containing the proper coupling of aerodynamic, gyroscopic, and Coriolis effects. This analysis of the disc motion shows two components; one member of the solution pair is strongly damped, while the other is a slightly unstable oscillation. The amplitude information in the solution shows the connection, in the small-angle limit, between the launch conditions and the initial wobble, and may be useful in minimizing the latter. The sections that follow contain descriptions of the coordinate frames and method of analysis used, and the conclusions reached. Key Words: Spinning Discs, Flight Dynamics.
1- Coordinate Frames The problem considered here involves flight of a disc which is spinning about its axis of symmetry and moving through a constant-property atmosphere at rest above a flat nonrotating earth. The flight path followed by the center of mass is described in a coordinate system xF, yF, zF fixed to the earth. The xF coordinate lies in the earth plane; the yF and zF coordinates form a right-handed set, with zF positive down (parallel to the gravity vector). The initial velocity vector lies in the xF, zF plane. Prior to launch, a set of axes xB, yB, zB fixed to the disc is initially aligned with the earth-fixed axes. The orientation of the central axis of the disc relative to the earth-fixed set immediately after launch is defined by two Euler angles here called pitch ( ) and roll (ø). This sequence begins with 1. University at Buffalo, State University of New York, Buffalo, New York, USA 14260 - E-mail: [email protected] 2. University of California at Davis. Davis, California, USA 95616 - E-mail: [email protected]
624 The Engineering of Sport 7 - Vol. 1 a rotation through the pitch angle about yF the axis. A second rotation is then made, about the xB axis, by the roll angle ø. A final rotation can then made, by the yaw angle about the zB axis. The transformation matrix that calculates vector components in the space-fixed system from those in the body-fixed system is (shown here for the case of a position vector)
(1)
The spin angle itself plays no role in the dynamics of the problem, although the spin angular velocity does. To take this into account, it is useful to define a set of coordinates which do not spin with the disc: (2)
This is equivalent to omitting the third Euler-angle rotation and resolving all vector quantities in coordinates that do not spin (Fig 1).
Figure 1 - Unit vectors in non-spinning coordinates.
With this definition, the unit vectors defining the disc orientation are (which lies in the plane formed by the axis and the gravity direction and is perpendicular to the axis) and , which in turn is perpendicular to and . Vector components in the non-spinning axes are defined by (3)
The Engineering of Sport 7 - Vol. 1
625
Then the inertial velocities are related to the velocity components in body axes by (4) (5)
with time rates of change of the Euler angles related to the three components of the angular velocity expressed in the ~ frame by Following the work of Hubbard and Cheng (2008), a final rotation through the angle is made about the axis (Fig 2) so as to align the body axes with the wind plane formed by the velocity vector and the symmetry axis. Unit vectors in this frame (denoted by the subscript 3) are directed along , , and . In this frame there is no sideslip component of velocity; instead one uses and the velocity magnitude to balance the centrifugal force perpendicular to the wind plane. The aerodynamic loads are those measured at a wind-tunnel pitch angle of attack .
Figure 2 - Wind-Plane Coordinates.
2- Equations of Motion For purposes of the present paper, the full six degree of freedom equations of motion written in the wind-plane coordinates (which are principal axes for the disc) are approximated by assuming that all wind angles and Euler angles are small, that the speed is constant, and that the aerodynamic loads are a simplified version of those of a flat plate
626 The Engineering of Sport 7 - Vol. 1 where the drag, lift, and pitching-moment coefficients are the constant and linear approximations These components of the aerodynamic force and moment dependence on angle of attack were estimated using parameter identification by matching predicted and experimental trajectories of markers on a Frisbee (Hubbard and Hummel, 2000). Linearized equations for the two velocities u3 and w3 and the three components of angular velocity p3, q3, and r then become (6)
where the axial and transverse moments of inertia are related by IZ = 2IT and where Q = V2/2 Taking the Laplace transform of equations (6) gives
(7)
The equations for the tightly coupled states w3, p3, and q3 can be solved for the transforms of the state-vector components, each of which has the form N(s)/D(s). The denominator D(s) is a cubic involving only the dimensionless normal force and pitching moment and the scaled spin rate A, B and R, respectively (8) and the numerators are linear combinations of the initial conditions and the gravity vector. With a subscript denoting the state variable to which they apply, they are:
(9)
Denoting the roots of the denominator by D(s) = s(s–s1) (s–s2) (s–s3), the method of partial fractions gives (herede notes any of the state variables) where (Kreyszig, 1983)
.
The Engineering of Sport 7 - Vol. 1
627
The inverse is (10) For the range of the parameters A and B of interest here, the zeroth root is zero, s1 is real and negative, and s2 and s3 are a complex-conjugate pair with a small positive real part .Thus the aerodynamic effects modify the wobble frequency slightly from its value in a vacuum of twice the spin (2r) and make it unstable by contributing a small positive real part to the root. The real root is closely approximated by s1 = – A/(1–B) and Figure 3 shows the variation of the complex-conjugate pair for reasonably small values of A and B.
Figure 3 - Locus of complex-conjugate roots (Only the positive imaginary par is shown).
Numerical evaluation of these formulas shows satisfactory agreement with the full six degree of freedom calculation. Figure 4 shows the comparison for the w3 velocity component. It is clear that the solution has an early transient which decays in about a half second plus an oscillatory component that grows gradually with time. This behavior is similar to that of an American football (Rae, 2003). The parameters used for all examples are
Figure 4 - Comparison of Small-Angle and 6DOF Solutions.
628 The Engineering of Sport 7 - Vol. 1
3- Implications for Launch Conditions Hubbard and Cheng (2007) chose “wobble free” initial conditions in the discus throw that had minimal, but nevertheless nonzero, wobble by choosing the accelerations . Equation (10) shows however, that merely choosing initial conditions so that the initial derivative of the angle of attack is zero is not sufficient to suppress wobble. Instead the requirement must be that the wobble amplitude must be zero. The response functions N(s) suggest some possibilities for altering the wobble response by proper choices of the launch conditions w3(0), p3(0), q3(0) since the oscillatory portion of the Paper No. 117 motion is confined to the terms associated with the complex-conjugate pair of roots. One baseline case is that in which all three initial conditions are zero. Figure 5 shows two 6DOF calculations with p3(0) =q3(0) = 0 and w3(0) equal to zero and + 5 degrees (the latter is close to the value at which the lift balances the weight). The w3 response is fairly smooth for w3(0)=0, but the pitch responses show that the gravity term induces significant oscillations, which increase when lift is used to cancel the weight.
Figure 5 - Effect of Initial Angle of Attack.
The fact is that all four parameters (g and the initial values of w3(0), g, p3(0) and q3(0)) have a significant influence on the oscillatory motion. Further exploration of the parameter space of launch conditions is needed, using the small-angle expressions as a guide to six degree of freedom calculations. Part of that exploration includes extension to the aerodynamic properties of more complex disc shapes.
The Engineering of Sport 7 - Vol. 1
629
4- References Crowther, W.J. and J. R. Potts, Simulation of a spin stabilized sports disc, Sports Engineering, vol. 10, pp. 3-21, 2007. M. Hubbard and K. B. Cheng,“Optimal Discus Trajectories”, Journal of Biomechanics, vol. 40, pp. 3650-3659, (2007) M. Hubbard and S. A. Hummel, Simulation of Frisbee flight, In Proceedings of 5th Conference on Mathematics and Computers in Sport, G. Cohen, Ed., University of Technology, Sidney, New South Wales, Australia, 14-16 June, 2000. Kreyszig, E. Advanced Engineering Mathematics, Fifth Edition, John Wiley and Sons, New York, 1983. W. J. Rae, Flight dynamics of an American football in a forward pass, Sports Engineering, vol. 6(3), pp. 149-164, 2003.
A Study of the Influence of the Environmental Condition and the Garment in Skin Temperature in Sport Activity (P119) Natividad Martínez, David Rosa, Javier Gámez, Juan Carlos González, Carlos Chirivella, José María Gutiérrez, Jaime Prat, José Javier Sánchez1
Abstract: Choosing the adequate garment for sport practice in adverse weather condition, either cold or hot, is an aspect of great influence on activity performance. This paper presents the results of a study carried out in 2005 by the Institute of Biomechanics of Valencia (IBV) together with the Physical and Sports Performance Research Unit at the University of Valencia (UIRFIDE). The final goal of the project was generating the knowledge to provide enough information to select the most adequate garment for the sport practice in each situation. This study was aimed at comparing the thermal response of the body during a nonuniform activity test performed on a tread-mill under different conditions (provided by clothing and environmental conditions). In this sense, two commercial shirts made of the same textile (100% polyester) but with different thickness (different thermal properties: Rct and Ret) were tested by a sample of 8 trained-participants under controlled environmental conditions (25ºC/50%RH, 10ºC/60%RH) in laboratory. Throughout the activity test, physiological parameters of the thermal response such as skin temperature (at three locations: chest, arm and thigh) and microclimate variables in user-garment interface (armpit and upper-back) as well as individual work load indicators (heart rate) were registered. Simultaneously, the user perception was also collected at different times of the test. The results allowed measuring a significant influence (p < 0.05) of the environmental condition independently of the garment and the activity level on the average skin temperature. The influence of the shirt was only described as significant (p < 0.05) at low- middle intensity of the activity for each environmental condition. These results provide the possibility for sport equipment manufactures of giving recommendations for users which let them choosing the most suitable garment according to the environmental conditions and activity level in order to improve the performance and comfort providing the body with the adequate thermal conditions. Keywords: thermoregulation, thermal comfort, garment selection.
1. Instituto de Biomecánica de Valencia, Universidad Politécnica de Valencia- Edificio 9C, Camino de Vera s/n, E-46022 Valencia, Spain - E-mail: natividad.martinez,david.rosa, javier.gamez, juancarlos.gonzalez, carlos.chirivella, josemaria.gutierrez, jaime.prat, [email protected]
632 The Engineering of Sport 7 - Vol. 1
1- Introduction The textile industry’s current development has flooded various sectors (such as fashion, sports clothing, protection equipment, etc.) with a wide variety of technical textiles with sufficiently diversified properties to be capable of satisfying the user’s needs in each situation. This development in textiles has not, however, been accompanied by a development in accordance both with design criteria and methods for evaluating their performance. Lack of knowledge of the effect that the textile material has in the userproduct interaction makes it enormously difficult to adapt products to the user, thereby limiting the obtaining of added value that distinguishes the end product from other similar products in the sector in aspects such as user satisfaction, performance, safety and health. One of the first and most important applications for this textile development was solely aimed at obtaining textiles that adapt to the characteristics and thermal necessities of the human body and the environment so as to provide suitable safety body operation conditions and thermal comfort in each situation of use. Although the body has different actuation systems to adapt itself to the thermal conditions, there are limits in its thermoregulation capability, so risks due to inappropriate temperature can appear (Kurz 1994). Two different approaches exist for study the thermal interaction between body and human products: the safety and health level and the comfort level. The first approach is from the point of view of health. Keeping physiological variables in a safety range includes ensuring no health problems such as dehydration, stiffness, heat stroke or frostbites appear. The second approach supposes a narrower range for thermophysiological values. Thermal comfort has been defined as the state of mind that expresses satisfaction with the thermal environment (Ashrae Fundamentals 2001). Fanger was the first at defining thermal comfort conditions (Fanger 1970): • Thermal balance of the body. • Perspiration level is into a determinated range (body can be in thermal balance but uncomfortable due to excessive perspiration). • Skin temperature is in comfort range (body can be in thermal balance but uncomfortable due to excessive cold or heat sensation in the surface). Thermal comfort is depending (in a complex way), on many interrelated factors such as environment, individual characteristics, physical activity and obviously garment. Among these factors, garment is the one user can act on and it should be choose or design as a function of the rest. A correct selection will make lighter the work of the thermoregulation body systems and will contribute to avoid thermal risks and even to enhance activity performance. It is known that product should present a specific range of temperature and humidity between body and product depending on the situation of use. The adaptive and dynamic nature of the human thermoregulatory response means that studying the interaction between textile materials and the user is complex and there is a shortage of knowledge in this area that makes the development of new materials and
The Engineering of Sport 7 - Vol. 1
633
products difficult and more costly without ensuring their success in the market. In the study of thermal comfort knowledge about anatomy and physiology, particularly about thermoregulation and thermal body receptors should be involved in every moment (Zhu and Baker 1998). This creates an important barrier for the development and transfer of these materials as a drive for innovation in multiple industrial sectors. In order to advance in the design of new products based on these textile materials it is necessary to generate criteria for both design and selection as methods for evaluating their performance in situations in which they are used or through methodologies that simulate service conditions as adequately as possible. The aim of this work is to collect valuable information by means of measuring the user thermal response about textile materials participating in the study for giving recommendations of use according user comfort. The hypothesis driving the study is that for each situation of use defined by the environmental conditions and activity level, the suitability or not of the garment at providing the body with the adequate thermal conditions may have effect on user comfort and even on performance.
2- Methodology Two commercial shirts were tested by the sample of users (see description in next paragraph) in the environmental chamber. Both commercial shirts were made of the same textile (100% polyester) but with different thickness and consequently different thermal properties. Both were thermally characterized (thermal resistance Rct and evaporation resistance Ret) according to Skin Model test described in UNE-EN 31092 standard. Thermal properties are shown in the next table: Table 1 - Thermal properties of the shirts according UNE-EN 31092 standard.
Shirt 1 Low resistance 2 High resistance
Thermal resistance Rct (m2•K/W)
Evaporation resistance Ret (m2•Pa/W)
0.0342±0.0002 0.0512±0.0005
3.56±0.03 4.89±0.01
Eight trained male subjects (average age: 25.25 years (Std dev:1.38); average height: 1.76m (Std dev: 0.068); average weight: 72.45 kg (Std dev: 0.67)) performed a test on a treadmill which consisted of six phases at different activity level in two environmental conditions (25ºC/50% RH and 10ºC/60% RH) performed in an environmental chamber. The activity test designed lasted for 60 minutes and included different phases of activity and rest in order to cover the wide range of activity possibilities for the use of the shirts. The test is described below: 1.Acclimatization (10 minutes). This step includes a first phase of walking at a speed of 2-3 Mph during 5 minutes followed by a second phase of jogging at 5 Mph. 2.Rest (5 minutes). The subject rested but he was not allowed to seat at any moment. 3.Low intensity activity (15 minutes). This step includes three phases of running at different speeds (5 minutes of duration each): 6, 7 and 8 Mph.
634 The Engineering of Sport 7 - Vol. 1 4.Rest (5 minutes). The subject rested but he was not allowed to seat at any moment. 5.High intensity activity (15 minutes). This step includes three phases of running at different speeds (5 minutes of duration each): 8, 9 and 10 Mph. 6.Rest (10 minutes). The subject rested but he was not allowed to seat at any moment. The methodology proposed consists of measuring different kinds of variables regarded to the thermal state throughout the test described above. Objective measurements were done as thermal comfort variables (skin temperature and microclimate variables) or physiological (heart rate) together with subjective variables (thermal perception and comfort degree).
2.1 Thermal comfort variables. Skin temperature at three different locations (chest, arm and thigh) and microclimate variables in subject-garment interface (at the armpit and upper-backs) were measured each 2 seconds during the whole test by means of digital sensors (developed by the Institute of Biomechanics of Valencia). Figure 1 shows the exact locations of the sensors.
Figure 1 - Sensors location.
After checking the existence of a pattern in chest, arm and thigh temperature registers, average skin temperature has been calculated according to equation (Daannen 1997). (1) From the curve of average skin temperature, 13 variables were obtained by parameterization of the calculated registered. The selected values corresponded with the phase changes of the activity test.
2.2 Physiological variables Heart rate was measured simultaneously using a POLAR™ heart rate monitor. The heart rate was recorded every 5 seconds during the whole activity test and parameterized as the thermal variables.
The Engineering of Sport 7 - Vol. 1
635
Every of the registered variables (cited in 2.1. and 2.2.) were parameterized according the phase changes. Afterwards, a statistical descriptive analysis of average skin temperature, humidity at armpit and lower-back and heart rate has been done and the effect of the environmental condition and the shirt was analyzed by means of ANOVA (software SPSS 14.0).
2.3 Thermal perception The aim of the subjective test was to gather on the thermal perception of users and about the sensations of general comfort or the incidence of discomfort associated with a certain thermal state. In this way, the subject’s opinion was recorded at the same time as the temperature and humidity data during the different stages of the study. The survey was divided into a preliminary survey and another while the activity test was being carried out. Different questions were asked before and after the subject did the test, to assess their tendencies in relation to the thermal perception of a situation. Subjects were asked during the test about humidity and temperature perception on the body as a whole or by different zones. Global comfort degree was also registered in each case. The survey was based in a 5-points scale as it follows: • For temperature perception: Very cold – Cold -Neither warm nor cold -WarmVery warm. • For humidity perception: Very dry – Dry -Neither dry nor wet - Wet -Very wet. • For comfort degree (global thermal state assessment): Extremely uncomfortableVery uncomfortable-Uncomfortable-Slightly uncomfortable-Comfortable.
3- Results 3.1 Average skin temperature Figure 2 shows the evolution of the parameters obtained from the average temperature of the skin (see equation 1) for all the subjects at each condition for both shirts in cold and heat conditions (each point represents the average value for all the subjects).
Figure 2 - Evolution of the average skin temperature in each case of the study.
636 The Engineering of Sport 7 - Vol. 1 Analysis of the variance (ANOVA) results showed that the shirt with higher thermal resistance produces higher average skin temperature significantly (p<0.05) during the test with the exception for high level of activity (end of the test) independently from the environmental conditions. Figure 2 only shows average values for the whole sample of subjects in order to show clearly that considering each condition separately, shirts are ordered according their thermal resistances.
3.2 Armpit humidity Figure 3 shows the evolution of the parameters obtained from the armpit humidity registers for all the subjects at each condition for both shirts in cold and heat conditions (each point represents the average value for all the subjects).
Figure 3 - Evolution of the armpit humidity in each case of the study.
In this figure 3, only average values for the whole sample of subjects have been drawn in order to show clearly that considering each condition separately, shirts are ordered according their evaporation resistances. However, the figure shows that when the perspiration rate of the user increases, the shirt with lower Ret seems to be capable of providing lower humidity levels at 25ºC than the shirt with high Ret at 10ºC: it seems that choosing the adequate shirt can compensate the effect of the environmental condition in armpit humidity level. Nevertheless, this effect could not be contrast by ANOVA. According to this finding, ANOVA analysis was not able to find any significant effect due to the shirt at any moment of the test and similarly to what was happening in the skin temperature, environmental condition is only significant at low-middle activity levels.
3.3 Heart rate Regarding the heart rate and after the parameterization of the curves, a descriptive analysis was done for estimating the average heart rate values for each test configuration. After the descriptive and having checked that the ANOVA did not show significant diffe-
The Engineering of Sport 7 - Vol. 1
637
rences for any heart rate parameter due to the shirt or the environmental conditions, individual differences were considered. The data shows a great variability in heart rate between subjects. It has been discussed that in our case this variability can hide any possible difference due to the shirt. However, differences due to the environmental conditions can be detected but not in a significant way. ANOVA analysis was not able to find any significant effect due to the shirt at any moment of the test and similarly to what was happening in the skin temperature, environmental condition is only significant at low-middle activity levels.
3.4 Thermal perception Data representation for ‘Perception of temperature’ and ‘Thermal comfort degree’ was carried out by means of histograms done separately for each test condition (environmental condition and shirt) and for the survey at the end of each activity step. ‘Perception of temperature’ was systematically found ‘Warmer’ at 25ºC than at 10ºC regardless of the shirt and the instant of the test. In the next figures a histogram for the Perception of temperature and another one for the associated ‘Comfort degree’ are presented. The relationships between both allowed to know which sensation was pleasant for the user. Furthermore, adding the objective thermal comfort variables is possible to decide which level of thermophysiological variables as skin temperature are found satisfactory and give comfort sensation to the user (see perception scale in Section 2. Methodology).
