Noise and Vibration Mitigation for Rail Transportation Systems: Proceedings of the 9th International Workshop on Railway Noise, Munich, Germany, 4 - 8 ... Mechanics and Multidisciplinary Design)

99

Notes on Numerical Fluid Mechanics and Multidisciplinary Design (NNFM)

Editors E.H. Hirschel/München W. Schröder/Aachen K. Fujii/Kanagawa W. Haase/München B. van Leer/Ann Arbor M.A. Leschziner/London M. Pandolfi/Torino J. Periaux/Paris A. Rizzi/Stockholm B. Roux/Marseille Y. Shokin/Novosibirsk

Noise and Vibration Mitigation for Rail Transportation Systems Proceedings of the 9th International Workshop on Railway Noise, Munich, Germany, 4–8 September 2007 Burkhard Schulte-Werning David Thompson Pierre-Etienne Gautier Carl Hanson Brian Hemsworth James Nelson Tatsuo Maeda Paul de Vos (Editors)

ABC

Dr. Burkhard Schulte-Werning Deutsche Bahn AG Systemverbund Bahn DB Systemtechnik Völckerstr. 5 80939 München Germany E-mail: [email protected] Prof. Dr. David Thompson ISVR, University of Southampton Highfield Southampton SO17 1BJ United Kingdom E-mail: [email protected] Dr. Pierre-Etienne Gautier SNCF Research & Technology 45 rue de Londres 75379 Paris cedex 08 France E-mail: [email protected] Dr. Carl Hanson Harris Miller Miller & Hanson Inc. 15 New England Executive Park Burlington, MA 01803 USA E-mail: [email protected]

ISBN 978-3-540-74892-2

Brian Hemsworth B.Sc., CEng., FIOA Brian Hemsworth Noise Consultant LLP 16 Whistlestop Close Mickleover Derby DE3 9DA United Kingdom E-mail: [email protected] Dr. James Nelson Wilson, Ihrig & Assoc., Inc. 5776 Broadway Oakland, CA USA 94618-1531 E-mail: [email protected] Dr. Tatsuo Maeda Railway Technical Research Institute 2-8-38 Hikari-cho Kokubunji-shi Tokyo 185 8540 Japan E-mail: [email protected] Paul de Vos M.Sc. DHV BV, Environment and Transportation P.O. Box 1132, NL 3800 BC Amersfoort The Netherlands E-mail: [email protected]

e-ISBN 978-3-540-74893-9

DOI 10.1007/978-3-540-74893-9 Notes on Numerical Fluid Mechanics and Multidisciplinary Design

ISSN 1612-2909

Library of Congress Control Number: 2008922523 c 2008 

Springer-Verlag Berlin Heidelberg

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable for prosecution under the German Copyright Law. The use of general descriptive names, 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 protective laws and regulations and therefore free for general use. Typesetting Scientific Publishing Services Pvt. Ltd., Chennai, India. Printed in acid-free paper 543210 springer.com

NNFM Editor Addresses

Prof. Dr. Ernst Heinrich Hirschel (General Editor) Herzog-Heinrich-Weg 6 D-85604 Zorneding Germany E-mail: [email protected] Prof. Dr. Wolfgang Schröder (Designated General Editor) RWTH Aachen Lehrstuhl für Strömungslehre und Aerodynamisches Institut Wüllnerstr. zw. 5 u. 7 52062 Aachen Germany E-mail: [email protected] Prof. Dr. Kozo Fujii Space Transportation Research Division The Institute of Space and Astronautical Science 3-1-1, Yoshinodai, Sagamihara Kanagawa, 229-8510 Japan E-mail: [email protected] Dr. Werner Haase Höhenkirchener Str. 19d D-85662 Hohenbrunn Germany E-mail: [email protected] Prof. Dr. Bram van Leer Department of Aerospace Engineering The University of Michigan Ann Arbor, MI 48109-2140 USA E-mail: [email protected] Prof. Dr. Michael A. Leschziner Imperial College of Science Technology and Medicine Aeronautics Department Prince Consort Road London SW7 2BY U.K. E-mail: [email protected]

Prof. Dr. Maurizio Pandolfi Politecnico di Torino Dipartimento di Ingegneria Aeronautica e Spaziale Corso Duca degli Abruzzi, 24 I-10129 Torino Italy E-mail: [email protected] Prof. Dr. Jacques Periaux 38, Boulevard de Reuilly F-75012 Paris France E-mail: [email protected] Prof. Dr. Arthur Rizzi Department of Aeronautics KTH Royal Institute of Technology Teknikringen 8 S-10044 Stockholm Sweden E-mail: [email protected] Dr. Bernard Roux L3M – IMT La Jetée Technopole de Chateau-Gombert F-13451 Marseille Cedex 20 France E-mail: [email protected] Prof. Dr. Yurii I. Shokin Siberian Branch of the Russian Academy of Sciences Institute of Computational Technologies Ac. Lavrentyeva Ave. 6 630090 Novosibirsk Russia E-mail: [email protected]

Preface

This book contains the presentations given during the 9th International Workshop on Railway Noise (IWRN9) which took place in Munich/Feldafing, Germany, on 4th to 8th September 2007. This workshop was organised by the Acoustics and Vibration Department of DB Systemtechnik, the technical engineering office of Deutsche Bahn AG. More than 120 participants from 17 countries followed the invitation to the workshop. This great response showed the continuing interest in an important topic of railway technology and offered the opportunity to present the recent results of intense worldwide activities to the international community of railway noise and vibration experts and to share knowledge as well as experience. Because an efficient transportation network is indispensable to handle the general mobility increase and road networks have reached their socio-ecological limits, the railway network is to be strengthened. For example the European Commission has given distinct political signals to get more passengers onto the railways. This policy represents a clear challenge for the next few decades not only for European railway companies: the considerable increase in mobility will lead to a doubling of the railway traffic volume within the next 10 to 20 years. To reduce the environmental impact, the Directive on the Assessment and Management of Environmental Noise has been put into force in Europe, aiming at avoiding, preventing or reducing harmful effects of environmental noise on human health. This directive requires member states to produce strategic noise maps by using noise indicators assessing the number of people affected by noise, to inform the public about noise exposure, and to draw up action plans to reduce noise where necessary. These action plans will also tackle railway noise. Responses to these subjects simultaneously involve the variety of rolling stock vehicles, infrastructure and also operating conditions. Since often efficient and economically satisfying solutions to railway noise problems cannot be found on one single component of the system alone, they need to be studied by considering the railway system with its interdependencies. The IWRN9 contributions give state-of-the-art answers to such questions and address general rolling noise aspects, new noise reduction technologies, prediction tools and theoretical models, high-speed trains, ground-borne vibrations, cost-benefit considerations of noise abatement as well as rail grinding, corrugation and roughness. Following the tradition of the previous workshops, IWRN9 was held as a single session event with the aim of providing the optimistic atmosphere for informal and inspired exchange of information on all facets of railway noise and vibration mitigation. Over three and a half days, 64 papers were presented in 13 sessions and in a poster session additional 11 presentations were subject of lively discussion. There is no formal organisation behind the IWRN but rather an informal, committed International Committee. It supports the chairman during the preparation process with the experience and expertise of its members. Assistance is given to formulate the

VIII

Preface

scientific programme, to release the Call for Papers, to perform the paper selection process, to act as session chairmen at the IWRN9 workshop and to act as a peer review group for the IWRN9 proceedings. Special thanks are due to Andrea Sahner, Gisela Rothermel, Melanie Payer, Wolfgang Behr and Alfred Hechenberger of the local committee for all the hard work and care in organising the conference. The editors are grateful to Prof. E.H. Hirschel as the general editor of the “Notes on Numerical Fluid Mechanics and Multidisciplinary Design” and also to the staff of the Springer Verlag for the opportunity to publish the proceedings of the IWRN9 workshop in this series. We look forward to this volume being used as a “state-of-the-art” reference by scientists and engineers involved in solving noise and vibration problems related to railway traffic in the years to come.

December 2007

Burkhard Schulte-Werning David Thompson Pierre-Etienne Gautier Carl Hanson Brian Hemsworth James Nelson Tatsuo Maeda Paul de Vos

Table of Contents

Session 1: High Speed Trains (I) Environmental Noise Reduction of Tokaido Shinkansen and Future Prospect H. Kanda, H. Tsuda, K. Ichikawa, S. Yoshida . . . . . . . . . . . . . . . . . . . . . . . .

1

Distortion of Compression Wave Propagating through Shinkansen Tunnel T. Miyachi, T. Fukuda, M. Iida, T. Maeda, S. Ozawa . . . . . . . . . . . . . . . . .

9

The Influence of the Train Speed on Vibrations Due to High Speed Trains G. Lombaert, G. Degrande, J. Bekaert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19

High Speed Train Noise Effects on Wildlife and Domestic Livestock C.E. Hanson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

26

Wind Tunnel Tests on the Control of Aeroacoustic Noise from High Speed Train N. Yamazaki, T. Takaishi, M. Toyooka, K. Nagakura, A. Sagawa, H. Yano . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33

Session 2: High Speed Trains (II) Measures to Counteract Micro-pressure Waves Radiating from Tunnel Exits of DB’s New Nuremberg-Ingolstadt High-Speed Line Th. Tielkes, H.-J. Kaltenbach, M. Hieke, P. Deeg, M. Eisenlauer . . . . . . .

40

Acoustic Assessment of Micro-pressure Waves Radiating from Tunnel Exits of DB High-Speed Lines K.G. Degen, Ch. Gerbig, H. Onnich . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

48

High Speed Railway Noise: Assessment of Mitigation Measures F. L´etourneaux, J.F. Cordier, F. Poisson, N. Douarche . . . . . . . . . . . . . . . .

56

Noise Measurement Results of Shinkansen High-Speed Test Train (FASTECH360S,Z) Y. Wakabayashi, T. Kurita, H. Yamada, M. Horiuchi . . . . . . . . . . . . . . . . .

63

X

Table of Contents

Noise Sources for High Speed Trains: A Review of Results in the TGV Case F. Poisson, P.E. Gautier, F. Letourneaux . . . . . . . . . . . . . . . . . . . . . . . . . . . .

71

Session 3: Ground Borne Vibrations (I) Survey of Metro Excitation Frequencies and Coincidence of Different Modes S.J. Cox, A. Wang, A. Adedipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

78

Floating Slab Track above Ground for Turnouts in Tram Lines H.-G. Wagner, A. Herrmann . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

86

Vehicle/Track Impact Due to Passing the Transition between a Floating Slab and Ballasted Track Z.G. Li, T.X. Wu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

94

Recent Developments in Operational Rail Noise and Vibration in NSW, Australia D. Anderson, C. Weber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Experimental Validation of a Numerical Model for Subway Induced Vibrations S. Gupta, G. Degrande, G. Lombaert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 A Numerical Model for Re-radiated Noise in Buildings from Underground Railways P. Fiala, S. Gupta, G. Degrande, F. Augusztinovicz . . . . . . . . . . . . . . . . . . . 115

Session 4: Ground Borne Vibrations (II) The Influence of the Soil on Track Dynamics and Ground-Borne Vibration L. Auersch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 A User-Friendly Prediction Tool for Railway Induced Ground Vibrations: Emission – Transmission – Immission W. R¨ ucker, L. Auersch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 Using the PiP Model for Fast Calculation of Vibration from a Railway Tunnel in a Multi-layered Half-Space M.F.M. Hussein, H.E.M. Hunt, L. Rikse, S. Gupta, G. Degrande, J.P. Talbot, S. Fran¸cois, M. Schevenels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 Structure-Borne Noise and Vibration Control for Chatswood Interchange J.T. Nelson, M. Harrisson, M. Pettersson . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Table of Contents

XI

Measurements and Investigations at the Floating-Track-Bed System in the North-South Tunnel in Berlin T. Jaquet, R. Garburg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 Propagation of Vibrations Due to a Tramway Line M. Maldonado, O. Chiello, D. Le Hou´edec . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

Session 5: General Rolling Noise Aspects (I) Railway Noise Statistics by Monitoring Stations – Input for Dutch Prediction Method RMR and Track Access Charging E. Verheijen, M.S. Roovers, J.W. van den Brink . . . . . . . . . . . . . . . . . . . . . . 165 Measurement and Modelling of Noise from the Arsta Bridge in Stockholm A. Wang, O.G. Bewes, S.J. Cox, C.J.C. Jones . . . . . . . . . . . . . . . . . . . . . . . 172 Minimising Noise from Viaducts in the Borough Area of London for the Thameslink Programme C. Cobbing, C.J.C. Jones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 The New German Prediction Model for Railway Noise “Schall 03 2006” – Potentials of the New Calculation Method for Noise Mitigation of Planned Rail Traffic U. Moehler, M. Liepert, U.J. Kurze, H. Onnich . . . . . . . . . . . . . . . . . . . . . . . 186 Floating Slab Track Re-engineering: Experience Drawn from a Completely Renovated FST Damaged by Major Flooding in Sao Paulo Metro P. Carels, K. Ophalffens, P. Pinto, R. Kelly. . . . . . . . . . . . . . . . . . . . . . . . . . 193

Session 6: General Rolling Noise Aspects (II) In-Car Noise and Carriage Floor Vibration on Different Track Forms and Curvatures in a Metro System A. Wang, S.J. Cox, H. Huang, L. Liu, J. Jiang, J. Sun . . . . . . . . . . . . . . . . 201 Experimental and Theoretical Analysis of Railway Bridge Noise Reduction Using Resilient Rail Fasteners in Burgdorf, Switzerland K.P. K¨ ostli, C.J.C. Jones, D.J. Thompson . . . . . . . . . . . . . . . . . . . . . . . . . . . 208 Comparison of Two Metrics for Assessing Human Response to Vibration R. Carman, C. Reyes, G. Glickman, M. Schaeffler . . . . . . . . . . . . . . . . . . . . 215

XII

Table of Contents

A Study on Source Mechanism in the Interior Noise Problem of High Speed Trains H.I. Koh, H.B. Kwon, W.H. You, J.H. Park . . . . . . . . . . . . . . . . . . . . . . . . . 222 Reducing the Noise Emission by Increasing the Damping of the Rail: Results of a Field Test B. Asmussen, D. Stiebel, P. Kitson, D. Farrington, D. Benton . . . . . . . . . . 229

Session 7: Cost Benefit Considerations of Noise Abatement Railway Noise Abatement: The Case for Retrofitting Freight Vehicles with Composite Brake Blocks J. Oertli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236 A Systematic Approach for Arriving at Reasonable Heights and Locations for Noise Barriers Adjacent to Railway Lines C. Weber, K. Atkinson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243 How Can Infrastructure Manager Influence Noise Generation of Rolling Stock M.T. Kalivoda . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250 Acoustic Effectiveness of Damped Wheels and Impact on Life-Cycle Cost of Different Typologies of Passenger Trains A. Bracciali, S. Cervello, P. Gatti . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257 Mitigation Measures for Open Lines against Vibration and Ground-Borne Noise: A Swiss Overview R. M¨ uller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264

Session 8: Prediction Tools and Theoretical Models (I) Preliminary Analysis on Effect of Sleeper Pitch on Rail Corrugation at a Curved Track X. Jin, Z. Wen, Q. Liu, Z. Zhou . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 A Hybrid Model for Noise Generation from a Railway Wheel Due to Wheel/Rail Impact X. Xiao, X. Jin, X. Sheng . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278 A Time Domain Model for Wheel/Rail Interaction Aiming to Include Non-linear Contact Stiffness and Tangential Friction A. Pieringer, W. Kropp, J.C.O. Nielsen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285 Optimization of Track Parameters, Considering Their Physical Dispersion, to Minimize Rail Corrugation O. Oyarzabal, J. G´ omez, J. Santamar´ıa, E.G. Vadillo . . . . . . . . . . . . . . . . . 292

Table of Contents

XIII

Predicting the Effect of Temperature on the Performance of Elastomer-Based Rail Damping Devices N. Ahmad, D.J. Thompson, C.J.C. Jones, A.H. Muhr . . . . . . . . . . . . . . . . . 299 Estimation of Sound Transmission through Extruded Panels Using a Coupled Waveguide Finite Element-Boundary Element Method C.M. Nilsson, A.N. Thite, C.J.C. Jones, D.J. Thompson . . . . . . . . . . . . . . 306

Session 9: Prediction Tools and Theoretical Models (II) Squeal Prediction for a Bogied Vehicle in a Curve Z.Y. Huang, D.J. Thompson, C.J.C. Jones . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 Synthesis of Noise of Operating Vehicles: Development within SILENCE of a Tool with Listening Features E. Bongini, S. Molla, P.E. Gautier, D. Habault, P.O. Matt´ei, F. Poisson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320 IMAGINE Rail Noise Sources – A Practical Methodology M. Beuving, B. Hemsworth, R.R.K. Jones . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 Optimization of a Wheel Damper for Freight Wagons Using FEM Simulation W. Behr, S. Cervello . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334

Session 10: Grinding, Corrugation and Roughness (I) Types of Rail Roughness and the Selection of Vibration Isolation Measures H.E.M. Hunt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 Rail Roughness Monitoring in the Netherlands A.H.W.M. Kuijpers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348 Rail Roughness Level Assessment Based on High-Frequency Wheel–Rail Contact Force Measurements Jens C.O. Nielsen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355

Session 11: Grinding, Corrugation and Roughness (II) Testing the New Acoustic Rail Roughness Measurement Standard C. Jones, P. Fodiman, F. L´etourneaux, B. Croft . . . . . . . . . . . . . . . . . . . . . . 363

XIV

Table of Contents

Practical Implementations and Benefits of Highly Accurate Rail Roughness Measurements S. Lutzenberger, P. Holm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370

Session 12: New Noise Reduction Technologies (I) New Rail Dampers at the Railway Link Roosendaal-Vlissingen Tested within the Dutch Innovation Program E. van Haaren, G.A. van Keulen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378 Theoretical Study on Noise Reduction of Rail Component by Use of Rail Absorber T.X. Wu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384 Reducing Wheel-Rail Interaction Forces and Roughness Growth by Application of Rail Dampers B.E. Croft, C.J.C. Jones, D.J. Thompson . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392

Session 13: New Noise Reduction Technologies (II) Mitigation of Wheel Squeal and Flanging Noise on the Australian Rail Network D. Anderson, N. Wheatley . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399 How to Avoid Squeal Noise on Railways State of the Art and Practical Experience S. B¨ uhler, B. Thallemer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406 Noise Reduction Measures at Freight Train Locomotives “Blue Tiger” C. Czolbe, M. Hecht . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412 Noise Reduction at Urban Hot-Spots by Vehicle Noise Control U. Orrenius, S. Leth, A. Frid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419

Poster Session Directivity of Railway Rolling Noise X. Zhang . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426 Complex Eigenvalue Analysis of Railway Curve Squeal G.X. Chen, J.B. Xiao, Q.Y. Liu, Z.R. Zhou . . . . . . . . . . . . . . . . . . . . . . . . . . 433 Wave Propagation in Railway Tracks at High Frequencies J. Ryue, D.J. Thompson, P.R. White, D.R. Thompson . . . . . . . . . . . . . . . . 440

Table of Contents

XV

Stability and Transient Analysis in the Modelling of Railway Disc Brake Squeal X. Lorang, O. Chiello . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447 Large-Scale Fatigue Test of Stone Wool Based Anti-vibration Mats K.B. Gatzwiller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454 Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461

Environmental Noise Reduction of Tokaido Shinkansen and Future Prospect Hitoshi Kanda, Hideaki Tsuda, Kimihiro Ichikawa, and Shohei Yoshida Technology Planning Department, General Technology Division Central Japan Railway Co., 2-1-85 Kounan, Minato-ku, 109-8204 Tokyo, Japan Tel.: +81 3 6711 9580; Fax: +81 3 6711 9707 [email protected]

Summary This paper describes the technical review of the environmental noise reduction of Tokaido Shinkansen and its future prospect. Current environmental quality standards for Shinkansen railway noise were established in 1975 by the Environmental Agency, Japan. The Japanese Cabinet agreed on “general principles for countermeasures against Shinkansen railway noise” in 1976. Based on the principles, Central Japan Railway Company has spent a great effort to reduce the railway noise at source. Owing to measures, the wayside noise has been remarkably decreased. On the other hand, the land use planning has not been strongly enforced. So there are regrettably some problems on later inhabitants along the Shinkansen railroad. This paper also discusses the noise evaluation. A present Shinkansen noise index is given by LAmax (peak noise level). It is reasonable to assume that the index will be shifted to LAEQ basis, since it has been widely applied. In order to study the effect of LAEQ standard, it seems important to investigate theoretical and scientific aspects of noise evaluation in other countries.

1 Introduction Central Japan Railway Company (JR Central) commenced operations in April 1987 upon the privatization of the Japanese National Railways (JNR). The core of JR Central’s operations is the high-speed Tokaido Shinkansen, the main transportation line linking Japan’s metropolitan areas of Tokyo, Nagoya and Osaka. Providing a great mobility, it has played an important role in Japan’s economic growth and high living standards. Moreover, Tokaido Shinkansen has since offered large capacity, high frequency transportation services with high safety; such performance is drawing a strong attention from the worlds. JR Central has consistently responded to growing demand from business travellers and long-distance commuters since its founding. In March 1992, the timetable was revised in order to start the series 300 rolling stock into service with operating speeds up to 270 km/h. It has cut travel time drastically. The similar efforts have been made constantly to raise the service level. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 1–8, 2008. springerlink.com © Springer-Verlag Berlin Heidelberg 2008

2

H. Kanda et al.

In the summer of 2007, the new trainset of series N700 starts the commercial operation. Along with starting new services, the series N700 will offer an "even more comfortable space" that meets the various needs of passengers. In addition, the series N700 achieves a reduction in energy consumption by 25% compared with the series 300 owing to the latest technologies. It greatly contributes to the preservation of the global environment. When studying such service improvement, the highest consideration must be given to the prevention of noise and vibration alongside the railroad because the Shinkansen connects Japan’s most densely populated areas. As means of improving environmental conditions, controlling environmental noise has recently been gathering attention in Japan as well as other countries.

2 Environmental Quality Standards for Shinkansen Railway Current environmental quality standards for Shinkansen railway noise were established in 1975 by the Environmental Agency, Japan. The index of evaluation is given by LpA, Smax (slow max value of A-weighted noise level) not by LAeq (equivalent noise energy level). The standard values are classified into two categories (I and II) based on the condition of residential use (Table 1). In category I, standard value is 70 dB or less, and in category II, standard value is 75 dB or less. They are applied from 6am to 24pm, the service hours of commercial Shinkansen operation. Table 1. Environmental quality standards for Shinkansen railway noise

Category of Area

Standard Value (LpA,Smax)

Category I

70 dB or less

Category II

75 dB or less

Category I --- Mainly residential use Category II --- Other areas including commercial and industrial, where normal living conditions should be preserved One year later, the Japanese Cabinet agreed on “general principles for countermeasures against Shinkansen railway noise” in 1976. The principles addressed three measures that should be strongly promoted for achieving the environmental quality standards; (1) countermeasures at noise source, (2) compensation for the loss, and (3) land use planning along the Shinkansen railroad. Based on the principles, the Japanese National Railways and JR-Central have so far spent a great effort to decrease the railway noise at source. Successive efforts on measures at noise source will also be accumulated in future toward the environmental quality standards.

Environmental Noise Reduction of Tokaido Shinkansen and Future Prospect

3

3 Technical Development for Noise Source 3.1 Rolling Stocks In order to reduce the high-speed railway noise at source, technical development has been carried out for rolling stocks including pantograph covers and low noise pantographs, car’s ultimate shapes, smooth car bodies, and so on. Figure 1 presents the comparison of old and new type of pantographs and insulator covers. Figure 1(a) shows the series 300 former type, figure 1(b) shows the series 300 later type and the series 700, and figure 1(c) shows the new type insulator cover and low noise pantograph developed for the latest series N700.

(a) Series 300 former type (b) Series 300 later type and series 700

(c) New Series N700

Fig. 1. Comparison of pantographs and insulator covers

Figure2 shows a photograph of a wind tunnel test for designing the series 700 “aerostream form” ultimate shape. Figure 3 shows the “external flush gap shroud (all covering hoods)” [1] and boggie skirts which are newly developed and introduced to the series N700 Shinkansen.

Fig. 2. Ultimate shape of Series 700

Fig. 3. All covering hoods and boggie skirts of N700

3.2 Rail and Trucks Noise can be reduced at rail and trucks by installing an elastic lubber sleeper or a lubber ballast mat shown in Figure 4. These measures are especially effective to reduce structure noise, and are widely used in viaduct sections of Tokaido Shinkansen. This measure is effective to vibration reduction as well as noise reduction.

4

H. Kanda et al.

lubber ballast mat elastic lubber sleeper

Fig. 4. Elastic lubber sleeper and lubber ballast mat

3.3 Noise Barriers Noise barriers are built alongside of the Shinkansen railroad. Typical noise barrier of Tokaido Shinkansen is shown in Figure 5. Basic noise barriers are 2-meters high, and some additional barriers may be attached in a special area to increase the reduction effect.

Fig. 5. Photograph of noise barriers

Fig. 6. Photograph of a new-type noise barrier

Because noise barriers were not considered for the initial design of the Tokaido Shinkansen structures, there is a limitation of barrier’s height especially in a viaduct section. In addition, since a high noise barrier may interrupt the wayside scenery from the window, special noise barriers are also experimentally developed. Figure 6 shows an example of a new-type barrier with “mountain-type” slits in the upper part of the wall to secure the window view. 3.4 Simulation Techniques Measurement techniques and prediction methods for noise propagation are important to develop the new type Shinkansen trainset with better environmental quality. Figure 7 shows the “sound intensity counter maps” of Series 300 and 700, developed by a research group of JR-Central[2]. It is clearly demonstrated that the dark areas near the boggie skirts, which display large sound pressure, are widely decreased for series 700 in Figure 7(b) compared to the series 300 in Figure 7(a).

Environmental Noise Reduction of Tokaido Shinkansen and Future Prospect

(a) Series 300

5

(b) Series700

Fig. 7. Sound intensity counter maps

3.5 Environmental Supervision Depots Central Japan Railway Co. has four local environmental supervision depots alongside of the Shinkansen railroad. The main objective of these depots is the wayside environmental preservation of the Tokaido Shinkansen. There are two or three special engineers in each depot who are in charge of noise and vibration measurement, public relations and negotiation with inhabitants.

4 Noise Level Reduction Owing to these measures, the Shinkansen wayside noise has been remarkably decreased from the year of its inauguration in 1964, whereas the maximum velocity has been increased to 270 km/h. Figure 8 shows the transition of noise level from 1964 to 2000[3]. Typical counter-measures in each period are also noted below the figure. 90

300 290 280

85 270 260 80

250 240 230

75 220 210 70

200 1964

1967

Countermeasures Inauguration Noise of Tokaido barrier Shinkansen

1975

1986

1991

1994

2000

Series 300

Series 700

Year

Ballast mat

Rail grinding

Pantograph cover

Fig. 8. Transition of Shinkansen noise level and countermeasures

Max Velocity (km/h)

Noise Level (LpA, Smax) dB

Noise Level Max Velocity

6

H. Kanda et al.

It is expected in future that successive efforts on measures at noise source will be accumulated toward the environmental quality standards.

5 Land Use Planning On the other hand, the land use planning, which is another principle addressed in the Cabinet Agreement, has not been strongly enforced so far. So there are regrettably some problems on later inhabitants along the Shinkansen railroad. Figure 9 demonstrates some photographs that show the change of Shinkansen wayside land use. After the inauguration of Shinkansen railway, many residential houses have been built alongside of the railroad. Therefore target areas for countermeasures have been increased with the development of residential areas at waysides. It is hoped in future that a specific method is introduced to the land use planning in Japan to restrict the increase of residential houses alongside of the railroad.

Tokaido Shinkansen

Local Tokaido Line

(a) Under Construction (1963)

(b) Present Condition (2004)

Fig. 9. Change of wayside land use

6 Index of Noise Evaluation 6.1 LAEQ of Shinkansen The new standards for Shinkansen noise have been discussed in recent years in Japan. A present noise index is given by LAmax, not by LAEQ which is widely applied in foreign countries. However, LAEQ has already been applied in Japan to the guideline for newly-constructed local railway noise from 1995 and the environmental standards for road traffic from 1998. It is therefore reasonable to assume that the Shinkansen noise index will be shifted to the LAEQ basis in a future. There is a research paper that estimates the LAEQ of Shinkansen noise by the use of an indirect calculation from LAmax with a current trains’ timetable[4]. The results are summarized in Table 2. Tokaido Shinkansen offers 244 regular trains daily, in the daytime from 7am to 22pm. If all trains have 70dB peak noise level, LAEQ, 15H (7-22) is calculated to be 54.3dB. Similarly, if all trains have 75dB peak level, LAEQ, 15H (7-22)

Environmental Noise Reduction of Tokaido Shinkansen and Future Prospect

7

becomes 59.3dB [5]. Table 2 also addresses the environmental standards for road traffic and guidelines for newly-constructed local railway in Japan. 6.2 Discussion It becomes clear from Table2 that LAEQ of Shinkansen is smaller than that of road traffic in all categories. In both “exclusive residential” and “commercial & industrial” areas, Shinkansen keeps the environmental standards of road traffic even at night. In addition, since the grouping of “residential” is different, residential area of Shinkansen offers still smaller LAEQ than road. Table 2. Comparison between Shinkansen Environmental Noise Standards and Other Surface Traffic Shinkansen Railway (Environmental Standards)

Shinkansen Railway (Indirect Estimation from L Amax)

L Amax (dB)

L Aeq (dB)

Service hours (6-24)

Day (7-22) 244 trains

70

54.3

Exclusive Residential

Road traffic (Environmental Standards)

L Aeq (dB) Day (6-22)

Night (22-6)

60

55

Residential Commercial and Industrial

65 75

Newlyconstructed local railway (Guidelines) L Aeq (dB) Day (7-22)

Night (22-7)

60

55

60

59.3

Next, when Shinkansen is compared with local railway, Shinkansen keeps the newly constructed local railway guidelines even at night in exclusive residential and residential areas, and keeps the daytime guidelines in commercial and industrial areas. This research reveals that, on the LAEQ basis, the present noise standards for Shinkansen railway provides more preferable wayside noise environment than the road traffic. This is contrary to many European countries that have the “railwaybonus” on the environmental standards for traffic noise. In order to study the effect of LAEQ standard and the background of railway-bonus, it is important to investigate the theoretical and scientific aspects of noise evaluation in European countries and to make clear if the similar idea is applicable to the standards in other countries.

7 Conclusion In this paper, technical review of the environmental noise reduction of Tokaido Shinkansen and its future prospect are described. Typical countermeasures at noise source are summarized, and the reduction effect of wayside noise level during past 40

8

H. Kanda et al.

years is reviewed. Some problems are also mentioned on later inhabitants along the Shinkansen railroad. This paper also discusses the noise evaluation and comparison of environmental standard value between railway and road traffic. In conclusion, it is hoped that the Japanese Cabinet Agreement in 1976 is promoted more strongly toward the environmental quality standards. At the same time, railway companies will continue technical development on countermeasures at noise source.

References [1] Kitayama, S.: Noise reduction of high-speed trains by installing an external flush gap shroud. In: Inter-Noise 2003, vol. 1008 (2003) [2] Kawahara, M., et al.: Source identification and prediction of Shinkansen noise by sound intensity method. In: Inter-Noise 1997 (1997) [3] Maeda, T.: Japanese Shinkansen noise: Development of noise reduction technology. In: Inter-Noise 2006, Distinguished Lecture 2 (2006) [4] Nagakura, K., Zenda, Y.: Prediction model of wayside noise level of Shinkansen. Railway Technical Research Institute Report (in Japanese) 14(9), 9 (2000) [5] INCE/J, Research report on noise countermeasures of Shinkansen Railway. In: Ministry of the Environment, Japan (2004) (in Japanese)

Distortion of Compression Wave Propagating through Shinkansen Tunnel T. Miyachi1, T. Fukuda1, M. Iida1, T. Maeda1, and S. Ozawa2 1

Railway Technical Research Institute, 2-8-38, Hikari-cho, Kokubunji-shi, Tokyo, 185-8540, Japan Tel.: +81 42 573 7318; Fax: +81 42 573 7329 [email protected] 2 Tokyo University of Technology, 1404-1, Katakura-cho, Hachioji-shi, Tokyo, 192-0982, Japan Tel.: +81 42 637 2508; Fax: +81 42 637 2508 [email protected]

Summary When a compression wave generated by a train entering a tunnel is propagated through the tunnel and reaches the exit portal, a pressure pulse (‘micro-pressure wave’) is radiated from the exit portal. The distortion of the compression wave during propagation in the tunnel is important for estimating the magnitude of the micro-pressure wave radiated from the exit portal. In this paper, the dependence of the distortion of the compression wave on the initial waveform at the tunnel entrance is investigated. First, a simple system of equations concerning the wave propagation in the tunnel is described. Secondly, theoretical and numerical investigations into the effects of nonlinearity and unsteady friction on the tunnel wall on the distortion of the compression wave are made.

1 Introduction 1.1 Objectives A pressure pulse (‘micro-pressure wave’ [1,2]) radiated from the exit portal when a compression wave generated by a train entering a tunnel is propagated through the tunnel and reaches the exit portal has become one of the important environmental problems in high speed railways. The pressure of the micro-pressure wave is approximately proportional to the pressure gradient of the incident compression wave at the exit portal [2]. Therefore the distortion of the compression waveform during propagation in the tunnel is important for estimating the magnitude of the micro-pressure wave and considering the countermeasures. The compression waveform is distorted by the combined effect of nonlinearity, frictional losses on the tunnel wall, side branches in the tunnel and others. In Shinkansen, the micro-pressure wave is large in a long concrete slab-track tunnel (we call it simply ‘tunnel’ in this paper) because the compression wavefront at the entrance of the tunnel (we call it ‘initial wave’ in this paper) steepens during propagation in the tunnel due to the dependence of propagation speed on particle velocity (we call it ‘nonlinear effect’). The waveform steepening becomes significant when the pressure gradient of the initial wave becomes large. In the long tunnel, the B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 9–18, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

T. Miyachi et al.

3 2 1 0 10 5 0 0

0.5

1 t(s)

x = 0, Initial Waveform CASE 1, Field test CASE 2, Field test

1.5

p(kPa)

2 1

∂p/∂t (kPa/s)

∂p/∂t (kPa/s)

p(kPa)

10

0 15 10 5 0 0

0.1

0.2

0.

t(s) x = 9.5 km CASE 1, Field test CASE 2, Field test

Fig. 1. Distortion of compression waves measured by field test

countermeasures at the stage of generation of the compression wave at the tunnel entrance, which reduce the pressure gradient of the initial wave, are very effective; tunnel entrance hoods [1] and elongation and optimization of the train nose shape [3] have been applied in Shinkansen. The tunnel entrance hoods installed at the tunnel portals in Shinkansen usually have openings (‘windows’) on their sides. Till now, the window pattern has been optimized to minimize the maximum pressure gradient of the initial wave, which is assumed to be the only index for optimization. However the recent field tests have revealed that its waveform becomes an important index for optimization in the case of a long tunnel in addition to the pressure gradient of the initial wave. In this paper, (1) we show a simple system of equations concerning the wave propagation in a tunnel, and (2) we make theoretical and numerical investigations into the effects of nonlinearity and unsteady friction on the tunnel wall on the distortion of the compression wave and clarify the dependence of the distortion of the compression wave on the initial waveform at the tunnel entrance. 1.2 Background Figure 1 shows the waveforms of the compression wave measured at two locations 100 m inside from both portals of a Shinkansen tunnel 9.7 km in length and the waveforms of the pressure gradient calculated from the measured data. Tunnel entrance hoods are installed at both portals and there are two large shafts in this tunnel. The train speed is almost identical in CASE1 and CASE2, however the train geometry is different: the cross-sectional area and the nose length of the CASE1 train are smaller than those of the CASE2 train. The maximum pressure gradient of the compression waves at x = 9.5 km in CASE1 is approximately 70 % of that in CASE2, although the maximum pressure gradients of the initial waves are almost identical. This experimental result indicates the importance of the effect of the initial waveform on the distortion of the compression wave during propagation in the tunnel.

Distortion of Compression Wave Propagating through Shinkansen Tunnel

11

2 Theoretical Analysis The following one-dimensional equation has been proposed for the propagation of the compression wave through a tunnel using forward-propagating wave approximation [1]:

∂u * ⎛ * γ + 1 * ⎞ ∂u * 1 4τ * u ⎟ *+ + ⎜ c0 + = 0, ∂t * ⎝ 2 2 ρ * d H* ⎠ ∂x

(1)

where u: particle velocity associated with pressure wave, c: speed of sound, t: time, x: space coordinate along tunnel, dH: hydraulic diameter of tunnel, γ : ratio of specific heat, ρ : air density, τ : wall friction, superscript *: quantity with dimension, subscript 0: quantity at reference state. The pressure of the compression wave p is approximately given by p* ≈ ρ0*c0*u*. The following dimensionless values p, x, t, τ are introduced:

p=

p* , γ P0*

x=

τ* x* t* τ , t = , = , γ P0* d H* d H* c0*

(

)

(2)

where P*0: atmospheric pressure. The value of p satisfies p << 1. Substituting eq. (2) into eq. (1) with ρ∗ = ρ0* in the friction term, dividing it by 1+(γ +1)p/2 with the approximation 1/(1+(γ +1)p/2) ≈ 1−(γ +1)p/2 and performing coordinate transformation x′ = x, t′ = t − x, we obtain the following space evolution equation in which x′ and t′ are rewritten as x and t again:

∂p γ + 1 ∂p 1 ⎛ γ +1 ⎞ = p − ⋅ 4τ ⎜ 1 − p⎟ . ∂x 2 ∂t 2 2 ⎝ ⎠

(3)

Taking into account the loss of the main stream due to the heat transfer on the tunnel wall [3, 4, 5], we modify eq. (3) by introducing a multiplier factor of 4τ . Finally, we obtain a system of equations for the propagation of the compression wave through the tunnel:

∂p γ + 1 ∂p 1 ⎛ γ − 1 ⎞ ⎛ γ + 1 ⎞ = p − ⎜1 + p ⎟ , 4τ = 4 (τ s + τ us ) , ⎟ 4τ ⎜ 1 − ∂x 2 ∂t 2 ⎝ Pr 2 3 ⎠ ⎝ 2 ⎠ 4τ s = ε s

4τ us = ε us

4

π

64 1 2 32 p = εs p, Re 2 Rec

(4)

(5) ⎛ 14.3 ⎞

log10 ⎜ 0.05 ⎟ exp[−4 βφ / Rec ] ⌠ ∂p ⎝ Re ⎠ , t d ( − φ ) φ , β = 0.135Re ⎮ Rec ⌡0 ∂t φ

1

t

(6) where Pr: Prandtl number, subscript s: steady condition, subscript us: unsteady condition. The friction factors εs and εus are determined by comparing the results of field

12

T. Miyachi et al.

test with those of numerical simulation. Two types of Reynolds number appeared in eqs. (5) and (6) are Re =

(( p

*

)

ρ * c* ) ⋅ d H* ν * and Rec = ( c* d H* ) ν * . 0

0

0

In Ref. [4], it has become clear that the numerical results by this method using eqs. (4), (5), and (6) agree well with those of numerical simulation based on the general CFD method [5, 6] when the same models of friction term are employed. The advantage of this method is that eq. (4) is simpler than the equations of CFD and remarkably reduces the calculation time because the CFL condition is relaxed and the space evolution equation is suitable for calculating of the convolution of the unsteady friction term. As shown in eq. (5), the steady friction is assumed as laminar, not turbulent flow type. The validity of this will be described in Sec.3.1, where is shown good agreement between the results of field test and numerical simulation. With approximations 1−(γ+1)p/2 ≈ 1 in the friction term and exp[ −B*4φ /Rec] ≈ 1 in the convolution, eq. (4) can be further simplified as follows: ∂p γ + 1 ∂p 2 ⌠ ∂p 1 = p − Ap − B dφ , ⎮ ∂t (t − φ ) ∂x 2 ∂t π ⌡0 φ t

⎛ γ − 1 ⎞ 16 ⎛ γ −1 ⎞ 1 A = ⎜1 + 2 3 ⎟ ε s , B = ⎜1 + 2 3 ⎟ ε us Pr Re ⎝ ⎠ c ⎝ Pr ⎠ Rec

(7) Analysis in the following sections is based on eq. (7). We first consider the effect of the nonlinear term alone in eq. (7). The solution of eq. (7) with A = 0 and B = 0 obtained by the method of characteristics is

∂p ( x, t ) ∂t

=

h′ ( t0 ) γ +1 , t=− h ( t0 ) x + t0 , γ +1 2 1− h ′ ( t0 ) x 2

where h(t): initial waveform at x = 0, h′(t) = dh(t)/dt. Equation (8) shows that the maximum pressure gradient of the compression wave at the location x depends only on the maximum pressure gradient of the initial wave h(t). Next, we consider the effect of friction term alone; the effect of the nonlinear term is not taken temporarily into account to make the theoretical analysis easy. We consider the following space evolution equation:

(8)

1 0.8 0.6 Erfc ⎡⎣1

t ⎤⎦

0.4 0.2 0 0

5

10 t

Fig. 2. Erfc ⎡1

⎣

t⎤ ⎦

t

∂p 2 ⌠ ∂p 1 = − Ap − B (t − φ ) dφ . ⎮ ∂x π ⌡0 ∂t φ

(9)

Distortion of Compression Wave Propagating through Shinkansen Tunnel

13

The analytical solution of eq. (9) can be obtained by use of the Laplace transform, and the pressure p and its time derivative ∂p/∂t under the condition h(0) = h′(0) = 0 are

∂p ( x, t ) ∂t

t

⌠ ⎡ Bx ⎤ = exp [ − Ax ] ⎮ h′′(t − φ ) ⋅ Erfc ⎢ ⎥ dφ , ⎮ ⎢⎣ φ ⎥⎦ ⌡0

(10)

where Erfc[ x ]: complementary error function. The exp[ −Ax ] expresses the effect of the laminar steady friction and this term reduces the compression waveform at the location x according to a fixed ratio of exp[ −Ax ]. The convolution integrals in eq. (10) express the effect of the unsteady friction. The value of this term becomes small when the integrand becomes small. Figure 2 shows Erfc[ t−1/2 ] as a function of t. The function of Erfc[ Bxφ−1/2 ] in the integrand is a weighting function, which takes a small value when φ is small. This shows that the waveform of the pressure gradient which arises rapidly to a given maximum value in a short time range of φ tends to be greatly reduced; the initial waveform having large ∂2p/∂t2 tends to be greatly reduced by the unsteady friction. The effect of the friction term causes rounding of the single rectangular waveform of ∂p/∂t at the initial stage and the waveform becomes higher on the more right side. This is caused not by the steady friction but by the unsteady friction, because the effect of the steady friction is only to reduce the pressure gradient according to a fixed ratio exp[ −Ax ] irrespective of time. Therefore, the effective waveform of ∂p/∂t of the initial wave for reducing ∂p/∂t at the tunnel exit after propagation in a long tunnel is such that the initial waveform of ∂p/∂t is higher on the more left side and there exists a trough between two peaks.

3 Numerical Simulation We investigate the combined effect of the friction term and the nonlinear term of eq. (4) by numerical simulation. Table 1 and 2 show the geometrical parameters used for numerical simulation of the compression wave Table 1. Geometrical parameters of Shinkansen tunnel for propagating through a numerical simulation Shinkansen tunnel with norCross-sectional area (m2) 63.4 mal slab-track (not frameHydraulic diameter (m) 8.1 Main tunnel shaped slab). The Chakravarthy’s TVD scheme [7] Type of track Normal-slab is employed for the numeriCross-sectional area (m2) Short side 7.1 branch Type 1 cal computation. Length (m) 2.0 or 3.0 Figure 3 shows the reCross-sectional area (m2) 7.1 Short Side branch Type 2 Length (m) 5.0 sults of numerical simulaapprox. 15 Cross-sectional area (m2) tion by use of the initial Inclined Shaft Length (m) over 170 waveform measured in the field test (Fig. 1). The field test has been conducted at Shinkansen tunnel about 9.7 km in length, where there are many short side branches arranged at almost constant intervals of 500 m and two

14

T. Miyachi et al.

large inclined shafts in the tunnel. Table 2. Parameters for numerical calculation of The effects of these branches are compression wave propagating in tunnel calculated by one-dimensional ∆t *=1.0×10-3 Grid interval (s) acoustic analysis [1, 5]. Figure 3 Integration interval (m) ∆x*=10 shows that the numerical results of the waveform of p and ∂p/∂t at approx. 0.5 Courant Number ε x =9.5 km agree well with those Steady friction factor s 1500 of the field test. ε Unsteady friction factor us 5 We investigate the distortion of the model waveforms of ∂p/∂t with different initial shapes of ∂p/∂t: trapezoid and triangle. Numerical simulation in Sec.3.2 has been conducted at Δx* = 5 m to avoid numerical instability due to the non-smoothness of the waveform of ∂p/∂t of the model initial wave, this value is smaller than that shown in Table 2 (Δx* = 10 m). Figure 4 shows the results of numerical simulation conducted under the assumption that there are no short side branches and shafts in the tunnel. An initial wave with isosceles trapezoid shape of ∂p/∂t tends to steepen its wavefront as compared with that with triangle shape. An initial wave with triangle shape of ∂p/∂t with large rise time tends to steepen its wavefront as compared with that with small rise time of ∂p/∂t. These results agree with the results of theoretical investigation in Sec.2.2. These tendencies are mainly due to the effect of the unsteady friction term, being enhanced by the nonlinear term.

0.1

0.2

p(kPa)

0 15 10 5 0 0

2 1

∂p/∂t (kPa/s)

p(kPa)

1

∂p/∂t (kPa/s)

2

0 15 10 5 0 0

0.3

0.1

t(s)

0.2

0.3

t(s)

x = 9.5 km CASE 1, Field test CASE 1, Simulation

x = 9.5 km CASE 2, Field test CASE 2, Simulation

Fig. 3. Comparison of results of field test and numerical simulation

Theoretical study on the pressure rise of the compression wave generated by a train entering tunnel has been done by Hara [8] which gives

1 − (1 − R ) 2 1 = ρ *0 U * , 2 2 (1 − M ) M + (1 − R ) 2

* max

p

(

)

(11)

Distortion of Compression Wave Propagating through Shinkansen Tunnel

where R: cross-sectional area ratio of train to tunnel and M: train Mach number. As the first approximation, the pressure rise of the compression wave depends on only M and R, whether a hood exists or not. Three typical examples of waveforms of p and ∂p/∂t of the initial compression waves at different conditions of train

Table 3. Geometrical parameters of tunnel entrance hood and train for numerical simulation Cross-sectional area ratio to main tunnnel area Window pattern Length (m) Cross-sectional area ratio to main tunnel area Nose length ratio to equivalent diameter of main tunnel Speed (km/h)

Hood

Train

0 0

0.3

0.16 3.3

300

x = 4 km

0.6

∂p/∂t (kPa/s)

x = 2 km

∂p/∂t (kPa/s)

∂p/∂t (kPa/s)

Initial Waveforms

10

1.4

2 patterns 20

20

20

20

15

10

0 0

0.3

10

0.6

0 0

0.3

t(s)

t(s)

0.6 t(s)

Fig. 4. Distortion of waveforms of ∂p/∂t with different initial shapes: trapezoid and triangle

speed and hood are shown in Fig.5. These initial compression waveforms generated by a train entering a tunnel and a hood with windows are calculated by Howe’s method [9] with conditions in Table 3. The total pressure rise by a train at 300 km/h (CASE4-1 and no-hood case) is larger than that at 245 km/h (CASE3) irrespective of

10 ∂p/∂t (kPa/s)

p(kPa)

2 1 0 0

0.2

0.4

0.6

0.8

x=0 km Initial waveform

5

0 0

0.2

t(s)

0.4

0.6

0.8

t(s) 245km/h, without hood, CASE3 300km/h, with 20m hood, CASE4-1 300km/h, without hood

Fig. 5. Waveforms of initial wave of p and ∂p/∂t by trains with different speeds

16

T. Miyachi et al.

5

5

20

x=3 km

∂p/∂t (kPa/s)

∂p/∂t (kPa/s)

x=5 km

0

0 10 5 0 0

10

x=0 km Initial waveform

∂p/∂t (kPa/s)

∂p/∂t (kPa/s)

10

0.2

0.4

0.6

0.8

x=8 km

15 10 5 0 0.2

0.3

t(s)

0.4

0.5

t(s) 245km/h, without hood, CASE3 300km/h, with 20m hood, CASE4-1 300km/h, with 20m hood, CASE4-2

Fig. 6. Distortion of ∂p/∂t by trains with different speeds

the presence of the hood. The maximum values of ∂p/∂t in CASE3 (train speed: 245 km/h, without hood) and CASE4-1 (train speed: 300 km/h, with 20m long hood) are almost identical, but the waveforms of ∂p/∂t are different: triangle (CASE3) and trapezoid (CASE4-1). The distortions of the waveforms of ∂p/∂t in CASE3, CASE4-1, and CASE4-2 (train speed: 300 km/h, with 20m long hood in a window pattern different from CASE4-1) are shown in Fig. 6. The compression waveforms Short side 500 m 500 m branch after propagation at x are calcuType 2 lated by the method of Sec. 3.1. Short side branch Here we assume that the short Type 1 + Type 2 side branches are arranged at regular intervals of 0.5 km as Fig. 7. Arrangement of short side branches for numerishown in Fig. 7. The maximum cal simulation value of ∂p/∂t in CASE3 and CASE4-2 are about 50 % of that in CASE4-1 at x = 8 km. The results of the numerical simulation (Fig. 6) show that the waveform of ∂p/∂t of the initial compression wave, effective for reducing the micro-pressure wave, has a trough and higher on the left side. The similar situation will occur when the cross-sectional area of a train is enlarged. In the results of the field tests shown in Fig. 1, the pressure rise in CASE2 is larger than that in CASE1. The difference of the pressure rise is due to the cross-sectional area of two types of trains (the train speeds are almost identical). The maximum

Distortion of Compression Wave Propagating through Shinkansen Tunnel

17

values of ∂p/∂t of the initial compression waves in CASE1 and CASE2 are almost identical because the train nose length in CASE2 is larger than that in CASE1. The waveform of ∂p/∂t in the initial compression wave in CASE2 looks much like a trapezoid because of its shallower trough between the two peaks. Therefore the initial compression wave in CASE2 tends to steepen more as compared with that in CASE1, and this has been mentioned by the analysis in Sec.2.2 and numerical simulation in Sec.3.2.

4 Conclusion The distortion of the compression wave during propagation in a Shinkansen tunnel is investigated. The important results are summarized as follows. (1) A simple system of equations concerning the wave propagation in a tunnel was presented. (2) Theoretical and numerical investigations into the effect of nonlinearity and unsteady friction on the tunnel wall on the distortion of the compression wave were made. The results made clear that the effective waveform of ∂p/∂t of the initial wave in suppressing it to a smaller value after propagation in a long slab-track tunnel is such that the initial waveform of ∂p/∂t is higher on the more left side and there exists a trough between the peaks of ∂p/∂t. (3) The results of numerical simulation showed that the magnitude of the micropressure wave caused by a higher speed train tends to become larger than that caused by a lower speed train in the case of a long slab-track tunnel, even when the maximum pressure gradients of the initial wave at the tunnel entrance are identical.

References [1] Ozawa, S.: Studies of micro-pressure wave radiated from a tunnel exit. Rai. Tech. Res. Rep., Japanese National Railways (in Japanese) 1121 (1979) [2] Yamamoto, A.: Micro-pressure wave radiated from tunnel exit. Preprint of the Spring Meeting of Physical Society of Japan (in Japanese) (1977) [3] Iida, M., Matsumura, T., Nakatani, K., Fukuda, T., Maeda, T.: Optimum nose shape for reducing tunnel sonic boom. Int. Rail. Conf. Better Journey Time - Better Business, IMechE, 271–280 (1996) [4] Miyachi, T., Fukuda, T., Ozawa, S.: A new simple method for calculating distortion of compression wave (in Japanese). Proc. Mec. Eng. Cong., Jpn. Mec. Eng. 2, 347–348 (2006) [5] Fukuda, T., Ozawa, S., Iida, M., Takasaki, T., Wakabayashi, Y.: Distortion of compression wave propagating through very long tunnel with slab-tracks. JSME Int. J. 49(4), 1156–1164 (2006) [6] Fukuda, T., Miyachi, T., Iida, M., Ozawa, S.: Propagation of compression wave in a long slab-trucked tunnel and ballast-tracked tunnel. In: Proc. 12th Int. Symp. Aerodyn. Vent. Vehi. Tun., vol. 2, pp. 777–788 (2006) [7] Chakravarthy, S.R.: A new class of high accuracy TVD schemes for hyperbolic conservation laws. AIAA Paper 85(0363) (1985)

18

T. Miyachi et al.

[8] Hara, T.: Aerodynamic force acting on a high speed train at tunnel entrance. Bulletin JSME 4, 547–553 (1961) [9] Howe, M.S., Iida, M., Maeda, T., Sakuma, Y.: Rapid calculation of the compression wave generated by a train entering a tunnel with a vented hood. J. Sound Vib. 297, 267– 697 (2006) [10] Vardy, A., Brown, J.: An overview of wave propagation in tunnels. In: TRANSAERO – A European initiative on transient aerodynamics for railway system optimisation, Notes on numerical fluid mechanics, 79th edn., pp. 249–266. Springer, Berlin (2002) [11] Ehrendorfer, K., Reiterer, M., Sockel, H.: Numerical investigation of the micro pressure wave. In: TRANSAERO – A European initiative on transient aerodynamics for railway system optimisation. Notes on numerical fluid mechanics, vol. 79, pp. 322–341. Springer, Berlin (2002)

The Influence of the Train Speed on Vibrations Due to High Speed Trains G. Lombaert, G. Degrande, and J. Bekaert Department of Civil Engineering, K.U. Leuven, Kasteelpark Arenberg 40, B-3001, Leuven, Belgium Tel.: +32 16 321677; Fax: +32 16 321988 [email protected]

Abstract This paper deals with the influence of the train speed on vibrations induced by high speed trains. In the past 10 years, free field as well as track vibrations have been measured at several sites on the Belgian part of the European high speed rail network. These experiments were performed within the frame of homologation tests, which has allowed data collection for a wide range of train speeds. This paper concentrates on the analysis of track as well as free field vibrations for different train speeds at a site along the line L2 between Brussels and K¨oln. At this site, 11 passages of the IC train at speeds between 156 km/h and 225 km/h have been recorded and 11 passages of the Thalys HST at speeds between 218 km/h and 326 km/h. The experimental data for different train speeds are analysed and compared to numerical predictions that have been performed by means of a numerical model that accounts for the dynamic interaction between the train, the track and the soil.

1 Introduction The development of the high speed train (HST) network in Europe, the USA and Asia has increased the interest for ground-borne vibrations in the built environment. These vibrations are generated by a large number of excitation mechanisms. For HST tracks on soft soils, the train speed can be close to or even larger than the critical phase velocity of the coupled track-soil system. Quasi–static axle loads dominate both the track and the free field response. When the train speed is well below the wave velocities in the soil, the contribution of the quasi-static axle loads to the free field response is small compared to the contribution of the dynamic axle loads. The latter are generated by parametric excitation due to the discrete supports of the rails, transient excitation due to rail joints and wheelflats and the excitation due to wheel and rail roughness and track unevenness. The interaction between the train and the track cannot be disregarded when calculating the dynamic axle loads, as the resonance of the unsprung mass against the track stiffness is important. Recently, several three-dimensional numerical models have been developed that account for the dynamic interaction between the train, the track and the soil. Sheng et al. have modelled the dynamic train-track-soil interaction by means of a track model B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 19–25, 2008. c Springer-Verlag Berlin Heidelberg 2008 springerlink.com 

20

G. Lombaert, G. Degrande, and J. Bekaert

with an infinite layered beam on top of a layered halfspace [1]. Auersch [2] has coupled a finite element model of the track to a boundary element model for the soil to solve the vehicle-track interaction problem. Metrikine et al. [3] have studied the stability of a moving train bogie. In this paper, a numerical model for railway induced vibrations developed by Lombaert et al. [4] is validated by means of vibration measurements during the homologation tests of the HST track on the line L2 between Brussels and K¨oln. These tests have been performed with an IC train at a speed between 155.9 and 225.3 km/h and a Thalys HST at a speed between 218.1 and 326.1 km/h. This paper concentrates on the validation of track as well as free field vibrations during a train passage. A comparison of experimental and computed results for the track receptance and the transfer functions between the track and the free field can be found elsewhere [4]. The model accounts for the dynamic interaction between the vehicle and the track by means of a compliance formulation in the frame of reference that moves with the vehicle [5]. The railway track is modelled as a longitudinally invariant system, and is assumed to be located at the surface of a horizontally layered elastic halfspace. The translational invariance of the geometry enables an efficient solution in the frequency– wavenumber domain. The model is not reviewed in the following; the reader is referred to a recent journal paper for more details [4].

2

The Characteristics of the Train, the Track and the Soil

2.1 The Dynamic Train Characteristics The Thalys HST consists of 2 locomotives and eight carriages and has a total length of 200.18 m. It is an articulated train with shared bogies between carriages, except for the locomotives. The Thalys HST has 13 bogies and 26 axles. The carriage length Lt , the distance Lb between bogies, the axle distance La , the total axle mass Mt , the sprung axle mass Ms and the unsprung axle mass Mu of all carriages are summarized in table 1. Table 1. Geometrical and mass characteristics of the Thalys HST Lt [m] Locomotives 22.15 Side coach 21.84 Central coach 18.70

Lb [m] 14.00 18.70 18.70

La [m] 3.00 3.00 3.00

Mt [kg] 17000 17000 17000

Ms [kg] 14937 14937 14937

Mu [kg] 2027 2027 2027

Table 2. Geometrical and mass characteristics of the IC train Lt [m] Locomotive 19.11 Central coach 26.40 Back coach 26.40

Lb [m] 10.40 18.40 18.40

La [m] 3.00 2.56 2.56

Mt [kg] 22500 11610 11830

Ms [kg] 19677 10100 10286

Mu [kg] 2823 1500 1544

The Influence of the Train Speed on Vibrations Due to High Speed Trains

21

The IC train consists of one locomotive, 7 standard central coaches and 1 additional coach on the back. The locomotive and every coach have two independent bogies and four axles. The geometrical and mass characteristics of the IC train are summarized in table 2. 2.2 The Dynamic Soil Characteristics The test site is located in Lincent along the new HST track on the line L2 between Brussels and K¨oln. Five seismic cone penetration tests (SCPT) and two spectral analysis of surface wave (SASW) tests have been performed to determine the soil layering and the small strain dynamic soil characteristics. From the SASW tests, it is found that the site consists of a layer with a thickness of 3.0 m and a shear wave velocity of 150 m/s on top of a halfspace with a shear wave velocity of 280 m/s. The results of the SCPT tests confirm these observations. A fit between the experimental and numerical soil transfer functions has been used to estimate a value of 0.03 for the material damping ratio β of all soil layers in deviatoric and volumetric deformation. A value of 1/3 has been assumed for the Poisson’s ratio and a density ρ equal to 2000 kg/m3 . 2.3 The Dynamic Track Characteristics The track in Lincent is a ballasted track with UIC 60 rails supported every 0.60 m by monoblock concrete sleepers. The rails are continuously welded and are fixed with a Pandrol E2039 rail fastening system and supported by resilient studded rubber rail pads with a thickness of 11 mm. Each rail pad is preloaded with a clip toe load of about 20 kN per rail seat. The track gauge is 1.435 m. The track is modelled as a longitudinally invariant single track (figure 1), where the dynamic stiffness of the rail pads and the mass of the sleepers are uniformly distributed along the track. The width of the track-soil interface is 2.5 m and corresponds to the length of the sleepers. The rails are modelled as Euler-Bernoulli beams with a bending stiffness Er Ir = 6.45×106 Nm2 and a mass per unit length ρr Ar = 60.3 kg/m. The rail

lsl

l1

ur2

ur1 kp

kb

l2

cp

usl βsl

cb

us 2b

βs

Fig. 1. Cross section of a ballasted track model

rail rail pad sleeper ballast interface Σ

22

G. Lombaert, G. Degrande, and J. Bekaert

pads are modelled as continuous spring–damper connections. The stiffness and damping coefficients krp and crp of a single rail pad are used to calculate equivalent stiffness and damping coefficients k rp = krp /d and crp = crp /d, where d is the sleeper distance. The uniformly distributed mass msl of the sleepers is equal to msl /d or 500 kg/m. A porphyry ballast layer with a height hb of 0.35 m and a density ρb of 1700 kg/m3 is included in the track model. The ballast bed acts as a set of distributed, independent linear springs and dampers, and each sleeper is only supported by that part of the ballast in contact with the sleeper. The sleeper is usually only supported under the rails, so that the vertical spring stiffness kb per sleeper [N/m] is calculated from the effective support length esl per rail, the sleeper width bsl and the ballast stiffness Kb [N/m3 ] as 2esl bsl Kb . The smeared ballast stiffness k b [N/m2 ] is equal to kb /d. The equivalent ballast mass mb is calculated from the part of the ballast in contact with the sleepers as ρb hb lsl bsl /d. The dynamic stiffness k b and damping cb of the ballast and the equivalent stiffness k rp = krp /d and damping crp = crp /d of the rail pad are obtained by means of a rail receptance test, during which the track has been pre-loaded with a locomotive and a dynamic excitation has been applied on the rail at a sleeper position. The ballast stiffness kb and damping cb are estimated as 1534.5 × 106 N/m2 and 27.7 × 103 Ns/m2 , corresponding to a ballast stiffness kb per sleeper of 920.7 × 106 N/m and damping cb of 16.6 × 103 Ns/m. The equivalent rail pad stiffness k rp is estimated as 255.7 × 106 N/m2 , while a value of 22.5 × 103 Ns/m2 is found for the rail pad damping crp . This corresponds to a rail pad stiffness krp = 153.4 × 106 N/m, as for medium to stiff rail pads, and a damping crp = 13.5 × 103 Ns/m.

3 The Free Field Response during the Passage of a Thalys HST 3.1 The Track Response

0.005 0

0

20

40

60

80 Frequency [Hz]

100

120

0.015 TS02z

0.005 0

0

20

40

60

80 Frequency [Hz]

100

120

4

5

6

7 Time [s]

8

9

10

11

12

TS02z

2

3

4

5

6

7 Time [s]

8

9

10

11

12

0.05 TS02z

0.01 0.005 0

3

0

−0.05

140

0.015

(c)

2

0.05

0.01

(b)

TS02z 0

−0.05

140

Velocity [m/s]

Velocity [m/s/Hz]

(a)

Velocity [m/s/Hz]

0.05 TS02z

0.01

Velocity [m/s]

0.015

0

20

40

60

80 Frequency [Hz]

100

120

140

Velocity [m/s]

Velocity [m/s/Hz]

The numerical model is now used to compute the sleeper response during the passage of an IC train at a speed of 224 km/h and a Thalys HST at a speed of 218 km/h and

TS02z 0

−0.05

2

3

4

5

6

7 Time [s]

8

9

10

11

12

Fig. 2. The experimental (grey line) and computed (black line) frequency content (left) and time history (right) of the vertical sleeper velocity during (a) an IC train passage at v = 224.1 km/h (push mode) and a Thalys train passage at (b) v = 218.1 km/h and (c) v = 306.8 km/h

The Influence of the Train Speed on Vibrations Due to High Speed Trains

23

307 km/h. Only the quasi-static contribution, that depends on the total axle mass Mt (tables 1 and 2), is taken into account. Figure 2a shows the frequency content and time history of both the measured and computed sleeper velocity during the passage of the IC train at 224 km/h. A modulation of the frequency content is observed which is due to the repitition of the same (scaled) response for each axle. The modulation is therefore determined by the total mass per axle (Mt ), the distance between the axles and the train speed v. An excellent agreement between the measured and computed results is obtained at low frequencies, while the experimental response is slightly overestimated between 20 and 60 Hz and underestimated at higher frequencies, due to the neglection of the dynamic axle loads. The identification of the track characteristics in loaded track conditions has proven to be crucial for a good prediction of the sleeper response. In the time history of the response, larger vibration levels are observed during the passage of the locomotive, which has a much larger total mass Mt per axle. Figures 2b and 2c show similar results for the passage of the Thalys HST at speeds of v = 218 km/h and v = 307 km/h. In the case of the Thalys HST, the total mass Mt per axle is the same for all axles, as can be observed from the time history. For similar train speeds, the difference between the response due to the passage of the IC (figure 2a) and Thalys train (figure 2b) can be attributed to the difference in total mass per axle. When the results in figures 2b and 2c are compared, it can be observed how a higher train speed shifts the modulation of the frequency content to higher frequencies and increases the response for each axle in the time domain. 3.2 Validation of the Free Field Response Free field vibrations during the passage of the IC train and the Thalys HST are considered next. As the train speeds are relatively low compared to the shear wave velocity of 150 m/s in the top layer, quasi-static axle loads are neglected and only dynamic axle loads are accounted for. The dynamic axle loads are determined from the solution of the dynamic train-track-soil interaction problem. At frequencies of more than a few Hertz, the vehicle’s primary and secondary suspension isolate the body and the bogie from the wheelset and only the vehicle’s unsprung mass affects the vertical dynamic loads. The influence of the primary suspension is small at frequencies higher than 10 Hz, so that each axle can be represented by its unsprung mass Mu (tables 1 and 2). Figures 3 and 4 compare the measured and computed frequency content and time history of the vertical free field velocity at 8 m and 48 m from the track, during the passage of the IC train at a speed of 224 km/h and a Thalys HST at a speed of 218 km/h and 307 km/h. The frequency content is mainly situated in a range upto 100 Hz (figure 3), with a shift towards lower frequencies for an increasing distance from the track (figure 4). For a similar train speed, similar vibration levels are found for the IC (figure 3a and 4a) and Thalys train (figure 3b and 4b). When the results in figures 3b and 4b are compared with those in figures 3c and 4c, an increase of the vibrations with the train speed is observed. At 8 m from the track, the passage of individual bogies can be identified (figure 3), which is no longer the case at 48 m (figure 4). The duration of the experimental and computed results increases in a similar way for decreasing train speed. At small distances from the track, the response is overestimated, while the agreement is better at larger distances.

24

G. Lombaert, G. Degrande, and J. Bekaert

(a)

−3

0.5

0

x 10

2 FF02z

0

20

40

60

80 Frequency [Hz]

100

120

Velocity [m/s]

Velocity [m/s/Hz]

−3

x 10

1

0 −1 −2

140

FF02z

1

2

(b)

1

2 FF02z

0.5

0

0

20

40

60

80 Frequency [Hz]

100

120

7 Time [s]

8

9

10

11

12

FF02z

−1

2 FF02z

0

6

2

3

4

5

6

7 Time [s]

8

9

10

11

12

−3

x 10

0.5

0

5

0

−2

140

20

40

60

80 Frequency [Hz]

100

120

Velocity [m/s]

Velocity [m/s/Hz]

(c)

4

x 10

1

−3

1

3 −3

Velocity [m/s]

Velocity [m/s/Hz]

−3

x 10

FF02z

1 0 −1 −2

140

x 10

2

3

4

5

6

7 Time [s]

8

9

10

11

12

Fig. 3. The experimental (grey line) and computed (black line) frequency content (left) and time history (right) of the vertical free field velocity at 8 m from the track during (a) an IC train passage at v = 224.1 km/h (push mode) and a Thalys train passage at (b) v = 218.1 km/h and (c) v = 306.8 km/h

(a)

−4

0.5

0

x 10

2 FF07z

0

20

40

60

80 Frequency [Hz]

100

120

Velocity [m/s]

Velocity [m/s/Hz]

−4

x 10

1

0 −1 −2

140

FF07z

1

2

(b)

1

2 FF07z

0.5

0

0

20

40

60

80 Frequency [Hz]

100

120

7 Time [s]

8

9

10

11

12

FF07z

−1

2 FF07z

0

6

2

3

4

5

6

7 Time [s]

8

9

10

11

12

−4

x 10

0.5

0

5

0

−2

140

20

40

60

80 Frequency [Hz]

100

120

140

Velocity [m/s]

Velocity [m/s/Hz]

(c)

4

x 10

1

−4

1

3 −4

Velocity [m/s]

Velocity [m/s/Hz]

−4

x 10

x 10

FF07z

1 0 −1 −2

2

3

4

5

6

7 Time [s]

8

9

10

11

12

Fig. 4. The experimental (grey line) and computed (black line) frequency content (left) and time history (right) of the vertical free field velocity at 48 m from the track during (a) an IC train passage at v = 224.1 km/h (push mode) and a Thalys train passage at (b) v = 218.1 km/h and (c) v = 306.8 km/h

4 Conclusion The experimental validation of a numerical model for the prediction of railway induced vibrations has been presented. The model accounts for the dynamic interaction between the train, the track and the soil by means of a compliance formulation in the moving frame of reference. The results of track as well as free field vibrations during the passage of an IC and Thalys train are used to validate the model. The quasi-static contribution is used to calculate the track response, while only the dynamic contribution is taken into account for

The Influence of the Train Speed on Vibrations Due to High Speed Trains

25

the free field response. Given the large number of modelling uncertainties, the predicted vibrations show a good agreement with the experimental results.

References [1] Sheng, X., Jones, C.J.C., Thompson, D.J.: A theoretical model for ground vibration from trains generated by vertical track irregularities. Journal of Sound and Vibration 272(3–5), 937–965 (2004) [2] Auersch, L.: The excitation of ground vibration by rail traffic: Theory of vehicle-track-soil interaction and measurements on high-speed lines. Journal of Sound and Vibration 284(1–2), 103–132 (2005) (accepted for publication in press) [3] Metrikine, A.V., Verichev, S.N., Blauwendraad, J.: Stability of a two-mass oscillator moving on a beam supported by a visco-elastic half-space. International Journal of Solids and Structures 42, 1187–1207 (2005) [4] Lombaert, G., Degrande, G., Kogut, J., Franc¸ois, S.: The experimental validation of a numerical model for the prediction of railway induced vibrations. Journal of Sound and Vibration 297(3–5), 512–535 (2006) [5] Clouteau, D., Degrande, G., Lombaert, G.: Numerical modelling of traffic induced vibrations. Meccanica 36(4), 401–420 (2001)

High Speed Train Noise Effects on Wildlife and Domestic Livestock C.E. Hanson Harris Miller Miller and Hanson Inc., 77 South Bedford Street, Burlington, MA 01803, USA Tel.: +1 781 229 0707; Fax: +1 781 229 7939 [email protected]

Summary The planned introduction of high-speed trains in the US requires the assessment of potential corridors for such systems. U.S. Department of Transportation, Federal Railroad Administration (FRA) recognized that noise and vibration issues are frequently among the potential impacts of most concern to residents. In contrast to some countries, city pairs are quite far apart in the United States, which brings the potential for high-speed train operations in rural and wilderness areas. Noise from high-speed train pass-bys in rural and wilderness areas may have effects on animals. Concerns have been raised by environmentalists who suggest high noise levels adjacent to a rail corridor can cause impact such as interference with communication, masking predation, startle and fright. FRA has adopted criteria for identifying the potential impact of such effects, based on research on aircraft over-flights with similar noise signatures. This paper summarizes the research, makes comparisons with high-speed train noise signatures and proposes criteria.

1 Introduction A wide range of studies has been conducted concerning noise effects on animals. For humans, annoyance is considered to be the primary environmental effect. Thresholds of annoyance in terms of sound exposure have been determined by surveys throughout the world. However, for animals, the effects are not easily determined. Much of the research on noise effects on animals relates to aircraft over-flights, especially in military training areas. Some research on highway noise effects has been performed. But so far nothing has been reported on the potential effects of high-speed train noise on wild or domestic animals. As more high-speed rail systems are developed, there is a potential for tracks being laid through normally quiet wilderness areas where the intrusive noise from train pass-bys could result in a permanent change in the environment. In order to quantify the potential impact of a new rail alignment through rural and wilderness areas, the Federal Railroad Administration (FRA) of the United States Department of Transportation developed preliminary criteria for the effects of high-speed train noise on wildlife and domestic livestock.