Figure 4 - Histograms of the thermal perception (left) and the degree of comfort (right) at the end of the last phase of activity for each condition and garment.
A descriptive analysis of the data shows that users systematically report warmer and wetter perception during the use of shirt with higher thermal properties (for both conditions). Also by means of a descriptive analysis, the comfort degree reported for each shirt was depending on the environmental condition: shirt with higher thermal properties was perceived as more comfortable regarding temperature and humidity global percep-
638 The Engineering of Sport 7 - Vol. 1 tion at 10ºC (except at highest intensity of the activity) and it was on the contrary at 25ºC (in this case, the difference has the same direction for every activity level). Nevertheless, analysis of the variance (Kruskal-Wallis) showed that, only the environmental conditions caused significant differences (p<0.05) at any activity level in the global temperature and humidity perception.
4- Discussion and conclusions After this study, it has been concluded that it is possible to estimate the adequation of the garment for a particular situation of use defined by the user, the environmental conditions and the activity performed. Independently from the garment, the environmental condition caused differences systematically in skin temperatures and thermal perception at any activity level. The study of the different variables leads to that apart from the variability between users; it is possible to point out to different clothing effects in some occasions. As it has been indicated in section 3. Results, at low-middle activity level is possible to detect an effect of the clothing at each environmental condition on the average skin temperature (high resistance shirt produces higher skin temperature on average), even though both shirts are apparently not very different regarding their thermal properties. However, such differences due to the shirt are not reflected on the armpit humidity, not even at low-middle intensity as far as this variable has presented higher variability inter-subject. Moving on the thermal perception of the user, only environmental condition has been reported as a cause of differences in thermal and humidity perception throughout the whole test (independent of the activity level). However, comfort degree as a function of the condition presents interactions with the activity level. Regarding to the shirt, it was no possible finding significant differences at any variable of thermal perception but users seem to report warmer and wetter perception during the use of shirt with higher thermal properties (for both conditions). In both cases, subjective perception seems to be in accordance with the tendencies in objective measurements in skin temperature and humidity. The study presented shows in one hand that it is possible to have objective measurements for predicting user thermal comfort but in the other hand it also shows that there is a high variability between users that does not let us have enough accuracy to generalise the results. Nevertheless, this inter-individual variability far from being neglected, should be strongly taken into account in order to satisfy user needs by means of selecting an adequate garment according his particular characteristics. Connecting these results with thermal perception of the user may improve the accuracy in the selection of the suitable garment (Page 1994).
5- Aknowledgement Collaborations of the Physiology Department of the University of León (Spain) and the Research Unit for Physical and Sports Performance of the University of Valencia (Spain) have been much appreciated in this work.
The Engineering of Sport 7 - Vol. 1
639
6- References [A1] ASHRAE. Ashrae Fundamentals Handbook 2001. [D1] Daanen, H. Central and peripherial control of finger blood flow in the cold. Thesis. Vrije Universiteit, 1987. [F1] Fanger, P.O. Thermal comfort. McGraw-Hill, 1992. [K1] Kurz, B. The physiology of comfort. Presentation at SATRA - Conference 4/5th May 1994. [PT1] Page, A., Tortosa, L., García, C., Moraga, R., Verde, V. Furniture design based on subjective tests. Application of Discriminant Analysis Methods. International Ergonomics Association Congress IEA’94. Toronto, 1994. [U1] UNE-EN 31092. Textiles. Determination of physiological properties. Measurement of thermal and water-vapour resistance under steady-state conditions (sweating guarded hotplate test). (ISO 11092:1993). [ZB1] Zhu, F., Baker, N. A thermoregulatory model for predicting transient thermal sensation. Contemporary ergonomics. Hanson M. A. (Eds) (Taylor and Francis, London), pp 515-519, 1998.
Compression Sleeves Significantly Counteracts Muscular Fatigue During Strenuous Arm Exercise (P124) Thibaud Thedon1,2, Nicolas Belluye2, Stéphane Perrey1
Topics: Exercise physiology, muscular performance, textile, engineering processes. Abstract: The principle of external compression (EC) regularly used in people with peripheral venous insufficiency has been shown to exhibit an increased oxygenation in healthy subjects at rest on gastrocnemius muscle (Bringard et al. 2006). During exercise, EC allows decreasing electromyography activity (EMG) of hamstrings (sprint; Ringaud et al. 2003) and increasing arterial flow in forearm (task of handgrip; Bochmann et al. 2005). However, no study has underlined the benefit of EC during strenuous exercise, known to alter neuromuscular (force, EMG) and haemodynamic (blood volume, oxygenation) function. After a standardized warm-up, 4 healthy subjects were required to maintain with or without forearm EC (randomised order) a handgrip force equal to 60% of their maximal voluntary force (MVF) until exhaustion. Before and after this task, they realized three MVF. The blood volume (Hbtot) and oxygenation (HbO2) of forearm flexor digitorum superficialis muscle was recorded by near infrared spectroscopy. The EMG RMS activity of flexor carpi ulnaris muscle was recorded with A/D system (MP30, Biopac systems, Inc, USA). With EC, the time to exhaustion was improved of 8% (P > 0.05). After fatigue, our results showed a less important decrease of MVF (11%, P < 0.05) and a higher neuromuscular efficiency (force/EMG) of 22% (P < 0.05) with EC. Hbtot was higher of 61% (P = 0.06) and HbO2 value increased of 90% (P > 0.05) during exercise with EC. As a conclusion, EC over an active muscular region seems to counteract efficiently the deleterious effects of muscular fatigue. Keywords: Muscular fatigue, blood circulation, exercise, compression.
1- Introduction Muscle fatigue can be caused by repeated or prolonged isometric or concentric muscle contractions. It is commonly defined as a decrease of the physiological efficiency during a prolonged task that can lead to the stop of the exercise after a certain time. This “failure time” or “endurance time” strongly depends on the force exerted and more generally on 1. Motor Efficiency and Deficiency Laboratory EA 2991, UFR STAPS, 700 avenue du Pic Saint Loup, 34090 Montpellier, FRANCE - E-mail: thibaud.thedon, [email protected] 2. Décathlon, 4 Bd de Mons, 59650 Villeneuve d’Ascq - E-mail: nicolas.belluye, thibaud.thedon}@decathlon.com
642 The Engineering of Sport 7 - Vol. 1 the physiological and mechanical work performed by the subject. During sustained submaximal contractions, three main markers of fatigue can be observed (BiglandRitchie et al. 1986): the increase of the electromyography (EMG) amplitude, the decrease of the frequency content of the EMG signal, and the gradually increasing voluntary effort needed to maintain the force output. During a sustained voluntary submaximal contraction there is a progressive increase in motor unit activity that can include a change in the number of active motor units and a modulation of discharge rate (Bigland-Ritchie at al. 1986, Lippold et al. 1960). The mechanism by which the slowing of motoneuron firing occurs has been the subject of many investigations but remains unclear. During muscle contraction, it is well known that if the supply of O2 is not adapting quickly enough to the needs, because of low perfusion pressure, adaptation of oxydative metabolism will decrease (Perrey et al. 2001). Changes in muscle oxygen delivery and muscle perfusion are known to affect force or power output of the muscle (Wright et al. 1999). When blood flow is diminished, both muscle endurance and oxygenation decrease, leading to muscle fatigue (Tachi et al. 2004). Motor unit firing and recruitment patterns have been shown to be altered during ischemia (Moritani et al. 1992), suggesting that the lack of oxygen availability increases motor unit discharge rate of high- threshold units (i.e. which are more likely to lead to a fatigue state). The technique of near infrared spectroscopy (NIRS) has been used for several years to evaluate oxygenation during physical work. NIRS has been used initially as a research tool to assess dynamic changes in the status of tissue oxyhaemoglobin (HbO2), deoxyhaemoglobin (HHb), total blood haemoglobin (Hbtot) in brain and muscle. It is a noninvasive method for monitoring oxygen availability and use by the tissues. Human forearm muscle blood flow by NIRS has been compared with venous occlusion plethysmography showing a good correlation (Edwards et al. 1993, DeBlasi et al. 1994). In a fatigue task induced by sustained isometric contraction, Hicks et al. (1999) displayed a reduction of venous blood O2 saturation of the forearm at 30% of maximal voluntary force (MVF). With the elbow flexor muscles (brachioradialis muscle) the maximal deoxygenation was obtained at 50% MVF (Kahn et al. 1998). On the back muscles, significant variations of blood volume and tissue oxygenation were also demonstrated (Yoshitake et al. 2001). In their study, McGill et al. (2000) suggested that the oxygenation decrease arises from an alteration of blood flow in capillary bed during isometric contraction. Finally, Tachi et al. (2004) demonstrated that isometric exercise performed in a leg up condition (i.e. with reduced perfusion pressure) induced a decrease in muscle oxygenation. A recent study on external compression (EC) products usually used in people with venous insufficiency showed an increase in tissue oxygenation of calf muscle, measured by NIRS, in healthy subjects at rest (Bringard et al. 2006). In this study, an appropriate but moderate pressure of 20 mmHg applied over the calf resulted in an increased muscle oxygenation and decreased Hbtot and HHb in comparison with regular shorts and elastic tights. Their findings suggest that EC may improve significantly blood supply to muscles and could enhance muscle function when fatigue occurs. If most studies showed a positive effect of EC on lower extremity, few have worked on upper limbs. Recently, Bochmann et al. (2005) have tested the hypothesis that EC increased forearm blood flow.
The Engineering of Sport 7 - Vol. 1
643
According to these authors, external compression pressure increases tissue pressure and consequently decreases transmural vascular pressure. This decreased transmural pressure may trigger a myogenic response resulting in vessel relaxation. This may in turn lead to a flow rise in small arteries and arterioles. With different applications of pressures from 20 to 30 mmHg, Bochmann et al. (2005) found a significant increase of arterial inflow. In addition, when this range of pressure was used during a task of handgrip (510% MVF, 1 s contraction / 2 s relaxation duty cycle for nearly 70 min), the arterial flow increased significantly. However their measurement was not continuous. Based on the aforementioned studies, this study aimed to evaluate continuously haemodynamic changes and muscle oxygenation with NIRS during a fatiguing task of handgrip. EMG and forearm flexor muscle force were measured to assess fatigue. We hypothesised that the use of EC may counteract muscle fatigue by an enhanced O2 delivery to the working muscle.
2- Methods 2.1 Subjects The effects of compressive sleeves on haemodynamics and muscle fatigue during and after a fatiguing forearm exercise was tested during a pilot study in 4 healthy young and normotensive males (age of 28 ± 3 years). Subjects were requested to refrain from physical exercise and any treatment for muscle soreness or damage for 48 hours prior to and during the study. The protocol complied with the Helsinki declaration for human experimentation and was approved by the local human ethics committee. Possible risks and benefits were explained and written informed consent was obtained from each subject before the experimentation.
2.2 Experimental procedures The subjects seated on a comfortable adjustable chair in front of a table, such that their forearms were laid in a resting pronation position on the table at roughly heart level. For all conditions, the subjects rested in a quiet room with constant temperature (~22 °C) to avoid changes in blood volume. The right side of the upper body was used for all subjects. The elbow angle was flexed 90°. Subjects grasped a handle equipped with a force transducer (Dec 200, Captels, France). The two conditions (with and without EC) were randomly performed in 2 separate sessions. The compression sleeves were made by a commercial manufacturer specialised in that field to deliver an appropriate compression profile to the largest cross-section area of the forearm (Bochmann et al. 2005). External pressure applied by sleeves on the skin was controlled by a validated pressure transducer (Kikuhime, TT Medi Trade, Soleddet 15, Sorro, DK). This small and flexible pressuremeasuring device has an air-filled pressure bladder of 30 x 38 mm dimension and 3 mm thick when calibrated to zero. After a short standardized warm-up (i.e., repeated contractions of moderate-intensity of the wrist muscles), subjects realized three MVFs separated by 60 s of passive rest.
644 The Engineering of Sport 7 - Vol. 1 The mean of the three MVFs values was used for analysis. The fatigue protocol consisted in a sustained isometric contraction of wrist flexors until exhaustion (i.e., task of handgrip). The workload was fixed at 60% of the MVF measured at the beginning of the experiment. A visual feedback was projected on a computer screen to allow subjects to control their force level. Subjects had to continue the fatigue task until exhaustion, when they were unable to maintain the workload for at least 5 s. MVFs were also performed at the end of the experiment, in order to evaluate the effect of fatigue on force generation capabilities.
2.2.1 EMG and force measurements The force transducer was connected to an acquisition A/D board (MP-30 Bipoac Inc., USA). Force signal was recorded simultaneously with EMG activity. EMG of the flexor carpi ulnaris muscle was picked-up using 9-mm diameter bipolar Ag/AgCl electrodes (Contrôle Graphique Medical, Brie-Comte-Robert, France) with an inter-electrode distance of 25 mm. The reference electrode was placed on the wrist. Low impedance between the two electrodes (<5 k) was obtained by abrading slightly the skin with emery paper and then by cleaning it with modified alcohol. All signals were sampled at 2,000 Hz and amplified and filtered (band pass 30– 500 Hz, gain = 1,000). The MVF was quantified as the average value over a 1 s interval centered around the peak force. Root mean squared (RMS) and mean power frequency (MPF estimated by a fast Fourier transform with Hanning window processing) of EMG were determined over the same 1 s interval.
2.2.2 Muscle oxygenation measurements Muscle oxygenation was assessed using the NIRS technique. The NIRS signal provides continuous, non-invasive monitoring of the relative concentration changes in HbO2 and HHb. Hbtot are the sum of HHb and HbO2 concentrations and give an index of the blood volume of the interrogated tissue region. In the present study, changes in Hbtot, HHb and HbO2 of the right forearm flexor muscles were continuously monitored at 2 Hz using a near-infrared spatially resolved spectroscopy oximeter (NIRO-300, Hamamatsu Photonics, Japan). Data were simultaneously transmitted to a personal computer using a RS-232C wire. NIRO-300 optodes were housed in an optically dense plastic holder, ensuring that their position relative to each other was fixed and invariant. The probe (i.e. the optodes support) was secured on the cleaned skin surface with tape. The probe was placed over the muscle belly. The position of the probe on the muscle was marked carefully. The detector in the NIRS probe was separated from the light source by 40 mm. The light emitted by the near infrared probe is assumed to depth tissues at 50% of the interoptode spacing (space between emitting and receiving probes). Skinfold thickness was measured between the NIRS optodes using a skinfold caliper (Holtain Ltd., Crymmych, UK), and was divided by 2 to determine the adipose tissue thickness (i.e. fat + skin layer) covering the muscle. The obtained values of adipose tissue thickness were 2.8 ± 0.9 mm, allowing the NIRS photons to penetrate through the muscle. The absorption of light at different wavelengths (775, 810, 850 and 910 nm) was analysed according to the modified Beer-
The Engineering of Sport 7 - Vol. 1
645
Lambert’s law. Because the differential path length factor (DPF) reported in the literature is estimated from a particular group of subjects, and hence it just represents a mean value for that group, no DPF was utilized in the present study. Changes in HbO2, HHb and Hbtot concentration were reported as a change from baseline values in micromolar units per centimetre (μM.cm). The HHb signal can be regarded as being essentially bloodvolume insensitive during exercise, thus it was assumed to be a reliable estimator of changes in intramuscular oxygenation status and O2 extraction in the field of interrogation. Moreover, the NIRO-300 provides directly tissue haemoglobin oxygen saturation [tissue oxygenation index, TOI=HbO2/(HbO2+HHb) 100, expressed in percentage] calculated independently by using the spatially resolved spectroscopy method, which exploits the source detector multidistance approach. Since subjects realized different times of exercise, NIRS-derived variables were averaged each epoch of 10% of total time to exhaustion. Then, individual data each 10% were averaged together to obtain a mean group response.
2.2.3 Statistical analysis Differences between EC and fatigue conditions (with and without) on NIRS-derived variables, EMG and force values were assessed using a two-way (timeÍcondition) repeated measures analysis of variance (ANOVA). The difference in time to exhaustion with and without EC was tested using a paired t-test. Any differences were further analysed with a Newman-Keuls post-hoc test. A two-way ANOVA was systematically performed if data distribution fails the normality or equal variance tests. All values are presented as mean ± SE. Statistical significance was accepted at P < 0.05.
3- Results 3.1 Endurance time No significant difference was found among EC conditions although the time to exhaustion tended to increase with EC by 8%.
3.2 Neuromuscular fatigue After fatigue, a decrease of MVF of 40% was noted. The decrease of force was less pronounced with EC than without (-12%, P < 0.05). For the flexor muscles, EMG MPF tended to decrease less with EC than without (-14% vs. -24%, respectively, P > 0.05). Meanwhile EMG RMS normalised by RMS measured during MVF before exercise increased less with EC (+68% vs. +90%, respectively, P > 0.05). Finally, neuromuscular efficiency (Force/EMG RMS) was higher by 22% (P < 0.05) with EC.
3.3 NIRS variables During exercise, HbO2, HHb, Hbtot, TOI were greater with EC than without (difference of +110%, +1.5%, +61%, +0.7%, respectively). Statistical analysis revealed no significant difference (P > 0.05) for HbO2, HHb and TOI while P = 0.06 for Hbtot.
646 The Engineering of Sport 7 - Vol. 1
4- Discussion In numerous occupations today the challenge on whole body human muscle power is very small. Instead, small muscles in the upper extremities are repetitively and for prolonged periods of time performing low static forces relative to their maximum strength. Consequently, the aim of this study was to evaluate haemodynamic changes (especially, Hbtot and TOI) during sustained handgrip exercise until exhaustion with and without an ergonomic interplay (i.e. EC) over the forearm flexor muscles. The main result of this study was that the use of EC sleeves during the sustained forearm exercise significantly diminished muscle fatigue without improving significantly endurance performance (+8 %). Subjects were able to hold the initial 50% MVF for 98.1 s without EC and for 106.1 s with EC. These endurance times are well in the range of those previously published in the literature in the absence of EC (Blackwell et al. 1999). It is worth noting that even if the endurance time was higher with than without EC in some subjects, this value was greater than or equal to that observed with EC in the other subjects, and there was therefore no statistically significant difference between the overall endurance times observed with and without EC. This finding is in agreement with the study of Maton et al. (2006) but is partly in contradiction with the decrease in the endurance time reported by Styf (1999) in subjects wearing contention braces. The discrepancy between these data and ours may be attributable to differences in the pressure exerted by the garments tested and the limbs studied. Here, we used deliberately the human forearm because it is not influenced to hydrostatic pressure differences as much as lower limbs. During sustained moderate force contractions, the decreased oxygen availability is associated with decrease in muscle activation and force production, i.e. fatigue development. Measurement of MVF is one direct method to assess muscle fatigue. The lower decline in MVF with EC (-12%) for equivalent endurance time suggests that EC counteracts muscle fatigue during strenuous forearm exercise. In the present study, muscle fatigue was based on EMG recordings during the MVF and spectral analysis of these recordings. During a long-lasting isometric force maintenance task, a decrease in EMG MPF accompanied by an increase in EMG RMS is known to reflect muscle fatigue (Bigland-Ritchie et al. 1986). This was the case in all trials tested. But again, EC tends to better overcome fatigue development with a lower decrease in MPF (10%) and a lower increase in RMS (22%). The decrease in MPF suggested that a decrease in muscle fibre conduction velocity had been induced by local metabolic changes (anaerobic component), i.e. those resulting from peripheral muscle fatigue. However, the possibility that these changes may have been of central origin (firing of the groups III and IV afferents sensitive to metabolic products and O2) cannot be ruled out, since MPF depends partly on motor unit recruitment. In the present study, we hypothesized that an increase in external pressure by sleeves and a subsequent increase in oxygenation and blood volume, measured within the muscle tissue by NIRS, resulted in lower muscle fatigue and possibly underlined changes seen in the EMG responses during the sustained maximal force contractions. Bochmann et al. (2005) showed an increase of arterial inflow as soon as EC is applied on forearm. In our study we have measured tissue oxygenation by NIRS
The Engineering of Sport 7 - Vol. 1
647
method. At the beginning of exercise HbO2 and Hbtot were greatly enhanced with EC. This can explain why subjects with EC increased their endurance times (+8%) because more oxygen was available. Moritani et al. (1992) suggested that if O2 availability decreases, time of endurance decreases too, because fast twitch fibres are recruited preferentially. Supply in O2 is very important at the beginning of exercise to meet the energetic demand. Finally, during the MVC, we calculated the neuromuscular efficiency as the ratio of the force developed divided by RMS activity of the flexor carpi ulnaris muscle. This ratio was significantly higher with EC and confirms the real benefits of compressive sleeves to delay muscle fatigue during sustained forearm exercise. In conclusion, this pilot study tends to show that EC might have a positive influence on muscle function of upper-extremity. The underlying mechanisms explaining this finding are unclear and further studies need to be performed by increasing the number of subjects.