2 Controversy There is disagreement among interested parties whether noise from high-speed trains is an important issue. It is fair to consider the arguments on both sides of the issue. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 26–32, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

High Speed Train Noise Effects on Wildlife and Domestic Livestock

27

2.1 Sceptics Say “No Effect” A sceptic’s point of view is based on observations that habituation to transportation noise is common among wildlife. The following examples are often cited as proof. Birds, especially starlings, congregate around airports despite airport authorities’ best efforts to discourage their presence for fear of ingestion into jet engines or damage to propellers. Vineyard owners have a long history of trying to develop sounds that scare birds and other animals who come to feed on valuable crops. Loud bangs similar to gunshot sounds appear to work for a short time, but rapidly lose their effectiveness. Invasion of wildlife such as bears, coyotes, deer, moose and raccoons is not uncommon in cities of the United States. Animals happily live near highways, railroad and maglev tracks: deer are plentiful along super-highways in the United States, antelope herds roam the FRA’s railroad test track complex in Colorado, bears feed on spilled grain near railroad tracks in Canada, and cows are seen grazing peacefully next to the maglev test track in Germany. 2.2 Environmentalists Say “Effect” On the other hand, environmentalists insist that transportation noise has a serious effect on wildlife and domestic animals. Among the reasons given are the following. Loud noise from any source interferes with communication, especially bird calls. External noise tends to mask predation, causing interference for both the predator and the potential prey. Sudden loud noises frighten wild and domestic animals, just as it does humans, and brings out the “fight or flight” response. Upon fleeing the source of noise, animals can be injured during an impulsive reaction or a stampede of a herd. 2.3 Researchers Say “Maybe There Is an Effect” Research has been inconclusive regarding noise effects on wild and domestic animals from transportation noise sources. Most of the research has focused on low-flying aircraft. Responses are well-documented. Since there are observed effects from lowflying aircraft, there may be similar effects from high-speed trains. The reasoning is based on similarities in sound signatures between low-flying aircraft and high-speed trains, including rapid onset rates and maximum sound levels. Unfortunately, the lack of uniformity in describing the noise sources in many of these studies has weakened any conclusions to be drawn, and the long-term effects remain a matter of speculation.

3 Studies of Noise Effects on Animals in the Outdoors Many studies have been conducted on hearing response of laboratory animals, but relatively little work has been done regarding noise effects on animals in the natural environment. Much of the research has been done in an effort to determine the effects of low-flying military aircraft on training flights over wilderness areas. Some research on road noise effects has been sponsored by Federal Highway Administration and state highway departments in the United States in connection with environmental impact assessments. General studies have been conducted by various universities throughout the world. Surveys of current information on this topic are available on the internet.

28

C.E. Hanson

3.1 Aircraft Noise Effects Most of the research on noise effects in the natural environment has been aimed at determining the response of birds and animals to aircraft noise. Studies have been conducted on reindeer, caribou, antelope, bears, rodents, waterfowl, raptors, and songbirds in the wild; and cows, goats and swine on ranches and farms [1] [2] [3] [4] [5]. Usually the studies involved the introduction of a specific noise event like an aircraft over-flight or a recording of same with a subsequent observation of animal response. Observations of response to aircraft noise in the wild have ranged from no reaction or mild response such as slight changes in body position to extreme responses such as panic and attempts to escape. Few of the wilderness studies documented the actual noise levels that were generated by aircraft, which diminishes the usefulness of the observations for developing criteria. Similarly, effects on domestic animals were often measured in terms of productivity, such as milk yields, hormonal changes, or heart rate variation in response to aircraft noise, but, again, the noise descriptors have not been standardized among the various research agencies. One study stands out in terms of documenting actual noise levels related to effects on domestic turkeys [6]. Many of the effects relevant to productivity (weight gain, mortality and meat quality) could be related to the Sound Exposure Level (SEL), with a level of 100 dB coming up as a threshold of effect. 3.2 Highway Noise Effects A recent review of road noise effects on wildlife populations was conducted by the Federal Highway Administration in the United States [7]. Studies of road noise on invertebrates, fish, reptiles, amphibians, birds, and mammals were included in the review. In most cases, the studies were flawed by lack of noise measurements to quantify the effects observed. Audible interference with bird calls was often cited as an effect from road noise. Mammals tended to avoid roads with heavy traffic, but the road acting as a physical barrier may have been a confounding variable in many of the studies. 3.3 General Noise Effects A wide variety of sounds besides aircraft flyovers and highway noise has been used to elicit responses from animals [2][3]. Some of these sources include white noise, music, pure tones, exploding paper bags, motorboat noise, bird calls, and distress calls. Once again, the lack of uniformity in testing protocols and measurement make it difficult to compare the results among the various studies. 3.4 Limitations of Research on Noise Effects Conclusions to be drawn from the existing research are limited for several reasons: • •

Confounding disturbance factors – the visual effect of low-flying aircraft in the wild may outweigh the auditory effect. Noise levels seldom quantified – most studies adequately described the source of noise and the animal response, but the actual noise levels on the ground were unknown or roughly estimated.

High Speed Train Noise Effects on Wildlife and Domestic Livestock

• •

29

Observers not trained in acoustics – levels, frequency content, duration often not reported. Wide variety of species – hearing acuity differences among species prevents uniform treatment.

Despite the above mentioned limitations, there are some commonalities to the research results that can be viewed as useful in establishing criteria. A summary of some relevant studies is shown in Table 1. Table 1. Summary of Noise Levels Associated with Effects on Animals and Birds. [1][5].

Animal Category

Domestic mals

Species

Mam- Dairy Cow

Swine

Sheep

Wild Mammals

Domestic Birds Wild Birds

Reindeer Caribou Antelope Grizzly bears Dall sheep Chicken Turkey Quail Canary Seabirds (general) Tern California Condor Eagles Falcons

Noise Level Descrip- Effect tor Associated with Effect 105 dB Reduction in milk production 97 dB Changes in blood composition 110 dB, 1kHz Changes in blood composition 108 – 120 dB Hormonal changes 93 dB Hormonal changes 120 – 135 dB Increased heart rate 100 dB “white noise” Increased heart rate, respiration 90 dB “white noise” Decreased thyroid activity 100 dB, 4 kHz Increase in lambs per ewe Sonic booms Startle Aircraft noise Startle, panic running 77 dBA, helicopter Running Aircraft, low altitude Startle, running >95 dB, military jets 115 dB, aircraft 100 dB SEL 80 dB 95- 100 dB Sonic booms

Disturbed, running Interrupt brooding Panic crowding Accelerated hatching Hearing loss Statle, flush from nest

Sonic booms Blasting, drilling, etc.

Reduced reproduction Flush from nest, abandon area Alarm Nest abandonment

Sonic booms 100 dB SEL

30

C.E. Hanson

4 Comparisons between High-Speed Train Noise and Low-Flying Aircraft Noise 4.1 Similarities If the research results from low-flying aircraft are to be applied to high-speed trains, there should be similarities in the noise characteristics of the two types of sources. The comparisons are restricted to subsonic low-flying aircraft; sonic booms from supersonic aircraft are significantly different from the noise generation from high-speed trains. Following are some of the similarities: • • •

Rapid onset of noise time history – rise times more than 30 dB per second. Frequency spectra – maximum sound levels in the 50 Hz to 500 Hz range. High sound levels – high-speed train levels can be 95 dBA or greater at 25 meters; low-flying aircraft generate 100 dBA or greater at ground level.

These similarities suggest that noise from high-speed train pass-bys could elicit similar responses from animals in their natural habitat if they are similarly exposed. 4.2 Differences While similarities in noise characteristics occur, there are some profound differences between high-speed trains and low-flying aircraft. These differences include: • • •

Routes – high-speed trains have a fixed alignment; aircraft vary their flight corridors. Schedules – high-speed train pass-by on a regular schedule; aircraft highly variable. Exposure – high-speed trains have repeatable levels with relatively long duration; low-flying aircraft have variable levels with short durations.

These differences may limit the application of current research to the development of criteria relating to startle effects and habituation.

5 Proposed Criteria for High-Speed Trains The FRA has proposed criteria for the impact of high-speed train noise on animals in wilderness and farming areas [8]. Following are the steps leading to development of the proposed criteria. 5.1 Noise Descriptor A descriptor for noise effects on wildlife and domestic animals has not been adopted. Few of the studies used a uniform approach to measuring and assessing the effects. One of the best documented studies indicated that the sound exposure level (SEL),

High Speed Train Noise Effects on Wildlife and Domestic Livestock

31

which accounts for both maximum noise level and duration of the event, is the most useful predictor of responses [6]. What frequency weighting to use is the next question. Environmental noise is commonly expressed in terms of dBA, but that frequency weighting may not apply to animals since their hearing thresholds differ form humans. However, since no uniform weighting has been established for representing the hearing characteristics of wild animals, the A-weighted sound pressure level continues to be recommended. 5.2 Threshold for Disturbance Observation of response to noise levels in all the research studies could be expressed in terms of some sort of disturbance, even if the effects varied widely from study to study. The aforementioned well-documented study on noise effects on turkeys identified a level of SEL = 100 dB as a significant threshold of disturbance. Although the descriptors were not the same as shown in Table 1, many of the studies report levels in the vicinity of 100 dB as associated with an observable effect. Until more definitive information can be developed, the preliminary criterion of SEL=100 dBA is the proposed criterion for disturbance by high-speed train operations. 5.3 Habituation Some of the research studies indicate that some animals habituate to noise after several repetitions of exposure. Previous exposure to noise levels below 100 dB served to eliminate panic among turkeys, and swine showed initial alarm followed by indifference to aircraft noise greater than 100 dBA. In contrast, some species apparently do not become accustomed to high noise levels. Since habituation is found to be speciesdependent, a general criterion cannot be proposed at this time.

6 Conclusions There has been a demand on the part of environmental interest groups in the United States to consider the effect of noise from high-speed trains on wildlife and domestic animals in their natural habitat. Despite the lack of relevant research, FRA has been able to adapt the research of noise effects from low-flying subsonic aircraft to the potential of similar effects of noise from high-speed trains. Preliminary criteria have been developed as: • • • •

Noise metric – A-weighted sound pressure level (dBA) Noise descriptor – sound exposure level (SEL) Threshold for impact – 100 dBA Habituation – no general criterion.

Acknowledgements The development of this preliminary criterion for high-speed train noise effects on wildlife and domestic animals was performed by Harris Miller Miller & Hanson Inc.

32

C.E. Hanson

under contract to Parsons Transportation Group and the U.S. Department of Transportation, Federal Railroad Administration. David Valenstein served as FRA’s Program Manager.

References [1] Manci, K.M., Gladwin, K.N., Villella, R.: Effects of Aircraft Noise and Sonic Booms on Domestic Animals and Wildlife: A Literature Synthesis. In: U.S. Fish and Wildlife Service, National Ecology Research Center, Colorado, NERC-88/29 (1988) [2] U.S. Environmental Protection Agency, Effects of Noise on Wildlife and Other Animals, NTID300.5 (1971) [3] U.S. Environmental Protection Agency, Effects of Noise on Wildlife and Other Animals: Review of Research Since 1971, Report No. 550/9-80-100 (1980) [4] Larkin, R.P., Pater,L.L., Tazik, D.J.: Efffects of Military Noise on Wildlife: A Literature Review, U.S. Army Corps of Engineers, Construction Engineering Research Laboratories, Technical Report 96/21 (1995) [5] Roby, D.D., Murphy, S.M., Ritchie, R.J., Smith, M.D., Palmer, A.G.: Nordmeyer-Elmore, D., Pruitt, E., Krull, R.C.: The Effects of Noise on Birds of Prey: A Study of Peregrine Falcons in Alaska, U.S. Air Force Research Laboratory, Report No. AFRL-HE-WP-TR2002-0190 (July 2002) [6] Bradley, F, Book, C., Bowles, A.E.: Effects of Low-altitude Aircraft Overflights on Domestic Turkey Poults, U.S. Air Force, Noise and Sonic Boom Impact Technology Program, Report No. HSD-TR-90-034 (June 1990) [7] Kaseloo, P.A., Tyson, K.O.: Synthesis of Noise Effects on Wildlife Populations, Office of Research and Technology Services, Federal Highway Administration, Report No. DTFH61-03-H-00123 (September 2004) [8] U.S. Department of Transportation, Federal Railroad Administration, High-Speed Ground Transportation Noise and Vibration Impact Assessment, Office of Railroad Development (October 2005), http://www.fra.dot.gov/us/content/253

Wind Tunnel Tests on the Control of Aeroacoustic Noise from High Speed Train N. Yamazaki1, T. Takaishi1, M. Toyooka2, K. Nagakura1, A. Sagawa1, and H. Yano2 1

Railway Technical Research Institute, Tokyo, 185-8540, Japan Tel.: +81 42 573 7353; Fax: +81 42 573 7418 [email protected] 2 West Japan Railway Company, Osaka, 530-8341, Japan

Summary In this study, we propose techniques for reducing the noise from gaps between Shinkansen cars based on the results of noise source localization in wind tunnel testing. In order to obtain the accurate noise source distributions, the microphone array is installed near the train model. The influence of the shear layer around the main flow on the directivity of the microphone array is clarified so that the microphone array should be set in the shear layer rather than the outside of the flow. Analysis of the noise source localization reveals the principal noise sources around the gap, which suggests efficient approaches to the noise reduction. Firstly, we found that the noise level of the gap section with rounded edge can be effectively reduced by approximately 7 dB compared with that of the case with current condition. We also confirmed qualitatively the effect of noise reduction techniques for the gap section by the field test.

1 Introduction The Japanese government’s Environmental Quality Standards for Shinkansen Superexpress Railway Noise prescribes that the maximum A-weighted sound pressure level (slow) of a train set passing LpA,Smax at the wayside of the track shall be 70 dB(A) or less in areas that are mainly residential, and 75 dB(A) or less in other areas including commercial and industrial areas, where normal living conditions should be preserved, as referred to in [1]. Although spark noise from the pantograph and rolling noise were the typical noise sources when Shinkansen commenced operation in October 1964, most of these sounds have now been effectively reduced through a number of noise reduction techniques. However, the new issue of aerodynamic noise has attracted a considerable amount of attention lately. This type of noise is generally caused by flow disturbance around uneven areas on the train body, and its power increases in proportion to the 6th powers of train speeds. A number of aerodynamic approaches have proposed optimized shapes for noise reduction such as the streamlined pantographhead [2]. Against this background, the contributions of the noise from gaps between adjacent cars to the environmental noise has been a relative increase. The Railway Technical Research Institute (RTRI) used the Large-scale Low-noise wind tunnel in Maibara City (MWT) to study the generation mechanisms of aeroacoustic B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 33–39, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

34

N. Yamazaki et al.

noise and techniques for its reduction. As a basis for the process toward achieving proper noise reduction techniques, accurate noise source localization is essential to determine which parts are responsible for noise generation, especially in the case of using small-scale model in wind tunnel testing. In this study, we first attempted to improve the accuracy of noise source localization by installing a microphone array near the model, and closely examined the influence of the free shear layer around the main flow on the acoustic characteristics. Secondly, noise reduction techniques for the gap between adjacent cars were presented on the basis of noise source localization results. Finally, the effect of noise reduction technique for the gap section was verified in a field test as a joint work between RTRI and West Japan Railway Company.

2 Wind Tunnel Tests 2.1 Development of a Microphone Array Adjacent to the Main Flow In taking measures to reduce aerodynamic noise, accurate noise source localization at the initial stage is of great importance. Particularly high spatial resolution and accurate position estimation are essential in investigating noise distribution around a narrow gap section. In the MWT, noise source distribution has so far been obtained by using directional microphone systems such as an acoustic mirror and an X-shaped microphone array specially designed for wind-tunnel testing ([3] and [4]). In particular, the microphone array is useful because it enables the desired focal point to be set and measurement time to be shortened. In the signal processing of the microphone array, the acoustic phase is corrected on the basis of the distance from the focal point to each microphone [5]. Therefore it is necessary to estimate the exact propagation distance for accurate source localization.

Ya

:LQGYHORFLW\ :LQGYHORFLW\ D 

E 

Fig. 1. Experimental setup for performance tests of the microphone array with a standard source device (a) Entire view of the test section and coordinate system (b) Comparison of location error of the microphone arrays installed at Ya=1,500 mm (near flow microphone array) and 5,000 mm (X-shaped microphone array). ■ Ya=5000 mm, ■ Ya=1500 mm.

Wind Tunnel Tests on the Control of Aeroacoustic Noise from High Speed Train

35

The conventional directional microphone systems mentioned above are installed in the area outside the free shear layer, which includes strong turbulence. However, it is difficult to accurately predict the sound propagation path precisely in practice because the free shear layer changes its width and flow velocity with respect to time and space. This problem results in a deterioration of the source localization accuracy. Furthermore, the signal-to-noise ratio decreases due to the acoustic scattering in the free shear layer. We therefore designed a microphone array to be installed adjacent to the main flow (referred to as a near-flow microphone array) so as to avoid being affected by the free shear layer. We conducted performance testing to investigate the relationship between the directivity of the microphone array and its setting position in the free shear layer. Figure 1 shows the experimental test set-up using a standard source device, which was installed at X = 1,750 mm and Y = 0 mm. The microphone array was installed at X = 2000 mm and distant from the centreline by Ya = 1500mm. 64 microphones were arranged in a spiral pattern and mounted flush on the board surface with a width of 1.3 m and the height. Flow-induced noise was suppressed by the cross spectrum method obtained by the outputs of two pairs of microphone arrays [6]. The board was also coated with urethane foam to a thickness of 10 mm. The analysis region is a square of 600 mm x 600 mm, and its lattice size is adjusted from 1 mm to 10 mm depending on the desired accuracy. Figure 1 (b) shows a comparison of source location errors measured using the near-flow microphone array installed at Ya = 1500 mm in the free shear layer and Xshaped microphone array [4] installed outside at Ya = 5000 mm. The accuracy of source localization using the near-flow microphone array is improved to a location error of better than 10 mm, which is suitable for narrow areas such as the gap sections of a scaled train model. 2.2 Wind Tunnel Testing Using a 1/8 Scale Train Model 2.2.1 Experimental Method We investigated the noise characteristics of the gap sections using a 1/8-scale train model. Figure 2 shows the experimental set-up of the train model and microphones. Based on the results outlined in Section 2.1, we installed the microphone array at Ya = 1,500 mm as shown in Fig.2 (a). The A-weighted sound pressure level (the noise level) was also measured using an omni-directional microphone. As shown in Fig. 2 (b), it was set Y = 3,125 mm from the nozzle centre to correspond to the typical noise assessment point of Shinkansen trains (i.e. the 25-m point) on a real scale. Frequency f appearing in this chapter is expressed as a frequency corrected to the real scale. Figure 3 show the test conditions of the gap. The train model consisted of the first car and a shortened second car with lengths of 3,275 mm and 1,087.5 mm respectively. These cars were set at an interval of 62.5 mm, corresponding to 1/8 of the real gap interval. The model had supports of 50 mm in height, and was installed on a rail 50 mm above the floor. This interval of 100 mm between the floor and the base of the train model corresponds to twice the distance between the top of the rail and the ground on a real scale. The sections below outline the details of the measures against noise reduction. Noise source distributions around the gap were analysed on a plane that included the train body side and the wheels.

36

N. Yamazaki et al.

1/8 scale train model

Microphone array

Omni-directional microphone

D 

E 

Fig. 2. Experimental setup for noise measurements in the open test section of Large-Scale LowNoise Wind Tunnel. (a) Setup for noise source localization tests with microphone array. (b) Setup for noise level measurements with omni-directional microphone.

Rounded edge Outer hood

D 

E 

F 

Fig. 3. Shapes of gap section of 1/8 scaled train model. (a) Current condition with outer hood. (b) Case without outer hood (c) Case with rounded edge shape.

The difference between the shear layer thickness of the model and a real train, which affects the aeroacoustic noise generation, should be examined in advance. The thinner shear layer around the 1/8-scale model can cause strong tonal noise from the gap. From previous studies in which pressure fluctuations inside a gap section were examined, it is known that treatment involving roughness or wire on the upstream edge is effective in controlling the shear layer thickness. We therefore placed a wire with a diameter of 2 mm at point 50 mm upstream from the edge of the first car. 2.2.2 Noise Reduction Techniques for the Gap Section In the current condition, the outer hood covers the gap up to its shoulder as illustrated in Fig. 3 (a). Shown in Fig. 4 (a) is the noise source distribution around the gap with the outer hood in an f = 315 Hz one-third octave band. It is clear that a noise source exists below the hood, and the phenomenon is observed more clearly without the outer hood. Fig.4 (b) shows a strong noise source along the downstream edge. This result suggests that the impinging vortex motion on the downstream edge is related to strong aeroacoustic noise generation. Sharp edges should also be considered as strengthening acoustic wave generation because of their acoustic radiation efficiency.

Wind Tunnel Tests on the Control of Aeroacoustic Noise from High Speed Train

37

10dB

Wind Direction

(a)

(c)

(b)

Fig. 4. Sound pressure level distributions around the gap (f = 315Hz one-third octave band) (a) Case with outer hood (b) Case without outer hood (c) Case with rounded edge shape

D 

E

Fig. 5. Experimentally measured A-weighted sound pressure level of the gap section obtained using an omni-directional microphone. (a) Results without background noise correction (b) Results with background noise correction. - -Current condition, - - Case without outer hood, - - Case with rounded edge, - - Smoothed condition (background noise).

□

◇

△

●

Although several measures to counter these noise problems have been adopted so far, most of the focus has been on conventional solutions such as smoothed shapes, exemplified by a cover-all hood to completely enclose the gap [7]. However such techniques seem to present difficulties in terms of maintenance and durability. We therefore examined noise reduction techniques involving alteration of the shape of the train itself, and on the basis of several attempts, a rounded-edge shape was found to be effective. Figure 3 (c) shows a gap featuring the rounded edge. The radius of the curvature at the edge is 25 mm on the model scale and 200 mm on the real scale. As shown in Fig. 4 (c), the noise source along the downstream edge is considerably less than that found in the case of a right-angled edge. These results suggest the following three noise reduction factors: firstly, the turbulence level of the shear layer around the gap is reduced; secondly, the impact of the impinging vortex on the downstream edge is lessened; thirdly, the acoustic radiation efficiency is reduced. Figures 5 (a) and (b) indicate the noise level measured using the omni-directional microphone and the results with background noise correction. It is clear that the noise level with the rounded

38

N. Yamazaki et al.

edge decreases in entire frequency region except for 80Hz band and almost the same with the case of smoothed condition. According to the results shown in Fig. 5 (b), the noise level by the rounded edge shape is estimated to be approximately 7 dB lower than that of the current condition.

3 Field Tests In the previous chapter, it was found that the rounded edge shape is suitable to reduce the noise level from the gaps between adjacent cars. The noise reduction effect by these techniques was verified in the field test as a joint work between RTRI and West Japan Railway Company. Figure 6 shows the countermeasures applied to a gap section between 7th and 8th cars of real Shinkansen. The rounded edges between adjacent cars were connected along the whole side with recessed diaphragm except for the floor.

(a)

(b)

Fig. 6. Photograph of the gap section of Shinkansen train used for the field tests. (a) Current gap condition. (b) Gaps with rounded edge shape. An enlargement is shown in right figure. 10dB High

Low Current shape

6th car

7th car

Rounded edge shape

7th car

8th car

Train direction U = 300 km/h Current shape

8th car

9th car

Fig. 7. Noise source distributions around the Shinkansen train (f = 250 Hz octave band). The countermeasures of the rounded edge are performed for the gap section between 7th and 8th cars.

The noise source distributions around the train were measured with a microphone array specially designed for the field test. The distance between the microphone array, which consists of 64 microphones, and the side of train car body was 9 m. These microphones were arranged so widely with 9 m width and 5.1 m height that the directivity

Wind Tunnel Tests on the Control of Aeroacoustic Noise from High Speed Train

39

of the microphone array is sharp enough even at a low frequency region. The signal processing was performed based on the conventional beam forming methods, in which the influence of moving train was considered [8]. Figure 7 shows the typical noise source distributions at f = 250 Hz octave band around the train including gaps between 7th and 8th cars. As shown in the gaps between 6th and 7th and between 8th and 9th cars, there exists a noise source around the current gap. It should be noted that sound power is comparable to that of the noise from the bogie section. This result indicates that the gap section is one of the main noise sources in a low frequency region. On the other hand, it is clearly shown the sound power at the gap between 7th and 8th cars is reduced by the countermeasures with rounded edge. This result consistent with that of the wind tunnel test as shown in Fig. 4 qualitatively.

4 Conclusion We examined effective noise reduction techniques for the gaps between adjacent cars section of high-speed Shinkansen trains through wind-tunnel testing. Accurate noise source distributions around the scaled train model were obtained using a microphone array installed inside the free shear layer. Based on the results of these tests, we proposed a number of noise reduction techniques. For the gaps between adjacent cars, rounded edges is effective for noise reduction because they cause a reduction in the strength of the vortex at the upstream edge, its impact on the downstream edge and acoustic radiation efficiency. The amount of noise reduction level obtained using this technique was estimated approximately as 6 to 7 dB. Finally, we verified the effect of noise reduction technique for the gap section in the field tests qualitatively.

References [1] http://www.env.go.jp/en/air/noise/railway.html [2] Ikeda, M., Suzuki, M., Yoshida, K.: Study on Optimization of Panhead Shape Possessing Low Noise and Stable Aerodynamic Characteristics. Quarterly Report of RTRI 47(2), 72– 77 (2006) [3] Nagakura, K.: Method of analyzing the Wind Tunnel Test data measured with directional microphone systems. Quarterly Report of RTRI 2(42), 104–109 (2001) [4] Yamazaki, N., Nagakura, K., Ikeda, M.: Acoustic Source Localization of Pantograph Using a 2-D Microphone Array (in Japanese). RTRI REPORT 18(11), 19–24 (2004) [5] Johnson, D.H., Dudgeon, D.E.: Array signal processing. Prentice Hall, Englewood Cliffs (1993) [6] Yamazaki, N., Nagakura, K., Ikeda, M., Sagawa, A.: Methods to Measure Acoustic Sources in a Closed Wind Tunnel Test Section. In: 11th AIAA/CEAS Aeroacoustics Conference, Monterey (2005) [7] Central Japan Railway Company, Annual Report 2006, p. 12 (2006) [8] Nordborg, A., Martens, A., Wedemann, J., Willenbrink, L.: Wheel/rail noise separation with microphone array measurements, inter noise, Hague (August 2001)

Measures to Counteract Micro-pressure Waves Radiating from Tunnel Exits of DB’s New Nuremberg-Ingolstadt High-Speed Line Th. Tielkes, H.-J. Kaltenbach, M. Hieke, P. Deeg, and M. Eisenlauer Deutsche Bahn AG, DB Systemtechnik, Department of Aerodynamics and Air Conditioning, Völckerstraße 5, 80939 Munich, Germany Tel.: +49 89 1308 5433; Fax: +49 89 1308 6795 [email protected]

Summary In December 2005, when first test runs were carried out on the new high speed line Nuremberg-Ingolstadt, significant sonic boom incidents occurred at the portals of the 7700 m long Euerwang tunnel and the 7260 m long Irlahüll tunnel. Both double-track tunnels with a cross section of 92 m2 were originally planned to be built with ballasted track. When a change to slab track was decided, no changes in the design of the already built tunnels were introduced. In order to reduce micro-pressure wave emissions and to ensure an in-time start of commercial railway operation, DB decided to take immediate countermeasures by equipping the two tunnels with acoustical track absorbers. These absorbers are designed to counteract railway noise, but also affect via dispersion and friction the pressure wave steepening process. With the installation of these absorbers the wave steepening process was significantly inhibited and, thus, commercial operation could start successfully in May 2006 without operational restrictions. A simplified formula based on the solution for the vibrating piston which was developed at RTRI appears to be well suited for the prediction of the amplitude and frequency content of the emitted micro-pressure wave in the vicinity of the tunnel portal.

1 Introduction In May 2006, the new high speed line connecting Nuremberg with Ingolstadt went into operation, thereby meeting the self-imposed deadline to be ready for the additional passenger traffic resulting from the FIFA soccer world cup carried out in Germany [1, 2]. In Dec. 2006, the upgraded line Ingolstadt-Munich was opened. Both lines are part of the international railway connection from Stockholm to Verona within the Trans-European High-Speed Railway Network (TEN). 77 km of the 89 km long double-track line connecting Nuremberg with Ingolstadt were newly built. From this, a total of 27 km is situated in tunnels with a nominal value of the semi-circular cross section of 92 m2. Most parts of the line are covered with slab track. In the original planning the tunnel interior as well as the entire line was to be built in ballasted track. Based on previous experience with wave attenuation B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 40–47, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Measures to Counteract Micro-pressure Waves Radiating from Tunnel Exits

41

in ballasted tunnels it was concluded that no micro pressure wave emission was to be expected on this line which is designed for regular train speeds up to 300 km/h. Experience on the Japanese high-speed network since the late 1970s [3, 4] has shown that the non-linear process of wave steepening during propagation along the tunnel depends on the tunnel length and on the friction effects. Thus, with the change from ballasted to slab track the likeliness of emission of significant micro-pressure waves increased for the two longest tunnels on the line, the Euerwang tunnel with a length of 7700 m and the Irlahüll tunnel with a length of 7260 m. In fact, during test runs in December 2005 with train speeds up to 330 km/h noticeable sonic boom incidents were observed near the portals of both tunnels. It was decided to install countermeasures before the line going into operation in the mid of 2006.

2 Tunnel Geometry and Setup of Measurements 2.1 Tunnel Geometry The tunnel walls have a smooth concrete surface which is interrupted in regular intervals by seams. The interior is equipped with installations holding the catenary, signals, lighting, and other items related to safety and emergency handling. In intervals of approximately 1000 m there are small niches with a depth of 2 m that are closed by pressure-tight doors leading to emergency exits. The track bed consists in a flat concrete surface from which shallow ends of sleepers protrude on which the rails are mounted. The two tunnels differ with respect to the portal geometry. Both the southern portal of the Euerwang tunnel and the southern portal of the Irlahüll tunnel are inclined by an angle of 45o. The opening is equipped with a circumferential flange of variable width. The southern portal of the Euerwang tunnel protrudes from a gently sloped hillside into the flat floor of the valley of the river Schwarzach. On one side, the area immediately ahead of the portal is bordered by the motorway A9 in a distance of approximately 50 m whereas on the other side the terrain consists in a sloped hillside. The northern portal of the Euerwang tunnel consists in a steep wall at the end of a 180 m long cutting with vertical side walls. Measurements have shown that entry in the northern portal of the Euerwang tunnel causes the largest pressure gradient for a given train speed. In view of other given parameters, thus, the most intense micro-pressure waves are emitted from the southern portal of this tunnel. Therefore, in the description of measurements we focus from now on entirely on the Euerwang tunnel. 2.2 Installation of Acoustic Absorbers During high-speed test runs in December 2005 employing the ICE-S test train at speeds up to 330 km/h audible micro pressure waves were noticed at the portals of the Euerwang and the Irlahüll tunnels. There are no close by residential areas but some public routes and hiking pathes. Although there are no applicable German regulations for the sonic boom and although other acoustic regulations had been complied with, Deutsche Bahn AG decided to take measures to reduce the micro-pressure wave emissions at these four portals.

42

T. Tielkes et al.

After a brief survey on the available countermeasures and after consideration of all relevant legal, temporal, and other constraints it was decided to equip both tunnels with acoustical track absorbers. This was achieved in a remarkably short time period in the beginning of 2006. The sound absorbing plates of brand name LIAKUSTIK had been manufactured by Bausteine Briest GmbH and already held a homologation by German railway authorities with respect to civil engineering demands. They consist in expanded clay with a porosity of 25 %, with an average density of 1027 kg/m3 and with a flow resistance which was found to be 9540 Pa/(s m2). The upper surface is formed by transversal ribs with a height of 30 mm. The plates in between the tracks have a thickness of 150 mm, and a characteristic width of 1000 mm. Additional plates cover the ends of the sleepers outside of the track. The absorbers are glued to the underground. An estimate for the entire cross section of the sound absorbing plates (for two tracks) yields 0.54 m2. 2.3 Measurement Set Up and Organisation of Test Runs A total of six pressure transducers were installed in a height of 2 m above top of rail along the wall inside the Euerwang tunnel. Two transducers were located at about 200 m from the portals and the other four were situated at approximately equidistant positions in between. Thus, the steepening of the compression waves could be observed over five consecutive segments, each with a length of ca. 1465 m covering a total stretch of 7300 m. The signals from the pressure sensors were recorded with sampling rates in the order of 1800 Hz. The train speed was determined by a light barrier installed at the northern end of the tunnel and by reading it from the display in the train cockpit. In average, the difference was about 2%. For the measurement of the emitted micro-pressure wave three pressure transducers were installed in lateral distances of approximately 1.5 m from the track in approximately 0.5 m height above the ground at axial distances of 5 m, 25 m, and 50 m from the base of the southern portal of the Euerwang tunnel. Prior to the equipment with absorbers 9 test runs in each direction were recorded for an ICE3 cruising with speeds in between 270 and 300 km/h. The installation of sound absorbing plates was suspended when approximately the first 2000 m of track counting from each portal inwards were equipped. During two days an additional 32 runs of ICE 3 with nominal speeds between 270 km/h and 300 km/h were monitored, thereby allowing measurements of the efficacy of the plates in a partially equipped tunnel. Finally, after complete installation of sound absorbers an additional 26 test runs with nominal train speeds from 280 km/h to 300 km/h were conducted.

3 Results 3.1 Pressure Wave Generation during Tunnel Entry When a train enters the portal, a system of pressure waves forms. Here we focus only on the leading compression wave. Its amplitude Δ p = p − p0 , i.e. the difference from

Measures to Counteract Micro-pressure Waves Radiating from Tunnel Exits

43

the ambient pressure p0 , depends primarily on the train speed Vtr and the blockage R = Atr / Atun = 1 − Φ . From isentropic theory follows [5]

Δp =

1 2

ρ Vtr

1 − Φ2

2

Φ + (1 − Φ ) Ma − Ma 2

2

=

2

1 2

ρ Vtr f ( Ma, Φ ) 2

(1)

where Ma = Vtr / c0 denotes the Mach number. Fig. 1 (left) shows measurements from a pressure transducer placed inside the tunnel in a distance of approximately 200 m from the entry portal. For the ICE 3 train with a blockage of R ≈ 0.11 equation (1) yields in the speed range 270 km / h < Vtr < 300 km / h for the non-dimensional ampli2 tude Δp /( ρ Vtr ) = f / 2 ≈ 0.128 since the Mach number dependence of f ( Ma, Φ ) is weak. Fig. 1 (left) implies that the pressure wave generated on the north portal has a somewhat higher amplitude than on the south side, thereby exceeding the prediction from eq. (1) by 6%. Assuming that the duration of the tunnel entry is proportional to D h / Vtr where Dh = 4 A tun / Utun denotes the hydraulic diameter, the maximum gradient of the entry wave scales as dp dt max

≈ξ

1

ρ Vtr f ( Ma, Φ ) 3

2

1 Dh

.

(2)

Fig. 1 (right) shows that this scaling holds in the considered speed range. As to be expected, for the same entry speed the resulting pressure gradient is higher (by 18%) for the steep northern portal than for the inclined southern portal of the Euerwang tunnel. This difference between a steep and an inclined portal is qualitatively similar to what was found in laboratory experiments for an axisymmetric model train head. However, the absolute value of the constant ξ in eq. (2) differs by more than 20% from existing laboratory results – the reason for this mismatch is not yet fully understood.

0.80

0.15

dp/dt (normalized)

Δp (normalized)

0.16

0.14 0.13 0.12 South North

0.11 0.10 260

270

280 290 Vtr [km/h]

300

310

0.70 0.60 0.50 0.40 260

South North

270

280 290 Vtr [km/h]

300

310

Fig. 1. Measurements of the tunnel entry of ICE 3 in 200 m distance from both portals of the Euerwang tunnel. Left: Scaling of the amplitude of the entry pressure wave according to 2 Δp /( ρ Vtr ) . Right: Scaling of the maximum entry pressure gradient according to 3 ξ = ∂p / ∂t /(0.5 ρ Vtr f ( Ma , Φ ) / Dh ) .

(

)

44

T. Tielkes et al.

3.2 Wave Propagation Inside the Tunnel

A large subset of the available measurements has been assembled in Fig. 2 where the ratio of the maximum pressure gradients at the end and at the begin of a segment of length of 1465 m is shown as a measure for wave steepening versus the characteristic frequency f w = (∂p / ∂t ) / Δp of the wave. Here, we also included measurements taken in the partially equipped tunnel. From the latter, only the pairs of sensors are considered which fall in a segment with homogeneous surface conditions, i.e. which are either fully equipped with absorbers (sensor pair 1-2 and pair 5-6) or which do not contain absorber plates (sensor pair 3-4). Since the wave steepening in the partially equipped tunnel is less attenuated than in the fully equipped tunnel, we obtain valuable information on the selective influence of absorbers on waves with high characteristic frequencies. 3.5

25.0

3.0

wave steepening

wave steepening

20.0 15.0 10.0 5.0

2.5 2.0 1.5 1.0 0.5 0.0

0.0 0.0

10.0

20.0

30.0

characteristic frequency [1/s]

40.0

0.0

10.0 20.0 30.0 characteristic frequency [1/s]

40.0

Fig. 2. Comparison of the wave-steepening prior to (left) and after (right) installation of sound absorbing plates in the tunnel. Shown is the ratio of maximum pressure gradients for wave propagation over a 1465 m long segment versus characteristic frequency of the wave at the begin of the segment.

In Fig. 2 a substantial scatter can be noted in the data which is related to the difficulty involved in the determination of a derivative of a non-smooth pressure recording. Furthermore, few samples are available for steep waves which are characterized by high frequencies. Thus, the continuous line fit has to be regarded as a rather crude approximation. The wave steepening in the new 92 m2 tunnels on the Nuremberg-Ingolstadt line is in line with experimental data obtained for the 4.5 km long Schulwald tunnel on the Cologne-Rhein/Main high-speed line [6]. However, it seems to proceed at a higher rate than in 64 m2 slab-track tunnels on the Japanese Shinkansen network. It is not clear whether this can be solely contributed to the relative increase in friction effects due to the reduction in the hydraulic diameter by approximately 20 %.

Measures to Counteract Micro-pressure Waves Radiating from Tunnel Exits

45

3.3 Emission of Micro-pressure Waves

Based on the exact solution for the sound field emitted by a vibrating piston in a baffle plate Ozawa et al. [7] derived a prediction formula for the pressure disturbance Δ pMPW in the far field in an axial distance s from the portal which covers the whole frequency range. Using the abbreviations p ′ = ∂p / ∂t , T1 = 1.4 r / c , T2 = r / c , k1 = 1 /( 4 π T1 ) , and k2 = 11 /(50 π T2 ) it reads t t ⎤ ⎛ −τ 2 ⎞ ⎛ −τ 2 ⎞ 2 Atun ⎡ 1 τ ⎛ s⎞ ΔpMPW ⎜ t + ⎟ = ⎢ p′(t ) + k1 ∫ exp⎜⎜ 2 ⎟⎟ p′(t − τ )dτ + k 2 ∫ exp⎜⎜ 2 ⎟⎟ p′(t − τ ) dτ ⎥ 4 T T 4 T ⎝ c ⎠ Ω c ( s + s0 ) ⎣ 2 ⎝ 1 ⎠ ⎝ 2 ⎠ 0 2 0 ⎦

(3) Here, r denotes the (hydraulic) tunnel radius, c the speed of sound, and Ω is the solid angle which depends on the geometry of the portal and its surroundings. Equation (3) has been evaluated for three pressure signals recorded close to the Euerwang south portal. Fig. 3 (top left) shows the signals which cover gradients in the range 9 kPa/s to 60 kPa/s. 1000

300 240 180 120 60 0 -60 -120 -180 -240

600 9 kPa/s

400

18 kPa/s 200

60 kPa/s

0 9.6 9.7 9.8 9.9

[60 kPa/s] Omega = 5/4 pi s0 = 8m

pressure [Pa]

pressure [Pa]

800

5m 25m 50m 2.55

10 10.1 10.2 10.3 10.4 t [s]

2.6

2.65

2.7 t [s]

2.75

2.8

2.85

120 [18 kPa/s]

pressure [Pa]

pressure [Pa]

5m

40

60

Omega = 5/4 pi s0 = 8m

80

0 25m -40

[9 kPa/s]

Omega = 5/4 pi s0 = 8m 5m

40 20 0

25m

-20

50m

50m

-40

-80 10.4

10.5

10.6

10.7 t [s]

10.8

10.9

10.2

10.3

10.4

10.5

10.6

10.7

t [s]

Fig. 3. Top left: Pressure signal inside the Euerwang tunnel in 200 m distance from the south portal. Others: Comparison of measurements outside of the tunnel near the track at distances s of 5 m, 25 m, and 50 m from the portal base with predictions using formula (3) with an offset s 0 = 8m and a solid angle Ω = 5 / 4π . Curves for 25 m (50 m) locations are plotted with offsets of -120 Pa (-240 Pa), -40 Pa (-80 Pa), and -20 Pa (-40 Pa), respectively.

46

T. Tielkes et al.

The data taken inside the tunnel is somewhat spoiled by a 50 Hz noise from the AC supply to the measurement system. This noise is carried through eq. (3) and shows up clearly in the prediction of the micro-pressure wave. Nevertheless, we find that eq. (3) is a good approximation both for the amplitude and the time signature of the emitted wave. Since the portal is inclined by 45˚ it is not obvious where to put the origin from which the axial distance s is measured. We found that s0 = 8 m together with a solid angle of Ω = 5 / 4 π yields good agreement with measurements taken in the vicinity of the portal. Since the first measurement station is only 5 m from the base of the portal it might not be appropriate to apply eq. (3) there since it strictly holds only in the far field. Thus, the mismatch of prediction and measurements taken at this station should probably not be overrated. The agreement between prediction and measurements for the two stations at 25 m and 50 m is satisfactory. With the installation of the acoustic track absorbers the micro-pressure wave emissions had changed advantageously. Besides a reduction of the amplitude, a significant change in the frequency spectrum towards lower frequencies was observed. Experimental data for the acoustic parameters and the acoustic assessment of the remaining micro-pressure wave emissions at Nuremberg-Ingolstadt high speed line are presented in a complementary paper [8].

Acknowledgements The authors wish to express their gratitude to all those who contributed to the installation of the track absorbers and to the various test campaigns. Special thanks are due to DB Netz AG and DB Projektbau GmbH for their leadership of the project and the very close and efficient collaboration.

References [1] Feldwisch, W., Schülke, H.: Die Inbetriebnahme der Großprojekte der Bahn zur Fußballweltmeisterschaft 2006. ETR Eisenbahntechnische Rundschau 5, 289–300 (2006) [2] Tielkes, T.: Aerodynamic aspects of Maglev Systems. In: Schach, R., Witt, M. (eds.) Proceedings of MAGLEV 2006, pp. 641–649 (2006) ISBN: 3-86005-535-6 [3] Yamamoto, S.: Micro-pressure wave issued from a tunnel exit. In: Abstract of the Spring Meeting of the Physical Society of Japan (April 1977) [4] Ozawa, S.: Studies on the micro-pressure wave radiated from a tunnel exit, Report 1121 of the Railway Technical Reseach Institute (RTRI), Japan (March 1979) [5] Maeda, T., Matsumura, T., Iida, M., Nakatani, K., Uchida, K.: Effect of Shape of Train Nose on Compression Wave Generated by Train Entering Tunnel. In: Conference Proceedings STECH 1993, Paper PS3-8, pp. 315–319 (1993) [6] Herb, J., Deeg, P., Tielkes, T.: Assessment of Possible Sonic Boom Effects in German High-Speed Railway Tunnels – Experimental and Numerical Data for the Wave Steepening Process. In: 11th Int. Symp. on Aerodynamics and ventilation of vehicle tunnels, Conference Proceedings BHR Group 2003, pp. 775–782 (2003)

Measures to Counteract Micro-pressure Waves Radiating from Tunnel Exits

47

[7] Ozawa, S., Murata, K., Maeda, T.: Effect of ballasted track on distortion of pressure wave in tunnel and emission of micro-pressure wave. In: 9th Int. Symp. on Aerodynamics and ventilation of vehicle tunnels, Conference Proceedings BHR Group 1997, pp. 935–947 (1997) [8] Degen, K.G., Gerbig, C., Onnich, J.: Acoustic assessments of micro-pressure waves radiating from tunnel exists of DB high-speed lines. In: 9th International Workshop on Railway Noise, Munich (2007)

Acoustic Assessment of Micro-pressure Waves Radiating from Tunnel Exits of DB High-Speed Lines K.G. Degen, Ch. Gerbig, and H. Onnich Deutsche Bahn AG, DB Systemtechnik, Department of Acoustics and Ground Vibration, Völckerstraße 5, D-80939 München, Germany Tel.: +49 89 1308 5273; Fax: +49 89 1308 2590 [email protected]

Summary In December 2005, when first test runs were carried out on the new high speed line Nuremberg-Ingolstadt, significant sonic boom incidents occurred at the portals of the 7700 m long Euerwang tunnel and the 7260 m long Irlahüll tunnel. Both doubletracked tunnels with a cross-section of 92 m2 were originally planned to be built with ballasted track. When a change to slab track was decided, no changes in the design of the already built tunnels were introduced. DB decided to take immediate countermeasures by equipping the two tunnels with acoustical track absorbers. These absorbers are designed to counteract railway rolling noise, but also affect the pressure wave steepening process. Thus a significant reduction of the micro-pressure emission was achieved and the commercial operation on Nuremberg-Ingolstadt line started in Mai 2006 successfully in time and without operational restrictions. This paper presents the acoustical measurements outside the tunnels and their assessment based on national traffic noise regulations for the neighbourhood, LCpeak for possible health hazards nearby the tunnel portals and LCE for comparison the micropressure effect with a train pass-by.

1 Introduction During test runs in December 2005 on the new German high-speed line NurembergIngolstadt clearly audible effects, recognizable as sonic boom, occurred at the tunnels Euerwang and Irlahüll, when the test trains entered the tunnels at the opposite entrance with speeds up to 330 km/h. When a high-speed train enters a tunnel, a micro-pressure air wave is raised moving along the tunnel at the speed of sound. Under certain circumstances the rising edge of the wave front becomes increasingly steeper on its path through the tunnel (Fig. 1). At the tunnel exit the pressure wave is partly reflected back into the tunnel and partly emitted outwards. The emitted part can be clearly audible even at distances far away from the tunnel portal (up to about 1 km). A detailed discussion of the physical mechanism and the influencing parameters is given in [1]. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 48–55, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Acoustic Assessment of Micro-pressure Waves Radiating from Tunnel Exits

Entrance

Tunnel

49

Exit

Train Generation of compression wave

Propagation of compression wave

Radiation of micropressure wave

Fig. 1. Evolution of the shape of a micro-pressure wave while propagating inside a tunnel

Clearly audible micro pressure wave effects were reported before only for the Shinkansen-lines in Japan, see e. g. Yamamoto [2]. At regular traffic of European high-speed lines this phenomenon did not show up in the past due to the use of ballasted track and the specifications for length and cross section of the tunnels. In order to ensure the successful opening of the new high-speed line planned for end of Mai 2006 it was very important, to efficiently reduce the micro pressure wave as well as to work out a suitable acoustic assessment procedure.

2 Local Situation at the Tunnels Euerwang and Irlahüll The tunnels Euerwang and Irlahüll (Fig. 2) with a length of more than 7 km are the longest ones of the line Nuremberg-Ingolstadt constructed for train-speeds up to 300 km/h. The inner tunnel walls with a cross-section of 92 m2 consist of smooth concrete. Only the Irlahüll northern portal shows a larger cross-section, as there are switches for two additional tracks located within the first 100 tunnel-meters. Schellenberg-Tunnel km 57,844 - km 58,496

Nuremberg

7700 m

km 49,154

km 56,854

Euerwang-Tunnel northern portal

Ingolstadt

7260 m

southern portal

km 59,564

km 66,824

Irlahüll-Tunnel northern portal

southern portal

Fig. 2. Schematic representation of the high-speed line Nuremberg-Ingolstadt in the area of the two longest tunnels Euerwang and Irlahüll and pictures of their portals

After the test runs countermeasures were decided by DB in order to avoid the emission of a clearly audible micro pressure wave. Therefore the tunnels were equipped with porous sound absorbing elements on each track (Fig. 3). The absorbers

50

K.G. Degen, C. Gerbig, and H. Onnich

were manufactured by Bausteine Briest GmbH, brand named LIAKUSTIK (for details see [1]), originally developed for reducing the rolling noise along open lines with slab track. In a first step the absorbers were installed over a length of approximately 2000 m inwards the tunnels starting from each tunnel portal (called “partly equipped”). In a second step also the middle tunnel section was equipped, thus having now absorbers located over the full tunnel length (called “fully equipped”).

Fig. 3. Photographs of the absorber elements on top of the slab-track inside the tunnel

In the close vicinity of the 4 tunnel portals there are rescue areas without public access. Nearest public areas are small roads with distances of 50 m or more to the portals. The nearest residents live at much larger distances of at least 1000 m.

3 Assessment Modes and Standards Commonly accepted regulations suitable for an assessment of tunnel micro-pressure effects actually do neither exist on a national level in Germany nor on a European level. On this background the following standards or directives were considered to assist an assessment, even though they were not expressly designed for railway noise: the directive 2003/10/EC [4] considering health and safety requirements of workers, the standards ISO 1996-1 [5] and ISO 7196 [6], the draft of a national directive dedicated to military gun noise [7] and the national directive TA-Lärm dedicated to industrial noise [8]. As the latter two directives are focused on the protection of areas dedicated to the regular stay of associated people, e.g. residential areas, and as these locations have distances of at least 1 km from the tunnel portals, no relevant noise limits of these directives were exceeded. Just as little the ISO 7196, dedicated to infra sound and the G-frequency weighting procedure, led to a limiting assessment aspect for the measured sound levels. The assessment thus was focused on the subsequently explained modes. 3.1 Legal Regulations in Germany The Directive 16. BImSchV [3] sets upper limits for the A-weighted sound pressurelevel in residential areas (time averaged over day and night period, respectively). No limits are given for the immediate surroundings of railway lines if these do not belong

Acoustic Assessment of Micro-pressure Waves Radiating from Tunnel Exits

51

to residential areas even though they might be accessible by public. For impulsive noise components, e.g. for buffer pulses in shunting-yards, a correction factor in the range of 0 to 8 dB(A) apply depending on sharpness and frequency. 3.2 Prevention of Effects Injurious to Health The Directive 2003/10/EC [4] is dedicated to the protection of workers. If the Cweighted peak sound pressure level LCpeak exceeds the “lower trigger value” at 135 dB(C), hearing protection has to be made available for workers. If the “upper trigger level” at 137 dB(C) is exceeded, hearing protection has to be applied and additional protection measures have to be performed. For working staff a stay beside the track in the near vicinity of the portals can not be excluded when the line is in operation. At this area the highest micro-pressure wave levels are expected. Public access is given only beyond the railway ground. Therefore next roads or other public areas have to be considered. 3.3 Comparison of Micro-pressure Sound and Train Pass-By For the comparison of the human experience of different sounds no common approved method is suitable. In particular this holds for sounds very different in time shape and frequency composition. In the opinion of the authors best regard to an assessment of pronounced micropressure wave effects is available by the Annex B (informative) of ISO 1996-1 [5], which is especially dedicated to high-energy impulse sounds. They are characterized by the adjusted sound exposure level LRE, based on the C-weighted sound exposure level LCE, with LRE = 2 LCE – 93 dB for LCE > 100 dB and LRE = 1.18 LCE – 11 dB for LCE < 100 dB. The adjustment of LCE depends directly on the energy content of the impulsive sound, while the C-weighting takes the contribution at low frequencies of micro-pressure wave effects into account.

4 Acoustic Measurements 4.1 General Remarks and Measurement Set Up The acoustical measurements were performed during test runs of a 200 m long ICE 3 unit at train speeds up to 300 km/h in March and April 2006. The microphones were located close to the tracks (distances 4 m and 7.5 m from the axis of the nearest track) at different distances from the portals (5 m, 25 m and 50 m) as well as at the next public road. The height of the microphones was 1.7 m above ground. The original time-signals with a measuring frequency band-with of about 3 Hz to 16 kHz were digitalized recorded, thus allowing the evaluation of any acoustical dimension. The train speed at tunnel entry was determined by a radar-pistol. 4.2 Characterisation of Micro-pressure Wave Sound The most pronounced micro-pressure sound effects were obtained for the Euerwang tunnel at the southern portal. As shown in Fig. 4 (left) the received sound starts abruptly with a micro-pressure caused sonic boom. In the following time-period the approaching train inside the tunnel causes a continuously increasing sound pressure

52

K.G. Degen, C. Gerbig, and H. Onnich

120

110

Sound pressure level, dB re 20 µPa

C-weighted sound pressure level, dB(C) re 20 µPa

level. When exiting the tunnel the train causes the “ordinary” pass-by noise, rapidly vanishing while the train disappears.

110 100 90 80

100 90 80 70 60 50

70 0

10

20

30

40 Time [s]

50

60

70

80

40 8

16 31.5 63 125 250 500 1k

2k

Frequency [Hz]

Fig. 4. Measured C-weighted sound pressure level as function of time at a distance of 65 m (next public road) in front of the tunnel without absorbers (left) and third octave band spectra at a distance of 50 m to Euerwang southern portal (right) ⎯ without absorbers; − − − partly equipped with absorbers; − ⋅ − ⋅ fully equipped with absorbers for an ICE 3-type train with 300 km/h at Euerwang southern portal

The spectra in Fig. 4 (right) are characterized by strong contributions at low frequencies (f < 125 Hz), also including infra-sound contributions below 20 Hz. For the original tunnel equipment without absorbers in addition pronounced sound pressure levels at frequencies > 250 Hz were measured (solid line), which correspond to a hearing impression as sharp and bright bang. Already the partial equipment with absorbers led to a clear reduction of the high frequency contributions, while low frequencies < 100 Hz were not significantly altered. Only when the tunnel was fully equipped with absorbers, reductions up to 9 dB in the low frequency range were obtained. This led to a considerably different hearing impression: After the absorbers were completely installed the micro-wave sound effect at the Euerwang portals was clearly reduced and experienced as a dull dump. At the Irlahüll portals the remaining acoustical micro-pressure effects are negligible. 4.3 Results and Assessment At the next buildings and living areas of people in the vicinity of the tunnel portals the calculated assessment-level defined by the 16. BImSchV [3] showed an increase by only 0.4 dB(A) at day and 0.1 dB(A) at night time, when the measured micro-pressure effects including a correction level of 8 dB(A) for impulsive noise were added. Thus it could be stated that no significant increase of the assessment-levels by micropressure effects results in the neighbourhood of these two tunnels. In order to assess the possibility of negative health effects, the C-weighted peak sound pressure level LCpeak was evaluated for the maximum train speed of 300 km/h at a portal distance of 5 m. The highest levels of both tunnels were obtained at the southern portal of the Euerwang tunnel (Fig. 5). Before installation of the track absorbers, the lower trigger level LCpeak = 135 dB(C) of standard 2003/10/EC was exceeded by

C-weighted peak sound pressure level, dB(C) re 20 µPa

Acoustic Assessment of Micro-pressure Waves Radiating from Tunnel Exits

53

150 without absorber partly equipped

140

fully equipped

130

120

Fig. 5. Maximum C-weighted peak sound pressure level LCpeak of the micro-pressure wave for ICE 3-type trains with 300 km/h at a distance of 5 m to Euerwang southern portal. Left bar without absorbers; middle bar partly equipped with absorbers; right bar fully equipped with absorbers. For comparison: − − − lower trigger level LCpeak = 135 dB(C) of standard 2003/10/EC.

C-weighted sound exposure level with adjustment for impulsive sound, dB(C) re 20 µPa

up to 9 dB(C). After the tunnel was fully equipped with absorbers the maximum measured levels were clearly below this limit thus excluding health hazards. Fig. 6 shows the adjusted sound exposure levels LRE of the micro-pressure wave according to Annex B of ISO 1996-1 [5] as defined in chapter 3.3. For comparison, the measured sound exposure level for a pass-by of a 400 m long ICE 3 type train with 300 km/h is reported. 130

120 without absorber 110

partly equipped fully equipped

100

Fig. 6. C-weighted sound exposure level LCE including a correction factor for impulsive noise according to ISO 1996-1 of the micro-pressure wave for a ICE 3-type train with 300 km/h at a distance of 65 m to Euerwang southern portal (nearest public road). Left bar: without absorbers; middle bar: partly equipped with absorbers; right bar: fully equipped with absorbers. For comparison: − − − LCE of the pass by for a full length ICE 3-type train at the same measuring point (distance about 16 m to the next track).

54

K.G. Degen, C. Gerbig, and H. Onnich

Without absorbers the micro-pressure wave causes about 5 dB(C) higher levels. However, after the tunnels were fully equipped with absorbers the micro-pressure level is about 6 dB(C) below the train pass-by level. This indicates the micro-pressure effects for the fully equipped tunnels to be not a new quality of noise, reaching or exceeding the typical noise loads in the neighbourhood of high-speed railways. Thus it can be stated, that a significant reduction of the micro-pressure emission was achieved after both tunnels were equipped with absorbers. While audible effects nearly vanished at the Irlahüll tunnel, the remaining micro-pressure effects at the two Euerwang tunnel portals are uncritical, even as they still are clear perceptible. This assessment was accepted by the German railway authority as well.

5 Conclusion and Outlook Corresponding to an efficiently delayed wave-steepening process along the wave propagation path inside the tunnel the acoustical effects outside the tunnels were successfully reduced by the porous track absorbers. In Mai 2006 the commercial operation on the Nuremberg-Ingolstadt line started in time and without operational restrictions. Nevertheless, for future railway infrastructure projects additional options for countermeasures are necessary and have to be developed, particularly if one-tracked tunnels with a much smaller cross-section are planned. A quite large mitigation potential is expected from optimizing the tunnel portal construction, as it is practiced at most Japanese high-speed lines. In addition an optimization of the front construction of future high-speed trains should be taken into account.

Acknowledgements The authors wish to express their gratitude to all those who contributed to the installation of the track absorbers and to the various test campaigns. Special thanks are due to DB Netz AG and DB Projektbau GmbH for their leadership of the project and the very close and efficient collaboration.

References [1] Tielkes, T., Kaltenbach, H.-J., Hieke, M., Deeg, P., Eisenlauer, M.: Measures to counteract micro-pressure waves radiating from tunnel exits of DB’s new Nuremberg-Ingolstadt highspeed line. In: 9th International Workshop on Railway Noise, Munich (2007) [2] Yamamoto, S.: Micro-pressure wave issued from a tunnel exit. In: Abstract of the Spring Meeting of the Physical Society of Japan (April 1977) [3] German federal traffic noise protection directive: Sechzehnte Verordnung zur Durchführung des Bundes-Immissionsschutzgesetzes (Verkehrslärmschutzverordnung – 16. BImSchV) (1990) [4] Directive 2003/10/EC of the European Parliament and of the Council, minimum health and safety requirements regarding the exposure of workers to the risks arising from physical agents (noise) (2003)

Acoustic Assessment of Micro-pressure Waves Radiating from Tunnel Exits

55

[5] International Organization for Standardization ISO 1996-1, Acoustics – Description, measurement and assessment of environmental noise – Part 1: Basic quantities and assessment procedures (2003) [6] International Organization for Standardization ISO 7196, Acoustics – Frequencyweighting characteristic for infrasound measurements (1995) [7] Draft version of a German gun noise management directive: Entwurf – Bestimmungen zum Schießlärm-Management auf Truppenübungsplätzen (Lärmmanagement-Richtlinie), German Ministry of Defence (2005) [8] German federal industrial noise protection directive: Sechste Allgemeine Vorschrift zum Bundes-Immissionsschutzgesetz (Technische Anleitung zum Schutz gegen Lärm – TALärm) (1998)

High Speed Railway Noise: Assessment of Mitigation Measures F. Létourneaux, J.F. Cordier, F. Poisson, and N. Douarche SNCF Agence d’Essai Ferroviaire, 21 avenue du Président Allende, F94407 Vitry-Sur-Seine Cedex, France Tel.: +33 1 47 18 82 32; Fax: +33 1 47 18 82 30 [email protected]

Summary Stringent limit values have been introduced in the HS-TSI for the pass-by noise of HS trains, as demonstrated by the NOEMIE project. Moreover, as further reductions of these limits are already planned, new designs of high speed rolling stock will have to have a lower acoustic signature to be TSI compliant. In that context, SNCF get involved, with some industrials partners, in different R&D projects that aims at the development and assessment of solutions, both for the track and the rolling stock. This has lead, in october 2006, to a field test measurement campaign on a French High Speed Line. This test involved a TGV trainset equipped with wheel damping rings on some motor and trailer bogies, and a section of track equipped with rail dampers. A cumulated efficiency up of about 3 dB have been achieved at 300 km/h. Due to the high speed, the solution dedicated to the reduction of the wheel acoustic radiation had the strongest effect on the overall pass-by noise (~2 dB). It has also been highlighted that the noise attenuation is directly dependent on the wheel and rail tread unevenness.

1 Introduction Stringent limit values have been introduced in the HS-TSI for the pass-by noise of HS trains, as demonstrated by the NOEMIE project [1]. Moreover, as further reductions of these limits are already planned, new designs of high speed rolling stock will have to have a lower acoustic signature to be TSI compliant. In that context, SNCF get involved, with some industrials partners, in different R&D projects that aims at the development and assessment of solutions, both for the track and the rolling stock. This has lead, in october 2006, to a field test measurement campaign on a French High Speed Line. This test involved a TGV trainset equipped with wheel damping rings on some motor and trailer bogies, and a section of track equipped with rail dampers. The objectives of the measurement campaign were twofold: at first assess if wheelrings could be an efficient way to reduce the acoustic radiation of the rolling stock even on tangent tracks, and second to demonstrate that a reduction of the rolling noise contribution lead to a reduction of the overall noise emitted during a pass-by. The paper will first present a description of the solutions that were tested. Then an overview of the measurement campaign will be provided and its main results analysed. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 56–62, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

High Speed Railway Noise: Assessment of Mitigation Measures

57

2 Description of the Mitigation Measures 2.1 Wheel Rings The DAAVAC system developed by the company ENAC in collaboration with the wheel supplier VALDUNES, is a damping device which aims at removing the vibrations of a railway wheel. It is an expansive multilayered ring made of a piece of elastomeric material in between steel circular segments. This ring is put into a groove especially machined in the wheel tyre and is installed by a simple ratchet mechanism (see a picture Fig. 1). 2.2 Rail Dampers The system designed by SHREI & VEIT and distributed in FRANCE by SOCITEC consists of three active parts fixed to the rail through a baseplate: Two active parts are located each side of the railweb and one is underneath the railfoot. Each of them is made of a stack of alternating layers of steel pieces and elastomer. To broaden the frequency range efficiency the steel masses have different widths. The overall added mass to the rail is nearly 70% (see a picture Fig. 2).

Fig. 1. DAAVAC system

Fig. 2. Rail damper

3 Field Test Description The acoustic efficiency of both solutions was assessed in october 2006, on a high speed line in the south of France. A piece of tangent track, 300m long, was fitted with rail dampers (up track only, both rails, one damper in each sleeper bay). It is hereafter named ZA section. A contiguous track section of 300m with conventional track equipment was considered as a reference section for comparison (ZR section). Rolling stock: Two different trainsets were measured during their pass-bys over the measurement site:

58

-

F. Létourneaux et al.

a single-deck TGV (TGV-Réseau) under test at speeds up to 360 km/h with no acoustic modifications, a double-deck TGV (TGV-Duplex) in operation but equipped with wheel rings on two bogies. The precise location of the rings is provided in a diagram Fig. 3: 2 powered and 2 trailed bogies have been fitted with this damping device. A reference trailed bogie has also been defined for the assessment of rings acoustic performance. The DAAVAC and reference trailed bogies were reprofiled one week before the acoustic test in order to have a similar wheel roughness. All the other bogies were also recently reprofiled.

Instrumentation: Only trackside acoustic measurements were performed. Two microphones (at 7, 5 and 25m from the track axis) were installed in the centre of each track section ZA and ZR to assess the rail damper effect. A near field microphone at 3m from the track axis and an acoustic antenna (2D array of 73 microphones) were also put in the section ZR to better focus on each wheel acoustic radiation. To measure the train speed and to identify the position of wheels in the acoustic time histories signals, wheel detection systems was used. (electromagnetic treadles fixed on the rail). M1 PB

PB

R2

R1 TB

TB

R4

R3 TB

TB

wheelsets fitted with DAAVAC (reprofiled) Reference trailer boggies (reprofiled)

R6

R5 TB

TB

R7 TB

M2

R8 TB

TB

PB

PB

TB : Trailed bogie PB : motored bogie

Fig. 3. Location of wheel rings on the TGV-Duplex

4 Experimental Results 4.1 Rail Dampers Efficiency The rail damper performance has been investigated using a comparison between LAeq,Tp issued from measurements at 25 m, at the two track sections (ZA & ZR). The case of the TGV-R under test, running in single and multiple unit configurations at speeds up to 360 km/h was chosen. At the beginning of the measurement campaign, the noise reduction was less than 1 dB. But it grew up to reach a maximum level of 1.5 to 2.3 dB at 350 km/h after a series of events which caused a significant increase of the wheel unevenness: following track works, some ballast dust might have been run over by the wheels (see [2]). The one-third octave band spectra confirmed that the rail dampers lower the spectral components between 630 and 5000 Hz, with a peak of attenuation in the range 630-1250 Hz where it is known that the track acoustic radiation is maximum (Fig. 4).

High Speed Railway Noise: Assessment of Mitigation Measures 120

40 dB (A)

5000 6300 8000 10000 12500

1000 1250 1600 2000 2500 3150 4000

One-third octaveFréquence band centre frequency in Hz (Hz)

One-third octaveFréquence band centre frequency in Hz (Hz)

dB (A)

40

20 25 31,5

50

dB (LIN)

50

200 250 315 400 500 630 800

60

50 63 80 100 125 160

60

dB (LIN)

70

6300 8000 10000 12500

70

80

2000 2500 3150 4000 5000

80

90

500 630 800 1000 1250 1600

SPL in dB

90

20 25 31,5 40

SPL in dB

100 Niveau de pression acoustique en dB réf. 20 µPa

100

351 km/h avant d'envols damage de ballast Track sec ZR1ère – série no wheel 351 km/h avant d'envols damage de ballast Track sec ZA1ère – série no wheel Track sec ZR1ère – série low d'envols wheeldedamage 342 km/h après ballast Track sec ZA1ère – série low d'envols wheeldedamage 342 km/h après ballast Track sec ZR3ème – high 352 km/h après sériewheel d'envols damage de ballast Track sec ZA – high wheel damage 352 km/h après 3ème série d'envols de ballast

110

10 dB

10 dB

110

160 200 250 315 400

300 km/h avant d'envols damage de ballast Track sec ZR1ère – série no wheel 300 km/h avant d'envols damage de ballast Track sec ZA1ère – série no wheel Track 303 km/h sec après ZR1ère – série low d'envols wheeldedamage ballast Track 303 km/h sec après ZA1ère – série lowd'envols wheeldedamage ballast Track sec ZR3ème – high damage 301 km/h après série wheel d'envols de ballast Track sec ZA3ème – high damage 301 km/h après série wheel d'envols de ballast

40 50 63 80 100 125

120

Niveau de pression acoustique en dB réf. 20 µPa

59

Fig. 4. Influence of rail dampers on LAeq,Tp (whole trainset, 25m) at 300 km/h (left) and 350 km/h (right)

4.2 Wheel Rings Efficiency

110

100

100 Niveau sonore en dB(A) re. 20µ Pa

110

90

SPL in dB(A)

SPL in dB(A)

Niveau sonore en dB(A) re. 20µ Pa

The influence of rings on the overall acoustic levels can be highlighted directly on the acoustic signature of the train as illustrated in Fig. 5.

80

70

60 Niveau sonore SPL @ 3m

wheel detector Tops essieux DAAVAC wheelset Essieux Daavac REFERENCE wheelset Essieux porteurs de référence

50

40

9

10

11

12

13

14

90

80

70

60 Niveau sonore SPL @ 25m wheelessieux detector Tops DAAVAC Daavac wheelset Essieux REFERENCE wheelset Essieux porteurs de référence

50

15

16

40

9

10

Tempsin (s)s Time

11

12 13 Temps Time (s) in s

14

Fig. 5. Acoustic signature of the TGV-Duplex partially equipped with DAAVAC

R3

R4 BP

BP

R2 BP

Fig. 6. Time signal used to assess the noise from bogies

15

16

F. Létourneaux et al.

0

0

-1

-1

-2

-2

-2

-3 -4 -5 -6 -7 -8 -9 -10 200

-3 -4 -5 -6 -7 -8 -9

@ 3 m - no dampers 250

300

350

Acoustic Gain in dB(A)

0 -1 Acoustic Gain in dB(A)

Acoustic Gain in dB(A)

60

-10 200

-3 -4 -5 -6 -7 -8 -9

@ 7.5 m - no dampers 250

300

350

@ 25 m - no dampers

-10 200

250

300

350

Train speed (km/h)

Train speed (km/h)

Train speed (km/h)

Fig. 7. DAAVAC acoustic efficiency 120 Remorque 5 - RÉFÉRENCE

Reference bogies Daavac bogies

Remorque 3 - DAAVAC 110

SPL in dB

Niveau de pression acoustique en dB réf. 20 µPa

100

90

80

70

60

50

dB (A)

One-third octave centre frequency (Hz) Fréquence (Hz)

dB (LIN)

8000 10000 12500

315 400 500 630 800 1000 1250 1600 2000 2500 3150 4000 5000 6300

50 63 80 100 125 160 200 250

40

Fig. 8. Spectral effect of wheel rings DAAVAC (LAeq,Tp @3m)

The acoustic attenuation achieved by the DAAVAC system on trailed bogies is calculated as the difference of LAeq,T. For each kind of bogie (DAAVAC or REFERENCE), the corresponding portion of time signal selected begins at the middle of the previous coach and ends at the middle of the following coach as shown in Fig. 6. The results of all recorded pass-by are given in Fig. 7 for the conventional track section ZR (no rail dampers). The noise reduction achieved lies between -1 and -4 dB with a mean of -2.3

High Speed Railway Noise: Assessment of Mitigation Measures

61

dB for the microphone at 7.5m and a train speed up to 300 km/h. At a greater distance (25m), the results are less relevant as extra bogies contribute to the measured noise. The quite high dispersion obtained is due to the too short integration duration of LAeq and also to the measurement uncertainties. When looking at the one-third octave spectra (see Fig. 8), the noise reduction is mainly achieved for frequencies above 1600 Hz, in the range where wheel acoustic radiation is significant. The added damping of the wheel modes introduced by the ring is then real. This finding is confirmed by the acoustic map issued from the acoustic antenna (see Fig. 9) which shows that noise sources from wheels fitted with DAAVAC are less emissive than those from the other bogies. For the powered bogies, no attenuation at all is observed on the overall acoustic levels due to the predominance of traction and aerodynamic noise. Nevertheless, an effect on the wheel radiation has been extracted from the one-third octave band spectra and also from the noise sources map.

DAAVAC

REFERENCE

Fig. 9. Noise source map

4.3 Combined Effect of Solutions for the Track and the Rolling Stock The combined effect has been assessed as the difference between: -

on one hand the LAeq,Tp relative to the DAAVAC bogies passing-by the track section WITH SHREI & VEIT rail dampers (ZA), and on the other hand the LAeq,Tp relative to the REFERENCE bogies passing-by the track section without rail dampers (ZR).

For a pass-by at 238 km/h, the results measured at 7.5m from the track axis show that the noise reduction of both solutions is cumulative: -

effect of wheel rings only = 2.3 dB, issued from LAeq,Tp (Rings, ZR) - LAeq,Tp (no Ring, ZR) , effect of rail dampers only= 0.9 dB, issued from LAeq,Tp (no Ring, ZR) - LAeq,Tp (no Ring, ZA ).

62

F. Létourneaux et al.

This overall noise reduction may be slightly underestimated as the acoustic descriptor used has a very short duration and consequently focus on a small part of the total track radiation length. Then the improvement of the rate of decay of track vibration introduced by the rail dampers is not fully taken into account.

5 Conclusion Within the frame of a field test campaign on TGVs which aimed at exploring speeds up to 360 km/h, different noise mitigation measures have been investigated : one solution for the track (rail dampers) and one solution for the rolling stock (wheel damping rings) have been tested, both being designed to attenuate rolling noise. A cumulative efficiency up of about 3 dB have been achieved at 300 km/h. Due to the high speed, the solution dedicated to the reduction of the wheel acoustic radiation had the strongest effect on the overall pass-by noise (~2 dB). It has also been highlighted that the noise attenuation is directly dependent on the wheel and rail tread unevenness.

References [1] Fodiman, P.: Project NOEMIE final Report (Project n° 2002/EU/1663) (July 2005) [2] Poisson, F., Gautier, P.E., Létourneaux, F.: Noise sources for high speed trains: A review of results in the TGV case. Proceedings of the 9th International Workshop on Railway Noise, Berlin, Germany (2007)

Noise Measurement Results of Shinkansen High-Speed Test Train (FASTECH360S,Z) Y. Wakabayashi, T. Kurita, H. Yamada, and M. Horiuchi Advanced Railway System Development Center, Research and Development Center of JR East Group, East Japan Railway Company, 2-0 Nisshin-cho, Kita-ku, Saitama-shi, Saitama 331-8513 Japan Tel.: +81 48 651 2460; Fax: +81 48 651 2492 [email protected]

Summary East Japan Railway Company has developed the high-speed test train “FASTECH360” for future speed-increase of the Shinkansen. Noise reduction is one of the most important issues to be resolved in such a case. The test train equips several countermeasures to reduce noise such as low-noise pantograph, pantograph noise insulation plates, sound absorbing panels. Moreover, the car body has been made as smooth as possible to reduce aerodynamic noise. Snow plow covers, circumferential bellows, smoothed doors on the drever’s cab are some of the examples of the smoother surface. As a result of running tests up to 360km/h, it is confirmed that the noise from the test trains at a speed of approximately 330km/h is as almost equal to that of present commercial trains at a speed of 275km/h (at a distance of 25m from the track).

1 Introduction East Japan Railway Company has developed 2 high speed test trains (FASTECH 360S and 360Z) with service operation at 360km/h as a goal in technology development. FASTECH360S has 8 cars and runs only on Shinkansen line. FASTECH360Z has 6 cars and runs both on Shinkansen lines and on conventional lines converted to Shinkansen gauge. Currently, these trains are being provided for high-speed running tests between Sendai and Kitakami on the Tohoku Shinkansen Line and improved to pursue optimal solutions to various technical challenges. Noise reduction is one of the most important issues for further speed-up. In Japan, the Shinkansen noise standard requires the wayside noise at the distance of 25m away from the track to be less than 75dB(A). Therefore, the noise level at higher speed has to be still the same as present commercial trains running at 275km/h. In order to overcome this technical challenge, the test trains incorporate several countermeasures to reduce noise. In this paper, details of these countermeasures are explained to give an overview. Furthermore, effectiveness of each countermeasure is also discussed by showing some of measurement results through running tests. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 63–70, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

64

Y. Wakabayashi et al.

2 Sources of Shinkansen Noise Shinkansen noise can be classified into five categories as shown in Fig.1: (1) pantograph noise (aerodynamic noise emitted from pantograph itself, sliding noise between overhead contact lines and pantograph, and so on), (2) aerodynamic noise from the train nose (noise from the door to the driver’s cab, noise from snow plow cover, and so on), (3) aerodynamic noise from the upper parts of cars (aerodynamic noise between cars, noise generated from bumps of doors and windows, and so on), (4) noise from the lower parts of cars (rolling noise from wheels, aerodynamic noise around bogies, and so on), (5) structure-borne noise (noise induced by vibration of structure). Although these cannot be specified, this paper describes outlines of countermeasures and their effectiveness along with this classification. Both FASTECH360S and 360Z have similar countermeasures. This paper mainly refers to FASTECH360S.

Aerodynamic noise from the upper parts of cars Pantograph noise Aerodynamic noise from the train nose

Noise from the lower parts of cars

Structure-borne noise

Fig. 1. Noise sources of Shinkansen

3 Countermeasures to Each Sound Source 3.1 Pantograph Noise Most of the Shinkansen track has sound barriers 2m in height. Therefore, noise from pantographs is a great contributor of overall noise whereas noise from the lower parts of cars is not distinguishable beyond barriers. Moreover, the faster a train runs, the louder aerodynamic noise from pantographs. The currently operated Shinkansen has two pantographs per train. (4 pantographs when 2 trains are connected.) These two pantographs are connected by cable for special high-voltage to control arcs. FASTECH360 has succeeded in diminishing generation of arcs by adopting flexible structure at the contact part with overhead contact line. As a result, it became possible to run with only one pantograph per train (Fig. 2). The folded pantograph could no longer be seen from the measurement point and the noise is effectively reduced. Noise insulation plates with Z-shaped cross section are attached in limited space to house for further reduction of pantograph noise (Fig. 3). It is confirmed by simulation

Noise measurement Results of Shinkansen High-Speed Test

65

and model tests that great diffraction damping effect is obtained by this type of sound insulation plate.[1] Down

Up

Up

Down

FASTECH360S

FASTECH360Z Travel direction

Fig. 2. Pantograph positions (FASTECH360Z and FASTECH360S (connected))

Series E2 that is used in daily operation has PS207 type low-noise pantographs (Fig. 4(a)). Noise generated from this pantograph is less than that from lozenge pantographs used before. However, there is one conspicuous noise source at the center of the base frame. Therefore, we developed a pantograph called “mid-hinge type pantograph” whose component is arranged to one side and covered with windscreen covers so that the angular pipes of base frame that are thought to cause noise are no longer exposed (Fig. 4(b)). We also developed a pantograph called “Single arm type pantograph” that has no mid-hinge, another noise source, for further noise reduction (Fig. 4(c)).

Fig. 3. Pantograph noise insulation plates (Z-shaped type)

(a) PS207 type pantograph (series E2)

(b) Mid-hinge type

(c) Single arm type

Fig. 4. Low-noise pantograph for series E2 and new low-noise pantograph for FASTECH360

66

Y. Wakabayashi et al.

3.2 Noise from Lower Parts Although lower parts of car bodies are hidden behind sound barriers, their contribution to overall noise becomes greater because of improved pantograph noise. In addition, because noise is much more prominent in the high-speed range for which FASTECH360 is aiming, the scope should be extended to noise from lower parts. FASTECH360 equips bogie covers at the side of all bogies. Sound absorbing panels are also attached to the lower side and the bottom of car bodies to absorb noise emitted from the lower parts and from the ground (Fig. 5). Noise is diminished in multi-reflection process between car bodies and sound barriers. 3.3 Aerodynamic Noise from the Train Nose Aerodynamic noise from the train nose is mainly composed of noise from the front bogie, the door to the driver’s cab, handrails for drivers, and the snow plow cover. In order to reduce this noise, snow plow covers and smoothed doors to the driver’s cab are adopted in addition to bogie covers mentioned above. Fig. 6 shows snow plow covers.

(a) Lower side of car bodies

(b) Bottom of car bodies

Fig. 5. Sound-absorbing panels for lower side and bottom of cars

Fig. 6. Snow plow covers

Fig. 7. Circumferential bellows

3.4 Aerodynamic Noise from Upper Parts One of the most dominant noise source from upper parts is gaps between cars. A countermeasure against this noise is circumferential bellows. Moreover, doors and windows are made as smooth as possible to avoid aerodynamic noise from bumps. Fig. 7 shows snow plow covers.

Noise measurement Results of Shinkansen High-Speed Test

67

4 Measurement Methods and Results 4.1 Measurement Using Spiral Microphone Array [2] A spiral microphone array was used in order to specify sound sources of FASTECH360S. Fig. 8 shows a schematic diagram of the measurement. Fig. 9 shows a sample of measurement results. As shown in Fig. 9(a), sound sources around pantographs which are generally thought to generate louder noise are distinguished and rear edges of pantograph noise insulation plates. It was also clarified that louder noise was generated around some parts of wheels and circumferential bellows. 10m Spiral microphone array

Line sensor camera

5m

4m Slab track

Noise barrier removed

Fig. 8. Schematic diagram of measurement using spiral microphone array

(a) Primary result of high speed tests

(b) After improvement

Fig. 9. Noise source distribution of FASTECH360S using spiral microphone array (340km/h)

Based on these results, improvements of FASTECH360S were made. The shape of noise insulation plates was changed to plates with thirty degree slope at both edges. The shape of the brake disk was also changed because it was found that the noise from wheels was due to aerodynamic noise generated from cooling fins behind brake disks. Moreover, gaps of circumferential bellows were adjusted so as to diminish the noise sources. As for pantographs, the mid-hinge type pantograph was replaced with a single arm type pantograph since it was confirmed that the single arm type pantograph was superior to the mid-hinge type pantograph in noise reduction.

68

Y. Wakabayashi et al.

As can be seen in Fig. 9(b), noise sources were greatly reduced after these improvements. Almost the same improvements were made with FASTECH360Z. 4.2 Measurements Using a Super Directional Microphone and a Normal Sound Level Meter Both a super directional microphone (time constant 35ms) and a normal sound level meter were set at 25m away from the nearest track. The schematic diagram of measurement using these microphones is shown in Fig.10. An example of measurements using the super directional microphone is shown in Fig. 11.

1.8m Non-directional microphone

Slab track

R.L

Microphone array

8.9m 1.2m 25m Fig. 10. Schematic diagram of measurement using non-directional microphone and microphone array Inter-car gap peak

1sec

10dB

sou n d pr essu r e level [dB]

Pantograph peak

FASTECH360Z

FASTECH360S Time

Fig. 11. An example of measurements using the super directional microphone

Pantograph peak level using the microphone array and inter-gap peak level (between the first car and the second) using microphone array are shown in Fig.12 and Fig.13 respectively. Fig.12 shows that with both low noise pantographs and the improved sound insulation plate, the pantograph peak level of FASTECH360S is reduced by more than 2dB compared to series E2 (commercial train used for Shinkansen line).

Noise measurement Results of Shinkansen High-Speed Test

69

When it comes to FATECH360Z, the level is reduced by more than 5dB compared to series E3 (commercial train for through-operation). Note that the peak level of the folded pantograph is less than unfolded one because of the difference of hidden areas behind Noise insulation plates. Fig.13 shows that the inter-car gap peak level of FASTECH360S is reduced by 1 to 2 dB compared to E2 and that of FASTECH360Z is reduced by approximately 4dB compared to E3. This is attributed by the noise reduction effect of circumferential bellows and sound absorbing panels. It is considered that sound absorbing panels show their ability especially around slab track.

] B d[ l e v e L k a e P hp ar g ot na P

.

1dB

FASTECH360Z (Up pantograph)

.

FASTECH360Z (Folded pantograph)

.

FASTECH360S (Up pantograph)

.

FASTECH360S (Folded pantograph)

.

Series E3 (Up pantographr)

.

Series E2 (Up pantograph)

240

260

280

300

320

340

360

380

Train speed [km/h]

Inter-car Gap Peak Level [dB]

Fig. 12. Pantograph peak level using microphone array

FASTECH360Z FASTECH360S Series E2 Series E3

1dB

240

260

280

300

320

340

360

380

Train Speed [km/h] Fig. 13. Inter-car gap peak level (between the first car and the second) using microphone array

70

Y. Wakabayashi et al.

A-weighted Sound Pressure Level [dB]

Fig.14 shows A-weighted sound pressure level using non-directional microphone. The time constant is 1s, which generally used to evaluate wayside noise level. The noise level of FASTECH360S+360Z (connected) is 4 to 5 dB less than that of E3+E2 (connected). Although the goal “360km/h” is still ahead, it is confirmed that the noise from the test trains at the speed of approximately 330km/h is as almost equal as that of the present commercial train at the speed of 275km/h.

1dB

FASTECH360Z+FASTECH360S FASTECH360S Series E3+Series E2(Commercial Train) Series E2(Commercial Train)

240

260

280

300

320

340

360

380

Train Speed [km/h]

Fig. 14. A-weighted sound pressure level using non-directional microphone

5 Conclusions (1) The noise level of FASTECH360S and 360Z (connected) at a speed of approximately 330km/h is as almost equal to that of the present commercial train at a speed of 275km/h. (2) With both low noise pantographs and the improved sound insulation plate, the pantograph peak level of FASTECH360S is reduced by more than 2dB compared to series E2, while that of FASTECH360Z is reduced by more than 5dB compared to series E3. (3) The inter-car gap peak level of FASTECH360S is reduced by 1 to 2 dB compared to E2 and that of FASTECH360Z is reduced by approximately 4dB compared to E3.

References [1] Ido, A., Kurita, T., Wakabayashi, Y., Hara, M., Shiraishi, H., Horiuchi, M.: Development of Technologies for Minimizing Environmental Impacts. In: Proceedings of 7th World Congress on Railway Research (2006/6) [2] Takano, Y., Sasaki, K., Satoh, T., Murata, K., Mae, H., Gotoh, J.: Development of Visualization System for High-Speed Noise Sources with a Microphone Array and a Visual Sensor. In: Proceedings of Inter-Noise 2003, vol. 930 (2003/8)

Noise Sources for High Speed Trains: A Review of Results in the TGV Case F. Poisson, P.E. Gautier, and F. Letourneaux SNCF Innovation and research department, 45 rue de Londres 75379 Paris, France Tel.: +33 1 53 42 92 39; Fax: +33 1 53 42 92 84 [email protected]

Summary For the past 15 years, a number of tests at high speed were carried out for different trains in different countries (France, Germany, Italy, Japan, Spain …). In the TGV case, apart from test at operating commercial speeds, various acoustical measurements were carried out, either within dedicated programs (Deufrako Cooperations K and K2, 350 kph measurement campaign a few years ago on TGV Duplex, and more recently 360 kph test on TGV Reseau), or at the occasions of very high speed campaigns. From the data gathered in each measurement campaign with a single microphone and an antenna, successive methods of sources characterisations were developed. After a review of the global noise level of the high speed trains in Europe, the regression law of the sound pressure level according to the train speed is investigated. Limits of the antenna measurement are presented and several ways of improvement are proposed. Source characterisation becomes a more and more important issue as these data are used as input for acoustical models of train pass-by in the design stage.

1 Introduction Noise from high speed trains is a sensitive issue, as high speed train lines are built either in densely populated areas, or conversely in zones where the pre existing noise was very low. High interest was then given, from the 1990’s, to the measurement of pass-by noise from high speed trains, and especially in France to the TGV case, with emphasis on the evolution with running speed. More recently, train pass-by noise was considered as an “interoperability parameter” and thence limited in the high speed and conventional rail Technical Specification for Interoperability (TSI) [1] concerning the subsystem Rolling stock. The goal was to limit train pass- by noise for all trains in Europe. Then, for the past ten years, pass-by noise measurement campaigns at high speed were carried out for different trains in different countries (France, Germany, Italy, Japan, Spain…): “Deufrako cooperation K” [2] and “K2” [3], “NOEMIE”, “V350” and “V360” SNCF measurement campaigns and more recently at the occasion of the very high speed campaign “V150”. From the data gathered in each configuration, regression laws have been extracted and source characterization performed. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 71–77, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

72

F. Poisson, P.E. Gautier, and F. Letourneaux

This article presents a review of the pass-by noise of high-speed trains in Europe. Results of a measurement campaign carried out with a TGV with high roughness wheels are also discussed. Then the evolution of the pass-by noise of a TGV according to the speed is investigated. Main conclusions of the characterization of the noise sources through antenna measurement campaigns are presented. The source models are also addressed through the limitations of the existing array processing methods.

2 Equivalent Sound Pressure Levels 2.1 Review of the Pass-by Noise Levels for the European High-Speed Trains A review of the results of the measurement campaigns carried out the past ten years is presented in table 1. The dispersion of external noise from the different series of TGV’s is around 1.5 dB(A), when only analyses on the same track are taken into account. Moreover, different series of trains from different countries (TGV, ICE, ETR…) show very close values when measured at the same site: identical values for 300 kph and above, and up to 2 dB(A) at 250 kph. It then can be asserted that, the overall dispersion of the pass-by noise of different types of high speed trains is narrow. Table 1. Pass-by noise values of high speed trains measured at 25m

Pass-by noise values Test site measured at 25m in dB(A) TSI+ tracks except Belgium

TGV Thalys

TGV Duplex TGV Atlantique TGV Réseau ICE3 AVE ETR480 ETR500 TSI limits

Belgium France Germany France France France France Germany Spain Italy Italy TSI+

Train speed (kph)

250

300

320

88.5 85.5 85.5 87

92 90

93 92

91 90.5 91.5 90 89 90

92 94 (330kph) 91.5 92 91

90.5 92

94

89 87.5 85.5 86 90.5 88 -

350

95 94.7 97

-

2.2 Influence of the State of the Wheels During the “V360” measurement campaign, pass-by noise of a TGV-Réseau was recorded at different stages of the measurement campaign in the same location. It appeared that following track works at a few defined dates during the test campaign, some ballast dust might have been run over by the wheels, the roughness of which significantly increased on the following days. The measured noise values were then

Noise Sources for High Speed Trains: A Review of Results in the TGV Case

73

increased by 1.5 to 2.5 dB(A) after each track work episode, and the influence of the increase of the measured pass-by level can be noticed throughout the whole investigated speed range : 250 to 360 kph (see table 2). Table 2. Influence of the wheel surface quality on pass-by noise (TGV Réseau)

Pass-by noise values at 25m (dB(A)) TGV Réseau TGV Réseau “corrupted wheel state”

250 89 93

Train speed (kph) 300 91.5 95.5

330 94 97

It can also be inferred from the latter observation that the transition speed between rolling noise and aerodynamic noise for the TGV Réseau is higher than often previously claimed, when it had been said to lie around or under 300kph. Following that hypothesis, the influence of rolling noise created by the wheel roughness up to 360 kph would not have been so significant: for the TGV Réseau case, the upper values of measured pass-by noise (for corrupted wheel state) went up to 97 dB(A), whereas values measured on the TGV Duplex in a former campaign were not higher than 94 dB(A) at 350 kph. These observations also still confirm in this case the applicability of the “30 logV” regression rule (1), which was shown in [3] and which is characteristic of rolling noise dominated behaviour. 2.3 TGV Pass-By Noise Versus Speed In the last SNCF measurement campaign, a TGV POS which is composed of Duplex power cars and 8 single floor coaches has been measured from 100 kph to 380 kph. The LAeq,tp measured at 25m are presented figure 1. A linear regression has been performed between the increase of LAeq,tp and the logarithm of the train speed. The general equation is: LAeq,tp(V) - LAeq,tp(V0) = K log(V/V0)

(1)

with V the train speed, V0 the reference train speed. Between 200 kph and 380 kph, the regression coefficient is K=30.4 with a correlation coefficient R2 equal to 0.93. This regression coefficient has already been checked within the frame of the Deufrako projects [2] [3] and reconsidered in [4]. It is nearby the value of 30 commonly used in the prediction formula for rolling noise which is widely used to extrapolate noise emission of classical trains. It confirms that the contribution of the rolling noise, which is the main noise source for conventional speeds, remains high at 380 kph for a TGV train set which comply with the TSI limits. Then, a significant reduction of the pass-by noise of a TGV train set running at commercial speed (~320 kph) can be reached by acting both, on the aerodynamic sources and the rolling noise sources, as presented in [4]. The “V150” measurement campaign carried out in France with Réseau Ferré de France and Alstom will be a good opportunity to check the upper limit of this regression law.

74

F. Poisson, P.E. Gautier, and F. Letourneaux

Fig. 1. LAeq,tp for TGV POS from 200 kph to 380 pkh

3 Source Models 3.1 Deufrako Measurement Campaigns Many of the measurement campaigns were, as soon as the DEUFRAKO K and K2 campaigns accompanied by acoustic array measurements. The acoustic array Table 3. Estimation of the level of the sources of a TGV A (DEUFRAKO project) [1]

Level at 5m dB(A) Wheels, coach Wheels, forward power car Wheels, rear power car Pantograph Cooling fan, front Cooling fan, rear Front window / Roof Intercar gap Bogie (aero)

100 kph

200 kph

300 kph

350 kph

82.4 82.9

89.7 89.4

97.2 102.5

98.1 100.4

81.7

89.6

102.7

98

82.1 78.9 78.3 77.9

91.3 87.9 79.6 88.5

103.5 101.3 102.1 102.2

104.3 104 100.5 104.9

81.7 76.8

87.1 78.6

92 90.1

98.1 93.3

Noise Sources for High Speed Trains: A Review of Results in the TGV Case

75

measurements in DEUFRAKO K project was further analysed by INRETS, where a first estimation of the level of the sources was attempted, by summing the acoustical energy received on a zone surrounding the assumed position for the source. Later in DEUFRAKO K2 some monopole fitting to the array results was attempted. Unfortunately, in order to prove the method, a loudspeaker was put onboard the train. It appeared that the loudspeaker source was not retrieved by the method. Then, the exploitation of the method on other sources less precisely located on the train could not be reliably done. 3.2 “V350” Measurement Campaign The tests carried out with TGV Duplex at 350 kph also involved acoustic array measurements. An example of the obtained results is given in figure 2.

Fig. 2. Example of a TGV Duplex noise maps obtained with an antenna of microphones Above: third octave band 500 Hz / Below: third octave band 4000 Hz

For processing the TGV Duplex results at 350 kph, a different method was developed [4], where sources contents in third octave bands could be estimated. A summary of the results is given in figure 3. The comparison of table 3 and figure 3 leads to the following conclusions: -

-

At 200 kph, the rolling noise is an important source but the noise radiated by the area located around the first bogie (bogie and windscreen) is of the same order, At 300 kph, the area around the first bogie is the main source but the noise radiated by the cooling and the pantograph can not be neglected, At 350 kph, the area located around the first bogie and the pantograph radiates much more.

These data are currently used as input of a pass-by noise simulation software like MAT2S [5] and VAMPASS [6] to assess the contribution of each source to global pass-by noise. The most radiating source is not always the main source in terms of contribution to the pass-by noise due to the number of sources, their location, their spectra, …

76

F. Poisson, P.E. Gautier, and F. Letourneaux

Fig. 3. Example of source power identification on TGV Duplex running at 350 kph (SPL of the most energetic third octave band of the source)

3.3 “V360” Measurement Campaign Measurement devices have been improved in the last five years which allow to use more easily much more microphones. Now, SNCF is using a star shape antenna of 72 microphones located near the track. The beamforming technique is still used after the removing of the Doppler effect to characterise the sources. Noise maps are presented for each third octave band and narrow band spectrum can be extracted for a given position. An example is given figure 4 where the noise emitted by the electrical equipment of the head power car is highlighted around 1580 Hz. Work is now in progress to improve the characterisation of moving sources using the antenna. The main drawbacks of the existing methods are: -

The estimation of the sound pressure level of the source is carried out during the tracking of the source. The directivity pattern of the antenna is evolving during the tracking, with a poor selectivity on both sides and a better one in the front. Error is then introduced in the assessment of the sound pressure level of the source.

Noise Sources for High Speed Trains: A Review of Results in the TGV Case

77

Fig. 4. Noise map of a TGV POS running at 318 kph (1580 Hz)

-

Several size of the antenna must be used to cover the whole frequency band from 100 Hz to 8 kHz to avoid aliasing, which lead to different directivity patterns according to the frequency band. Then, noise maps can not be compared from one configuration to another one.

Recent developments in array processing give interesting ways of improvement to get rid of this drawbacks; simulations must be carried out to develop these methods in the field of a railway application.

4 Conclusion Different series of high speed trains from different countries (TGV, ICE, ETR…) show very close values on the same measurement site. The regression law of the global sound pressure level according to the train speed was investigated during the last measurement campaign with a TGV POS train set. The regression coefficient 30 log(Vtrain) is valid up to 380 kph which confirms that the contribution of the rolling noise remains important for TGV running at commercial speeds (300 kph or 320 kph). For antenna measurement, the recent improvement of the measurement devices must be accompanied by an improvement of the array processing itself to develop an accurate and robust method to characterize the sources. The sources characteristics becomes more and more important data as they are used as input of pass-by simulation software as one’s developed in [6]. This software is needed to carry out parametric studies to define the most relevant combination of noise reduction solutions to reduce the noise of existing and future trains.

References [1] Technical specification for interoperability relating to the rolling stock subsystem of the trans European high-speed rail system, 30/05/2002. official journal of the EC 12 (September 2002) [2] German-French cooperation, Annex K, Final Report (December 1994) [3] German-French cooperation, Annex K2, Final Report (December 1999) [4] Mellet, C., Létourneaux, F., Poisson, F., Talotte, C.: High Speed Train emission: Last investigation for the aerodynamic / rolling noise contribution. In: IWRN (2005) [5] Poisson, F., Gautier, P.-E., Fortain, A., Margiocchi, F.: Pass-by noise reduction at 350 kph: A parametric study. In: WCRR (2006) [6] Bongini, E., Molla, S., Gautier, P., Habault, D., Mattéi, P.O., Poisson, F.: Synthesis of noise of operating vehicles: development within SILENCE of a tool with listening features. In: IWRN (2007)

Survey of Metro Excitation Frequencies and Coincidence of Different Modes S.J. Cox, A. Wang, and A. Adedipe Pandrol Rail Fastenings, 63 Station Road, Addlestone, Surrey KT15 2AR, UK Tel.: +44 1932 834500; Fax: +44 1932 850858 [email protected]

Summary Six characteristic frequencies that occur on metro systems have been investigated, with the objective of determining whether there is likely to be a coincidence of any of these that might lead to adverse behaviour of the vehicle–track system. Data and measurements drawn from a number of metro systems have been considered – all of these were at locations with direct fixation track. From the characteristic frequencies investigated, it has been found that only the fastener-passing frequency and the P2 loaded track resonance are likely to coincide. The parameters for which this is most likely to occur are discussed.

1 Introduction Trains passing along railway tracks excite vibrations because of the roughness on the wheels and rail. Because this roughness contains components with a wide range of wavelengths, a wide range of frequencies is excited. In addition, there is parametric excitation of vibrations at particular frequencies that are related to the geometry of the system – for instance the frequency with which axles, bogies, and vehicles pass above a given position; with which a wheel passes over the fasteners; and so on. Here the frequencies excited are determined by the dimensions of the track and vehicles, and by the train speed. There are also particular frequencies associated with resonances in the train-track system. These characteristic frequencies of the system are not affected by train speed. Where behaviours with different underlying causes occur at the same or similar frequencies, the results are likely to be difficult to predict - no attempt is made here to do so - and may be undesirable. The subject of this paper is how some of the vehicletrack system resonance frequencies on metro systems relate to the frequencies of parametrically excited vibration, and whether it is possible to avoid coincidence of frequencies by appropriate design.

2 Characteristic Frequencies on Metro Systems A number of characteristic frequencies are excited on any conventional metro system. To assist in the following discussion, here these have been (arbitrarily) divided into those B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 78–85, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Survey of Metro Excitation Frequencies and Coincidence of Different Modes

79

occurring at low frequencies below 10 Hz; medium frequencies between 10 and 100 Hz; and high frequencies above 100 Hz. The frequencies that are considered here are: • •

• •

•

The frequency with which bogie-centres pass a given position on the track. This is determined by the bogie-centre spacing and the train speed, and is usually a low frequency. The frequency with which axles pass a given position on the track. The spacing and frequency considered here is that between adjacent axles on one bogie – that between the last axle on one vehicle and the first on the next may differ slightly, so that two similar frequencies will be excited. The frequency is determined by the axle spacing and the train speed. For metro traffic, this is usually a low frequency. The frequency with which a given axle on the train passes above regularly spaced rail fastenings. This is determined by the fastener spacing and the train speed. For metro traffic, this is usually a medium frequency. The ‘P2’ or ‘loaded track resonance’1. This frequency does not depend on the train speed, but on the unsprung mass and the stiffness of the track. On directly fastened track on a rigid base slab, the track stiffness depends in turn on the stiffness and spacing of the fasteners and the resonance usually occurs at medium frequencies, but for very soft track forms such as floating slab track, it may be at low frequencies. The ‘pinned-pinned’ frequency at which the rail vibrates with nodes at the fastening positions. There are at least two modes of this type – those that are considered here are the vertical mode and the lateral mode. These are usually high frequencies.

These frequencies were selected because the data required to estimate them had already been gathered during a number of measurement exercises; because they will all occur on all metro systems; and because there is at least one possible coincidence of frequencies that might be avoided by making appropriate design decisions. Other frequencies that are no less significant will also be present on metro systems, but are not considered here. These include: • •

•

1

The many resonance frequencies associated with the vehicle, including the low frequency bounce, pitch, and yaw modes of the body of the vehicle on its suspension. Bogies have a kinematic wavelength over which they naturally oscillate as they travel along the track. The wavelength is determined by geometry of the bogie, the wheels, and the track. The frequency increases with train speed, but usually occurs at low frequency. Other periodic features in the track system such as tunnel elements or discrete track slabs will generally give rise to low frequency vibrations as metro trains pass over them.

The ‘P1’ and ‘P2’ forces were originally identified by Jenkins [1] as the forces generated at a dipped joint and dipped welds respectively. The terms are used here for convenience because they succinctly identify the loaded track resonance, which is that associated with high forces at dipped joints, and the contact spring resonance, which is that associated with high forces at dipped welds.

80

S.J. Cox, A. Wang, and A. Adedipe

•

• •

•

Additional resonances (occasionally called ‘P1½’ – for example by Tunna [2]) within the track system. These occur where two or more layers of resilience are separated by intermediate mass elements. Examples include floating slab track (FST), with bearings beneath and rail fastenings above; resilient booted blocks or sleepers with railpads; and resilient baseplates with railpads. The frequencies generated vary widely, depending on the particular design. There are also bending resonances of resiliently supported track elements such as sleepers or slab panels, which usually occur at high frequencies. Torsional and bending resonances of the wheelset associated with wind-up of the axle. The lowest resonant frequency on most metros is usually in the medium frequency range. Rail roughness may include specific frequencies with high amplitudes, such as corrugations. In many cases the wavelength fixing mechanism for the corrugation will be one of the system characteristic frequencies or resonances described, so that the particular frequency excited by corrugations will, by definition, be coincident with that frequency. On metro systems corrugations usually give rise to high frequency excitation. The ‘P1’ resonance of the wheel mass on the wheel-rail contact spring1 always occurs at high frequencies.

3 Data Collected on Typical Metro Systems The measurements on metro systems cover a wide geographical distribution in Europe, North America, and Asia. A total of 26 separate measurements have been considered. There are measurements on light rail systems with 8-10 tonne axle loads and heavy metros with 16-18 tonne loads. The largest rail had a mass of a little over 60 kg/m, and the smallest a mass of a little less than 40 kg/m. All of the measurements were made on directly-fastened tracks with a rigid concrete base slab to which the rail was fixed with either baseplates, simple fasteners, booted blocks, or embedded sleepers. None of the measurements discussed here were made on floating slab track, and none on ballasted track. When each set of measurements was made, some basic parameters relating to the trains and to the track were collected. These included the vehicle bogie centre spacing and axle spacing, the rail sections and fastener spacing, the axle load, and where available, the unsprung mass of the vehicles. The measurements included dynamic deflections of the rail and, in some cases rail and base slab vibrations. The speed of the trains at the measurement site was calculated from the time history of the data recordings. The dynamic stiffness of each rail fastening was estimated from the dynamic deflection using a beam-on-elastic-foundation model [3]. For each measurement location and track configuration, six characteristic frequencies were calculated using the formulas given in the Appendix. These calculated frequencies are identified in Figure 1 on a vibration spectrum obtained from measurements made on one typical metro. There are a number of peaks in the spectrum, and as would be expected, several of these correspond well with the characteristic

Survey of Metro Excitation Frequencies and Coincidence of Different Modes

81

-30 -40

Fastener Passing

Velocity dB (ref 5E-8m/s)

-50

P2

-60

PinnedPinned

Bogie Passing

-70

PinnedPinned

-80 Axle Passing

-90 -100 -110 -120 -130 0.1

1

10 Frequency (Hz)

100

1000

10000

Fig. 1. Spectrum of vibration on one metro system

Measurement Configuration

frequencies calculated from the train speed and dimensions and parameters for the track and the trains. The figure confirms that resonances in the vehicle track system lead to high levels of track vibration, and that the frequencies at which a number of these occur can readily be calculated from a number of simple parameters. Many of the factors that affect these six characteristic frequencies lie within fairly narrow bands. Except for one particular measurement location where the operating speed was atypically low, the maximum and minimum values of bogie centre, axle spacing, fastener spacing, and train speed collected on metros all lie within about 30% of the mean. Consequently, many of the frequencies calculated lie within fairly narrow bands. The six characteristic frequencies calculated for each measurement

0.1

1.0

10.0

100.0

1000.0

Frequency (Hz) Bogie-passing P2 resonance

Axle-passing Pinned-Pinned Lateral

Fastener-passing Pinned-Pinned Vertical

Fig. 2. Some characteristic frequencies occurring on metro systems

10000.0

82

S.J. Cox, A. Wang, and A. Adedipe

configuration on all of the metro systems considered, are plotted below in Figure 2. There are 26 horizontal lines, each of which represents one measurement configuration. In Figure 2, the data have been arranged in order of increasing P2 frequency. Where the unsprung mass of the vehicle was not known, a fixed value of 800kg was used (and in many cases the calculated P2 frequency was then confirmed by reference to the spectrum of rail vibration recorded at the site). At these frequencies, the unsprung mass can be assumed to be the mass of the wheel plus a proportion - often taken to be 1/3 - of the axle mass. It can be seen that the bogie-passing, axle-passing, fastener-passing, lateral pinnedpinned and vertical pinned-pinned frequencies always occur in this order, and are generally well spaced from each other. The only confusion is between the P2 frequency and the fastener-passing frequency. Where the P2 resonance occurs at frequencies below about 40Hz, the P2 and fastener-passing frequencies may be similar. There are two instances in the data where the P2 resonance occurs at a higher frequency than fastenerpassing, and two where the two frequencies are nearly coincident.

4 Discussion In practice, at none of the locations in Figure 2 where the P2 and fastener-passing frequencies are similar were any obvious ill effects (for example rail-head wear or damage) visible on the track. At one site, there were short wavelength corrugations, but the corresponding frequency was much higher than the P2 frequency and appeared to fixed by the vertical pinned-pinned resonance [4]. Similar corrugations also appear elsewhere on the same line but where the track stiffness is much higher, so that here there is no coincidence between the P2 and fastener-passing frequencies. It is difficult to predict the combined effect of parametric excitation such as that generated by fastener passing with that of excitation of track resonances by wheel-rail roughness. Where attempted, the approach is usually to estimate the effects separately and assume that they can be combined by linear addition, as for example in the work to predict tunnel floor vibration through different mechanisms by Vibratec [5]. Nevertheless, an assumption is made here that a coincidence of the P2 and fastener-passing frequencies is undesirable. Is it practical to consider tuning the P2 resonance or the fastener-passing frequency to avoid a clash between the two? Before addressing this question, some situations are considered where resonances in railway track systems have had undesirable effects, and proposals have been made (and in some cases implemented) to tune the system to avoid these. For example, the mechanism that fixes the wavelength of the rail-head corrugations that develop on some metro systems is often a resonance in the track system. This may be the vertical pinned-pinned resonance; the P2 resonance; or others - a number of examples are given in Elkins [6]. A modification to the system that moves the resonance out of a given frequency range may prevent development of the corrugation. For instance, the vertical pinned-pinned resonance is strongly dependent on fastener spacing. Reducing the spacing increases the pinned-pinned frequency until, at a point when the corresponding wavelength becomes shorter than the contact patch between the wheel and rail (usually 10-15mm in length), in principle corrugations will no longer develop. Another example of tuning a single resonance to a different frequency range to avoid corrugation development is the introduction of softer railpads to reduce the P2 frequency [7]. There are also more

Survey of Metro Excitation Frequencies and Coincidence of Different Modes

83

complex situations where the co-incidence of two resonances has undesirable results. Grassie notes [8] that co-incidence between the P2 resonance and the first torsional resonance of the wheelset can give rise to corrugation, and an example is described by Tassilly [9], where a change in the railpad stiffness to reduce the P2 resonance had beneficial results on corrugation growth rates in similar circumstances. So there are some precedents for the principle that the loaded track resonance can be chosen to promote or avoid certain wear patterns of the track. Returning to co-incidence between the P2 resonance and the fastener passingfrequency, the former will increase if the track stiffness is increased (either by increasing the stiffness of each fastener, or by reducing the fastener spacing) or the vehicle unsprung mass is reduced. As Figure 2 shows, coincidence between the P2 resonance and fastener-passing frequencies can be avoided on metros when the former occurs at frequencies of 40 Hz or more. However, this cannot be achieved simply by increasing the track stiffness, because this has an adverse effect on control of ground vibration and on wheel-rail contact forces. One of the prime reasons for lowering the P2 resonance of the track system is to reduce the transmission of railway vibrations into the surroundings. Lower dynamic track stiffness results in a lower P2 resonance frequency, and generally results in lower transmitted vibration levels at frequencies greater than this [10]. A co-incidence between the P2 resonance and the fastener passing-frequency is likely to have adverse consequences on vibration transmission into the environment at the P2 frequency. Scope for other measures to avoid a clash is limited. A low unsprung mass will tend to increase the frequency of the P2 resonance and to reduce wheel-rail contact forces. However, there are limits as to what can be achieved, and most modern rail vehicles are already designed to minimize unsprung mass. Nor are adjustments in the fastener spacing particularly effective in avoiding coincidence between the fastenerpassing frequency and the P2 resonance, because an increase in fastener spacing will not only decrease the fastener-passing frequency but also the P2 resonance (though the changes occur at different rates). Fasteners spacings are also limited by the need to maintain safe working limits on rail deflections and stresses and on fastener loading. In many situations it may be that all that can be done, given these constraints, is to be aware of the circumstances in which the potential for clashes is greatest. In practice on those metro systems where the rail is fixed directly to a rigid base slab with socalled ‘direct fixation’ fastener systems, a coincidence between the P2 resonance and the fastener-passing frequency is most likely to occur where the fastener stiffness is low (< 20 kN/mm); or the fastener spacing is small (< 0.6m); or the unsprung mass is high (> 1000kg); or the train speed is relatively high (> 60 km/hr); or particular if two or more of these conditions apply. The potential for clashes between the P2 resonance and the fastener-passing frequency may be greater on track systems with significant added mass such as FST, or ballasted track with ballast mats. Here the P2 resonance will generally occur at lower frequencies, and there may also be the possibility of coincidence between the P2 resonance and the axle passing frequency. Similarly on tracks where the vehicle speeds are significantly higher than on metro systems, axle-passing and fastenerpassing will occur at higher frequencies, and there may be increased likelihood of coincidence between the P2 resonance and both of these frequencies.

84

S.J. Cox, A. Wang, and A. Adedipe

5 Conclusions Six characteristic frequencies that occur on all metro systems have been discussed, with a view to determining whether there is likely to be a coincidence of any of these that might lead to adverse behaviour of the vehicle–track system. No attempt has been made to predict what these effects might be if they do occur. There are other characteristic frequencies present in the system that have not been considered here. Of the characteristic frequencies investigated, only the fastener-passing frequency and the P2 loaded track resonance are likely to coincide on metro systems with direct fixation track in which the rail is fixed to fixed slab with resilient fasteners. This is most likely to occur when the fastener stiffness is low, the fastener spacing is small, the vehicle unsprung mass is high, the train speed is relatively high, or particular if two or more of these conditions apply. Coincidences of the P2 resonance with the axle-passing or the fastener-passing frequency is more likely to occur on metro systems on track that has added sprung mass and lower P2 resonant frequencies such as FST, or on railway systems with higher speeds than those prevalent on metros.

References [1] Jenkins, H.H., Clayton, G.A., Morland, G.W., Lyon, D.: The effect of track and vehicle parameters on wheel/rail dynamic forces. The Railway Engineering Journal, IMechE 3(1) (January 1974) [2] Tunna, J.M.: Wheel/Rail forces due to wheel irregularities. In: Proceedings 9th International Wheelset Congress, Montreal, paper 6–2 (1988) [3] Grassie, S.L., Gregory, R.W., Harrison, D., Johnson, K.L.: The dynamic response of railway track to high frequency vertical excitation. J. Mech. Engng. Sci. 24, 77–90 (1982) [4] Grassie, S.L., Edwards, J.W.: Development of corrugation as a result og varying normal load. In: Proceeding sof 7th International Conference on Contact Mechanics and Wear of Wheel Rail Systems Brisbane, Australia (2006) [5] CONVURT Project: Development of the excitation model Vibratec Report 450.003.RA.03B (December 2003) [6] Transit Cooperative Research Program - Research Results Digest. Rail corrugation mitigation in transits (26) (June 1998) [7] Grassie, S.L., Kalousek, J.: Rail corrugation: characteristics, causes and treatments. Journal of Rail and Rapid Transit, Proc. of Inst. Mech. Eng. 270F, 57–68 (1993) [8] Grassie, S.L.: Rail Corrugation: advances in measurement, understanding and treatment. Wear 258, 1224–1234 [9] Tassilly, E., Vincent, N.: Rail corrugations: Analytical model and field tests. Wear 144, 163–178 [10] Cox, S.J., Wang, A.: Effect of track stiffness on vibration levels in railway tunnels. Journal of Sound and Vibration 267, 565–573 (2003)

Appendix

Formulae used to calculate characteristic frequencies fb,a,s = vs / xb,a,s fp ≈ (1/2π).(1/ mw.αr)½ fx,y = (1/2π).(π / xs)2.(EIx,y / mr)½

Survey of Metro Excitation Frequencies and Coincidence of Different Modes

85

bogie-centre, axle, fastener-passing frequency fb,a,s loaded track resonance ‘P2’ frequency fp pinned-pinned resonance in the appropriate direction f x,y bogie-centre, axle, fastener spacing* xb,a,s train speed† vs mw vehicle unsprung mass* dynamic track receptance (deflection per unit applied load)† αr rail bending stiffness in the appropriate direction* EIx,y mr rail mass per unit length* * input data collected from static measurements, drawings, etc. † input data calculated from measurements of dynamic track deflections.

where

Floating Slab Track above Ground for Turnouts in Tram Lines Hans-Georg Wagner and Axel Herrmann GERB Schwingungsisolierungen GmbH & Co. KG, Ruhrallee 311, 45136 D-Essen, Germany Tel.: +49-(0)201-26604-21; Fax: +49-(0)201-26604-50 [email protected]

Summary There is undoubtedly a remarkable revival of trams. Trams are an effective means of mass transit that can move large crowds directly to the downtown shopping and work areas. Inevitably, some tracks will have to be routed through narrow streets in close proximity to buildings. In these cases, abatement measures are required to avoid the transmission of noise and vibration which otherwise would annoy residents. Then the installation of a floating slab system should be taken into consideration. Especially in the areas of crossings and turnouts, high-performance floating slabs are often the best choice. The present paper highlights low-frequency floating slabs using steel springs thus achieving the high reduction in noise and vibration levels. Two projects commissioned in 2006 are described.

1 Introduction Crossings and turnouts are critical components of railroad systems. They generate sound and vibration emissions far higher than regular track segments. In some cases, elastic rail bearings are not viable solutions. Furthermore, nearby buildings with “soft” slabs are often easily excited by low frequency rail activity. Here, vibration control systems with higher isolation efficiency should be used. Floating track slabs, if designed properly, provide the highest possible degree of isolation. There are cases, where, due to cost reasons, within a track section only turnouts and crossings are isolated.

Fig. 1. Typical floating slab for a turnout in a tram line at grade level

2 The Mass-Spring-System (MSS) The MSS consists of a concrete slab supported by an elastic vibration isolation layer. The rail system is mounted directly on the concrete slab usually fixed by means of B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 86–93, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Floating Slab Track above Ground for Turnouts in Tram Lines

87

elastic fastenings. The elastic vibration isolation layer can be an elastomer material or consist of helical steel springs. The degree of vibration isolation greatly depends on the material selected in addition to the weight and geometry of the slab. Effective vibration isolation can only be achieved when the system natural frequency of the trackbed is, by a factor of > 2 , below the expected excitation spectra. Railroad systems typically generate a wide spectrum of high energy excitations between 40 - 80 Hz. However, the concentration of excitation can even be as low as 10 Hz. A MSS with a 5- 8 Hz vertical natural support frequency will provide effective isolation in the entire excitation spectra, even in the case of excitations in the 10 Hz range.

Fig. 2. Plan of a FST turnout slab in a tram line at ground level

Fig. 3. GSI spring element embedded in a track slab

A low tuned MSS with a vertical system natural frequency below 12 Hz is often described as a “heavy” MSS. “Heavy” refers to the large mass that is necessary to achieve a high, static deflection of the elastic support layer. However, selection of proper materials with high elasticity can aid in the design of a low tuned frequency system with significantly less mass. Train dynamics will strongly dictate the technical requirements and hence the above-mentioned reduction in slab mass. Requirements vary from train system to train system. Therefore, MSS have to be carefully designed to meet both isolation efficiency and also technical requirements due to train dynamics. Dimensions of elastically supported slabs are therefore specific to each project, considering local limitations and guidelines. Furthermore, special requirements are necessary in the design of a MSS for trams when the tracks are at street level and part of the street. This requires that the tracks support the load of motor vehicles and even heavy trucks. Especially the gap between slab and street has to be carefully considered. Here the choice of a slab mass heavier than usually required, eg. in a tunnel, will be advantageous. Because of the high amount of spring stiffness required, the dynamic vertical displacement of the system will be sufficiently limited to a few mm without sacrificing the low frequency tuning.

88

H.-G. Wagner and A. Herrmann

3 Helical Steel Springs in the MSS Helical steel springs have been used since the early 1990’s to provide the elastic layer in the MSS. Yielding a high level of isolation against structure borne noise and vibrations in the rail vicinity, they are used in the isolation of tracks in tunnels, at ground level, or in bridges and viaducts. Steel spring elements are even used as vibration isolation bearings for bridges. Furthermore, helical steel springs have been used as bearings in the permanent way of subways, trams, cargo and passenger trains and high speed trains. For example, parts of the Frankfurt International Airport Skytrain viaduct are isolated with steel spring devices in spite of the fact that the train is equipped with rubber tires. The viaduct is isolated in parts that are close to VIP lounges and waiting areas where high levels of structure borne noise control and vibration reduction is desired. The helical steel springs are contained in steel housings embedded in the slab. This so-called GSI-System (company designation) provides access from above. The GSISystem has proven very effective for the isolation of crossings and turnouts. The benefits of the system are: -

Spring elements can be placed between ties and rails so that the spring support system does not interfere with the operation of crossings and turnouts. Spring housings are embedded in the slab when it is poured. The springs are later placed comfortably from above and activated. The concrete slab can be poured on top of the foundation slab. The concrete slab is later lifted by a small jacking device. The system allows levelling of the slab within a few millimetres. No large-scale jacking mechanisms are required.

4 Isolated Turnout Slab in Heidelberg Two MSS were installed in turnouts 92 and 93 of the new single track tram system in the narrow streets of the German city of Heidelberg (Kirchheim suburb). The systems were installed in turnouts before and after the “Schwetzinger Straße-Odenwaldplatz” stop. The vibration isolation measure of the 25-meter long turnout slabs (Fig. 5) limits structure borne noise and vibration emissions to buildings in the nearby area (Fig. 7,8). The width of the slab varies from 2.8m to 5.2m.

Fig. 4. Arrangement of spring elements in a Heidelberg turnout slab

Floating Slab Track above Ground for Turnouts in Tram Lines

89

According to the project specifications, the unloaded system was to be designed with a theoretical vertical natural support frequency of 8 - 9 Hz. The isolation layer was to be provided by helical steel springs contained in steel housings embedded in the concrete slab. Dynamic and kinematic rail calculations were provided by the consultant IBU Ingenieurbuero Uderstädt + Partner [2] determining the maximum rail stress and the fitness of purpose for the spring elements . The analysis included prove of spring deflections in areas of transition to the adjacent systems due to train transit (10 t load per axle) to ensure a smooth change in stiffness, as well as braking, lateral impact, and temperature variations. Both MSS are identical, but are arranged in mirror-image to each other. Each turnout is supported by 59 spring-elements type GSIR20 (Kv=5.3kN/mm) and 12 spring-elements type GSI-R21 (Kv=6.6kN/mm). The spring-element layout is shown in Fig. 4. In addition, all horizontal loads are carried by the steel springs. Lateral restraints are typically not necessary. The system was installed October 2006, including the housing placement, mounting of spring inserts and the slab hoisting. The gap beneath the slab was set at about 40 mm. Hoisting and levelling of slabs was performed by 3 riggers in 3 working days (Fig. 5). Hoisting of the slab was started only one week after placement of concrete. The contractor handed over the turnouts to the owner “Heidelberger Straßen- und Bergbahn AG” in December 2006, who has been operating the turnouts since. The purchase cost for the GSI elements required for one turnout slab was €€ 35,000, including all installation including works.

Fig. 5. Floating slab under construction, being lifted, and in use

5 The New MSS in Basle/Switzerland The Basle Music Hall was edified in 1876 and is considered one of the finest ten Concert Halls in the world. However, the joy of a perfect concert experience was heavily infringed many years ago by structure borne noise and vibrations generated by nearby tram transit. The dramatic conditions even made some famous artists and orchestras to refuse to perform. Ten out of the eleven Basle tram lines run next to the Music Hall, approximately 60 times per hour during the evening. The T-crossing at Theaterstrasse and Steinenberg is a high source of sound and vibration emissions due to the arrangement of crossings, turnouts, and the tight curves. Even damage to the building structure has been attributed to the tram layout [1]. The planned improvement and renovation

90

H.-G. Wagner and A. Herrmann

Fig. 6. Plan of the Basel FST

works in the hall and the reconstruction of the adjacent City Casino provided an opportunity to deal with the unfortunate situation using a MSS. A generous donation of a local trust also contributed to the project. The main goal was to reduce the noise level inside the empty concert hall from about 46 dB(A) by at least 20 dB(A). A good deal of the registered noise in the empty Music Hall was in the 40-60 Hz range [1]. The low frequency range of the spectra is especially critical, since the distinct resonance frequencies of the hall are about 38-40 Hz. Therefore a system with a vertical natural support frequency of 5 Hz was selected for the application. This low tuning frequency could only be expected from “soft” helical steel springs supporting a stiff and massive concrete slab. The system, if planned and executed properly, guaranties the maximum mechanical vibration and structure borne noise isolation in the entire relevant spectra of 10-120 Hz and beyond. Additionally, a disturbing noise level in the frequency range from 200 - 2000 Hz could not be ruled out. Therefore a “light” MSS was “planted” in the main system with a designed vertical natural support frequency of 17 Hz. The light MSS consists mainly of a 30 cm thick concrete slab with lateral and base polyurethane mat.

Fig. 7. FST cross section

Consequently, the total height of the spring-supported concrete slab is 1.05 m resulting in an overall, spring supported mass of about 3,000 t. In spite of the low tuned

Floating Slab Track above Ground for Turnouts in Tram Lines

91

system with a 10 mm static spring deflection, the total stiffness of the resilient interface is quite high. Therefore there is only a small, almost imperceptible additional deflection of a few millimetres when tram and vehicle traffic runs over the slab. This small dynamic deflection is also a design advantage because it simplifies sealing of gaps.

Fig. 8. Installation of the elastic joint sealing during tram operation, detail [3]

To avoid an abrupt change in the stiffness, stiffer springs were placed in the transition areas between the spring-supported slab and the adjacent unisolated track. These special springs provide double the stiffness of “normal springs”. However, the high vertical stiffness allows for very limited dynamic horizontal deflections. Therefore, to account for thermal expansion displacements, here the spring system has been furnished with sliding contact bearings. The floating concrete slab consists of 3 main elements and 3 end elements at the transition zones. A total of 762 springs were used for the application. The springs were arranged outside the rail range and placed from above into the spring housings

Fig. 9. Lifting and adjustment of the track slab

Fig. 10. Spring units prepared for installation

92

H.-G. Wagner and A. Herrmann

once the concrete was cured. Then, the concrete slab, which was poured on site over a carpet-like surface, was lifted 50mm from the pit slab and levelled using a relatively simple tool especially manufactured with regard to the exceptional high housings. The hoisting process was performed in several stages to protect the concrete slab. A special highlight of the project was the fact that construction of the approximately 170 m long MSS was completed within the 6 weeks school summer break. This was required of the contractor because the tram circulation could only be completely halted during this time of the year. The entire scope of work completed in the 6 available weeks included: Demolition of the old rail system - Removal of old road surface/material - Excavation work Construction of base slab Installation of pre-cast pit walls - Construction of adjustment-foot directly below spring elements to compensate for street slope of up to 5% Fig. 11. Transmission curves (above: before and after; Placement of bond-breaking below: insertion loss) measured in front of the Basle layer - Installation of spring Concert Hall [2] casings - Installation of slab reinforcement and placement of concrete - Construction of the light MSS - Installation and commissioning of rail system. Finally, extensive but worthwhile procedure resulted in a noise and vibration level reduction of 22 dB(A) inside the concert hall. This world famous building has now, and after decades of exterior encroachments, been made available at its full value to the demanding cultural scene in Basle. The additional overall cost due to the “heavy” MSS totalled 3 Mio €€ .

6 Conclusion Within cities, tram lines are often located in close vicinity to sensitive buildings. Here, floating slab track technology has proved to be the most effective and reliable

Floating Slab Track above Ground for Turnouts in Tram Lines

93

solution to abate noise and vibration, especially, but not only, when radiated from turnouts and crossings. Helical steel springs used as the supporting element in a floating slab provide highest attenuation levels due to their low stiffness. The advantages are shown in the light of 2 projects having recently successfully been commissioned.

References [1] Bopp, U., Despotovic, D., Herrmann, A.: Erschütterungsschutz für den denkmalgeschützten Musiksaal von 1876 in Basel durch Sanierung der Straßenbahngleisanlagen. In: Railn.o.i.s.e 2007, Berlin (February 01–02, 2007) [2] Liesenfeld, B., Stummeyer, H.-J.: Masse-Feder-System, Steinenberg/Theaterstraße in Basel, Schweiz, Meßtechnische Untersuchung im Außenbereich des Theaters. In: Teil 2: Messung nach dem Umbau Ingenieurbüro Uderstädt + Partner, Essen (2006) [3] LeCo Lagertechnik AG, Schweiz

Vehicle/Track Impact Due to Passing the Transition between a Floating Slab and Ballasted Track Z.G. Li and T.X. Wu State Key Laboratory of Mechanical System and Vibration, Shanghai Jiao Tong University 800 Dong Chuan Road, Shanghai 200240, P.R. China Tel.: +86 21 34206332 Ext. 819; Fax: +86 21 34206006 [email protected]

Summary When a train runs over the transition between a floating slab track (FST) and ballasted track, wheel/rail impact forces arise because of the stiffness difference between the two kinds of track. A time-domain model is developed for the vehicle/track interaction and the characteristics of the impact are analyzed by numerical simulation using the model. Calculation results show that the wheel/rail impact load at the transition is moderate and its peak-to-peak value is only about 24 percent of the static load at a vehicle speed of 60 m/s.

1 Introduction The floating slab track (FST) is designed to reduce transmission of the track vibration to the infrastructure to reduce the ground-borne vibration [1-2] or re-radiated noise [3]. Generally the FST uses high compliance supports and heavy slabs to achieve low natural frequency, whereas at its ends the adjacent track usually is the ballasted or non-floating slab track with stiffer supports. When a train runs over the transition between the FST and ballasted (or non-floating slab) track, vehicle/track force occurs at the transition due to a sudden change in the track stiffness. As a result, large wheel/rail dynamic loads may arise. To study the impact due to passing the transition between the FST and ballasted track, the vehicle/track coupling model is needed. Two kinds of model are usually used to study the wheel/rail dynamic interaction, a moving irregularity between the stationary wheel and rail, and a wheel moving on the track [4]. In this study a time-domain model is developed for the track interacting with a moving vehicle to investigate the vehicle/track impact force due to the stiffness change at the transition of the FST and ballasted track. The vehicle/track impact at the transition is simulated. Influences of the vehicle speed, natural frequency of the FST, and length of the single slab are investigated.

2 Vehicle/Track Interaction Model Since the track is symmetric, only half a track (single rail) is considered here, as shown in Fig. 1. A long, simply supported Euler-Bernoulli beam is employed to represent the B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 94–100, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Vehicle/Track Impact Due to Passing the Transition

95

rail, with the bending stiffness ErIr and mass per unit length ρrAr. The slab is simplified as a free-free Euler-Bernoulli beam with length Ls, the bending stiffness EsIs and mass per unit length ρsAs. The rail pad is modeled as a discrete spring-damper and so are the slab bearing and the ballast. Rail pads and slab bearings are laid with spacing d and D respectively. For the ballasted track the sleeper is modeled as a lumped mass ms. A 1/8 vehicle model running at a constant speed is used to interact with the track via the Hertzian contact stiffness. The 1/8 vehicle model includes a wheel mass mw, bogie mass mb and car body mass mc. They are connected through the primary and secondary suspensions that are characterized by spring-damper k1, c1 and k2, c2 respectively. The infrastructure is treated as a rigid body and, only the vertical vehicle/track interaction and vibration are considered. mc

car body secondary suspension

mb

bogie primary suspension z

wheel

xr

xc k2, c2 v k 1, c 1

xb

mw

Hertz contact stiffness

CH

xw

ErIr, ρrAr

rail cp kp

pad slab 1 EsIs, ρsAs bearing

…

xsh

zsh

slab h

…

slab N

1

cb1 kb1

2

…

n

…m N s

xsn cb2 kb2

2

sleeper ballast

Fig. 1. Vehicle/track coupling model

The equations of motion of the rail and slab are described using modal coordinates, and thus the equation of motion of the vehicle/track system can be represented by a set of the ordinary differential equations, given in the form of a matrix:

&& + Cu& + Ku = F Mu

(1)

where M, C, K are the (modal) mass, damping, stiffness matrix respectively, u is the displacement vector composed of the modal displacements of the rail and slab and the physical displacements of the sleeper and vehicle, and F is the force vector composed of the wheel/rail contact force and the gravity of the wheel, bogie and car. The wheel/rail interaction force is calculated by

⎧ C [ x ( z, t ) − xr ( z , t )]3/ 2 , xw ( z , t ) − xr ( z , t ) > 0 f c (t ) = ⎨ H w , xw ( z , t ) − xr ( z, t ) ≤ 0 ⎩0

(2)

where xw and xr represent the wheel and rail displacement, respectively, and CH is the Hertzian constant, CH = 9.37×1010 N/m3/2. The fourth degree Runge-Kutta method is used to solve equation (1).

96

Z.G. Li and T.X. Wu

3 Simulation of Vehicle/Track Impact at the Transition The vehicle/track impact due to the stiffness change at the transition is simulated using the model. The main parameters are listed in Table 1 for half a track and an eighth vehicle. The rail used is UIC 60. The natural frequency of the floating slab system is 18 Hz. For the floating slab the softer bearing is preferable to gain better vibration isolation performance. However it leads to a larger difference in the stiffness between the FST and ballasted track. When a vehicle passes the transition, the stiffness difference between the two kinds of track may cause the vehicle/track impact. Fig. 2 shows the variation in the track stiffness around the transition between the FST and ballasted track. The static stiffness of the ballasted track is about three times as high as that of the FST, and the stiffness difference increases with reducing the single slab length. It is also found that the stiffness of the FST periodically varies with the single slab length, and this may cause parametric excitation when a vehicle runs over the FST. Table 1. Parameters used for calculation

Static Stiffness [MN/m]

Young’s modulus of slab Es, N/m2 Area moment of slab cross-section Is, m4 Slab density ρs, kg/m3 Cross-section area of slab As, m2 Mass of sleeper ms, kg Stiffness of rail pad kp, N/m Damping of rail pad cp, N·s/m Stiffness of slab bearing kb1, N/m Damping of slab bearing cb1, N·s/m

3.5×1010 1.44×10-3 2500 0.30 162 6.0×107 1.12×104 5.76×106 1.32×104

Stiffness of ballast kb2, N/m Damping of ballast cb2, N·s/m Mass of wheel mw, kg Mass of 1/4 bogie mb, kg Mass of 1/8 car body mc, kg Stiffness of primary suspension k1, N/m Damping of primary suspension c1, N·s/m Stiffness of secondary suspension k2, N/m Damping of secondary suspension c2, N·s/m

5.0×107 9.96×104 825 833 4000 1.18×106 4.00×104 2.00×105 2.50×104

80 60 floating slab track

ballasted track

40 20 45

50

55

60 65 Vehicle Position [m]

70 72

75

80

Fig. 2. Static stiffness around the transition of the floating slab track and ballasted track. Slab length: ––– 12 m; – – 6 m; · 3 m; ······ 1.8 m.

‐‐

Vehicle/Track Impact Due to Passing the Transition

97

Displacement [mm]

Fig. 3 shows the displacement of the wheel and rail at the contact point and the interaction force when the vehicle runs over the transition from the FST to ballasted track at a speed of 60 m/s. The displacement of the wheel and rail can be seen to decrease by about 2 mm because the track stiffness increases from the FST to the ballasted track. The wheel/rail impact occurs at the transition and the interaction force reaches its maximum. After a few oscillations the interaction force is attenuated to the static load level. The peak-to-peak impact load at the vehicle speed of 60m/s is estimated to be about 24 percent of the static load. Fig. 4 shows the results when the vehicle runs over the transition from the ballasted track to the FST. The displacement of the wheel and rail increases by about 2 mm as the track stiffness decreases from the ballasted track to the FST. A sudden drop in the wheel/rail interaction force can be observed from Fig. 4(b) due to the decrease of the track stiffness at the transition. Subsequently, the interaction force rises rapidly to its maximum. The peak-to-peak impact load is about 17 percent of the static load. The parametric excitation can also be observed from Figs. 3 and 4 caused by the track stiffness variation with the period of the slab length.

0.5 1 1.5

2.5 3 45 (a)

Interaction Force [kN]

ballasted track

floating slab track 2

50

55

60 65 Vehicle Position [m]

70 72

75

80

peak

60

floating slab track

ballasted track

55 50 45 (b)

trough 50

55

60 65 Vehicle Position [m]

70 72

75

80

Fig. 3. Wheel and rail displacement and interaction force when the vehicle runs over the transition from FST to ballasted track at speed 60 m/s, slab length Ls=1.8 m. · rail displacement; ······ wheel displacement.

‐‐

98

Z.G. Li and T.X. Wu

Displacement [mm]

0.5 1 1.5 2

ballasted track

2.5 3 80 (a)

Interaction Force [kN]

floating slab track

75

72 70

60

65 60 Vehicle Position [m]

peak ballasted track

55

50

45

50

45

floating slab track

55 50

trough

80 (b)

75

72 70

65 60 Vehicle Position [m]

55

Fig. 4. Wheel and rail displacement and interaction force when the vehicle runs over the transition from ballasted track to FST at speed 60 m/s, slab length Ls=1.8 m. · rail displacement; ······ wheel displacement.

‐‐

4 Parametric Study on Impact at the Transition Three factors affecting the vehicle/track impact at the transition between the floating slab and ballasted track are investigated through simulations. These factors are running speed of the vehicle, natural frequency of the floating slab system, and length of the single slab. The ratio of the wheel/track dynamic load to the static load is shown in Fig. 5 for the vehicle running over the transition at different speeds. The single slab length is 6 m. The upper curves correspond to the peak value of the dynamic load, while the lower curves correspond to the trough value. Both the impact load at the transition and the parametric excitation induced load increase with the vehicle speed. The wheel/rail impact force at the transition is larger for the vehicle running from the FST to ballasted track than that from the ballasted track to FST. The parametric excitation due to the periodic variation in the stiffness of the FST is smaller than the impact at the transition in terms of the vehicle/track dynamic load. In general, the impact at the transition is moderate and the peak-to-peak value of the impact load is only about 24 percent of the static load at a speed of 60 m/s for the vehicle running from the FST to ballasted track. Three kinds of FST are chosen with different natural frequencies designated of 18, 10, and 6 Hz, corresponding to the cross-sectional dimension of the slab (width/2×height) 1.25×0.24 m, 1.25×0.4 m and 1.40×0.5 m, respectively. The single

Vehicle/Track Impact Due to Passing the Transition

99

slab length is 6 m and the bearing stiffness of the slab is determined in accordance with the natural frequency designated and the slab mass.

Dynamic load ratio

1.15 1.1

peaks

1.05 1 0.95

troughs

0.9 0.85 10

15

20

25

30 35 40 45 Vehicle Speed [m/s]

50

55

60

Fig. 5. Ratio of the wheel/rail dynamic load to static load at different speed. ––– vehicle runs from the FST to ballasted track; ······ from the ballasted track to FST; · due to the parametric excitation from FST. Interaction Force [kN]

‐‐

60

55

45

ballasted track

floating slab track

50 50

55

60 65 Vehicle Position [m]

70 72

75

80

Fig. 6. Wheel/rail interaction force when the vehicle passes the transition at 40 m/s from FST to ballasted track. Natural frequency of FST: ––– 18 Hz; · 10 Hz; ······ 6 Hz. Interaction Force [kN]

‐‐

60 floating slab track

ballasted track

58 56 54 52 45

50

55

60 65 Vehicle Position [m]

70 72

75

80

Fig. 7. Wheel/rail interaction force when the vehicle passes the transition from FST to ballasted track at 40 m/s. Single slab length: ––– 12 m; · 6 m; ······ 3 m.

‐‐

100

Z.G. Li and T.X. Wu

It can be seen from Fig. 6 that the vehicle/track impact force at the transition increases when the natural frequency of the FST decreases. This is because the stiffness of the slab bearing is decreased in order to reach lower natural frequency, so that the track stiffness difference becomes large at the transition between the two kinds of track. As a result the vehicle/track impact load increases with decreasing natural frequency of the FST. Fig. 7 shows the wheel/rail interaction force when the vehicle runs over the transition between the FST with different single slab lengths and the ballasted track at a speed of 40 m/s. The impact load increases with decreasing single slab length, because a short slab leads to increasing stiffness difference at the transition, refer to Fig. 2.

5 Conclusions A time-domain model is developed to simulate the vehicle/track interaction at the transition between the FST and ballasted track. Due to the stiffness difference of the two kinds of track, vehicle/track impact force occurs when the vehicle passes the transition between the two tracks. The impact load is moderate and increases with the vehicle speed. The interaction force is larger when the vehicle passes the transition from the FST to the ballasted track, compared with the vehicle running from the ballasted track to the FST. The wheel and rail displacements decrease or increase by about 2 mm when the vehicle passes the transition due to the track stiffness change. This is not crucial as the 2 mm displacement variation arises gradually during a couple of metres passage of the wheel. The impact load increases with reducing natural frequency of the FST and, decreasing the single slab length leads to increasing vehicle/track impact load. The wheel/rail parametric excitation is due to the variation of the stiffness of the FST with the period of the single slab length. The vehicle/track load due to the parametric excitation increases with the vehicle speed but is smaller, compared with the impact load due to passing the transition.

Acknowledgements This study has been supported by the Science and Technology Development Foundation of Shanghai, under the project ‘Study on Vibration Isolation of the Floating Slab System for Reducing Structure-born Noise of Railway Transit’, Grant No. 042312011.

References [1] Nelson, J.T.: Recent developments in ground-borne noise and vibration control. Journal of Sound and Vibration 193, 367–376 (1996) [2] Lombaert, G., Degrande, G., Vanhauwere, B., Vandeborght, B., François, S.: The control of ground-borne vibration from railway traffic by means of continuous floating slabs. Journal of Sound and Vibration 297, 946–961 (2006) [3] Crockett, R., Pyke, J.R.: Viaduct design for minimization of direct and structure-radiated train noise. Journal of Sound and Vibration 231, 883–897 (2000) [4] Knothe, K., Grassie, S.L.: Modelling of railway track and vehicle/track interaction at high frequencies. Vehicle System Dynamics 22, 209–262 (1993)

Recent Developments in Operational Rail Noise and Vibration in NSW, Australia D. Anderson1 and C. Weber2 1

Rail Corporation NSW, Level 4, 18 Lee Street, Chippendale, 2008 Sydney NSW Australia [email protected] 2 Heggies Pty Ltd, Sydney NSW Australia [email protected]

Summary This paper provides a brief overview of the history of operational noise and vibration management on the NSW rail network and then gives a detailed description of a number of recent mitigation projects, programs and techniques. It covers mapping of rail noise on a number of urban lines for the prioritisation of mitigation options; detection of “noisy” wheel defects; and the application of source noise control. It also describes mitigation measures applied to new rail projects, including both ground-borne and airborne noise mitigation for tunnel and surface tracks. The mitigation techniques include both design measures (such as track support systems) and ongoing maintenance of both track and rolling stock during operation.

1 Introduction Noise and vibration emerged as important issues in NSW during the 1970s. The Eastern Suburbs Railway (ESR) opened in 1979 and included a form of floating slab track to control ground-borne noise in a nearby theatre. The tunnels were also retrofitted with sound absorptive panels shortly after opening, to reduce in-tunnel car interior noise for passengers and train staff. The 1980’s saw an increase in sensitive land development near the rail corridor, an increase in general rail noise complaints, and the emergence of issues with wheel squeal [1]. The class 81 locomotive required muffler redesign to reduce low frequency tonal noise, following complaints regarding airborne-noise induced vibration of lightweight homes. Before redesign, levels were approximately 100 dB (re 2 x 10-5Pa) at 40 to 50 Hz (depending on notch setting) at 15m distance. Following a technical review in 1995 [2], a number of significant steps were taken to manage noise more strategically, including a more scientific approach to investigating wheel squeal; a noise (and associated complaint) management procedure; a database of NSW rolling stock noise; and publication of rail noise guidance documents for the community and land developers [3,4]. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 101–107, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

102

D. Anderson and C. Weber

2 Examples of Noise Reduction Programs for Existing Operations 2.1 Sydney Harbour Bridge Residential locations near to the bridge approach spans were subject to rail noise levels of over 75 dBLAeq(24hour) and 90 dBLAmax,fast [6]. The opportunity was therefore taken to install resilient base-plates (Cologne Egg type, with nominal static stiffness of 9 kN/mm) during rail replacement carried out in 1994. Noise level reductions matched predictions [7], with 2dBA attributed to improved rail surface condition and 5dBA to the resilient base-plates (a reduction of 7dBA overall). 2.2 Hunter Valley Coal Line Approximately $10m was spent over 3 years from 1997, primarily on noise barriers, to address increased complaints about noise from freight rail operations in the Hunter Valley. Noise reductions of around 10 dBA were typically achieved, although some community members were still annoyed by low frequency noise and tonal squeal noise (from both brakes and curves). Top-of-rail friction modifiers were also applied at curve locations. Subsequent efforts have been made to address low frequency “booming” noise from empty coal wagons. Stiffening and constrained layer damping techniques have both been successfully applied, providing benefits of around 5 dB in the frequency range of concern and allowing compliance with a regulator requirement that the overall linear noise levels should not exceed the A-weighted values by more than 15 dB. 2.3 Development of Noise Pollution Reduction Programs on Sydney Metropolitan Lines A five year noise Pollution Reduction Program (Noise PRP) was established in 2001 for five priority lines in the Sydney metropolitan area [5]. These lines comprise 130km of rail corridor with approximately 67,000 dwellings or other sensitive receiver locations within 100m of the tracks. The methodology for the study is summarised in reference [8] and included the preparation of technical working papers covering source noise levels for general reference and for use in the noise modelling [9], noise mitigation measures that could be considered for inclusion in the PRPs for each line (including track measures, corridor measures, operational measures, rollingstock and buildings) [10] and a methodology for evaluating the Community Noise Burden (CNB) adjacent to the five priority lines [11]. The Community Noise Burden (CNB) was developed specifically for the PRP study to assist in ranking priority areas for noise mitigation along the five lines. It is based on the percentage of people “highly annoyed” by noise at various levels of exposure together with factors for tonal or impulsive characteristics (from sources such as turnouts, wagon bunching, curve and brake squeal, and stationary locomotives). The CNB was summed over a track length of 100 m account for the number of receivers in local community groupings. Each of the three components of the CNB formula (see below) correspond approximately to the percentage of people highly annoyed, however the overall CNBi is not directly comparable to any other rating.