5- References [BF1] Bigland-Ritchie B., Furbush F. and Wood J.J. Fatigue of intermittent submaximal voluntary contractions: central and peripheral factors. In Journal of Applied Physiology, 61: 421-429, 1986 [BD1] Bringard A., Denis R., Belluye N. and Perrey S. Effects of compression tights on calf muscle oxygenation and venous pooling during quiet resting in supine and standing positions. In Journal of Sports and Medicine in Physical Fitness, 46: 548-554, 2006 [BS1] Bochmann R.P., Seibel W., Haase E., Hietschold V., Rödel H. and Deussen A. External compression increases forearm perfusion. In Journal of Applied Physiology, 99: 2337-2344, 2005 [DF1] DeBlasi R.A., Ferrari M., Natali A., Conti G., Mega A. and Gasparetto A. Noninvasive measurement of forearm blood flow and oxygen consumption by near-infrared spectroscopy. In Journal of Applied Physiology, 76: 1388-1393, 1994 [ER1] Edwards A.D., Richardson C. and van der Zee P. Measurement of hemoglobin flow and blood flow by nearinfrared spectroscopy. In Journal of Applied Physiology, 75: 1884-1889, 1993 [HM1] Hicks A., McGill S. and Hughson R.L. Tissue oxygenation by near-infrared spectroscopy and muscle blood flow during isometric contractions of the forearm. In Canadian Journal of Applied Physiology, 24: 216-230, 1999 [KJ1] Kahn J.F., Jouanin J.C., Bussière J.L., Tinet E., Avrillier S., Ollivier J.P. and Monod H. The isometric force that induces maximal surface muscle deoxygenation. In European Journal of Applied Physiology, 78: 183-187, 1998 [LR1] Lippold O.C., Redfearn J.W. and Vuco J. The electromyogram of fatigue. In Ergonomics, 3: 121-131, 1960 [MH1] McGill S.M., Hughson R.L. and Parks K. Lumbar erector spinae oxygenation during prolonged contractions: implications for prolonged work. In Ergonomics, 43: 486-493, 2000 [MS1] Moritani T., Sherman W.M., Shibata M., Matsumoto T. and Shinohara M. Oxygen availability and motor unit activity in humans. In European Journal of Applied Physiology, 64: 552-556, 1992 [MB1] Maton B., Thiney G., Ouchène A., Flaud P., Barthelemy P. Intramuscular pressure and surface EMG in voluntary ankle dorsal flexion: Influence of elastic compressive stockings. In Journal of Electromyography and Kinesiology, 16: 291-302, 2006
648 The Engineering of Sport 7 - Vol. 1 [PT1] Perrey S., Tschakovsky M.E. and Hughson R.L. Muscle chemoreflex elevates muscle blood flow and O2 uptake at exercise onset in nonischemic human forearm. In Journal of Applied Physiology, 91: 2010-2016, 2001 [S1] Styf J. The effects of functional knee bracing on muscle function and performance. In Sports Medicine, 28: 77-81, 1999 [TK1] Tachi M., Kouzaki M., Kanehisa H. and Fukunaga, T. The influence of circulatory difference on muscle oxygenation and fatigue during intermittent static dorsiflexion. In European Journal of Applied Physiology, 91: 682-688, 2004 [YU1] Yoshitake Y., Ue H., Miyazaki M. and Moritani T. (2001). Assessment of lower-back muscle fatigue using electromyography, mechanomyography, and near-infrared spectroscopy. In Journal of Applied Physiology, 84: 174-179, 2001 [WM1] Wright J.R., McCloskey D.I. and Fitzpatrick R.C. Effects of muscle perfusion pressure on fatigue and systemic arterial pressure in humans subjects. In Journal of Applied Physiology, 86: 845-851, 1999
Development of a New System for Measuring Tennis Court Pace (P126) Simon Goodwill1, Steve Haake1, James Spurr2, Jamie Capel-Davies2
Topics: Tennis & other Rackets Sports; Testing, Prototyping, Benchmarking. Abstract: Tennis can be played on a variety of surfaces including clay, grass and acrylic. The performance of each surface is currently classified using the surface pace rating (SPR). This measurement is meant to quantify the speed of a tennis court. The current method of measuring SPR involves the use of an air cannon to propel the ball onto the surface at a nominal velocity and angle of 30 ms-1 and 16 ° respectively. A Sestée device is used to measure the inbound and rebound velocities and angles of the ball. From these measurements, the SPR value can be determined. The International Tennis Federation (ITF) are the governing body of tennis, and one of their roles is to protect the nature of the game. To fulfil part of this role, they need to monitor the surface pace rating for tennis surfaces all over the world. This is difficult to achieve using only the Sestée as this device is (1) relatively expensive, (2) requires a skilled operator to use and (3) difficult to transport. The aim of this current study is to develop a low cost, portable system that can be shipped to various locations around the world. A study was conducted to compare SPR values measured using the Sestée (inbound velocity of 30 ms-1) with a new system (inbound velocity of 13 ms-1). It was found that the SPR values were in good agreement, for the two methods. The new system requires no external power and uses a minimal number of sensors. Keywords: tennis, court, speed, ball, surface.
1- Introduction The ‘speed’ or surface pace rating (SPR) of a tennis court is derived from Brody’s model for a non-spinning, rigid ball (Brody, 1984), as shown in figure 1. It assumes that players associate the speed of the surface with the amount of horizontal deceleration caused by the bounce, i.e. ‘slower’ surfaces lose more horizontal velocity than ‘faster’ ones.
1. Centre for Sport and Exercise Science, Sheffield Hallam University, Collegiate Campus, Sheffield S10 2BP, United Kingdom E-mail: s.r.goodwill, [email protected] 2. International Tennis Federation, Bank Lane, Roehampton, London SW15 5XZ, United Kingdom E-mail: Jamie.Capel-Davies, [email protected]
650 The Engineering of Sport 7 - Vol. 1
Figure 1 - Definition of the velocity components used to calculate the surface pace rating.
The surface pace rating, SPR, is defined as, (1) where μ is the coefficient of friction. The aim of this current study is to develop a low cost, portable system that can be used to measure the SPR of a tennis surface.
2- Design considerations
Figure 2 - A ball cannon and the Sestée.
The current device used to measure SPR is called the Wassing Sestée, and is shown in figure 2. The Sestée consists of two boxes, which are placed on the surface to be measured. At both ends of each box is an array of laser-receiver pairs. When a ball passes into a box, it interrupts a number of laser-receiver pairs. As the relative position of the laser-receiver arrays is known, both speed and angle can be calculated. A ball cannon is used to fire a non spinning tennis ball at 30 m/s and 16 degrees to the horizontal. These impact conditions have been chosen as they represente a typical ball impact found in the game of tennis. By performing the test at realistic speeds, the pace calculated by the Sestée is representative of that perceived by players. The Sestée also measures the coefficient of restitution COR (Vo(y)/Vi(y)) for the surface. This is another parameter that is used to characterise the performance of the surface.
The Engineering of Sport 7 - Vol. 1
651
The design criteria for the new system is listed below. It must: 1- be able to measure the SPR of a tennis surface to with in 5 points of that measured by the Sestée, 2- be able to measure the COR of the tennis surface, 3- require no external power source, 4- be intuitive to operate so that it can be used with minimal training, 5- have no user dependency, 6- be portable and be able to be transported in the ‘hold’ luggage of an airplane. A number of different methods of obtaining SPR have been tested by previous authors. The most comprehensive review of these is given in Hamilton (2000). Most methods aim to measure the ball/surface coefficient of friction μ, and then use equation (1) to calculate SPR. The two main approaches to measuring μ are (1) pendulum and (2) friction sled. Hamilton tested a couple of pendulum designs, but the principle was the same for both. A pendulum was allowed to swing across the surface, and the amount of energy lost in the swing was equated to the coefficient of friction. In general, the pendulum method was very prone to user dependency errors, and was difficult to set up and use. Teasdale (2003) developed a friction sled in which three tennis balls were clamped in a rig, and a vertical load was applied to the sled. The sled was pulled at a constant speed, and a load cell used to measure the force. The coefficient of friction (and therefore SPR) could be calculated from this force reading. For the small number of surfaces tested in that study, the sled results correlated well with the Sestée results. However, internal research conducted by ITF more recently has shown that the friction sled method was unable to accurately measure the coefficient of friction of many carpet surfaces such as Teraflex. Neither the pendulum nor friction sled method are able to measure the COR of the surface and therefore they do not satisfy the design criteria listed above. In order to measure COR, the ball needs to be impacted on the surface, and therefore the new system must essentially recreate the functionality of a Sestée. The main differences between the new system and the Sestée are as follows: 1- the ball will be launched at a lower speed because the new system must have no external power supply, 2- the new system must be significantly smaller/lighter so that it can be carried as ‘hold’ luggage on an airplane, 3- the new system must have more robust sensors than those used on the Sestée. In terms of the ball launch speed, the new system would ideally be able to propel the tennis ball at 30 ms-1, as in the Sestée test. However, the Sestée test uses an air cannon, and the this is not available to the new system. Preliminary testing was conducted by the authors and it was found that the maximum ball speed achieved with no external power source was approximately 13 ms-1. Details of this method are given in a following section.
652 The Engineering of Sport 7 - Vol. 1
3- Velocity dependence of Surface Pace Rating (SPR) In the previous section it was stated that the Sestée measures the SPR for a ball inbound velocity of 30 ms-1. However, it was also stated that the maximum ball inbound velocity which can be generated for the new system is 13 ms-1. Therefore, it is necessary to validate that the SPR of a surface is independent of the ball inbound velocity. In this current section, an experiment which measured the SPR (and COR) of a range of tennis surfaces for inbound velocities of 13 and 30 ms-1 is described.
3.1 Experiment The Sestée apparatus (and air cannon) was used to measure the SPR and COR for a range of surfaces. The ball was propelled from the air cannon at 30 m/s and 16 ° respectively. The air cannon was also used to propel the balls at 13 m/s, however it was found that the Sestée was unable to track the ball correctly for this lower ball inbound velocity. Therefore, a Vision Research Phantom v4.3 high speed video camera was used to film these impacts. The camera was operated at 1000 frames per second with a shutter speed of 100 μs and a resolution of 800 x 600 pixels. The camera was used to calculate the inbound and rebound velocities. Seven different artificial tennis surfaces were tested. These cover the range of typical surfaces used in tennis. No particulate surfaces (eg clay) or natural surfaces (grass) were tested. The surfaces tested were a range of different constructions. Each sample was approximately 1 m2 in size. The samples tested included two different samples of Taraflex carpet, one sample of Rebound Ace and four samples of acrylic based paint with a varying amounts of sand content. A smooth rigid acrylic sheet was also tested. Whilst this is not a realistic tennis surface, it is represents an extreme low friction surface.
3.2 Results
Figure 3 - Comparison of (a) SPR and (b) COR, for two different ball inbound velocities (13 and 30 m/s).
The Engineering of Sport 7 - Vol. 1
653
Each surface was tested nine times, and the average SPR and COR value for each surface was calculated from this data. The data for this study is presented in figure 3 (a) and (b), with each data point representing one of the eight surfaces. To give a direct comparison of the values obtained by each method, the values are plotted against each other. Furthermore, a 1:1 line is also shown (equivalent of y = x). A data point that lies on (or close to) this line indicates a good correlation between the two methods. Figure 3 (a) shows that the SPR values measured at the two different velocities correlate very closely. The only data point which lies significantly away from the 1:1 line represents the impact on a smooth acrylic sheet. This is clearly not representative of a tennis surface so is of limited relevance in this study. Figure 3 (b) shows that the COR values measured at 13 m/s are of a similar order of magnitude to those measured at 30 m/s. It is difficult to conclude whether there is a good correlation between the two methods because the COR is only measured over a small range.
4- New system - Sprite 4.1 Design
Figure 4 - Schematic and photograph of the new system (Sprite).
Figure 4 shows a schematic diagram and photograph of the prototype device. This device has been codenamed Sprite. The basic operation of the Sprite is summarised below, 1- The user pulls a cord which rotates the arm to its starting position. A trigger holds the arm against the restoring force of a torsional spring. 2- The user drops a ball into the carrier. The carrier comprises of a concave platform for the ball to rest on, and a sprung finger which holds the ball on the concave platform. 3- When ready, the trigger is released and the torsional spring accelerates the arm, carrier and tennis ball in an arc. The arm is brought to rest suddenly by two stiff rubber stoppers. At this point the ball is released.
654 The Engineering of Sport 7 - Vol. 1 4- The ball breaks two light beams as it travels from the carrier to the surface. It then rebounds and hits a position sensor. The light beams and position sensor are powered by the 5V output from a National Instruments USB-6009 data acquisition device. The USB-6009 is sampled using a laptop PC. The outputs from the two light beams are sampled on a single channel in differential mode. The position sensor is 145 mm long, and it outputs a voltage between 0 and 5V depending on the location that the ball hits the sensor. Further details of the sensor can be found in Infusion Systems (2008). The output of the position sensor is sampled in single ended mode (on a separate channel). A sampling rate of 20 kHz was used for the USB-6009.
4.2 Trajectory reconstruction model
Figure 5 - Trajectory reconstruction model.
4.2.1 Inbound ball velocity and angle The trajectory of the ball in the Sprite device is shown in figure 5. The positions P1 & P2 (and times T1 & T2) are easily obtained from the two light beam sensors, assuming that the ball cuts the beams along the centre line of the ball. The ball inbound velocity can be calculated from the time interval (T2-T1), and the ball inbound angle is defined by the light beam layout.
4.2.2 Rebound ball velocity and angle Unlike the Sestée, the Sprite device does not measure the ball rebound velocity and angle directly. The method used to obtain the rebound velocity and angle is more complicated, and requires a series of assumptions. The position P3 and time T3 can be accurately predicted by knowing the P1 , P2 , T1 and T2. The position P3 represents the ball immediately prior to impacting the surface and P4 is the instant that the ball leaves the surface. Position P4 and time T4 can be calculated if the contact time (ball in contact with surface) and contact distance are known, or can be assumed. A separate study was conducted where the contact time and contact distance were measured for a range of tennis surfaces. It was found that the minimum and maximum
The Engineering of Sport 7 - Vol. 1
655
contact times were 4.6 and 5.0 milliseconds respectively, for the surfaces tested. Therefore an average value of 4.8 milliseconds was assumed for the model and therefore T4 = T3 + 4.8 . It is unsurprising that the contact time exhibits little variability between surfaces because the surface is typically significantly stiffer than the tennis ball, and therefore the contact time magnitude is primarily dependent on the ball. The contact distance was found to typically range between 40 and 48 mm. An average value of 44 mm was assumed for the model, and therefore P4(x) = P3(y) + 44. The position P5 and time T5 were directly obtained from the signal from the position sensor. By knowing the positions P1-5 and the times T1-5, the complete trajectory of the ball can be recreated, and therefore the SPR and COR values can be calculated.
4.2.3 Sensitivity analysis Three major assumptions were made in the trajectory reconstruction model: 1- the contact time is assumed to be a constant and equal to 4.8 ms, 2- the contact distance is assumed to be a constant and equal to 44 mm, 3- the ball passes perfectly through the two light beam sensors. Table 6 - Illustration of the effects of varying the model inputs (contact time and contact distance).
Table 6 illustrates the effects of the assumptions made regarding contact time and contact distance. It can be seen that adjusting the value of the assumed contact time has negligible affect on both SPR and COR values calculated by the model. However, the value of the assumed contact distance does affect the calculated SPR value. This will clearly have an influence on the accuracy of the Sprite device to measure the correct SPR value for the surface it is being used to test. However, it should be noted that for the full range of contact distances, the maximum variation in SPR is only ±1.6 . Table 7 - Illustration of the effects of varying the assumed value of ball height P2(y).
656 The Engineering of Sport 7 - Vol. 1 The effect of the assumption that the ball passes perfectly through the two light beam sensors is difficult to quantify. A high speed video camera was used to record the position of the ball as it passed through the light gates. A first order polynomial trend line was plotted through this trajectory, and the positional accuracy was calculated as ±0.5 mm. It was found that the ball passed through the first beam with a repeatability that was equal in magnitude to the measurement accuracy (±0.5 mm). However, the ball exhibited more height variation when passing through the second beam. Typically it was found that the ball could pass through the second beam with a maximum variability in height of ±1 mm (equivalent to ±0.4° inbound angle). The sensitivity of the model to the fact that the ball must pass perfectly through the light beams is shown in Table 7. It can be seen that the COR values are highly sensitive to the small variability in the ball height. Therefore, each test will exhibit a large uncertainty in the COR value. However, for each surface, the test is repeated nine times, and providing that the scatter in the ball height has no bias in one direction, the average COR value should exhibit a higher accuracy than indicated by the scatter in table 7.
5- Validation of Sprite v1 In section 3 it was shown that the SPR and COR values measured at 13 and 30 m/s are similar (using high speed video and Sestée equipment respectively). In this section, the results obtained at 30 m/s (Sestée) are compared with those obtained from the Sprite.
Figure 8 - Comparison of (a) SPR and (b) COR, for two different ball inbound velocities (13 and 30 m/s).
Figure 8 (a) compares the SPR values measured using the Sestée with those obtained from the Sprite. Each data point represents the average of nine tests. It can be seen that there is a good correlation between the two methods. Apart from one data point (smooth acrylic sheet), the maximum difference between the two methods is approximately ±5 points. However, it should be noted that the uncertainty in the Sestée measurements is ±2 points, and therefore some of the difference between the two methods is due Sestée measurement error.
The Engineering of Sport 7 - Vol. 1
657
Figure 8 (b) compares the COR values measured using the two methods. The maximum difference between the two methods is 0.04. Initially this seems to indicate a poor performance of the Sprite (assuming the Sestée is the ‘gold’ standard). However, this may easily be explained by considering the error analysis in Table 7. In figure 8 (b) the COR values are, on average, 0.02 higher for the Sprite. It was shown in Table 7 that an error of 1 mm in the ball height as it passes through the second light beam, leads to an error of 0.02 in COR. Therefore, the systematic shift in COR results in Figure 8 (b) could simply be due to the ball passing 1 mm lower than the intended position. If the model was adjusted to account for this, then the COR results comparison in Figure 8 (b) would show no systematic difference between the two methods. Clearly, this indicates that in the future, a quality control system must be implemented to ensure that the model settings are correct (specifically ball height through second beam).
6- Conclusions and Future Developments In this paper it has been shown that the a simple device comprising of a pair of light beams and a position sensor can be used to estimate the surface pace rating (SPR) and coefficient of restitution (COR) of a tennis court. A trajectory reconstruction model was used to reconstruct the inbound and rebound trajectories of the ball. It was found that the calculated value of SPR was not sensitive to the assumed values of contact time and contact distance. Furthermore, the SPR value was not sensitive to the height of the ball as it passed through the second light beam. However, it was found that the calculated value of COR was highly dependent on the height of the ball as it passed through this beam. In the future, a quality control system must be implemented into the Sprite to ensure that an accurate value of ball height (through the second beam) is used in the trajectory model.