Recent Developments in Operational Rail Noise and Vibration

103

CNBi = [(LAeq(24hour) - 53 + c)1.24 + (LAmax - 78 + c)1.24 + (LAeq,stationary - 43)1.24]/100 (1) where CNBi is the Community Noise Burden for an individual receiver, LAeq(24hour) is the equivalent continuous noise level due to all train operations over a typical 24 hour period, LAmax is the indicative maximum noise level predicted for normal train operations, LAeq,stationary is the LAeq noise level due to stationary locomotives where the cumulative period regularly exceeds 1 hour per day and c is an adjustment of up to 5 dB to account for the tonal or impulsive character of the noise. The CNB100 is the sum of the individual CNBi values for all receivers evaluated over a track length of 100 m and can be broken down to the portion of CNB100 on each side of the track. Computer noise modelling of five priority lines was undertaken with SoundPLAN software [12], which incorporated the ground topography, building locations and relevant train operations data. The presentation of the data included noise contour diagrams, point receiver calculations and plots of the CNB100 versus track location. Sample output data are provided in Fig. 1.

Fig. 1. Typical noise modelling output from PRP study showing CNB100 Calculations

The modelling and mapping confirmed the extent of locations where noise levels exceeded goals set by the regulator, and, for the first time, allowed a prioritised approach to noise reduction together with a broad indication of the noise contributions from key components of the source (rolling noise, locomotive noise, bridges, track joints etc). Following cost-benefit analysis, the implementation phase of the program concentrated on source noise controls and on improved planning control and acoustic design of adjoining land-use developments. Noise barrier options were also reviewed in detail at three “hot-spot” locations where source controls were not considered practical, but, in all three cases, the costs of the concept barrier designs proved excessive relative to the expected benefits. The source controls included rail grinding, the use of an improved gauge face lubricant, and the implementation of top-of-rail friction modifier at three additional sites. It was also determined that further effort should be spent in managing wheel roughness (see below). Updated land-use planning guidelines [3,4] were issued to all local councils in the Sydney metropolitan area, accompanied by a communication and awareness program. In conclusion, noise mapping was useful for prioritising mitigation treatments, but improved monitoring methods are necessary for properly quantifying the noise reduction results achieved in practice.

104

D. Anderson and C. Weber

"Good" Condition Suburban (tread brakes)

+5 dBA Tangara (disc brakes)

+10 dBA Intercity (tread brakes)

+15 dBA

LAmax Sound Pressure Level at 15m (dBA)

110

100

90

80

70

Benchmark Curve LAmax = 30xlog(speed/80) + 80

60 30

40

50

60

70

80

90

100

110

120

130

Train Speed (km/h)

Fig. 2. Measured LAmax,fast Noise Levels at 15 m

Noise monitoring confirmed that wheel defects (such as spalling and flats) were a significant cause of elevated rolling noise. A section of track was instrumented with eight accelerometers located on the rail flange between sleepers, a microphone located mid-track (between the rails), and a photoelectric sensor to identify each wheel of the passing train. The measurement results indicated a good correlation between the LAeq noise levels at 15 m from the track centreline, the mid-track LAeq noise levels, and the corresponding track-based RMS vibration levels. On this basis, it was concluded that the track-based vibration measurements provided a reliable indication of the 15 m noise levels during train passby events. Fig. 2 summarises LAmax,fast noise levels versus speed for three different train types from attended noise measurements at more than ten measurement sites across the Sydney metropolitan rail network between 1997 and 2002. A “benchmark” curve was established for train passbys without audible wheel defects, representing the expected LAmax,fast noise levels at 15 m for a train with freshly turned wheels (after initial bedding in). At 80 km/h, the LAmax,fast noise level for a “benchmark” train in NSW is 80 dBA at 15m. This compares to approximately 78 dBA at 15m for disc braked passenger rolling stock in the UK [2]. The study results indicated that trains with wheels in very poor condition generated LAmax,fast noise levels up to 20dBA higher than a “benchmark” train. Since the 2002 study, RailCorp has installed a number of Wheel Impact Load Detection (WILD) systems on the network. These include a low-pass filter on the force transducers (which removes much of the wheel/rail vibration energy relevant to wayside noise), because the systems are primarily for asset management (not for rail noise management). The result is a rather weak correlation between wheel noise level and WILD force outputs. However, ongoing operation of the WILD systems has led to a significant improvement in wheel condition, primarily through more proactive and

Recent Developments in Operational Rail Noise and Vibration

105

timely attention to defects. Recent measurements confirm that average noise and vibration levels have reduced by 2 to 3 dB and the proportion of trains with audible “noisy” defects has also reduced.

3 Noise and Vibration Mitigation for New Operations 3.1 Introduction Various noise and vibration design guidelines have been developed over recent years, culminating in the issue of a new guideline by the environmental regulator [13]. Previous guidelines required new rail projects to take all practical means to achieve 55dBLAeq(24hour) and 80dBLAmax,fast at the nearest residential facades. The new guideline takes a slightly different approach, requiring assessment of impacts, community consultation and consideration of mitigation options in cases where levels are predicted to: • •

increase above existing rail noise by more than 2dBLAeq(1hr) or 3dBLAmax,fast, and exceed 60dBLAeq(15hour DAY), 55dBLAeq(9hour NIGHT) or 80dBLAmax,fast

For the first time in Australia, the guideline also introduces criteria for groundborne noise. For residential premises, the night-time criterion is 35dBLAmax,slow (which is to be met by 95% of train movements). 3.2 Woollahra Cutting For historical reasons “artificial platforms” (or “platform profile barriers”) are used at this location to minimise airborne noise, lined with sound absorptive glass fibre material. By 1995 (after over 15 years’ service) this lining was showing signs of deterioration due to exposure to weather. Despite this, measurements at that time showed approximately 12 to 14 dBA noise reduction relative to standard slab track. Measurements were repeated in 2007 following removal of the sound absorptive lining. The noise reduction had reduced to approximately 7 to 10 dBA. 3.3 New Southern Railway (Sydney Airport Rail Link) The Airport Rail Link opened for service in 2000, prior to the Sydney Olympics. It includes 10 km of twin track rail tunnel, much of it passing near residential buildings. The tunnel is constructed in both hard rock and soft ground. Track support is conventional ballast with concrete sleepers. A ground-borne noise limit of 35 dBLAmax,fast was specified in the project consent conditions. Design studies [14] highlighted the influence of ballast quality on primary track response frequency and noted the influence of ground response on low frequency source vibration levels. Mitigation took the form of 3.6 km of tunnel treated with two grades of Phoenix ballast mat with dynamic stiffness values of 0.025 N/mm3 and 0.04 N/mm3 respectively. A further 2km of the tunnel required “marginal” ground-borne noise reduction of around 1 dBA. It was predicted that low stiffness rail pads (with a nominal dynamic stiffness of 100 kN/mm) would reduce the primary track response to approximately 50 Hz and thereby reduce ground-borne noise levels

106

D. Anderson and C. Weber

by 2dBA. In-tunnel vibration measurements were carried out on completion of the project and matched (or improved on) predicted levels [14]. 3.4 Epping to Chatswood Rail Line Many of the consent conditions for this new line [15] deal with operational noise and vibration. Of particular interest are the ground-borne noise criteria, which include requirements to meet 30 dBLAmax,fast for 50% of trains, 35 dBLAmax,fast for 95% of trains, and action to identify and manage trains that exceed 35 dBLAmax,fast (the noisiest 5%). Approximately 80 % of the alignment comprises slab track with Delkor “Egg” fasteners at 700mm spacing, while FST is used for the remaining 20 % of the line [16]. Monitoring near an existing underground line was carried out to review distribution of ground-borne noise levels (estimated from A-weighted vibration velocity and proposed track designs). The results indicate that the noisiest 5% of trains are between 10 and 15 dB noisier than the median train. In order to identify trains that exceed 35 dBLAmax,fast when the new line is operational, an automated system is proposed to monitor ground vibration levels from each passing train and to determine transfer functions during the commissioning stage to translate the results into equivalent groundborne noise levels at relevant receiver locations above the tunnels. Studies are currently underway to determine the costs, benefits and timescales for corrective action, should the automated system identify certain trains causing repeat events over 35dBLAmax,fast. Wheel maintenance typically accounts for up to 25% of total rolling stock maintenance costs, and overall wheel life can be significantly reduced by excessive turning or lathing. Fleet availability is also critical to maintaining service reliability, so the emphasis will be on preventative measures (such as wheel slide prevention) and proactive measures (such as trim or scrubber blocks for discbraked stock). A developer has completed several medium and hi-rise residential buildings above the line to the north of Chatswood, where the track configuration is slab track with 5 crossovers and 10 turnouts. A technical study identified two key issues. Firstly, the existing tracks were found to generate unusually high levels of low frequency vibration at 25 Hz and 31.5 Hz. This was attributed to poor ground conditions, but suggested that the use of very resilient track support for ground-borne noise control could be unwise due to amplification at these frequencies. Secondly, the turnouts and crossovers were likely to generate additional vibration, including at the frequency range of concern. It was determined that the optimum design for this location involved a combination of moderately resilient base-plates (20 kN/mm static stiffness); excavation and replacement of poor fill and formation material; and the use of swing-nose crossings. Vibration measurements were carried out at source and receiver locations at various stages in the construction to confirm that results met expectations.

4 Discussion and Conclusions The management of rail noise and vibration in NSW has become increasingly important over the last 30 years or so. At the same time, mitigation of noise and vibration has matured from a largely reactive approach to a more balanced and proactive one, including cost effective treatments at source. Many of the criteria and mitigation

Recent Developments in Operational Rail Noise and Vibration

107

techniques are consistent with those used overseas, but some examples are somewhat unique and may be of particular interest to other practitioners.

Acknowledgements The authors would like to acknowledge the substantial contribution made by Andrew Wearne to almost all of the work reported in this paper. Andrew died in a tragic accident in March 2006. As part of his legacy, he leaves rail noise and vibration in NSW greatly improved in terms of understanding and impact. He is sorely missed.

References [1] Anderson, D., Wheatley, N.: Mitigation of wheel squeal and flanging noise on the Australian rail network. In: Proceedings of the 9th International Workshop on Railway Noise, Munich, Germany (September 2007) (in press) [2] Hemsworth, B.: Assessment of Noise and Vibration on the Network of State Rail Authority of NSW. State Rail Commissioned Report (1995) [3] Rail Infrastructure Corporation and State Rail Authority, Interim Guideline for Applicants – Consideration of Rail Noise and Vibration in the Planning Process (November 2003) [4] Rail Infrastructure Corporation and State Rail Authority, Interim Guideline for Councils – Consideration of Rail Noise and Vibration in the Planning Process (November 2003) [5] Rail Services Australia, Prioritising Lines for Noise Management of the RAC Network (1999) [6] Richard Heggie Associates Report 4053R1, Sydney Harbour Bridge - Investigation of Railway Noise Control Options (1994) [7] Richard Heggie Associates Report 4053R2, Sydney Harbour Bridge Noise Control Works - Noise Reductions at Stage 1 and Stage 2 (1995) [8] Wearne, A., Weber, C.: Development of a Line Based Rail Noise Pollution Reduction Programme. In: Proceedings of the Australian Acoustical Society Conference, Brisbane, Australia (2004) [9] Richard Heggie Associates Report 10-1142R1 (rev 1), RAC Line-based Noise PRP Study - Noise Source Working Paper (2000) [10] Richard Heggie Associates Report 10-1142R2 (rev 1), RAC Line-based Noise PRP Study - Noise Mitigation Working Paper (2000) [11] Richard Heggie Associates Report 10-1142R3 (rev 1), RAC Line-based Noise PRP Study - Community Noise Burden Working Paper (2000) [12] Braunstein + Berndt, SoundPLAN English Users Manual (1996) [13] New South Wales Department of Environment & Climate Change, Interim Guideline for the Assessment of Noise from Rail Infrastructure Projects (2007) [14] Anderson, D., Harris, M.: New Southern Railway, Sydney: Noise and Vibration Attenuation Systems. In: Proceedings of Expo Rail (Asia), Hong Kong (2000) [15] NSW Government, Minister’s Conditions of Approval for Epping to Chatswood Rail Link [16] Transport Infrastructure Development Corporation, Epping to Chatswood Rail Link Track Design Factsheet, http://www.tidc.nsw.gov.au/ArticlePage.aspx?PageID=766

Experimental Validation of a Numerical Model for Subway Induced Vibrations S. Gupta, G. Degrande, and G. Lombaert Department of Civil Engineering, K.U. Leuven, Kasteelpark Arenberg 40, B-3001, Leuven, Belgium Tel.: +32 16 321677; Fax: +32 16 321988 [email protected]

Summary This paper presents the experimental validation of a numerical model for the prediction of subway induced vibrations. The model fully accounts for the dynamic interaction between the train, the track, the tunnel and the soil. The periodicity or invariance of the tunnel and the soil in the longitudinal direction is exploited using the Floquet transformation, which allows for an efficient formulation in the frequency-wavenumber domain. A general analytical formulation is used to compute the response of threedimensional invariant or periodic media that are excited by moving loads. The numerical model is validated by means of several experiments that have been performed at a site in Regent’s Park on the Bakerloo line of London Underground. Vibration measurements have been performed on the axle boxes of the train, on the rail, the tunnel invert and the tunnel wall, and in the free field, both at the surface and at a depth of 15 m. Prior to these vibration measurements, the dynamic soil characteristics and the track characteristics have been determined. The Bakerloo line tunnel of London Underground has been modelled using the coupled periodic FE-BE approach and free field vibrations due to the passage of a train have been predicted and compared to the measurements. The correspondence between the predicted and measured response in the tunnel and in the free field is reasonably good, given the large amount of uncertainties involved.

1 Introduction Ground-borne vibrations induced by underground railways are a major environmental concern in urban areas. These vibrations propagate through the tunnel and the surrounding soil into nearby buildings, causing annoyance to people. Vibrations are perceived directly or they are sensed indirectly as re-radiated noise. The frequency range of interest for subway induced vibrations is 1-80 Hz and for the re-radiated noise it is 30-200 Hz. To quantify these vibrations, great efforts have been made in recent years to develop the prediction models [1,2,3,4,5,6] that account for the three-dimensional dynamic tracktunnel-soil interaction. These advanced models take advantage of the invariance (or the periodicity) of the geometry along the tunnel axis using a Fourier or Floquet transformation. This paper concentrates on the coupled periodic finite element-boundary element (FE-BE) model [1,2] that was developed within the frame of the CONVURT project [7]. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 108–114, 2008. c Springer-Verlag Berlin Heidelberg 2008 springerlink.com 

Experimental Validation of a Numerical Model for Subway Induced Vibrations

109

Elaborate in situ vibration measurements have been performed at a site in Regent’s Park situated above the north- and south-bound Bakerloo line tunnels of London Underground [8]. These tunnels are deep-bored segmented tunnels with a cast iron lining and a single track, embedded in London clay at a depth of 28 m. The reference section is situated at kilometer post 46.306, which is 581 m west of Regent’s Park station or approximately 200 m east of Baker Street station. Vibration measurements have been performed during engineering hours at night for 35 passages of a test train in the north-bound Bakerloo line tunnel at a speed between 20 and 50 km/h. In addition, rail and wheel roughness have been measured, while the track characteristics have been determined by rail receptance measurements [9]. The dynamic soil characteristics have been determined by in situ tests (SCPT, SASW) and by laboratory testing [10]. The results of the vibration measurements are presently used to validate the coupled periodic FE-BE model.

2 The Numerical Method Within the frame of the CONVURT project [7], a coupled periodic FE-BE model has been developed that exploits the longitudinal invariance or periodicity of the tracktunnel-soil system [1,2]. The response to the moving loads in the periodic domains is given by Chebli et al. [11]. The response to moving loads is deduced from the transfer function in the frequency-wavenumber domain and the frequency content of the axle loads. 2.1 Modelling of the Track-Tunnel-Soil System The transfer functions are computed using the classical domain decomposition approach based on the finite element method for the tunnel and the boundary element method for the soil [1,2]. The Floquet transform is used to exploit the periodicity of geometry and to restrict the problem domain to a single bounded reference cell. The track-tunnel-soil interaction problem is solved in the frequency-wavenumber domain and the wave field radiated into the soil is computed. Reader is referred to complimentary literature [1,2] for more details on the coupled periodic FE-BE model. The Bakerloo line tunnel of London Underground is a deep bored tunnel with a cast iron lining and a single track, embedded in London clay at a depth of 28 m. The tunnel has an internal radius of 1.83 m and a wall thickness of 0.022 m. There are six longitudinal stiffeners and one circumferential stiffener at an interval of 0.508 m, resulting in a periodic structure. The tunnel’s reference cell is modelled using finite element method, where shell elements have been used for the cast iron lining, while the longitudinal and circumferential stiffeners are modelled using beam elements. The concrete on the tunnel invert has been modelled using 8-node brick elements with incompatible bending modes. Dynamic soil characteristics have been determined by in situ and laboratory testing [10,12]. The testing revealed that the tunnel is embedded in a layered soil consisting of a shallow layer with a thickness of 5 m on top of a homogeneous half space consisting of London clay. The top layer has a shear wave velocity Cs = 275 m/s, a longitudinal wave velocity Cp = 1964 m/s, a density ρs = 1980 kg/m3 and a material damping ratio

110

S. Gupta, G. Degrande, and G. Lombaert

β s = 0.042. The underlying half space has a shear wave velocity Cs = 220 m/s, a longitudinal wave velocity Cp = 1571 m/s, a density ρs = 1980 kg/m3 and a material damping ratio β s = 0.039 [8]. The track is a non-ballasted concrete slab track with Bullhead rail supported on hard wooden sleepers nominally spaced at d = 0.95 m with cast iron chairs. Both ends of a sleeper are concreted into the invert and the space between the sleepers is filled with shingle. The rails have a mass per unit length ρr Ar = 47 kg/m and a bending stiffness Er Ir = 3.04 × 106 Nm2 . The rails are not supported by rail pads and the resilience is mainly provided by the timber sleepers, which have a varying stiffness depending on its moisture content. The track model consists of two infinite Euler beams representing the rails and the mass elements representing the sleepers. The mass of the sleepers is distributed in the longitudinal direction with a mass per unit length ms = Ms /d = 70 kg/m. As there are no rail pads, a stiff connection is assumed between the rails and the sleepers, while the sleepers are continuously supported on the tunnel invert with springs of vertical stiffness k¯s = ks /d = 100 MN/m2 . The continuous elastic support below the sleepers accounts for the resilience of the track. A high value of the damping c¯s = 15 × 104 Ns/m2 is assumed in the elastic support to suppress the resonance, as no resonance has been observed in the rail receptance measurements within 200 Hz. 2.2 The Train-Track Interaction Forces Ground vibrations generated by moving trains arise from the combination of various excitation mechanisms. For the experimental validation of the numerical model, two main excitation mechanisms are considered: the quasi-static excitation and the unevenness excitation. The test train employed for vibration measurements on the Bakerloo line consisted of seven cars: a driving motor car, a trailer car, two non-driving motor cars, two trailer cars and a driving motor car. The length of a motor car is 16.09 m, while the length of the trailer car is 15.98 m. The bogie and axle distances on all cars are 10.34 m and 1.91 m, respectively. The distance between the first and the last axle of the train is 108.33 m. The tare mass of a motor car is 15330 kg, while the bogie mass is 6690 kg and the mass of wheelset is 1210 kg. The tare mass of a trailer car is 10600 kg, while the bogie mass is 4170 kg and the mass of a wheelset is 950 kg. The quasi-static excitation occurs when the axles of the train pass over the track and can be modelled as constant forces moving along the track with the train speed v. The constant load is equal to the total weight of the train distributed on the axles. Roughness is the main generation source of vibrations from moving trains. Rail roughness has been measured on the site using M¨ uller BBM rail roughness measurement equipment (RM1200E) [8]. Unevenness of the rail with wavelength longer than 0.10 m could not be measured with the available equipment, thus restricting the analysis to frequencies above 130 Hz for a train speed of 47.6 km/h [8]. Due to this limitation of the roughness measurements, axle box vibrations are used to estimate the wheel-rail ˆ d (ω) [13]: interface forces g ˆ v (ω)ˆ C gd (ω) = −ˆ ua (ω)

(1)

Experimental Validation of a Numerical Model for Subway Induced Vibrations

111

Acceleration [m/s2]

100 0 −100 −200 52

54

56 Time [s]

(a)

58

60

RMS force [dB ref 1 N]

Displacement [dBref μ m]

80 200

60 40 20 0 −20 0 10

1

2

10 10 Frequency [Hz]

(b)

3

10

80 60 40 20 0 10

1

2

10 10 Frequency [Hz]

3

10

(c)

Fig. 1. (a) Time history and (b) one-third octave band spectrum of the axle box response for a period of time corresponding to the passage of a train length over the test section at a train speed of 47.6 km/h. (c) The one-third octave band spectrum of the contact force at the front axle of the train for a speed of 47.6 km/h.

ˆ v (ω) is the vehicle’s compliance and u ˆ a (ω) is the response of the axle of the where C train. For the computation of the dynamic forces, the train can be well represented with the vehicle’s unsprung mass [13]. In this case, the vehicle compliance matrix is equal to ˆ v (ω) = diag{−1/(Muω 2 )} of order 28. It should be mentioned the diagonal matrix C that the vehicle compliance is significantly influenced by the vehicle’s suspensions at low frequencies and therefore, the estimation of the forces at frequencies below 10 Hz will be inaccurate. Moreover, the effect of wheel-axle bending resonance, which occurs at higher frequencies has not been accounted for. Axle box vibrations have been measured on six axle boxes during the whole journey of the test train on the section between Regent’s Park and Baker Street stations [8]. Figure 1a shows the time history of the acceleration of axle box 1 for a period of time corresponding to the passage of a train over the test section at a speed of 47.6 km/h. The advantage of using equation (1) to estimate the interaction force is that the various excitation mechanisms such as the unevenness excitation, parametric excitation and excitation due to rail joints and wheel flats are intrinsically accounted for. Visual inspection of the track revealed that there were a number of rail joints in the vicinity of the reference section, which could generate significant impact forces at the wheel-rail interface. The peaks in the axle box response shown in figure 1a are clearly due to the passage of the axle over joints in the rail. Figure 1b shows the one-third octave band spectra of the axle box displacements. The high response of the axle boxes indicate that the rails are of very poor quality. This has also been confirmed by the observation on the site that the rails were indeed heavily corrugated. Figure 1c shows the contact force at the front axle of the train for a train speed of 47.6 km/h. The frequency content of these forces exhibits a clear maximum near the train-track resonance frequency between 50 and 60 Hz.

3 Response during the Passage of a Train in the Bakerloo Line Tunnel The response can be calculated by adding the contribution of the dynamic forces and the quasi-static forces in the frequency domain. In the following, the running RMS and

112

S. Gupta, G. Degrande, and G. Lombaert

5

m/s]

0 Time [s]

120 A6 100

80

60 −5

0 Time [s]

A1 120

100

80 2 4 8 16 31.5 63 125 250 One−third octave band center frequency [Hz]

120 A6

−8

RMS velocity [dB ref 10

(b)

RMS velocity [dB ref 10

100

5

RMS velocity [dB ref 10

RMS velocity [dB ref 10

120

80 −5

140

−8

A1

−8

m/s]

(a)

m/s]

140

−8

m/s]

the one-third octave band RMS spectra of the measured and predicted vertical vibration velocity are compared. The running RMS has been calculated with an averaging time of one second according to ISO 2631-2.

100

80

60 2 4 8 16 31.5 63 125 250 One−third octave band center frequency [Hz]

Fig. 2. The experimental (gray line) and computed (black line) running RMS and one-third octave band RMS spectra of the vertical velocity on (a) the rail (A1) and (b) the tunnel invert (A6) during the passage of a test train at a speeds 47.6 km/h

Firstly, the response in the tunnel is compared to the experimental results on the rail and the tunnel invert [9]. Figure 2 compares the predicted and measured vertical velocity on the rail (A1) and tunnel invert (A6) during the passage of the test train at a speed of 47.6 km/h. On the rail, the contribution of the quasi-static forces and the dynamic forces can be distinguished. The quasi-static forces are important at low frequencies below 10 Hz. A peak corresponding to the axle passage frequency of fa = v/La = 6.92 Hz (La = 1.91 m) is observed in the predicted as well as the measured spectra. The vibration levels are maximum on the rail and decrease on the tunnel invert. The contribution of the quasi-static forces has diminished on the tunnel invert, and the dominant frequency content is situated in the frequency range above 30 Hz, where the dynamic forces are significant. The response at low frequencies below 12 Hz is underestimated by the numerical model. This could be due to the underestimation of the dynamic forces at these low frequencies. Also the influence of the train suspensions is significant at low frequencies, which has been disregarded in the model. The running RMS values show that the prediction accuracy on the rail and the tunnel invert is within 10 dB. Free field vibration measurements have been performed in Regent’s Park above the Bakerloo line tunnels. The vibrations measurements have been performed on the surface as well as at a depth of 15 m, where tri-axial accelerometers have been installed in a seismic cone [10]. In this paper, the computed vertical response is compared to the measurements on the surface (FF02z) and at a depth of 15 m (FF03z), at a distance of 23.5 m from the tunnel.

−2.5

0 Time [s]

2.5

90 FF03z 70

50

30 −5

−2.5

0 Time [s]

2.5

60

40

20 2 4 8 16 31.5 63 125 250 One−third octave band center frequency [Hz]

5

m/s]

30 −5

FF02z

80 FF03z

−8

RMS velocity [dB ref 10

(b)

RMS velocity [dB ref 10

50

5

RMS velocity [dB ref 10

RMS velocity [dB ref 10

70

113

80

−8

FF02z

−8

m/s]

(a)

m/s]

90

−8

m/s]

Experimental Validation of a Numerical Model for Subway Induced Vibrations

60

40

20 2 4 8 16 31.5 63 125 250 One−third octave band center frequency [Hz]

Fig. 3. The experimental (gray line) and computed (black line) running RMS and one-third octave band RMS spectra of the vertical velocity on the surface (FF02) and (b) at the depth (FF03) during the passage of a test train at a speed of 47.6 km/h

Figure 3 compares the experimental and computed running RMS and the one-third octave band spectra of the vertical free field vibration at points FF02 and FF03 during the passage of a test train at a speed of 47.6 km/h. Both the experimental and numerical results show that the dominant frequency content is around the wheel-track resonance frequency of about 50 Hz. A reasonably good agreement between the experimental and numerical results is observed for the response in the free field. The quasi-static contribution at low frequencies is not important in the free field and only the dynamic forces prevail. The vertical response at the surface (FF02z) has approximately the same magnitude as the vertical component at depth at the same location (FF03z). Furthermore, the amplitude decreases for increasing distance from the tunnel due to geometrical damping, while higher frequency components at increasing distance are significantly attenuated by material damping in the soil.

4 Conclusions In this paper, the experimental validation of a numerical model for the prediction of subway induced vibrations has been presented. An elaborate measurement campaign has been conducted at a site in Regent’s Park situated above the Bakerloo line tunnels of London Underground. In situ vibration measurements have been performed on the axle boxes of the test train, in the tunnel and in the free field. Apart from these measurements, other tests and measurements have also been performed to determine the soil properties and track characteristics. The coupled periodic FE-BE model fully accounts for the dynamic interaction between the train, the track, the tunnel and the soil. The response to moving loads (trains) is computed by first estimating the excitation forces and then solving the track-tunnelsoil interaction problem to compute the vibrations in the free field. The interaction force determined from the axle box vibrations accounts for various excitation mechanisms

114

S. Gupta, G. Degrande, and G. Lombaert

such as unevenness excitation, excitation due to rail joints and wheel flats as well as parametric excitation. The free field vibrations for the passage of a test train in the Bakerloo line tunnel have been predicted and validated. The correspondence between the predicted and experimental results is reasonably good, given the large number of modelling uncertainties. This paper demonstrates the applicability of the state-of-the-art 3D model in accurate prediction of vibrations from underground railways. The advanced model enables to investigate the inherent physics of ground vibrations, which is not possible with empirical methods or simplified deterministic models.

References [1] Clouteau, D., Arnst, M., Al-Hussaini, T.M., Degrande, G.: Freefield vibrations due to dynamic loading on a tunnel embedded in a stratified medium. Journal of Sound and Vibration 283(1–2), 173–199 (2005) [2] Degrande, G., Clouteau, D., Othman, R., Arnst, M., Chebli, H., Klein, R., Chatterjee, P., Janssens, B.: A numerical model for ground-borne vibrations from underground railway traffic based on a periodic finite element - boundary element formulation. Journal of Sound and Vibration 293(3–5), 645–666 (2006) [3] Forrest, J.A., Hunt, H.E.M.: A three-dimensional tunnel model for calculation of traininduced ground vibration. Journal of Sound and Vibration 294(4–5), 706–736 (2006) [4] Hussein, M.F.M.: Vibration from underground railways. PhD thesis, Department of Engineering, University of Cambridge (2004) [5] Sheng, X., Jones, C.J.C., Thompson, D.J.: Prediction of ground vibration from trains using the wavenumber finite and boundary element methods. Journal of Sound and Vibration 293, 575–586 (2006) [6] Andersen, L., Jones, C.J.C.: Coupled boundary and finite element analysis of vibration from railway tunnels-a comparison of two- and three-dimensional models. Journal of Sound and Vibration 293, 611–625 (2006) [7] (2003), http://www.convurt.com [8] Degrande, G., Schevenels, M., Chatterjee, P., Van de Velde, W., Hölscher, P., Hopman, V., Wang, A., Dadkah, N.: Vibrations due to a test train at variable speeds in a deep bored tunnel embedded in London clay. Journal of Sound and Vibration 293(3–5), 626–644 (2006) [9] Wang, A.: Track measurements on London Underground Bakerloo Line. Report 16487-2, Pandrol. CONVURT EC-Growth Project G3RD-CT-2000-00381 (May 2003) [10] Hölscher, P., Hopman, V.: Test site Regent’s Park London. Soil description. Report 381540104, Version 2, GeoDelft. CONVURT EC-Growth Project G3RD-CT-2000-00381 (December 2003) [11] Chebli, H., Ramzi, O., Clouteau, D.: Response of periodic structures due to moving loads. Comptes Rendus Mécanique 334, 347–352 (2006) [12] Pyl, L., Degrande, G.: Determination of the dynamic soil characteristics with the SASW method at Regent’s Park in London. Report BWM-2003-17, Department of Civil Engineering, K.U. Leuven. CONVURT EC-Growth Project G3RD-CT-2000-00381 (December 2003) [13] Lombaert, G., Degrande, G., Kogut, J., François, S.: The experimental validation of a numerical model for the prediction of railway induced vibrations. Journal of Sound and Vibration 297(3–5), 512–535 (2006)

A Numerical Model for Re-radiated Noise in Buildings from Underground Railways P. Fiala1,2, S. Gupta2, G. Degrande2, and F. Augusztinovicz1 1

Laboratory of Acoustics, Budapest University of Technology and Economics Magyar tudósok körútja 2. H-1117, Budapest, Hungary 2 Department of Civil Engineering, K.U. Leuven, Kasteelpark Arenberg 40, B-3001, Leuven, Belgium Tel.: +36 1 463 2543 [email protected]

Summary A numerical prediction model is developed to quantify vibrations and re-radiated noise due to underground railways. A coupled FE-BE model is used to compute the incident ground vibrations due to the passage of a train in the tunnel. This source model accounts for three-dimensional dynamic interaction between the track, tunnel and soil. The incident wave field is used to solve the dynamic soil-structure interaction problem on the receiver side and to determine the vibration levels along the essential structural elements of the building. The soil-structure interaction problem is solved by means of a 3D boundary element method for the soil coupled to a 3D finite element method for the structural part. An acoustic 3D spectral finite element method is used to predict the acoustic response. The Bakerloo line tunnel of London Underground has been modelled using the coupled periodic FE-BE approach. The free-field response and the reradiated noise in a portal frame office building is predicted.

1 Introduction Ground-borne vibrations induced by underground railways are a major environmental concern in urban areas. These vibrations propagate through the tunnel and the surrounding soil into nearby buildings, causing annoyance to people. Residents in buildings are affected both by vibrations of the structure (5-80 Hz) and through the re-radiated noise (20-200 Hz) from the walls and ceilings of the rooms. For the prediction of ground-borne vibrations and re-radiated noise in buildings, a modular architecture is adopted, which consists of the following subproblems: the dynamic vehicle-track-tunnel-soil interaction problem, the dynamic soil-structure interaction problem, and the prediction of re-radiated noise in the structures. In the first subproblem, discussed in details in reference [4], the free field vibrations are predicted, by computing the contact force generated by the wheel/track interaction and then solving the dynamic track-tunnel-soil interaction problem. The dynamic track-tunnel-soil interaction problem is tackled using the coupled periodic FE-BE model developed within the framework of CONVURT [1]. A finite element method is used to model a periodic unit of the tunnel, while a boundary element method is used to model the soil as a horizontally layered elastic half space [1]. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 115–121, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

116

P. Fiala et al.

Once the incident wave field in the soil has been determined, the dynamic response of a three-dimensional building due to this incident wave field is computed. Similarly to the tunnel-soil interaction, a subdomain formulation is employed where a finite element method is used for the structure and a boundary element formulation is used for the soil. This approach allows investigation of influence of dynamic soil-structure interaction on the structural response [2]. In the third subproblem, the computed structural displacements are used as a vibration input for the computation of ground-borne noise in the building's enclosures. A spectral finite element method is applied to the acoustic problem, which, for the case of low wall absorption, can lead to a direct integral representation of the internal sound pressure. To demonstrate the efficiency of the approach, the tunnel on the Bakerloo line of London Underground is modelled. The free field response is predicted in the frequency range 1-150 Hz, and subsequently the re-radiated noise in a hypothetic nearby multi-story portal frame office building is estimated.

(a)

(b)

Fig. 1. (a) Scheme of the numerical example. (b) Finite element model of the three-story portal frame office building.

2 The Incident Wave Field The thory of computing the incident wave field is described in reference [4]. Here, only a brief introduction is presented. The model of a moving vehicle on a track periodic or invariant in the longitudinal direction is used to compute the incident wave field. The periodicity of the tunnel and the soil in the longitudinal direction is exploited using the Floquet transform, limiting the discretization effort to a single bounded reference cell and formulate the problem in the frequency-wavenumber domain [3][1]. The vehicle is modelled as a set of concentrated masses representing the train's unsprung masses. The reference cell of the track and the tunnel is modelled by means of a finite element method. The tunnel is embedded in a horizontally layered soil. The geometrical and material characteristics of the soil, the tunnel and the track, as well as methodology for the determination of the incident wave field are decribed in [4] and [5].

A Numerical Model for Re-radiated Noise in Buildings

117

Fig. 2 shows the vertical component of the free-field incident velocity computed in the corner point of the building at coordinates {-5 m, -7.5 m, 0 m}. The vibration source is the rail roughness expressed as a stochastic process with a power spectral density decreasing with the longitudinal wave number. The train speed is 50 km/h. The dominating part of the frequency content is between 20 Hz and 80 Hz with a peak at 50 Hz, which corresponds to the wheel-track resonance frequency. The bogie passages are not clearly visible in the time history as the tunnel is situated at a considerable depth, however, due to the specific train composition, the observed velocity spectrum is quasi-discrete. The computed incident wave field has been validated by means of the experiments performed on the Bakerloo line [4] [6]. −5

4

−4

Velocity [m/s/Hz]

1

x 10

Velocity [m/s]

0.5

0

−0.5

−1 −10

−5

0 Time [s]

5

x 10

3

2

1

0 0 10

10

(a)

1

2

10 10 Frequency [Hz]

3

10

(b)

Fig. 2. (a) Time history and (b) frequency content of the vertical component of the incident wave field in the soil at coordinates {-5 m, -7.5 m, 0 m}

3 Dynamic Soil-Structure Interaction A weak coupling between the incident wave field and the structure is assumed, meaning that the presence of the building has no effect on the vibration generation mechanism and the free field displacements are applied as an excitation on the coupled structure-soil model. The decomposition method proposed by Aubry et al. and Clouteau [7] is used to formulate the dynamic soil-structure interaction problem. The structure is modelled in the frequency domain by a 3D structural finite element method. The equation of motion of the building is:

(K − ω M + Kˆ )uˆ = fˆ 2

g

(1)

ˆ, where the structural displacements are denoted by u

M and K denote the finite g ˆ element mass and stiffness matrices, and K stands for the frequency dependent stiffness matrix of the soil. The force fˆ and the stiffness of the soil are computed with a 3D boundary element model in the frequency domain, using the Green's functions of a layered half-space.

118

P. Fiala et al.

The portal frame office building, shown in Fig. 1b, has the dimensions 15x10x9.6m and is symmetrically placed on the free surface above the tunnel. The three story superstructure is supported by a 0.3 m thick reinforced concrete raft foundation. The basic structure consists of a reinforced concrete portal frame structure containing vertical columns of cross sectional dimensions 0.3x0.3 m and horizontal beams of dimensions 0.3x0.2 m. This frame structure supports 0.3 m thick horizontal slabs. The structure has a reinforced concrete central core which surrounds the stair-case. The thickness of the core walls is 0.15 m. The structural model is extended with the in-fill walls of three rooms besides the core. Room 1 has dimensions 5x6x3 m, and is located on the first floor, behind the core wall; room 2, which has the dimensions, is located on the second floor; a smaller room 3 with dimensions 5x4x3 m is located on the first floor, besides the core. The masonry in-fill walls are 0.06 m thick. The finite element size is chosen as 0.5 m, which is fine enough for computations up to 150 Hz. In the following, the structural response of the office building to the passage of the metro is presented. Fig. 3 displays the structural velocities at two points (Q1 and Q2) of the building. The point Q1 is located on the ground level, Q2 is located on the floor of room 1, both at horizontal coordinates x = -3 m, y = 0 m. The vibration levels at the point Q1 are very similar to the incident wave field, presented in Fig. 2. This indicates −5

6

−4

1

x 10

x 10

Velocity [m/s/Hz]

5 Velocity [m/s]

0.5

0

−0.5

4 3 2 1

−1 −10

−5

(a)

0 Time [s]

5

0 0 10

10

1

2

1

2

10 10 Frequency [Hz]

3

10

−5

6

−4

1

x 10

x 10

Velocity [m/s/Hz]

5 Velocity [m/s]

0.5

0

−0.5

−1 −10

(b)

4 3 2 1

−5

0 Time [s]

5

10

0 0 10

10 10 Frequency [Hz]

3

10

Fig. 3. Time history and frequency content of the vertical structural velocity in points (a) Q1 and (b) Q2

A Numerical Model for Re-radiated Noise in Buildings

119

that, in the present case, dynamic soil-structure interaction plays a negligible role in the vibration transmission between the soil and the building. A significant vibration amplification can be observed between the foundation and the first floor due to the first local bending modes of the floor slab in the frequency range 20-30 Hz. The ground vibrations above 70 Hz are not transmitted up to the first floor, which is an effect of structural damping.

4 Re-radiated Noise in the Structure After determining the structural response of the building, the acoustic radiation problem can be solved. As the impedance of the radiating walls is much larger than that of the internal acoustic space, a weak coupling between structural and acoustic vibrations is assumed: The acoustic pressure inside the room has no effect on the vibration of the walls and the computed structural vibration velocity is applied as a boundary condition in an acoustic boundary value problem. The internal acoustic space is characterized by the speed of sound Ca = 343 m/s and the density of the air

ρ a = 1.2 kg/m 3 . The absorbing surfaces of the rooms are charac-

terized by the acoustic impedance

Z a , relating the acoustic pressure pˆ a to the difference

of normal structural and acoustic velocities of the acoustic boundary. At relative low frequencies, the acoustic impedance can be computed from the walls' acoustic absorption coefficient α , which gives the ratio of the absorbed and the incident acoustic energy when a normal incident acoustic plane wave is reflected from the surface. An acoustic spectral finite element method [8] is used to express the internal presˆ a (x, ω ) in terms of the acoustic room modes: sure p

pˆ a (x, ω ) = ∑Ψ n (x)βˆ n (ω )

(2)

n

where Ψ n (x) denotes the n-th acoustic mode of the shoe-box shaped interior domain with rigid boundary conditions and

βˆn (ω )

is the corresponding modal coordinate.

The application of the spectral finite element method results in a system of linear equations for the acoustic modal coordinates:

(Λ + iωD − ω I )βˆ = iωFˆ 2

{ }

(3)

Λ = diag ωn2 contains the eigenfrequencies of the acoustic domain, I is a unit matrix, D is the modal damping matrix related to the wall absorption, and F where

denotes the modal load vector [2]. Two different absorption coefficients are considered, assumed to be constant on the rooms' surface and over the whole frequency range: α = 0.03 stands for a strongly reflecting room with uncovered concrete walls and an uncarpeted floor, while α = 0.15 is typical for an unfurnished, carpeted room. A modal base including all the acoustic modes up to 200 Hz has been used in the spectral finite element method.

120

P. Fiala et al. 80

Velocity [dB ref 2e−5 Pa]

Pressure [Pa]

0.2 0.1 0 −0.1 −0.2 −10

−5

0 Time [s]

5

10

70 60 50 40 30 20

16 20 25 31 40 50 63 80 100 125 One−third octave band center frequency [Hz]

(a)

(b)

Fig. 4. (a) Time history and (b) one-third octave band spectra of the sound pressure in room 1 during the passage of the train, with α = 0.03 (solid black) and α = 0.15 (gray dash-dotted)

Fig. 4 shows the pressure response in room 1 during the passage of the train for the two absorption coefficients. The dominant one-third octave bands are those containing the room's resonance frequencies at 28.6 Hz, 34.3 Hz, 57.2 Hz, 61.25Hz and 68.6Hz. Due to the frequency dependent sensitivity of the human ear, the apparent noise is determined by the 63 Hz peak. The one-third octave band spectra show a difference of 5 dB between the two wall absorptions above the first acoustic resonance of the room. As an application, the effect of base isolation of the building on the re-radiated noise is investigated. The base isolation is performed by placing springs between the foundation and the columns of the first floor. 9 springs of equal stiffness kz have been inserted below the structure's columns which are separated from the central core, and a distributed spring of total stiffness 3 kz has been inserted under the core. In the example, the total mass of the superstructure and the stiffness of the springs result in an isolation frequency of 10 Hz. Fig. 5 displays the re-radiated noise in Room 1 during the passage of the metro for the unisolated and the isolated buildings. The base isolation appears to be very 80 Velocity [dB ref 2e−5 Pa]

Pressure [Pa]

0.2 0.1 0 −0.1 −0.2 −10

−5

0 Time [s]

(a)

5

10

70 60 50 40 30 20

16 20 25 31 40 50 63 80 100 125 One−third octave band center frequency [Hz]

(b)

Fig. 5. (a) Time history and (b) one-third octave band levels of the pressure in room 1 for the unisolated building (solid black) and the base isolated building (gray dash-dotted), for the case of α = 0.03

A Numerical Model for Re-radiated Noise in Buildings

121

effective, as the noise level in the room is reduced by 15-20 dB in the higher frequency range (above the 63 Hz band), where the human ear is more sensitive to the noise.

5 Conclusions A 3D numerical model has been presented that is capable of predicting subway induced vibrations and re-radiated noise in buildings. The Bakerloo line tunnel of London Underground has been modelled using the coupled periodic FE-BE model and subsequently the structural and acoustic response in a hypothetic three-story portal frame office building has been predicted in the frequency range 1-150 Hz. The dominant frequencies of the traffic induced acoustic response are basically determined by the first acoustic resonances of the room. The effect of wall absorption on the sound pressure has been investigated, and above the first acoustic resonance, a difference of 5 dB has been found between typical wall absorptions for concrete and carpeted walls. It has been shown how base isolation affects the re-radiated noise in the room.

References [1] Degrande, G., Clouteau, D., Othman, R., Arnst, M., Chebli, H., Klein, R., Chatterjee, P., Janssens, B.: A numerical model for ground-borne vibrations from underground railway traffic based on a periodic finite element - boundary element formulation. Journal of Sound and Vibration 293(3–5), 645–666 (2006); In: Proceedings of the 8th International Workshop on Railway Noise, Buxton, U.K., September 8–11 (2004) [2] Fiala, P., Degrande, G., Augusztinovicz, F.: Numerical modelling of ground-borne noise and vibration in buildings due to surface rail traffic. Journal of Sound and Vibration 301, 718–738 (2007) [3] Clouteau, D., Arnst, M., Al-Hussaini, T.M., Degrande, G.: Free field vibrations due to dynamic loading on a tunnel embedded in a stratified medium. Journal of Sound and Vibration 283(1–2), 173–199 (2005) [4] Gupta, S., Degrande, G.: Experimental validation of a numerical model for subway induced vibrations. In: 9th International Workshop on Railway Noise, Munich, Germany (September 2007) (accepted for publication) [5] Gupta, S., Degrande, G., Chebli, H., Clouteau, D., Hussein, M.F.M., Hunt, H.: A coupled periodic fe-be model for ground-borne vibrations from underground railways. In: Mota Soares, C.A. (ed.) Proceedings of the 3th European Conference on Computational Mechanics, Lisbon, Portugal (June 2006) [6] Degrande, G., Schevenels, M., Chatterjee, P., Van de Velde, W., Hölscher, P., Hopman, V., Wang, A., Dadkah, N.: Vibrations due to a test train at variable speeds in a deep bored tunnel embedded in London clay. Journal of Sound and Vibration 293(3–5), 626–644 (2006); In: Proceedings of the 8th International Workshop on Railway Noise, Buxton, U.K., September 8–11 (2004) [7] Aubry, D., Clouteau, D.: A subdomain approach to dynamic soil-structure interaction. In: Davidovici, V., Clough, R.W. (eds.) Recent advances in Earthquake Engineering and Structural Dynamics, Ouest Editions/AFPS, Nantes, pp. 251–272 (1992) [8] Gustafsson, T., Pota, H.R., Vance, J., Rao, B.D.: Estimation of acoustical room transfer functions. In: Proceedings of the 39th IEEE Conference on Decision and Control, Sydney, Australia (December 2000)

The Influence of the Soil on Track Dynamics and Ground-Borne Vibration L. Auersch Federal Institute of Materials Research and Testing, Unter den Eichen 87, D-12200 Berlin, Germany Tel.: +49 30 8104 3290; Fax: +49 30 8104 1727 [email protected]

Summary The dynamic compliance of the railway track is the central part in the vehicle-track interaction and the excitation of ground vibration. The track-soil interaction is calculated by the integral transform method (ITM) for infinite track-beams on the soil or by the combined finite-element boundary-element method (FEBEM) for detailed 3dimensional track models. The finite element model of the track can include the ballast as continuum elements, what is necessary for the investigation of under ballast mats, or the ballast can be treated in a simpler way as an additional soil layer. Railway tracks on different homogeneous and layered soils are investigated. The influence of the layering of the soil and namely the influence of the ballast on the dynamic compliance of the track is analysed. The top soil layer (the ballast) is more important at high frequencies whereas the underlying (stiffer) soil is dominant at low frequencies. If the stiffness of the underlying soil is considerably greater than that of the top soil/ballast, the damping of the track-soil system is reduced and track resonances are visible. The compliances of the track-soil systems are introduced into the vehicle-track interaction analysis and the dynamic loads, which are acting on the soil and generating the ground vibration, are determined. Vehicle-track resonances (wheel-set mass on track-soil compliance) occur in the range of 40 to 120 Hz where

a

c

e

b

d

f

Fig. 1. Calculated and measured track-soil situations, a) homogeneous soil, b) layer on a rigid base, c) variation of the sub-soil, d) variation of the ballast/top soil, e) slab track, f) ballast-plate track, parameters: shear wave velocities vS, vS1, vS2, layer and plate thicknesses d B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 122–128, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

The Influence of the Soil on Track Dynamics and Ground-Borne Vibration

123

the strongest resonances at the lowest frequencies are highly influenced by a deep soil layering with a strong wave velocity contrast. Measurements of ground vibrations show some similarities for ballasted tracks in the ground vibration, but also significant differences for different types of tracks. The stiffness and damping of the soil has a strong influence on the amplitude and frequency content of the ground vibrations. In case of a very soft soil, the amplitudes of the track and the soil are increased considerably and the effect of a critically moving train can be studied.

1 Methods for Calculating the Wave Propagation and the Soil-Structure Interaction Track dynamics and ground-borne vibrations are related subjects which are analysed by the following methods: Integral transform methods, boundary element methods and finite element methods. The dynamic point-load solutions of layered soils are the base of ground vibration and boundary element studies. They are achieved by a Fourier integral in wave-number domain [1]. Similar one-dimensional Fourier integrals could also be used for circular load areas, but square or rectangular load areas need the calculation of two Fourier integrals. This 2-dimensional Fourier method is used to calculate (track) beams on the soil under stationary or moving loads [6]. Threedimensional track models can be calculated by the combination of the finite and boundary element method [5]. The point-load solutions are used to calculate a boundary stiffness matrix of the soil which is added to the FE stiffness matrix of the track/structure. Finally, the track-soil transfer functions are coupled with the multibody model of the vehicle [4], and the forces on the track or ground can be calculated which are generated by the irregularities of the train and track and which are the source of the ground vibration.

2 Track-Soil Dynamics and Vehicle-Track Interaction A 3D track model of 11 concrete sleepers and 2 UIC60 rails is analysed for different ballast and soil models (Fig. 1). In Figures 2 and 3, the calculated displacements of the rail (track compliances, left) and the forces on the soil (vehicle-track transfer functions, right) are shown. Figures 2a+b demonstrate the influence of the soil under the ballast which is very strong for the static and low-frequency region. At high frequencies, the differences are smaller and the resulting vehicle-track resonance frequencies (Fig. 2b) are nearly the same for the different soils. Figure 2c+d shows the results for a similar variation of the ballast stiffness. Although the compliances look similar (Fig. 2c), the vehicle-track resonance frequencies clearly increase with the stiffness of the ballast f0 ~ vS0.5 (Fig. 2d). Figures 3a+b show the effect of different thicknesses of the top soil layer. A soft top layer yields a layer resonance which is more pronounced and at lower frequencies for thicker layers. In Figure 3c+d, the stiffness contrast between the top layer and the

124

L. Auersch

underlying half-space is increased. This leads to clearer resonances in the displacements and forces. But generally, the layer resonances are more pronounced for the displacements and have minor effects on the force transfer functions which can be regarded as fluctuations around the homogeneous solution. More track-soil resonances are found in layered situations [2]: A layer on a rigid base (Fig. 1b, d = 0.5 m, vS = 100, 150, 200 m/s) yields resonance frequencies of f0 = 70, 105, 140 Hz which are proportional to f0 ~ vS but much lower than the layer resonance fL = vS/2d. If the rigid base is replaced by a concrete plate (Fig. 1f, d = 0.5 m), the soft layer vS = 100 m/s yields f0 = 65 Hz. If the rigid base is replaced by a stiffer half-space of vS = 300 m/s (Fig. 1c), the track resonance is at f0 = 55 Hz. In [7], a deeper layer (d = 1 m) with less contrast (vS = 150 over 200 m/s) resulted in a clear resonance at 40 Hz. All these results can be interpreted in the way that there is a track resonance in the frequency range of 20 - 80 Hz which is hidden by the strong radiation damping of the homogeneous soil and which is visible for the reduced damping of a layered situation.

3 Ground-Borne Vibration Calculations of the ground vibration [1] yield similar rules for layered situations. The low-frequency response is determined by the deep soil, whereas the high frequencies are ruled by the top soil layers. The amplification of the ground vibration due to the critical train speed was analysed in [6], showing that the influence increases for lower frequencies and is strongest for the static train load. Measurements at a very soft soil (vS = 30 m/s) with trains at the critical speed showed a completely different response compared to a normally stiff soil [1]. Due to the very soft soil, there are much higher amplitudes which appear at much lower frequencies. But no critical speed effects appeared as calculated for the static load on homogeneous soil, which is explained by the layering and the random variation of the soil [6]. Figure 4 shows results of normally stiff soils of vS = 100 - 300 m/s. The softer soil (150 m/s, Fig. 4a) has higher amplitudes than the stiffer soil in Figure 4b (vS = 225 m/s), but the frequency content is almost the same with two specific ranges at 8-16 Hz and around 50 Hz. The next two subfigures are from a gravel soil with even higher stiffness (vS = 275 - 300 m/s). They compare a standard ballast track to a ballast-plate track (Fig. 1f) where a plate is placed under the ballast to reduce the vibration. A clear reduction is found in the far-field at the typical frequency range between 10 and 16 Hz. The near-field of Figure 4d shows some irregularities due to a locally softer soil (dam material). Figures 4e+f compare the ground vibrations of a ballasted and a slab track (Fig. 1e). The frequency content of the ground vibration is very different. The low-frequency amplitudes of the slab track are reduced due to the better track quality, whereas the high-frequency amplitudes are increased due to the less damped vehicle-track resonance, which can be concluded from axle-box measurements [4, 5]. Because of the higher speed, the typical frequency ranges are shifted to 20 – 32 Hz (16 – 25 Hz) and 100 – 125 Hz (sleeper distance). Measurements at the same line but with a different ballast track (different rail-pad stiffness) and with a different type of slab track show the same results [3].

The Influence of the Soil on Track Dynamics and Ground-Borne Vibration

125

Fig. 2. Amplitude and phase of the track compliance (left) and the vehicle-track force transfer to the soil (right), variation of the sub-soil vS2 = 100..200..500 m/s (top) and the ballast vS1 = 100..200..300 m/s (d = 0.35 m, bottom)

126

L. Auersch

d

vS

vS

Fig. 3. Amplitude and phase of the track compliance (left) and the vehicle-track force transfer to the soil (right), variation of the top layer thickness d = … 2, { 1.5, U 1, © 0.7, ° 0.5, ‘ 0.35 m (top) and the sub-soil contrast 150 m/s to ‘ 150, … 170, { 200, U 300, © 500, ° 700 m/s (bottom)

The Influence of the Soil on Track Dynamics and Ground-Borne Vibration

127

Fig. 4. Measured train induced ground vibrations at r = … 3, { 5, U 10, ©20, ° 30, ‘ 50 m (6, 16, 32, 64 m, Fig. e and 4, 12, 24, 48, 96 m, Fig. f, after [8, 9]), soils with vS = 150, 225, 275, 300, 180, 200 m/s (Figs. a to f) and passenger trains with vT = 100 (Figs. a-d), 250 and 200 km/h (Figs. e+f)

128

L. Auersch

Acknowledgements Some of the research works were supported by German Railways (DB Systemtechnik, München), and the good cooperation is kindly acknowledged.

References [1] Auersch, L.: Wave propagation in layered soil: Theoretical solution in wavenumber domain and experimental results of hammer and railway traffic excitation. Journal of Sound and Vibration 173, 233–264 (1994) [2] Auersch, L.: Vehicle-track-interaction and soil dynamics. Vehicle System Dynamics 29, 553–558 (1998) [3] Auersch, L.: The influence of the track on railway induced ground vibration. In: Proc. 30. Deutsche Jahrestagung für Akustik (DAGA)/ 7. Congres Francais d’Acoustique (CFA), Strasbourg, pp. 1079–1080 (2004) [4] Auersch, L.: The excitation of ground vibration by rail traffic: Theory of vehicle-track-soil interaction and measurements on high-speed lines. Journal of Sound and Vibration 284, 103–132 (2005) [5] Auersch, L.: Dynamics of the railway track and the underlying soil: The boundary-element solution, theoretical results and their experimental verification. Vehicle System Dynamics 43, 671–695 (2005) [6] Auersch, L.: The moving-load effect of railway tracks on soft soil and isolation elements. Journal of Sound and Vibration (submitted) [7] Rücker, W., Auersch, L., Baeßler, M., et al.: A comparative study of results from numerical track-subsoil calculations. In: System Dynamics and Long-Term Behaviour of Railway Vehicles, Track and Subgrade, pp. 471–488. Springer, Berlin (2003) [8] Savidis, S., Bode, C., Bergmann, S., Schepers, W., Deift, S., Löwis, P.v., Schneider, S.: Analyse und Integration der Minderungsmaßnahmen. Schlussbericht zum BMBFVorhaben Nr. 19U0039C, TU Berlin (2006) [9] Staiger, Martens, Meunier, Heimberger, Freystadtl: Erschütterungsmessungen begleitend zum Projekt Noemie. Prüfbericht 12.1-PR-0086-03, DB Systemtechnik, München (2005)

A User-Friendly Prediction Tool for Railway Induced Ground Vibrations: Emission – Transmission – Immission W. Rücker and L. Auersch Federal Institute of Materials Research and Testing, Unter den Eichen 87, D-12200 Berlin, Germany Tel.: +49 30 8104 3290; Fax: +49 30 8104 1727 [email protected]

Summary The excitation and propagation of railway vibration has been thoroughly investigated in the German research project ‘Praxisgerechtes Prognoseverfahren für Schienenverkehrserschütterungen’. The aim of this project is to establish simple models and rules, which are theoretically founded and experimentally proven, and to put them in one comprehensive computer programme. The following work has been done in the three parts of the problem 1. Excitation by railway traffic Irregularities of the vehicle and track are introduced as the main sources of railway vibration. The compliance of the track on the soil is calculated and combined with the dynamic stiffness of a vehicle model. The vehicle-track interaction results in dynamic loads acting on the track or on the soil respectively. 2. Wave propagation through the soil The transfer functions of a layered soil are calculated for a point load in an appro-ximate manner. A superposition process leads to the transfer functions for a train load. 3. Transfer from the soil to the building For the last step, the reduction and amplification effects at the transfer from the soil to the building structure and elements are investigated by different numerical methods:

Fig. 1. Railway induced vibration, excitation by dynamic loads (emission 1), wave propagation through the soil (transmission 2), transfer from soil to building (immission 3), front page of the prediction tool B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 129–135, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

130

W. Rücker and L. Auersch

Conventional finite elements, finite elements combined with the thin-layer-method for the soil, and the combined finite and boundary element method for 3-dimensional structures on the soil. At the end, the floor amplitudes must be predicted by simplified rules which are also checked by measurements. The different components of prediction are combined in a comprehensive computer code with an easy graphical user interface. So a complete prediction of railway vibration is available for specialists and non-specialists.

1 Introduction Train induced vibrations propagate through the soil and excite neighbouring buildings (Fig. 1). The problem is divided into the parts emission, transmission and immission of vibration. The prediction of vibrations due to railway traffic has been investigated at BAM for a long time. In a research project sponsored by the German government (BMBF) and partners (German Railways, Hochtief and Porr), the results of the former research work as well as other published results have been condensed in a unique consistent prediction model [6]. The problem is analysed in detail with complex numerical methods and by extensive measurements. The prediction scheme does not work with these detailed models and information. The results have been simplified, smoothed and averaged, and simple models have been found, which are restricted to the most important parameters and physical rules.

2 Emission – Excitation by Railway Traffic Ground vibration near railway lines is caused by dynamic loads. These dynamic loads are the result of the interaction of the railway vehicle and the track. The whole system consisting of vehicle, track and soil is excited by irregularities. There are very different

Fig. 2. The excitation of railway vibration by the effective irregularities in track alignment …, rail roughness, out-of-roundness of the wheels {, sleeper passage at 50 Hz, train speed v = 100 km/h. Models of vehicle on ballasted track with sleeper and sleeper mat (a); Alternative models of a slab track (b) and a conventional track without vibration reduction measures (c).

A User-Friendly Prediction Tool for Railway Induced Ground Vibrations

131

irregularities of the vehicle, the track and the soil (Fig. 2). At low frequencies, there are the track irregularities due to imperfect track alignment. At high frequencies, there is the unevenness of the rail, (for very high frequencies, it is called the rail roughness). A specific excitation due to the track is the excitation with the sleeper-passage frequency. It is a regular parametric excitation as the stiffness of the track is varying on and between the sleepers [2]. The solution is approximated by introducing a special component of irregularity at the sleeper-passage frequency. The main irregularity of the vehicle is the out-of-roundness of the wheel. The outof-roundness has discrete Fourier components as it is a strictly periodical excitation. The first component, which is often the strongest component, is found at lower frequencies whereas components of higher order, which are slightly decreasing with frequency, are found at the same higher frequency range as the unevenness of the rail. When calculating the dynamic loads, the most important parameter of the railway vehicle is its unsprung mass which is specific for freight cars and powered units. The track usually has an influence on the excitation of railway vibration at high frequencies around and above the vehicle-track resonance. For special tracks such as tracks with ballast mats or mass-spring-systems, there is also an influence at lower frequencies [7]. The method of investigation of the vehicle-track interaction is the combined boundary and finite element method [1], but for the prediction scheme simpler models are used, multi-body models for the vehicle and beam-on-support models for the track [8].

3 Transmission – Wave Propagation through the Soil The wave propagation through the soil is the important link between the emission of the railway line and the immission into buildings. It provides the transfer from the excitation force P to the freefield vibration v0 at the place of the building. This transfer function of the soil can be very different in amplitude and frequency content for different soil situations [3]. The soil ampitudes v increase at least inversely with the shear stiffness G of the soil or v ~ vS-2 with vS the speed of the shear wave. As the stiffness of the soil can vary by decades from very soft to very stiff soil, corresponding strong differences in the vibration amplitudes occur. The frequency content of the soil vibration is influenced by the material damping of the soil. The material damping reduces high-frequency amplitudes, especially strong at longer distances from the source. This damping effect is stronger for soft soils (Fig. 3a). Therefore, softer soils have in general more low-frequency vibrations whereas stiffer soils have a more high-frequency content. A layering of the soil generates an even more pronounced individual frequency distribution for each soil (Fig. 3b). These effects of different soil situations are included in the prediction programme in an approximate way. The base of the prediction is the transfer function of a homogeneous half-space. In a layered situation, there is a general rule that deeper soil layers determine the low-frequency vibrations whereas the upper layers determine the high-frequency vibration. This holds for the amplitudes of the wave-field as well as for the frequency dependent wave velocity. The approximate solution for a layered

132

W. Rücker and L. Auersch

Frequency [Hz]

Frequency [Hz]

Fig. 3. Transfer function for a standard train excitation at at r = … 4, { 8, U 16, ©32, ° 64, a) homogeneous soil with vS = 100 m/s and D = 3 %, b) layered soil with vS = 100 m/s over 300 m/s, h = 3 m

soil makes use of this simple rule. The amplitudes are calculated as the amplitudes of a half-space, but the material properties of the half-space are different for different frequencies. In general, the properties of a deep layer are used at a low frequency. To get a proper choice of the parameters, the dispersion v(f) of a certain soil profile is calculated approximately [5]. For each frequency f, a homogeneous half-space with the corresponding v(f) is used to calculate the amplitudes of the wave-field of the soil. The transfer function between the excitation force P and the wave-field velocity amplitudes v(r) is the central concept of the prediction of the transmission. The calculated transfer function can be used if the excitation forces have been calculated within the emission module. A number of axle loads are superposed to yield the response of a train load. Compared to a single axle load, the amplitudes are increased, especially at the far-field.

4 Immission – Transfer from the Soil to the Building For the last step, the reduction and amplification effects at the transfer from the soil to the building structure and elements have been investigated by different numerical and experimental methods. Conventional finite elements have been used to study different types of buildings – low and high wall-type and column-type buildings [9]. Finite elements combined with the thin-layer-method for the soil [10] have been used to investigate the influence of the depth of the foundation and its type – single, strip or plate foundation [9]. The combined finite and boundary element method has been used to analyse the wave-field excitation of building structures on the soil [1]. A number of small 1-storey to 3-story residential buildings have been measured in and around Berlin. Some more complex buildings have been measured in detail. 60 measuring points have been installed to get an insight in the vibration modes of these buildings, an old 4-storey residential building in Berlin-Friedrichsfelde and a newly built 24-storey office tower in Vienna. The theoretical and experimental results are used to define the prediction model.

A User-Friendly Prediction Tool for Railway Induced Ground Vibrations

133

The immision of vibration in a building is calculated with a wall-floor model [4]. It consists of a number of wall and floor elements as well as a soil (spring+damper) element which are calculated by transfer matrices. By tuning the parameters and by averaging and smoothing procedures the model is fitted to the experimental and theoretical knowledge. The influence of the parameters is demonstrated in Figure 4. The starting point is a 6-storey concrete building on a medium soft soil (vS = 200 m/s). Figure 4a shows the influence of soft soils on the amplitudes of the building. The softer the soil is, the lower is the structure-soil resonance frequency. It is below 4 Hz for the softest soil of vS = 100 m/s. At higher frequencies, the amplitudes of the building are reduced compared to the exciting soil. The strongest reduction is for the softest soil. The influence of the foundation is expressed by the parameter FF/FB which is the ratio of the foundation area to the total area of the building. The standard building with strip foundation has the value FF/FB = 0.25, the smaller values are for single footings. The results in Figure 4b show different structure-soil resonance frequencies. wall

wall

Frequency [Hz]

Frequency [Hz]

Frequency [Hz]

Frequency [Hz]

Fig. 4. Soil-building transfer functions of a 6-storey concrete building, variation of a) the soil stiffness, b) the foundation (ratio of foundation to building area), c) the (average) mass density of the building, d) the resonance frequency of the floor

134

W. Rücker and L. Auersch

The lowest resonance frequency is at 4 Hz for the single footings, where the resonance amplitude is increased due to the reduced foundation area and foundation damping. The differences at higher frequencies are very strong. The single footings yield the strongest reduction lower than 0.1. Another important parameter of the building-soil interaction is the (average) mass density of the building ρ* (Fig. 4c). The standard concrete building has a value of ρ* = 300 kg/m3. More massive buildings have a lower building-soil resonance and a stronger amplitude reduction at a medium frequency range (up to 40 Hz). The higher frequency range shows no clear tendency. The variation is due to the elasticity of the walls which also prevents a stronger reduction of the building amplitudes. Some of the frequency dependent variations are shifted with the mass density of the building. The resonance frequency of the floors is a very important parameter for the prediction of railway induced vibration. The maximum amplitudes are generally found at the resonance frequency of the floor and its level can be very high if the frequency meets a frequency range of high excitation amplitudes. Figure 4d shows the influence of the floor resonance frequency on the soil-building transfer function. The amplification at the resonance frequency decreases with increasing resonance frequency. The resonance amplification is V = 5 for f0 = 12 Hz and continuously decreases to V = 2 for f0 = 32 Hz. The prediction code provides a tool for calculating the floor resonance frequency in dependence of material, dimensions and support conditions. As a standard, the simply supported plate is used. In addition to the emission, transmission and immission module, there are assessment modules for the vibration and the secondary noise. For situations where the vibration level exceeds the threshold value, reduction measures can be calculated in all three prediction modules.

5 Conclusion The research project “Praxisgerechtes Prognoseverfahren für Schienenverkehrserschütterungen” established a computer programme which combines all necessary modules for the prediction of railway induced vibration. Complete predictions are made possible, starting with the excitation by irregularities of the train and track, calculating the forces that act on the ground, determining the transfer through the soil and yielding the response of floors in buildings. The solution includes a chain of transfer functions for the emission, transmission and immission, but the parts vehicletrack-soil and soil-building-floor are solved with complete interaction. So the models used in the prediction programme are both, simple to use and theoretically founded. After a testing, calibrating and validating phase, when practicability and accuracy will be established, the software will be available for use by others.

References [1] Auersch, L., Schmid, G.: A simple boundary element formulation and its application to wavefield excited soil-structure interaction. Earthquake Engineering and Structural Dynamics 19, 931–947 (1990) [2] Auersch, L.: Parametric excitation of rail-wheel-system: calculation of vehicle-tracksubsoil-dynamics and experimental results of the high speed train Intercity Experimental. Arch. Applied Mechanics 60, 141–156 (1990)

A User-Friendly Prediction Tool for Railway Induced Ground Vibrations

135

[3] Auersch, L.: Wave propagation in layered soil: theoretical solution in wavenumber domain and experimental results of hammer and railway traffic excitation. Journal of Sound and Vibration 173, 233–264 (1994) [4] Auersch, L., Said, S., Schmid, W., Rücker, W.: Erschütterungen im Bauwesen: Messergebnisse an verschiedenen Gebäuden und eine einfache Berechnung von Fundament-, Wand- und Deckenschwingungen. Wand- und Deckenschwingungen 79, 185–192 (2004) [5] Auersch, L.: Simplified methods for wave propagation and soil-structure interaction: The dispersion of layered soil and the approximation of FEBEM results. In: Proc. 5th Eurodyn, Paris, pp. 1303–1309 (2005) [6] Auersch, L., Gerstberger, U., Meinhardt, C., Rücker, W.: Praxisgerechtes Prognoseverfahren für Schienenverkehrserschütterungen. Schlussbericht zum BMBF-Vorhaben Nr. 19U0039B, BAM, Berlin (2006) [7] Auersch, L.: Dynamic axle loads on tracks with and without ballast mats – numerical results of three-dimensional vehicle-track-soil models. Journal of Rail and Rapid Transit 220, 169–183 (2006) [8] Gerstberger, U.: Prediction of dynamic loads beneath tunnels due to railway traffic. In: Proc. 5th Eurodyn, Paris, pp. 2053–2058 (2005) [9] Meinhardt, C.: Relevante Einflussgrößen auf das Schwingungsverhalten von Gebäuden zur Berücksichtigung für ein praxisnahes Verfahren zur Prognose von Erschütterungsimmissionen. Dissertation TU Berlin (in preparation, 2007) [10] Rücker, W.: Ermittlung der Schwingungserregung beim Betrieb schienengebundener Fahrzeuge in Tunneln sowie Untersuchung des Einflusses einzelner Parameter auf die Auswirkung von Erschütterungen im Tunnel und dessen Umgebung. BAM-Bericht 64, Berlin (1980)

Using the PiP Model for Fast Calculation of Vibration from a Railway Tunnel in a Multi-layered Half-Space M.F.M. Hussein1, H.E.M. Hunt2, L. Rikse2,3, S. Gupta3, G. Degrande3, J.P. Talbot4, S. François3, and M. Schevenels3 1

School of Civil Engineering, University of Nottingham, University Park, Nottingham, NG7 2RD, UK Tel.: +44 1159 513904; Fax: +44 1159 513898 [email protected] 2 Engineering Department, Cambridge University, Trumpington Street, Cambridge, CB2 1PZ, UK Tel.: +44 1223 332730; Fax: +44 1223 332662 3 Department of Civil Engineering, K.U. Leuven, Kasteelpark Arenberg 40, B-3001 Leuven, Belgium Tel.: +32 16 321667; Fax: +32 16 321988 4 Atkins Consultants, Brunel House, RTC Business Park, London Road, Derby, DE1 2WS, UK Tel.: +44 1332 225617; Fax: +44 1332 225649

Summary This paper presents a new method for calculating vibration from underground railways buried in a multi-layered half-space. The method assumes that the tunnel’s near-field displacements are controlled by the dynamics of the tunnel and the layer that contains the tunnel, and not by layers further away. Therefore the displacements at the tunnel-soil interface can be calculated using a model of a tunnel embedded in a full space. The Pipe-in-Pipe (PiP) model is used for this purpose, where the tunnel wall and its surrounding ground are modelled as two concentric pipes using elastic continuum theory. The PiP model is computationally efficient on account of uniformity along and around the tunnel. The far-field displacement is calculated by using another computationally efficient model that calculates Green’s functions for a multi-layered half-space using the direct stiffness method. The model is based on the exact solution of Navier's equations for a horizontally layered half-space in the frequency-wavenumber domain. The results and computation time of the new method are compared with those of an alternative coupled Finite-Element-Boundary-Element (FE-BE) method that accounts for a tunnel in a multi-layered half-space. It is shown that the results of the two methods are in a good agreement for typical parameter values of a tunnel. The new method is computationally more efficient, i.e. requires significantly less running-time on a personal computer with much less use of memory.

1 Introduction A number of models have been presented in the literature for calculating vibration from underground railways. These models range from simple models (e.g. based on a few degrees-of-freedom or two-dimensional plane-strain models) to threedimensional comprehensive models. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 136–142, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Using the PiP Model for Fast Calculation of Vibration

137

The coupled FE-BE method accounts for a tunnel embedded in a multi-layered half-space. This method has significantly improved in the last few years in terms of its computational efficiency. This has been achieved by accounting for periodicity in the tunnel direction by incorporating the discrete wavenumber method [1] or by using the Floquet transformation [2,3]. Even with this development, the running time of the model is still long and it requires significant computational resources. The model is valuable for research purposes but still too computationally expensive to be used by engineers as a practical prediction tool. The PiP model is presented to account for a tunnel embedded in a full-space [410]. The model consists of two concentric pipes, both of infinite length. The inner pipe represents the tunnel wall and can be modelled using thin shell theory [4] or elastic continuum theory [7]. The outer pipe, with its outer radius being set to infinity, represents an infinite soil with a cylindrical cavity and is modelled using elastic continuum theory. The PiP model is computationally efficient on account of the uniformity along and around the tunnel. The model is validated against a coupled FE-BE model for the case of a tunnel embedded in a full-space [7]. A good agreement is achieved between results of the two models. A model based on the PiP model, Green’s functions for a full-space and Green’s functions for a half-space has been presented to calculate vibration from a tunnel embedded in a half-space [11]. The model assumes that the tunnel near-field vibration is not influenced by the existence of a free-surface. This allows calculating the displacements at the tunnel-soil interface using a model of a tunnel embedded in a fullspace. This model is further developed in [12] to improve the computations

x z

y

d

~ Fe i (ξx +ωt ) rt rc

Fig. 1. Layout of the model. Analysis is performed for a tunnel embedded in a multi-layered half-space. The figure shows a single layer on a half-space.

138

M.F.M. Hussein et al.

involved. A variant to the PiP model is employed to model a continuum, and Green’s functions in closed-form are derived in the wavenumber-frequency domain for a load on the free-surface. These functions are used to calculate vibration only on the freesurface by using Betti’s theory of reciprocity. This results in a significant improvement in the running-time of the model. In this paper, the work presented in [12] is extended to account for a tunnel embedded in a multi-layered half-space. This paper falls into three sections. Section 2 presents an outline of the model and describes the steps followed to calculate vibration in the far field. In Section 3, the main assumption used in these calculations is checked and results of the model are compared to those of an alternative coupled Finite-Element-Boundary-Element (FE-BE) method. Finally, Section 4 presents the conclusions of this work.

2 Description of the Model In this paper, the vertical displacement at any point in a multi-layered half-space due to a load applied on the tunnel invert is calculated. The load takes the form

~ F = Fe i (ξx +ωt ) , as shown in Fig. 1. The displacement takes the form u z = u~z e i (ξx +ωt ) and is measured along any line parallel to the x-axis and passing

through a point (y,z). These forms of force and displacement are the basis of the analysis in the wavenumber-frequency domain, which can be used to calculate the displacement for any sort of loading, such as a concentrated harmonic load, see [13] for more details. To calculate the far-field displacement from a tunnel embedded in a multi-layered half-space, three steps are followed. These steps are described in the following subsections. 2.1 Calculating Displacements at the Tunnel-Soil Interface The first step in the calculations is described in this section. The PiP model is employed to calculate the displacements at the tunnel-soil interface using a model of a tunnel embedded in a full-space. The PiP model consists of two concentric pipes. The inner pipe accounts for the tunnel wall with inner radius rt and outer radius rc . The outer pipe accounts for the surrounding soil with an inner radius

rc and outer radius

of infinite extent. In this paper, both the tunnel wall and the surrounding soil are modelled using the theory of elastic continuum in cylindrical coordinates. Note that the tunnel wall can be modelled as a thin cylindrical shell as in reference [4]. The theory of elastic continuum is adopted here as it gives more accuracy with no significant loss of computational efficiency. A summary of the elastic continuum equations is given in [4, 6, 12]. 2.2 Calculating the Internal Source in a Full-Space The PiP model is used in this step to model a full-space. The inner pipe represents a solid cylinder with radius ri < rc and the outer pipe is an infinite domain with a

Using the PiP Model for Fast Calculation of Vibration

139

ri and external radius of infinite extent. The objective of the second step is to calculate the stresses at r = ri that produce the same displacement at r = rc as calculated by the first step. Full details about this step cylindrical cavity with internal radius

will be published in [12]. 2.3 Calculating the Far-Field Displacements The Green’s functions for a multi-layered half-space are calculated following the direct stiffness method [14,15]. The dynamic equilibrium of the layered half-space is expressed in the frequency-radial wavenumber domain ( k r , ω ) as

~~ ~ Ku =p

(1)

~ where K is the stiffness matrix and it is assembled from element stiffness matrices ~ K e in a similar way to the Finite Element method. The element stiffness matrices ~ K e relate the displacements and tractions at the boundaries of a homogeneous layer ~ denote the displacements and external ~ and p or a half-space element. The vectors u tractions at the interfaces between elements. The Green's functions

uˆ ijG ( z ' , x, y, z , ω ) in the frequency-spatial domain are

subsequently obtained by means of an inverse Hankel transformation in the radial direction and an inverse Fourier series expansion in the circumferential direction (about the vertical axis x=0, y=0)[14]. Finally, the Green's functions

uˆ ijG ( z ' , x, y, z , ω )

are

transformed

to

the

frequency-wavenumber

domain

(k x , k y , ω ) . The Green’s functions are used along with the forces calculated from step 2 to calculate the displacements at any point in the ground. This is done in the same way as described in [12] taking advantage of symmetry around the tunnel centreline.

3 Results and Discussions In this section, the near-field and far-field displacements of a tunnel embedded in a layered half-space are calculated for a harmonic load applied at the invert of the tunnel, at (x=0 m, y=0 m, z=16.75 m), see Fig. 1 for information about the coordinate system. The displacements are calculated by the new model and then compared with those calculated by the coupled FE-BE model. Parameters of the tunnel and the ground are given in Tables 1 and 2 respectively. The distance between the tunnel centre and the free-surface is 14 m. The ground consists of a single layer on a half-space. The surface layer has properties of type 1 of soil with thickness of 6 m and the halfspace has properties of type 2 as presented in Table 2. The loss factors given in the tables are associated with both Lame’s constants. The displacements are calculated at 4 points: A (x=0 m, y=0 m, z=17 m); B (x=0 m, y=0 m, z=11 m); C (x=10 m, y=10 m, z=6 m); and D (x=10 m, y=10 m, z=0 m).

140

M.F.M. Hussein et al.

The frequency range of the results presented here is 1-80 Hz with a step of 1 Hz. Results of the new model are calculated using a personal computer (PC) with 2 GB RAM and 2.0 GHz processor. The FE-BE results are calculated by one processor of the high performance cluster (HPC) at K.U. Leuven. Higher frequencies are not attempted due to the high computational cost of the FE-BE model as a result of the finer mesh required at high frequencies. The current frequency range is the perceptible range for ground-borne vibration in buildings and it is sufficient for the sake of validation of the new model. For frequencies above 80 Hz, the new model is believed to produce more accurate results because the stiffness at the tunnel-soil interface is more dominated by the stiffness of the tunnel rather than the stiffness of the soil. The stiffness of the tunnel and the tunnel-soil interaction are well accounted for by the PiP model. Table 1. Tunnel Parameters External radius (m) 3

Thickness (m) 0.25

P-wave velocity S-wave velocity (m/s) (m/s) 5189 2774

Density (kg/m3) 2500

Loss factor 0.03

Table 2. Soil Parameters Type 1 2

P-wave velocity (m/s) 1964 1571

S-wave velocity (m/s) 275 220

Density (kg/m3) 1980 1980

Loss factor 0.08 0.08

Figs. 2.a and 2.b show the displacement at the tunnel-soil interface at the tunnel invert and the tunnel apex respectively. The results are calculated by the PiP model and the coupled FE-BE method. A good agreement is observed which confirms that the near-field vibration is not influenced by layers away from the tunnel. point A

point B −160

disp. [dB ref m/N]

disp. [dB

ref

m/N]

−160 −180 −200 −220 0

20 40 60 frequency [Hz] (a)

80

−180 −200 −220 0

20 40 60 frequency [Hz]

80

(b)

Fig. 2. The near-field displacements at (a) point A with coordinates (0, 0, 17) and (b) at point B with coordinates (0, 0, 11) as calculated by the PiP model (continuous) and the coupled FE-BE model (dotted)

Using the PiP Model for Fast Calculation of Vibration

point C

point D −200 disp. [dB ref m/N]

disp. [dB

ref

m/N]

−200 −220 −240 −260 0

141

20 40 60 frequency [Hz]

80

−220 −240 −260 0

(a)

20 40 60 frequency [Hz]

80

(b)

Fig. 3. The far-field displacements at (a) point C with coordinates (10, 10, 6) and (b) at point D with coordinates (10, 10, 0) as calculated by the new model (continuous) and the coupled FEBE model (dotted)

Figs. 3.a and 3.b show the far-field displacements at point C, at the interface of the two types of soil and point D at the free-surface. These results confirm that the new model calculates the far-field vibration for a tunnel embedded in a layered-ground with good accuracy. The running time for the new model to produce the results in Fig. 3.a is approximately 1.5 minutes on the personal computer described above. This involves all computations including those of Green’s functions. The same results are calculated using the coupled FE-BE method in approximately 17 hrs on the HPC. This demonstrates clearly the computational efficiency of the new method. The example above confirms that the near-field vibration is controlled by the dynamics of the tunnel and the soil layer surrounding the tunnel. Work is currently under development to study the effect on the near-field vibration of decreasing the depth of the tunnel and/or the distance between the tunnel and the nearest layer that does not contain the tunnel. This will help determine the limitations of the method. This method will be put forward for the new version of the PiP software [9].

4 Conclusions A new method for calculating vibration from a railway tunnel embedded in a multilayered half-space has been presented. The method is based on the PiP model where the near-field vibration is calculated using a model of a tunnel embedded in a fullspace. The far-field displacements are calculated using Green’s functions for a multilayered half-space. The new method is validated against the coupled FE-BE method for a tunnel embedded in a layered half-space. The new method shows good accuracy with a significant reduction in the running time and the computational requirements compared to the coupled FE-BE method.

142

M.F.M. Hussein et al.

References [1] Sheng, X., Jones, C.J.C., Thompson, D.J.: Prediction of ground vibration from trains using the wavenumber finite and boundary element methods. Journal of Sound and Vibration 293(3–5), 575–586 (2006) [2] Clouteau, D., Arnst, M., Al-Hussaini, T.M., Degrande, G.: Freefield vibrations due to dynamic loading on a tunnel embedded in a stratified medium. Journal of Sound and Vibration 283(1–2), 173–199 (2005) [3] Degrande, G., Clouteau, D., Othman, R., Arnst, M., Chebli, H., Klein, R., Chatterjee, P., Janssens, B.: A numerical model for ground-borne vibrations from underground railway traffic based on a periodic finite element–boundary element formulation. Journal of Sound and Vibration 293(3–5), 645–666 (2006) [4] Forrest, J.A., Hunt, H.E.M.: A three-dimensional model for calculation of train-induced ground vibration. Journal of Sound and Vibration 294(4–5), 678–705 (2006) [5] Forrest, J.A., Hunt, H.E.M.: Ground vibration generated by trains in underground tunnels. Journal of Sound and Vibration 294(4–5), 706–736 (2006) [6] Hussein, M.F.M., Hunt, H.E.M.: A numerical model for calculating vibration from a railway tunnel embedded in a full-space. Journal of Sound and Vibration 305(3), 401–431 (2007) [7] Gupta, S., Hussein, M.F.M., Degrande, G., Hunt, H.E.M., Clouteau, D.: A comparison of two numerical models for the prediction of vibrations from underground railway traffic. Soil Dynamics and Earthquake Engineering 27(7), 608–624 (2007) [8] Hussein, M.F.M., Hunt, H.E.M.: The PiP model, a software application for calculating vibration from underground railways. In: Proceeding of the fourteenth International Congress on Sound and Vibration (ICSV14), Cairns, Australia, July 9–12 (2007) [9] The PiP model (2007), http://www.pipmodel.com [10] Rikse, L., Hunt, H.E.M., Hussein, M.F.M., Degrande, G., Gupta, S.: A model for calculating vibration from a railway tunnel buried in a full-space including rigid bedrock. In: Proceeding of the fourteenth International Congress on Sound and Vibration (ICSV14), Cairns, Australia, July 9–12 (2007) [11] Hussein, M.F.M., Gupta, S., Hunt, H.E.M., Degrande, G., Talbot, J.P.: An efficient model for calculating vibration from a railway tunnel buried in a half-space. In: Proceeding of the thirteenth International Congress on Sound and Vibration (ICSV13), Vienna, Austria, July 2–6 (2006) [12] Hussein, M.F.M., Gupta, S., Hunt, H.E.M., Degrande, G., Talbot, J.P.: A computationally efficient model for calculating vibration from a railway tunnel buried in a half-space. International Journal for Numerical Methods in Engineering (submitted for publication) [13] Hussein, M.F.M., Hunt, H.E.M.: Modelling of floating-slab tracks with continuous slabs under oscillating-moving loads. Journal of Sound and Vibration 297(1–2), 37–54 (2006) [14] Kausel, E., Roesset, J.M.: Stiffness matrices for layered soils. Bulletin of the Seismological Society of America 71(6), 1743–1761 (1981) [15] Schevenels, M., Degrande, G.: ElastoDynamics Toolbox (EDT) for Matlab. Version 2.0. User’s manual BWM-2007-07, Department of Civil Engineering, K.U. Leuven (May 2007)

Structure-Borne Noise and Vibration Control for Chatswood Interchange J.T. Nelson1, M. Harrisson2, and M. Pettersson2 1

Wilson, Ihrig & Associates, 5776 Broadway, Oakland, California 94618, USA Tel.: 510-658-6719; Fax: 510-652-4441 [email protected] 2 Bassett Acoustics, Level 11, 44 Market Street, Sydney, NSW 2000, Australia

Summary The Chatswood Interchange project is a transit-oriented development combining multi-family residential and commercial development with a commuter rail station in Chatswood, NSW, Australia. The train station and garage structure would directly support 40-storey residential condominium towers and commercial spaces without benefit of structural breaks or building isolation. The vibration control provisions incorporated into the design include floating slab track, a massive and thick concrete deck, and large cross-section columns. Structure-borne noise and vibration levels were predicted with a finite element model, using an enforced relative displacement between wheel and rail. The predicted structure-borne vibration levels were compared with vibration levels measured at the top of the station structure during partial completion and found to be in reasonable agreement with prediction.

1 Introduction The Chatswood Interchange project is a joint development consisting of a commuter rail transit station, commercial development, and residential high-rise condominium tower in Chatswood, New South Wales. Two rail lines connect at the interchange: the existing North Shore Line running from Central Station in Sydney to Hornsby in the North, and the new Epping to Chatswood rail line, previously known as the Paramatta Rail Link. The Chatswood station would thus be a major interchange for commuter trains. The design effort included both basic analyses of mechanical impedances and detailed finite element analysis of the combined vehicle, track, track deck and support structure, using a wheel/rail relative roughness spectrum as the input. The project is still under construction as of this writing, with two out of four tracks completed.

2 Design Criteria The design criterion for the residential interior noise was 40dBA, measured with a precision sound level meter set to “fast meter response”. Structure-borne noise levels were required to be less than this level 95% of the time. The fast meter response corresponds to a root-mean-square sound pressure level over a time period of 0.125seconds. Considerable fluctuation of A-weighted sound levels may occur with B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 143–149, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

144

J.T. Nelson

the fast meter response, especially when controlled by low frequency vibration and structure radiated noise. A limit of 35dBA for the energy mean sound level measured during train passage was employed for analysis, assuming that the “fast” A-Weighted level would be 5dB higher.

3 Structural Design Figure 1 is an elevation of the station, parking garage, and building. The project design included commuter trains running on tracks supported on the concrete structure that would support the building. The track deck would be built into the support structures of Tower One and Tower Two, and thus there would be no benefit from isolation joints or other physical separation. The garage’s concrete floors were assumed to absorb vibration and damp support structure vibration. A number of structure-borne noise control provisions were included in the design, the most important of which was a floating slab track. Additional provisions included a thick, massive, high mechanical impedance train deck to support the floating slab, and large cross-section columns. These latter design features were limited in size by architectural and parking space requirements.

Fig. 1. Conceptual Cross-Section of Station, Parking, and Building Structures (COXDesignInc)

3.1 Track Isolation Design The track isolation is illustrated in Figure 2 and Figure 3. A normal weight concrete discontinuous “double tie” floating slab of approximately 2600Kg mass, natural rubber main support pads, side pads, and separation pads, was selected to isolate the track from the structure.

Structure-Borne Noise and Vibration Control for Chatswood Interchange

DF FASTENER AND PLINTH

ELASTOMER SEPARATION PAD

ELASTOMER SIDE PADS

145

DF FASTENER AND PLINTH

ELASTOMER SIDE PADS

MAIN SUPPORT ELASTOMER PADS

Fig. 2. Plan of the Pre-Cast Discontinuous Double-Tie Floating Slab

Fig. 3. Elevation of Pre-Cast Discontinuous Double-Tie Floating Slab and Deck

The design resonance frequency was 9 to 10Hz, based on the mass per unity length of the slabs and bogies, and stiffness per unit length of the isolators. The slab measured 1350mm longitudinal by 2500mm transverse to the track. The nominal thickness was 300mm, with a stepped up centre section to add bending stiffness and mass. This design supports four rail fasteners per slab, two for each rail, at 700mm pitch. The bending modes of the floating slab were calculated with a finite element model to be 223 Hz for normal weight concrete. A key feature of this double-tie slab crosssection is that the nodal points for the first bending mode would be at the approximate locations of the direct fixation fastener supporting the rail, minimizing excitation of the first bending. Further, locating the floating slab main support pads at the nodal points would further reduce the transmissibility of the slab at the first bending mode. The next bending mode would be at a much higher frequency, well above the typical frequency range of structure borne noise from track with resilient fasteners. The main support pads were spread further apart during final design to improve track stability. The slab’s first torsional mode of vibration was computed to be 277Hz, again assuming normal weight concrete. In this case, the rail fasteners and main support pads would not be at the nodal points. However, the dynamic forces of adjacent pads or fasteners would be of opposite phase and cancel, so that the mode would have less effect on insertion loss than the fundamental bending mode would. The foregoing suggests that the floating slab should act like a rigid mass up to frequencies in excess of 200 Hz. This is an important feature of the slab design, as bending resonances may adversely affect slab performance. 3.2 Deck Input Mechanical Impedance The train deck was designed to be as massive and as stiff as practicable to reduce its response to forces transmitted by the main support pads. The deck provides a high driving point mechanical impedance to react against the floating slab track isolation. The driving point impedance of an infinite plate is given by Cremer [1] in Eq (1) Z = 8 (B’ ρ h)1/2 ∼ 2.3cLρh2

(1)

B’ is the bending stiffness per unit width of the slab, ρ the density of the material, h the plate thickness, and cL the longitudinal wave propagation velocity. The mechanical impedance increases as the square of the plate thickness. Hence, increasing

146

J.T. Nelson

the thickness of the slab from, perhaps, 0.6m to 1.0m increases the input mechanical impedance by a factor of about 2.8. This substantial increase reduces the response of the slab by a factor of about 9 decibels, the effects of column impedance and span notwithstanding. The span between columns greatly influences the response of the slab, but, at the frequencies at which structure-borne noise is significant, namely in excess of 30 Hz, the infinite plate characteristics of the deck would control the deck response. The columns and boundaries of the slab introduce resonances in the deck that reduce the input mechanical impedance at discrete frequencies. 3.3 Column Stiffness The structure-borne noise in the condominium towers would be conducted primarily through the columns supporting the train deck, garage, and concourse floors. A reduction of the response of the columns to deck vibration would tend to reduce structureborne noise in the towers. The response of the columns to deck vibration would be inversely proportional to the column impedance, which would be proportional to the column’s cross-sectional area. Thus, increasing the diameter of the columns from perhaps 0.6m to 0.9m would increase the column impedance by a factor of about 2.25, potentially reducing the response of the columns by a factor of about 2.25, or 7dB. The actual reduction would depend on the relative impedances of the deck slab and column. However, the source mechanical impedance of the track deck would largely control the response of the columns. 3.4 Direct Fixation Fasteners The track resonance (sometimes called the P2 resonance) would degrade the vibration isolation performance of the floating slab at the resonance frequency. The track resonance frqeuency would be controlled by the direct fixation fastener stiffness, rail mass per unit length, and unsprung mass. Several fastener stiffnesses were investigated for this application, and a stiffness of about 11MN/m was selected to give a rail support modulus of 18MN/m2 and rail vertical resonance frequency of about 87Hz for 60Kg rail.

4 Finite Element Model The vibration transmission properties of the structure from the train way to the main transfer beams that would support the condominium tower were evaluated with the finite element model illustrated in Figure 4. This model incorporated parabolic brick elements to reduce model stiffness and more accurately characterize vibration transmission at high frequencies than would linear interpolation functions. The structure includes the train deck, the columns supporting the train deck, and the upper columns and transfer beams supporting the condominium tower. The tower structure was not included in the model due to the large number of degrees of freedom that would have been required, and because an estimate of vibration at the train deck

147

20

-6

1/12 OCTAVE ROUGHNESS LEVEL - dB RE 10 M

Structure-Borne Noise and Vibration Control for Chatswood Interchange

Fig. 4. Finite Element of Train Deck and Tower One Support Structure and 12Hz Structural Mode

10

0

-10

-20

-30 10

5

2.5

1.25

0.63 0.316 0.16

WAVELENGTH - CM

Fig. 5. One Twelfth Octave Roughness Spectrum

without the Tower One included would provide a conservative estimate of vibration response at the transfer beams. Addition of the tower structure would have added more mass to the model, and greater damping due to losses in the structure. Thus, the limited model was assumed to provide a conservative prediction of main support structure vibration. The columns were constrained against lateral horizontal response at the garage deck and concourse deck floors to simulate the constraint that these floors would apply to the columns. A mode of vibration of the transfer beam structure, supporting columns, and train deck beneath the structure, was identified at about 12Hz. The mode shape is also illustrated in Figure 4. The mode appears as a fundamental mode of the transfer beam structure, dish-panning in the vertical direction. The mode also involves the track deck. The added mass of the tower building structure would reduce the fundamental modal frequency of vibration, and transmission of vibration into the tower would further damp the response of the structure. The vehicle and track were incorporated into the model to compute the response to an enforced displacement between the rail and wheel. The bogies and car body were represented by lumped parameter models. Three bogies were assumed to be on the floating slab track, directly beneath the tower transfer beams. A relative displacement was enforced at one of the wheels of the centre bogie by applying apposed loads across a very stiff spring element connecting the top of rail and wheel centre. A 1/12 octave roughness spectrum was derived by subtracting 6 dB from the 1/3 octave band ISO 3095 octave band roughness spectrum for rail [2] and adding 10dB to account for imperfections in the rail and wheel. The 1/12 octave roughness spectrum, shown in Figure 5, was used in the model, assuming a train speed of 36KPH, or 10m/s. The responses at seven locations on the transfer beam structure were computed with a 1/12 octave resolution. The maximum velocity response of these locations was selected as the principal transfer function for estimating structure-borne noise and vibration. The resulting spectrum was summed to develop 1/3 octave band vibration velocity levels at the transfer beam structure. Finally, 11dB were added to account for contributions by the 12 wheels of three bogies.

148

J.T. Nelson

The floor vibration in the occupied spaces was computed by adding an additional 8dB to the predicted transfer beam 1/3 octave vibration velocity levels to account for floor resonance amplification, assuming a relatively weak floor with resonance frequency of about 4 Hz. Finally, the structure-borne noise in dB re 20 micro-Pascal was computed from the estimated floor vibration velocity levels by adding 34dB to the predicted floor vibration velocity levels in dB re 1 micron/sec. This is more conservative than assumed above in the calculation of structure-borne noise from measured train vibration, and is consistent with rooms with low acoustical absorption.

5 Predicted and Measured Structure Vibration

-6

50

40

30

20

10

0 OA

1/3 OCTAVE VELOCITY LEVEL - dB RE 10 M/SEC

-6

1/3 OCTAVE VELOCITY LEVEL - dB RE 10 M/SEC

Vibration velocity levels were measured with a seismic accelerometer at the columns and mid-span between columns of the station roof structure that would support the condominium tower. Data were collected for both northbound and southbound trains. The results are compared with the estimated transfer beam vibration velocity levels in Figure 6. The predicted and measured velocity spectra agree reasonably well with the maximum levels occurring at the 12.5Hz 1/3 octave band. The prediction is slightly less than measured at this frequency. At other frequencies, the measured levels were less than predicted. Both predicted and measured data indicate a knee in the spectrum at 40Hz, presumably due to the track resonance. The measured levels are very much less than predicted below 8Hz. Assuming an 8 dB floor resonance amplificaton, the A-Weighted noise in the first floor condominiums based on the measured data should be about 25dBA (slow meter response), well below the computed level of about 31dBA. These levels would correspond to about 30 and 36 dBA for a fast meter response, respectively. The estimated noise level based on measured roof vibration would thus be about 10 dB below the criterion.

4 8 16 31.5 63 125 250 OCTAVE BAND CENTER FREQUENCY - HZ ROOF SB

ROOF NB

500

ESTIMATED

Fig. 6. Comparison of Measured and Estimated Structure Vibration at Transfer Beams

50

40

PLATFORM-4-TEST PLATFORM-5_TEST ROOF-TEST

30

20

10

0 OA

4 8 16 31.5 63 125 250 OCTAVE BAND CENTER FREQUENCY - HZ

500

Fig. 7. Measured Structure Vibration - Energy Average of Northbound and Southbound Trains

Structure-Borne Noise and Vibration Control for Chatswood Interchange

149

Platform and station roof structure vibration velocity levels were measured for northbound and southbound trains. The results are compared in Figure 7. The data clearly indicate an amplification of about 15dB at about the 10 and 12.5Hz one-third octave bands, corresponding to the resonance frequency for the tower support structure vibration mode shown in Figure 4. The amplification of about 15dB of roof structure vibration relative to platform vibration suggests a damping factor of about 6%, less than assumed. Addition of the tower would add mass and damping to the support structure. Assuming an 8 dB floor resonance amplification factor, the condominium floor vibration would be at about 40dB re 1micron/second, at the typical limit of floor vibration velocity for living areas. The floor vibration level should be less if the floors are designed to avoid this resonance frequency.

6 Conclusion The foregoing measurements indicate that the design objectives were met. The structural resonance at 12Hz predicted by the finite element analysis was confirmed by measurement. The results suggest that the finite element method can be useful at high frequencies. While an FEM model might not capture the exact response of a complex structure such as the Chatswood Interchange, the results do indicate that such models are capable of reasonable predictions and modeling of vibration energy flow through the structure. The predictions are not necessarily less accurate than empirical estimates obtained from data measured in buildings of similar structural design. The reason that the FEM model is effective is that it accounts for the transmission of energy and losses in the structure in a realistic way.

References [1] Cremer, L., Heckle, M.: Structure-Borne Sound. Uncar, E.E. (translated), p. 264, 280. Springer, New York (1973) [2] Kalivoda, M.T.: Measurements of Railway Noise. In: Krylov, V., Telford, T. (eds.) Noise and Vibration from High Speed Trains, p. 155 (2001)

Measurements and Investigations at the Floating-TrackBed System in the North-South Tunnel in Berlin Thomas Jaquet1 and Rüdiger Garburg2 1

Ingenieurbüro Dr. Heiland, Bergstraße 174, D - 44807 Bochum, Germany Tel.: +49 234 950206 [email protected] http://www.baudynamik.de/ 2 Deutsche Bahn AG, DB Systemtechnik TZF 12, Caroline-Michaelis-Straße 5-11, D - 10115 Berlin, Germany Tel.: +49 30 297 57186 [email protected]

Summary The new north-south connection is the heart of the redesigned railway lines in Berlin. The line consisting of four to eight tracks runs in a tunnel system through the government district and the “Potsdamer Platz” with modern and very sensitive buildings. The immediate vicinity of these buildings required complex vibration-reduction measures in order to meet the very strict noise and vibration limits that had been contractually agreed with the owners of the neighbouring buildings. One of the major problems was that no experience had previously been gained with vibrations of running trains in interaction with one-shell-tubbing tunnels in such dimensions. Therefore, extensive experimental investigations were necessary to validate the first vibration prognosis and to optimise the vibration-reduction measures. Moreover, measurements for technical approval and quality assurance were carried out.

1 Introduction The centrepiece of the north-south connection is located in a 3.5 km tunnel in the centre of Berlin. The tunnel begins to the north of the main station, Berlin Hauptbahnhof, and ends south of the Landwehrkanal. In tandem with construction of the north-south connection, a lot of buildings are being planned and realised around the tunnel. Parts of the new buildings are located close to the new tunnel. The vibration level in the planned building depends on structure properties and the soil/building interaction. Up to the time of the noise and vibration proposals under the required environmental review process, these conditions were not clear because of the concurrent planning processes of the tunnel and the building development. Therefore it was not possible to meet the general standards (e.g. DIN 4150-2 [6]) for groundborne vibrations inside buildings. This fact led to an agreement between DB AG and the building investors to adopt emissions at the tunnel wall as a vibration criterion. The situation requires the technical potential to obtain a maximum of insertion loss. This means a very stiff tunnel in combination with a low tuned mass spring system (MSS), also often called a floating track bed system. For the purpose of ensuring B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 150–157, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Measurements and Investigations at the Floating-Track-Bed System

151

these high requirements a 3-step dynamic quality assurance system was adopted. The following sketch shows the basic procedure involved. Besides the principle of isolation efficiency of the MSS serviceability is another issue. The level of isolation efficiency is affected by the following properties: • • • • • • • •

Dynamic properties of the bearings Cross section and length of the slabs Design of the joints Arrangement of the bearing (single bearing or full surface system) Influence of lateral mat Change of the dynamic spring rate of the full surface layer due to absorption of water (perched water) Drainage concept of the slab Fatigue of the bearings

These properties have to be clarified to ensure isolation efficiency for a long period. The designs of the different MSSs are well described in [1]. Planning approval Definition of the areas with prevention system

STEP 1

STEP 2

STEP 3

After completion of the tunnel: Measuring of the transfer function and the mobility of the tunnel

Adjustment of the parameters of the MSS Serviceability of the MSS After completion: • Natural frequencies and checking the deflection • Vibration level at the tunnel wall

Dynamic calculation of the MSS and the bearings Checking bearings in the laboratory

Quality assurance during creation of the MSS

Fig. 1. Steps in the structural dynamic quality assurance procedure

2 Step 1: Dynamic Measurement to Check the Vibration Prognosis Before the requisite mass spring systems were finally designed the requirements for the prevention systems had to be checked. Measurements had to be carried out to clarify how vibration transmits to adjacent buildings. The basis of this investigation is an application of a vibration scanner. The vibration scanner used consists of 4 axle unbalance discs. The 4 axle constellation allowed generation of a train-quality load spectrum on the bottom plate of the tunnel. The advantage of this scanner is its ability to generate harmonic excitation with a changing rate of 0.25 Hz/s. The high energy

152

T. Jaquet and R. Garburg

density at each frequency is definitely identifiable in buildings. Disturbance caused by other sources (e.g. underground trains or road traffic) can be separated from the scanner signal. The investigation identifies the following dynamic properties: − − −

Mobility of the tunnel construction Tunnel-building transfer function Isolation from the soil.

2.1 Mobility Mobility M(f) is defined using the following formula:

M(f ) =

v( f ) F( f )

(1)

where v(f) is vibration spectrum, F(f) is force spectrum

2.OG-z,x,y A+T BT5 EG-z UG-z

1

3

Mobility Tunnel

Track 3 SO ~+17.13

2

Track 4 SO ~+17.09

Transfer function Tunnel / Building

Fig. 2. Left: vibration scanner, right: sketch of the transfer functions North-South Connection Berlin

Average of Mobility all values 1/3 octave spectra, Peak-Hold g, df=0,25Hz, Window = 4s, Hanning, Overlap=0

Mobility [m/s / N]

1,0E-06

One Shell Tubbing Tunnel

1,0E-07

Rectangular Tunnel (1m thick bottom)

1,0E-08

1,0E-09 10

100 frequency [Hz]

Fig. 3. Mobility in different tunnel types

1000

Transfer function Basement-Floor

Measurements and Investigations at the Floating-Track-Bed System

153

The degree of mobility provides the gauge for excitation of the tunnel construction. The investigation shows that mobility is heavily dependent on the tunnel design and the foundation in the soil. The one-shell tubbing tunnel (d=40 cm) shows a level of mobility 10 times higher than a cut-and-cover tunnel with a thick bottom. 2.2 Transfer Function to Adjacent Buildings The buildings around the “Potsdamer Platz” area feature basements with thick dimensions. This stiff construction provides high resistance to vibration. This leads to an attenuation of vibration and therefore to a convenient transfer function.

3 Step 2: Quality Management of the Bearings in the Laboratory There are currently no comprehensive design codes for the design and calculation of floating slabs (MSS). It was necessary, therefore, to coordinate means of proof for structural analyses and testing procedures with the Federal Railway Authority (EBA), the applicable supervisory and licensing authority, and the awarding body Deutsche Bahn. The following checks on and testing procedures for the bearing properties of MSS by an independent institute were agreed and are recommended for future projects: • • • •

Static stiffness (to DIN 45673-1 [7]) Dynamic stiffness (to DIN 45673-1 [7]) Load capacity Upper and lower load for dynamic testing (3.0 million cycles)

After passing this testing procedure the bearings are approved for installing on site. Table 1. Overview of dynamic properties of the MSS tested

Length of MSS Number of single bearings checked Full surface layer of bearings checked

15,428 km 198 approx. 20 sqm

Number of measuring sections Natural frequencies Measurement of deflection lines Vibration level

94 282 measuring points 876 passings 45 measuring points

4 Quality Management at Site Verification of the theoretical calculations was achieved by means of a measurement campaign. The first property approved was the natural frequency of the mass spring system. The natural frequencies of all floating slabs were checked at an interval of 200 m. Single bearing-supported slab track showed a good degree of agreement with the theoretical calculations. Even at the joints, where a higher required level of stiffness is

154

T. Jaquet and R. Garburg

installed locally, the natural frequency is only marginally higher than the theoretical one. This effect is due to the fact that the natural frequency roughly encompasses a wide slab length. This effect depends on bending length and the bedding factor. In some sections the natural frequencies measured showed a wider deviation from those calculated. Additional investigations showed that they were caused by exhaustive water absorption in the full surface layer. After partially dehydration of the gap beneath the slab with the full surface layer, the measured values were significantly improved. Table 2. Comparison of the design frequencies and natural frequencies measured

Single bearing system Single bearing system Single bearing system

Required frequency 7 Hz 8 Hz 9 Hz

Full surface layer

12 Hz

Full surface layer

15 Hz

Full surface layer

23 Hz

Type

Design frequency 7.1 Hz 8.1 Hz 8.3 Hz 11.5Hz 10.8 Hz 12.8Hz 11.74 Hz 16.7 Hz

Natural frequency measured 6.5 Hz – 7.5 Hz 8.25 Hz 8.5 Hz – 8-75 Hz Full surface layer partially dry: 12.25 – 12.75 Hz Full surface layer wet: 14 Hz – 17.0 Hz Full surface layer partially dry: 14.0 – 14.5 Hz Full surface layer wet: 13.7 Hz – 15.7 Hz Full surface layer wet: 21 Hz – 22.2 Hz

Additionally deflection measurements were carried out. The issue of this campaign was to check the static properties arrived at in structural and dynamic analysis of the mass spring system. The calculation of mass spring systems draws on the theory of a continuously elastic supported beam.

dy 4 b ⋅ q b ⋅ cst ⋅ y = − dx 4 E ⋅ I E⋅I

(2)

North South Connection Deflection of the slab track KM 05+77 track 4 2

410

1

deflection [mm]

133

125

0 -1 -2 -3

Load 1 Load 2 Load 4 theoretical deflection at Load 4 linear extrapolation on UIC 71

-4 -5 -6 -7 -8 -60

-40

-20

0 distance [m]

20

40

60

Fig. 4. Cross-section and the respective deflection curve (measurement/theory)

The deflection curve is calculated as follows:

y=

Q 4⋅ E ⋅ I sin ξ + cos ξ ⋅η where L = 4 and η = K ⋅ 2 ⋅ b ⋅ l ⋅ cst b ⋅ cst eξ

(3)

Measurements and Investigations at the Floating-Track-Bed System

155

and where: b is width of slab, E is Young Modulus of the slab track, I is inertia, cst is the bedding factor of the slab and Q is the load. The result will extrapolate to the theoretical UIC 71 load picture. The procedure was performed in the following way (fig. 5.): Top View

Detail

1)

Measuring deflection at 3-4 loads and simultaneously measuring velocity

2)

Presentation of the deflection curve for different loads.

3)

Calculating theoretical deflection (after Zimmermann) Parameters: stiffness, bedding factor.

4)

Calculation of the deflection curve with variation of the parameters: Option 1: linear extrapolation. Application of the bedding factor at the highest load level Option 2: nonlinear extrapolation. Application of the bedding factor at the UIC load. This can be obtained from diagrams of elastomeric bearings.

Mounting Position Sensor

Measurement Points

Deformation

Cross Section Additional Weights DB

Time

a)

b)

Comparison of the deflection line incl. safety factor phi with the permissible bending criterion d < L / 2000 Comparison: dynamic calculation with measurement

Fig. 5. Fundamental procedure for measuring deformation of the MSS

The comparison of the measurement with the theory shows good conformity, see fig. 4. Application of the Zimmerman method is confirmed. For straightforward systems the Zimmermann method is a good calculation tool. More difficult systems (e.g. wide slabs with different tracks on one common slab, slabs with turnouts etc.) have to be subjected to FEM. G4_KM_0668_3_L1_V1/K:6/S:1/G4_KM 0668_3_W3

1,0

Weg / mm

0,5 0,0 -0,5 -1,0 0

25

50

75

Zeit / s

Fig. 6. Detection of shorts between floating slab and tunnel

156

T. Jaquet and R. Garburg

Shortcuts can be well detected by this procedure. The deflection curve shows a branch if a shortcut occurs (fig. 6.). This is totally different to systems with undisturbed deflections.

5 Step 3 Vibration Isolation Efficiency Finally the efficiency of the MSS was measured during passage of two types of train: the ICE-T and a Rapid Light Rail RE. The sensors were installed at the tunnel wall. The vibration level in the tubbing tunnel was compared with and without the 7 Hz tuned MSS. Isolation efficiency (insertion loss) is shown in the next figure. Isolation starts at 10 Hz and increases up to 25-30 dB at 50 Hz. measured isolation efficiency of a 7 Hz MSS comparison to a slab track system (1/3 octave spectra) 40

isolation efficiency [dB]

30

20

10

0

ICE-T passing Rapid Light Rail passing

-10

-20 4

5

6,3

8

10 12,5 16

20

25 31,5 40

50

63

80 100 125 160 200 250 315

frequency [Hz]

Fig. 7. Isolation efficiency of the 7 Hz MSS

The vibration level complies with the stringent criteria at all locations.

6 Conclusion The new north-south connection includes 15 km of MSS. Because of the high performance required of these systems, continuous inspection of all properties was necessary. The three-step dynamic quality assurance system is a successful means of checking the properties of MSS on site and adapting the slab system to specific local characteristics.

Measurements and Investigations at the Floating-Track-Bed System

157

The requisite strict vibration levels, defined at the tunnel wall, are adhered to at all locations. The natural frequencies measured agree well with the design frequency for single bearing systems. Deviation is greater for MSS with full surface layer. Deflection calculation adopting the analytical Zimmermann method is well confirmed by measurements.

References [1] Jaquet, T., Heiland, D., Rutishauser, G., Garburg, R.: Nord-Süd-Verbindung in Berlin Eisenbahningenieur 57, vol. 9 (2006) [2] Jaquet, T., Heiland, D., Flöttmann, H.: Tiefabgestimmtes Masse-Feder-System bei der Flughafenanbindung Köln / Bonn, ETR Eisenbahntechnische Rundschau, 6 (2004) [3] Jaquet, T., Krüger, F., Repczuk, A.: Schwingungsminderung im Schienenverkehr, Schallund Erschütterungsschutz im Schienenverkehr. In: Krüger, Friedrich (eds.) Schall- und Erschütterungsschutz im Schienenverkehr, 2nd edn., ch. 7, Expert Verlag (2006) ISBN 38169-2494-8 [4] Heiland, D., Jaquet, T., Flöttmann, H.: Einbau eines leichten Masse-Feder-Systems unter Betrieb. In: Proceedings VDI conference, Kassel (May 2003) [5] Jaquet, T., et al.: Ausbildung eines tieffrequenten Masse-Feder-Systems mittels Stahlfederelementen bei U- und Vollbahnen als Schutz gegen Erschütterungen und Körperschalleinwirkungen. VDI Report No. 1345, pp. 143–160 (1997) [6] DIN 4150-2: Structural vibration - Human exposure to vibration in buildings [7] DIN 45673-1: Mechanical vibration - Resilient elements used in railway tracks - Part 1: Laboratory determination of static and dynamic characteristics

Propagation of Vibrations Due to a Tramway Line M. Maldonado1, O. Chiello2, and D. Le Houédec1 1

GeM Laboratory, Ecole centrale de Nantes, 1 rue de la Noë, 44321 Nantes Cedex 3, France Tel.: +33 (0)2 41 49 16 64; Fax: +33 (0)2 41 49 16 69 [email protected] 2 INRETS, 25 av. F. Mitterrand, 69675 Bron Cedex, France Tel.: +33 (0)4 72 14 24 05; Fax: +33 (0)4 72 37 68 37 [email protected]

Summary Tramway traffic may produce vibrations propagating in soil leading to vibration annoyance for people living or working in neighbouring buildings. Thus vibration is an important parameter to be considered when planning new lines and dynamic performance evaluation of tramway tracks is necessary to validate or modify the existing means that reduce vibrations. This paper presents experimental and theoretical investigations of vibrations caused by tramway passages in Nantes, France. It focuses on the control of ground-borne vibrations for the whole system, taking into account important elements such as the dynamic vehicle characteristics, the track and the soil behaviour. A complete track-soil-ground model is proposed to predict ground-borne vibrations, so as to estimate a trouble gauge concerning - for example - the impact of a future tramway line.

1 Introduction Nowadays, the tramway is more and more developed so as to reduce traffic jams in large cities. It offers a fast and alternative means of transport but nevertheless can produce new kinds of vibration (and sound) annoyance for people walking in the street or staying in buildings near the tramway line. The level of noise and vibration depends highly on the soil and track characteristics, and also on the vicinity of houses, buildings and in some cases on streets profile (narrow or wide). Within the frame of this problem, this proposal investigates only the vibration field with the objective to deduce the response of the track and of the surrounding soil, due to a tramway passage. The modelling of ground-borne vibrations is a general subject that is widely dealt with in literature (see review in Lombaert et al. [1]). Difficulties arise when the whole structure (i.e. tramway, track and soil) has to be considered, therefore rarely is the complete vehicle-track-ground system taken into account. In the first part of the paper, the model proposed for the track and the coupling conditions with the ground is drawn from Sheng et al. [2]. These authors coupled a two-dimensional track model to a layered halfspace model of the soil, using the Haskell-Thomson transfer matrices approach. Here a semi-analytical model is adopted (firstly introduced by Jones [3]) to obtain the flexibility matrix of the layered soil in the wavenumber domain (extended model in [4]). In the second part of the paper, track and soil characteristics are fitted B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 158–164, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Propagation of Vibrations Due to a Tramway Line

159

using both measurements, and manufacturer data. Then the tramway passage is calculated, using results from the first part concerning the point load solution. Hence, tramway dynamic load is implemented assuming that wheel-rail contact forces behave as uncorrelated fixed point loads. A comparison between calculations and measurements is finally made, in the particular case of a concrete slab track.

2 Track-Soil-Ground Model 2.1 Receptance of the Ground The ground is modelled as a multilayered and semi-infinite medium (figure 1), each layer being elastic, homogeneous and isotropic. For each layer, Navier's dynamic equation and linear behaviour law between stresses and displacements are written. Next, the use of the double Fourier transform

f ( β ,γ , z ) =

+∞+∞

∫ ∫ f ( x , y , z ).e

−i .( β . x +γ . y )

.dxdy

(1)

− ∞− ∞

provides the writing of the motion equation into the domain of the wave numbers β in the x direction and γ in the y -direction. x, y and z-directions correspond to the longitudinal (infinite) track direction, the transversal track direction, and the subsoil direction, respectively. Finally, it yields for each layer

{u

*

{- σ

}

(0), v * (0), w * (0), u * (h), v * (h), w * (h) * xz

T

= [T ].{A , B , C , D , E , F } (2) T

}

(0),-σ yz* (0),-σ zz* (0),+σ xz* (h),+σ yz* (h),+σ zz* (h)

T

= [S ].{A , B , C , D , E , F }

T

(3) where

σ xz* , σ yz* , σ zz*

indicate the stresses, and

u * , v * and w * the displacements.

All these elements are defined in the wave number domain in the x, y and z-direction respectively. The mark 0 corresponds to the top of the considered layer, and h to its bottom. In the following, a bar above a variable denotes its representation in the wave number domain, therefore β ,γ are omitted. S and T are 6× 6 matrices and

[ ]

[ ]

A , B , C , D , E , F are unknown constants. Thus, one obtains the following matrix equation (which eliminates constants)

[S ][. T ]−1.{u * } = {Σ } [ ][ ]−1

(4)

where S . T is the exact dynamic stiffness matrix of the layer derived in the transform domain, and its inverse is called the Fourier-transformed dynamic flexibility matrix of the layer. A dynamic stiffness matrix can also be derived for the halfspace. These matrices are then assembled to form a flexibility matrix representing the response at the surface of the whole ground. That is the reason for the minus or plus in the definition of displacement and stress matrices (equations (2)(3)) so that the

160

M. Maldonado, O. Chiello, and D. Le Houédec

stresses at the layer interfaces other than the top and bottom of the ground cancel (for more details see [5]).

Fig. 1. Track-soil-ground model

The vertical direct receptance of the ground due to a surface harmonic load of wave number β in the x-direction is given by [6]

H (β ) = where

1

π

+∞

∫Q

33

( β , γ ).

0

sin(γ .b) dγ γ .b

(5)

Q33 is deduced from the flexibility matrix, γ represents the wave number

associated with y-direction, and b is the half width of the track-soil interface. This receptance will be introduced in the track equations to obtain the coupled model. 2.2 Coupled Model Rails and concrete slab are modelled by means of infinite Euler beams and rail pads by continuous spring elements (equivalent continuous track models provide a reliable prediction of track receptance, more details available in [7] and Cui et al. [8]). Damping is taken into account as hysteretic one. All equations are derived in the wave number domain, using a one dimensional Fourier transform. The receptance of the soil is then introduced to complete these equations providing the "stiffness" of the soil. Precisely, concrete slab vertical displacement equals subsoil vertical displacement for the whole concrete slab/subsoil interface in the x-direction, assuming that displacements are identical in the y-direction at the interface. Finally, coupled model equations can be written in the wave number domain as

⎡ A1 ⎢A ⎢ 2 ⎢⎣ 0

A2 A3 1

* * ⎤ ⎧wr ⎫ ⎧ P ⎫ ⎪ ⎪ ⎪ ⎪ 1 ⎥⎥.⎨wc* ⎬ = ⎨0 ⎬ − H ( β )⎥⎦ ⎪ Fc* ⎪ ⎪⎩0 ⎪⎭ ⎩ ⎭

0

(6)

Propagation of Vibrations Due to a Tramway Line

161

where

⎧ A1 = − mr .ω 2 + k p .(1 + i.η p ) + Er I r .(1 + i.η r ).β 4 ⎪⎪ ⎨ A2 = −k p .(1 + i.η p ) ⎪ 2 4 ⎪⎩ A3 = − mc .ω + k p .(1 + i.η p ) + Ec I c .(1 + i.η c ).β

(7)

All these elements are considered per unit length of the track in the x-direction: mr or mc , Er I r or Ec I c and η r or η c are mass, bending stiffness and hysteretic damping ratio of the rail or concrete slab, respectively ;

k p and η p are the stiffness

and hysteretic damping ratio respectively, concerning the continuous rail pad element; the unknown quantities of equation (6) are

wr* (rail center vertical displacement), wc*

(concrete slab center vertical displacement) and

Fc* (force acting between the con-

*

crete slab and the soil), and P stands for the one-dimensional Fourier transform of the point-force P.δ ( x − x0 ) acting on the rail at the point x = x0 . Finally equations are solved, first in the wavenumber domain and then the real displacements are obtained by means of inverse Fourier transforms (one or two dimensional, as concerning the track or the soil respectively).

3 Comparison with Measurements The parametrical analysis mainly includes the tramway speed (from 20 to 50 km per hour), the tramway type (two manufacturers) and the track type (classical track, special track with softer rail pads, or lying on an elastic slab). In this paper, we only present results for a classical tramway track, without specific insulation. First, the soil is characterised by measurements using the Spectral Analysis of Surface Waves (S.A.S.W.). The objective is here to define shear moduli, layer thicknesses, scattering parameters and material damping. Secondly, the track receptance and the transfer functions between the track and the free field give information concerning the track parameters. Finally, measurements on the moving vehicle are used to obtain data relative to the real dynamic forces acted by the tramway on the rails. Besides, for each site, measurements were recorded to obtain an important data base concerning vibration levels induced by tramway passages, both on the rail and the ground surface. 3.1 Soil Characteristics Soil is characterised by measurements using the usual two-stations S.A.S.W. test with hammer impacts. The phase of the cross-power spectrum of the two receiver responses is used to obtain the phase velocity (of Rayleigh waves) vs. frequency. Attention must be paid to the frequency range of acceptable data (near field and attenuation effects) [9]. An inversion algorithm using a least squares optimisation routine [10] is then implemented to obtain a layered soil which corresponds to the experimental dispersion curve. The inversion procedure only gives Young's moduli, so the material

162

M. Maldonado, O. Chiello, and D. Le Houédec

soil damping η has to be chosen using a different method. Hence, the soil admittance is used to fit the hysteretic damping ratio for each layer. The results are presented in figure 2 (a). A good agreement is found for the comparison between the numerical and the experimental results.

(a)

(b)

Fig. 2. (a) Experimental (solid line: 2m - dashed line: 4m - dotted line: 8m) and computed (triangle markers) soil admittance - (b) Experimental (solid line) and calculated (dashed line) track-soil transfer functions (mobility) at a distance 2.5m from the track

3.2 Track-Soil Behaviour Track characteristics are obtained using experimental investigations with hammer impacts on the rail, these being performed to measure both the ground surface response at different distances from the track and the vertical response (receptance) of the rail. In particular, the rail pad stiffness is fitted with track receptance, and bending stiffness and mass of the concrete slab are fitted with track-soil transfer functions. Each procedure is solved using a nonlinear least-squares method. Figure 2 (b) illustrates the acceptable fitted response of the soil-track model. The comparison between calculated and measured transfer functions of the track provides a validation of the complete track-soil-ground model. 3.3 Dynamic Forces on the Rails In order to estimate the dynamic axle loads acting on the rails, measurements on a carrying bogie have been performed. Accelerometers have been fixed on axle boxes providing measurements of vertical vibrations close to each wheel of the bogie. Poor correlation is observed between the signals coming from the wheels of different axles whereas the coherence between the signals coming from two wheels of a same axle is very high around two frequency values (about 63 Hz and 125 Hz), for which the accelerations are also high. These peaks probably correspond to the resonance of the axle rigid modes (translation and rotation) on the track stiffness. The total vertical force acting on the two rails by an axle is calculated by adding linearly the accelerations measured close to each wheel and by multiplying it by the total axle mass (unsuspended mass). With this method, the second peak, corresponding to the rotation of the axle, is removed from the spectrum.

Propagation of Vibrations Due to a Tramway Line

163

3.4 Tramway Passage Validation of the complete model uses vertical vibrations of the ground surface which are considered in the frequency domain. A Fourier spectrum is performed concerning measurements, for a precise duration which corresponds to the time between two bogie passages (hence, this period depends on the tramway velocity) and the power RMS spectra vs. one third octave band is calculated. Calculation of the tramway passage is first implemented assuming that each axle induces a force on the rail which could be modelled as a point load. The calculated dynamics axle loads are supposed uncorrelated, according to measurements. The tramway passage is modelled assuming an initial position for the tramway, far enough from the receiver point. Next, a time increment gives another tramway position, and so on. Thus, for each tram position, vertical displacement of a considered point on the ground (receiver point) is obtained by summing each axle contribution, as power spectra. Maximum vertical velocity values are also obtained, figure 3 (a), whereas the same precise duration is used to define the average values (figure 3 (b)). A rather good agreement can be found for frequencies up to 100Hz. For frequencies larger than 100Hz, the difference between computed and experimental vertical velocities is probably due to the anti-symmetric excitation of the rails by the axle rigid rotation mode. Indeed, this excitation is not taken into account in the model since the slab rotation (torsion) is not possible. However, this phenomenon is less important when large distances from the track are considered (compared to the 2.5m case described here), since Rayleigh waves propagate in a lower frequency range.

(a)

(b)

Fig. 3. Computed (dashed line) and experimental (solid line) wave field amplitudes (dB ref. 5.10-8 m/s) - Tramway passage at 50km per hour - distance from the track: 2.5m - (a): maximum values - (b): average values

4 Conclusion In this paper, a complete model for the vehicle-track-ground is proposed to perform ground-borne vibration calculations. Vertical axle forces acting on the rail are estimated using measurements. A semi-analytical model for the track and ground is used and important parameters are fitted using measurements. Results presented here show

164

M. Maldonado, O. Chiello, and D. Le Houédec

that the proposed model is acceptable for a frequency range up to 100Hz. For frequencies above 100Hz, it seems that the anti-symmetric excitation of the rails cannot be neglected to predict the soil vibrations near the tracks. Current research is now focusing on the validation of the present model for other track-soil configurations (floating slab for example) and tram type (independent wheels). The calculation of dynamics axle loads from the measured unevenness of the wheel-rail surfaces is also in progress. The aim is to obtain a complete numerical model, which can be used without further measurements for the analysis and validation of new planned lines.

Acknowledgements In addition to GeM Laboratory and INRETS, this project involves three other partners: SerdB, SEMITAN and C.D.M. The authors wish to acknowledge the ADEME Agency for its financial support, and J.P. Regoin for his great contribution concerning with the experimental investigations.

References [1] Lombaert, G., Degrande, G., Vanhauwere, B., Vandeborght, B., François, S.: The control of ground-borne vibrations from railway traffic by means of continuous floating slabs. Journal of Sound and Vibration 297, 946–961 (2006) [2] Sheng, X., Jones, C.J.C., Petyt, M.: Ground vibration generated by a harmonic load acting on a railway track. Journal of Sound and Vibration 225, 3–28 (1999) [3] Jones, D.V.: The surface propagation of ground vibration. PhD thesis, University of Southampton (1987) [4] Picoux, B., Rotinat, R., Regoin, J.P., Le Houédec, D.: Prediction and measurements of vibrations from a railway track lying on a peaty ground. Journal of Sound and Vibration 267, 575–589 (2003) [5] Picoux, B., Le Houédec, D.: Diagnosis and prediction of vibrations from railway trains. Soil Dynamics and Earthquake Engineering 25(12), 905–921 (2005) [6] Sheng, X., Jones, C.J.C., Thompson, D.J.: A theoretical study on the influence of the track on train-induced ground vibration. Journal of Sound and Vibration 272, 909–936 (2004) [7] Lombaert, G., Degrande, G., Kogut, J., François, S.: The experimental validation of a numerical model for the prediction of railway induced vibrations. Journal of Sound and Vibration 297, 512–535 (2006) [8] Cui, F., Chew, C.H.: The effectiveness of floating slab track system - part one. receptance methods. Applied Acoustics 61, 441–453 (2000) [9] Foti, S.: Multistation methods for geotechnical characterisation using surface waves. PhD thesis, Politecnico di Torino, Italy (2000) [10] Lai, C.G., Rix, G.J.: Simultaneous inversion of Rayleigh phase velocity and attenuation for near-surface site characterisation. National Science oundation and U.S. Geological Survey (1998)

Railway Noise Statistics by Monitoring Stations – Input for Dutch Prediction Method RMR and Track Access Charging E. Verheijen1, M.S. Roovers2, and J.W. van den Brink2 1

dBvision, Vondellaan 104, 3521 GH Utrecht, The Netherlands Tel.: +31 6 29076165; Fax: +31 30 281 9844 [email protected] 2 ProRail, Postbus 2212, 3500 GE Utrecht, The Netherlands

Summary The infrastructure management organisation ProRail has installed five noise monitoring stations on the railway network in the Netherlands for the purpose of monitoring the progress of the Dutch Noise Innovation Programme. These monitoring stations record noise spectra and rail vibration spectra of train pass-bys. A weather station is integrated to allow the exclusion of data recorded during windy and rainy periods. As the stations are also capable of identifying trains that are equipped with tags. This creates a unique possibility for statistical analysis of the noise of such trains. A statistical data analysis programme has been conducted by Railway Noise Knowledge Centre of ProRail. In this programme, the performance of silent test trains has been monitored. These passenger and freight test trains have been equipped with LL braking blocks, which are expected to keep the running surface of the wheel smooth. The monitoring stations provide information on the long-term performance of these trains in terms of noise reduction. The analysis programme also involves an evaluation of the source parameters of the Dutch railway noise prediction method (RMR). These source parameters which depend on the type of rolling stock have been derived and established in the mid 1990s. It has been established that the average noise emission level of most train types is still close to the RMR values; however, for some train types a significant lower noise emission level has been measured. This will lead to adjustment of source parameters in RMR, which enables a more efficient use of noise abatement resources in the future. Furthermore, the potential usability of the noise monitoring in noise-based railway access charging is considered. For this purpose, the reliability and accuracy of individual pass-by measurements is assessed. Special attention is paid to the measurement uncertainty caused by local track properties, which, of course, should be eliminated before imposing a fine on noisy trains.

1 Introduction 1.1 Noise Monitoring and Source Recognition Noise monitoring is common practice in the assessment of industrial and aircraft noise sources. For such applications that generally aim at long-term Lden-assessment a B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 165–171, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

166

E. Verheijen, M.S. Roovers, and J.W. van den Brink

large range of reasonably priced stand-alone equipment is available. Spurious noise events (related to other sources than the ones under investigation) are eliminated automatically or manually during post-processing. This can be done if the sound characteristics of the source are known in sufficient detail (temporal and spectral behaviour) and if video systems are used. Noise monitoring is a much more complex task if the noise events and sources need to be classified into (sub)types. In the case of trains the principle of axle pattern recognition is often used to classify the type of rolling stock [1, 2]. The principle is based on the fact that the distance between successive axles is characteristic for each type of train, although this method is inadequate for the monitoring goals set by the Noise Innovation Programme. 1.2 Noise Innovation Programme and Project Goals Within the railway part of the national Noise Innovation Programme [3], retrofit solutions are applied to trains with cast-iron braking blocks. In order to study the long-time noise effects caused by the new braking blocks, monitoring stations have been developed that are capable of identifying the pilot trains and measuring their respective noise levels. Apart from this task the stations are also used to check the average and spread of the noise emission of regular trains and compare these to the values that are incorporated in the Dutch railway noise prediction method (RMR, [4]). The third goal for monitoring is to assess the usability of the stations for noise-based access charging.

2 The Monitoring Stations 2.1 Design Goals and Specifications The design properties of the monitoring stations have been described elsewhere [5], and will only briefly be repeated here. Five noise monitoring stations have been designed and built in conformity with Procedure A of the Reken- en Meetvoorschrift Geluidhinder, a national regulation for calculation and measurement of railway noise1. This basically means that the monitoring stations meet the requirements for the measurement environment, track condition, microphone position, rail roughness and weather conditions. Each station consists of a stand-alone computer, two microphones and accelerometers (one per track), a weather station and a data transmission modem (GSM). The monitoring stations make use of the Radio Frequency Identification (RFID) tags for train recognition. Most Dutch trains are equipped with these tags for maintenance purposes [6]. This allows for the statistical analysis of the pass-by noise of individual trains. 2.2 Measurement Parameters The following data are generated automatically after each train pass-by: 1

This regulation has become part of the European interim computation method for railway noise [4].

Railway Noise Statistics by Monitoring Stations –Input

167

• Pass-by time and date; • Noise: A-weighted noise level + octave spectrum 63 Hz – 8 kHz (microphone position at 7,5 m from the centre of each track, 1,2 m above the rail head); • Acceleration level of the rail + octave spectrum 31,5 Hz – 8 kHz, (accelerometer position underneath the rail foot, measuring in vertical direction); • Pass-by duration from buffer to buffer2, speed at the front and speed at the tail; • Number of vehicles, their tag identification numbers and rolling stock type; • Wind direction and speed, precipitation, air temperature. Besides train-related data, each station produces status and quality information to facilitate maintenance and control. The measured data and status information is transmitted each night to a central database which can be accessed through an internet application [7]. In order to acquire reliable data from the unmanned stations, extensive quality checking is required. This is done by automated rail vibration measurements, pass-by time checking and daily microphone calibration. Invalid data due to bad weather conditions and non-constant speed are excluded manually before further analysis. The rail roughness is measured manually once a year. Depending on the analysis task, the noise levels are adjusted for rail roughness differences. The measurement stations have been developed in 2004, and installed and tested over 2005. Full operation has started in April 2006. The monitoring stations are located on the main lines in the railway network (near Utrecht, Amsterdam, Rotterdam and Eindhoven). Both tracks of each line are Dutch standard ballasted track: 54E1 rail profiles on concrete sleepers (NS90) with stiff pads (FC9). As the project has only national objectives, the sites have not been tested against TSI requirements. Nevertheless, only one of the sites has TSI+ roughness and, although decay rates have not been determined, it is known from sites with similar track systems that the vertical decay rates can be out of TSI+ range between 400 and 630 Hz [8]. With lateral decay rates there is usually no problem. 2.3 System Performance The performance of the stations is monitored by means of key performance indicators. The most meaningful indicator, availability, counts the number of days that a station is fully operational. The average availability of the five stations lies between 15% and 60%. This rather poor performance is mainly due to problems related to the accelerometers, which frustrate the triggering of the measurements (e.g. deterioration of mechanical contact). Unfortunately, the repair times for accelerometer failures are long due to safety regulations, as track closure is required. Other failures, that occur incidentally, are related to instable hardware components like the modem, temperature control, weather station, back-up battery (UPS) and hard disk. The repair of failures has in some cases been postponed or prioritised because of the needs of the analysis programme. 2

The buffer position is read from a look-up table with all tagged trains types and combined with rail peak vibrations of the first and last axle, giving the time window for the noise analysis and also the pass-by time.

168

E. Verheijen, M.S. Roovers, and J.W. van den Brink

3 Monitoring Individual Trains This section describes the results of monitoring an individual test train in normal service. The cast-iron braking blocks of three test coaches of type “ICR” have been replaced by composite braking blocks (LL type) as a part of the Noise Innovation Programme. ICR trains use disc brakes for low braking power and, additionally, castiron blocks for high braking power. The trains that include these test coaches also include regular ICR coaches with cast-iron blocks. They run on several lines and will sometimes pass a monitoring station. Figure 1 shows the development of the pass-by noise level of the middle coach of the three coaches with LL blocks, starting from the moment of reprofiling and retrofitting3. Each circle represents one pass-by at a speed between 120 and 135 km/h. The measurements are all taken from the same site. The rail roughness of that site lies between the ISO3095 limit and the TSI+ limit. The graph shows a gradual increase of the noise emission during the first two months (mileage 40.000 km) and then the noise emission stabilises. This period is about 4 times as long as with standard cast-iron blocks after reprofiling. Because of the monitoring stations it is now known that LL blocks may require a much longer period before representative noise levels are measured than for instance cast-iron blocks. This information is important for type testing programmes and also for measurements that feed the noise prediction method.

L_Aeq [dB(A)]

95

90

85

80 -10

0

10

20

30 40 50 60 Days after retrofitting

70

80

90

100

110

Fig. 1. Development of A-weighted equivalent sound pressure level LAeq after retrofitting and reprofiling for LL block equipped wagons

4 Checking the Source Parameters of the Noise Prediction Model 4.1 Train Categories At present eleven categories of trains exist in the Netherlands on the basis of their noise emission. Each category has a fixed set of source parameters for rolling noise 3

With respect to measurement accuracy, only three silent coaches is not preferable, but it is an operational condition that we have to face and which may lead to a small overestimation of the noise levels of the middle coach.

Railway Noise Statistics by Monitoring Stations –Input

169

and, if applicable, braking noise, traction noise and aerodynamic noise. The parameters were originally determined in the early 1980s and updated and extended in the mid 1990s [9]. More than 80 train pass-bys for each category were used to derive the parameters. Although there has been no particular reason to doubt the values of the parameters, it is considered that the noise monitoring stations offer a good possibility to check these parameters and to update them, if necessary. Under the Noise Abatement Act (Wet geluidhinder), an update of these parameters would affect the required intensity of noise reduction measures in future construction plans for houses and railways. 4.2 Results A dataset of 10.000 pass-bys has been analysed. Most of these are from the first half of May 2006 from three monitoring stations. The measured rolling noise of fifteen sub-types of rolling stock out of six categories has been compared with the theoretical values. The results for the main train types are given in Table 1. Fortunately none of the examined trains have become noisier over the last decade. On the contrary, three types of trains turned out to be quieter than expected. Also, for two types of trains a steeper slope was found for the noise emission level as a function of train speed. These trains are about 3 dB less noisy at a speed of 80 km/h, but still show good agreement at maximum speeds of the conventional network (130-140 km/h). All measurements have been adjusted for rail roughness differences in conformity with the national method (see reference [10], equation (1)). Table 1. Comparison of the old and new noise measurement cat. 1 2

decade

type, brakesa

# meas.

match with results of 1995 at 80 km/h

at 130 km/h

Mat64

60s

EMU, b1

1600

within 1 dB

ICR

80s

coach, b2

1500

within 1 dB

70-80s

EMU, b2

1400

within 1 dB

ICM3

3

SGM

70s

EMU, b3

450

now 3 dB less

within 1 dB

4

freight

any

all types, b1

600

within 1 dB

–

DDM2

80s

EMU, b3

850

8

9 a

name

within 1 dB

ICM4

90s

EMU, b3

700

IRM

90s

EMU, b3

900

now 3 dB less

now 2 dB lessb

within 1 dB

TGV-PBA

90s

EMU, b3

70

now 4 dB lessc

ICE-3M

90s

EMU, b3

70

now 4 dB lessd

b1 = cast-iron blocks, b2 = disc brakes + additional cast-iron blocks, b3= no cast-iron blocks (i.e. disc, Kblocks or magnetic). b IRM was already slightly quieter than category 8 in 1995. c Category 9 has been based on high speed measurements of a older TGV type (Atlantique) in France. d ICE has been assigned in 2006 to the same category as TGV-PBA.

170

E. Verheijen, M.S. Roovers, and J.W. van den Brink

4.3 Conclusion The monitoring stations can be used to assess the source parameters of the national noise prediction method. Monitoring stations can provide a larger statistical basis than manned measurements. Monitoring stations can also be used to determine the category of newly introduced rolling stock. They form an independent source of information which is much more relevant to noise prediction than the usual noise type testing report that is based on fresh new trains. It is believed that significant changes of the actual noise emission will not occur on the time scale of one year. Therefore, the source parameters need only be checked every two or three years.

5 Usability for Noise-Based Track Access Charging Railway operators have to pay for the use of the railway infrastructure for their services. ProRail aims to reflect the weight of trains and the scarcity of rail capacity in the access charges. One of the environmental factors of scarcity is the noise emission. It is considered that noise-based access charges would stimulate silent rolling stock (new of retrofit). In principle noise monitoring stations could play a role in the administration of noise emission. A few principal requirements must be met: • Reliability of the measurements must be high. This has been evaluated through analysis of repeatability and comparison with manned measurements. It appears that the unmanned stations perform as well as manned measurements. • Accuracy of the judgement (which noise charge?) must be high. The overall accuracy is limited due to differences between measurement locations that cannot (yet) be compensated for. Even after compensation for rail roughness, one individual train yields a standard deviation between around 1,4 dB(A) between different, but highly similar sites. • Sufficient stations must be installed over the railway network. In the Netherlands, approximately 40 monitoring locations are needed for full coverage [6]. A matter of concern for access charging is the poor up-time of the stations (less than 60%). Also, the question if noise measurements are useful for access charging depends on the method of charging. If the charge is related to the actual noise emission of a train, then noise measurements are useful. If the charge is related to the number of silent vehicles in a train, then noise measurements are not sufficiently discriminating. This will be problematic for mixed freight trains of which some vehicles may be retrofitted while others are not. In that case a tag reader would suffice in order to identify and count the number of silent vehicles; noise measurements would not be required then.

6 Conclusions The noise measurement stations have proved to be excellent tools to study railway noise emission. Especially, the train identification feature creates unique research possibilities.

Railway Noise Statistics by Monitoring Stations –Input

171

The monitoring stations provide reliable information about the performance of silent pilot trains. Because of the monitoring stations it is now known that LL blocks may require a much longer period before representative noise levels are measured than for instance cast-iron blocks. The monitoring stations allow for regular updates of source parameters for the noise prediction method. It turned out that most of the noise values of the prediction model were still valid after ten years. The stations can also be used for track access charging, however it will be difficult to measure accurate noise levels if silent vehicles are sandwiched between noisy vehicles within the same train.

Acknowledgements The Noise Innovation Programme (Innovatieprogramma Geluid) develops measures to tackle traffic noise at the source, to make Dutch railways and highways quieter. Rather than inventing new technologies, this programme uses available knowledge from earlier (European) research projects and prototypes from industrial research projects and puts them into practice. The programme includes the development and data analysis of the noise measurement stations and is initiated and sponsored by the Ministries of Transport and Environmental Affairs.

References [1] Automatic vehicle classification by axle pattern recognition is a widely used technique for the railways [2] Monitoring Eisenbahnlärm, Jahresbericht, Bundesamt für Verkehr, Switzerland (Robert Attinger) (2006) [3] The Noise Innovation Programme website (also English), http://www.innovatieprogrammageluid.nl [4] Reken- en Meetvoorschrift Geluidhinder, Bijlage IV en Technische Regeling, Ministerie van VROM (a translated earlier version, 2006), http://circa.europa.eu/Public/irc/env/noisedir/library [5] van den Brink, J.W.: Permanent Measurement Stations for Railway Noise. In: Proceedings of Euronoise 2006, Tampere, Finland, p. 235 (June 2006) [6] den Buurman, G., Zoeteman, A.: A vital instrument in asset management. European Railway Review (3), 80–85 (2005) [7] Lub, J.: The Web Processing and Presentation of Railway Noise Monitoring Data. In: Proceedings of Euronoise 2006, Tampere, Finland, p. 238 (June 2006) [8] Verheijen, E.: Geluidreducties raildempers en slijpen, proef Veenendaal, report AEAT/02/1400041/031 (December 2002); Geluidreducties raildempers, proef Oudenbosch, report AEAT/03/1400041/033 (January 2003) [9] Janssen, G.: De geluidemissieformules van nieuw en bestaand spoorwegmaterieel voor de standaard rekenmethoden I en II; een voorstel op basis van metingen, NSTO report 9471096/9571114, The Netherlands (December 21, 1995) [10] Verheijen, E.: A survey on roughness measurements. Journal of Sound and Vibration 293(3–5), 784–794 (2006)

Measurement and Modelling of Noise from the Arsta Bridge in Stockholm A. Wang1, O.G. Bewes2, S.J. Cox1, and C.J.C. Jones3 2

1 Pandrol Limited, 63 Station Road, Addlestone KT15 2AR, England Arup Acoustics, Blythe Gate, Blythe Valley Park, Solihull B90 8AE, England 3 ISVR, University of Southampton, Southampton SO17 1BJ, England Tel.: +44 1932 834514; Fax: +44 1932 850858 [email protected]

Summary This paper describes the measurement and modelling of noise on a steel railway bridge over the Arstaviken bay in Stockholm, Sweden. The track on the bridge was refurbished with the objective of reducing noise emission from the bridge. A mathematical model of the steel bridge structure was developed. This includes rolling noise from the train and track, and noise resulting from vibration of the structure. The model was used to identify the noise spectra from different sources within the system and these were compared with direct measurements. The model was then used to predict the effect of changing the stiffness of the rail fastening system on the steel bridge, first on the structural noise spectrum and then on the total noise level. The predictions showed a reduction in the low frequency rumbling component of noise. Finally, baseplates of the stiffness modelled were installed on the bridge, and the measurements were repeated. The installation of the new baseplates was successful in controlling low frequency noise from the bridge. The reduction in the low frequency component of noise was approximately in agreement with that predicted.