7- References [B1] Brody, H. That’s how the ball bounces. The Physics Teacher, 22, 494 – 497, 1984. [H1] Hamilton, G. The Development of Test Methods to Characterise Sports Surfaces, MSc Thesis, University of Sheffield, 2000. [T1] Teasdale, C Friction of Artificial Sports Surfaces, MEng Thesis, University of Sheffield, 2003. [I1] Infusion Systems http://infusionsystems.com/catalog/product_info.php/products_id/46 , 2008.
A Feedback System for Coordination Training in Double Rowing (P127) Arnold Baca, Philipp Kornfeind1
Abstract: A measuring station for assessing the coordination in double rowing on land has been developed. Two rowing ergometers are put on slides allowing to imitate the on-water team situation. Pulling forces are measured for both ergometers simultaneously. One ergometer is equipped with a device for measuring reaction forces at the foot stretcher. A set of parameters comprising time values, stroke length and force values characterizing the individual strokes and their synchrony are calculated. The system may be used to give extrinsic feedback to the rowers on their technique and on their coordination during training. In addition, seat specific differences can be analyzed (e. g. possible differences in the rowers’ adaptibility). Coaches can therby be assisted in the team selection. Keywords: Ergometer rowing, team selection, force patterns, biomechanics, coordination.
1- Introduction Boat propulsion in team rowing is strongly affected by the level of synchrony between the members of the team. Most biomechanists assume that optimal boat propulsion results from a uniform force application from all crew members ([MS1]). Successful rowers show similar force patterns when they have been rowing together for a long time ([H1]). In double sculls, rowers experience a common motion of the shell providing the rowers with feedback cues that might help to reduce differences in their individual forcetime profiles ([BH1]). More experienced rowers are assumed to better be able to adapt their own force-time profile when rowing with another person ([WW1]). Strategies, in which athletes row out of phase, in order to provide a more continuous impulse to the system – thus minimizing shell velocity fluctuations – have also been suggested ([BH1]), but did not gain practical relevance, which may be explained by construction characteristics of the boat. Pairing of athletes should therefore be determined on the basis of their ability to match force time profiles and to generate a balanced cumulative blade force ([BH1]). 1. Department of Biomechanics, Kinesiology and Applied Computer Science, ZSU, University of Vienna, Auf der Schmelz 6A, 1150 Wien, Austria - E-mail: [email protected], [email protected]
660 The Engineering of Sport 7 - Vol. 1 Dynamic analyzes in the boat are difficult to realize and demanding in time and instrumentation. In many cases, analyzes are therefore based on rowing simulators (ergometers) on land ([PH1], [LB1]). One typical simulator is the Concept 2 ergometer. If put onto slides, a construction allowing the ergometer to roll back and forth during the rowing stroke, the situation in the boat is better imitated ([BK1]). This setup may also provide the basis for offering feedback on certain aspects of the rowing technique. A cascaded double ergometer system (Figure 1) has been developed for this purpose. It was constructed in order to • give feedback on reaction forces and selected parameters quantifying coordination in (almost) real time • serve as diagnosis system to analyze the adaptability of athletes or to identify a master/slave behaviour in specific double pairs (depending on the position of the athletes on the cascaded ergometer) • assist coaches in the team and position selection
2- Methods A cascaded ergometer system was assembled. Two rowing ergometers (Concept2, Vermont, USA) are connected by putting them on three slides (Figure 1).
Figure 1 - Cascaded ergometers in feedback session.
Force transducers (U9B, Kistler, Winterthur, Switzerland) are fixed between the chain and the handle (Figure 2) of each ergometer in order to measure the pulling forces.
Figure 2 - Handle with integrated force sensor.
The Engineering of Sport 7 - Vol. 1
661
The handle position relative to the ergometer is determined from a precision potentiometer combined with a gearing unit (ratio 2:1) which is mounted on the axle of the flywheel (Figure 3).
Figure 3 - Determination of the handle position.
In the current version, only one of the ergometers may be used to measure reaction forces at the foot stretcher (the device, the feet are strapped into). It is equipped with two identical constructions based on load cells and strain gages, which allow determining reaction forces perpendicular and in parallel to the platforms (Figure 4) for both feet separately ([BK2]). Based on these two force components the vertical and horizontal reaction forces can be calculated. At the moment only the horizontal reaction forces are used for further calculations. Consequently it is not possible to directly compare those parameters (foot stretcher) between the bow and the stroke during the double trials. A LabVIEW® (National Instruments, Austin, USA) program has been developed to record the data measured and to calculate characteristic parameters quantifying synchrony. The pulling force curves and a selectable set of these parameters may be presented to the rowers stroke by stroke in almost real time or as summative feedback after a series of strokes.
Figure 4 - Foot stretcher equipped with force plates.
662 The Engineering of Sport 7 - Vol. 1 In particular, the following variables are computed for each stroke (from catch to catch) and each rower resp. from the force curves of both rowers (cf. [H1]): Parameter calculated for both rowers independently (see Figure 5): • Duration of the stroke ts [s] • Duration of the drive phase td [s]: end time - onset time of pulling force • Maximum of the pulling force FPmax [N]) • Instant of this maximum tFPmax with regard to the duration of the drive phase [%] • Area under the pulling force curve AFP [Ns] • Centre of the force graph tFP50 [%]: instant at which the force graph is divided into two halves of equal area related to the duration of the drive phase • Maximum of the power Pmax: maximum of pulling force times pulling velocity [W] (A light bungee cord tension keeps the carriage, where the legs of the ergometer are put on, centered on the base of the slides. This tension is neglected in the power calculation) • Instant of this maximum with regard to the duration of the pulling phase tPmax [%] • Average power per stroke Pavg [W] • Maximum of the total horizontal reaction force at the foot stretcher during the drive phase FFmax [N] • Instant of this maximum with regard to the duration of the pulling phase tFFmax [%] • Area difference between the horizontal reaction forces of left and right foot during the drive phase ALR [Ns] – if positive, the impulse from the left foot is larger • Stroke length sl [m] All values are calculated from the ergometer equipped with the device measuring foot stretcher forces. Combined parameters calculated: • Time difference between the onsets of the pulling forces of the two rowers ton [s] (bow - stroke) • Time difference between the finishes of the pulling forces of the two rowers toff [s] • Area difference between the pulling forces AP [Ns] (if positive, area of bow is larger) • Form difference defined as the area between the normalized (area = 1) pulling forces APN These parameters are calculated from the pulling forces of both rowers measured simultaneously.
The Engineering of Sport 7 - Vol. 1
663
Figure 5 - Example of pulling and foot stretcher force curves. Note that the foot stretcher force actually acts into the opposite direction.
3- Application The application of the system for giving stroke by stroke feedback is illustrated in Figure 1. A screen is positioned in front of the rowers. Pulling forces of both rowers are presented simultaneously. The rowers try to alter their movement pattern in order to better match the curves. Subjects exercising on the system have confirmed the assumtion that the feeling is similar to that in the boat. Two case studies shall demonstrate the use of the system for diagnosis purposes.
3.1 Men’s double The influence of the seat position on the synchrony was investigated for two male rowers aged 26 and 27 years with body masses of 90 kg (rower A) and 85 kg (rower B). Both are well experienced rowing athletes (several years of serious sport, winner of national championships) and exercise currently at mass sport level. Among others they have rowed together in a quadruple scull for about 3 and in a double scull for about 2 seasons. Compared to a fixed ergometer on the floor no subjective influences regarding their rowing motion on the moving slides were observed from them. In addition, the seat specific differences between single and double rowing were analyzed. One ergometer
664 The Engineering of Sport 7 - Vol. 1 equipped with force transducer, handle locator and instrumentation for foot stretcher reaction force measurement was used for obtaining the respective parameter values for single rowing. Exercises were performed with 18 as well as 30 strokes per minute. After a warm-up the rowers started exercising on the ergometer(s). After running in, the measurements were started. Each rower began with the trials on the single ergometer. Next, double ergometer sessions were analyzed. Since only one ergometer was equipped with the device for measuring foot stretcher reaction forces, four combinations of ergometers and rowers were evaluated (position change of ergometers as well as of rowers). 15 consecutive strokes were recorded for each variant. Parameters characterizing individual rowing performance in single ergometer rowing and as stroke and bow were calculated from the trials on the ergometer with the device only (15 strokes). Combined parameters were determined from 2 sequences of 15 strokes (position change of the ergometers). Mean values and standard deviations of selected parameter values are presented in Tables 1-7. Table 1 - Time parameters of rower A (18 and 30 strokes per minute).
Table 2 - Time parameters of rower B (18 and 30 strokes per minute).
The Engineering of Sport 7 - Vol. 1
665
Table 3 - Time differences between onsets and finishes (estimated from 2 sequences of 15 consecutive strokes).
Table 4 - Stroke lengths (18 and 30 strokes per minute, Rower A and B).
Table 5 - Force and power parameters of rower A (18 and 30 strokes per minute).
Table 6 - Force and power parameters of rower B (18 and 30 strokes per minute).
666 The Engineering of Sport 7 - Vol. 1 Table 7 - Area and form differences (estimated from 2 sequences of 15 consecutive strokes).
Considering the time parameters, it can be observed that nearly all values characterizing the behaviour of rower A precede those from rower B. The values presented in Table 3 demonstrate that rower A started and finished the strokes before rower B independently from the seat position. Moreover, from the values calculated for tFpmax, tFP50, tPmax and tFfmax (Tables 1 and 2), it can be concluded that the corresponding points in time occur even more earlier for rower A compared to rower B than the onset times of their strokes. At 30 strokes per minute, tFP50, for example, occurs at 48 % of the driving phase for rower A and 52 % for rower B on average, if rower A is stroke, as well as 49 % for rower A and 49 % for rower B, if A is bow. The values for ton and toff (Table 3) are smaller, if rower B is stroke. At the higher stroke rate (30 strokes per minute), this sequence also results in smaller differences values for tFpmax, tFP50, tPmax and tFfmax. Hence, time synchrony appears to be better in this case. However, at the lower stroke rate (18 strokes per minute) these difference values are slightly larger at this order (B / A). In this situation, a clear decision, which sequence might result in more synchrony (A / B) or (B / A) can not be made from the time parameters, although (B / A) might be slightly, but insignificantly, better. Differences in stroke lengths are smaller for rower A being stroke and rower B being bow. Considering force, impulse and power parameters derived from the pulling forces (FPmax, AFP, Pmax and Pavg), it can be observed that nearly all values are more similar for rower A as stroke. At 30 strokes per second, AFP, for example, is 315 (±8) for rower A and 327 (±8) for rower B, if rower A is stroke, but 329 (±12) for rower A and 297 (±8) for rower B, if A is bow. From these values, it can be followed that more synchrony can be achieved with rower A being stroke. This conclusion is also confirmed by the results from the calculation of the force differences AP (Table 7). On the contrary, the values given for the form differences APN in this table suggest a reverse sequence of rowers (B / A). Since these differences are strongly affected by the timing of the rowers, this result is not surprising with reference to the time parameters discussed above. Considering the reaction forces at the foot stretcher, similar maximum values can be observed for FFmax for both stroke rates and positions. Compared to the single ergometer situation, smaller values were calculated for rower A (exception: SR 30, stroke A) and larger values for rower B (same exception). In 5 of the 6 conditions investigated, the mean value of the maximum pulling force of rower A (FPmax) was smaller than that of FFmax (Table 5). Quite the opposite could be observed for rower B (Table 6). The sequence (B / A) resulted in more similar values.
The Engineering of Sport 7 - Vol. 1
667
In order not to draw wrong conclusions and to eliminate disturbing factors such as the actual personal condition of fitness, tests of the kind described have to be repeated. This is particularly important, if subjects are not used to ergometer rowing on slides. For the present case study a confirmation of the results could then be basis for feedback training. One strategy could be to position rower A as stroke and rower B as bow. In the first feedback sessions the rowers could then try to improve the synchrony of the time parameters, in particular that of the onsets of the strokes.
4- Discussion The case study presented is a good example for demonstrating the potential and the limits of the method presented. On the one hand, all relevant parameters describing the dynamic force pattern can be obtained rapidly and easily without disturbing the athletes in performing their motions. Force curves and parameters can vividly be presented and the athletes can try to change their motion pattern in the desired direction. On the other hand, identifying this direction is a challenging task. Even the question, which rower should be stroke and which bow is sometimes difficult to answer. In the example given, the time parameters indicate that a sequence B / A could be the variant to prefer. The pulling force and impulse parameters rather suggset to select rower A as stroke and rower B as bow. However, the system presented allows an easy and continous control of the change of the different characteristic parameters in time. Thus, the effectiveness of recommendations can be quantified. In order to simplify the system’s application and to be able to present the foot stretcher reaction forces from both rowers simultaneosly during exercising, a second measuring device is required. This device, which is currently under construction, will also allow comparing the reaction forces on a stroke by stroke basis. The devices for measuring the reaction forces at the foot stretcher are also applicable in the boat ([BK2]). In combination with dynamometric oarlocks, paramters can be meassured and calculated for the boat and compared with those obtained for the double ergometer (cf. [BK1]). This comparison between boat and ergometer situation has not yet been performed.
5- Conclusion Assuming that synchrony of the force-time profiles is an essential factor for the coordination in double rowing, a feedback system based on two Concept 2 rowing ergometers put on slides has been developed. Reaction forces on the foot stretcher of one ergometer as well as the pulling forces are measured; parameters describing the shape and magnitude of the force curves are calculated. The system has proven to be easily applicable in practical feedback training. Athletes were able to observe, how changes in their movement pattern altered the force curves, which were displayed on a screen in front of them stroke by stroke. In order to identify promising parameters, which should be changed by (feedback) training, the system is helpful to diagnose peculiarities in the force-profiles. Profound understanding of the biomechanics of rowing is required for this purpose.
668 The Engineering of Sport 7 - Vol. 1
6- Acknowledgements We thank Richard Roscher for his contribution in constructing the cascaded ergometer measurement system and for assisting us in performing the measurements.
7- References [BK1] A. Baca, P. Kornfeind & M. Heller, Comparison of foot-stretcher force profiles between onwater and ergometer rowing, Proc. XXIV Int. Symposium on Biomechanics in Sports, Vol. 1, Salzburg, Austria: Univ. Press. 2006, 347-350. [BK2] A. Baca, P. Kornfeind & M. Heller, Feedback systems in rowing, The Engineering of Sport 6, Vol. 1, Developments for Sports, 2006, New York: Springer, 407-412. [BH1] A. Baudouin & D. Hawkins, A biomechanical review of factors affecting rowing performance, British Journal of Sports Medicine, 36, 2002, 396-402. [H1] H. Hill, Dynamics of coordination within elite rowing crews: evidence from force pattern analysis, Journal of Sports Sciences, 20, 2002, 101-117. [LB1] J.M. Loh, A.M.J. Bull, A.H. McGregor & R.C. Schroter, Instrumentation of a Concept II rowing ergometer for kinetic and kinematic data acquisition, The Engineering of Sport 5, Vol. 2, Sheffield, 2004, 173-179. [MS1] M.E. McBride, D.J. Sanderson & B.C. Elliott, Seat specific technique in pair oared rowing, Proc. 19th Int. Symposium of the Society of Biomechanics in Sports, San Francisco, CA. 2001, 263-266. [PH1] P.N. Page & D. Hawkins, A real-time feedback system for training rowers, Sports Engineering, 6, 2003, 173-179. [WW1] A.M. Wing & C. Woodburn, The coordination and consistency of rowers in a racing eight, Journal of Sports Sciences, 13, 1995, 187-197.
Modelling the Oblique Impact of Golf Balls (P128) James Cornish1, Steve Otto2, Martin Strangwood1
Topics: Golf, Modelling, Materials. Abstract: The performance of golf balls during various shots occurs under a range of strains and strain rates, which results in varying ball stiffness and energy loss responses due to the viscoelastic nature of the polymers used in ball construction. Modelling of ball performance is often based on fitting FE models to ball rebound data, although some groups have started to successfully use the material properties of the various components to account for normal impact. The latter approach is more predictive as it allows the potential effects of varying material properties and ball constructions to be estimated before balls are actually fabricated. The normal impact is often represented through a normal coefficient of restitution (CoR), which is strain and strain rate dependent. In oblique impacts the situation is complicated as both normal and tangential forces act with the bending moment of the tangential force causing backspin generation. Modelling of oblique impact will need to be able to predict contact time of the ball on the club face as this is the time during which the tangential force is generating backspin. One route to this contact time could come from the normal part of the impact as the ball centre of mass velocity normal to the face is decelerated and then accelerated. This paper will report results of impacts of a number of commercial ball types with a series of angled plates at a number of impact speeds. These tests have been analysed in terms of normal and tangential CoR values and show that the normal part of the oblique impact can be treated independently of the tangential. In addition the relationship between backspin rate and loft angle, determined by the tangential component can be characterised using an effective friction coefficient, which has been related to the material properties and dimensions of the cover and sub-surface regions of the golf ball. Keywords: Golf, oblique impact, viscoelastic behaviour, coefficient of restitution, backspin.
1. Sports Materials Research Group 2. R et A Rules Ltd, The University of Birmingham, Dept of Met and Mat, 6 Pilmour Links, Elms Road, Edgbaston, Birmingham. B15 2TT. UK, St Andrews, Fife, Scotland. KY16 9JG - Email: [email protected] - Email: [email protected]
670 The Engineering of Sport 7 - Vol. 1
1 - Introduction Impact sports, such as tennis, cricket, squash, hockey and golf, often involve interaction between linear or visco-elastic implements (bat, club, stick) and visco-elastic balls (solid or hollow). The development of greater quantitative understanding of these interactions is an active area of research often directed towards equipment optimisation with respect to the athlete to aid manufacture and coaching. One popular method of relating equipment properties to performance is via finite element (FE) modelling, which can be fitted to experimentally determined ball behaviour (Fuss, 2007). This approach is not suitable for predictive modelling of the effects of different materials or dimensions in the ball, for which the visco-elastic properties of the individual materials need to be determined and used in visco- or hyper-elastic FE models (Duris, 2004;,Tanaka et al., 2006). Models (FE and other types) have been successfully developed for hollow (Dignall et al., 2004; Ashcroft & Stronge, 2004) and solid (Duris, 2004; Tanaka et al., 2006; Fuss, 2007) sports balls for normal impact, where a single dominant force is acting and can be related to the strain and strain rate dependence of materials behaviour. The application of modelling to oblique impact is less well established as the occurrence of normal and tangential forces along with non-Coulombic friction relationships increase the complexity of the process. There is a need to develop verified relationships between normal and oblique impacts and materials properties. This has been addressed by ‘slip ratios’ for oblique impacts of hollow tennis balls (Goodwill & Haake, 2004), which show limited visco-elastic behaviour so that the rigid body approximation of the ‘slip ratio’ is still valid. The solid nature of golf balls increases their visco-elastic nature so that the ‘slip ratio’ approach is less valid; this paper reports characterisation of normal and oblique impacts for commercial golf balls in order to establish relationships between them.
2- Experimental A number of commercial multi-piece golf balls, Table 1, were obtained in two batches (A, C – E) and (3, 4 and 6). These were characterised by whole ball compression tests using a Zwick Z100 universal testing machine at crosshead speeds of 0.02, 0.2 and 2 mm/s to total ball deformations of between 2.25 and 10.5 mm allowing tangential ball stiffness to be measured as a function of deformation and deformation rate. Ball dimensions and component properties were measured as reported previously (Johnson et al., 2004). The normal impact behaviour was determined by firing balls (equilibriated at 25 C for several hours) at a fixed 200 kg vertical steel target at speeds of 20 – 50 m/s using an ADC Supercannon 2000. Horizontal speeds prior to and after impact were measured using a pair of light screens and were used to determine coefficient of restitution (CoR) values as the ratio of speed after impact (VOUT) to speed before impact (VIN). The same cannon was used to carry out oblique impact tests using the same balls and grooved plates at effective lofts of 20 – 70 and inbound speeds of 20, 30 and 40 m/s. The speed after impact, launch angle and spin rate were all determined using a commercial stereoscopic launch monitor.