1 Introduction The old Arsta bridge, which was opened to traffic in 1929, is on the main railway route running south from Stockholm Central Station. The bridge is about 650 metres long and carries two tracks. It consists of a reinforced concrete arch viaduct, with a short lifting span towards the northern end, and a longer riveted steel structure towards the southern end where the bridge crosses the shipping channel. The track on the viaduct is ballasted. The lifting bridge and the steel bridge have open deck structures with transverse timber bearers. The steel bridge section is the main object of this study. Microphones were set up at positions on a new parallel bridge, opposite the concrete and the steel sections of the Arsta bridge. In 2004, measurements were made under normal service traffic. The trains running across both sections of the Arsta bridge were of the same type, and were travelling at the same speed. The assumption was made that the re-radiated noise due to the steel structure could be estimated as the difference between the two measured noise spectra. In addition, a theoretical study was performed to predict the noise reduction that would result if the existing fastening system on the steel section were replaced with B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 172–178, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Measurement and Modelling of Noise from the Arsta Bridge in Stockholm

173

≈ 650 m 150 m

concrete section

steel section

Fig. 1. Arsta bridge in Stockholm

new baseplates that have a lower dynamic stiffness. When the bridge was refurbished in 2005, the existing timber bearers were replaced and PANDROL baseplates were installed. Further measurements were then made in 2006 and these have been compared to the predictions.

2 Noise Modelling 2.1 Modelling Approach - NORBERT The bridge modelling approach used is based on work by Janssens and Thompson [1] and Harrison, Thompson and Jones [2]. The approach is to model the bridge with an analytical track model to calculate the power input to the bridge, then couple this to a simple SEA model to calculate the power distribution amongst the various components of the bridge. Based on this previous work, an improved model that takes better account of the finite length of the bridge and contains a better calculation for the power input to the bridge has been developed [3, 4]. The combined wheel and rail roughness is used as the input to the model. Roughness values are either used from a database or specifically measured roughness spectra can be input. Using the roughness, the vibration of the rail and wheel is calculated from the rail and wheel mobility and the speed of the rolling stock. Rail vibration is considered in the vertical direction only. At low frequencies, when there is strong coupling between the rail and bridge, the track is modelled with two finite simply supported Timoshenko beams that represent the rail and bridge support girders, continuously connected via a resilient layer that represents the rail fastening system. The power input to the bridge resulting from the induced rail vibration is then calculated from the relative velocity in the resilient layer and the stiffness of the resilient layer. At high frequencies, when coupling between the rail and bridge is weak, the rail is modelled as a Timoshenko beam continuously supported on top of a rigid surface by a resilient layer. The force input to the bridge is assumed to be the resulting force transmitted through the resilient layer if the rail is subject to the calculated vibration. Using either track model, the power input to the bridge structure per wheel can be calculated from the force acting at the output side of the resilient layer and the mobility of the bridge. The calculated power input to the bridge is then distributed amongst the components of the structure using a simple form of statistical energy analysis (SEA), which assumes equipartition of energy among all of the components in the bridge. The SEA model of the bridge structure itself is made up of only plate components. Using the

174

A. Wang et al.

radiation efficiencies of each component, the total sound power radiated by the bridge is calculated. Finally a rolling noise database is accessed that contains calculations for the rolling noise of several wheel and track combinations calculated by using the TWINS software [5, 6]. The correct values are accessed and corrected for differences in the rail fastener stiffness and the rolling noise is output in terms of sound power per unit length of bridge. The sound powers radiated by the bridge, rail and wheels are then summed to give a total noise from the bridge. 2.2 Input Parameters A summary of the input parameters used and assumptions made is given below. Full details of the measurement parameters can be found in [4]. Rolling stock: Traffic consisted of various commuter trains made up of 4 to 8 carriages. Rolling stock parameters where obtained from Banverket for one of the most common types of train passing over the bridge, the type M3 commuter train. The speed of each train was assumed to be 70 km/h, the design speed of the track. Track: The Arsta bridge supports two tracks. Prior to refurbishment, the running rails were BV50 rails fastened to wooden sleepers with Hayback fastenings. On the steel section of the bridge, each sleeper is fastened directly to two steel I-section support girders. The dynamic stiffness of the pad and the sleeper in series has been taken as 265 MN/m, which is typical [7]. The rail and wheel roughness at the site were unknown. A combined wheel/rail roughness spectrum has been assumed, based on typical disc braked wheels and rail roughness in the Netherlands [8]. Bridge: Part of the steel section of Arsta Bridge is shown in Figure 2. The bridge is made up of an assemblage of approximately 2,400 steel plates. Each of these plates has been grouped by type, with common cross-sections and dimensions, and the quantities of each plate type were used in the SEA model of the bridge.

Fig. 2. Part of the steel section of Arsta bridge showing its component “plates”

Measurement and Modelling of Noise from the Arsta Bridge in Stockholm

175

3 Noise Measurements Noise was recorded at positions opposite the centres of different sections of the Arsta bridge, including a section of the concrete bridge over the water, and the steel bridge over the water. In each case the microphone was positioned at approximately 1.5m above rail head level and between 45 and 55 m from the bridge. For all measurements reported in both 2004 and 2006, the weather was fine and clear with no wind, and the temperature was between 5˚C and 7˚C. Noise recordings were taken between 9:00 and 11:00am. Among the noise recordings made were 9 trains of type M3 before the track was refurbished, and 9 trains of the same type afterward. Each recording was individually checked for quality so that recordings with high background noise could be eliminated. Between 3 and 5 good quality recordings were identified for each section with similar train type and at similar speeds. These were averaged to provide the 1/3 octave A-weighted spectra shown.

4 Comparison between Predicted and Measured Results Measurement results for 2004 are shown in Figure 3 as 1/3 octave band spectra and total A-weighted levels. The noise spectra on the steel bridge are similar to those measured on the same bridge by Ingemasson [9]. The highest levels occur between 400 Hz and 1 kHz on the steel bridge section and 400 Hz and 2 kHz on the concrete bridge section. Above 1 kHz, the spectra recorded on both bridge sections are similar. The noise in this frequency range is likely to be dominated by airborne rolling noise from wheels and rails. The results suggest that the majority of the wayside noise measured from the steel bridge is secondary noise emanating from the bridge structure, and that this contributes 3-7 dB(A) to the total level below 1 kHz.

85

81.4

80

77.2

75 70.5 70 65 60 55 50 45

50 00 on cr et e Ba Ste el ck gr ou nd C

40 00

31 50

25 00

20 00

16 00

12 50

10 00

80 0

63 0

50 0

31 5

40 0

25 0

20 0

16 0

12 5

40

10 0

Sound Pressure Level dB(A) (ref 20E-6 Pa)

90

1/3 Octave Central Frequency (Hz) Concrete bridge

Steel bridge

Estimated noise component from steel bridge

Fig. 3. Noise from the same train on both the concrete and steel bridges, 2004

176

A. Wang et al.

80

80

75

75

S o u n d P re s s u r e L e v e l d B (A ) (re f 2 0 E -6 P a )

S o u n d P re s s u r e L e v e l d B (A ) (re f 2 0 E -6 P a )

Assuming that the noise levels radiated by the massive concrete arched sections can be neglected and that the rolling noise is the same in each case, the component of the secondary noise radiated by the steel structure alone can be estimated by subtracting the concrete section spectrum from the steel section spectrum. This is also plotted in Figure 3, and suggests that on the steel section, the structure-radiated noise is the dominant noise source between 200 Hz and 800 Hz, whereas above 800 Hz rolling noise is the dominant source. The estimated re-radiated noise component and the re-radiated and rolling noise components predicted using the NORBERT software are shown in Figure 4. Good agreement can be seen between the estimate and the prediction of the re-radiated noise as shown in Figure 4(a), thereby validating the model and the assumptions regarding the relative contribution of the rolling and re-radiated noise components to the total level. Also plotted in Figure 4(a) is the predicted structure radiated component if the fastener stiffness is reduced to that of the Pandrol VIPA SP baseplate. Significant reductions in the structure radiated noise are predicted. The predicted rolling noise component is slightly increased with Pandrol VIPA SP baseplates as shown in Figure 4(b).

70

65

60

55

50

65

60

55

50

45

45

40 100

70

125

160

200

250

315

400

500

630

800 1000 1250 1600 2000 2500 3150 4000 5000

Re-radiated noise - before

Re-radiated noise - estimated before

(a) Re-radiated noise

40 100

125

160

200

250

315

400

500

630

800 1000 1250 1600 2000 2500 3150 4000 5000

1/3 Octave Central Frequency (Hz)

1/3 Octave Central Frequency (Hz) Re-radiated noise - aftrer

Rolling noise - before

Rolling noise - after

(b) Rolling noise

Fig. 4. Predicted re-radiated and rolling noise spectra

In 2005, the track on the steel section of the bridge was refurbished with the aim of reducing wayside noise levels adjacent to this section. The rail, cross-timbers, and the existing fastening system were replaced with a UIC60 rail fastened with Pandrol VIPA SP baseplates. The significant change is an overall reduction in fastener stiffness from 265 kN/mm to 19.5 kN/mm. Noise measurements were made in the same locations after the refurbishment had been implemented, and the results of the simulations for the ‘before’ and ‘after’ conditions are compared to the corresponding measurements in Figure 5. There is generally good agreement between measurement and prediction. Measured results show that the reduction in wayside noise is most significant for frequencies less than 500 Hz with reductions of more than 10 dB(A) seen in some frequency bands. The overall reduction in A-weighted noise level of 3 dB(A) was less significant, as the rolling noise is clearly still a significant contribution to the overall noise at high frequencies. Similar results have been reported by Ingemansson [9].

Measurement and Modelling of Noise from the Arsta Bridge in Stockholm

177

80

Sound Pressure Level dB(A) (ref 20E-6Pa)

75

70

65

60

55

50

45

40 100

125

160

200

250

315

400

500

630

800

1000 1250 1600 2000 2500 3150 4000 5000

1/3 Octave Central Frequency (Hz) Measured - before

Measured - after

Predicted - before

Predicted - after

Fig. 5. Steel bridge noise levels before and after refurbishment

5 Discussion The work has shown that in the right circumstances, fitting low stiffness baseplates to steel bridges can reduce wayside noise. This is because vibration of the structure, and so the re-radiated noise component, is reduced. However, this reduction has to be balanced against a possible increase in the rolling noise component. This is sometimes associated with reductions in track stiffness [6]. It has previously been noted that the wayside noise level near steel bridges depends on the detailed design of the bridge [10]. The relative importance of the rolling and re-radiate noise components will vary according to the bridge design. So it might be expected that the degree of noise reduction that might be achieved by fitting low stiffness baseplates would also depend on the bridge design. The stiffness of the fasteners that were replaced is obviously also relevant to the reduction achieved. The authors have been involved in fitting similar baseplates to a number of different bridges [11 - 16]. While the number that has been treated is small, it has been noted that in general greater reductions are achieved on bridges with plate decks that radiate high levels of noise and side–walls that block rolling noise than is the case on more open structures with cross bearers. The Arsta bridge is typical of the latter, and the total noise reduction, at 3 dB(A), was relatively modest. Reductions of up to 15 dB(A) have been recorded on other structures [10].

6 Conclusions The noise measured during the pass by of commuter trains opposite a steel bridge structure was approximately 4 dB(A) higher than opposite a concrete bridge. The measurements suggest that for frequencies between 400 Hz and 1 kHz, the majority of the wayside noise from the steel bridge was secondary noise radiated from the structure. Predictions were performed to estimate the noise reduction that would be

178

A. Wang et al.

achieved when the bridge was refurbished by fitting low stiffness baseplates and a new rail. Good agreement was found between the measured and predicted noise spectra for both the ‘before’ and ‘after’ conditions. A significant reduction in wayside noise for frequencies below 500 Hz was measured, and there was an overall reduction in the noise level of approximately 3 dB(A).

References [1] Janssens, M.H.A., Thompson, D.J.: A calculation model for noise from steel railway bridges. JSV 193, 295–305 (1996) [2] Harrison, M.F., Thompson, D.J., Jones, C.J.C.: The calculation of noise from railway viaducts and bridges. In: Proceedings of the IMechE, vol. 214, Part F, pp. 125–134 (2000) [3] Bewes, O.G., Thompson, D.J., Jones, C.J.C., Wang, A.: Calculation of noise from railway bridges and viaducts: Experimental validation of a rapid calculation model. In: Proceedings of the Eighth International Workshop on Railway Noise, Buxton, pp. 447–458 (2004) [4] Bewes, O.G.: Calculation of noise from railway bridges and viaducts, EngD Thesis, Southampton University (2005) [5] Thompson, D.J., Hemsworth, B., Vincent, N.: Experimental validation of the TWINS prediction program for rolling noise, part 1: Description of the model and method. JSV 193, 123–136 (1996) [6] Thompson, D.J., Fodiman, P., Mahé, H.: Experimental validation of the TWINS prediction program for rolling noise, part 2: Results. JSV 193, 137–148 (1996) [7] Thompson, D.J., Verheij, J.W.: The dynamic behaviour of rail fasteners at high frequencies. Applied Acoustics 52(1), 1–15 (1997) [8] Dings, P.C., Dittrich, M.G.: Roughness on Dutch railway wheels and rails. JSV 193, 103– 112 (1996) [9] Ingemasson Report No. 31-02138-F. The Old Arsta bridge over the Arstabay, Stockholm: Result of acoustical damping (September 2006) [10] Hardy, A.E.J.: Noise from railway bridges. In: Proceedings of the IMechE, vol. 213, Part F, pp. 161–172 (1999) [11] Wang, A., Cox, S.J., Gosling, D., Prudhoe, J.E.W.: Railway Bridge Noise Control with Resilient Baseplates. Journal of Sound and Vibration 231, 907–911 (2000) [12] Wang, A., Li, M.: Railway bridge noise and vibration control - Karlstad Bridge in Sweden. Pandrol Journal (2002) [13] Brekke, A.: Measurements after noise reducing measures (Norwegian), The Institute of Transport Economics Report 7501-10 (2003) [14] Poisson, F.: Reducing the noise from steel bridges, http://www.recherche.sncf.fr/uk/projets/uk_bruit_ponts.html [15] Koestli, K.: Experimental and theoretical analysis of railway bridge noise reduction using resilient rail fasteners; A case study. In: IWRN9 (2007) [16] Pandrol Report 85163, Test of Arad Bridge – Romania (2004)

Minimising Noise from Viaducts in the Borough Area of London for the Thameslink Programme C. Cobbing1 and C.J.C. Jones2 1

Temple Group Ltd, Barnards, Station Road, Horsted Keynes RH17 7ED, West Sussex, UK Tel.: +442084220258 [email protected] 2 Institute of Sound and Vibration Research, University of Southampton, Highfield, Southampton SO17 1BJ, UK Tel.: +44 2380 593224; Fax: +44 2380 593190 [email protected]

Summary Network Rail’s Thameslink Programme is designed to deliver a substantial increase in passenger capacity north-south across London. Part of the measures necessary to achieve this is the construction of new viaducts in central London south of the river Thames parallel to existing 19th century wrought iron viaducts. The route passes through a busy city area and it is vital for the project to demonstrate that it will minimise noise from the railway. Studies are reported here that assess bridge and track design to reduce noise - these use the ‘Norbert’ model for bridge noise rolling noise. A combined view of results for of old and new bridges has been used to show the important noise sources for ten sensitive receivers. The acoustic design for the area has then been optimised by looking at the effects of the combination of different track and bridge designs.

1 Introduction The Thameslink Programme will build a new set of viaducts in the Borough Market/London Bridge Station area of central London. These have been modelled with ‘Norbert’ [1] to assess the noise mitigation options for bridge noise and rolling noise. It has been necessary to predict the relative noise contributions for each structure, both new and existing at each receiver location. A summary presentation of otherwise detailed predictions enables practical decisions to be made to optimise the acoustic design. For this process, ten sensitive receiver locations were chosen by the project and 15 different bridge structures were modelled for the proposed track alignment with the scheme. These are a mixture of the existing steel bridges, for which the model has been checked, and proposed concrete composite viaduct spans. The results for the relative sound levels at each of the receivers have been presented to show which bridges are the greater sources for that particular receiver. This combination takes into account the distances to different track structures, the effects of barriers, the track design on each structure and number of trains on each route. A set of structural and track designs for the new structures already existed at the beginning of the current exercise. They had already been optimised acoustically B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 179–185, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

180

C. Cobbing and C.J.C. Jones

within the project constraints and represent a snapshot at a stage of the project development circa 1999 - these were used in the preparation of the Environment Statement [2] published in 2004. The reference for the assessment of improvements to the acoustic design in the current work is taken as this set of designs. Since 2004 the engineering of the project has been reviewed and many of the constraints applying to noise mitigation revisited. This provides opportunities for improved acoustic design in many parts of the Thameslink Programme and an area around Borough Market in particular. The process of improving the noise mitigation in this part of the project is the subject of this paper.

2 The Bridges and Buildings in the Borough Market and London Bridge Station Area Fig. 1 shows a diagram of the area that has been studied. This portion of the Thameslink route passes south of the river Thames and is situated to the west of London Bridge Station. The tracks in the whole area are supported about 11 m from ground height on a series of viaducts. The lighter blocks in Fig. 1 show the existing bridges constructed in 1865 and widened in the 1890’s. Cannon Street lines

R10

Bridge 47 Bridge 47A Existing Stoney Street Borough Viad. Bridge 45B Bridge

Railway App. Bridge 51

R3

North Side Park Street Viaduct

R5

Thameslink lines

R1 R4

New Borough Park Street Viaduct Market Viaduct

R2

Southwark St Bridge

R6

New Borough High St Bridge R7

New Railway Approach

New Station Approach

R8

Fig. 1. Diagram of the layout of structures and receiver locations

Southwark Street Bridge is a wrought iron structure with ballasted track, this leads onto Park Street Viaduct which is a series of brick arches. This is to be widened in the new scheme so that the track can split onto two routes; the existing structures taking a double track to the north and a set of new spans taking a double track to the south of the existing route. The existing wrought iron structures of Stoney Street Bridge and Borough Viaduct take the railway over an area of market that is itself a

Minimising Noise from Viaducts in the Borough Area of London

181

London attraction. These structures carry a track based on longitudinal timber bearers. Bridge 45B consists of three spans of ballasted track. It is on this structure that two more lines join from the north out of Cannon Street Station. There is a cross-over on the northern two tracks of this four track structure which are not part of the new alignment associated with the Thameslink Programme. Bridges 47 and 47A again have the timber bearers but Railway Approach Viaduct and Bridge 51 revert to ballasted track. The widening of Park Street Viaduct is to be made with masonry-supported concrete arches. The new Borough Market, Borough High Street Bridge, Railway Approach Viaducts are to be made as concrete decks supported at the sides by steel box girders. On these new structures, the 1999 design prescribed concrete slab track with rail baseplates having a dynamic stiffness of about 50 MN/m.

3 The Bridge Noise Model The ‘Norbert’ bridge model has been developed as a prediction tool for both the structure radiated and vehicle-track components of rolling noise. Fig. 2(a) shows a simplified diagram of the modelling process. The input is the combined roughness expected of the wheel and rail running surfaces. In the present work, roughness typical of the UK system has been used [3]. The principles of calculating the wheel and track response to their combined roughness are well established in rolling noise models [4]. In comparison with the existing TWINS model for rolling noise an extra component, the bridge structure, is excited through the track. The analytical interaction model is depicted in Fig 2(b). The part of the bridge structure, either a thick plate representing a concrete deck, or a Timoshenko beam representing longitudinal support is incorporated in the model used for frequencies below the rail decoupling resonance frequency. The power input to the supporting member of the bridge is calculated using this model as well as the response of the track components and the wheel. Only vertical dynamics are considered. For frequencies above the rail resonance, the power input to the bridge is calculated as that transmitted though an equivalent point source into the mechanical impedance of the track-supporting bridge component. For this high-frequency region, a special approximation for the mechanical impedance of an Isection beam has been derived [5]. The components of the bridge structure are modelled as a Statistical Energy Analysis (SEA) network of plate-bending subsystems and the special assumption of ‘equipartition of energy’ is made. This allows the mean square velocity response of each plate to be calculated only on the basis of the vibrational energy being shared amongst the plates according to their area, Sj, and thickness, hj. The resulting expression for the mean squared velocity of a particular plate,

< v j > , of the bridge structure is shown 2

in Fig 2(c) [6]. With the vibration response per one-third octave band for each of the bridge plates, the radiated sound power can be calculated according to the formula shown in Fig 2(d) using simple analytical approximations for the radiation ratios, σj, for each plate. The component sound powers radiated by the wheels of the rolling stock, the various track forms and 15 different bridge structures have been calculated. For the noise

182

C. Cobbing and C.J.C. Jones

at each receiver, these sound powers are used in a propagation calculation that takes into account different paths for those sources inside the parapets and noise barriers of the bridge and those outside such as the outside of parapets and the underside of the deck. The height and location of the receiver in relation to the sources are taken into account using a standard finite line-source approximation [7]. The effects of the parapets and barriers have been accounted for using a Maekawa calculation [7] of a frequency dependent insertion loss. This is important for these sources that have predominantly low frequency content. The models for the existing bridges were validated against measurements to ensure that the models formed a good basis for considering variations to the track designs.

4 Engineering Design Process Since 2004, in the design review process, a number of track designs have been considered for the new structures as alternatives to the slab track. In addition, designs for renewal of track on the existing structures have been investigated. In recent years, rail dampers have been shown to reduce rail noise where soft supports are used. Thus, they have been brought into consideration. Whereas the slab track and 50 MN/m support stiffness may have been optimum previously, the introduction of rail damping provides further opportunity for overall noise mitigation using softer supports. A systematic process has been used to assess noise mitigation options. In this work several different options have been considered in regard to (i) technical feasibility and roughness

(a) (b)

vehicle suspension

wheel/rail interaction

wheel contact spring

roughness input

rail vibration

wheel vibration

rail radiation

wheel radiation

rail power input to bridge

(c) v

2 j

=

Σ

transmission to bridge deck

bridge deck or beam

vibration of component 1

vibration of component i

vibration of component n

radiation from component 1

radiation from component i

radiation from component n

Win

ωη j ρ j h 2j ∑ i

Si hi

(d) Win = ρ 0c0 ∑ σ i Si vi2 i

Σ

shielding

bridge noise

train noise

Σ

Propagation calculation including barriers.

total noise

Fig. 2. Diagram of the major components of the Norbert bridge noise model

Minimising Noise from Viaducts in the Borough Area of London

183

practicality, (ii) acoustic performance, (iii) costs (capital and maintenance), (iv) operational and safety considerations and (v) other matters such as visual impacts. A series of workshops were held with the project engineers and studies based on different track designs were discussed. In the first place, typical existing and proposed structures were used in conjunction with a large number of track models. Tracks types considered at this stage included slab tracks with resilient baseplates (including Vanguard), booted sleeper slab tracks, ballasted tracks with ballast mat or with sleeper soffit pads and floating slab track. A range of dynamic support stiffness for each of these was considered with and without rail dampers. This gave information on the relative noise reductions that could be achieved with different track types that are feasible for certain of the bridges. The noise predicted from different bridge structures varies considerably and, the balance of the wheel-track noise in comparison to the bridge noise varies according to the propagation path and the presence of parapet noise barriers. Therefore each bridge must be considered using a separate Norbert model and its effect at each receiver also modelled individually. A presentation of the information arising out of the model was therefore devised to summarise the large amount of data in a way that could be used by everybody involved in the selection process to see the result of different choices of track design on different bridges. This data for a particular example receiver, the Globe Public House (Fig. 3) is presented in Fig. 4 for the track designs at the beginning of the review process. The Globe sits at the centre of the area in question immediately adjacent to the Fig. 3. The Globe Public House, receiver 5 with existing Bridge 45B (three spans) on the spans of Bridge 45B behind and the cathedral the opposite side to Southwark Cabeyond thedral. The new viaduct will pass close on the other side of the Globe. Fig. 4 shows the predicted sound level during a train pass-by at three heights of the building: near the ground (below the viaduct), at track level and at the floor level overlooking the tracks. For each of the bridges identified as significant sources here, the wheel-track noise and bridge structure noise are identified separately. It is clear that the existing viaduct is noisier than the new viaduct is expected to be; the largest component radiated by the structure. In order to assess the overall noise at the Globe, the relative number of trains during the day must be used along with the predictions of pass-by noise from each structure to calculate the period LAeq. Using representations like Fig. 4 for each of the ten receiver locations considered, a number of track design choice ‘scenarios’ were assessed. These were chosen on the grounds of feasibility, maintenance and consistency in engineering along the track as well as acoustic efficiency.

184

C. Cobbing and C.J.C. Jones =================================== R5: Globe Public House ============================== Ground level, 1.6m

Track level, 11.6m

76.0

1.Bridge

76.8

1.Bridge

58.8

45B

14.8m

76.1

1.Masonry-arch

53.9

supported track

56.3

supported track

53.9

61.8

2.New-Borough

56.3

62.8

2.New-Borough

57.3

Market Viad.

40

50

60

70

80

30

40

50

76.1 dBA 68.5 dBA

50

80

30

40

60

70

80

40

50

80

68.9 dBA 74.4 dBA

LAeq day

56.0 dBA

30

70

82.1 dBA

Total route 2

70.1 dBA

LAeq day

60

Total route 1

66.4 dBA

LAeq night

50

dBA 77.6 dBA

Total route 2

54.2 dBA 40

70

Total route 1

63.1 dBA

LAeq day

60

68.9

dBA

Total route 1

LAeq night

67.5

Market Viad.

66.4

dBA

Total route 2

63.1

2.New-Borough

63.9

Market Viad.

63.1

30

82.1

1.Masonry-arch

38.9 38.9

30

80.3

45B

77.6

1.Masonry-arch supported track

77.2

1.Bridge

70.0

45B

60.1 dBA

LAeq night 60

70

80

30

40

50

60

70

01-Jul-2005

80

Structure

Bridge45B b ow n trk.mat Lw roll std.mat new bv s50.mat

Wheel-rail Total

Fig. 4. The summary of noise sources significant at the Globe Public House at the reference design stage =================================== R5: Globe Public House =================================== Ground level, 1.6m

Track level, 11.6m

73.5

1.Bridge

1.Bridge

55.6

45B CS spur

34.3 37.2 39.0

2.Masonry-arch

supported track

supported track

52.3 51.2 54.8

2.Bridge 45B T'link lines

Market Viad.

30

40

74.9

50

34.0

70

80

30

40

dBA

40

50

60

54.4

3.New-Borough

62.3 63.0

Market Viad.

70

80

30

40

dBA

Total route 1 73.6 dBA Total route 2 54.9 dBA Total route 3 55.6 dBA LAeq day 63.2 dBA LAeq night42.2 dBA 30

50

72.2 72.2

45B T'link lines

54.1 58.7 60.0

Market Viad.

53.3

2.Bridge

62.1 62.6

3.New-Borough

55.3 55.3

supported track

53.1

2.Bridge

60

2.Masonry-arch

52.8 52.9

45B T'link lines

53.0 52.2 55.6

3.New-Borough

45B CS spur

34.1

2.Masonry-arch

74.7 77.1 79.1

1.Bridge

66.8

45B CS spur

73.6

14.8m

74.2

60

70

30

40

50

60

70

80

dBA

Total route 1 74.9 dBA Total route 2 63.0 dBA Total route 3 60.0 dBA LAeq day 64.9 dBA LAeq night 46.1 dBA 80

50

60

70

Total route 1 Total route 2 Total route 3 LAeq day LAeq night 80

30

79.1 dBA 72.3 dBA 63.0 dBA 69.5 dBA 51.9 dBA 40

50

60

70

80

14-Jun-2007 Bridge45B b typeA abs semijointed.mat new bv dec06.mat Bridge45B b dec06.mat new bv dec06.mat Bridge45B b own track jointed.mat

Structure W heel-rail Total

Fig. 5. The summary of noise sources significant at the Globe Public House with also feasible noise mitigation measures implemented

Minimising Noise from Viaducts in the Borough Area of London

185

5 Example Result Fig. 5. shows the results at the Globe for one of the scenarios being considered. Here Bridge 45B is split into two sources; the lines from Cannon Street Station are subject only to limited noise mitigation measures because of the presence of the crossover. The Thameslink lines have been considered with sleeper soffit pads (dynamic stiffness 10 MN/m per sleeper end), soft rail pads and rail dampers. The new Borough Market Viaduct is considered with ballasted track and also with sleeper soffit pads and rail dampers, the preferred solution. A reduction in the day-time LAeq of over 5 dB is seen compared with the reference designs. This is constrained by the limited treatment available on the Cannon Street lines.

6 Conclusion This project is a case example of the use of a detailed theoretical model within the framework of a complex railway project to achieve an overall reduction of noise in a specific environment. The scenario predictions have enabled the direction of appropriate spending on different structures to achieve a balanced and optimised control of noise for this mixed residential, high profile area of central London.

Acknowledgements The work reported was conducted by Temple Group Ltd for Network Rail as part of its Thameslink Programme.

References [1] Bewes, O.G.: The calculation of noise from railway bridges and viaducts, EngD Thesis, University of Southampton (2005) [2] Thameslink 2000, Environmental Impact, Main report, inner areas. Prepared for Network Rail by Temple Environmental Ltd (June 2004) [3] Hardy, A.E.J.: Draft proposal for noise measurement standard for ERRI committee C163. Report RR-SPS-97-012 published through ERRI [4] Thompson, D.J., Jones, C.J.C.: A review of the modelling of wheel/rail noise generation. Journal of Sound and Vibration 231, 519–536 (2000) [5] Bewes, O., Thompson, D.J., Jones, C.J.C.: Calculation of noise from railway bridges: The mobility of beams at high frequencies. Structural dynamics: Recent advances. In: Proceedings of the 8th International conference, Institute of Sound and Vibration Research, Southampton, paper 64, (CD ROM) July 14–16 (2003) [6] Janssens, M.H.A., Thompson, D.J.: A calculation model for noise from steel railway bridges. Journal of sound and vibvration 193, 295–305 (1996) [7] Bies, D.A., Hanson, C.H.: Engineering noise control, theory and practice, 3rd edn. Spon Press, London (2003)

The New German Prediction Model for Railway Noise “Schall 03 2006” – Potentials of the New Calculation Method for Noise Mitigation of Planned Rail Traffic U. Moehler1, M. Liepert1, U.J. Kurze2, and H. Onnich3 1

Moehler + Partner, Paul-Heyse-Str. 27, D-80336 München, Germany Tel.: +49 89 544217 0; Fax: +49 89 54421799 [email protected] 2 Müller-BBM, Robert-Koch-Str. 11, D-82152 Planegg bei München, Germany 3 Deutsche Bahn AG, DB Systemtechnik, TZF 12, Völckerstraße 5, D-80939 München, Germany

Summary The German prediction method for railway noise from new railway lines was revised by an expert team during the last five years. The draft issue of “Schall 03 2006”[1] is available now. The calculation model is based on octave-band sound power levels describing the emission in different heights of different vehicles, noise sources and parts of noise sources, e.g. roughness of wheels and rails, pantograph noise, and engine noise. The description of sound propagation follows the methods of ISO 9613-2 [2]. The new calculation method allows taking noise reductions into account that are based on technical progress or improvements. To this end, a scheme is described in “Schall 03 2006” [1] for introducing new measures of noise abatement and environmental protection. This includes measures for rolling stock, for rail and track, for bridges, and even measures in the propagation path. The effects of the parameters used in “Schall 03 2006” [1] on noise immission in general and on special existing noise abatement measures, e.g. low noise composite breaks and rail grinding, are presented.

1 Introduction A calculation method for Railway noise, including the noise of tramways and shunting yards, for new constructions of railway tracks was updated and revised; the new calculation method “Schall 03 2006” [1] shall be implemented in 2008. In comparison to the “Schall 03” this calculation method introduces octave band emission levels and three different heights of sound sources. The sound sources taken into account are distinguished by their mechanisms of appearance. The calculation method of sound propagation now considers buildings as noise barriers. The guide line is structured according to German and European standards and contains a comprehensive description of acoustic calculation methods for sound emission and sound propagation as well as a calculation of a rating level which shall be used for comparison with the applicable noise limits. Furthermore this directive includes a method to take into consideration new railway technology and innovations which will have effects on sound emissions. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 186–192, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

The New German Prediction Model for Railway Noise “Schall 03 2006”

187

The consequences of possible noise reduction measures will be shown after a short introduction of the calculation method. Finally the procedure of issuing new technologies will be illustrated.

2 Main Features of New “Schall 03” The calculation methods are detailed in „Schall 03 2006“[1] (see also [3]). In “Schall 03 2006” [1] seven types of powered vehicle units, three types of un-powered vehicle units and three types of trams are distinguished. For these types of vehicle units four types of sources consisting of a total of nine types of partial sources are distinguished. The parameters to be considered are listed in Tab. 1. Table 1. Parameters of sound emission, used in Schall 03 2006 ,

Type of vehicle high-speed traction unit high-speed coach high-speed train-set high-speed tilting tech. rapid transit train set electric locomotive diesel locomotive passenger coach freight wagon Low - floor tram high - floor tram Metro

Type of source rolling noise aerodynamic noise equipment noise propulsion noise

Partial sources rail roughness wheel roughness Structure-borne sound of tank wagons Pantograph grills of cooling systems Bogies Ventilators exhaust gas system Engine

The partial sources of each type of vehicle are assumed to be located and energetically summarised at three different heights: 0 m, 4 m and 5m above railhead. In figure 1 the different sound sources of an ICE1 traction unit are shown as an example. aerodynamic noise

equipment and aerodynamic noise 5,0 m 4,0 m

0,0 m rail rolling noise

influence of track type

wheel rolling noise

Fig. 1. Noise sources and corresponding heights of an ICE traction unit

188

U. Moehler et al.

Rolling noise is dominating under free-field conditions in the velocity range from 70 to 250 km/h. To keep the calculation procedure practicable, a minimum fictitious velocity of 70 km/h for railway vehicles and 50 km/h for trams is compulsory to account for all additional sound sources in and next to stations, e.g. brake squeal. Total noise data collected during the pass-by of about 10,000 trains at various tracks in Germany have been supplemented by special measurements involving phased microphone arrays and by enveloping measurements on powered vehicles at rest [4, 5, 6]. Rolling noise and aerodynamic noise were determined by evaluating the pass-by of different trains varying parameters like train type, track type etc. [4]. Other types of noise sources (such as equipment noise) were determined by measurements of vehicles at rest. The results were compared to the results from a theoretical model for noise emission, which is described by equation (1) and provides the basis for determining the sound emission from “Schall 03 2006” [1]: LW ' A, f , h ,m , Fz = a A,h ,m , Fz + Δa f ,h ,m , Fz + 10 lg

nQ nQ , 0

⎛v dB + b f ,h ,m lg ⎜⎜ Fz ⎝ v0

⎞ ⎟⎟ dB + ∑ c f ,h ,m + ∑ K ⎠

(1)

with

a A,h ,m, Fz

A-weighted overall level of the sound power per unit track length emitted from the height range h for the source type

m of one vehicle unit

Fz , equipped with the reference number nQ , 0 of sound sources, running at the reference velocity

Δa f , h , m, Fz

v 0 = 100 km/h, on a ballasted track with

average condition of the rail surface, neither on a bridge nor on a curved track, in dB level difference between overall level and octave band level in the octave band

f

, in dB

nQ

number of sound sources of a vehicle unit

b f ,h,m

velocity factor

v Fz Σ c f ,h, m ΣK

train velocity, in km/h level corrections for type of track and rail surface condition, in dB level corrections for bridges and particular nuisance of noise, in dB

The speed dependence of these factors is taken into account for rolling noise and aerodynamic noise. In the case of equipment and propulsion noise the speed dependence is disregarded, because no relevant influence of speed on the total noise is expected. The validity of equation (1) is documented in [4] in terms of small mean values and standard deviations for the differences between measured and calculated data. The acoustic parameters were summarized on data sheets for each of the ten types of vehicles (table 1). Correction terms for the track type or for bridges were determined by comparing the results of pass-by measurements on ballasted track to those on different track types or types of bridges ([4], [10]).Examples for the velocity dependence are plotted for complete trains in Fig. 2.

The New German Prediction Model for Railway Noise “Schall 03 2006”

189

70

noise level at 25 m distance LpAeq [dB(A)]

ICE 1 ICE 3 IC NV Gz

60

50 50

100

150

200

250

300

train speed v [km/h]

Fig. 2. Dependency of velocity on calculated noise levels LpAeq at 25 m distance for different train types with 1 train per hour according to “Schall 03 2006”

3 Effects of Noise Mitigation According to “Schall 03 2006” “Schall 03 2006“[1] allows to deduce noise reducing measures for particular brakes or for especially maintained tracks. In the following a description of the effect of disc brakes, cast iron block brakes and composite block brakes in co-action with rail grinding is shown: 70 Gz cast iron brake noise level at 25 m distance LpAeq [dB(A)]

Gz cast iron brake, rail grinding Gz comp. brake Gz comp. Brake, rail grinding ICE 1 ICE 1, rail grinding

60

50 50

100

150

200

250

300

train speed v [km/h]

Fig. 3. Calculated noise levels LpAeq for trains with cast iron brake and composite brake with and without rail grinding according to “Schall03 2006”

From figure 3 it can be deduced that e.g. the usage of composite block brakes instead of conventional cast iron block brakes for freight trains results in a reduction of about 4 dB(A) for an average maintained track. Using rail grinding in combination with cast iron block brakes allows a reduction of only approx. 1 dB (A), however in combination with disc brakes (ICE1) approx. 4 dB(A). Combining composite block brakes with rail grinding will result in a reduction of about 8 dB(A) compared to conventional cast iron block brakes.

190

U. Moehler et al.

From figure 3 can also be deduced the poorer performance of rail grinding for ICE 1 at higher speeds (above approx. 200 km/h) and at low speed (approx. 100 km/h). In these cases the influence of other sound sources, such as equipment noise at low speed and aerodynamic noise at high speed, can be seen. These effects are negligible in the case of freight trains, where rail grinding causes a constant decrease of noise. Figure 4 shows the influence of the kind of track and the possible measures to reduce noise level. 70

noise level at 25 m distance LpAeq [dB(A)]

ICE 1, ballasted track ICE 1, ballested track, railgrinding ICE 1, slab track ICE 1 slab track with absorber ICE 1 slab track with rail grinding 60

ICE 1 l b t

k ith b

b

d

il

50 50

100

150

200

250

300

train speed v [km/h]

Fig. 4. Calculated noise levels LpAeq for trains on different tracks according to “Schall03 2006”

From figure 4 it can be deduced that slab tracks with absorbers will have approximately the same acoustic properties as ballasted tracks. Comparing ballasted tracks and slab tracks (with or without absorbers) the effect of rail grinding is approximately the same. Slab tracks with absorbers and rail grinding cause the same noise levels as the ballasted track with rail grinding. The reducing effect of sound insulation walls for freight trains and ICE trains is shown separately in the following chart: Gz, v=100 km/h

noise reduction ΔLpAeq [dB(A)]

ICE1, v= 240 km/h

-5

-15 10

100

1000

distance [m]

Fig. 5. Difference in calculated noise levels LpAeq with and without noise barrier depending on the distance to the track according to “Schall 03 2006” (h=3,5 m, reference level 6,0 m)

The New German Prediction Model for Railway Noise “Schall 03 2006”

191

From figure 5 it can be deduced that the reducing effect of sound insulation walls is significantly higher for freight trains compared to high speed ICE trains. This effect can essentially be put down to the reduced effect on high positioned noise sources as aerodynamically caused noises by the pantographs. Experimental results of this effect can be seen in [5].

4 Procedure of Introduction of Innovations New railway technologies, which are not mentioned in spreadsheets and data sheets of “Schall 03 2006” [1], can acoustically evaluated according to a defined method. These railway technologies are including: • • • • • • •

Types of vehicles Components of vehicles Components of railroad shunting yards or container terminals Types of tracks Bridges Maintenance methods of tracks and wheels Railway specific sound insulations in the direction of sound propagation whose shielding effectiveness is not appropriately described in “Schall 03 2006” [1].

The acceptance procedure requires a formal application, the confirmation of the acoustic improvements by a defined measuring method subject to railway techniques mentioned above and passes the following steps: • • • • •

Application Certification of acoustic improvements Survey by authorised institution Certificate Publication

To confirm the acoustic improvements, the results of e.g. at least three measurement of pass-by noise testing and stationary noise measurement according to DIN EN ISO 3095 [7] and TSI [8] are required. Shielding devices and similar measures whose effects can not be calculated, are to be described in accordance to existing regulations. To verify the modifications measurement results are to be named as differences in the eight octavebands with centre frequencies of 63 Hz to 8 kHz to the calculated differences according to existing calculation methods. The authorised institution has to verify if the object of application differs from the directive. As a rule, a significant deviation is on hand if for a part of the source the deviation of the A-weighted overall level of the sound power is greater than 2 dB or in single octave-bands greater than 4 dB. The authorised institution has to issue a certificate for the subject of application if a significant deviation is on hand. This certificate allocates the subject of application to the existing spreadsheets and textual specifications of this guide line and describes the deviation of sonic effect.

192

U. Moehler et al.

5 Conclusion With the new calculation method of “Schall 03 2006” an up to date method for the calculation of noise immissions caused by railway lines is available. Noise emission is separated in different types of sound sources at different types. Each vehicle can be described by these sound sources with their specific contribution. Therefore noise abatement measures, which are implemented to a specific sound source, can be taken into account. The noise reducing effect can be taken into account by developing a new data sheet for the improved vehicle.

References [1] SCHALL 03 2006, Richtlinie zur Berechnung der Schallimmissionen von Eisenbahnen und Straßenbahnen (Draft, 21.12.2006) [2] DIN ISO 9613-2:1999 Dämpfung des Schalls bei der Ausbreitung im Freien In: Teil 2: Allgemeines Berechnungsverfahren [3] Möhler, U., Liepert, M., Kurze, U., Onnich, H.: The new German prediction model for railway noise SCHALL 03 2006. In: Some proposals for the harmonised calculation method in the EU directive on environmental noise Euronoise, Tampere, Finland (2006) [4] Kurze, U.J., Weißenberger, W.: Der Aufbau einer Datenbank als Grundlage für eine neue Schall 03 (Development of a data base for a new Schall 03), Müller-BBM Report No. 52253/9 for Deutsche Bahn (March 28, 2003) [5] Barsikow, B., Hellmig, M.: Bestimmung des Einfügungsdämpfung einer Schallschutzwand anhand von Messungen in derselben Ebene (Determination of the barrier insertion loss from measurements in the same cross section), Report fror Umweltbundesamt (2000) [6] Möhler + Partner: Schallmessungen zur Bestimmung der unterschiedlichen Wirkung absorbierender und reflektierender Schallschutzwände (Sound measurements for determining the different performance of absorptive and reflective noise barriers), Report No. 1011882 (February 2005) [7] DIN EN ISO 3095: Bahnanwendungen - Akustik - Messung der Geräuschemission von spurgebundenen Fahrzeugen (ISO 3095:2005) [8] TSI: Entscheidung der Kommission 2002/735/EG vom 30. Mai 2002 über die technische Spezifikation für die Interoperabilität des Teilsystems "Fahrzeuge" des transeuropäischen Hochgeschwindigkeitsbahnsystems nach Artikel 6 Absatz 1 der Richtlinie 96/48/EG (Bekannt gegeben unter Aktenzeichen K(2002) 1952), Amtsblatt L245 vom 12, S. 402 (September 2002) [9] Kurze, U.J., et al.: Outdoor sound propagation. In: Ver, I.L., Beranek, L.L. (eds.) Noise and Vibration Control Engineering, 2nd edn., ch. 5, Wiley, Hoboken (2006) [10] Stiebel, D., Behr, W., Brandl, W., Degen, K.G.: Silent Railway Bridges. In: CFA/DAGA 2004, pp. 971–972 (2004)

Floating Slab Track Re-engineering: Experience Drawn from a Completely Renovated FST Damaged by Major Flooding in Sao Paulo Metro P. Carels, K. Ophalffens, P. Pinto, and R. Kelly CDM, Reutenbeek 9, 3090 Overijse, Belgium [email protected]

Summary This paper describes the FST re-engineering of a Linha Leste track section in Sao Paulo Metro after major flooding in 2000 and the related experiences which have subsequently influenced novel design concepts. Due to major floods in the track section, excessive rail deflections and some fastening clip failures were noted, but while the floating slab track still provided the expected vibration isolation level the operational speed was reduced for safety reasons. The combination of the tunnel flooding and the pumping effect at rolling stock passage caused a migration of the elastic bearings under the slab track. After an intensive re-engineering program (including FEM simulations, material tests in harsh conditions, vibration and track stability measurements) a new FST system was installed in 2006 with easy to access and replaceable resilient strips, using the original concrete slabs. The work was carried out during the nighttime maintenance window resulting in no disruption of traffic. This re-engineering experience has allowed us to understand the complex behaviour of FST in tunnels and to devise solutions for improving future FST design.

1 Introduction In 1997 MSP (Metro Sao Paulo) was planning for the extension of Line 6 - Leste (Arthur Alvim – Guaianazes). In order to gain experience and understanding of vibration isolation mitigation techniques they decided to trial different types of FST solutions: 1) solutions with discrete bearings where high performances and maintainability of the support systems were required and 2) solutions with full surface mats in areas with requirements for less vibration isolation performance. MSP had already built up experience in track vibration isolation systems, for example in the Paulista Line under the MASP building, where the structural borne noise and vibration levels in the auditorium room caused by the metro passage was above the acceptable limits. As a “correction” intervention MSP introduced in the existing direct fixation track system (Landis) a softer base plate pad in order to obtain a track system with lower stiffness. As a result, the vibration levels were reduced in MASP auditorium room to acceptable levels, but the behaviour of the track created many problems, especially with the track gauge [1]. Based upon the studies to determine the vibration transmissibility between the tunnel and the neighbouring buildings for Linha Leste extension and the acceptable values for noise and vibration, there was a strong B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 193–200, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

194

P. Carels et al.

request to implement high performance FST systems. MSP also requested an FST system where it was possible to inspect and easily replace the resilient support without disrupting the metro operation. Focused on those requirements CDM provided solutions at 2 intervention levels, i.e. solutions using full surface mats (CDM-DFMA systems) in less critical areas; and solutions using discrete bearings with a FST system designed for MSP, based on discrete elastomer bearings (CDM DFPA).

2 CDM FST - Initial Solution CDM made proposals for MSP Linha 6 - Leste extension an FST system, according to the requirements defined by the Project Department of MST, following three main design parameters - Track stability - Vibration isolation performance - Track maintainability. The CDM DFPA solution was based on concrete slabs, each supported by 128 bearings CDM-81 70x70x50mm, as already applied successfully in Metro Lines (Antwerp-BE) and in Tramway lines (Nantes-FR) [2]. The working conditions were: Dead load/slab = 130 kN; Live load/slab (2 axles) = originally 380 kN but after operation modified to max. 420 kN; Load window for the bearings: 0,21 MPa<σ<0,88 MPa; Design K-stat-total/slab = 47,4 MN/m or 3 MN/m³ static bedding module (with distribution as indicated); Design K-dyn-total/slab = 71,7 MN/m (with distribution as indicated). Because of the replaceability requirement the joint between slabs didn’t include any shear beam (shear load transfer to be taken by the rail). The FST design took into consideration reinforcement of this area by adding a bearing with a higher stiffness under the joint between 2 slabs. Between FST and non-FST zones a transition zone is defined where total track stiffness gradually varied from 4xK to 2xK and finally to the standard FST stiffness K. This was realised by increasing the number of bearings over 12,5m track length (2 slabs) further to UIC guidelines for transition stiffness length of Vmax (m/s) x 0,5 sec; if track stiffness increase is > 2. The FST system was designed to f-res < 10 Hz (loaded) in order to reach an IL around 30–35 dB (1/3 octave 63 Hz). MSP defined 3mm as acceptable limit for deflection of the system. This was incompatible with the requirement for an IL 30–35 dB and at the time of the decision MSP considered the acoustic performance the more important of the two criteria. A small scale CDM-DFPA prototype was approved and considered fit for use after a test program at UNICAMP [3].

3 First Experiences /Comparison of Solutions The MSP extension Leste with different FST solutions became operational in August 2000. On site vibration measurements were conducted by UNICAMP and showed good results (report nr. 305-05CT2001). In the meantime MSP however reported some concerns on the vertical deflections being too high (9 mm). It is important to understand though that the actual axle load was higher than the load stated during the original design phase (210 kN vs. 190 kN).

Floating Slab Track Re-engineering

195

4 Flooding Hazard In December 2000, torrential thunderstorms hit the Sao Paulo region and flooded a large part of the MSP Linha Leste FST section with CDM-DFPA. The trains continued their service and the FST started showing erratic deflection patterns with abnormally high slab deflections, ripped off lateral isolation strips and an unexpected high amount of eclip breaks, particularly around the slab joints. Because of safety concerns MSP immediately imposed a speed restriction from 90 kph to 20 kph in the affected area. MSP reported the incident to CDM early January 2001 requesting an in-depth study, explanation and action plan. The first step was to inspect the area below the slab. Since the original FST design specifications imposed easy intervention for maintenance, inspection and replacement, this area could be inspected by a pre-established procedure during the night-time maintenance window.

5 Measurements Required to Ascertain Current Status Based upon these first experiences MSP asked for an indepth technical evaluation of the existing track compliance at different operation speeds. This evaluation process ran from 09/2001 until 06/2002 under the expert guidance of TU Munich [10]. All measurements were conducted by IPT. The typical clip fracture pattern indicated that the clip failed due to excessive lateral movement rather than vertical (internal communication CDM/Pandrol 2004, see Fig. 1).

Fig. 1. Clip break in FST

6 FE-Model Parallel to the in-situ measurements a FE-model of the existing situation was made [4]. The para-metric 2D FE-model has been composed with “x” in the longside of the track and “y” in the vertical direction. For each node 3 degrees of freedom are foreseen, being displacement in x and y and rotation around z axis. The model uses 10 concrete plates. Each plate is supported by 16 isolation blocks in the longitudinal direction, so 15 sections can be considered. Each section is divided into 5 elements. The model has 5040 degrees of freedom with 1838 beam and springdamper elements. A detail of the model is shown in Fig. 2. FE Model Fig. 2. The model

196

P. Carels et al.

was validated by the measurements and the study of differ-rent intervention strategies by looking into an integrated approach, i.e. in the existing condition keeping the track stable within acceptable vibration limits (Lv < 70 dBV, ref. 25,4 E-6 mm/s). For more in-depth in-formation of the description and imple-menttation of the FE model, the reader is referred to [4].

7 What Caused the FST Failure? The TUM report [10] concluded in general that the present FST design was not critical under normal conditions but that increased inspection was required specifically for speeds exceeding 50 kph. The flooding was a new element that brought the existing FST into non safe operating conditions. As reported earlier, the failure of the original FST solution became visual by abnormal high deflections and unexpected high amount of e-clip breaks around the concrete slab joints in the area affected by the flooding. Originally MSP showed concerns regarding the bearing quality and asked for inspection of their mechanical behaviour. For that reason some bearings were recuperated from below the FST and compared with the original expected mechanical characteristics. The elastomer bearing recuperated from sector 17 showed a K-stat of 0,44 MN/m in the typical load window (0,22-0,88 MPa). A new bearing was tested in the same conditions, which resulted in a K-stat of 0,41 MN/m. The original design stiffness of the AV mount was 0,37 MN/m. It was concluded that the bearings showed no significant degradation, the basic elastomer material remained intact and performing to acceptable standards; therefore material degradation was not the cause for track malfunctioning. After in depth study of the visual inspection findings it was concluded that increased hydraulic pressure and water pumping (generated by the high FST deflection of the completely flooded slab, with water drainage not having been sufficiently provided under the sealed off FST) caused water to penetrate the contact area between the glued bearings and the lost formwork supporting the slab, resulting in the lateral isolation being ejected from its position. As the glue joint deteriorated between bearing and supporting surface it caused the bearings to migrate under the slab, with some bearings slipping into the reservation left by the adjustment panels on top of the lost formwork. Once the bearings were trapped in this free space, they lost their load bearing function thus altering the slab load characteristics; causing an erratic pattern of slab deflection. Since the original design introduced separate concrete slabs with no shear connection between them and 2x higher local stiffness bearings concentrated in the joint areas (bearings supporting the joint between 2 consecutive slabs, a proven technique in urban tramway surface track, see e.g. Tramway Nantes in 1997-1998), the shear force and bending moment are being borne by the rail. Because of this, there is a differential deflection at the joint between consecutive slabs. The technique of using the rail as “shear beam” in separate FST slabs was not new; see e.g. double tie FST – Toronto TTC (TTC Engineering department, 1982), San Francisco BART and Hong Kong KCRC-West Rail [5] and 3 tie systems like Milano Massivo [6]. During rolling stock passage the double axle bogie (with normal axle distances ranging from 1,6 to 2,1m) is either fully on the slab or with 1 axle on the edge of the slab, the other axle already being on the next slab (causing as such a “pitching” effect of the slab) with the rail

Floating Slab Track Re-engineering

197

acting in both cases as stabilising element. Because of the differential movements of rail supports, the loads are transmitted via the fasteners. In other words, flooding caused an erratic deflection pattern that increased the already high differential deflection at the joint between consecutive slabs.

8 Solutions Researched

Vrms-global

In the meantime MSP had initiated temporary repair works in the areas most affected by the flooding in order to warranty safe operation. Since the FST design allowed access to the AV mounts, different solutions were installed and could be compared to the original one. This provided the opportunity to compare 3 different stiffness set-ups from a track stability and vibration isolaSetor 10A tion (vibration measurements Vrms-global vs.K-stat Setor 17 PUR Vrms = 65.01 + 0.03*K-stat in different locati-ons directly Setor 17 original 90 above the tunnel) point of 80 70 view. Sector 17WA with very 60 stiff PU pads 45 MN/m, total 50 K-stat 360 MN/m; Sector 40 30 10FA repaired with bearings K-stat (v-rms=70 dBV) = 157 20 CDM-81050 135x135mm, 10 0 total K-stat 55 MN/m and fi0 100 200 300 400 nally the original solution K-stat slab Sector 17DA. Fig. 3 shows the N&V measu-rements in Fig. 3. Vibration level in function of the static support houses neighbouring the 3 stiffness areas. The vibration threshold set out by MSP was 70 dBV, corresponding to a K-stat of 167 MN/m (see Fig. 3). It is interesting to see this threshold is valid because during the measurements only vibration levels above 74 dBV (= 0,127 mm/sec) gave rise to complaints (i.e. pts. in 17WA).

9 Final Solution Design Strategy Between 2003 and 2004 the FE-Model [4] was used to fine-tune solutions of this type; many designs were presented and discussed. By using N&V measurements from the different temporary repair set-ups and the FE-model validation, a new and final strategy was put forward. Although the original FST solution with small AV mounts behaves acceptably in a wide dynamic load window (0,21-0,88 MPa or 5-21% of compressive strain) for present load conditions (rolling stock axle load and present inertial loads) which cannot be changed, the present “normal” system deflection of approx. 9mm was judged to be too high because of increased probability for fatigue clip breaks. It was therefore suggested to increase the stiffness until a point still acceptable for N&V mitigation. The original FST performance, i.e. mean insertion loss (50-80Hz) was >35 dBV; the requested min. IL being 25 dBV. The proposal was to increase the stiffness by a factor 2, which would reduce the IL by approx. 6-8 dBV, still keeping the IL within the contractual range. Various solutions could have been proposed to reach this stiffness

198

P. Carels et al.

increase allowing for a quick installation; however, one had to take into account that the support system would need to perform in immersed conditions without jeopardising water drainage. In order to have an idea where the stiffness may be increased to avoid vibration problems, the influence of v-rms in function of the K-stat/slab of the different set-ups was checked (see Fig. 3). By setting out the points at a more or less same radial distance D (54m 80 MN/m → lift-up forces ≈ 2500 N. The final solution was defined based upon the vibration comfort requirements, which gave rise to a max. allowable slab stiffness and the static and dynamic stability requirements of the floating slab track (clip failure safety, track deflection

Fig. 4. Final FST design using CDM-RR strips

Floating Slab Track Re-engineering

199

minimisation, avoidance of shear beam connectors to keep easy access and safeguard maintainability), which gave rise to a minimal needed stiffness per slab. A solution has been developed using CDM-43 material, a homologated material with proven track record in under ballast mats and under sleeper pad applications (CDM-RR material family; resin bonded rubber base material with good alkaline and acid resistance, excellent for tunnel conditions). The system type is called CDMDFSA-M9-MSP. The strips proposed in this solution are equipped with an integrated lateral buffer, manufactured in 1 single piece without gluing, which is positioned between the lateral wall and the slab. This lateral buffer serves 3 purposes: 1) fixing the strip in place and avoiding migration under abnormal conditions, 2) providing additional lateral stiffness reducing lateral movement, 3) serving as location device to help quick and correct positioning of the strip. The shape of the strips (L-shape in plan view, see Fig. 4) is such that, at the joint between slabs, stiffness is locally increased in order to reduce the differential movement between the slabs. Each slab is supported by 4 L-shaped strips and 4 straight strips all in 1 single layer of 50mm. With the known load conditions and a total support surface per slab of 3,72 m², the work window for the strips becomes 0,04-0,15 MPa. The static stiffness of the resilient strips is 28 MN/m³, which for 1 slab element comes to a total static stiffness of 104 MN/m (with adapted distribution, see Fig. 4). This is approximately 2 times stiffer than the original solution; which was the objective of the new design. In order to homologate this new design and new AV strips, MSP requested a complete series of tests. The resilient strips, type CDM-43, were tested by IPT [7] further to a pre-defined program. K-stat and K-dyn were measured before and after a fatigue test, a creep test and an immersion in water. All tests were satisfying (< 10% stiffness variation) resulting in an approval of the new type of strips. CDM then supplied the material to replace the existing discrete bearing system by the newly approved strips in sector 10. In this test zone IEME performed N&V measurements [8] which proved all registered levels were within the acceptable ranges. Based on both the IPT and the IEME reports, MSP has eventually approved and officially homologated the CDMDFSA-M9-MSP FST system. At the writing of this paper (April 2007), all remaining CDM-DFPA areas, are being replaced by the CDM-DFSA system.

10 Conclusions An FST with discontinuous concrete slabs and supported by AV bearings installed in MSP failed after flooding. After extensive research a new solution was proposed, tested and eventually homologated by MSP. The re-engineering work, discussed in this paper, allowed gaining valuable experiences for future FST design, where more emphasis will be put on new aspects such as maintainability and accessibility. Some important guidelines for future FST design: ¾

Although discontinuous FST systems offer solutions for easy inspection and maintenance and limits the standing wave propagation (reduction of radiated noise), they impose additional mechanical constraints in the rail fastening system (not only vertical, but also lateral, pitching, etc. and in general more deformation modes may occur). Where discontinuous FST systems are desired by the permanent way owner, the direct fastening plates on the slab should be avoided and use of booted sleeper with soft interface and classic rail fastening is to be preferred.

200

¾

¾

P. Carels et al.

In order to make a track fastening system behave in a homogeneous and solid way, continuous FST systems are to be preferred. They can also be made inspectable, by concentrating the resilient support system on the FST edge for easy access from either side (reservation for jacks). All FST system designs in tunnels should take the presence of water in all its possible forms (flooding water, infiltration water with certain degree of acidity or alkalinity under influence of e.g. stray current ionization of local chemical environment, …) into consideration.

Note: This paper is a reduced version of a technical case study with in-depth description of all steps of the FST re-engineering process. The full version technical case study is available upon request.

References [1] de Sousa, A.J.C., et al.: Método para estudo preliminar do impacto de vibrações oriundos de linhas metroviárias. Article Engenharia (2004) [2] Carels, P.: Low vibration & noise track systems with tunable properties for modern LRT/street car track on surface in urban areas. In: Proceedings of IWRN 7, Portland (2001) [3] UNICAMP, Ensaios estáticos e dinâmicos no sistema de fixação direta com massa mola cdm em apoios discretos, bem como seus respectivos conjuntos de fixações e componentes. Relatorio Tecnico CT DESF Nº 013/98(1998) [4] dba-consult, Simulation floating subway track Sao Paulo. Report 35-96560-390 r11 (2005) [5] Crockett, A., Pyke, J.: Viaduct design for minimization of direct and structure radiated train noise. In: Proceedings of IWRN 6 Ile des Embiez (1998) [6] Dott. Ing. Marco Acquati, Dott. Ing. Bruno Cavagna, Geom. Luigi Di Ilario, Dott. Ing. Sergio Vigano, Armamenti antivibranti per metropolitane e tranvie. Ingegneria Ferroviaria 07/2000 (2000) [7] IPT, Ensaios laboratoriais em mantas elastomericas do tipo CDM-DFSA-M9. Relatorio tecnico N° 90121-205 (2006) [8] IEME Brasil, Monitoração das vibrações e ruídos secundários em residencies lindeiras à extensão leste do Metrô de São Paulo (2006) [9] IPT, Mediçao de deslocamentos e de esforços nas fixaçoes e detecçao de passagem de roda na extensao leste – linha E, setor 10 da CPTM / Trecho em massa mola CDM. Relatorio tecnico N° 91 349-205 (2a via) (2007) [10] Technische Universität München, Behaviour of CDM-DFPA Floating Slabtrack, MSP – Linha Leste. Research report N° 1952 (2002)

In-Car Noise and Carriage Floor Vibration on Different Track Forms and Curvatures in a Metro System A. Wang1, S.J. Cox1, H. Huang2, L. Liu2, J. Jiang3, and J. Sun3 1

Pandrol Limited, 63 Station Road, Addlestone KT15 2AR, England Tel.: +44 1932 834514; Fax: +44 1932 850858 [email protected] 2 Guangzhou Metro Co., 8 Huadidadao Nan Road, Liwan, Guangzhou 510380, P.R. of China 3 Beijing Metro Co., A2 Baiwanzhuang Street, Xicheng District, Beijing 100037, P.R. of China

Summary In-car noise and carriage floor vibration measurements were carried out on Guangzhou Metro (GZM) Line 1 in January 2006 to assess the effect of different track forms on noise and vibration levels within the carriage. The track forms tested included two different types of baseplate fastenings directly fixed to a rigid base slab (DFF), and one type of floating slab track (FST) systems. The measurements showed that the in-car noise and the carriage floor vibration levels were higher on curved sections of track than that on the straight. Comparing different track forms, there was no significant difference between the standard GZM baseplate track, and a PANDROL VANGUARD baseplate. However, the in-car noise level and the carriage floor vibration on the FST sections were much higher. The carriage floor vibration levels are generally higher in the vertical direction than in the lateral direction on both curved and straight track.

1 Introduction Guangzhou Metro has several installations of different track forms, each extending over signify-cant lengths. This paper describes in-car noise and vibration measurements made at two locations on Line 1 in January 2006. The directly fastened ‘DFF’ site is between Changshoulu and Huangsha stations and consists of two curves, each approximately 200 m each in length, with tangent track between. The track has two different fastening systems – the majority is fitted with standard GZM baseplates, but there is a 192 m long trial section of the PANDROL VANGUARD fastening on the southbound track [1, 2]. These low-stiffness vibration-control rail fastening baseplates were installed in January 2005. The other, ‘FST’ site, is between Tiyuzhongxin and Tiyuxilu stations, and has with one long curve on which there is an installation of FST that is about 291 m long. There is track directly fastened with GZM baseplates at either end of this on the same curve. In-car noise and carriage floor vibration measurements were made near the centre of trains running between both pairs of stations on both the southbound and northbound tracks. The noise level in the carriage was measured at 1.5 m above the floor. The carriage floor accelerations were measured in both vertical and lateral directions at the centre of the carriage. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 201–207, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

202

A. Wang et al.

2 Test Locations and Measurements The measurements were made on the Line 1 of the Guangzhou Metro system in China. This line runs in a broadly northeast to southwest direction, as shown in Figure 1. At the ‘DFF’ site, the PANDROL VANGUARD baseplates were installed on the southbound track between kilometre posts ZDIK5+887 and ZDIK5+695 in the tunnel between the Changshoulu and Huangsha stations. At the ‘FST’ site, a heavy mass-spring FST has been constructed on both tracks between kilometre posts DIK15+680 and DIK15+971 in the tunnel between the Tiyuxilu and Tiyuzhongxin stations. The GZM baseplates used to fix the rail to the FST are also used at either end of the same curve to fix the rail directly to a rigid base slab. The traffic is 6 car EMUs, and the maximum traffic frequency was 15 trains per hour during peak hour operation when measurements were made. The normal track speed was about 70 km/h at the DFF site and 60 km/h at the FST site. The measurements were taken in the third car of each train. Figure 2 shows the measurement positions. All recordings were made under normal service passenger traffic (with an axle load of approximately 16 tonnes).

FST

VANGUARD

Fig. 1. Guangzhou Metro System Map

In-Car Noise and Carriage Floor Vibration

203

M1 – Microphone; AC1 – Accelerometer in the vertical direction; AC2 – Accelerometer in the lateral direction Fig. 2. Measurement Positions

Floor vibration measurements were made in both vertical and lateral directions at the centre of the carriage. All test equipment was calibrated before the data recordings. A PCB seismic ICP type 393B12 accelerometer was used with PCB480E09 conditioning amplifier. For noise measurements a B&K type 4189 microphone was used with a B&K 5935L dual microphone power supply. The measurement system was calibrated with a B&K type 4231 sound calibrator. The microphone was fitted with a spherical windscreen during noise recordings. The in-car noise and vibration recordings were made with a laptop PC with a National Instruments 6036E DAQ-Card (16-channel, 16bit, 200k samples/s) running at a sample rate of 24 kHz. The recordings and analysis were controlled with custom Pandrol data logging software. Vibration records were analysed to give frequency spectra. One-third octave band spectra were obtained, and the linear and A-weighted levels calculated. The vibration results presented here are of average vibration levels for 3 recordings made over each of the specified sections of track.

3 Measurement Results and Discussion For the DFF site, noise and vibration recordings were made on both the southbound (from Changshoulu to Huangsha) and northbound trains. Four sections of the track of similar lengths were evaluated, the section T1 with PANDROL VANGUARD installed, a second straight section of track with GZM fasteners T2 and the two curved sections of track, C1 and C2, both also fitted with GZM fasteners. For the FST site, noise and vibration recordings were only made on northbound (from Tiyuxilu to Tiyuzhongxin) trains. Three sections of the track were evaluated, the middle part of the curved section with FST and GZM standard track fastenings C4, and two curved sections with a rigid track base and standard GZM track fastenings C3 and C5.

204

A. Wang et al.

The carriage floor vibration measurements contained some external interference, mainly from passengers walking along the train. When these were eliminated, the vibration levels were found to be the highest in the frequency range up to 1 kHz. The total vibration levels have been calculated for a frequency range from 20 Hz to 1 kHz. Since the concern here is airborne noise, the velocity values are A-weighted. Table 1 shows the average vertical and lateral velocity levels of the carriage floor for six sections C1 to C4 plus T1 and T2. Sections C3 and C5 were found to give very similar results. The noise measurement also contained some external interference, mainly from passengers talking and announcements being made during the recordings. The average A-weighted noise levels for the six sections of track for the frequency range 20 Hz to 2500 Hz are also given in Table 1. Table 1. Carriage floor vertical and lateral velocity and in-car noise levels

Carriage floor velocity dB(A) (ref. 5E-8m/s) Vertical Lateral Train direction South North South North Curve C1 95.5 91.4 84.1 81.4 88.5 77.4 * 79.0 DFF Tangent T1* 89.8 * 70 km/h Tangent T2 85.2 86.6 74.2 78.6 Curve C2 93.2 92.4 83.5 83.3 Curve C3 80.9 75.7 FST 60 km/h Curve C4** 93.9 ** 87.2 ** Location / Section

In-car noise dB(A) (ref 20E-6 Pa) South North 89.0 82.4 81.3* 81.5 77.2 79.0 81.2 83.1 79.8 86.4 **

* DFF track fitted with PANDROL VANGUARD fastenings. ** FST track fitted with GZM fastenings.

Comparing the two baseplates used on direct fixation track, there is generally little difference between the in-car noise levels or the floor vibration levels on the standard GZM track and the PANDROL VANGUARD track. Noise levels on curves are generally significantly higher, about 3-4 dB(A), than on tangent track. Vertical vibration levels on the carriage floor are generally much greater than lateral vibration levels on both tangent and curved track. Both vertical and lateral floor vibration levels increase on curves compared to tangent track. In-car noise levels generally increase and reduce with carriage floor vibration levels. However, it should not be inferred that in-car noise is always dominated by radiation from vibration of the car body itself. There may also be significant higher frequency components of in-car noise originating from the wheels and rail, and lower frequency components from the track base and tunnel walls. Some indication of this may be seen from the fact that the in-car noise and floor vibration levels are much greater on the FST than on the DFF track sections. This is also clear from Figure 3, which shows time histories for northbound trains passing over the FST site. There is a marked and drama-tic increase in both floor vibration and in-car noise that exactly corresponds to the train running on to and off the C4 section of FST track on curve. The carriage floor vibration levels are about 12-13 dB(A) higher than those on the directly fastened GZM standard track on curve sections C3 and C5. The in-car noise is about 7 dB(A) higher. This is likely to be the result of the high level of track slab vibration [3] on the FST track and the low frequency rumbling noise radiated from it [4].

Acceleration dB (ref 1e-6 m/s2)

In-Car Noise and Carriage Floor Vibration

205

150 140 130 120 110 100 0

10

20

30

Time (s)

Vertical (AC1) - Northbound

40

50

60

70

Lateral (AC2) - Northbound

(a) Carriage floor acceleration dB Tiyuxilu

Tiyuzhongxin Train running direction (712m between two stations including 580m curved track and two 60m transitions near the stations, curvature R=320m)

120 SPL dB (ref 2e-5 Pa)

Floating Slab Track Section 250 m C3

110

50 m C5

291 m C4

100 90 80 70 0

10

20

30

40

50

60

70

Time (s)

(b) In-car noise dB Fig. 3. Vibration and Noise Measurements at FST site

Returning to the DFF site, another significant feature is the particularly high peak in both floor vibration and in-car noise on curve C1. Figure 4 shows the averaged vibration and noise measurements in both northbound and southbound directions at this site. To make a comparison with the southbound trains easier, the time scale for the northbound trains has been reversed. The two peaks on the graph represent curves C1 and C2 respectively. The reason for this peak on curve C1 was investigated. The rail condition was inspected, and it was found that there was a heavy wear pattern on the rail on this section on the southbound track. Figure 5 shows the rail condition. The rail fasteners here are the standard GZM type. There is evidence of a relatively long-pitch corrugation, with a wavelength of approximately 0.4 m. It is very probable that the high levels of vehicle floor vibration and in-car noise on this curve are caused by this long wavelength corrugation on the track. Indeed it can be surmised that the generally increased levels of in-car noise and vibration observed on all curves are always associated with some degree of increased rail roughness on curves as compared to tangent track. This in turn may be related to the increased slippage and wear in the wheel-rail interface that is likely to occur on curved track [5]. This leads to higher vertical vibrations. Lateral vibrations on the carriage floor, on the other hand, seem to be less closely related to rail corrugation – as might be expected.

Acceleration Level dB (ref 1e-6 m/s2)

206

A. Wang et al.

150 140 130 120 110 100 0

10

20

Vertical (AC1) - Southbound

30 Time (s)

Lateral (AC2) - Southbound

40 Vertical (AC1) - Northbound

50

60

Lateral (AC2) - Northbound

(a) Carriage floor acceleration dB Changshoulu 120

SPL dB (ref 2e-5 Pa)

Curvature R=400m

VANGUARD Section (Southbound only) Curvature R=300m 200 m C1

110

208 m T2

192 m T1

Huangsha

200 m C2 Track Profile

100 90 80 70 0

10

20

30

40

50

60

Time (s) Southbound

Northbound

(b) In-car noise dB Fig. 4. Vibration and Noise Measurements at DFF site

Fig. 5. Rail Corrugation on the Southbound Track with GZM Rail Fastening

Note that there is some ‘leakage’ of the higher in-car noise levels that are associated with curved track into the time period during which the measurement position on the train is traversing along the tangent track. This is because the train is approximately

In-Car Noise and Carriage Floor Vibration

207

140 m long, so that at a time the carriage in which measurements are being made is on tangent track, the back or front of the train may still be on curved track. The noise in the carriage is of a “line source” type rather than a “point source”, with contributions from further along the train in each direction. This is more relevant in the case of noise, which may reverberate through the system for a longer time than is the case for the vibration. This leads to some ‘blurring’ at the ends of the curves in the time histories shown in Figures 3 and 4. Noting that the time axis has been reversed for northbound trains in Figure 4, the figures also indicate that there may be a small time lag before the highest in-car noise level is established as trains run through a curve, with the peak reached only towards the end of the curve. In order to avoid pollution of the results for the tangent sections of track from the higher noise levels on curves, the overall average levels report in Table 1 are for 10 second long tranches of the recordings for the curves, but for shorter central 8 second long sections for the straight track. The time to traverse a 200m long section at 70 km/h is approximately 10 seconds, so the shorter sections on tangent correspond to about 160m of track.

4 Conclusions • • • •

• •

In-car noise and vertical floor vibration levels are higher on curved sections of track than on straight sections of track. In-car noise and vertical floor vibration levels are at a similar level for directly fastened track with PANDROL VANGUARD and with standard GZM baseplates. In-car noise and vertical floor vibration levels are higher on curved sections of track with FST fitted than on curved sections of track that are directly fastened. In-car noise and vertical floor vibration levels on a curved section of the southbound track near Changshoulu station are higher on the northbound track, and this difference is likely to be caused by a long wavelength rail corrugation on the rail. Both tracks are directly fastened and are fitted with standard GZM baseplates. Vertical floor vibration levels are higher on both curved sections and straight sections of track than lateral floor vibration levels. Lateral floor vibration levels are higher on curved sections of track than on straight sections of track.

References [1] Methodology for the installation and track testing of PANDROL VANGUARD fasteners on Guangzhou Metro Line 1 trial, Pandrol Report No. 85171-21 (November 2004) [2] In-Car noise and vibration measurements on Guangzhou Metro Line 1 fitted with PANDROL VANGUARD system, Pandrol Report No. 85171-34 (April 2006) [3] Wang, A., Cox, S.J., Liu, L., Huang, H., Chan, S.: Ground vibration control using PANDROL VANGUARD in a tunnel on Guangzhou Metro Line 1. In: ISEV 2005, Japan (September 2005) [4] Crockett, A.R., Pyle, J.: Viaduct design for minimization of direct and structure radiated train noise. In: 6th IWRN, France (November 1998) [5] Thompson, D.J., Jones, C.J.C.: A review of the modeling of wheel / rail noise generation. In: 6th IWRN, France (November 1998)

Experimental and Theoretical Analysis of Railway Bridge Noise Reduction Using Resilient Rail Fasteners in Burgdorf, Switzerland K.P. Köstli1, C.J.C. Jones2, and D.J. Thompson2 1

Swiss Federal Railways SBB, I-FW-PS, Schanzenstr. 5 CH-3000 Bern 65, Switzerland Tel.: +41 (0)51 220 4699; Fax: +41 (0)51 220 5014 [email protected] 2 University of Southampton, ISVR, Highfield Southampton, SO17 1BJ, Southampton, UK Tel.: +44 (0) 2380 593224; Fax: +44 (0) 2380 593190 [email protected]

Summary The increased noise level as trains travel over bridges is, in many situations, a source of disturbance for nearby residents. As well as the rolling noise radiated by the wheel and track, the vibration generated at the wheel-rail interface also propagates into the bridge structure and the vibration response of the components of the bridge is an important extra source of noise compared with tracks at-grade. Vibration isolation of the bridge structure from the rail is therefore used to reduce noise. This often takes the form of resilient rail fasteners. Two different elastic rail fastenings were therefore tested on a twin track bridge by the Swiss Railways (SBB). The bridge over the river Emme at Burgdorf, is a ballastless steel bridge with timbers between the rail fastener and the bridge. Hanging steel sleepers have been added between the wooden sleepers on which the track is supported to form a continuous deck under the track. To find the best elasticity for the rail fasteners, predictions of the bridge noise were made using the Norbert model. Measurements were made on the bridge with the track in its original state to provide parameters for the model. These included rail and sleeper vibration as well as pass-by noise from service passenger and freight trains at different speeds. For the two tracks, elastic rail fasteners from two suppliers were installed. The measurement after installation showed a clear noise reduction for the frequency range from 80 to 400 Hz of about 10 dB. However the reduction in A-weighted overall noise level is in the range of 2 to 4 dB, as indicated by the model. The results show similar reduction for both systems.

1 Introduction The main reason for the increase in noise as trains cross a bridge is the vibration of the structure. Here the rolling noise radiated by the track and bridge is studied for the twin track bridge over the River Emme at Burgdorf, Switzerland. Two photographs of the 57m long bridge are shown in Fig 1. The track on the bridge takes an unusual B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 208–214, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Experimental and Theoretical Analysis of Railway Bridge Noise Reduction

209

form in that steel sleepers have been added between the wooden sleepers on which the track is supported in order to form a continuous deck. The steel sleepers are hung from the rail and are not otherwise supported by the bridge structure. Resilient rail fasteners were installed and the steel sleepers replaced with wooden sleepers in October 2006. The reconstruction was accompanied with noise measurements in May, October, December 06 and January 07. One track was equipped with resilient fasteners from Pandrol, the other, from Vossloh. The elasticity of the fasteners in both cases is about 20MN/m. To identify the effect of the hanging steel sleepers, additional noise measurements were carried out on one track after installing the resilient fasteners but before replacing the steel sleepers. To find the best elasticity for the rail fasteners, predictions of the bridge noise were made using the ISVR software Norbert (Noise of Railway Bridges and Elevated Structures) [1]. Measurements were made on the track in its original state to provide parameters for the model.

Fig. 1. Left: The Burgdorf bridge and track. Right: Arrangement of wooden bearing sleepers and suspended sleepers.

2 Modelling ISVR has developed a bridge noise prediction model called Norbert based on the combination of an analytical model of the track and a statistical energy analysis method (SEA) for the bridge [2, 3]. In this method, the bridge structure is described using a single subsystem type; plates in bending. It has been found sufficient to use a simplified version of SEA which assumes equipartition of energy. In this assumption the energy is distributed to each of a number of subsystems according to their resonant energy capacity resulting in an estimate of the mean squared vibration velocity on each susbsystem. For parts of the structure such as the wooden walkways to the side of the tracks on the Burgdorf Bridge, additional networks of subsystems can be added to the model that are excited by the velocity of particular susbsystems of the main network. The bridge bearings and piers are not important for the noise radiation. For the track, the rail support stiffness is an important parameter. The combined roughness of the wheel and rail running surfaces excites the wheel and rail into vibration according to their dynamic properties. Models for the rolling

210

K.P. Köstli, C.J.C. Jones, and D.J. Thompson

noise emitted by the wheel and track are well established [4, 5]. The way in which the track is assumed to transmit energy to the SEA model of the bridge is an important aspect of the calculation. For the current bridge, a model of the rail coupled via a continuous spring-mass-spring support to the longitudinal beam of the bridge has been used up to the rail-on-sleeper resonance frequency. Above this frequency the force from the track is assumed to drive the local input mobility of the longitudinal beam. The latter is based on an approximate formula for an I-section beam [6]. The sound power is calculated from the vibration velocity of each component via simply calculated radiation ratios for each plate subsystem. Simple propagation calculations are then used to estimate the sound pressure level at particular receiver locations. The principles and results of the analysis were compiled in a report for Swiss Railways [7]. In the case of the Burgdorf Bridge, the steel sleepers will have a significant effect on the track decay rates and provide an additional radiating component. These effects have been accounted for with a specially constructed track model. This includes the modal behaviour of the wooden and steel sleepers modelled as beams. The effect on the decay rates on noise is calculated as correction to the prediction (see [8]). 2.1 Parameters Derived from Measurements on the Bridge Using Service Trains Fig. 2(a) shows the measured vertical direct receptance at the rail-head. Although the measurement quality is poor, especially below 300Hz, the main resonance peak due to the stiffness of the rail support can be clearly seen. The dashed line on Fig. 2(a) shows the calculated receptance using a support stiffness of 240MN/m and a mass per sleeper end of 28kg. In Fig. 2(b) the calculated decay rates are shown along with the decay rates measured by SBB according to the method described in reference [8]. In the measurements of the decay rate responses up to a distance of 7.2m from the excitation have been used. The minimum decay rate measureable with this length of baseline is 0.6dB/m [8]. The measured decay rate is well above this. 10-7 Measured Calculated

10-8

Decay rate, dB/m

Receptance (m/N)

101

10-9

10-10

wooden sleepers wooden + steel sleepers

100

measured -11

10

102

102

103

Frequency (Hz)

(a)

103

Frequency, Hz

(b)

Fig. 2. Comparison of measurement and calculation: left side, vertical point receptance of the rail; right side, decay rates of the rail

Experimental and Theoretical Analysis of Railway Bridge Noise Reduction

211

Equivalent combined roughness level [dB re 1 m]

An estimation of the combined effective roughness (i.e. combined wheel and rail roughness with the contact filter already accounted for) is needed for the predictions. A method to determine the combined roughness from rail vibration measurements under traffic is described in reference [9]. This method uses the spectrum of vibration during a train pass-by and three correction factors. It has been applied to rail vibration measured by the SBB. The roughness is a function of the brake type of the train. Here, the most important trains to consider are the cast iron block tread-braked freight trains. Two measurements of vertical rail vibration are available from trains travelling at steady speeds of 77km/hr and 72km/hr. Three more are available for lower speeds that vary during the measurements from about 40 to 60km/hr. The estimates of the combined effective roughness from these records are presented in Fig. 3. They are plotted as a function of frequency corresponding to a train speed of 100km/hr. These roughness spectra are plotted in comparison with the typical spectra for smooth rail and either cast iron block tread-braked, or disc-braked trains. These are the standard roughness spectra used in the Silent Freight and Silent Track EU projects [10]. The roughness assumed for the present calculations is also shown on Fig. 3. -80

-90

-100

-110

-120

-130

-140

-150

10 2

103

Frequency corresponding to 100 km/hr [Hz]

Fig. 3. Combined effective roughness spectra derived from measurements compared with those used in the Silent Freight and Silent Track Projects (──, calculated from the two faster freight trains; - - - -, calculated from the slower freight trains - - - , SF/ST project combined roughness for tread braked and (lower) for disc braked wheel; ── mean spectrum assumed for current calculations)

3 Noise Measurements The microphone positions were, for both tracks: beneath the bridge; 7.5m to the side, 1.2m above the rail head and 25m to the side, 2m above the rail head. These were all in a plane 6m from one end of the bridge. An additional measurement was made in

212

K.P. Köstli, C.J.C. Jones, and D.J. Thompson

the 7.5m position adjacent to the track at grade. The details of the measurements can be found in a report [11] for Swiss Railways. Both freight and passenger trains were measured. Only freight trains fully equipped with cast iron brakes are included. All passenger trains were either equipped with composite-block or disc brakes. All passenger trains stopped at the nearby station of Burgdorf resulting in a large range of velocities and some changes of velocity (up to 20% in some cases) during measurement.

4 Results and Discussion Fig. 4 presents the measurements and predictions for the freight trains for the north and south tracks before and after the installation of the resilient baseplates. Fig. 5 presents the corresponding measurement results for the passenger trains. No predictions were made for the passenger trains. The measurements are the average of freight trains travelling between 66 and 72km/hr in each case. The predictions are the nearest available at 80km/hr. It can be seen that both baseplate types perform similarly resulting in 5 to 10dB reduction of noise in the 80Hz to 400Hz one-third octave frequency bands. However, around the peak of the noise spectrum near 500Hz, smaller reductions are achieved. It is at this frequency and above that the rail noise dominates over the bridge-structure radiated noise. At 1.6kHz and above the wheel is the dominant noise source and there is very little variation before and after the change was made to the track. Additionally, Fig. 4(a) shows the mean of measurements on the track at-grade. These were made at the slightly lower average speed of 60 km/hr. The noise from the wooden sleepers, which was identified by Twins modelling [7] to dominate below 800Hz, is in this case ‘baffled’ by the ballast and the ground reflection is expected to be fairly absorbing. This is in contrast to the open bridge structure and the reflection -5

-5

Linear sound pressure level

Linear sound pressure level

dB(lin) re 2x10 Pa 100 90 80 70 60 50

31.5 63 125 250 500 1k Frequency [Hz]

(a)

2k

4k

dB(lin) re 2x10 Pa 100 90 80 70 60 50

31.5 63 125 250 500 1k Frequency [Hz]

2k

4k

(b)

Fig. 4. Comparison of predictions with measurements of average of freight trains; (a) north track; (b) south track (······, measured before; ·−·−·, measured after, ──, predicted before; - - -, predicted after, ··−··− (a) measured at at-grade track, (b) measured at the bridge with the steel sleepers and resilient baseplates)

Experimental and Theoretical Analysis of Railway Bridge Noise Reduction -5

-5

dB(lin) re 2x10 Pa 100

Linear sound pressure level

Linear sound pressure level

dB(lin) re 2x10 Pa 100 90 80 70 60 50

213

31.5 63 125 250 500 1k Frequency [Hz]

2k

4k

90 80 70 60 50

(a)

31.5 63 125 250 500 1k Frequency [Hz]

2k

4k

(b)

Fig. 5. Average of measurements of passenger trains; (a) north track; (b) south track (──, before; - - -, after, ······ right graph, measured at the bridge with the steel sleepers still in place but with resilient baseplates installed)

of the water surface. Additionally there is no control of the rail roughness accounted for in the comparison. Despite these differences, it can be observed that the bridge noise has been lowered to levels similar to those of the at-grade track in the frequency bands up to 250Hz. The rail noise component is clearly much lower from the at-grade track than from the bridge (800Hz to 1.25kHz). Fig. 4(b) presents the noise measured when the resilient baseplates had been installed but the steel sleepers had not yet been replaced. It shows that the reduction achieved by the baseplates in the 80 to 400Hz range is compromised by the steel sleepers by around 2 to 3dB. The results for the passenger trains shown in (Fig. 5) indicate the same trends as the freight trains. The 1kHz peak in the spectrum of noise from the south track is probably equipment noise from the trains travelling in this direction.

5 Conclusions The bridge is shown to give much higher levels of noise than nearby track at grade. The installation of resilient baseplates has reduced the overall level difference from about 11dBA to about 8dBA. However, the baseplates make of greater noise reduction of 5 to 10dB where the bridge structure-radiated noise dominates between 80 and 400Hz. The Norbert model has predicted the reduction in noise in the 80 400Hz bands reasonably well although the measured spectra are smoother than those predicted. In the 630 and 800Hz bands, where the rail noise dominates, Norbert has predicted a reduction that is greater than that actually achieved. This leaves the overall noise reduction to be only about 3dB (4dB predicted) for the freight and passenger trains alike. The removal of the hanging steel sleepers was worthwhile to gain the full benefit of the structure noise reduction Promising locations for similar bridge treatments will be identified according to their cost benefit. The baseplates from different suppliers to the same specification produce similar results.

214

K.P. Köstli, C.J.C. Jones, and D.J. Thompson

Acknowledgements The authors are grateful to the SBB for permission to publish this work.

References [1] Bewes, O.G., Thompson, D.J., Jones, C.J.C., Wang, A.: Calculation of noise from railway bridges and viaducts: Experimental validation of a rapid calculation model. Journal of Sound and Vibration 293, 933–943 (2006) [2] Janssens, M.H.A., Thompson, D.J.: A calculation model for the noise from steel railway bridges. Journal of Sound and Vibration 193, 295–305 (1996) [3] Harrison, M.F., Thompson, D.J., Jones, C.J.C.: The calculation of noise from railway viaducts and bridges. Proc. Institution Mechanical Engineers, Part F (Journal of rail and rapid transit) 214, 125–134 (2000) [4] Thompson, D.J., Hemsworth, B., Vincent, N.: Experimental validation of the TWINS prediction program for rolling noise, part 1: Description of the model and method. Journal of Sound and Vibration 193, 123–135 (1996) [5] Thompson, D.J., Jones, C.J.C.: A review of the modelling of wheel/rail noise method. Journal of Sound and Vibration 231(3), 519–536 (2000) [6] Bewes, O., Thompson, D.J., Jones, C.J.C.: Calculation of noise from railway bridges: The mobility of beams at high frequencies. Structural dynamics: Recent advances. In: Proceedings of the 8th International conference, Institute of Sound and Vibration Research, Southampton, (paper 64 on CD ROM) July 14–16 (2003) [7] Jones, C.J.C., Thompson, D.J.: Acoustic analysis of Burgdorf bridge, ISVR contract report no 06/03, University of Southampton (2006) [8] Jones, C.J.C., Thompson, D.J., Diehl, R.J.: The use of decay rates to analyse the performance of railway track in rolling noise generation. Journal of Sound and Vibration 293(3–5), 485–495 (2006) [9] Janssens, M.H.A., Dittrich, M.G., de Beer, F.G., Jones, C.J.C.: Railway noise measurement method for pass-by noise, total effective roughness, transfer functions and track spatial decay. Journal of Sound and Vibration 293(3-5), 1007–1028 (2006) [10] Bouvet, P., Vincent, N., Coblenz, A., Demilly, F.: Optimisation of resilient wheels for rolling noise control. Journal of Sound and Vibration 231(3), 765–777 (2000) [11] Muff, W., Grolimund & Partner AG, SBB Stahlbrücke Burgdorf, Lärmmessungen vor und nach der Sanierung, Bern (2007)

Comparison of Two Metrics for Assessing Human Response to Vibration R. Carman1, C. Reyes1, G. Glickman2, and M. Schaeffler2 Wilson, Ihrig & Associates 1 5776 Broadway Oakland, CA 94618 2 65 Broadway, Suite 401 New York, NY 10006 USA [email protected], [email protected], [email protected], [email protected]

Summary The United States Federal Transit Administration has recently recognized the vibration criteria contained in ISO 2631 Part 2 as the criteria to be used in assessing environmental impacts for new rail projects in the USA. This is specified in the latest version (May 2006) of the FTA publication Noise and Vibration Impact Assessment for Rail Transit. The FTA preferred prediction model for vibration is the empirical based model of Nelson and Saurenman, in use for over 20 years. The force density level (FDL) component of their model is obtained from field measurements of train induced ground vibration. Analysis of the data to obtain an FDL can be achieved by measuring either the Leq of the train passby signature or by recording one-second RMS averages (the metric specified by ISO 2631) over the duration of the passby and determining the “maximum value” for a passby. Both approaches have been used by different practitioners to evaluate new transit lines and assess human response to groundborne vibration and noise. Obviously the two methods can produce different results, but to what degree. Ground vibration data for two different transit train systems (one light rail and one heavy rail) were analyzed both ways and the results compared. The implications for the prediction of groundborne noise and vibration are presented and discussed.

1 Introduction The means of measuring and evaluating environmental groundborne vibration from transit train operations is provided in a recently revised publication of the United States Federal Transit Administration (FTA). The criteria for vibration inside buildings related to annoyance and interference, the FTA publication Noise and Vibration Impact Assessment for Rail Transit [1] to be used is that of ISO Standard 2631 Part 2 [2]. The former document is commonly referred to as the “Guidance Manual.” The Guidance Manual indicates that the vibration criteria are specified in terms of the maximum RMS vibration velocity level with a one-second averaging time. This presumably comes from the ISO Standard 2631. We shall refer to this as the MaxRMS with one second averaging implied. Specifically what is meant by “maximum” is not B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 215–221, 2008. springerlink.com © Springer-Verlag Berlin Heidelberg 2008

216

R. Carman et al.

stated either in the Guidance Manual or the ISO Standard. However, the ISO vibration criteria are specified in terms of 1/3-octave band vibration acceleration levels. One interpretation of the MaxRMS could be that the MaxRMS spectrum is the one-second RMS average spectrum with the highest 1/3-octave band level for the entire passby. An alternative interpretation could be that the MaxRMS spectrum is made up of the maximum level in each 1/3-octave band level regardless of when it occurs during the train passby (i.e., max “hold”). ISO 2631 appears to suggest that the MaxRMS is the RMS spectrum with the highest “overall” level. Conversations with other practitioners have indicated varying interpretations as well. One practitioner prefers to discard the “highest” spectrum and average the rest, the result of which he calls the MaxRMS, but was not clear on the averaging method. At Wilson, Ihrig & Associates (WIA) the procedure for over 30 years has been to characterize wayside groundborne vibration from rail transit by obtaining the energy equivalent level (Leq) of the train passby. The period of time over which the analysis is conducted is the time between the “3 dB down points” of the overall unweighted vibration level preceding and following the train passby. Given the possibility of different results depending on the metric used (i.e., MaxRMS or Leq), it is worthwhile exploring how much the differences might be between the two metrics. Prior to performing the analysis reported herein it was thought by at least one practitioner that there was little if any difference between the two, which we found not to be true as we will show.

2 Discussion of the Issues There are basically two issues we would like to explore concerning these two metrics. One simply involves determining whether a vibration measurement achieves a specified vibration criterion. The other involves the prediction of groundborne vibration (and groundborne noise) for new rail transit lines. While these may seem like different subjects, if one is to make realistic predictions, it is not unreasonable to expect there should be some relation to what one expects to measure. When the ISO 2631 criteria, which are specified in acceleration levels, are converted to velocity levels, the criteria are independent of frequency above 8 Hz. If the significant vibration is greater than 8 Hz, which it usually is, then it would not matter when the highest level in any one band occurred, only that it did and that would determine whether the criterion had been satisfied. However, the MaxRMS metric, as a measure of impact to humans, does not address how long the vibration lasted nor does it tell us whether the highest vibration level lasted for longer than a second. For that matter, neither does the Leq, but it does account for the variation in levels during the passby. The question in this first issue is whether an approach using the MaxRMS adequately addresses how humans respond to groundborne vibration or is there another metric such as the Leq that correlates better with response. Speaking strictly from experience, WIA has found that the Leq is a reasonable predictor of how people respond to groundborne vibration, at least for passbys of sufficient duration. Admittedly attitudinal survey data on the subject is sparse, but there is currently research in the USA on the subject now in progress [3], which will hopefully provide an answer to this question or at least a better understanding of the phenomenon. When it comes to making predictions of groundborne vibration inside buildings that will be adjacent to a new rail transit line yet to be built, the issue of MaxRMS or

Comparison of Two Metrics for Assessing Human Response to Vibration

217

Leq is more complicated. In general terms, the process of predicting vibration for new transit lines attempts to address the typical or average vibration that can be expected and not necessarily a worst case scenario. At least in the USA, this is generally the manner in which practitioners approach the problem. The Leq is clearly more representative of the average vibration level experienced during a passby than the MaxRMS unless the suite of one-second RMS spectra for a passby are averaged as some practitioners apparently do. The Guidance Manual recommends using the prediction model developed by Nelson and Saurenman [4], which is based on an empirical approach. The model has three basic ingredients: force transmitted to the ground from the track, propagation of vibration through the surrounding soil and geologic strata, and dynamic response of the affected buildings. The vibration inside a building is given by the following formula: Lv = FDL + LSR + BVR in dB (re: 1x10-6 inches/second). The FDL is obtained by measuring ground surface vibration at several locations adjacent to a representative portion of the transit system and “normalizing” the vibration levels obtained from a particular site by subtracting the LSR. The BVR models the response of the building relative to the ground surface. The FDL is affected by the metric used to characterize the train vibration. An FDL determined from the MaxRMS would tend to emphasize a worst case scenario, whereas the Leq would account for more energetic events in a passby, but only in an average sense. Wheel flats due to emergency braking are generally the cause of higher than average vibration during a passby. Discarding the highest MaxRMS spectrum when calculating the FDL is apparently the way in which some practitioners deemphasize a worst case using an average representation instead. The predicted vibration levels inside a building are also used to predict groundborne noise. Presumably one would not alter the FDL when predicting groundborne noise. However, the spectrum that produces the highest 1/3-octave band of vibration does not always produce the highest level of groundborne noise, since it involves higher frequencies that may peak at a different moment than the groundborne vibration. This is one of the potential problems with using MaxRMS as the metric to characterize the vibration generated by a train.

3 Measurement of Transit Train Groundborne Vibration WIA conducted vibration measurements at two different transit systems using geophones as transducers, which were mounted on ground stakes or attached to a concrete street curb. The geophone signals were preconditioned and amplified for proper gain control before recording on a DAT recorder. Though there were several geophone channels (typically eight per measurement site) at different distances from the track, we have focused on just two of those distances per site, one relatively close (i.e., 9 m to 15 m) and one farther away from the track (e.g., 30 m). WIA evaluated ground vibration data for a Light Rail Transit (LRT) system in Southern California which operates two-car trains (total length of 55 m) at moderate speeds (generally 75 kph or less, though at times up to 90 kph). The LRT measurement data provided a sample of seven trains to be included in the analysis. Train speeds ranged from 67 to 83 kph and all trains had a two-car consist. The combination

218

R. Carman et al.

of train length and speed produced passby times that were from 1.1 to 3.0 seconds long. We define passby time as the time it takes for the entire length of the train to pass a point on the track. WIA also measured ground vibration for a Heavy Rail Transit (HRT) system in Northern California with longer trains of up to 10 cars (total length of 230 m) that generally travel at greater speeds (up to 112 kph). Data from two different measurement sites for the HRT were analyzed. One site provided a sample of 10 trains and the other a sample of 13 trains. Train lengths at the two sites varied considerably from 3 to 10 cars and speeds varied from 65 to 125 kph. The combination of train length and speed produced passby times that were from 2.2 to 11.6 seconds long. Between the two different transit systems, we obtained a variety of train lengths, speeds and passby times. The MaxRMS 1/3-octave band spectrum was determined with a real time analyzer (RTA) for each train passby as was the Leq spectrum of the vibration for that passby. ISO Standard 2631 Part 1 [5] recommends an exponential averaging to obtain the one-second RMS spectra. In our experience this results in an undue influence by data at the end of the interval. Instead we use linear averaging, which results in equal influence over the interval period. The linear difference between the 1/3-octave band spectra for the two metrics was calculated and a statistical analysis performed on the differences to determine mean values and a standard deviation for each 1/3-octave band.

4 Presentation of Data and Statistical Analysis Results As would be expected, a train passby produces wayside groundborne vibration that varies with time. The overall vibration level varies as does the vibration spectrum as

Fig. 1. Vibration Spectra (1 Second RMS) During a Train Passby

Comparison of Two Metrics for Assessing Human Response to Vibration

219

the train passes due to the passage of individual wheels and the condition of each wheel. An example of the time varying velocity level spectrum for one-second RMS averages is shown in Fig. 1. It is important to note that the buildup and decay of each 1/3-octave band occur at different times. A representative sample of the MaxRMS and the Leq passby spectra for an HRT train passby is presented in Fig. 2. Although the overall (unweighted) vibration levels are virtually the same for the two metrics as are the highest 1/3-octave band levels (although occurring in different frequency bands), the two spectra at frequencies above 25 Hz are quite a bit different with the MaxRMS being less than the Leq. We note that the Leq spectrum shown in Fig. 2 would produce a higher level of groundborne noise than the MaxRMS spectrum would. Linear differences (Δ) were calculated between the spectral values for the ith train passby of the MaxRMS and the Leq spectral values for each 1/3-octave band. The difference is given by Δ(f)i = MaxRMS(f)i – Leq(f)i. The mean value Δmean = (1/N) ΣΔi and standard deviation σ of the differences for all N passbys were calculated. The mean values for one of the HRT measurement locations are plotted in Fig. 3. Mean values that are negative indicate that the Leq is greater than the MaxRMS.

Fig. 2. MaxRMS and Leq Spectra for Single HRT Train Passby

220

R. Carman et al.

5.00 4.00 3.00 2.00 1.00

Mean Value (dB)

0.00 -1.00 Ch 3 Ch 8 Comb

-2.00 -3.00 -4.00 -5.00 -6.00 -7.00 -8.00 -9.00

z

z 5H

0H 40

31

0H

z

z

20

25

z

0H

z

0H 16

z

5H 12

z

z

0H

10

80 H

z

63 H

50 H

z

z 40 H

z

z

H .5

31

25 H

20 H

z

z H .5

16 H

12

z 10 Hz

z H

8H

6. 3

z 4H

5H

z

-10.00

Frequency (Hz)

Fig. 3. Mean Value of Differences for HRT1

The mean value of the MaxRMS and Leq differences for the three measurement sites (two HRT and one LRT) are indicated in Table 1, where we include just the data from 6.3 Hz to 200 Hz. However, the data in Fig. 3 indicate the general trend of the standard deviation outside that frequency range. The data in Table 1 indicate that the standard deviation vary with the site and the transit system. For HRT1 and the LRT, the mean values are positive below 25 Hz, whereas for HRT2 all of the mean values are negative. The mean value varies from -4.4 to 6.7 dB, which is quite a large range. For most of data the absolute value of the mean value is greater than 1. Table 1. Mean Value of Differences (dB) Data For HRT1 HRT2 LRT

6.3 8 10 12.5 0.7 0.0 0.7 1.0 -4.4 -1.8 -2.2 -1.3 1.9 5.4 6.7 2.8

1/3-Octave Band Center Frequency (Hz) 16 20 25 32 40 50 63 80 1.1 1.3 0.3 1.6 0.4 -1.4 -1.5 -1.4 -2.3 -0.7 -0.5 -0.8 -2.3 -2.3 -1.9 -1.3 1.4 0.4 -1.7 -0.7 -2.7 -3.2 -3.5 -3.9

100 -1.7 -1.2 -3.6

125 -2.0 -1.6 -4.2

160 -2.3 -1.2 -3.5

200 -2.8 -1.1 -1.7

The standard deviations of the differences are indicated in Table 2, where we can see that the standard deviation is non-trivial and varies from 1.3 to 4.9 dB. Taken together with the mean value, we see that the mean value plus σ (or in most cases minus σ) can be quite large. For example, at 10 Hz for the LRT, this means Δmean+ σ = 10.6 dB. In fact for most all frequencies, we see that |Δmean± σ| > 3. Clearly there can be major differences between the MaxRMS and Leq metrics and the metric used can

Comparison of Two Metrics for Assessing Human Response to Vibration

221

influence groundborne noise and vibration predictions. For lower frequency vibration, which usually determines perceptible vibration, there are substantial differences that would effect whether the ISO 2631 criteria for vibration is achieved. Table 2. Standard Deviation of Differences (dB) 1/3-Octave Band Center Frequency (Hz) 12. 16 20 25 32 40 50 63 80 5 HRT1 3.9 4.2 3.5 3.0 3.5 2.7 2.1 1.8 1.8 3.0 3.2 3.0 HRT2 3.8 2.8 2.1 2.4 2.0 1.7 2.0 2.2 2.7 1.3 2.5 2.1 LRT 2.1 3.6 3.9 4.9 2.4 2.8 3.0 2.1 3.4 3.7 3.9 4.0 Data For

6.3

8

10

100 125 160 200 2.8 3.8 3.4 3.7 1.4 1.3 2.3 2.6 3.7 3.9 3.7 2.0

5 Conclusions By examining groundborne vibration data covering a wide variety of transit system operating parameters (train length, speed and type of vehicle), we have identified significant differences between the two metrics in general use for measuring and predicting vibration. These differences are not trivial as some might think. There are definite implications for the prediction of groundborne vibration and especially noise depending on how the data are used. It is hoped that these results will prompt further investigations of this nature. It may be instructive to separate out parameters such as passby times or to study more closely the effect that distance from the track has on the differences in the two metrics instead of lumping data as we have done for this study.

References [1] Noise and Vibration Impact Assessment for Rail Transit, FTA-VA-90-1003-06, Federal Transit Administration, Office of Planning and Environment (May 2006) [2] Evaluation of Human Exposure to Whole-Body Vibration, Part 2: Continuous and ShockInduced Vibrations in Buildings (1-80Hz), ISO-2631-2, International Standards Organization (1989) [3] Saurenman, H.J.: Status: TCRP Project D-12 Ground-Borne Noise and Vibration in Buildings Caused by Rail Transit. In: APTA Rail Transit Conference, American Public Transit Association (June 2006) [4] Nelson, J.T., Saurenman, H.J.: A prediction procedure for rail transportation groundborne noise and vibration, Transportation Research Record 1143. In: A1F04 Committee Meeting on the Transportation Research Board (January 1987) [5] Mechanical vibration and shock – Evaluation of human exposure to whole-body vibration, Part 1: General requirements, ISO-2631-1, International Standards Organization (1997(E))

A Study on Source Mechanism in the Interior Noise Problem of High Speed Trains H.I. Koh, H.B. Kwon, W.H. You, and J.H. Park Korea Railroad Research Institute #360-1, Woramdong Uiwang City Kyonggido 437-757 Korea Tel.: +82 31 460 5207; Fax: +82 31 460 5279 [email protected]

Summary In this paper the noise generating mechanism for the interior area of passenger coaches is studied. It is primarily aimed at understanding the noise source mechanism for the narrow band noise problem in the range of 70 Hz~90 Hz. Though these phenomena differ with respect to the track construction inside the tunnel and vehicles, it could be concluded that this problem appears during the whole high speed lines basically. First the problem is investigated in regard to different driving conditions, the aerodynamic noise mechanism and the acoustical distribution near the inter space between passenger coaches are investigated numerically as well as experimentally. These results are analysed also in respect to the structure borne noise tests.

1 Introduction High speed train has started revenue services since April 2004 with the maximum speed of 300 km/h. Since the revenue service noise sources of the high speed line has been being investigated to enhance the passenger comfort and for the application to several on-going low noise vehicle development projects. One of the investigated noise issues was the pronounced tonal noise components of interior noise inside the tunnels installed with slab track systems, which are mostly found below 250 Hz. The possible source mechanisms depend on various structures and factors simultaneously and have been researched in several aspects. In this paper this problem is discussed in relation to the behaviour of the car body structure at high speed runs as one approach. The Vehicle consists of 20 compartments in total, including 2 power cars, 2 motorized cars and 16 passenger compartments. Trains are designed with wheels installed between the compartments, the articulated bogie constructions, between two passenger compartments there exist inter-coach spacing. The Korean high-speed line consists of 412km double track, including 112km (27%) of at-grade sections, 109km (27%) of viaducts and 191km (46%) of tunnels. Rails have no joints, and thus produce less impulsive joint noise and vibration, the 300m long rails were made using a special welding technique, individual rails are connected each other creating a continuously welded track. However several long tunnels are constructed with slab tracks through which trains pass at their maximum speed of 300 km/h. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 222–228, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

A Study on Source Mechanism in the Interior Noise Problem of High Speed Trains

223

First the low frequency noise phenomena in consideration of speeds, car types, characteristics of railway sections are discussed. The noise and vibration behaviour of various parts of car structures are shown, as one possible source mechanism the aeroacoustical problem at the boundary and inside the cavity between passenger coaches are investigated.

2 Noise Characteristics At the beginning of the revenue service measurement results showed in general high energy components in 80 Hz area when analyzing in 1/3 octave band. These occur in most high speed railway sections (Figure 1). Therefore basically it should be primarily related to the general driving mechanism of the train which is possibly results from the rolling dynamics and vehicle constructions.

v=262km /h v=263.6km /h

) A B d( L P S

v=282km /h v=293.7km /h v=296km /h v=297.3km /h v=299.3km /h v=299.8km /h

10dB

10

100

1000 10000 Frequency(H z)

100000

Fig. 1. Sound pressure levels measured inside passenger coach at a high speed line (one year after revenue service)

Investigations on booming noise effects especially in the range of 70 Hz ~ 300 Hz interior of the coach can be found in other studies [1][2][3], in which the rolling dynamic of wheels and vibration transfer are the main objects. As a slab track structure dynamic problem there are wheel/rail impedance model calculations explain wheel/rail resonant frequency area between 40 Hz and 100 Hz, which is influenced by damping and stiffness of intermediate plate of the rail fastener. At contact resonance area between 300 Hz and 800 Hz, affected by stiffness of the intermediate plate and masses coupled with rail, the impedance of the wheel is bigger than that of the rails, there exists eigenfreqeuncy with maximal velocity of the rail [4]. In this study we concentrated rather on the additional level increase effects than the basic source mechanism of the interior booming noise. The low frequency noise components became quite dominant interior of the passenger coach than inside motorized

224

H.I. Koh et al.

coaches and especially in the tunnels installed with slab tracks, which have a length of 5 km and more. As in Fig. 2 and Fig. 3 shown according to the measurement one year after the start of the revenue service interior noise level differences between passing tunnels with slab track system and with ballast track system could be observed, at high speed operation. The sound level difference is distinctive mainly between passenger coaches and inside the passenger coach. The vibration behaviour of the car body structures showed also differences in both types of tunnels mainly in the range of 60 Hz~100 Hz. When train passed through the tunnel with slab track system interior noise increased considerably at this frequency range and it could be observed that this is caused not only by resonant effects of car body structures, such as floor, side walls and windows but also by noise and vibration near the cavity between passenger coaches. It could be assumed that for this emergence not only the dynamic behaviour of the articulated bogie and car body system with slab track system but also the aerodynamic factors inside a tunnel with slab track system can be possible affecting factors.

3 Low Frequency Noise Source Investigation One possible noise increasing path is the aerodynamic effects at the cavity boundary with a gap opening (Figure 4) through which pressure increases and acoustical resonance is induced. Whether this effect can affect the resonant vibration of the car body structure in the are of 60Hz ~ 100 Hz and cause the high noise level inside the passenger coach the inter coach space has been investigated. At the vicinity of the gap opening the flow results pressure increase on the surface (shear layer) and results in acoustical feed back phenomena inside the cavity. The principle and related resonance frequencies which depend on passing flow speed and gap dimensions are shown in Figure 4 and Figure 5. The resonance frequency induced from the pressure feedback inside the cavity can be obtained by means of the Rossiter’s semi empirical formula [5][6]:

f =

U∞ m − α L 1+M ∞ k

(1)

12 cavity between coaches under vehicle interior of the coach

10 8

).l b( L P S)b al s( L P S

6 4 2 0 -2 -4 -6

20

40

80

160

315

630 1.25k Frequency(Hz)

2.5k

5k

10k

20k

Fig. 2. Differences between sound levels measured inside a tunnel with slab track system and inside a tunnel with ballasted track system (one year after revenue service)

A Study on Source Mechanism in the Interior Noise Problem of High Speed Trains

225

12 window side wall floor

10 8

sat la bba sl( L A V

6 4 2 0 -2 -4 -6 -8

20

50

125

315 800 2k Frequency(Hz)

5k

12.5k

Fig. 3. Differences between vibration levels measured inside a tunnel with slab track system and inside a tunnel with ballasted track system (one year after revenue service)

Fig. 4. Cavity between adjacent passenger coaches and acoustical feedback principle

where U∞ is flow speed, L is cavity length and k = 0.57 and α =0.25 is values obtained empirically. According to Fig 5 for example at speed of 300 km/h with cavity length of 0.3m~0.4m, calculated frequency area is 78Hz ~ 104 Hz. The sound pressure due to the pressure around the open gap of the cavity becomes lower with the reduced gap size. The sound level reduction effects increase if the resonance frequency of the flow feedback inside the cavity resulted from this pressure falls off. This can be one of the explanation for the amplified noise level inside the cavity between the coaches in the tunnel, but it has not clearly defined which factor is responsible for the big difference of the noise level of two different type of the tunnels. It can come from the different flow speed and different vibrating level of the car body or higher noise level inside the tunnel with slab track system due to the reflection at the concrete slab but the different track dynamic behaviour can be also affecting factor or all factors together simultaneously.

226

H.I. Koh et al.

Fig. 5. Resonance frequencies inside cavities with various lengths

Fig. 6 shows the numerically predicted sound pressure level near cavity openings with different lengths, at the same flow speed. The calculation model of the cavity structure of the high speed train is obtained by Gambit program and for analysing the aerodynamic noise effect using Fluent program LES (Large-Eddy Simulation) Model is used. As parameter train speed, dimension of the cavity gap are varied. The sound level reduction inside the cavity achieved by the size change is mainly in area 50 Hz ~ 250 Hz. In Fig. 7 the measured sound levels inside the passenger coaches with reduced gap size are analysed, which show reduced level difference between results of tunnel with slab track system and tunnel with ballasted track system. The reduced differences in interior noise level between both types of tunnels are mainly in 70 Hz ~100 Hz, however since it is related to resonance frequency change there exist areas where noise levels are rather increased. In overall measurement results show somewhat reduced interior noise levels for both types of tunnels and these depend on vehicles and measuring locations. But it is to see that this is not the principal approach for eliminating of the interior noise sources and besides, inside tunnels with slab track system the interior noise in the area

Sound Pressure Level (dB)

L=0.39(m) L=0.26(m)

20dB

0

100 200 300 400 500 600 700 800 900 10001100 1200

Frequency (Hz)

Fig. 6. Calculated sound pressure level inside cavity

A Study on Source Mechanism in the Interior Noise Problem of High Speed Trains

227

5 4 3

)l. b- 2 ba ls 1 (L P 0 S -1 -2 -3 29

47

74

118

187 297 472 Frequency(Hz)

749

1189 1888 2997 4757

Fig. 7. Differences between sound levels measured inside a tunnel with slab track system and a tunnel with ballasted track system (cavity openings = 0.26m) 15 10 5 ). l b 0 b al (sL -5 A V -10

bogie floor wheelset

-15 -20

29

47

74

118

187

297 472 Frequency (Hz)

749

1189

1888

2997

4757

Fig. 8. Differences between vibration levels measured inside a tunnel with slab track system and a tunnel with ballasted track system (cavity openings = 0.26m)

of 300 Hz~1000 Hz has still higher energy portion than inside tunnels with ballasted track system. In Fig. 8 the measured vibration level of the bogie, wheelsets and floor of the passenger coach are shown, the gap size of the cavity is 0.26m. The differences between levels from both types of tunnels are displayed. This doesn’t represent tendencies of all tunnels with slab track system or ballasted track system but this result shows that the interior noise level differences of two types of tunnels in 300 Hz ~ 3000 Hz are more related with the structural dynamical behaviour under the car body than that in 70 Hz ~ 100 Hz after the change in resonance frequency of the acoustical feedback inside the inter coach space.

228

H.I. Koh et al.

4 Concluding Remark This paper dealt with the interior noise phenomena inside tunnels with slab track system at train speeds of 260 km/h ~ 300 km/h. Interior noise of passenger coaches results from rolling dynamics of wheel and rail and vehicle constructions which are dependent on track systems and train speeds. At high speeds where the aerodynamic effects become dominant it can influence the structure borne noise sources of vehicles and results in amplification of noise components in certain frequency areas. Various noise sources exist simultaneously at the high speed operation. In this study the increased noise and vibration level inside tunnels with slab track system compared to the level inside tunnels with ballasted track system are investigated, which occurred mainly at low frequencies and could be heard as Booming noise interior of the passenger coach especially. Measured sound levels in other high speed sections, such as open fields show also relatively high energy portion at that frequency area but are not so pronounced as in tunnels. As one aspect aerodynamic effects near the inter space cavities between passenger coaches where the articulated bogie systems are placed are investigated. Maybe it is not directly related to the tunnels with “slab track system” itself but considering the noise increased in frequency areas 70 Hz~100 Hz, acoustical feedback effects due to the flow disturbance along the cavity gaps are studied. The effects of resonance frequency change due to the gap size variation on the interior noise level of passenger coaches in two types of tunnels are compared. In the related frequency band the interior noise level could be reduced and the differences between noise levels inside two types of tunnels could be reduced as well. However it is still remained to identify other parameters which cause the noise increase inside the tunnels with slab track system and to reduce it efficiently.

References [1] Schirmacher, R., Hoelzl, G., Redmann, M., Scheuren, J.: Active noise and vibration control for a high speed railcar: A case study. In: Proceedings of Active 1997, pp. 557–564 (1997) [2] Peiffer, A., Storm, S., Röder, A., Maier, R., Frank, P.G.: Active vibration control for high speed train bogies. Smart materials and structures 14, 1–18 (2005) [3] Morys, B., Kuntze, H.B.: Simulation and analysis and active compensation of the out-ofround phenomena at wheels of high speed trains. In: Proceedings of the World Congress on Railway Research WCRR 1997, Florence (Italy), November 16-19 (1997) [4] Mueller, G., Moeser, M.: Taschenbuch der technische akustik. Springer, Heidelberg (2004) [5] Kerschen, E.J., Alvarez, J.O., Tumin, A.: A theoretical model for resonances in flow past a cavity. In: XXI ICTAM, Warsaw, Poland (August 15–21, 2004) [6] Rossiter J.E.: Wind Tunnel Experiments on the Flow over Rectangular cavities at subsonic and transonic speeds. Royal Aircraft Establishment, TR No. 64307 (October 1964)

Reducing the Noise Emission by Increasing the Damping of the Rail: Results of a Field Test B. Asmussen1, D. Stiebel1, P. Kitson2, D. Farrington2, and D. Benton2 1

Deutsche Bahn AG, Technik/Beschaffung, DB Systemtechnik, Völckerstr. 5, D-80939 München, Germany Tel.: +49 89 1308 7547; Fax: +49 89 1308 2590 [email protected] 2 Corus Rail France SA 2 avenue President Kennedy, 78100 St Germain en Laye, France Tel.: +33 139 04 63 24 [email protected]

Summary The noise emission from railways can be reduced by increasing the damping of the rail. A system of rail dampers has been optimized and tested within the EU-funded project SILENCE. Besides the direct noise mitigation effect these rail dampers can also indirectly reduce noise emission by reducing the roughness growth rate on the rail. The paper will describe the damping system and the field tests, which were carried out in order to directly measure the achieved noise mitigation.

1 Introduction Rolling noise is the dominant contribution to the overall noise emission from conventional rail traffic. It is caused by the roughness of the surfaces of wheel and rail and is radiated by both the rail and the wheels with the radiation from the rail dominating in the frequency range below 1000 Hz and the radiation from the wheels dominating above 1500 Hz. This implies that special attention has to be paid to the noise emission from the rail when discussing noise reduction measures at the source for railway lines in urban areas where the speeds are typically < 120 km/h. SILENCE is an integrated project funded by the EU within the sixth frame work program. It focuses on noise reduction in urban areas [1]. One work package within SILENCE aims at reducing the noise emission by increasing the damping of the rail. The target is twofold: (1) Increased damping shall reduce the direct noise emission of the rail by modifying the vibration behaviour of the rail and (2) indirectly by reducing the roughness growth rate. In this paper we will introduce an optimized rail damping system and the results of a field test focussing on the achieved direct noise mitigation.

2 The Concept of Reducing Noise Emission by Increasing the Damping of the Rail It is known that the rail vibration spectrum differs according to the form of track design (defined by rail type, pad stiffness, rail fastening and sleeper design) and the type B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 229–235, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

230

B. Asmussen et al.

of traffic passing over it as well as the speed of that traffic. A rail damping system can be adapted to suit these variations - for any type of track form and any rail size. It is possible to design the damper to absorb vibration at the frequencies that are producing the greatest sound power. Rail damping has been shown to increase track decay rates, reducing the distance travelled by the vibration energy and thus reducing the noise generated by the track [2]. The rail dampers produced by Corus consist of steel masses and elastomer-based materials. The elastomer functions as damped springs in a mass-spring system, absorbing energy through internal friction in the elastomer as the steel elements vibrate in response to vibrations in the rail. The mass-spring system is tuned to respond to the frequency range where the greatest noise producing vibration is found. To achieve this tuning, the mass of the steel elements and their spacings are altered such that the resonant frequencies of the mass-spring system correspond to the middle of the frequency range being targeted. At the resonant frequency the movement of the steel mass relative to the rail is at its greatest, and thus the largest amount of energy possible is absorbed as internal friction in the elastomer spring. The energy absorbed by the damper is therefore maximised, and at the frequencies that are producing the greatest sound power, the vibrations are minimised. This effect is readily seen in the changes of decay rate from un-damped to damped rails (see figure 1). The rail dampers designed by Corus use two masses, one above the other, thus generating two dominant frequencies at which energy absorption is maximised (as well as a number of other modes). The resonant peaks are very wide due to the high damping, so that the final system behaves as a broad band damper as the tails of the peaks overlap. Compared to a single mass system [3], this widens the frequency range at which the damper operates, and enables an effective noise reduction to be obtained.

Decay rate (dB/m)

100

10

1

0.1

0.01 100

10 1000

10000

One third-octave band centre frequency (Hz) Fig. 1. Decay rate tests of rails with (solid line) and without dampers (dashed line), mounted on low impedance supports to remove the effect of the track

Reducing the Noise Emission by Increasing the Damping of the Rail

231

3 The Test Track at Gersthofen A suitable test site had to be chosen for performing field tests. It had to meet the following requirements: • • • •

Representative superstructure for tracks in urban areas Easy access for installation of dampers and roughness measurements Representative traffic mix: regional trains, high-speed trains, freight trains Suitable for microphone measurements

Finally it was decided to install the dampers near Gersthofen on the railway line Augsburg – Donauwörth. The relevant track parameters are listed in Tab 1. Pad stiffness and damping were measured separately on a test rig. Table 1. Relevant parameters of the test track

Track Rail Sleeper Pad Dynamic pad stiffness dynamic pad loss factor Max. train speed

Ballast UIC60 B70 (concrete) Zw 700 (Wirthwein) 400 MN/m 0.17 200 km/h

A section of 45 m length was equipped with dampers provided by Corus (see Fig. 2). A neighbouring section with identical track parameters but without dampers served as reference.

Fig. 2. Test section near Gersthofen equipped with rail dampers

232

B. Asmussen et al.

Vertical track decay rate (dB/m)

Both the test section and the reference section were characterized by measuring their surface roughness and their track decay rates. The rail roughness measured according to EN ISO 3095:2005 in both sections was well below the TSI+ limit curve. It is slightly higher in the damped section, which means that the effect of the dampers is slightly underestimated (at least for high speed trains with low wheel roughness; for freight trains the differences in rail roughness are irrelevant). The measured track decay rate is shown in fig.3 (lower curve for the un-damped rail). In a first step, the performance of the existing design was examined in order to define the potential for further optimization. A clear increase in track decay rate could be recorded (see middle curve in fig. 3) but the target for the noise reduction of ~3dB(A) or better was not yet reached. Further investigation showed that the performance of the damper could be improved by the use of an acoustic coupling material. This enabled the damper to be more closely coupled to the rail, such that it can better respond to the vibrations in the rail. The coupling material is applied between the damper and the rail as a thixotropic paste which cures to produce a material which is stiff with respect to the elasomer, and thus more effectively transfers vibrations into the damper. Therefore a new clipping system in combination with the application of a filler material between damper and rail was developed by Corus. This improved damping system was installed at the Gersthofen test track in September 2006. The corresponding track decay rate shown as the upper curve in fig. 3 is significantly enhanced as result of the optimization. It is now by 5-10 dB higher in the acoustically relevant frequency range than the decay rate of the un-damped rail.

100

10

1

0.1 125

250

500 1000 2000 Frequency (Hz)

4000

8000

Fig. 3. Measured track decay rates; lower curve: un-damped reference section, middle curve: damped section with the original coupling system, upper curve: damped section with the optimized system

Reducing the Noise Emission by Increasing the Damping of the Rail

233

4 Results of the Field Tests In an extensive measurement campaign in September 2006 the performance of the optimized dampers was investigated by measuring pass-by noise levels on the damped section and on the un-damped reference section for the regular train traffic (ICE trains, regional trains both with block brakes and with disc brakes, and freight traffic). Two microphone positions at distances of 3m and 8.5 m respectively were chosen in each section. Fig. 4 as an example shows the spectrum of the pass-by noise of a regional train. It should be noted that both curves in fig. 4 were recorded for the same train. A reduction of the noise emission by up to 4 dB(A) is obtained in the acoustically relevant frequency range f > 500 Hz. The cross-over in fig. 4 occurs exactly in the frequency range where according to fig. 3 the dampers begin to significantly increase the decay rate of the rail vibrations. The maximum performance is obtained around f=1000 Hz, which is the tuning frequency of the dampers (see section 2). Tab. 2 lists in dB(A) the measured differences in the pass-by sum levels averaged for each of the four train categories ICE (high-speed train), IC (long-distance trains), regional trains, and freight trains. The highest mitigation was recorded for freight trains (average 4,1 dB(A) over all measured trains in this category). This is due to the relatively low speeds of the freight trains. This means that the maximum in the sound emission spectrum occurs near 1000 Hz, where the sound radiation from the rail predominates. Additionally, this maximum matches the tuning frequency of our dampers. At higher speeds the maximum in the spectra shifts to higher frequencies, where the sound power radiated from the wheel is usually larger than the sound power radiated from the rail. 1000.0

100

8000

Sound pressure level in dB(A)

95

90

85

80

75

70

65

60

125

250

500 1000 2000 4000 8000 Frequency (Hz)

Fig. 4. Sound pressure spectra of the pass-by noise of a regional train on the damped track (lower curve) and on the reference track (upper curve)

234

B. Asmussen et al.

Table 2. Measured differences in the pass-by sum levels in dB(A) averaged for each of the four train categories ICE (high-speed train), IC (long-distance trains), regional trains, and freight trains. The final line gives the average speed of the trains in the respective train category. microphone distance from the centre of the track 3m

ICE

IC

regional

Freight

2,9

3,2

3,4

4,1

8.5 m

2,1

2,7

3,0

3,4

average speed in km/h

162

165

138

90

Train category

More detailed information is contained in the spectral representation of the difference in noise emission between damped and un-damped track (see Fig. 5). According to Fig. 5 reductions up to 6 dB(A) are obtained in certain one third octave bands. The maximum noise mitigation occurs at the tuning frequency of the damper near 1000 Hz. It is the higher the lower the average speed of the respective train category is. On a second test section near Gersthofen only one rail was equipped with dampers over a length of 30m. This section is dedicated to investigate the influence of the dampers on roughness growth by regularly monitoring the difference in roughness between left and right rail. The parameters differ from those of the first test section given in Tab. 1 in that it has a stiffer rail pad (Cemafer Zw700, dynamical stiffness 800 MN/m). Since roughness growth is predicted to be higher on a stiffer rail fastening system, this section of track with stiffer rail pad is more likely to show significant differences between damped and un-damped rail within the time frame of the SILENCE project. However, quantitative measured results are not yet available. Therefore this paper has focussed on the direct mitigation effect of the dampers. In

Fig. 5. Spectral representation of the difference in noise emission between damped and undamped track. The average speed of the trains in the different catagories is given in Tab. 2. “RE with Dosto” denotes regional trains with double-decker coaches.

Reducing the Noise Emission by Increasing the Damping of the Rail

235

order to understand in detail and to enable the optimization of the effect of rail dampers on roughness growth, this part of the project involves extensive model development and computer simulation, which is carried out by ISVR, University of Southampton.

5 Conclusion and Outlook A noise reduction up to 4 dB(A) due to the increased damping of the rail was measured on the test track. The mitigation potential of the dampers is dependent on the rail fastening system: It is larger for softer rail pads because the contribution of the rail to the overall noise is higher than for stiffer pads. By variation of the tuning frequency (mainly by changing the size and spacing of the steel elements, or the stiffness of the elastomer of the damper) the dampers can be optimized with respect to the speed of the trains. Future work will aim at broadening the frequency range of a damper, in such a way as to combine the anti roughness growth dynamic characteristics with decay rate reduction in an optimal manner. This will be done by an integrated approach of development, computer simulation and validation measurements. Within SILENCE a field test will be carried out to measure the combined effect of damped freight wheel on a damped rail.

Acknowledgements The authors like to thank C.J. Jones and B.C. Croft (University of Southampton) for many fruitful discussions. This work was supported by the European Union within the Project SILENCE.

References [1] Project homepage, http://www.silence-ip.org [2] Vincent, N., Bouvet, P., Thompson, D.J., Gautier, P.E.: Theoretical optimization of track components to reduce rolling noise. Journal of Sound and Vibration 193, 161–171 (1996) [3] Thompson, D.J., Jones, C.J.C., Waters, T.P., Farrington, D.: A tuned damping device for reducing noise from railway track. Applied Acoustics 68, 43–57 (2007)

Railway Noise Abatement: The Case for Retrofitting Freight Vehicles with Composite Brake Blocks J. Oertli Swiss Federal Railways, Infrastructure, Schanzenstrasse 5, CH-3000 Bern 65, Switzerland Tel.: +41 51 220 39 40; Fax: +41 51 220 51 09 [email protected]

Summary Basically rolling noise in railways is created by rough wheels and tracks. If both can be kept smooth, noise can be reduced significantly. Smooth wheels can be achieved by replacing cast-iron brake-blocks with composite brake blocks. Currently two types of composite brake blocks are being discussed: K- and LL-blocks. K-blocks probably have a higher noise reduction than LL-blocks, but require adapting the braking system while wagons can be retrofitted with LL-blocks without adapting the braking system. Several economic studies show that railway noise reduction in retrofitting the freight wagon fleet with composite brake blocks has the highest cost-effectiveness. Also, if composite brake blocks are combined with other measures, the overall costeffectiveness is increased. However in many European countries extensive noise barrier construction programmes are being implemented. Due to the harsh competitive transport market, retrofitting is not possible without external financial support for railway operators. Currently EU funding is only likely for pilot or demonstrator projects. Direct national subsidies are the most efficient short term solution. Differential track access charges are considered too complicated for a short term implementation, they may prove useful to maintain a silent freight fleet in the long term.

1 The European Framework The European Commission is concerned about the impact of transportation on the environment. It realises that railways are the most environmentally friendly and sustainable means of transportation, both for freight and passenger traffic. In a white paper the European Commission therefore proposed to increase the market share of the railways. The stated aim is to attain the levels of 1998 by the year 2010 [1]. In a green paper [2] the EU considers noise one of the main local environmental problems, saying that noise abatement should therefore be given high priority. As a consequence new and stricter noise legislation is being implemented considering both transport noise creation as well as ambient noise reception. Several working groups (WG) advise the Commission on noise questions. One of them was the WG Railway Noise, which concluded its work in 2004. This WG included participation of all major stakeholders, analysed many different noise abatement scenarios and produced a position paper proposing retrofitting rolling stock with silent braking systems and noise limits for new rolling stock as a first priority [3]. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 236–242, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Railway Noise Abatement

237

2 Legal Framework for Railway Noise With Technical Specifications for Interoperability (TSI) the EU enacts noise creation limits for railway vehicles, both for new rolling stock and for renewed or upgraded rolling stock. Different values are defined for the various types of rolling stock (e.g. freight wagons, locomotives, multiple units, coaches) as well as for different operation situations (e.g. pass by, stationary, starting and interior noise). For conventional railways the limit values for pass-by noise came into force on 23 June 2006. TSI regulations undergo a regular revision process every three years. This TSI acknowledges that retrofitting is desirable to accelerate reduction of rail freight traffic noise. In addition, all European countries as well as Norway and Switzerland have noise reception thresholds for new lines. Many countries also have limits for upgraded lines, while a few, such as Switzerland and Italy, also have reception thresholds for existing lines. The directive 2002/49/EC relating to the assessment and management of environmental noise requires strategic noise maps and action plans for major railways for major railways by 2007 (maps) and 2008 (action plans). Other railways will be treated five years later.

3 The Railway Framework Railway freight traffic is the main source of noise on existing railway networks. In order to maintain a sustainable transport system, the railways must reduce noise as their main environmental problem. If this is not done, the favourable view on railways may decline. In addition noise issues may prevent a traffic increase and therefore hinder the implementation of the European transport policy and its focus on increasing the railways’ traffic share. The particular circumstances in which railways operate must be taken into account when considering solutions for railway noise: • • • •

The railways operate in a very tight competitive economic environment. Each investment influences competitiveness and must be considered very carefully. Normally freight wagons are only replaced after a very long life span. A satisfactory noise reduction therefore cannot be achieved merely through the normal replacement of existing wagons. Many stakeholders with different agendas are involved. These include operators, infrastructure owners, governments, regional authorities, and line side inhabitants. Action planning will be the responsibility of the infrastructure departments. These should include retrofitting in the currently ongoing action planning process in the framework of implementing the END.

4 The UIC Action Programme The railways recognize the need for noise reduction. Therefore the UIC (International Union of Railways), the CER (Community of European Railways and Infrastructure Companies) and the UIP (International Union of Private Car Owners) initiated the “Freight Traffic Noise Reduction Action Programme” in 1998. This project aims to

238

J. Oertli

equip new freight wagons with composite brake blocks and to achieve the retrofitting of the existing European freight fleet.

5 The Available Technology Railway rolling noise is the result of small irregularities or roughness on the wheel and on the track. When in motion, this causes both the wheel and the track to oscillate, thus creating noise. A significant portion of the noise can be eliminated, if both the wheels and the track are smooth. Cast-iron braked wheels cause rough wheels. On the other hand, wheels remain smooth using composite brake blocks. Therefore, the choice of brake blocks has a large effect on rolling noise. Currently there are two types of composite brake blocks in discussion: The K- and the LL-blocks. Table 1. Comparison of K- and LL-brake blocks

Rolling noise reduction

Retrofitting possibilities Braking characteristics Homologation Approval of braking system

K-blocks 8 – 10 dB

Requires adapting braking system Independent of velocity Definitive homologation of three types since 2003 New approval required

LL-blocks Not yet sufficiently quantified, 2 dB less than Kblocks expected Minor adapting of braking system Velocity dependent (similar to cast iron brake blocks) Provisional homologation since 2005 for three types No new approval required

In addition to braking performance, homologation requires safety and operating issues, such as performance under severe winter conditions and studying possible effects on track circuits. Not all technical issues have been resolved, but the increase in efficiency in overall noise abatement is well worth the effort to continue development.

Fig. 1. Picture of wagon retrofitted with k-blocks. Old wagons such as these are now as silent as modern passenger vehicles.

Railway Noise Abatement

239

6 The Economics 6.1 Studies Undertaken Anticipating the need to optimize noise control strategies on a European level, both the railways and the EU have undertaken several cost-effectiveness analyses. The most comprehensive study was the STAIRRS (Strategies and Tools to Assess and Implement noise Reducing measures for Railway Systems)[4] project, co-financed by the EU fifth framework programme and by the UIC. In this project the acoustically relevant geographic, traffic and track data were collected for 11’000 km of lines in seven European countries. Standard cost-benefit methodologies were adapted to fit the requirements of the project. An extrapolation mechanism allowed studies on Europe as a whole and, in an approximate manner, also on each individual country or region of interest. E u r o p e , 2 1 c o u n t r ie s , P C n o w in d o w s / P B U IC S t e e r in g G r o u p S c e n a r io

P r e s e n t b e n e fits (p e r s > 6 0 d B

2 5 0 .0 0

w ith k - b lo c k s a n d o p t.w h e e ls

2 0 0 .0 0

a ll m e a s u r e s 4m

w ith k - b lo c k s

w ith k - b lo c k s

N o is e b a r r ie r s 2m

1 5 0 .0 0

1 0 0 .0 0

T u n e d a b so rb ers

w ith g r in d in g

F r e ig h t r o llin g s to c k im p r o v e m e n t

5 0 .0 0

G r in d in g 0 .0 0 0

10

20

30

40

50

60

70

80

P r e s e n t c o s t s , E U R O ( in b illio n s )

Fig. 2. Main results of the STAIRRS project. The graph shows that solutions using composite brake blocks save considerable amounts of money in comparison to noise abatement with only noise barriers.

Major conclusions are: • • • •

Good cost-effectiveness can be achieved by combining measures Freight rolling stock improvement has the highest cost-effectiveness both on its own and in combination with other measures. Noise barriers, in particular high ones, have a low cost-effectiveness. The conclusions for Europe as a whole are also true for individual countries.

In sum, STAIRRS shows that solutions using composite brake blocks save considerable amounts of money (billions of Euros in many European countries) in comparison to noise abatement including only noise barriers.

240

J. Oertli

6.2 Overall Cost Estimates Purchasing new wagons with K- or LL-blocks instead of cast iron blocks does not increase the overall costs of a vehicle. On the other hand, retrofitting existing wagons with K-blocks gives additional costs of €€ 4’000 to €€ 10’000 per vehicle [5], depending on the number of axles and wagon type. Retrofitting using LL-blocks is significantly less expensive and may even be cost-neutral. It must be noted that considerable costs occur for each wagon type in the homologation process. Wagon classes consisting of only few vehicles are therefore not the primary focus for retrofitting. Retrofitting is most cost-effective if carried out during compulsory freight wagon inspection, which must be undertaken at least every 6 years. In total about 600’000 wagons must be retrofitted in all of Europe. First studies indicate that maintenance costs are probably not affected when castiron blocks are replaced with composite brake blocks. Some studies indicate a small increase in costs while others show a small decrease. The main cost drivers are wheel and brake shoe wear. There is potential for optimization in maintenance cycles, so that an overall decrease in costs is expected. 6.3 Regulations and Possibilities for Funding and Financing Due to the harsh competitive transport market, railway freight companies currently do not have the financial possibilities for investments in composite brake blocks. Retrofitting the freight fleet will therefore require financial help from outside, which might be supplemented by incentives for railway undertakings and wagon owners. Possibilities include: • • •

EU funding: Possible funding of pilot or demonstrator projects is being investigated. The EU is planning a communication on the retrofitting process by the end of 2007. A public hearing on the issue was held in May 2007. National funding: The EU is in the process of developing state aid rules regulating subsidies by member states including retrofitting freight wagons. Differential track access charges: Due to the large number of stakeholders involved - e.g. the operator is not necessarily the same entity as the wagon owner it is doubtful that differential track access charges will lead to a large scale short term retrofitting. Furthermore there is a risk that differential track access charges may increase the overall cost or rail transport which in turn runs counter to the stated policy of promoting rail traffic due to environmental reasons. Most railways therefore do not support differential track access charges and prefer direct subsidies instead – at least for a short term retrofitting programme.

7 Current State of Noise Abatement in Europe Currently many countries have started extensive programmes to install infrastructure related noise protection. According to a study carried out by the UIC [6], about 1000 km of barriers have already been constructed along European railway lines, together with noise insulation in some 60’000 buildings. Related annual costs are put at €€ 200 million.

Railway Noise Abatement

241

In terms of retrofitting, the current situation (June 2007) can be described as follows: •

• • • •

Switzerland: All Swiss rolling stock will have been retrofitted by 2010. Financing comes from funds to promote public transportation consisting mostly of money received from taxes on road traffic. Currently, several thousands of wagons have been retrofitted since the start of the programme in 2000. All retrofitting is being undertaken with K-blocks. Each wagon group requires a separate engineering. No technical or security problems have been encountered with retrofitted wagons. Germany: The German legislative bodies are in the process of deciding on financing the retrofitting of freight wagons. The German traffic minister supports the retrofitting. Italy: Currently extensive financial means are being spent on noise barriers. Italy, however, plans to reconsider the policy by the end of 2007. The Netherlands: Extensive tests with LL-brake shoes are being conducted. Other countries: Some other countries such as the Czech Republic are considering pilot projects with composite brake blocks.

In sum there is a distinct risk that non-cost-effective noise abatement measures will be implemented for most major railway lines, resulting in additional costs of several billion euros. Ideally, therefore, finances would be transferred from infrastructure measures to retrofitting of freight vehicles. Fig. 3 illustrates the potential savings, if rolling stock is retrofitted in comparison to the planned expenditure on infrastructure. These savings largely result from a decrease in the required infrastructure measures. The earlier the retrofitting process is started, the larger the potential savings. C o s ts

B o th p o lic ie s h a v e th e s a m e b e n e fit

1 0 ’0 0 0 M io

re m a in in g e x p e n d itu re fo r in fra s tru c tu re

P la n n e d e xp e n d itu re F o r in fra s tru c tu re

S a vin g d u e to re tro fittin g

In v e s tm e n ts re tro fittin g ro llin g s to c k

7 0 0 M io M o n e y s p e n t to d a te O n in fra s tru c tu re T o d a y ’s p o lic y

W ith re tro fittin g

Fig. 3. Potential savings by transferring financial means from infrastructure to retrofitting freight rolling stock. Investments in rolling stock result in a reduction in infrastructure measures.

8 Conclusions •

Retrofitting saves money: Noise abatement solutions using freight wagons with composite brake blocks are cost-effective and save considerable amounts of money (billions of Euros in many European countries) in comparison to solutions including only noise barriers.

242

•

• •

J. Oertli

Large risk of implementing non-effective infrastructure measures: There is a risk that non-cost-effective infrastructure based noise abatement measures will be implemented for most major railway lines, costing some additional billions of euros. Currently many countries have infrastructure based noise abatement strategies in place. Therefore retrofitting must be started soon to prevent inefficient use of funds. Remaining technical difficulties with composite brake blocks do not offset economical advantage: Retrofitting has such a large overall economic advantage that technical obstacles should not be used as an excuse to postpone retrofitting. Outside financial support necessary for railway operators: Due to the harsh competitive transportation market the railways are currently not in a position to finance retrofitting. Direct subsidies are favoured by the railways as the best method to achieve a rapid retrofitting. Differential track access charges may provide a long term incentive to maintain a quiet freight fleet.

References [1] European Commission, White Paper, European transport policy for 2012, time a decide, 370 (Com 2001) [2] http://ec.europa.eu/environment/noise/greenpap.htm [3] http://ec.europa.eu/environment/noise/pdf/railway_noise_en.pdf [4] Oertli, J.: The STAIRRS project, work package 1: Cost-effectiveness analysis of railway noise reduction on a European scale. Journal of Sound and Vibration 267, 431–437 (2003) [5] AEAT Technology, Status and options for the reduction of noise emissions from the existing European rail freight wagon fleet – including a third-party assessment of the UIC/UIP/CER Action Programme Noise reduction in Freight Traffic (2001) [6] UIC, Status report noise abatement on European railway infrastructure (2007)

A Systematic Approach for Arriving at Reasonable Heights and Locations for Noise Barriers Adjacent to Railway Lines C. Weber and K. Atkinson Heggies Pty Ltd, Sydney NSW Australia Tel.: +612 9427 8100; Fax: +612 9427 8200 [email protected], [email protected]

Summary Communities living adjacent to railway lines and major roadways are demanding much higher transparency and assurance of an equitable outcome in the determination of noise mitigation measures for new and upgraded projects. As a result, it is now necessary for the acoustic consultant to apply methodologies that are fair, reasonable and clearly documented in order to provide communities with the required level of assurance. This paper describes a noise barrier design optimisation methodology that has been applied on recent rail upgrade projects in Sydney Australia. The noise barrier optimisation process uses project related noise goals (LAmax and LAeq) to inform the calculation of the Total Noise Benefit per Unit Area (TNBA) and the Marginal Benefit Value per Unit Area (MBVA). An optimal cost effective region based on the TNBA and MBVA has been proposed. As a result of continuing advances in PC processing power, the population density can also be included in the analysis by modelling all receivers that are potentially “noise affected”.

1 Introduction Communities living adjacent to railway lines and major roadways are demanding much higher transparency and assurance of an equitable outcome in the determination of noise mitigation measures for new and upgraded projects. This is particularly the case in New South Wales (NSW), Australia. For new and upgraded projects, proponents are required to undertake an operational assessment of the existing and future noise levels and implement feasible and reasonable mitigation measures to minimise project-related impacts. As part of the approval process, the operational noise assessment and mitigation report is made available to the public and other stakeholders. The community and planning agencies are encouraged to provide feedback on the proposal (including the proposed mitigation measures). In most instances, the extent and height of noise barriers is a topic of hot debate and the proponent is required to provide assurances that the proposed mitigation measures have been determined in a fair, open and reasonable manner. The optimisation of noise barriers is usually undertaken using an iterative approach whereby the height and extent is minimised to achieve compliance at all receiver locations, or if the noise goals cannot be achieved, the shape of the noise barrier is optimised to minimise costs [1]. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 243–249, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

244

C. Weber and K. Atkinson

In this paper, an alternative noise barrier optimisation approach is proposed whereby the extent and heights are based on the calculation of the Total Noise Benefit per Unit Area (TNBA) and the Marginal Benefit Value per Unit Area (MBVA). At locations where the TNBA and MBVA are above a certain threshold, the noise barrier construction is considered to be reasonable and cost effective. At locations where the TNBA and MBVA are below the threshold, the noise barrier construction is not considered to be reasonable or cost effective.