The Engineering of Sport 7 - Vol. 1
671
3- Results and Discussion Oblique impact tests showed that, initially, backspin rate and launch angle increased with increasing loft angle, up to a maximum loft angle beyond which the backspin rate decreased, Figure 1. This has been noted previously (Monk et al., 2005; Cornish et al., 2007) and both the rate of backspin increase on the rising portion and the angle of maximum backspin were dependent on ball type. The oblique impact can be treated as a normal force (bringing the ball to rest in a direction normal to the face and then accelerating it off the face), FN, and a tangential force, FT, which causes the torque to generate backspin. The normal; and tangential forces can be related by an effective coefficient of friction, μeff (Monk et al., 2005). FN is related to the normal deceleration rate, which will depend on ball compression and the normal inbound speed, i.e. (1) (Full derivation of Eqn 1 in Monk et al., (2005)). Backspin is generated by nonnormal forces, i.e. the tangential force. The backspin rate, , will be related to the torque ( FT) acting and the time for which it acts (the contact time, which is related to the compression and de-compression of the ball). From Eqn 1 and the relationship between FN and FT (Monk et al., (2005)), this would give: (2) Therefore, if this analysis is correct, μeff should be proportional to /cos(Loft angle). This analysis is shown for balls A, C, D and E for a constant inbound speed of 30 m/s in Figure 2. Table 1 - Characteristics of ball types studied.
Figure 1 - Backspin variation with loft angle.
Figure 2 - Relationship between μeff, and loft angle.
672 The Engineering of Sport 7 - Vol. 1 Figure 2 shows a generally straight line relationship confirming, within experimental error, the hypothesis above. The deviation at high backspin rates for ball D is due to the angle of maximum backspin rate being approached. The analysis above, therefore only applies, as expected, to the rising portion of the backspin / loft angle curve. This relationship between backspin rate, loft angle and μeff allows backspin rates to be predicted if the ball and material characteristics controlling the proportionality can be identified. The normal deceleration of the ball is caused by compression of the ball, which is the mechanism acting during normal impact and, if the normal and tangential forces can be dealt with independently, then the speed normal to the face after impact should be related to the speed normal to the face before impact and the normal CoR. The normal CoR was determined as a function of inbound speed for the balls in Table 1. From plots such as in Figure 3, the normal CoR was determined for each ball type at the ball speed normal to the face, i.e. at VIN cos(Loft angle, ). For the oblique impact tests the outbound speed normal to the face was determined (VOUT cos(Launch angle, )) as defined in Figure 4 for the angled plate tests, so that the normal CoR from the oblique test can be determined from: (3) The CoR values from equation (3) were then compared with the CoR value for the same ball at an inbound speed of VINcos( ) determined from normal impact tests. The resulting ratio was determined over a range of loft angles, e.g. Figure 5, for all the balls studied.
Figure 3 - Normal and angled (50°) plate CoR values for ball 3 as a function of inbound speed.
Figure 4 - Definition of loft angle ( ) and launch angle ( ) for angled plate oblique impact tests.
The Engineering of Sport 7 - Vol. 1
673
Figure 5 - Ratio of normal CoR determined in oblique impact tests to normal CoR determined at same normal speed from normal impact tests as a function of loft angle.
As shown in Figure 5, up to the loft angle corresponding to the maximum backspin rate, the normal CoR values in both types of test are equal indicating that the normal and tangential aspects can, within experimental error, be treated as independent. This will allow the contact time and normal outbound speed for oblique impact to be determined from the normal impact behaviour, which, itself can be related to the compression behaviour of the core materials and the Shore D hardness of the mantle / cover (Strangwood et al., 2006). The tangential behaviour is more complex as it depends on meff, which (Cornish et al., 2007) is a function of the cover interaction with the grooves on the face / plate and so depends on the compression of the cover, mantle (if present) and core during oblique impact. FN will be greater at lower loft angles for which the cover effectively sticks to the face / plate, not moving up the plate, but remaining fixed whilst compression and shearing in the mantle / core resulting in ‘winding-up’ of the backspin rate under the action of FT. Therefore, the assumption can be made that frictional forces are equal to the forces tending to slide the ball up the face, and that this situation exists until sliding starts. This latter situation has been shown, from high speed imaging (Cornish et al. 2007), to occur only as the backspin rate decreases for loft angles beyond that corresponding to the maximum backspin rate. Thus, on the rising backspin portion of Figure 1 the conditions holding are that tangential forces are equal to sliding forces up the face and that, for each particular ball type, the loft angle at which maximum backspin rate occurs defines the limiting value of tangential force, which can be assumed to be that acting at lower loft angles. This could be found empirically, as in this study, but is related to the deformation of the ball cover into an effectively non-deforming groove in the face / plate under the constraint of an underlying mantle / core. Hence, the dimensions of core / mantle and cover and their properties should control the frictional force along the face / plate and the loft angle corresponding to maximum backspin rate. Sectioning of the balls in this study along a diameter allowed the dimensions of the components along with their Shore D values (determine strength and modulus (Qi et al., 2003) to be determined. Considering the deformation of a thin layer between two harder substances suggested
674 The Engineering of Sport 7 - Vol. 1 that the interaction between the cover and the face / plate would be related to subsurface (mantle or core) hardness (SH), cover hardness (CH) and cover thickness (CT). The relationship then becomes: (4) The relationships between these ratios and the loft angle corresponding to maximum backspin rate are shown in Figure 6 and indicate a general straight line relationship. The fit is good for two-piece balls and seems to be good for three- and four-piece balls, although the spread in data for the latter types of ball is not good, requiring values between 0.5 and 1.2 for the parameter in equation (4) to be sure that the relationship holds and to establish the reason for the apparently different slopes for the two-piece and multi-piece ball lines. This is not due to cover material type as both PU and ionomer covered two-piece and multi-piece balls were studied, Table 1. Thus, relationships have been found empirically that relate the construction (materials and dimensions) of a solid golf ball to the normal and tangential components of force acting during oblique impact. These relationships should allow backspin and launch angle to be estimated for a given loft angle below the loft angle corresponding to maximum backspin rate, which can also be estimated from these relationships. This study now needs to be taken forward to predict the backspin and launch angle for a given construction ball before experimentally measuring it.
Figure 6 - Variation in loft angle corresponding to maximum backspin rate and ball material property and dimension ratio.
4- Conclusions and Future Work The experimental data for oblique golf ball impacts have confirmed increasing backspin rate and launch angle with increasing loft angle up to a ball-dependent maximum. For loft angles below that corresponding to the maximum backspin rate the behaviour of the ball normal to the face / plate is the same as that for an entirely normal impact so that modelling of the normal and tangential components of the impact can be carried out
The Engineering of Sport 7 - Vol. 1
675
independently (for the range of ball types studied). The tangential behaviour can be dealt with using the effective friction coefficient, μeff, which has been related to the material properties and dimensions of the cover and sub-surface region of the ball. μeff has been shown to account for backspin / loft angle variation up to the loft angle corresponding to maximum backspin rate so that it should be possible to predict the backspin rate behaviour of golf balls based on their component material properties and dimensions. The data currently obtained for multi-piece balls needs to be supplemented to prove the generality of the relationships determined and a model based on these relationships needs to be tested by predicting then measuring the backspin / loft relationships, which would form the next stage of this research.
5- Acknowledgements The authors would like to thank R & A Rules Ltd and EPSRC for provision of samples, equipment and financial support during this study.
6- References [As1] Ashcroft A.D.C. and Stronge W.J., Dynamic dissipation during bounce of tennis balls. In The Engineering of Sport 5, vol. 1, M.Hubbard, R.D.Mehta and J.M.Pallis (eds), ISEA, Sheffield, 2004, pp 386-392. [CO1] Cornish J.E.M., Otto S.R. and Strangwood M., The Influence of Groove Profile, Ball Type and Surface Conditions on Golf Ball Backspin Generation. In The Impact of Technology in Sport II, F.K.Fuss, A.Subic and S.Ujihashi (eds), Taylor & Francis, London, 2007, pp 229-234. [DG1] Dignall R.J., Goodwill S.R., Haake S.J. and Miller S., Tennis GUT – modelling the game. In The Engineering of Sport 5, vol. 2, M.Hubbard, R.D.Mehta and J.M.Pallis (eds), ISEA, Sheffield, 2004, pp 382-388. [D1] Duris J., Experimental and numerical characterization of softballs. MS Thesis, Washington State University, 2004. [F1] Fuss F.K., Non-linear Viscoelastic Properties of Golf Balls. In The Impact of Technology in Sport II, F.K.Fuss, A.Subic and S.Ujihashi (eds), Taylor & Francis, London, 2007, pp 207-221. [GH1] Goodwill S.R. and Haake S.J., Ball spin generation for oblique impacts with a tennis racket. Experimental Mechanics, 44(2): 195-206, 2004. [JO1] Johnson A.D.G, Otto S.R. and Strangwood M., Radial property variations in solid golf balls and their effects on impact performance. In The Engineering of Sport 5, vol. 2, M.Hubbard, R.D.Mehta and J.M.Pallis (eds), ISEA, Sheffield, 2004, pp 10-16. [MD1] Monk S.A., Davis C.L., Otto S.R. and Strangwood M., Material and surface effects on the spin and launch angle generated from a wedge. In Sports Engineering, 8(1): 3-11, 2005. [QJ1] Qi H.J., Joyce K. and Boyce M.C., Relationship between durometer hardness and the stressstrain behaviour of elastomeric materials. In Rubber Chemical Technology. 76: 419-435, 2003. [SJ1] Strangwood M. Johnson A.D.G. and Otto S.R., Energy losses in viscoelastic golf balls. In Proc. I. Mech. E.,Part L: Journal of Materials, Design and Applications, 220(1): 23-30, 2006. [TS1] Tanaka K., Sato F., Oodaira H,, Teranishi Y. and Ujihashi S., Construction of the finite-element models of golf balls and simulations of their collisions. In Proc. I. Mech. E.,Part L: Journal of Materials, Design and Applications, 220(1): 13-22, 2006.
Modelling and Stability Analysis of a Recumbent Bicycle with Oscillating Leg Masses (P131) Brendan Connors1, Mont Hubbard1
Topics: Bicycle; Modelling. Abstract: It has been observed in the testing of a recumbent bicycle with a very low centre of gravity that the pedalling cadence can affect the rider’s ability to control the vehicle. To understand the relationship between cadence and bicycle stability, a multibody dynamic model is created. This model has nine massive rigid bodies: the bicycle frame with fixed rider torso (with head & and arms), the front fork with handlebars, the front wheel, the rear wheel, the left thigh, the right thigh, the left shank with foot, the right shank with foot, and the cranks. Nonlinear equations of motion are compiled in Autolev, a symbolic calculator using Kane’s method for multibody dynamics (Autolev, 2005). A simulation of the bicycle slowly accelerating from its starting speed (5 m/s) to its target speed (35 m/s) is run iteratively over several gear ratios. A steering controller is implemented to stabilize the bike outside its stable stable speed range. The simulation displays the lean and steer angles as well as steering control torque. Lean angle and control torque increase significantly with cadence, and steer angle increases slightly with cadence. This relationship is used to create a shifting strategy to reduce the control effort needed by the pilot during top top-speed speed-record attempts. Keywords: recumbent; bicycle; modelling; stability; cadence.
1- Introduction Since the invention of the bicycle, riders and frame builders have been pushing the limits of top speed. While the configuration of a bicycle used in professional racing is strictly governed by the Union Cycliste Internationale, the International Human Powered Vehicle Association places very few restrictions on the configuration of human powered vehicles used in human powered land land-speed speed-record attempts. The most timetested and successful configuration for such a record attempt is a low-slung, fully faired recumbent bicycle. This design has many benefits: a two-wheeled vehicle is more stable at high speed than a narrow three- or four-wheeled vehicle, and a low, reclined recumbent has minimal frontal area, is less susceptible to disruption by crosswinds than a taller 1. University of California, Department of Mechanical and Aeronautical Engineering, One Shields Ave., Davis, CA 95616, USA - E-mail: bwconnors, [email protected]
678 The Engineering of Sport 7 - Vol. 1 bicycle, and still allows the rider to exert a great deal of power. However, the low-slung recumbent has significantly different handling qualities from a traditional upright bicycle, mostly due to the low centre of gravity and decreased moment of inertia along the roll axis. Furthermore, pilots of bicycles at the World Human Powered Speed Challenge have reported that in many record attempts, their own power output was not the limiting factor for top speed; rather, the controllability of the vehicle kept them from going faster, or at the least required a significant portion of their energy (Cook, 2003.[C2]) Road tests of a low-slung recumbent have shown that pedalling cadence affects the ease with which a rider can control the bicycle. It is assumed that the mass in the rider’s legs, which is a substantial portion of the entire vehicle’s mass, excites the lateral dynamics of the bicycle-rider system in an oscillatory manner. In order to To understand the relationship between the oscillations of the legs and the dynamics of the bicycle, a multibody dynamic model is created using the aid of computer software. Iterative simulation of the model with varying gear ratios illustrates the conditions under which the bicycle is difficult to control and the severity of the oscillations. This information can allow racers to formulate a shifting strategy for top top-speed runs that avoids undesirable oscillations in lean and steer.
2- Methods The most commonly referenced mathematical bicycle model is the Whipple model, named for Francis Whipple who first proposed the model more than a century ago [W1](Whipple, 1899). Since its introduction, it has been revised and altered, and the model was benchmarked by Meijaard et al. (2007) [MP1]. The model consists of four rigid bodies: the bicycle frame with fixed rider, the front fork with handlebars, the front wheel, and the rear wheel. The wheels have knife-edged tires that roll without slip and without rolling resistance. The revolute joints between the frame and fork, frame and rear wheel, and fork and front wheel are all frictionless, and no control torque is applied to the steer axis. The absence of control torque can be thought of as hands-free riding. While the hands-free condition may not be a realistic condition for study, it is assumed that a bicycle that is stable with no steering control will be easier to control than one that is unstable in the hands-free condition. The model must have both tires in contact with the ground at all times, necessitating a configuration constraint on the pitch of the frame. In addition, the no-slip condition requires four nonholonomic constraints on the longitudinal and lateral motion of the front and rear contact points. Autolev, a symbolic manipulator, is used to enforce the constraints and derive equations of motion for the bicycle [A1](Autolev, 2005). Eight generalised coordinates describe the configuration of the model, as seen in Ffigure 1. The four rigid bodies associated with the frame, fork, rear wheel, and front wheel are X, F, WR, and WF, respectively. Starting with an inertial reference frame N, variables xP1 and yP1 define the point of contact of the rear wheel in the ground plane. Two auxiliary reference frames are used to rotate the bike frame in yaw and lean, with the angles of rotation given by and , respectively. Since Because the bike has nonzero trail (the distance of the front frontwheel contact point behind the intersection of the steer axis and the ground plane in the
The Engineering of Sport 7 - Vol. 1
679
upright, unsteered configuration), the bike frame must be allowed to pitch slightly about the rear axis to maintain contact between the front wheel and ground. The pitch angle is given by y. The fork rotates through the steer angle relative to the frame around a steering vector s, rotated up from a horizontal line in the sagittal plane of the frame by constant angle (commonly referred to as the head angle). Finally, the rear wheel rotates relative to the frame about its axle through angle r, and the front wheel rotates relative to the fork about its axle through angle f.
Figure 1 - The bicycle model, with four rigid bodies, generalised coordinates, and points of interest. Some unit vectors are resized or hidden for clarity.
A four-body system starts with 3 three translational and 3 three rotational degrees of freedom for each body – a total of 24 degrees of freedom. A hinge (revolute joint) restricts a body to rotation in one direction relative to another body, so each hinge eliminates 5 five degrees of freedom. Both wheels must remain in contact with the ground, thus eliminating 2 two more degrees of freedom. This reduces to 7 seven the number of coordinates required to describe the system. The model described here uses 8 eight coordinates. Since Because the relationship between pitch and other variables cannot be expressed practically [K1](Kane, 1975), it is eliminated via a motion constraint: if the pitch is such that the contact points p1 and p2 in the front and rear and front wheels, respectively, are at zero height, then the time derivative of their difference in height must remain zero for all time. (1)
680 The Engineering of Sport 7 - Vol. 1 . The derivative relation in Eequation (1) allows the pitch . angle rate to be expressed . in terms of the lean angle rate and the steer angle rate . Finally, four nonholonomic rolling constraints are defined, two for each wheel. While the contact point in the rear wheel can be calculated with any point in the ground plane, the velocity of this point must be zero for all time to satisfy the condition of rolling without slip., (2) (3) Likewise, the front wheel contact point must also have zero velocity in the ground plane., (4) (5) Autolev uses the nonholonomic constraint equations together with the constraint . . . equation for pitch rate to solve for three independent generalised speeds: , , and r, the time derivatives of lean, steer, and rear wheel rotation angles, respectively. Thus, the system has three degrees of freedom. Seven variables fully describe the configuration of the model at any time, and three generalised speeds fully describe its motion at any time. After validating the Autolev model with results from the benchmark model [MP1](Meijaard et al., 2007), the model is modified by adding five rigid bodies: the cranks, the left thigh, the right thigh, the left shank with foot, and the right shank with foot. The ankles are fixed to avoid static indeterminacy – it is assumed that the flex of the ankles does not have a great effect on the motion of the leg masses. The motion of the cranks is directly related to the motion of the rear wheel, so no new degrees of freedom are introduced to the model. Parameters for the recumbent bicycle with four bodies and for the recumbent bicycle with moving legs come from a 3-D solid model derived from measurements from an actual bicycle and rider and formulated in Solidworks [S1](2005). Meijaard et al. [MP1](2007) linearised the equations of motion for the Whipple model and calculated eigenvalues and eigenvectors for three modes of motion: caster, which is always stable with positive velocity; weave, which is unstable at zero velocity and becomes stable as speed increases; and capsize, which starts out stable but becomes unstable as speed increases. A bicycle model with oscillating leg masses, on the other hand, cannot exhibit asymptotic stability – there will always be some oscillation of the lean and steer due to the oscillation of the leg masses. Therefore, an eigenanalysis of the oscillating-leg model cannot be undertaken. Instead, a nonlinear simulation is executed in which the bicycle slowly accelerates from a low speed (5 m/s) to a high speed (35 m/s)
The Engineering of Sport 7 - Vol. 1
681
and lean and steer could be examined over the simulation. This speed progression is accomplished by adding a constant accelerative torque to the rear wheel in the model. Carvallo [C1](1900) discovered that the four-body model of a bicycle has a stable stable speed range, below which the weave mode is unstable and above which the capsize mode is unstable. Since Because the nonlinear simulation occurs at speeds outside this stable stable speed range, a controller for steering torque must be devised. Define a scalar performance index for the controller as, (6) . . whereis the vector ➝ x = [, , , ] of state variables, is the steering torque applied between the rider and the handlebars about the steer axis, Q is a constant positive definite weighting matrix with only diagonal elements aii = 1/xai2 where xai is the maximum allowable value of the ith element of the state vector. Likewise, r is a positive scalar equal to 1/(.a)2 where a is the maximum allowable value of control torque. The state variables
r, r are omitted from this formulations since because it can be shown [MP1] . (Meijaard et al., 2007) that speed u0 = rr r is constant to first order in such a linearisation. It is well known [BH1](Bryson and Ho, 1969) that the optimal control that minimises the performance index J is a linear state feedback control law of the form,
= – K ➝ x (t)
(7)
where the constant feedback gain matrix K is calculated using standard techniques in Matlab [M1](Matlab, 2007). The state equations for the recumbent bicycle are linearised numerically in Matlab using a centre-differencing method around = 0, = 0, and forward speed = 20 m/s (halfway between 5 m/s and 35 m/s). Maximum . allowable values . are chosen arbitrarily to be a = 0.1 rad, a = 0.1 rad, a = 1 rad/s, a = 1 rad/s, a = 2 N-m. The optimal gain matrix is K = [22.4185, 71.2568, 1.9909 s, 1.2619 s] rad-1. While these values for K may present an optimised controller for the system, they may be unrealistic. It is hypothesised that a human rider would control a bicycle primarily based . on lean angle and rate. Thus, control torque is selected to be = – 20 • – 20 • . The controller was first benchmarked on the Whipple model, stabilising it at most speeds below the stable stable speed range and at all speeds above it. The controller also stabilised the four-body bicycle model with the parameters of the recumbent bicycle over the range of speeds of interest for the simulation. That a controller can stabilise two bicycle models with dramatically different parameters over the range of 5 m/s to 35 m/s is indicative of how easily a bicycle can be stabilised at high speed. Once it is verified that the non-oscillating recumbent model would be stable across the speed range in question, simulations are executed with the cranks (and legs) moving at a set gear ratio – one of four gear ratios implemented on the actual bike. In addition to amplitudes of the periodic lean and steer angles, the magnitude of the steering control torque (t) is plotted to show the amount of rider input necessary to control the bike.
682 The Engineering of Sport 7 - Vol. 1
3- Results Figures 2 and 3 show the amplitude of oscillations in lean angle, steer angle, and steering control torque in a nonlinear simulation in top gear accelerating from 5 m/s to 35 m/s over 300 s. The amplitude of oscillations in lean angle increases almost linearly with speed, while the amplitude of oscillations in steer angle has relatively little change (Figure 2). The steering control torque is also oscillatory, gradually increasing in amplitude (Figure 3). Changes in gear ratio primarily result in a change in the rate of increase in lean angle amplitude with time, as well as the rate of increase in steering control. The largest recorded lean or steer angle with a controlled model in top gear was approximately 0.018 rad, or 1 degree. The steering torque reached a maximum value of more than 8.5 N-m. In top gear, this steering torque oscillates at 1.8 Hz. For comparison, the largest lean angle in first gear is approximately 0.035 radians (2 degrees), and the steering control torque is more than 21 21 N-m at 3.35 Hz.
Figure 2 - Amplitudes of the lean angle and steer angle for a nonlinear simulation in top gear with constant acceleration from 5 m/s to 35 m/s. Oscillations have been removed for clarity.
Figure 3 - Amplitude of the steering control torque for a nonlinear simulation in top gear with constant acceleration from 5 m/s to 35 m/s. Oscillations have been removed for clarity.
The Engineering of Sport 7 - Vol. 1
683
4- Discussion The response shown in Figures 2 and 3 to the leg excitation gradually increases with speed and corresponding pedalling frequency. If the pedalling frequency were ever to approach the frequency of a poorly damped weave mode, one would expect resonant behaviour, i.e. large resonant peaks of both lean and steer angles as well as steering torque. This behaviour is not apparent in the simulation results shown because of the fact that riders cannot pedal economically at frequencies approaching the frequency of the closed closed-loop weave mode. For example, in first gear at 35 m/s the pedalling cadence would be 3.35 Hz, or approximately 200 rpm, which is less than half the frequency of the closed-loop weave mode at this speed. The top speed modelled in this exercise is 35 m/s, which is a lofty goal for all but a few pilots at the World Human Powered Speed Challenge. The magnitude of oscillation in lean angle at this speed (0.018 rad) is not significant enough to disrupt a rider from accelerating a bicycle in a top top-speed run. On the other hand, the magnitude of steering control torque at this speed (8.5 N-m in top gear) combined with the frequency of oscillation (1.8 Hz) would be a sizeable input task for a pilot, especially during maximal exertion. The magnitude of the control torque is most likely due to the weighting of the control torque in the linear quadratic regulator, but nonetheless it provides a clear choice for shifting criteria. If we set the allowable amount of steering control torque is set at 7.161 N-m, then the logical shift points would be 16.369 m/s, 20.406 m/s, and 25.540 m/s, as shown in Figure 4. This also allows a top speed of 31.283 m/s, a respectable top speed that only an elite few have attained [C2](Cook, 2003). Note that these shift points are selected to minimise lean oscillation and the control torque necessary to stabilise it, and their effect on power production is not considered. Shifting strategy should still depend heavily on power production and efficiency. The data on lateral oscillation should be integrated into a total shifting strategy that maximises power production while minimising the effort a rider puts into controlling the bicycle. Another strategy for reducing the magnitudes of lean oscillation and control torque necessary for stabilisation is counterbalancing. If crank arms could be designed with inertial properties (as are designed for automotive engine crankshafts) such that each side’s thigh, shank with foot, and crank arm have combined centre of gravity and moments of inertia that did not change much with crank angle, then oscillations in lean and steer due to oscillating legs could be negated. This strategy presents not only a mathematical design challenge but a significant production challenge, as the cranks would have to be several times larger in size and mass to counterbalance the rider’s legs. However, a smaller counterbalance could be implemented to reduce the oscillations without altering the cranks to completely unfeasible proportions.
684 The Engineering of Sport 7 - Vol. 1
Figure 4 - Amplitude of steering control torque when gears are shifted so as to avoid exceeding a control torque of 7.161 Nm.
5- Conclusions With the aid of a symbolic manipulator and numerical integration software, a nine body model of a bicycle with oscillating leg masses was created and simulated at varying speeds and gear ratios. The results of these simulations show a direct correlation between pedalling cadence and the control torque required to stabilise the bicycle at high speeds, confirming observations made in road tests. As cadence increases, steering control torque increases, becoming substantially large at the speeds and cadences seen in human powered land-speed-record attempts. Two strategies are proposed to ameliorate the large control torque required to stabilise the bicycle under these conditions: a shifting strategy that minimises the control torque to a reasonable maximum value, and the possibility of a counter-balanced crankset that reduces the excitations induced by the oscillating leg masses.
6- References [A1] Autolev 4.1. Online Dynamics, Inc. 2005. [BH1] Bryson, A. E. and Ho, Y. C. Applied Optimal Control. Blaisdell, Waltham, MA, 1969. [C1] Carvallo, E. Theorie du movement du monocycle et de la bicyclette. In Journal de L’Ecole Polytechnique, Series 2, Part 1, Volume 5, “Cerceau et Monocycle”, pp. 119-188, 1900. Part 2, Volume 6, “Theorie de la Bicyclette”, pp.1-118, 1901. [C2] Cook, B. World human powered speed challenge 2003. Barcroft Cycles Video, Falls Church, VA, 2003. [K1] Kane, T. Fundamental kinematical relationships for single-track vehicles. In International Journal of Mechanical Science, Pergamon Press, UK, Volume 17, 1975. [M1] Matlab, version 6.5.1. Control Toolbox, lqr.m. The Mathworks, Inc., 2003.
The Engineering of Sport 7 - Vol. 1
685
[MP1] Meijaard, J.P., Papadopoulos, J.M., Ruina A., and Schwab, A.L. Linearized dynamics equations for the balance and steer of a bicycle: A benchmark and review. Proceedings of the Royal Society A463, pp. 1955-1982, 2007. [S1] Solidworks 2005. Solidworks Corporation, 2004. [W1] Whipple, F.J.W. The stability of the motion of a bicycle. The Quarterly Journal of Pure and Applied Mathematics, Vol. . 30, pp. 312-348, 1899.
Computerised Games for Balance Training: A Pilot Study on Collegiate Females (P135) Jonathan S. Wheat, Ben Heller, Stephanie Lovick1
Topics: Virtual Reality & Computer application in Sports, Biomechanics. Abstract: Sporting and everyday tasks often require effective control of posture. Many studies have assessed interventions used to improve balance and postural control and, recently, interventions based on computer games have been reported (e.g. Betkar et al., 2006). Often, the equipment required for such interventions is expensive and inaccessible. In this study, we present the details of a relatively inexpensive posturally-controlled computer games system suitable for home use and the results of an initial four week intervention using the system. Sixteen healthy, collegiate-aged, adult volunteers were assigned to either an exercise or control group. The exercise group completed a four-week, posture-controlled, computergame-based intervention. The custom-built system comprised a set of modified electronic weighing scales interfaced to a laptop computer. Custom software calculated the centre of pressure (COP) from the vertical force transducers in the weighing scales. Participants completed three 15 minute training sessions per week in which they played various games (e.g. Pong, Breakout) controlled via the COP. The same games were played by the control group during the same number of sessions but controlled via a mouse. Measures of static (force platform) and dynamic (game score) balance were taken both pre- and post-intervention to evaluate the effectiveness of computer games in improving balance and postural control. Participants in the exercise group improved their performance at the computer games (p < 0.05). The exercise group also showed improvements in static balance with, for example, a reduction in anterior-posterior COP variability during two-legged eyes-open stance (p < 0.05). These improvements were not reflected in the control group. Therefore, this initial study suggests that posture-controlled computer games could be effective in improving balance and postural control. The potential benefits of this approach, in the context of the limitations of the study, will be discussed. Keywords: postural control, exercise, sway, game play, intervention.
1. Sports Engineering Research Group, Sheffield Hallam University, UK - Email: j.wheat, b.heller, [email protected]
688 The Engineering of Sport 7 - Vol. 1
1- Introduction Effective control of balance is important in meeting the demands of many everyday tasks and sporting activities. Impairments in balance and postural control have been reported in older adults (e.g. van Wegen et al., 2002) and people with various neuromuscular disorders - e.g. stroke (e.g. Peurala et al., 2007) and Parkinson’s disease (e.g. van Wegen et al., 2001) Also, postural control impairments have been associated with the occurrence of injuries in sports performers such ankle sprains/functional ankle stability (e.g. McGuine et al., 2000). Furthermore, better balance ability has been reported in expert performers in sports with extensive demands on balance and postural control (e.g. gymnastics, Robertson et al., 1994). The importance of balance and postural control in sports and during everyday activity has led to the development of many exercise-related programs to improve balance and postural control. The efficacy of such programs, including physiotherapy and occupational therapy (e.g. Paillex and So, 2005), Tai Chi (e.g. Wong et al., 2001), home-based exercise (e.g. Campbell et al., 1997) and balance board training (e.g. Verhagen et al., 2004) has been the subject of much research. Lord et al. (2001) suggested that an intervention must involve pushing the individual to, or near to, the limits of the participant’s equilibrium. Various interventions that achieve this have been assessed and suggest potentially beneficial effects. For example, Paillex and So (2005) investigated the effect of hospital based physiotherapy exercise on balance and fall occurrence in a group of stroke patients. They reported improvements in balance – as quantified using various force platform measurements – as a result of the intervention. Beneficial effects of balance training have also been reported in cohorts of sports people. For example, McGuine and Keene (2006) reported the effect of a balance training program on the occurrence of ankle sprains in high school soccer and basketball players in a large scale randomised control trial (n = 765). During the course of a season, the rate of ankle sprains was significantly lower for subjects who participated in the balance training (half the risk). Importantly, Campbell et al. (1997) suggested that, to be beneficial, balance-based exercise programmes need to be easy to implement, cost effective and simple. For these reasons, Campbell et al. (1997) explored the efficacy of home based balance rehabilitation exercises in older females in a six-month study. The home-based exercise included walking, strength and balancing activities. The results suggested improved balance and decreased risk of falls in the exercise group but Campbell et al. (2005) recently suggested that a significant mediating factor in the success of many such balance-based exercise programs is participant compliance. Furthermore Betker et al. (2006) suggested that, in many participants, a lack of interest can impair the potential effectiveness of therapeutic exercise. Betker et al. (2006) highlighted that rewarding activities can improve a participant’s motivation to practice. A potential approach to making balance-based exercises rewarding is to introduce a gaming element to the activity. This can be achieved by exploiting real-time biofeedback information which has been used to augment therapeutic exercises for many years (e.g. Sihvonen et al., 2004). Betker et al. (2006) presented the results of an initial study in which a system based on video game-based balance exercises was developed. The system comprised a flexible pressure mat interfaced with a laptop computer that was capable of
The Engineering of Sport 7 - Vol. 1
689
monitoring the position of the users centre-of-pressure (COP) in real-time. The position of the COP is used to control various computer games which require users to explore their boundaries of stability and which challenge equilibrium. For a small number of participants (n = 3) with various pathologies (excised cerebellar tumor, single right cerebrovascular accident, traumatic brain injury), Betker et al. (2006) reported trends towards improved balance (decreased COP excursion limits) and reduced risk of falls after periods of computer games-based balance exercise. Furthermore, participants also demonstrated increased attention during training and were motivated to increase their practice volume (Betker et al., 2006). The preliminary results of Betker et al. (2006) suggest several potential benefits of computer game-based balance exercise. However, the sample in the study was small and participants had specific balance-related pathologies. More work is warranted to determine if similar benefits to balance and postural control are present in a larger sample from different populations. Furthermore, the balance computer game system used by Betker et al. (2006) was relatively expensive (~$8700) which might limit the widespread use of this and similar systems in non-lab-based settings e.g. clinic, gym, home. Therefore, the aim of this study was to develop a relatively inexpensive balance computer game system that could be suitable for large scale use in clinics, fitness gyms and the home to improve balance and postural control. Furthermore, we present the results of an initial, pilot, four week intervention using the system with healthy collegiate-aged females.
2- Methods 2.1 The balance computer game system In developing the system, our aim was to design a simple, cost-effective means of obtaining centre-of-pressure (COP) data, displaying these data in real-time and allowing the user to control a series of computer games through whole body weight shift movements (Figure 1). As such, we chose to use a set of customised commercially available weighing scales (Glass Precision Electronic scale, Weight Watchers ~ £20) to obtain the COP signal. The scales (275 mm X 300 mm) were adapted to output voltage data (0-5 V) from each of the four force transducers positioned at the corners of the plates. When the system is in use, these voltages are transferred to a laptop computer via a 12-bit analogue-to-digital converter (NI USB6009, National Instruments Corporation, Austin TX, USA) sampling at 400 Hz. Subsequently, the data are downsampled to 25 Hz by averaging 16 consecutive samples – which acts as a simple moving average filter – and are converted to force data using calibration values determined for each transducer. The transducers were statically calibrated by placing on them a series of known weights (10 N, 50 N, 100 N and 200 N). The offsets and gains for each transducer were obtained from the linear regression line fitted to the force vs voltage data. Once the forces from the four transducers are acquired, the COP, relative to the origin of the scales, is calculated in the x and y directions by resolving moments acting on the scales about the y and x axes, respectively. The COP data are then used to control a suite of balance computer games.
690 The Engineering of Sport 7 - Vol. 1
Figure 1 - The System setup.
In designing the COP-controlled computer games we followed similar principles to those outlined by Betker et al. (2006). Importantly, to maintain the interest of the participants, the games were designed to be competitive and to be challenging in terms of both magnitude and speed of movement. In all games, the difficulty and demands on the participants can be scaled (e.g. faster/more precise movements required) to allow progression. This, in conjunction with a “High Score Table” for each of the games helps foster both personal and group competitiveness. According to Betker et al. (2006) games designed in this way offer the following: 1) goal-directed and intended behaviour, 2) multitasking and 3) rewards with moment to moment goal attainment. The following sections outline details of the three balance-based computer games that have been developed.
2.1.1 Balance-based computer game descriptions Balloon Burst: The aim of Balloon Burst is to move a dart around the screen to burst as many balloons as possible in a preset period of time (Figure 2a). The dart position is determined from the position of the user’s COP so the user can control the position of the dart using whole body weight shifts. Balloons appear at random positions in the blue circular area (movement space - see Figure 2a). If a balloon is not popped within a set period of time it disappears and reappears at another random location within the movement space. When a balloon is popped, positive reinforcement is provided via a sound and an increment in the score displayed on the top right of the screen. To allow manipulation of the difficulty of the game the following parameters are configurable: 1) ‘Move’ – the time between balloon location changes, 2) the size of the balloon, 3) the area in which the balloons can appear, 4) ‘Hover Time’ – the time the user has to dwell before the balloon pops. Pong: This game is an adaptation of the well known early computer game (Figure 2b). The aim of the game is to keep the ball in play for as long as possible. The ball rebounds off the two sides and top of the screen but the game is over if the ball is allowed to hit the bottom of the screen. To prevent this, the user moves the paddle (off which the ball rebounds) to intercept the ball. Similar to Balloon Burst, the position of the
The Engineering of Sport 7 - Vol. 1
691
paddle is determined from the position of the user’s COP so the user can control the paddle using whole body weight shifts. When the paddle makes contact with the ball a sound is played and a point is added to the score – the measure of performance in this game. Furthermore, each time the paddle makes contact with the ball, the speed of the ball is increased meaning that the game becomes progressively more challenging. The game can also be made more difficult by increasing the size of the movement space and decreasing the size of the paddle. Break Out: Again, this is an adaptation of a well known computer game. The goal is to guide the ball to the top of the screen by destroying the rectangular blocks initially blocking the path (Figure 2c). Similar to Pong, the ball rebounds off the sides of the screen but the game is over if the ball is allowed to hit the bottom of the screen. To prevent this, the user moves the paddle (off which the ball rebounds and which is controlled by the user’s centre of pressure) to intercept the ball. When the ball makes contact with a rectangular block at the top of the screen the block explodes. Once the user clears a path and guides the ball to the top of the screen they progress to the next level. As the level increases, the speed of the ball increases. Achievement on this game is determined by the level attained and the number of blocks destroyed at that level. Similar to Pong, Break Out can be made more challenging by increasing the size of the movement space and decreasing the size of the paddle.
Figure 2 - Screenshots of (a) Balloon Burst, (b) Pong and (c) Breakout.
2.2 Computer game based exercise program 2.2.1 Research design This study was a small-scale, pilot randomised-controlled trial to investigate the effects of a four week computer game based exercise program on balance and postural control in collegiate-aged females. Sixteen healthy females were randomly assigned to either the exercise (n = 8) or control groups (n = 8). Over the four week period, both groups were required to attend 12, 15 minute computer game play sessions. During the sessions, the exercise group was required to play four games using the balance computer game system: 1) Balloon Burst 1 - the movement space was set to 150 X 150 mm, balloon size was 15 X15 mm, Move was 3 seconds and Hover Time was 0.5s; 2) Balloon Burst 2 - same
692 The Engineering of Sport 7 - Vol. 1 settings as Balloon Burst 1 but Hover Time was set to 0; 3) Pong - the movement space was set to 150 X 150 mm and the paddle width was 30 mm; 4) Break Out - same settings as Pong. Participants in the exercise group controlled the games using the position of their centre-of-pressure, whereas, participants in the control group used a mouse. The exercise program consisted entirely of computerised balance game play and, during the four-week study period, participants in both the control and exercise groups did not alter their normal day-to-day exercise activity. During the 15 minute sessions, participants spent an approximately equal time playing each of the four games. The effectiveness of the program of computerised balance game play exercise was assessed by testing participants’ balance and postural control both pre and post the exercise period. Details of data collection and analysis are given below.
2.2.2 Participants Before data collection began, all procedures were approved by the Faculty’s Ethics Committee and all participants provided written informed consent. Sixteen, healthy collegiate females volunteered to take part in the study. All participants had no history of neurological disorders or cardiovascular disease and were free from injury for at least two months prior to the start of the study. The participants were randomly assigned into a control group (n=8) and an exercise group (n=8). Subsequently, differences in the age, stature and mass of the groups were examined using an independent t-test ( = 0.05) to ensure that the groups were similar. There were no significant differences between the groups (Table 1). Table 1 - Control and Exercise group characteristics.
2.2.3 Experimental setup During pre and post exercise programme testing sessions a portable force platform (9286AA Kistler Instrumente AG, Switzerland) - covered with a sheet of paper - and amplifier (9865 Kistler Instrumente AG, Switzerland) were used to obtain the anteriorposterior and medio-lateral components of COP. The output of the force platform was collected using a 12 Bit A/D convertor that sampled data at 100 Hz. The force platform was positioned one metre from a wall, on which was positioned - at eye level - a visual target. In addition to the force platform measures, during pre and post exercise programme testing sessions participants scores on Balloon Burst 1 and Balloon Burst 2 were recorded, requiring the use of the balance computer game system. During testing, the scales of the system were placed 1 m in from the laptop screen – the top of which was positioned at a height of 1.45 m.
The Engineering of Sport 7 - Vol. 1
693
2.2.4 Protocol Testing procedures were identical in both the pre and post exercise testing sessions. The data collection sessions started with the participants being asked to assume a standing position on the force platform with feet parallel and 15 cm apart, with arms by their sides. Subsequently, an outline of the feet was drawn on the paper covering the plate and these markings were used to ensure consistent foot placement in all trials of the two foot stance conditions. Participants then completed five quiet stance trials in three conditions: two feet eyes open, two feet eyes closed and one foot (dominant leg) eyes open. The order in which the conditions were presented to the participants was randomised. Each trial lasted 60 seconds and a seated reset period of 50 seconds was provided between the trials. Before the start of a trial, participants were given approximately 10 seconds to allow them to settle into the stance before measurements of sway were taken. For trials in the two feet eyes open condition, participants were required to stand in the position defined at the beginning of the testing session and fixate on the visual target. In the one foot eyes open trials, participants were required to stand on the foot of their dominant leg, with the foot positioned in the centre of the force platform and with the knee of the non-dominant leg flexed to approximately 90? (Frandin et al., 1995). Similarly to the two feet eyes open trials, in the two feet eyes closed condition participants were required to stand in the position defined at the beginning of the testing session, fixate on the target but then slowly close their eyes. During all trials, participants were instructed to stand as still as possible keeping their arms by their sides. In the testing sessions, the scores achieved by participants on Balloon Burst 1 and Balloon Burst 2 were recorded using the balance computer game system. Three attempts of both computer games were given. Each attempt lasted 60 seconds, at the end of which the score was noted (number of balloons popped). A 60 second rest period was provided between each attempt.
2.2.5 Data analysis The position of the centre of pressure of the participants during the trials was calculated using Bioware (Kistler Instrumente AG, Switzerland) and exported for processing in custom written MATLAB (Mathworks, Natick, MA, USA) routines. The anterior-posterior and medio-lateral centre of pressure position data were filtered using a low-pass finite impulse response (FIR) filter with a cut-off frequency of 5 Hz. To quantify the magnitude of movement in the centre of pressure over a trial we analysed the movements in the anterior-posterior and medio-lateral directions separately, as advocated by Winter (1995). Two dependent variables were calculated in both directions: 1) centre of pressure standard deviation and 2) total length of path (path length). The means of these variables over the five trials for each participant in each stance condition were then calculated. Subsequently, a series of 2 x 2 mixed design Analyses of Variance (ANOVAs) were used to determine main effects and interactions for each dependent variable in each stance condition for the following factors: Test (pre vs post exercise programme) and Group (control vs exercise group). Furthermore, differences in mean game scores were assessed using two further two factor (Test, Group) mixed design ANOVAs. All statistical
694 The Engineering of Sport 7 - Vol. 1 tests were performed suing SPSS 13 (SPSS Inc, Chicago, IL, USA) and alpha was set at 0.05. Because of the relatively small sample size in this exploratory pilot study and to determine the magnitude of any differences, we also calculated Ë2 as a measure of effect size.
3- Results Table 2 - Postural sway data.
With the exception of the comparison of anterior-posterior standard deviation of sway in the eyes-open condition, all interactions and main effects in the ANOVAs applied to the postural sway data were not significant (p > 0.05). The test for anteriorposterior sway in the eyes-open condition revealed a significant interaction between the Group and Test factors (F(1,14) = 5.09, p = 0.04, 2 = 0.27).
Results of the ANOVAs applied to the game score data revealed significant interactions between the Group and Test factors for both Balloon Burst 1 (F(1,14) = 37.72, p < 0.001, 2 = 0.73) and Balloon Burst 2 (F(1,14) = 34.43, p < 0.001, 2 = 0.71).
4- Discussion We have designed and developed a relatively low-cost system which enables users to exercise using balance-based computer games. Furthermore, the results of an initial fourweek pilot intervention during which the participants took part in a balance-based exercise intervention using the system have been presented. Significant interactions existed between the Test and Group factors for the game scores on Balloon Burst 1 and Balloon Burst 2. The interactions suggest an improvement in the performance of the exercise group on both games that was not matched by the control group. Obviously, the improvements of the exercise group could be attributed to a learning effect due to the fact that they practised these games (using whole-body weight shift movements to control the games) throughout the intervention; the control group played the games but controlled them using a mouse. However, it is possible that
The Engineering of Sport 7 - Vol. 1
695
the improvement in performance of the exercise group reflects an improvement in dynamic balance function. Further work, using other measures of dynamic balance function, is required to determine whether balance-based computer game play improves dynamic balance. Results for the measures of quiet stance postural sway are much less clear. In the main, there were no significant effects of the balance-based computer game play exercise on measures of postural balance during quiet stance. However, the comparison of anterior-posterior standard deviation of sway, in the two-feet eyes-open condition revealed a significant improvement in the exercise group that was not reflected in the control group. Furthermore, although improvements were only significant for the anterior-posterior standard deviation, there was a trend towards improvement in all dependent variables in the two feet eyes open condition – effect sizes 0.07-0.17 (small to moderate effects). Improvements in postural sway as a result of computer game and biofeedback based exercise have been reported previously (e.g. Betkar et al., 2006; Sihvonen et al., 2004). Interestingly, Sihvonen et al. (2004) did not report differences in the eyes open condition; differences between the exercise and control groups were only apparent in more challenging postural conditions. In comparison, the results of the present study only revealed differences in the eyes open condition. Indeed, for the path length dependent variables, there were trends towards a decrement in performance after the intervention in the exercise group in the single leg stance condition – effect sizes > 0.02-0.19 (small to moderate effects). The reasons for the differences between this investigation and the study of the Sihvonen et al. (2004) and the potential decrease in performance of the exercise group in the single leg stance condition are unclear. However, it should be noted that Sihvonen et al.’s challenging postural condition required participants to stand on a foam surface not on a single leg as included in the present study. A larger-scale, follow-up study is required to further investigate the tentative results related to quiet stance postural sway. The present study has several limitations that need to be acknowledged. First, the small sample size used in this initial pilot study limits the statistical power of the investigation. However, very few controlled trials exist that explore the efficacy of balancebased computer game interventions and the initial findings of this investigation warrant further study. Second, the four week intervention was relatively short and we did not include any follow-up testing in the weeks after the intervention. Third, we included only ‘traditional’ measures of postural balance – standard deviation, path length. There is evidence to suggest that these measures might be less sensitive to changes in balance than more contemporary measures of postural sway – e.g. time-to-boundary (van Wegen et al., 2002). Potentially, the inclusion of more contemporary measures in future studies will reveal more subtle/relevant changes in balance. The aim of this study was to design and develop a cost-effective system for the provision of computer game based balance exercise. An initial four-week pilot intervention with a group of collegiate-aged females suggested potential benefits of such exercise. More work is required to further investigate the benefits of this type of exercise with larger samples from different populations.
696 The Engineering of Sport 7 - Vol. 1
5- References [BS1] Betker, A. L., Szturm, T., Moussavi, Z. K., and Nett, C. Video game-based exercises for balance rehabilitation: A single-subject design. In Archives of Physical Medicine and Rehabilitation, 87(8): 1141-1149, 2006 [CR1] Campbell, A. J., Robertson, M. C., Gardner, M. M., Norton, R. N., Tilyard, M. W., and Buchner, D. M. Randomised controlled trial of a general practice programme of home based exercise to prevent falls in elderly women. In BMJ (Clinical Research Ed.), 315(7115): 1065-1069, 1997 [CR2] Campbell, A. J., Robertson, M. C., La Grow, S. J., Kerse, N. M., Sanderson, G. F., Jacobs, R. J., et al. Randomised controlled trial of prevention of falls in people aged > or =75 with severe visual impairment: The VIP trial. In BMJ (Clinical Research Ed.), 331(7520): 817, 2005 [FS1] Frandin, K., Sonn, U., Svantesson, U., and Grimby, G. Functional balance tests in 76-yearolds in relation to performance, activities of daily living and platform tests. In Scandinavian Journal of Rehabilitation Medicine, 27(4): 231-241, 1995 [LW1] Lord, S. R., Ward, J. A., and Williams, P. Exercise effect on dynamic stability in older women: A randomized controlled trial. In Archives of Physical Medicine and Rehabilitation, 77(3): 232-236, 1996 [MG1] McGuine, T. A., Greene, J. J., Best, T., and Leverson, G. Balance as a predictor of ankle injuries in high school basketball players. Clinical Journal of Sport Medicine. In Journal of the Canadian Academy of Sport Medicine, 10(4): 239-244, 2000 [MK1] McGuine, T. A., and Keene, J. S. The effect of a balance training program on the risk of ankle sprains in high school athletes. In The American Journal of Sports Medicine, 34(7): 11031111, 2006 [PS1] Paillex, R., and So, A. Changes in the standing posture of stroke patients during rehabilitation. In Gait and Posture, 21(4): 403-409, 2005 [PK1] Peurala, S. H., Kononen, P., Pitkanen, K., Sivenius, J., and Tarkka, I. M. Postural instability in patients with chronic stroke. In Restorative Neurology and Neuroscience, 25(2): 101-108, 2007 [RC1] Robertson, S., Collins, J., Elliott, D., and Starkes, J. The influence of skill and intermittent vision on dynamic balance. In Journal of Motor Behavior, 26(4): 333-339, 1994 [SS1] Sihvonen, S. E., Sipila?, S., and Era, P. A. Changes in postural balance in frail elderly women during a 4-week visual feedback training: A randomized controlled trial. In Gerontology, 50(2): 87-95, 2004 [VB1] Verhagen, E., van der Beek, A., Twisk, J., Bouter, L., Bahr, R., and van Mechelen, W. The effect of a proprioceptive balance board training program for the prevention of ankle sprains: A prospective controlled trial. In The American Journal of Sports Medicine, 32(6): 1385-1393, 2004 [WE1] van Wegen, E. E., van Emmerik, R. E., and Riccio, G. E. Postural orientation: Age-related changes in variability and time-to-boundary. In Humun Movement Science, 21(1): 61-84, 2002 [WE2] van Wegen, E. E., van Emmerik, R. E., Wagenaar, R. C., and Ellis, T. Stability boundaries and lateral postural control in parkinson’s disease. In Motor Control, 5(3): 254-269, 2001 [W1] Winter, D. A. A.B.C) of balance during standing and walking. Waterloo, Ont: Graphics Services, 1995 [WL1] Wong, A. M., Lin, Y. C., Chou, S. W., Tang, F. T., and Wong, P. Y. Coordination exercise and postural stability in elderly people: Effect of tai chi chuan. In Archives of Physical Medicine and Rehabilitation, 82(5): 608-612, 2001
Effects of Turbo-jav Release Conditions on Distance of Javelic Throw (P136) M. Maeda1
Topics: Athletics; Biomechanics. Abstract: The turbo-jav is used in the javelic throw as an introduction to the javelin throw or as a technical tool with which to practice the javelin throw. The present study investigated the effects of turbo-jav release conditions on javelic throwing distance. Javelic throws of a turbo-jav by 14 university students were measured over 260 trials. Each turbo-jav throw was videotaped using two high-speed video cameras, and four standard-speed video cameras during flight. All conditions of turbo-jav release and flight were measured using the 3dimensional (3D) DLT method. The results revealed a significant positive correlation between initial velocity and distance thrown using the javelic throw (r = 0.775, p < 0.01). The turbo-jav was thrown far even when the angle of attack was > 25°. The actual distance covered by a thrown turbo-jav was less than the theoretical throwing distance without air. In other words, the flight characteristics indicate that the flight of the turbo-jav in the javelic throw differs from that of a thrown javelin. Keywords: turbo-jav, javelic throw, initial conditions at release, flight characteristics, angle of attack.
1- Introduction The javelic throw is a formal track and field event associated with the javelin throw in the Junior Olympic Games in Japan. A long, narrow polyethylene implement called the turbo-jav (length, 0.7 m; weight, 0.3 kg), resembling a javelin is thrown and athletes compete to achieve maximal distance. The men’s steel or duralumin javelin used in the javelin throw is 2.6 m long and weighs 0.8 kg. Although the javelic throw is a formal event, less in understood about the characteristics of the turbo-jav compared with the javelin. The developer of the turbo-jav is a previous world record holder of the javelin throw, and he seemed to consider that the turbo-jav has similar flight characteristics to the javelin insofar as the turbo-jav and the javelin appear similar. 1. Tsurukabuto Nada Kobe Japan - E-mail: mmaeda @kobe-u.ac.jp
698 The Engineering of Sport 7 - Vol. 1 Although two studies on the javelic throw have been published (Ae et al., 2001; Ohta et al., 2002), they both compare throwing movements of the javelin, and do not address the features of the turbo-jav. Because the javelic throw is an official competition, the characteristic of the implement thrown should be understood in detail. The present study investigated the effects of turbo-jav features and release conditions on javelic throwing distance and clarified the flight characteristic of the turbo-jav during the javelic throw.
2- Methods 2.1 Subjects Table 1 shows the characteristics of the 14 male university students including their experience of javelin throwing.
2.2 Experiment The javelic throws during 260 trials by 14 university students were measured. Each throw was videotaped using two synchronized high-speed video cameras (FASTCAM-PCI, Photron Ltd.; 250fps). Each turbo-jav thrown was videotaped and images of range during flight were obtained using four synchronized standard-speed video cameras (XC009 and DXC-200A, 2 sets each; 60 fps; SONY Co.). Table 1 - Physical characteristics of subjects and their experience in javelin throw.
2.3 Analysis The conditions under which each turbo-jav was released were analyzed using a 3D video motion analysis system (Frame-DIAS II, DKH Ltd.). The tip, center of gravity and edge of each turbo-jav, were digitized, measured using three-dimensional – direct linear transformation (3D-DLT) and the data were smoothed using a digital filter. The flight of each turbo-jav was then similarly measured using 3D-DLT. The x-axis was defined from a point 8 m from the origin of the foul line at the center of the runway in the direction of the throw, the y-axis as the right - left direction of the thrower and the z-axis as the vertical height of the thrower. The actual distance thrown (RA) was the horizontal distance from the point at which the turbo-jav was released to that where it landed, and this measurement was added to the horizontal distance from the release point to the foul line.
The Engineering of Sport 7 - Vol. 1
699
The present study defined the moment of release as that time just before the precise frame in which the turbo-jav left the hand of the thrower. The initial conditions of the turbo-jav, namely initial velocity v, throwing angle , height of release h, attitude angle and angle of attack were calculated at release. These angle parameters were defined as shown in Figure 1. These angles for release of the turbo-jav complied with the definition of Bartlett et al. (1996). Figure 1a defines the throwing angle , attitude angle and angle of attack on the sagittal plane (x-z plane). Figure 1b subsequently defines each angle on the horizontal plane (x-y plane) as the horizontal throwing angle h, horizontal attitude angle h? and horizontal angle of attack h.
Figure 1 - Definition of angle parameters in (a) sagittal and (b) horizontal plane.
The theoretical throwing distance RT (without air) was calculated from that substitutes measured initial velocity v, throwing angle and height of release h from the following equation (Hubbard, 1989).
where RT is theoretical throwing distance, v is initial velocity, is throwing angle, g is gravitational acceleration and h is height of release. The v, and h in this equation corresponded with the measured data in each.
3- Results Figure 2 shows the points at which all the thrown turbo-javs landed. The range thrown varied from about 15 to 50 m. These throws were useful as trials under various conditions of release. Figure 3 shows the relationship between the initial velocity of the released turbo-jav and the actual distance thrown. The relationships were very close with significant correlation (r = 0.775, p < 0.01).
700 The Engineering of Sport 7 - Vol. 1 Figure 4 shows the relationship between the throwing angle of the released turbo-jav and the actual distance thrown. The range was widely distributed between 25 and 50°, and the throwing angle was around 30° when the distance thrown was > 50 m. However, throws of over 40 m were achieved despite an increase in the throwing angle to between 40 and 50°. Figure 5a shows that the relationship between the angle of attack of the released turbo-jav and the actual distance thrown fluctuated widely in the range of -5 to 35°. The trial found that throwing distance tended to increase with a decreasing angle of attack. This trial found that the turbo-jav could be thrown over 40 m despite the larger attack angle of 25 to 35°. Figure 5b shows that the relationship between the horizontal angle of attack of the released turbo-jav and the actual distance thrown did not significantly correlate. Several trials indicated a long throw despite the horizontal angle of attack being not quite 0°. Figure 6 shows the relationship between the theoretical (without air) and the actual distance thrown. The dotted line indicates that the actual and theoretical throwing distances were quite similar. Most trials in the present study remained below the dotted line. Figure 7 shows examples of turbo-jav flight with an excess angle of attack or throwing angle. These distances reached approximately 40 m despite the angle of attack being > 30° or the throwing angle being excessive at about 40°.
Figure 2 - Landing points of turbo-jav during trials.
Figure 3 - Relationship between initial velocity and distance thrown.
The Engineering of Sport 7 - Vol. 1
Figure 4 - Relationship between throwing angle and distance thrown.
Figure 5a - Relationship between angle of attack and distance thrown.
Figure 5b - Relationship between horizontal angle of attack and distance thrown.
701
702 The Engineering of Sport 7 - Vol. 1
Figure 6 - Relationship between theoretical distance thrown and actual distance thrown.
Figure 7 - Examples of turbo-jav in flight with excessive angle of attack.
4- Discussion Tom Petranoff, a previous world record holder for the javelin throw, developed the turbo-jav as a means of practicing the javelin throw. Therefore, the features of the turbojav during the javelic throw and of the javelin during the javelin throw were apparently considered equal. The present study found that the distance thrown was associated with an initially higher initial velocity of the turbo-jav, which is similar to the findings of others: r = 0.97 (Komi and Mero, 1985) and r = 0.80 (Murakami and Ito, 2003), indicating that the relationship between initial velocity and distance thrown is similar in the javelic throw between the turbo-jav and the javelin. After all, the initial velocity affected javelic throwing distance using the turbo-jav almost as well as in the javelin throw.
The Engineering of Sport 7 - Vol. 1
703
Figure 4 shows that the throwing angle of turbo-jav ranged from about 25 to 50°. In the javelin throw, Murakami and Ito (2003) reported that this is distributed between about 25 to 40° and Wakayama et al. (1994) reported a distribution from about 30 to 45°. As compared with similar data from the javelin throw, the throwing angles in the present study were distributed more widely, since half of our participants were unskilled. Wakayama et al. (1994) reported that the optimum throwing angle of the javelin throw is 35° at the 60-m skill level, and 33° at that of 30 m. That is, a lower skill level is associated with a lower optimum throwing angle. The optimum throwing angle cannot be applied to the javelic throw, because if the characteristics of the turbo-jav and the javelin are equal, the thrower cannot throw further by increasing the throwing angle during the javelic throw. The present study recognized that some throws reached over 40 meters even though the throwing angle was excessive at 40 - 50°. That is to say, the excessive throwing angle in the javelic throw might not be the cause of a shorter throw. Maeda et al. (1996) reported that a slightly positive or negative excess angle of attack does not increase the thrown distance even in the javelin throw; that is, an angle as close as possible to 0° is desirable. The angles of attack in the present study ranged from about -5 to 35° and those who attempted to decrease the angle of attack threw the turbo-jav further. These results agree with those of others, whereas distance thrown always increased even when the angle of attack, the horizontal angle of attack and the throwing angle were large. That is, the turbo-jav differs from the javelin insofar as it could be thrown farther even when the angle of attack or throwing angle was excessive. Either of these conditions causes a stall in the javelin throw, which results in no gains in distance (Terauds, 1985). An excessive throwing angle or angle of attack during the javelic throw does not particularly affect the distance thrown. Figure 6 indicates that the actual distance travelled by the thrown turbo-jav did not significantly increase compared with the theoretical distance based on the initial conditions at release. In contrast, the actual distance achieved by the javelin throw tends to be farther than the theoretical distance in the absence of the air (Maeda, 1996). That is to say, the turbo-jav does not confer a significant benefit compared with the javelin. The flight characteristics of the turbo-jav in the javelic throw differ from those of the javelin in the javelin throw. Although the turbo-jav was introduced as a technical tool with which to practice the javelin throw, the increased distance achieved in the javelic throw would not always result in throwing the javelin further.
5- Conclusion The present study examined the influence of the initial release conditions on the distance travelled by a turbo-jav in the javelic throw and clarified the flight characteristics of the turbo-jav. Fourteen experienced and inexperienced individuals performed the javelic throw and the distance thrown was measured. The initial conditions at the moment of release were recorded during 260 turbo-jav throws by 14 throwers using two high-speed video cameras, and information about the flight conditions of the turbo-jav was obtained using four normal-speed video cameras after release. The recordings analyzed using 3D-DLT revealed the following results.
704 The Engineering of Sport 7 - Vol. 1 The correlation between the initial velocity and distance travelled during the javelic throw was close and positive, and the initial velocity was the critical influence on distance thrown. Since the flight characteristics differ between the turbo-jav and the javelin, the turbojav was thrown farther even when the angle of attack or throwing angle was comparatively large. The actual distance thrown was often shorter than the theoretical distance thrown during the javelic throw, suggesting that the turbo-jav does not confer a significant benefit on the distance thrown compared with the javelin. The turbo-jav should be regarded as having different flight characteristics from the javelin.
6- References [AS1] Ae, M., Shimada, K., Enomoto, Y. and Yokozawa, T. Comparison of the throwing motions of turbo-jav and javelin. In Research Quarterly for Athletics, (46): 16-24, 2001. (in Japanese) [BM1] Bartlett, R.M., Müller, E., Lindinger, S., Brunner, F. and Morriss, C.J. Three-dimensional evaluation of the kinematic release parameters for javelin thrower of different skill levels. In Journal of Applied Biomechanics, 12: 58-71, 1996. [H1] Hubbard, M. The throwing events in track and field. In Biomechanics of Sport. Vaughan, C. L. Ed., CRC Press, Florida: pp.213-238, 1989. [KM1] Komi, P.V. and Mero, A. Biomechanical analysis of Olympic javelin thrower. In International Journal of Sport Biomechanics, 1: 139-150, 1985. [M1] Maeda, M. Flying behavior of javelin in javelin throw. In Japanese Journal of Sports Sciences, 15(3): 207-213, 1996. (in Japanese) [MN1] Maeda, M., Nomura, H., Yanagida, Y. and Miyagaki, M. Optimum release conditions in javelin throwing considering human movement. In DESCENTE SPORTS SCIENCE, (17): 270277, 1996. (in Japanese) [MI1] Murakami, M. and Ito, A. Relationship between the performance and the throw is moment in javelin throwing. In Japanese Journal of Biomechanics in Sports & Exercise, 7(2): 92-100, 2003. (in Japanese) [OA1] Ohta, K., Ae, M. and Yokozawa, T. Effectiveness of turbo-jav as a training tool for the improvement of javelin throw technique. In Research Quarterly for Athletics, (50): 13-20, 2002. (in Japanese) [WT1] Wakayama, A., Tazuke, S., Kojima, T., Ikegami, Y., Sakurai, S., Okamoto, A., Ueya, K., and Nakamura, K. Biomechanical analysis of javelin throw. In: Scientific Report on the 3rd World Championships in Athletics Tokyo 1991. Japan Association of Athletics Federations. Baseball Magazine Ltd., Tokyo: pp.220-238, 1994. (in Japanese) [T1] Terauds, J. Biomechanics of the javelin throw, Academic Publishers, California: 1985.
Differences Between Leather and Sybthetic NBA Basketballs (P137) Hiroki Okubo1, Mont Hubbard2
Topics: Basketball, Handball & Volleyball. Abstract: Leather and synthetic basketballs of National Basketball Associations are compared in terms of their coefficients of friction between the balls and the rim and backboard, and their coefficients of restitution. We estimate the value of coefficient of friction between the balls and rim and backboard using a static model and incipient slip measurements. The friction coefficient for the synthetic balls is much larger than that of leather ones when dry. The coefficient of restitution is measured in an experiment in which the leather balls bounced higher than the synthetic balls inflated to the same pressure. Finally we investigate the effects of these differences between the balls for capture conditions using our dynamic model. Keywords: leather basketball, micro fiber composite basketball, coefficient of friction, coefficient of restitution.
1- Introduction The National Basketball Association (NBA) tried the new micro fiber composite basketball for the 2006-07 season. However, the old leather ball was restored on Jan. 1, 2007. Players complained that the synthetic ball bounced differently than the leather ball off the floor and the rim and that the synthetic material cut their fingers. The coefficient of restitution and coefficients of friction between the basketball and the rim and backboard are vital to play. In basketball simulations, real parameters are essential. A few studies have measured basketball parameters. Okubo and Hubbard (2004) measured the radial ball stiffness and an equivalent radial damping coefficient in an MTS test. They also showed the damping coefficient as function of impact velocity. Cross (2003) discovered that basketballs demonstrate tangential and radial compliance at the contact point. Okubo and Hubbard (2006) measured the coefficient of friction between a basketball and an acrylic board. Weiss (2006) reported that Horwitz and De at the University of Texas, Arlington did tests in which they showed that the NBA synthetic 1. Chiba Institute of Technology - E-mail: [email protected] 2. University of California, Davis - E-mail: [email protected]
706 The Engineering of Sport 7 - Vol. 1 balls bounced less elastically and resisted sliding much more strongly when dry on sheets of silicon, and that the leather balls are much easier to grip when wet. To our knowledge no study has measured the coefficient of friction between a basketball and rim. In this paper, we derive a formula calculating the coefficient by analyzing a model of two taped basketballs supported by two toroidal rims. We compare the basketball-backboard friction for the two types of balls. In impact tests the coefficients of restitution are also calculated for the leather and synthetic balls. Finally we investigate the effects of differences between balls for capture using our dynamic model.
2- Basketball equipment 2.1 Basketball The range of proper inflation pressures 7-9 lb/in2 is stamped on official basketballs. FIBA rules specify that the ball must rebound to a height, measured to the top of the ball, of between 1.2 m and 1.4 m when dropped onto the playing surface from a height of 1.8 m, measured from the bottom of the ball. We used three leather and synthetic balls (Fig. 1) for our tests and the parameters measured are shown in Table 1.
2.2 Rim and backboard Basketball hoops are made from baked, powder-coated solid steel with an inside major diameter of 45 cm and minor diameter of 1.6-2.0 cm. FIBA rules require that backboards must be made of a suitable transparent material (tempered safety glass for Level 1 and 2). We used an official rim (Mizuno 13ZA-200) and a poly carbonate backboard (Spalding 74307) in our experiments. Table 1 - NBA official basketballs.
Figure 1 - NBA official basketballs (a) leather and (b) synthetic types.
The Engineering of Sport 7 - Vol. 1
707
3- Coefficient of friction between basketball and rim 3.1 Theoretical model The geometry of the two taped basketballs (BL andBR) and two rims system is shown in Fig. 2. The Newtonian XYZ frame with origin at the middle of the hoop centre has XY axes horizontal and Z axis vertical. BO is the contact point between BL and BR. BL* and BR* denote the centres of BL and BR, respectively. is the angle between the horizontal plane and the line from O to BO. b is the angle between the lines OBO and OBL*, or OBR* . The tilt angle characterizes the deviation of the contact points into the intra-rim region. The equilibrium equations are derived as (1) (2) (3) where FL2 and FR2 are the normal reaction forces at the ball-rim contact point. FL3 and FR3 are the friction forces in the tangential plane to both the rim and ball surfaces and perpendicular to the sections AL – AL and AR – AR, respectively. When the taped balls is at the angle on the boundary between slip and stick, the forces satisfy relationships FL3 = μFL2 and FR3 = μFR2, from which Eqs (1)-(3) give the coefficient of friction (4) where A = cos sinb(sinb – cos), B = sin cos(1 – sinbcos), and C = –cos cos2bcos2
Figure 2 - Front, side, and section views of two taped basketballs supported by two vertical rims.
708 The Engineering of Sport 7 - Vol. 1
3.2 Measurement We calculate the coefficient of friction for the leather and synthetic balls as a function of using Eq. (4) for a range of distances between the two rims with w=0.05-0.15 m in Fig. 3(a). For larger , a more linear relationship exists between and the coefficient of friction μ. Figure 3(b) shows the measurement results of the coefficient of friction for the two kinds of balls. The coefficient of friction for the synthetic balls is much larger than that of leather ones. We average 0.51 for the leather and 1.20 for synthetic of the all numbers of the results, therefore estimate the coefficients to be μ = 0.5 for leather and μ = 1.2 for the synthetic. The error bars of the synthetic balls are longer than those for leather balls because the measurement is more sensitive at smaller angles of (Fig. 4) where a stronger relationship exists between μ and .
Figure 3 - Coefficient of friction μ. (a) as a function of critical slip angle for 5 values of separation distance w between the rims. (b) measurements for the two types of balls. μ = 0.5 for leather, and 1.2 for synthetic balls are estimated.
Figure 4 - Basketballs slip on the rims when they experience maximum possible friction force. Critical (minimum) inclination angles are approximately (a) = 68 deg and (b) = 54 deg for the leather and synthetic balls, respectively, with rims separated by w = 0.1 m.
The Engineering of Sport 7 - Vol. 1
709
4- Coefficient of friction between basketball and backboard The coefficient of friction between basketball and backboard μBo was determined by two methods. One is similar to Sawicki, Hubbard and Stronge (2003) who measured friction between a baseball and bat using a variant of the inclined plane experiment. Three basketballs, taped together to prevent rolling, were set on a backboard. A two-dimensional force balance at incipient slip shows that μBo is given by μBo = tan , where is the tilt of the board from horizontal. The angle was 34-35 deg for the leather ball, from which the static coefficient of friction is calculated to be μBo = 0.7. However, this was not successful for synthetic balls because they did not slip on the inclined board before rolling over. Nevertheless this provides a lower bound μBo >1.15. For this reason we measured the friction force between the three taped balls and a horizontal backboard using a spring balance. The force for the leather is approximately 1.2 kg, and for the synthetic 3.1-3.3 kg, from which coefficients of friction for the leather and synthetic balls were calculated as μBo = 0.7 and 1.8, respectively.
(a) (b) Figure 5 - Rebound height vs. initial height of (a) leather and (b) synthetic basketballs.
Figure 6 - Coefficient of restitution for leather and synthetic basketballs.
710 The Engineering of Sport 7 - Vol. 1
5- Coefficient of restitution Figures 5(a) and 5(b) show rebound height as function of initial drop height for both the leather and synthetic basketballs with 7-9 lb/in2 pressure. We approximate the rebound-initial height relation by a straight line in the range of 0.6-2.0 m initial height. Leather balls rebound higher than synthetic balls do. The coefficients of restitution of the two ball types are shown in Fig. 6 as a function of the velocity at ball-floor contact for three proper inflation pressures. The coefficient decreases strongly with impact velocity.
6- Discussion How do these measured differences affect performance? To learn this, we compare capture conditions as a function of release angle and velocity for leather and synthetic balls using our dynamic model (Okubo and Hubbard 2006). The model has six distinct sub-models: gravitational flight with air drag, and sub-models of ball-rim, ball-board, ball-bridge, ball-rim-board, and ball-bridge-board contact. Each contact sub-model has possible both slipping and non-slipping motions. With the same stiffness (42,000 N/m) damping coefficients of 32 Ns/m and 34 Ns/m, for leather and synthetic respectively, resulted from the dynamic model.
Figure 7 - (a) Capture conditions in release velocity and angle space for shots in a board normal plane including the hoop centre with 2 rev/s backspin and the release point 3m away from and 0.2 m below the hoop centre using leather and synthetic balls. (b) Examples of ball centre paths for shots in the normal plane with the same release conditions. The leather ball escapes after bouncing off the rim. But the synthetic ball is captured after bouncing off the rim and backboard.
The Engineering of Sport 7 - Vol. 1
711
Figure 8 - Capture combinations of release velocity and angle for angled (a) direct and (b) bank shots with 2 rev/s backspin from 3 m away from and 0.2 m below the hoop centre, and with floor angle of 30 deg.
Figures 7(a) shows capture combinations (darkened) for release in a vertical plane including the hoop centre and normal to the backboard, with the release point 3 m away from and 0.2 m below the hoop centre and 2 rev/s backspin angular velocity. The set of capture combinations for synthetic balls is slightly larger than that for leather. Especially, the ball trajectories are quite different in a many-bounce shot as shown in Fig. 7(b). The synthetic ball is captured after bouncing off the rim and board, but the leather one escapes. We also calculated successful release combinations for angled direct and bank shots using leather and synthetic balls as shown in Figs 8(a) and (b). The shot is from a release point 3 m away from and 0.2 m below the hoop centre, and has 2 rev/s backspin and a floor angle of 30 deg between the board surface and the horizontal projection of the line from the hoop centre to the release point. There is almost no difference between the leather and synthetic balls in their capture conditions for these shots.
7- Conclusions The coefficients of friction between the rim and the NBA leather and synthetic balls have been measured using a static model and incipient slip tests. The coefficients were μ = 0.5 and μ = 1.2 for leather and synthetic, respectively. Friction coefficients between balls and the backboard was also measured and are estimated to be μBo = 0.7 and μBo = 1.8 for leather and synthetic balls, respectively. In general, synthetic balls have much larger coefficients of friction than do leather balls when dry. We also confirmed that the coefficient of restitution for the leather balls is larger than that for synthetic ones. The set of capture combination for angled direct and bank shots are nearly the same but the balls behave differently in many-bounce shots.
8- References [C1] Cross R. Grip-slip behaviour of a bouncing ball. In American Journal of Physics, 70: 10931102, 2003.
712 The Engineering of Sport 7 - Vol. 1 [OH1] Okubo H. and Hubbard M. Dynamics of basketball-rim interactions. In Sports Engineering, 7(1) : 15-29, 2004 [OH2] Okubo H. and Hubbard M. Dynamics of the basketball shot with application to the free throw. In Journal of Sports Science, 24: 1304-1314, 2006 [SH1] Sawicki G.S., Hubbard M. and Stronge W.J. How to hit home runs: Optimum baseball bat swing parameters for maximum range trajectories. In American Journal of Physics, 71: 11521162, 2004 [W1] Weiss P. Dribble Quibble: Experiments find that new basketball gets slick. In Science News Online, 170(19), 2006
Subject Index ORAL COMMUNICATION : gras POSTER : maigre
SPORTS Acrobatics Athletics Badminton Basketball, Handball & Volley Baseball Bicycle Climbing Extreme Sports Fishing Fitness Golf Half-pipe Judo & Combat sports Lawn Sports (Hockey, Cricket) Outdoor Sports Rugby Running Sailing/water Sports Skate & other Urban Sports Ski & other Winter Sports
P3, P86, P95, P140, P171, P218, P237, P248 P27, P37, P117, P136, P151, P156, P179, P188, P191, P196, P208 P254 P137, P213 P34, P103, P138, P138, P181, P214, P234, P273, P274, P275 P39, P49, P51, P68, P76, P80, P81, P82, P83, P85, P112, P114, P131,P207, P226, P242, P247, P255, P267 P6, P31, P97, P142 P84, P252 P33, P57 P27, P176 P5, P11, P35, P79, P90, P128, P143, P147, P271 P237, P240 P43, P253 P9, P20, P26, P31, P70, P84, P88, P145, P178, P185, P77, P125, P198, P235 P7, P30, P41, P67, P125, P278 P52, P96, P145, P152, P189, P190, P193, P196, P250, P10, P15, P17, P18, P19, P56, P71, P127, P149, P174, P203, P215, P215, P244, P277 P24 P3, P12, P48, P65, P86, P95, P111, P119, P140,P153, P160, P162, P163, P171, P194, P212, P225, P228, P237, P239, P240, P245, P251, P243, P268, P269
714 The Engineering of Sport 7 - Vol. 1 Soccer Surf & other Sliding Sports Swimming Tennis & other Rackets Sports
P44, P45, P96, P106, P115, P125, P172, P186, P217 P56, P261 P10, P15, P17, P18, P56, P203, P277 P21, P22, P109, P110, P126, P172, P175
GENERIC Aerodynamics Anthropometry Apparel Biomechanics
Composite Finite elements analysis Handicap Industrial design Innovation & Design Instrumentation Kinematics Management Materials
Mechanical Engineering Medical Measurement Systems
Methodology
P3, P67, P68, P70, P76, P86, P106, P114, P117, P138, P140, P171, P172, P212, P214, P226, P234 P242 P30, P43, P72, P203, P240 P7, P18, P19, P22, P26, P27, P39, P37, P48, P52, P56, P57, P77, P80, P94, P95, P96, P99, P102, P105, P115, P116, P118, P119, P124, P127, P135, P136, P142, P145, P166, P174, P188, P190, P191, P192, P193, P208, P217, P218, P222, P228, P235, P241, P250, P260, P264 , P268, P276 P31, P137, P150, P274 P5, P21, P22, P35, P90, P97, P90, P175, P186, P245, P251, P271 P28, P62, P241, P277 P7, P12, P28, P30, P34, P62, P65, P143, P150, P162, P174, P178, P181, P186, P244, P245, P247, P261, P262, P30, P52, P179, P183, P145, P174, P215, P244, P247, P253, P261, P262 P27, P42, P52, P82, P135, P176, P216 P6, P57, P90, P94, P96, P189, P190, P217, P218, P235, P243, P276 P116, P183, P240, P261, P262 P5, P28, P48, P62, P68, P72, P88, P96, P105, P119, P124, P125, P128, P156, P137, P150, P152, P163, P168, P179, P193, P208, P213, P226, P239, P240, P241, P243, P248, P268 P20, P24, P30, P41, P70, P112, P156, P175, P178, P185, P194, P212, P215, P245, P275, P278 P31, P72, P105, P124, P208, P222, P231 P10, P17, P37, P42, P43,, P44, P48, P49, P51, P52, P56, P80 , P81, P84, P85, P111, P142, P156, P162, P184, P191, P194, P196, P216, P222, P228, P231, P239, P241, P242, P244, P245, P247, P251, P255, P268, P269 P45, P179, P183, P253
The Engineering of Sport 7 - Vol. 1
Modelling
Paralympics Performance Sports
Re-education, Rehabilitation, Prevention & Health Safety Shoes Social Sciences Testing, Prototyping, Benchmarking
Virtual Reality & Computer application in Sports
715
P15, P18, P19, P24, P57, P71, P79, P83, P90, P97, P103, P128, P131, P149, P147, P151, P166, P174, P193, P203, P207, P239, P248, P255, P254, P264, P267, P268, P276, P277 P277 P3, P5, P10, P11, P12, P17, P19, P26, P34, P48, P56, P76, P79, P81, P83, P85, P88, P94, P111, P112, P116, P145, P153, P160, P163, P171, P174, P179, P212, P218, P222, P225, P228, P235, P237, P239, P245, P251, P255, P268, P273, P274, P275, P277 P192, P264 P9, P181, P268 P48, P52, P145, P152, P179, P193 P99 P7, P12, P34, P45, P48, P65, P80, P88, P109, P110, P115, P125, P126, P149, P147, P152, P162, P178, P181, P188, P191, P194, P212, P213, P215, P245, P260, P278 P18, P27, P33, P42, P71, P81, P82, P86, P95, P102, P110, P135, P148, P149, P176, P184, P253