2 Operational Noise Goals Until April 2007, the noise goals for railway operations in NSW were as follows [2]: Planning Levels LAmax,fast 80 dBA LAeq(24hour) 55 dBA

Maximum Levels LAmax,fast 85 dBA LAeq(24hour) 60 dBA

The above noise goals were evaluated at a distance of 1 m from the most affected residential building façades. For the purposes of this paper, the noise levels above the “planning levels” are considered to be “noise affected” and noise levels above the “maximum levels” are considered to be “acutely affected” (note that these terms are not currently in use by either the rail operator or regulator). In terms of the noise barrier optimisation methodology presented in this paper, the “noise affected” terminology has two purposes: firstly, only “noise affected” receivers are included in the noise catchment area where the barrier is being optimised; and secondly, once the noise levels are below the “noise affected” levels, any additional noise benefit provided by an increase in barrier height is disregarded in the calculation of the TNBA and MBVA.

3 Computer Noise Modelling With the recent advances in PC technology, it is now possible to undertake point receiver noise modelling at thousands of receiver locations adjacent to a project area in a fast and efficient manner. Because the operational noise criteria in NSW are based on the LAmax and LAeq(24hour) noise parameters, Heggies have adopted the Nordic Rail Traffic Noise Prediction Method (1984) [3] within the SoundPLAN software program [4]. The software and calculation algorithms allow the calculation of both the LAmax and LAeq(24hour) noise levels. The calculation algorithms include the effects of distance attenuation, ground topography, ground absorption, noise shielding from cuttings, buildings and noise barriers, and train operation characteristics such as speed and the number of passbys in 24 hours. A sample scatter graph of the predicted future noise levels from a recent project in Sydney is provided in Fig. 1. The scatter graph indicates that there are a significant number of “acutely affected” receiver locations for the future noise modelling scenario without mitigation. There are also a much larger number of receivers that are within the “noise affected” zone.

A Systematic Approach for Arriving at Reasonable Heights and Locations

245

4 Locations Where Noise Mitigation Is Considered For rail upgrade projects, the determination of feasible and reasonable mitigation measures has typically been determined on a project-by-project basis with little available guidance being provided by the regulator or rail operator. More recently, there have been a substantial number of rail upgrade projects in Sydney and therefore it was considered prudent to develop a standardised approach for determining locations where noise mitigation should be considered. For recent projects, the determination of locations where noise mitigation should be considered have been based on two criteria: the first is that the future noise levels at residential receiver locations must be “acutely affected”; and the second is that the noise level increase as a result of the proposed upgrade must be greater than 2.0 dBA. In order for noise mitigation to be considered, both criteria must be satisfied. Similar criteria have been adopted by the regulator in the recently released Interim Guideline for the Assessment of Noise from Rail Infrastructure Projects [5]. Where possible, source control measures such as the introduction of new rollingstock, improved wheel and rail condition, and track lubrication are considered in preference to noise barriers or building façade treatments. For rail upgrade projects, however, source control measures usually provide only a small decrease in noise levels and may take a long tome to be realised. Because noise barriers create an immediate noise benefit, these have been favoured over recent years as the primary mitigation measure. LAmax Noise Levels LAmax Noise Goals (85/80)

LAeq(24hour) Noise Levels LAeq(24hour) Noise Goals (55/50)

Sound Pressure Level (dBA)

100 90 80 70 60 50 40 31.0

31.5

32.0

32.5

33.0

33.5

34.0

34.5

35.0

Track Chainage (km)

Fig. 1. Typical noise modelling output from a recent project in Sydney showing the point receiver LAmax and LAeq(24hour) noise levels compared with the noise goals (without mitigation)

246

C. Weber and K. Atkinson

5 Noise Barrier Optimisation At “acutely” affected locations where the increase in noise levels as a result of the upgrade project is greater than 2.0 dBA, noise barrier optimisation is undertaken to determine the extent and height of noise barriers. The noise barrier optimisation is similar to the procedure described in Practice Note iv of the NSW Roads and Traffic Authority’s Environmental Noise Management Manual [6]. Modifications to this procedure include more detailed guidance in relation to the interpretation of the TNBA and MBVA curves and the establishment of a lower cut-off point where the noise benefit of the barrier is no longer considered (i.e. where receivers are no longer “noise affected”). The assessment process is as follows: a. Identify all noise sensitive receivers in the catchment area (i.e. where the predicted future noise levels without noise mitigation are “noise affected”). b. Break-up the large catchment area into smaller, manageable sub-catchments. The sub-catchments are typically 200 m to 400 m long sections encompassing sensitive receiver locations having a similar geographic environment and exposure to railway noise emissions. This allows the barrier height in each sub-catchment to be optimised separately. c. For each sub-catchment, determine the number of noise sensitive receiver locations that are “noise affected”. Also determine the extent to which the predicted noise levels are “noise affected” without mitigation. d. For each sub-catchment, calculate the overall noise levels and noise benefit for barriers of varying height in 0.5 m increments (i.e. barriers of height 0.5 m, 1.0 m, 1.5 m, …). e. For each sub-catchment, calculate and plot the Total Noise Benefit (TNB) for each barrier height. The TNB is the sum of the dBA reductions (LAmax plus LAeq(24hour)) achieved at all noise-sensitive receivers within each sub-catchment for the barrier height. Note however, that the dBA reductions are only summed whilst the noise levels are above the “noise affected” thresholds. f. For each sub-catchment, calculate the Marginal Benefit Value per Unit Area (MBVA) and Total Noise Benefit per Unit Area (TNBA) for each barrier height. The MBVA represents the increase in TNB per unit increase in barrier area. (The method-ology assumes barrier costs are proportional to barrier areas, even though other factors such as barrier material will also have an influence on costs.) The TNBA represents the TNB per unit area of the barrier in the sub-catchment being examined. g. Following calculation of the above parameters (TNB, MBVA and TNBA), the data is plotted against barrier height to determine peaks in cost effectiveness. Peaks in the MBVA curve correspond to barrier options with the greatest marginal cost effectiveness. Peaks in the TNBA curve correspond to the barrier options with the greatest overall cost effectiveness, compared with the other barrier height options being considered. The cost effectiveness curves are reviewed to determine the required noise barrier height. In order for a noise barrier to be considered costeffective, both the MBVA and TBNA values must exceed 0.2 dBA per square metre of noise barrier (see Section 6).

A Systematic Approach for Arriving at Reasonable Heights and Locations

247

h. For each sub-catchment, determine whether the proposed noise barrier provides a minimum insertion loss of 5 dBA (LAeq or LAmax) at any of the “noise affected” receiver locations. This test is applied to avoid constructions with significant visual impact that are later criticised by the community as providing little or no acoustic benefit and is consistent with a barrier that breaks the line of sight between the source and receiver in a simple acoustic environment. The noise barrier is not considered if this minimum performance standard is not achieved. i. Review other project related feasibility and reasonableness considerations to determine whether the proposed noise barrier should be constructed or modified. These include: - Feasibility Issues - including constructability factors such as space limitations or safety concerns. - Community Considerations - including overshadowing effects, loss of outlook, damage to existing vegetation or vandalism concerns. - Maximum Barrier Height - maximum barrier heights are limited to 4 m in NSW. - Acoustic Treatment of Individual Dwellings - by adopting the procedures described in this paper, it follows that the extent of noise barriers may not be sufficient to adequately mitigate railway noise emissions at all noise sensitive receiver locations. At such locations, noise mitigation in the form of acoustic treatment of individual dwellings may be required to achieve a reduction in the internal noise environment. Number of Exceedances

Number of Exceedances

Number of 'noise affected' receivers

Number of 'acutely affected' receivers

50 40 30 20 10 0 0m

0.5m

1.0m

1.5m

2.0m

2.5m

3.0m

3.5m

4.0m

4.5m

5.0m

TNB and MBV (dBA)

Total Noise Benefit (TNB) and Marginal Benefit Value (MBV) Sum of dBA 'noise affected' exceedances

TNB

MBV

600 500 400 300 200 100 0 0m

0.5m

1.0m

1.5m

2.0m

2.5m

3.0m

3.5m

4.0m

4.5m

5.0m

4.5m

5.0m

Total Noise Benefit and Marginal Benefit per Square Metre of Barrier TNBA and MBVA (dBA/m2)

Non Cost-effective Region

TNBA

MBVA

3.0 2.5 2.0 1.5 1.0 0.5 0.0 0m

0.5m

1.0m

1.5m

2.0m 2.5m 3.0m Barrier Height (m)

3.5m

4.0m

Fig. 2. Typical noise barrier optimisation calculation. For this example, the selected noise barrier height is 3.0 m.

248

C. Weber and K. Atkinson

6 Selection of Noise Barrier Height Following the calculation of the TNBA and MBVA, and considering other feasibility and reasonable issues, the recommended noise barrier is the highest noise barrier for which the TNBA and MBVA are both greater than 0.2 dBA/m2. The cost-effectiveness value of 0.2 dBA/m2 was determined on the basis of experience and the typical range of urban and suburban environments within the Sydney metropolitan area. For a sub-catchment with an average distance of 20 m between dwellings, the cost-effectiveness value of 0.2 dBA/m2 equates to a 10 dBA noise reduction (LAmax + LAeq(24hour)) for a noise barrier height of 2.5 m. For residential spacings greater than 20 m or noise reductions less than 10 dBA for the same barrier, the construction is not considered to be cost-effective. Fig. 2 and Fig. 3 provide a summary of the noise barrier optimisation for two subcatchment areas adjacent to a recent project in Sydney. Number of Exceedances

Number of Exceedances

Number of 'noise affected' receivers

Number of 'acutely affected' receivers

15 10 5 0 0m

0.5m

1.0m

1.5m

2.0m

2.5m

3.0m

3.5m

4.0m

4.5m

5.0m

TNB and MBV (dBA)

Total Noise Benefit (TNB) and Marginal Benefit Value (MBV) Sum of dBA 'noise affected' exceedances

TNB

MBV

120 100 80 60 40 20 0 0m

0.5m

1.0m

1.5m

2.0m

2.5m

3.0m

3.5m

4.0m

4.5m

5.0m

4.5m

5.0m

Total Noise Benefit and Marginal Benefit per Square Metre of Barrier TNBA and MBVA (dBA/m2)

Non Cost-effective Region

TNBA

MBVA

0.3 0.2 0.1 0.0 0m

0.5m

1.0m

1.5m

2.0m 2.5m 3.0m Barrier Height (m)

3.5m

4.0m

Fig. 3. Typical noise barrier optimisation calculation. For this example, there are no barrier heights where both the TNBA and MBVA are greater than 0.2 dBA/m2.

A Systematic Approach for Arriving at Reasonable Heights and Locations

249

For the first example (Fig. 2), the TNBA indicates that the noise barrier is cost effective (i.e. greater than 0.2 dBA/m2) for all barrier heights up to 5 m. On the other hand, the MBVA is only cost effective for noise barrier heights up to 3.0 m high. The extra 0.5 m barrier height between 3.0 m and 3.5 m is not considered to be costeffective on the basis that the MBVA is less than 0.2 dBA/m2 (i.e. there is very little additional noise benefit per cost unit). For the second example (Fig. 3), the TNBA is less than 0.2 dBA/m2 for all barrier heights. On this basis, the construction of a noise barrier at this location is not considered to be cost-effective.

7 Conclusion The noise barrier optimisation process described in this report has been used by Heggies on a number of recent rail upgrade projects in the Sydney metropolitan area. By adopting the same assessment methodology, it has been possible to undertake a fair, reasonable and clearly documented assessment of feasible and reasonable noise barrier heights across different project areas and railway lines. The noise barrier optimisation process has by and large been accepted by the regulator as it is clearly documented and is seen to treat all locations in an equitable manner. In Heggies experience, the output of the noise barrier optimisation process (i.e. noise barrier height) passes the “common sense” test.

References [1] SoundPLAN wall design. Braunstein + Berndt GmbH, http://www.soundplan.com/modul14.htm [2] State Rail Authority of New South Wales, Rail Related Noise and Vibration Issues to Consider in Local Environmental Planning - Development Applications and Building Applications (1995) [3] Council of Nordic Ministries, Railway Traffic Noise – Nordic Prediction Method (1996) [4] Braunstein + Berndt, SoundPLAN English Users Manual (1996) [5] New South Wales Department of Environment & Climate Change, Interim Guideline for the Assessment of Noise from Rail Infrastructure Projects (2007) [6] NSW Roads and Traffic Authority, RTA Environmental Noise Management Manual (2001)

How Can Infrastructure Manager Influence Noise Generation of Rolling Stock M.T. Kalivoda psiA-Consult, Lastenstraße 38/1, A-1230 Wien, Austria Tel.: +43 1 8656755; Fax: +43 1 8656755 16 [email protected]

Summary Liberalisation of the European railway market dramatically changed the position of the railway infrastructure manager. A study for the Austrian Ministry of Transport came up with a proposal how to influence the noise generation of rolling stock by introducing a noise related track access charge. Although all cost benefit studies show that investments in noise reduction at the source are more effective than investments into noise barriers from the macro economic view, there is no economic benefit at the moment for train operators and rolling stock owners from investments on noise attenuation. This situation can be changed by including noise generation in the track access charge. Different approaches of relating track access charge to noise generation have been studied and assessed and will be presented by this paper. The system proposed includes the existing administrative and highlight the changes that are necessary as well as the acoustical aspects of basing the charge on type testing or on actual measured noise data.

1 Introduction Railway noise in general and rail freight noise at night in particular is a severe environmental problem. The Position Paper on the European strategies and priorities for railway noise abatement [1] states that “Railway freight traffic is the main contributor to the noise problems of the European Railways” and that “there is a high potential for the reduction of railway noise in Europe.” The authors of the position paper also say that “although the technical instruments for a considerable reduction of the freight noise problem are available, the main problem is the economically viable implementation of the noise abatement measures”. Austria has been the first European country limiting noise generation from rolling stock in 1993 by introducing the SchLV [2] ordinance. Only rolling stock registered in Austria was covered by this legislation so due to the international and interoperative character of rail transport there was only a limited effect. However, recent monitoring results (Kalivoda/Jaksch [3]) show that rolling stock that came into operation after 1993 is less noisy than the older one and that there was an international effect as well. New locomotive and multiple unit generations for Austrian Federal Railways ÖBB had to fulfil SchLV noise limits. Since industry delivers the same kind of rolling stock to all over Europe the local generation limits had a positive European effect as well. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 250–256, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

How Can Infrastructure Manager Influence Noise Generation of Rolling Stock

251

The technical and macro economic message is clear enough. Nevertheless, it is very difficult to get existing (freight) rolling stock acoustically retrofitted on a deregulated railway market with different responsibilities and interests of the stakeholders involved. The infrastructure manager generally is responsible to observe noise reception limits along his railway lines. To prevent costly noise barriers, the infrastructure manager is in favour of a fleet emitting hardly any noise. Low noise generation of rolling stock helps the infrastructure manager to save money but will cost money for the rolling stock owners respectively the rolling stock operators. Cost benefit studies such as STAIRRS [4] always show from a macro economic point of view that it is better to invest money in noise abatement at the source than to build hundreds of kilometres of noise barriers along the railway lines. So the macro economic message is clear. Noise barriers do cause cost for the infrastructure manager and leave the train operators and vehicle owners out of the game. If we want quiet vehicles to gain the macro economic benefit, train operators and vehicle owners will have to pay the price for it with no advantage or benefit for themselves. This is the main reason why noise barriers are built and almost no retrofit measures at the vehicle are made. We have to conclude, that political measures are necessary to reward train operators and vehicle owners if they use low noise rolling stock and to help to gain the macro economic benefits.

2 Rail Noise Management by the Infrastructure Operator There are several approaches for the infrastructure manager to stimulate and influence the use of low noise rolling stock. Together with the institute for rail engineering, traffic economics and ropeways of the Technical University of Vienna, psiA-Consult accomplished a study [5] for the Austrian Ministry for Traffic on how to manage and monitor railway noise. In a first step the most popular options for railway noise reduction at the source have been described and assessed. 2.1 Speed Reduction Reduction of pass-by speed is reasonably effective. Lowering speed of freight trains from 100 to 80 km/h will decrease the noise generation by about 3 dB(A), a decrease from 80 to 60 km/h will give a further 3.7 dB(A) reduction. Apart from the acoustic effect of this measure, a number of political and economic questions remain. Is it really the overall goal of infrastructure to reduce speed? There might be some specific situations where a speed reduction leads to a more harmonic operation and increases line capacity. Generally speaking speed reduction is no preferred option. 2.2 (Noise) Emission Ceiling Introducing a noise emission ceiling for a line will allow operating either a few noisy trains only or lots of low noise trains. Apart from the noise perception aspect - do some noisy trains cause the same perception as many quiet trains? - the big disadvantage of this option is that the infrastructure manager has no real chance to influence the fleet and to make use of his operative line capacity. If some noisy trains arrive early at night the ceiling could be reached and the line has to be closed. This is no good option in the light of transport policy.

252

M.T. Kalivoda

2.3 Ban of Noisy Vehicles Doubtlessly, to ban noisy vehicles is a very effective way of reducing noise. This option, however, raises a lot of critical questions. How to handle a train with few noisy vehicles in it? We need shunting yards at the beginning and end of such a line or a section to get rid of the noisy vehicles. Do we really want to stop a freight wagon this way? Will such a procedure stimulate the modal shift from road to rail? And finally, is it legally possible to reject a vehicle because it is (only) noisy? This is the reason why this option is not acceptable. 2.4 Noise Related Track Access Charge Including the noise generation into the track access charging would create an incentive to use low noise rolling stock. This option gives the vehicle owner and train operator a financial benefit if he invests into noise reduction. This approach is non-discriminatory since it only rewards the plus of noise reduction regardless if it is a new or retrofitted vehicle. Therefore, this option has been looked at more closely in our study.

3 Proposal for a Noise Related Track Access Charge (TAC-N) 3.1 Noise Classification of Rolling Stock 3.1.1 Classification According to Constructional Features Classification according to constructional features is the simplest method to classify rolling stock. No noise measurements are needed; only the knowledge of the braking technology. Vehicles with cast iron block brakes will be noisy ones and vehicles with disc, Sinter, K- or LL-bloc brakes will be the quiet ones. The simplicity of this method is its big advantage. However, the quiet technologies will have a spread of 5 to 8 dB(A) which is not taken into account. Also the actual noise generation which can differ due to damages or good maintenance is not taken into account. 3.1.2 Classification According to Type Testing Level Taking the type testing level as baseline for the calculation of noise related track access charge will make the system more selective. There are more than 2 classes and actual noise generation at the time of type testing and differences between several products are taken into account. There is no additional measurement effort for new vehicles since TSI Noise [6] certificates are needed anyway. For the existing fleet a default classification according to constructional features can be made. If a vehicle owner feels that his vehicle is less noisy than the default class he can present a noise certificate and be re-categorised. Actual noise generation, however, is not taken into account, still. 3.1.3 Classification According to Noise Generation in Service Actual noise generation during regular service is the most accurate way of classification. It does not take into account differences in noise generation due to the general design of the rolling stock but also the maintenance and irregularities like wheel flats. If one wants to use this method of noise classification one must be able to monitor noise generation permanently and reliably.

How Can Infrastructure Manager Influence Noise Generation of Rolling Stock

253

3.2 Principles of a Noise Related Track Access Charge 3.2.1 Noise Generation Is a Bonus of the Track Access Charge This principle is easy to explain: quiet vehicles pay less, noisy ones pay full track access charge. As a consequence the infrastructure manager will have less and less income when the fleet becomes less noisy. That means a regulatory body is needed which takes care of the macro economic effects, watches the development of TAC-N and refunds the lack of income from money that is saved from noise barrier constructions. 3.2.2 Noise Criterion A-weighted pass-by level measured according to TSI Noise [6] or ISO 3095 [7] at 80 km/h is the noise criterion used to determine the TAC-N. This level can be gained from a type testing certificate as well as from a measurement in service. The pass-by level is referenced to 80 km/h so speed is not taken into account since basically it is an operative parameter which can be influenced by the infrastructure manager anyway. 3.2.3 Height of the TAC-N Bonus The TAC-N bonus depends on the additional improvement in noise reduction. State of the art noise levels will not be rewarded. If a vehicle is better than the state of the art track access charge is reduced. The better the improvement the higher the bonus is. As a consequence an existing vehicle (pre-TSI) after retrofit will get a different bonus than a new vehicle (post-TSI) that has to comply with TSI-Noise anyway. That is easy to understand. An existing freight wagon that has a pass-by level of 82 dB(A) after retrofit has to be rewarded. A new freight wagon does not get a bonus if it meets the 82 dB(A) TSI-Noise limit. A bonus is paid only if the new freight wagon has a passby level lower than the 82 dB(A) required. Different vehicle types will also get different boni. It is easier to achieve 78 dB(A) pass-by level for a coach than for a Diesel loco so the bonus for a 78 dB(A) Diesel loco will be higher than for a coach. TAC-N bonus can be linear or progressive. There is some good reason to choose a non-linear bonus since the first dB of reduction is easy to achieve. The higher the reduction will be the greater the costs for it. A progressive bonus takes this effect into account.

4 Monitoring of Noise Generation from Trains in Regular Service 4.1 Railway Noise Monitoring System acramos® In July 2006 an automatic railway noise monitoring station, called acoustic railway monitoring system acramos®, has been installed near Deutsch Wagram at the Nordbahn north of Vienna. This project is sponsored by Austrian Federal Railways - Infrastructure and the Federal Ministry for Transport, Innovation and Technology. acramos® has been designed and installed by psiA-Consult in cooperation with Woelfel Messsysteme Software GmbH & Co and includes a number of innovative features to monitor noise and vibrations from trains and rolling stock in daily operation:

254

M.T. Kalivoda

• Axle patterns are recorded in parallel with pass-by noise and vibrations. The parallel recording allows matching noise and vibration data with the position of the train. • Rail vibrations are recorded in parallel with the pass-by noise signals. Pass-by noise can be recovered from vibration signals in adverse environmental conditions (wind, rain, snow). • Two identical measurement cross-sections are used in 4,2 m distance to increase reliability and reproducibility of results. The distance of 4,2 m represents about 1,5 times the wheel circumference. Each measurement section includes 1 microphone in 7,5 m distance from the track and 1,2 height above rail surface, 3 accelerometers for vertical and lateral railhead vibration and vertical sleeper vibration and 1 inductive wheel sensor (Fig. 1). At the moment three more signals are added from a triax geophone to include ground borne vibrations in 12 m distance from the track. One main feature is the automatic train categorisation that is based on the axle pattern. The measurement system includes a database with axle patterns of about 30 different train categories and locomotive classes. Between mid July 2006 and mid November 2006 about 11.900 trains with 478.000 axles have been recorded. 4.2 Results from acramos® 4.2.1 Speed Dependent Pass-By Level Per Train Category Fig. 2 shows the speed dependent pass-by level for different train categories. Results are gained from measurements under track and site conditions that comply with TSINoise. This diagram shows very clearly the progress in railway noise abatement technology. S-trains class S4024 and regional trains class 80-33 and 80-73 represent modern vehicles. Class 4024 is an EMU with small wheels and wheel disc brakes, class 80-33 is disc braked double deck regional coaches and class 80-73 is retrofitted (Kbloc) regional coaches. Average noise emission under normal operation conditions of these 3 fleets stay 2 to 4 dB(A) under the TSI-Noise limit for DMUs. The other 3 categories represent old vehicles with cast iron or combined brakes.

Fig. 1. Image from one cross section

Fig. 2. Speed dependent A-weighted pass-by level in 7,5 m distance of the track for TSI-Noise track conditions

How Can Infrastructure Manager Influence Noise Generation of Rolling Stock

255

4.2.2 Noise Generation during Night from Different Train Categories Categorised monitoring data easily can be used to generate a roadmap for noise abatement and to show where to invest money first. Fig. 3 shows an example of Aweighted equivalent level for the night per train category. It is obvious that mainly category I. is responsible for the total noise generation. It does not make any sense to reduce noise of any of the other categories before noise from category I. has been cut. Simulation shows that lowering pass-by level of category I. by 10 dB(A) will reduce the total level by 8 dB(A). Introducing best practice for all categories in a second step will lessen the total level further by 3 dB(A).

Fig. 3. Emission levels during night (22:00 - 6:00) for 5 train categories

4.2.3 Braking Noise Recordings of the speed per axle allow analysing braking noise. This analysis shows that braking leads to an average increase of A-weighted pass-by level by 3 dB(A). That number does not include the high frequency brake squealing before the wheel is stopping but the level increase according to the rubbing of the cast iron block on the surface of the fast rolling wheel. Further data analysis showed that it is not necessary to include the rate of deceleration. Results did not improve significantly after including deceleration in the regression.

5 Conclusion Increasingly, railway infrastructure operators will collect safety related as well as environmental data from the trains running on their networks. Especially noise creation data will be used in the future to charge for costs generated by noisy rolling stock causing the infrastructure manager to build noise barriers. Within the research framework “Innovative System Bahn” (innovative railway systems) the Austrian Federal Ministry of Transport, Innovation and Technology sponsors the development of a methodology for automated permanent railway noise monitoring of train in daily operation.

256

M.T. Kalivoda

For the Austrian Federal Railways (ÖBB) Infrastructure, psiA-Consult GmbH has developed the acramos® monitoring system, which is able to measure pass-by noise and ground borne vibration from trains in daily operation automatically and vehicle selectively. The acramos® system is able to recorded axle patterns in parallel with passby noise and vibrations. The parallel recording allows matching noise and vibration data with the position of the train and to assign emissions measured to generic train categories. The monitoring system has been installed in July 2006 at Nordbahn line next to Vienna and has gained a lot of useful data and information in the meantime.

Acknowledgements The author would like to thank the Austrian Federal Ministry for Transport, Innovation and Technology as well as the R&D unit of Austrian Federal Railways, Infrastructure, BauAG for the support of the projects presented.

References [1] European Commission (ed.): Position Paper on the European strategies and priorities for railway noise abatement. ISBN 92-894-6055-5, Brussels (2003) [2] SchLV: Bundesgesetzblatt der Republik Österreich, SchienenfahrzeugLärmzulässigkeitsverordnung, BGBL Nr. 414/1993, Wien (1993) [3] Kalivoda, M.T., Jaksch, M.: Safety – Instability – Noise (project), Bahnlärm-Monitoring, final report (DE), Vienna (March 2007) [4] STAIRRS: Strategies and Tools to Assess and Implement noise Reducing measures for Railway Systems; final report, Utrecht (2003) [5] Kalivoda, M.T., et al.: Studie zur Entwicklung der Methoden für ein automatisches Bahnlärm-monitoring und –management, final study report (DE), Vienna (May 2006) [6] TSI NOISE: Commission decision of 23 December 2005 concerning the technical specification for interoperability relating to the subsystem ‘rolling stock - noise’ of the transEuropean conventional rail system (notified under document number C(2005) 5666) (2005) [7] ISO 3095:2005: railway applications – Acoustics – Measurement of noise emitted by railbound vehicles. 2005-11-01 (2005)

Acoustic Effectiveness of Damped Wheels and Impact on Life-Cycle Cost of Different Typologies of Passenger Trains A. Bracciali1, S. Cervello2, and P. Gatti3 1

Università degli Studi di Firenze, via Santa Marta, 3 - 50139 Firenze - Italy Tel.: +39 347 2429240; Fax: +39 055 4564064 [email protected] 2 Lucchini Sidermeccanica S.p.A., via G. Paglia 45 – 24065 Lovere (Bg) – Italy Tel.: +39 035 963483; Fax: +39 035 963324 [email protected] 3 ALSTOM Ferroviaria S.p.A., via Ottavio Moreno 23, 12038 Savigliano (Cn) - Italy Tel.: +39 0172 718374; Fax: +39 0172 718383 [email protected]

Summary Railway noise reduction at the source is the preferred way indicated by the European legislator, although certainly the hardest to be obtained from an engineering point of view. A cost-benefit analysis of the available solutions should drive the political decisions. This paper describes the impact of a specific low noise wheel, with noise reduction effects proven in normal service, on the Life-Cycle Cost of different types of vehicles. The analysis is conducted on several types of passenger rolling stock, i.e. high speed trains, both conventional and tilting, Electrical and Diesel Multiple Units.

1 Introduction Rolling noise is the main source of railway noise in a wide range of speeds, approximately up to 250÷300 km/h where aerodynamic noise becomes prevailing. To reduce rolling noise it is necessary either to reduce the wheel and rail combined roughness or the radiation properties of the bodies involved. Unfortunately the combined roughness is not adjustable by changing transverse profiles but only by grinding the rails, a methodology that is valid only when wheel and rail roughness are similar, i.e. for disc braked wheelsets. Measure to reduce wheel roughness for tread braked vehicles are not described here. Measures acting on track radiation are quite various and can be complicated; it was shown that carefully designed rail dampers can noticeably reduce rail noise (up to approximately 5 dB) but that their effect on overall noise is much lower (in the order of 2÷3 dB) if not used in conjunction with low noise wheels [1]. This paper deals with damped wheels, which are used in commercial service with relevant overall noise reduction. One of the classical arguments against the use of “low noise” wheels is their extra cost as both first fitting and spare parts, neglecting not only all the other associated costs (maintenance, disposal) but, even more important, the advantages given by their use and the associated savings. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 257–263, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

258

A. Bracciali, S. Cervello, and P. Gatti

While it is clear that in very loud situations the use of low noise wheels can not solve the noise problem, in many “border” cases their use can save the erection of noise barriers, of which the cost is nowadays in the order of 1 M€€ /km of double track. The paper will show how the costs associated with the use of damped wheels are only a fraction of that value, even if considered for the entire life of a train.

2 The Lucchini Sidermeccanica Syope® Wheel: Description, Performances, Considerations Lucchini Sidermeccanica SpA approached the low noise wheels field in 1995, and the Syope® treatment was readily developed to provide levels of damping much higher than the rolling damping [2]. The treatment consists of a steel layer constraining a special adhesive polymer sheet developed by 3M attached to the wheel web. The Syope® treatment can be retrofitted on any existing axial symmetric disc-braked wheel mounted on the axle at room temperature (press-fit) [3]. A modified version of the treatment, named Syope Braw®, is under development for wheels with web-mounted brake discs [4]. The behaviour in service and during numerous test campaigns is described in a number of papers [5, 6, 7, 8, 9,10] to which the reader is referred for further details. Roughly speaking, the observed LpA,max (overall A-weighted maximum level, measured at 25 m from the track axis with the Fast time constant during passby) reduction, treating the wheels of an existing vehicle with the Syope® treatment is in the order of 3 to 5 dB, without any treatment or modification applied to the track. The use of “special” wheels is always debated as the inevitable associated costs and complications (maintenance, disposal, non-destructive testing, risks associated to objects mechanically mounted on the wheel, etc.) must be carefully addressed. Since the first tests in 1998 on the Fiat Ferroviaria SpA (now Alstom Ferroviaria SpA) train set ETR470-0, the experience gained in several years of testing and service allows to say that: • • •

some wheels travelled for more than 5 years before the end of their useful life and for more than 1.1 million kilometres without any reduction of safety (i.e. detachment of the steel constraining plate); no special attention or procedure was needed during the service. For the final user, the wheel treatment can be defined as a “fit and forget” measure that effectively reduces noise without any impact on maintenance procedures; withdrawal and disposal followed the usual procedures as no risks or additional costs are associated to the treatment. Scrap steel can be re-melt in the electric arc furnace without generating noxious gases (dioxin).

3 Life Cycle Cost Analysis and the Implication of Syope® Wheels Life-cycle costs (LCCs) are all the anticipated costs associated with a project or program alternative throughout its life. This includes costs from pre-operations through operations or to the end of the alternative. By applying the principles of LCC analysis, it is possible to evaluate several designs and select the one with the lowest LCC [11,12]. The definition of the typical system profile is crucial. It is generally recognized that it is

Acoustic Effectiveness of Damped Wheels and Impact on LCC

259

necessary to perform an LCC analysis early in a project’s life. This is particularly evident in those systems for which the operational support forms a substantial part of the LCC. An example of the impact of the R&D development is shown in Fig. 1, together with an LCC profile for system acquisition where operation and support costs are the greatest part of the life-cycle cost. Rolling stock purchasing is a complex activity that involves many technical, economical and environmental issues. Train operating companies (TOCs) have developed in the years, often with the help of consultancy experts (typically from defence), their own models to evaluate LCC. While R&D, engineering and manufac-turing costs are relevant to the industry, operating costs are associated with normal service and therefore are sustained by TOCs that, on their side, control the efficiency of their rolling stock in terms of RAM indices (Reliability, Availability and Maintainability). The procurement of new rolling stock is therefore normally accompanied with a Technical Specification including an in-depth decomposition of the vehicle and the allocation of availability, maintainability and reliability with associated indices (MTBF, MTTR, and the like).

Fig. 1. Actions affecting LCC (left) and typical LCC profile for system acquisition (right) (from [12])

At the same time, an analysis of the faults and of their impact on service has to be done with the classical FMECA (Failure Mode, Effects, and Criticality Analysis) applied to the Function Block Diagram (FBD) of the vehicle in order to find the Reliability Block Diagram (RBD). This last document is particularly important as redundancy is taken into account and the calculation must be performed on all the mission profiles identified by the customer. For the railway sector, criticalities of the failures are commonly classified on the type of service disruption that is consequent to the failure. Maintenance is commonly split in preventive (programmed, on-condition and predictive) and corrective maintenance. All the associated activities have an impact on the LCC, both in the design phase, where all the problems should be addressed at their best, and in the operation phase. The Syope® wheel is normally proposed as an additional feature to be applied on existing or new rolling stock. Limiting the analysis to new rolling stock, for which an LCC analysis can be conducted “from the cradle to the grave”, it is fundamental to

260

A. Bracciali, S. Cervello, and P. Gatti

evaluate the impact of the treated wheel on a vehicle’s cost, including all direct and indirect aspects. This identification is fundamental in order to understand if the use of such wheels requires a different LCC model or if it is sufficient to input in the model the extra cost of the treated solution. The activities that can be linked to the pre-delivery phase of a new rolling stock, already shown in Fig. 1, may be affected by the use of a damped wheel. They can be analyzed as follows: •

•

• •

concept formulation: if the new train typology allows the use of damped wheels, there are no additional activities or costs associated to their adoption as the desired wheel properties (strength, stiffness, weight) can be obtained independently from the damping; concept validation: i.e. the validation of the solution including safety assessment of the vehicle. In this case the use of the treatment is neutral as it does not affect the mechanical properties of the wheel [13]. No specific procedures need to be used to evaluate the behaviour of the wheel in service under the usual loads. No extra costs are therefore associated to this phase linked to the use of Syope® wheels; development: there is no impact on development costs of the vehicle, as the treatment introduces negligible masses (calculations remain valid); production: wheelsets installation is not affected at all by the use of the damping treatment on the wheels. There are therefore no additional costs associated in the vehicle production phase.

As shown in Fig. 1, operations can be a relevant part of the total LCC. Also in this case it is necessary to evaluate the effects of the use of a damped wheel on the forecast overall cost of the vehicle for its entire life. Activities and consequences can be analyzed as follows: •

•

•

availability: that the use of damped wheels does not affect in any way the fleet availability, i.e. there are no failures associated that can reduce this parameter. No extra costs are therefore anticipated to take into account the possible reduction in the availability associated to the use of damped wheels; reliability: no reduction in the reliability of the wheelset subassembly is to be expected by the use of damped wheels. Being a non-structural treatment, the constrained layer damping technique leaves unaltered the behaviour of the wheel, neither introducing new modes of failure nor changing the failure rates for the usual wheels. There are therefore no associated costs; maintainability: wheelset maintenance is a complex subject, but it can be said that for routine operations (for example ultrasonic testing or visual check), no modifications to the standard maintenance routines is requested. Wheel reprofiling can be done with the usual tools and machine tools (either underfloor or parallel lathe). The use of damped wheels leads to no modifications in the LCC calculation scheme.

As a consequence, Syope® wheels are “neutral” for the LCC calculation that can be used by simply increasing the cost of the wheel of the amount related to the damping treatment. Also for the disposal at the end of life, there are no additional costs associated.

Acoustic Effectiveness of Damped Wheels and Impact on LCC

261

4 Simulation of the Impact of Syope® Wheels on LCC Alstom Ferroviaria SpA, one of the companies of the Alstom Transport group, is the manufacturer of all the tilting trains running in Italy, and the same will apply with the new Trenitalia ETR600 and Cisalpino ETR610 (tilting train, max speed 250 km/h). For reasons that will not be analyzed here, there is a general tendency to purchase passenger rolling stock with distributed power with electric or diesel power (the so called EMU or DMU, i.e. Electrical or Diesel Multiple Units). These train sets are normally seen as a whole, simplifying the management and the effect of low noise vehicles. The calculation of the LCC of a train is a complex activity that can not be done with the goal of writing a paper. The only possibility to have an idea of the impact of the application of damped wheels is to use already prepared LCC calculation schemes, simply by introducing the corresponding extra cost of the wheel. A fully developed and validated LCC model was available for four types of vehicles: • • • •

a high speed (vmax=250 km) dual voltage tilting train with distributed power (EMU, 7 coaches, 8 wheels/vehicle), with an expected life of 25 years (three wheel changes expected); a high speed conventional trainset with distributed power (EMU, 7 coaches, 8 wheels/vehicle, vmax=250 km/h) , with an expected life of 25 years (three wheel changes expected); a regional train (EMU, 3 coaches on 2 motor bogies + 2 Jacobs bogies, 16/3=5.8 wheels/vehicle), with an expected life of 30 years (three wheel changes expected); a regional train (DMU, 3 coaches on 2 motor bogies + 2 Jacobs bogies, 16/3=5.8 wheels/vehicle), with an expected life of 25 years (two wheel changes expected).

The LCC model includes numerous sensible data that cannot be detailed here for confidentiality; it is anyway important to highlight that the estimation is conducted on trains already in service or in the delivery phase, and this ensures the maximum validity to the calculations. Different approaches to maintenance and different types of service are included in the simulations, and this gives the calculation an even greater validity, clearly slightly increasing the spread in the results. Both the impact on maintenance cost, which is important to economically compare the application of low noise wheels on an existing fleet and the use of noise barriers, and the impact on LCC on the new rolling stock, including the cost of the new train set, were evaluated. The latter value is especially most important as it quantifies the overall impact of the damped wheels for the entire useful life of the train. The results are shown in Table 1. From these values it can be concluded that: • •

costs associated to the trainset range from 0.26% to 0.87% of the LCC. It is likely that for more complex trains this value will further decrease, while for simpler trains it could increase; impact on operation costs only are similarly quite variable (from 0.54% to 2.3% of the LCC) mainly depending on the complication of the considered vehicle. Not surprisingly, the highest value is obtained for a rather simple vehicle (a trailing car with little equipment of a high speed non-tilting EMU) while for the lowest value is obtained for a motor car of a DMU trainset whose complication and associated operating costs are inevitably higher.

262

A. Bracciali, S. Cervello, and P. Gatti Table 1. Evaluation of the impact of damped wheels on operation costs and on LCC

Tilting EMU (250 km/h) Train set (7 cars) Motor car (#1) Trailer car (#4) EMU (250 km/h) Train set (7 cars) Motor car (#1) Trailer car (#4) Regional EMU Train set (3 cars) Motor car (#1) Trailer car (#2) Regional DMU Train set (3 cars) Motor car (#1) Trailer car (#2)

Impact on operation costs %

Impact on Life-cycle cost %

1.5% 1.3% 2.0%

0.74% 0.66% 0.87%

1.6% 1.4% 2.3%

0.79%

1.27% 1.08% 1.14%

0.35%

0.64% 0.54% 1.55%

0.26%

To provide absolute figures, the LCC cost of the fleet of 20 non-tilting EMU is estimated in the order of 405 M€€ , 5.8 M€€ of which are the extra cost associated with the use of the Syope® treated wheels.

5 Conclusions The use of damped wheels with low noise emission is often debated as their use is inevitably linked to extra costs for wheel purchase. The evaluation of this alternative is only partly satisfactory as wheels are an important component whose cost contributes to the total LCC of the train together with many other important factors. The evaluation of the extra costs associated to the use of damped wheels was therefore performed on different trainsets for high speed, long distance and regional passenger trains (either with electrical or Diesel traction) showing that the impact on costs is limited and that there is a distinct advantage in using low noise wheels in those situations where noise limits are not respected for a few decibels.

Acknowledgements The authors are grateful to Lucchini Sidermeccanica and Alstom Ferroviaria companies for the permission to publish the results.

Acoustic Effectiveness of Damped Wheels and Impact on LCC

263

References [1] Létourneaux (AEF), F., Margiocchi, F., Poisson, F.: Complete Assessment of Rail Absorber Performances on an Operated Track in France. In: Proceedings of the World Congress on Railway Research WCRR 2006, Montreal (September 2006) [2] Krylov, V. (ed.): Noise and Vibration from High-speed Trains. Thomas Telford Ltd., London (2001) [3] Bracciali, A., Cervello, S., Laganà, A., Villa, V.: Low Noise Wheel: From Design To Applications. In: Proceedings of the 15th International Wheelset Congress, Prague (September 2007) [4] Bracciali, A., Cervello, S., Moroder, H.: Application of noise reduction systems to wheels with web-mounted discs. In: Submitted for publication to World Congress on Railway Research 2008, Seoul (May 2008) [5] Cervello, S., Bracciali, A.: Development of a Vibro-Acoustical Methodology for the Design of Low Noise Railway Wheels. In: Proceedings of the World Congress on Railway Research WCRR 1997, Firenze, vol. E, pp. 157–163 (September 1997) [6] Bracciali, A., Bianchi, M.: Lucchini CRS Syope® damped wheels noise qualification. In: Proceedings of the 13th International Wheelset Congress, Roma, Italy (on CD) (2001) [7] Degen, K.G., Nordborg, A., Martens, A., Wedemann, J., Willenbrink, L., Bianchi, M.: Spiral array measurements of high-speed train noise. In: Proceedings of Internoise 2001, Den Haag (September 2001) (on CD) [8] Cervello, S.: Syope promises quieter running. Railway Gazette International, 571–575 (September 2002) [9] Bracciali, A.: Damped Wheels as an Efficient Measure to Reduce Railway Noise. In: Proceedings of Euronoise 2003, Napoli (September 2003) (on CD). Reprinted on Acta Acustica (Stuttgart), vol. 89 (supp.), p. S75 (May/June, 2003) [10] Bracciali, A., Piccioli, F.: Experimental analysis of wheel noise emission as a function of the contact point location. Journal of Sound and Vibration 267(3), 469–483 (2003) [11] Fabrycky, W.J., Blanchard, B.S.: Life-cycle cost and economic analysis. Prentice Hall, Englewood Cliffs (1991) [12] United States Department of Energy, http://www.directives.doe.gov/pdfs/doe/doetext/neword/430/g4301-1chp23.html [13] Italcertifer, Relazione tecnica per ruota Syope, prot. 140 (August 28, 2002)

Mitigation Measures for Open Lines against Vibration and Ground-Borne Noise: A Swiss Overview R. Müller SBB Rail Environmental Center, Hochschulstrasse 6, Bern 65, Switzerland Tel.: +41 512205118; Fax: +41 512204475 [email protected]

Summary A new Swiss ordinance, planned to come into force in autumn 2008, will demand mitigation measures for groundborne vibration over the whole existing Swiss railway network. Swiss Federal Railways (SBB) tested several mitigation measures on open lines and in tunnels. Under ballast mats: For open lines the tracks with under ballast mats need enhanced lateral support of the ballast to stabilize the ballast. The under ballast mats on open lines show some reduced insulation efficiency especially for dam situations in contrast to tunnels. Under sleeper pads: First results in Switzerland show that insulation efficiency of under sleeper pads is very similar for open lines and tunnels and comparable to under ballast mats on open lines. Mitigation measures for switches: Measurements of SBB demonstrate that so far a movable frog for switches is not a satisfying solution for vibration control. Two recent Swiss studies adjacent to a switch illustrate that the insulation efficiency of trenches is effective with about 6 dB over a large frequency domain (20/25 Hz to 250 Hz). Finally SBB proposes a simple cost-benefit-method for vibration and ground-borne noise similar to a method used in Swiss noise abatement.

1 Under Ballast Mats (UBM) 1.1 Tests in Raron and Gampel To improve track quality two UBM-tests were constructed in 1997 in southwest Switzerland in Raron (straight line) and Gampel (curved line). The installation of UBM (cstat = 0.06 N/mm3) was carried out with an asphalt layer to improve the formation. Furthermore two types of lateral support, intended to stabilize the ballast, were tested: at one location ballast was stuck together laterally by a special adhesive and at another location a concrete board held by mini-piles was used (fig. 1.a.). Both lateral supports have worked satisfactorily since the installation. Insulation efficiency of the under ballast mats was measured in immediate proximity to the dam in Raron. The performance was poor and only significant for frequencies over 60 Hz and remained smaller than 12 dB (fig. 1.b.). Further measurements on the dam in Raron and in Gampel in a curve showed similar results. Up to now there is only one B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 264–270, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Mitigation Measures for Open Lines against Vibration and Ground-Borne Noise

265

explanation of the unsatisfactory insulation efficiency: The underground of the dam is substantially softer compared to the tunnel, this may lead to a dramatically decrease of the insulation efficiency.

Insulation efficiency (dB)

15

Raron, 8m, measurements below dam, test trains

10

5

Raron, 8m, measurements on dam, Intercity

0

Gampel, 8m, Intercity

-5

250

200

160

125

80

100

63

50

40

32

25

20

16

12

8

10

6

-10

Frequency (Hz)

Fig. 1a. Test track Raron with concrete board

Fig. 1b. Insertion loss of UBM in Raron and Gampel

1.2 Tests in Rothrist In the station area of Rothrist, an endpoint of the new high speed line MattstettenRothrist (in use since end 2004), mitigation measures in form of under ballast mats (cstat = 0.06 N/mm3) against vibration and ground-borne noise were necessary (fig. 2.a.). Measurements showed promising results (see fig. 2.b). However the insulation efficiency in the tunnel (UBM cstat = 0.06 N/mm3) was not fully achieved on the open line neither (see fig. 3.).

Insulation efficiency (dB)

20

Rothrist II, 8m, test mesasurements

15

Rothrist II, 8m, Intercity 10

Rothrist III, 8m Intercity

5

Rothrist VII, Regional train

0

Mean UBM Rothrist

-5

Mean UBM Raron+Gampel 250

200

160

125

100

80

63

50

40

32

25

20

16

12

10

8

6

-10

Frequency (Hz)

Fig. 2a. Test track Rothrist with lateral support

Fig. 2b. Insertion loss of UBM in Rothrist, mean Raron/Gampel

This leads to the following conclusions: 1.

Compared to Raron, there is sufficient insulation efficiency and the under ballast mat performs already in lower frequency ranges.

266

R. Müller

2. 3. 4.

A broader under ballast mat including a lateral mat for the lateral supports, similar to UBM in tunnels, does not noticeably improve insulation efficiency. Switches do not alter noticeably the insulation efficiency of UBM. The different tests show that mitigation measures must be selected carefully according to the site specific underground conditions.

2 Under Sleeper Pads (USP) 2.1 Tunnel The SBB started its first test with 40 meters of very soft under sleeper pads for vibration reduction (cstat = 0.035 N/mm3) in 1986 in a tunnel with wooden sleepers [1]. No problems of track stability or material deterioration have been noted after more than 20 years of use. On the new high-speed line between Berne and Olten under sleeper pads (cstat = 0.13 N/mm3) should improve the life cycle of the ballasted track in the tunnels [2]. As a side-effect of this measure some buildings are protected against ground-borne noise (see fig. 3.). 16 km single track length was equipped with under sleeper pads and concrete sleepers. Since the end of 2004 the line has been in operation with speeds up to 160 km/h and from December 2007 the maximum speed will be 200 km/h. In the tunnel of Leuk a softer material (cstat = 0.1 N/mm3) for concrete sleepers was installed in a test in 2006 designed to obtain additional information about insulation efficiency (see fig. 3.) [3]. 2.2 Open Line In Zurich main station [4] the same under sleeper pads as for the high-speed line were used for wooden sleepers. There were no conclusive results due to very low velocities in the station. Test track new high-speed line: A short test track of 100 m was constructed with under sleeper pads inside and outside a tunnel (Rüdtligen). Measurement results are shown in fig. 3. Test track in Pratteln: The test section in Pratteln for open line with train velocities up to 140 km/h shall clarify feasibility and effectiveness of very soft under sleeper pads on open line. First measurements of five intercity trains before and directly after installation show good results but further measurements for concluding remarks should be awaited. 2.3 Comparison of USP with UBM, Tunnel and Open Line The insulation efficiency measurements of SBB on open lines and in tunnels for USP and UBM are compared in fig. 3. USP show similar insulation efficiency as UBM on open lines, whereas the UBM show clearly the best effects in the tunnel. The USP on open line perform as well as in the tunnel. For UBM the insulation efficiency on open lines can be reduced by up to 10 dB because of softer underground (Raron, Gampel) compared to tunnels. It can be concluded that USP are very useful for open lines whereas UBM are very useful for tunnels if high performance is needed.

Mitigation Measures for Open Lines against Vibration and Ground-Borne Noise

267

Insulation efficiency (dB)

25

20

USP open line, Rüdtligen+Pratteln

15

USP Tunnel, Leuk+Rüdtligen

10

UBM open line, Rothrist

5

UBM open line, Raron+Gampel

0

UBM Tunnel, St.Aubin+Concise 250

200

160

125

80

Frequency (Hz)

100

63

50

40

32

25

20

16

12

8

10

6

-5

Fig. 3. Comparison of insulation efficiency for USP and UBM on open line and tunnel

3 Switches 3.1 Introduction Vibration mitigation measures for switches are difficult to find but important: over 30% of the vibration problems are due to new or existing switches. An overview of 20

Hindelbank, EW 900 Lenzburg, EW 900 Lenzburg, EW 900

15

Othmarsingen, EW900

Amplification (dB)

Rothrist Be III, EW 300

10

Rothrist Be III, EW 300 Morges, EW 900 Rupperswil Gl1, EW1600

5

Rupperswil Gl2, EW1600 Wichtrach, EW 900 Gl1,IC Wichtrach, EW 900 Gl2,IC

0

Rothrist Gl1,EW1600 Rothrist Gl2,EW1600

-5

Rubigen Gl1, EW 900 Rubigen Gl2, EW 900 Mean EW 300

250

200

160

125

100

80

63

50

40

32

25

20

16

12

8

10

6

5

4

-10

Mean EW 900 Mean EW 1600

Frequency (Hz)

Fig. 4. Additional vibration of switches compared to track at grade in about 8 m distance to the open line near the unmoving frog. The numbers after “EW” indicate curve radii of the switch in deflection (e.g. EW 300 = radius 300 m).

268

R. Müller

frequency dependent emissions of several switches is given in fig. 4. The measurements of switches in Switzerland show a substantial increase of vibration compared to a normal track, worse for low curve radius. The highest increase of vibration appears on the average from 16 Hz to 40 Hz. 3.2 Moving Frogs At first SBB tested movable frogs (fig 5.a.) to reduce additional vibration of switches on open lines. A first estimation of the insulation efficiency of old mobile frogs of switches on wooden sleepers took place in Lenzburg [5] and Othmarsingen [6]. A comparison between switches with and without mobile frogs could be accomplished for both cases. Additionally in the year 2004 a pilot project was executed in Rueschlikon where 2 switches without mobile frogs were replaced by switches with mobile frogs [7]. The vibration reduction is small for all mobile hearts: on the average it amounts to between 0 and 3 dB for the frequency range of interest from 16 Hz to 80 Hz (see fig. 5.b.). The SBB measurements demonstrate that the installation of mobile frogs is not suitable for vibration mitigation.

Insulation efficiency (dB)

15 10 5

Lenzburg, 8m Lenzburg, 8m

0

Othmarsingen, 8m Rüschlikon, 2.5m

-5

Rüschlikon, 7m Mean movable frog

-10

250

200

160

125

80

100

63

50

40

32

25

20

16

12

8

10

6

5

4

-15 Frequency (Hz)

Fig. 5a. Movable frog

Fig. 5b. Insulation efficiency of 5 switches with movable frogs (reference: switches with fixed frogs)

3.3 Trenches Trenches could be an option to solve the problem of high vibration immissions. A first test was conducted by BLS in Berne with impressive results [8]. A further trench was constructed in Berg near a switch to protect primarily a machine and secondarily further living rooms [9]. Measurements at both places show similar results as other measurements of 3 open trenches [10] and 8 concrete trenches [12] near railway tracks in Europe and Japan (see fig. 6.). 3.4 Other Mitigation Measures of Switches Alternative mitigation measures of switches were successfully tested by SBB: a. switch on heavy mass spring system in the tunnel (Zurich-Rämistrasse) b. switch on under ballast mat (e.g. measurements in Rothrist)

Mitigation Measures for Open Lines against Vibration and Ground-Borne Noise

269

c. switch on slab track (tunnel Heitersberg): measurements showed substantially fewer vibrations compared to conventional switches on ballast. d. slab underneath the ballast for new lines (Visp 3rd track) e. under sleeper pads (tests SBB in Rubigen and Stadelhofen) could create a vibration reduction by better track alignment similar to c.

Bern, BLS, 4m depth, distance of flat 4 m, comparison before/after Berg, SBB, 3.5 m depth, distance 2 m, comparison to reference

10

Berg, SBB, 3.5 m depth, distance 2 m, comparison before/after

5

Mean of 3 open trenches, Germany+Japan (depth 5.5 m, distance 3 m)

0

250

200

160

125

80

100

63

50

40

32

25

20

16

12

8

10

6

-5

Mean of 8 concrete trenches in Japan (5.4 m depth, 0.8 m thick, 3 m distance) 5

Insulation efficiency (dB)

15

Frequency (Hz)

Fig. 6. Insulation efficiency of trenches in Switzerland, Germany and Japan

4 Cost Benefit Calculation As all mitigation measures are extremely expensive, costs and benefits should be weighted systemathically in a net-wide consideration. The Swiss environmental law includes that protection measures normally must be proportional. The Swiss authorities for vibration intend in the new regulation to define proportionality dependent on: a. effectiveness of the measures b. number of protected people c. costs of the measures. SBB proposes a model [13] based on the physical effectiveness, measured in dB(A). This methodology intends to guarantee an efficient use of the financial means, i.e. the amount of money is minimized to obtain an optimum protection of the people. Different options can be compared and the most efficient be selected. The proportionality of the mitigation measures against vibrations can be described with the cost benefit index (CBI): CBI =

YearlyCosts Costs = Benefits reduction × numberofpeopleconcerned

(1)

Yearly costs: “yearly costs” are the sum of the investment costs per year (depreciation, interest) and the yearly maintenance costs of a measure. Small costs result in a good (low) CBI. Number of people concerned: All people are counted who are disturbed by vibrations above the limit value and feel the defined reduction of the measure.

270

R. Müller

Reduction: The reduction of the measure is determined as the difference between the immission values with and without measures per building (see formula 2 and 3) expressed thereby, as for noise, in dB(A) (A-weighted decibels). Day- or night time is used, depending on which is more critical. The reduction of vibration, seen in formula (3), can be based on a relationship between subjective annoyance of noise and vibration of people living next to railway lines [14]. A CBI can be calculated by formula 1 for vibration or groundborne noise separately or together by adding both reductions of formula 2 and 3. ReductionGroundborne noise = LEQwithout measure - LEQwith mesure

(2)

ReductionVibration = 18.3*log[KBrwithout measure/KBrwith mesure]

(3)

For simplicity the benefit of the reduction is not weighted but it is obvious that reductions far above the limit values give more benefit than far below the limit values.

References [1] SBB Infrastruktur Umwelt/Altlastensanierung: 19 Jahre Erfahrung mit Unterschottermatten und Schwellenbesohlung im Zimmerberg-Scheiteltunnel. 20 (December 2005) [2] Rutishauser Ingenieurbüro: Schwellenbesohlung, Bestimmung des Einfügedämm-Masses mit Schwingungsmessungen. Report for UIC working group USP. 23 (January 2007) [3] Boget, E., Köstli, K.: Elastische Lagerung von Schwellen gegen Bahnerschütterungen – ein Pilotversuch in Leuk. Schweizer Eisenbahn-Revue, 174–176 (April 2007) [4] Rutishauser Ingenieurbüro: Schwellenbesohlung Gleis 3 bis 10. Report for SBB, 23 (June 2004) [5] Müller, R.: SBB-BahnUmwelt-Center: Bewegliche Weichenherzstücke. Messungen Lenzburg. SBB Report (2000) [6] Rutishauser Ingenieurbüro: Erschütterungsmessungen Othmarsingen, Weichen mit und ohne bewegliches Herzstück. Report for SBB (June 2000) [7] Müller, R.: SBB-BahnUmwelt-Center: Messbericht zu Erschütterungsmessungen in Rüschlikon, bewegliches Herzstück. SBB Report 3 (January 2006) [8] Gartenmann Engineering: BLS Doppelspurausbau Fischermätteli-Weissenbühl: Bodenschlitz als Abschirmung gegen Erschütterungen. Report for BLS. 27 (October 2005) [9] Rutishauser Ingenieurbüro: Bodenschlitz in Berg. Erschütterungsuntersuchung. Report for SBB. 12 (April 2007) [10] Krüger, F.: Schall- und Erschütterungsschutz im Schienenverkehr, p. 200. Expert verlag (2001) [11] Melke, J.: Erschütterungen und Körperschall des landgebundenen Verkehrs (1995) [12] Yoshioka, O.: Basic characteristics of Shinkansen-induced ground vibration and its reduction measures. Wave. Proceedings (2000) [13] Oertli, J., Wassmer, D.: Rail Noise Control in Switzerland: Legislation, Environment, Politics and Finances. Journal of Sound and Vibration 193, 403–406 (1996) [14] Zeichart, et al.: Kombinatorische Wirkung von Bahnlärm und Bahnerschütterung, Zeitschrift für Lärmbekämpfung, Nr. 1 (1998)

Preliminary Analysis on Effect of Sleeper Pitch on Rail Corrugation at a Curved Track Xuesong Jin, Zefeng Wen, Qiyue Liu, and Zhongrong Zhou State Key Laboratory of Traction Power, Southwest Jiaotong University, Chengdu 610031 China Tel.: +86-28-87634355; Fax: +86-28-87600868 [email protected]

Summary This paper investigates the effect of sleeper pitch on the initiation and development of rail corrugation at a curved track when a railway vehicle passes through the curved track using a numerical method. The numerical method considers a combination of Kalker’s rolling contact theory with non-Hertzian form, a linear frictional work model and a dynamics model of a half railway vehicle coupled with the curved track. The numerical analysis examines in detail the variations of wheel/rail normal loads, the creepages, and the rail wear volume. The numerical results show that the discrete track supports cause fluctuations in the normal loads and the creepages at different frequencies. These frequencies encapsulate the sleeper passing frequency and the track resonant frequencies higher than the sleeper passing frequency. Consequently rail corrugation with several wavelengths initiates and develops. Also the results show that the contact vibrating frequencies of the four wheels of the same bogie and the rails are different. The rail corrugations caused by the four wheels of the same bogie present different wavelengths and wave depth.

1 Introduction According to site observations and studies in many published papers it was found that the discrete track support is a key factor causing rail corrugation. In the 80s of last century, Clark found that the short wavelength corrugation occurred only in the second half of each sleeper bay, on high speed sections on British Railway [1]. Through theoretical analysis he also found that the normal force of wheel-rail in rolling contact oscillates at corrugation passing frequency around the wheel-rail static load, and the oscillation amplitude is greatest near sleeper positions [2]. Ref. [3] analysed the conditions necessary for the existence in the British Rail system of corrugations with wavelengths in the range 40-80 mm with two oscillators in reference to track discrete support structure, friction and locomotive traction characteristics. Such a type of rail corrugation was referred to as roaring rail by Grassie and Kalousek [4], and attributed to a wear mechanism and the concerned factors including vertical dynamic loads, creepages between wheel and rail. Knothe and his group made a detailed investigation into the material wear situation at the different positions in a sleeper bay with the concepts of receptance and the pinned-pinned mode [5, 6], and found that a high corrugation growth rate occurs at the position of the rail over a sleeper. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 271–277, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

272

X. Jin et al.

Vadillo et al. found that the corrugation with a wavelength of about 60 mm appeared at the mid-span of each sleeper bay in the Bilbao area, Spain where the sleeper spacing was 1000 mm. By inserting an intermediate sleeper between each pair of sleepers they found the corrugation could be eliminated [7]. To explain the unusual behaviour they observed further, Gomez and Vadillo utilized the linear model for corrugation developed by Frederick to analyse it, and demonstrated that a corrugation initially occurred at the mid-span and almost disappeared above the sleeper [8]. Jin et al investigated the effect of periodic variation of sleeper supports on rail corrugation on the curved track by inputting a vertical rail irregularity with a period of sleeper pitch [9]. Wu and Thompson found that the wheel/rail force spectrum had several peaks at different frequencies due to the wave reflection between the wheels on a rail [10]. They investigated the normal force behaviour, much concerned with rail corrugation wavelengths, under multiple wheel/rail interaction [11]. They also analysed the formation of short pitch corrugation on the railhead using an approach combining wheel/track dynamics, contact mechanics and wear, in which multiple wheel/rail interactions were taken into account in the wheel/track dynamics [12]. According to the opinion of the present authors, the passing frequency of track sleeper is one of key factors influencing rail corrugation. The discrete support of track by sleepers can excite the higher resonant frequencies of the track, and therefore leads to rail corrugation formation. So, it is very important to develop theoretical models in time domain [13-16], to investigate the effect of sleeper passing frequency on rail corrugation in detail. The track model in Refs. [9, 13, 14] is assumed to be in a static state with respect to the vehicle. Ref. [17] developed the track model with the moving discrete supports to investigate the short pitch corrugation formation on the tangent track when the vehicle is hunting. In the present study, the corrugation model in Ref. [17] is used to investigate the effect of sleeper pitch on the corrugation formation on a curved track.

2 Calculation Model of Rail Corrugation The present model is roughly depicted in Fig. 1. In Fig. 1 Ywi and ψwi are, respectively, the lateral displacement and the yaw angle of wheelset i, subscript i denotes the number of the wheelsets considered in the calculation model, Yrk is the lateral displacement of the rail head under wheel k, and subscript k is the number of the wheels, δk and hk are, respectively, the contact angle and the normal distance between wheel k and the rail. ξjk and Pwrzk are, respectively, the creepages and the vertical loads between the wheels and the rails, subscripts j = 1, 2, 3 indicate the longitudinal, lateral and normal directions at the contact point of wheel and rail, subscripts w and r denote wheel and rail, z indicates the vertical direction of the track. fwk denotes the frictional work density in the contact area under wheel k. Fig. 1 indicates a feed-back process between the transient coupling dynamics of railway vehicle and track and long-term wear processes. In the calculation of rail corrugation at a curved track, a coupled dynamic model of a half vehicle and a curved track is used to analyse the dynamical behaviour of the vehicle passing over the curved track.

Preliminary Analysis on Effect of Sleeper Pitch on Rail Corrugation

273

Calculation of Curving Dynamics of Vehicle

Initial Rail Profiles

Ywi, ψwi, Yrk Worn Profiles of Rails (Known)

Rail Corrugation

(Known)

(Known)

(Known)

Calculation of Wheel/Rail Contact Geometry

Calculation of Profiles

fwk

ξjk, Pwrzk

Calculation of Wheel/Rail Rolling Contact

δk, hk (Known)

Fig. 1. A general description of curved rail wear model

Through the dynamic analysis, ξjk, Pwrzk, Ywi, ψwi and Yrk are obtained [13, 14]. Using them and the calculation method of wheel-rail contact geometry [18], hk and δk are calculated. ξjk, Pwrzk, hk and δk are used in calculating the rolling contact mechanics of the wheel and rail, in which Kalker’s model of three dimensional elastic bodies in rolling contact is introduced and modified to calculate fwk, contact stresses, stick/slip areas, etc. [18, 19]. The frictional work density on the contact area is calculated by using the tangent traction components and the total slip components between a pair of the contacting particles in the contact area. The undulatory wear depth on rail running surface is determined with the material wear model by Clayton et al. [20]. Using the known wear depth at the present step, the existing rail profile is updated for the next loop calculation. After the repeated loop calculations the accumulated material wear and its pattern on the rail head appear.

3 Numerical Results and Discussions In the present analysis a half passenger car passing through a curved track with radius of 300 m is considered. The transition curve is 90 m long, the circle curve is 100 m long, the track gauge is 1435+2 mm, the rail cant is 1/40, and the super elevation of the track is 180 mm. The speed of the vehicle is 80 km/h. The nominal radius of the wheelsets used in the calculation is 457.5 mm, and the wheelbase is 2.4 m. The values of the other parameters used in the analysis are found in Refs. [13, 14, 18]. In order to present the numerical results clearly and conveniently, the left wheel and the right one of the leading wheelset are denoted as wheel 1 and wheel 2 respectively, and the corresponding wheels of the trailing wheelset indicated by wheel 3 and wheel 4. The leading and trailing wheelsets are denoted as wheelset 1 and wheelset 2, respectively.

274

X. Jin et al.

3.1 Dynamic Behaviour of Vehicle Curving It is essential to acquire the normal load and the creepages between the wheel and rail in the calculation of rail wear with the rolling contact theory by Kalker. Fig. 2 shows the normal loads with the wheel travelling distance along the curved track. The normal loads oscillate at high frequencies when the vehicle passes through the transition curve. Fig. 3 illustrates the lateral creepages between the wheels and the rails. The creepages also oscillate fiercely at the transition curves. On the transition curves the longitudinal and spin creepages oscillate at high frequencies, and they are not shown in the present paper. The lateral creepage oscillating at high frequencies indicates that the contacting surface of the wheel moves back and forth with respect to the rail contacting surface in the lateral direction. In such a situation the contact point of the wheel tread moves back and forth in the lateral direction. The instant radius of the wheel rolling circle varies at the high frequencies and therefore the longitudinal creepage includes the same high frequencies. The frequencies of the creepage oscillating depend on the excited resonant frequencies of the flexible wheelset and the elastic track. In order to clarify the frequencies of the oscillating normal loads and creepages, the curves of the normal loads and the creepages in the range of 90 m to 93 m are analysed in the frequency domain. It is found that the normal load of wheel 1 oscillates at different frequencies, the normal load of wheel 3 has the same frequency as that of the sleeper passing, and those of wheels 2 and 4 have higher frequencies. A similar situation occurs in the creepages. The oscillating amplitude of the lateral creepage of wheel 2 is very large. At these oscillating frequencies, deeper wear forms. The lateral creepage of wheel 3 has peaks over the sleepers, namely, reaches the maximum, therefore, serious wear is caused by wheel 3 over the sleepers. According to the normal loads and the creepages, it is judged that wheels 2 and 4 can cause uneven wear with wavelengths less than a sleeper pitch.

Fig. 2. Normal loads between rails and wheels

Preliminary Analysis on Effect of Sleeper Pitch on Rail Corrugation

275

Fig. 3. Lateral creepages between rails and wheels

3.2 Rail Corrugation Fig. 4 illustrates that the maximum depth of the rail wear caused by each wheel, after different numbers of passages. Curves 1, 2, 3, 4, and 5 indicate the wear depths after 1, 5, 10, 20, and 30 passages, respectively. Except for wheel 3, the other 3 wheels cause the uneven wears with the similar distribution of the troughs. The corrugation caused by wheel 1 has wavelengths from 180 mm to 300 mm. Compared with the other three wheels the average wear depth caused by wheel 1 is the largest. The corrugations caused by wheels 2 and 4 present the troughs with about 100 mm width and large depth, which leads to strong impacts between the wheels and rails. The

Maximum Depth(mm)

0.08

Wheel 1 5

0.15 0.12

4

0.09 3

0.06

2

0.03 0.00

1

Maximum Depth(mm)

Sleeper Positions

0.18

1800

0.05 0.04

5

0.03

4

0.02

3 2 1

0.01

2400

1200

x  mm (a)

5

0.003 0.002

4

0.001

3 2 1

1200

1800

x  mm (c)

2400

0.006

Maximum Depth(mm)

Maximum Depth(mm)

Wheel 3

0.004

0.000

1800

x  mm

2400

(b)

Sleeper Positions

0.005

Wheel 2

0.06

0.00 1200

Sleeper Positions

0.07

Sleeper Positions

Wheel 4

0.004 5

0.002

0.000

4 3 2 1

1200

1800

2400

x  mm

(d)

Fig. 4. Maximum depths of uneven wear caused by the four wheels after different passages

276

X. Jin et al.

corrugations caused by wheels 2 and 4 possess not only about 180 mm and 300 mm wavelengths, but also wavelengths from 20 mm to 100 mm. The corrugation with about 180 mm and 300 mm wavelengths is concerned with the resonant vibration of the rails and the sleepers on the ballast bed. The corrugation with wavelength from 20 mm to 100 mm is related to the excited track resonance frequencies of 447 Hz, 800 Hz, and 1220 Hz. The wavelength of the corrugation caused by wheel 3 equals a sleeper pitch. The largest wear depth occurs over the sleepers. Fig. 4 shows that corrugation occurred on the low curved rail is much more serious than that on the high curved rail at the same curved track, just as the observations at railway sites.

4 Conclusions (1) When a railway vehicle passes through a curved track the discrete rail support by sleepers causes the normal loads and the creepages to oscillate at different frequencies. These frequencies include the passing frequencies of the sleepers and the high excited track resonance frequencies. Therefore, rail corrugations initiate and develop with several wavelengths. (2) The patterns of rail corrugations caused by the four wheels of a bogie are different when it passes through a same curved track. The average wear depth on the high curved rail caused by the leading wheel is the largest, compared with the other three wheels. The corrugation occurring on the low curved rail is more serious than that on the high rail for the present cases. The corrugation occurring on the high curved rail caused by the trailing wheel forms with the passing frequency of the sleepers, and its maximum wear depth occurs at over the sleepers.

Acknowledgements This project is supported by NSFC (50521503), FANEDD (No 2002048), and National Basic Research Program of China (2007CB714702). The authors are grateful to thank Professor David Thompson (ISVR) and Professor Tianxing Wu (Shanghai Jiaotong University) for their valuable technical advice.

References [1] Clark, R.A., Dean, P.A., Elkins, J.A., Newton, S.G.: An investigation into the dynamic effects of railway vehicles running on corrugated rails. J. Mech. Eng. Sci. 24, 65–76 (1982) [2] Clark, R.A.: Slip-stick vibrations may hold the key to corrugation puzzle. Railway Gazette International 17, 531–533 (1984) [3] Brockly, C.R.: The influence of track support structure and locomotive traction characteristics on short wavelength corrugations. Wear 153, 315–322 (1992) [4] Grassie, S.L., Kalousek, J.: Rail corrugation: Characteristics, cause and treatments. Proceedings of the Institution of Mechanical Engineers 207(Part F), 57–68 (1993) [5] Knothe, K., Ripke, B.: The effects of parameters of wheelset, track and running conditions on the growth rate of rail corrugation. Vehicle System Dynamics (Supplement) 18, 345–356 (1989)

Preliminary Analysis on Effect of Sleeper Pitch on Rail Corrugation

277

[6] Hempelmann, K., Knothe, K.: An extended linear model for the prediction of short pitch corrugation. Wear 191, 161–169 (1996) [7] Vadillo, E.G., Tarrago, J.A., Zubiaurre, G.G., Duque, C.A.: Effect of sleeper distance on rail corrugation. Wear 217, 140–146 (1998) [8] Gomez, I., Vadillo, E.G.: An analytical approach to study a special case of booted sleeper track rail corrugation. Wear 251, 916–924 (2001) [9] Jin, X.S., Wen, Z.F., Wang, K.Y.: Effect of track irregularities on initiation and evolution of rail corrugation. Journal of Sound and Vibration 285(1–2), 121–148 (2005) [10] Wu, T.X., Thompson, D.J.: Vibration analysis of railway track with multiple wheels on the rail. Journal of Sound and Vibration 239, 69–97 (2001) [11] Wu, T.X., Thompson, D.J.: Behaviour of the normal contact force under multiple wheel/rail interaction. Vehicle System Dynamics 37, 157–174 (2002) [12] Wu, T.X., Thompson, D.J.: An investigation into rail corrugation due to micro-slip under multiple wheel/rail interactions. Wear 258, 1115–1125 (2005) [13] Jin, X.S., Wen, Z.F., Wang, K.Y., Zhou, Z.R., Liu, Q.Y., Li, C.H.: Three-dimensional train-track model for study of rail corrugation. Journal of Sound and Vibration 293(3–5), 830–855 (2006) [14] Jin, X.S., Wen, Z.F., Wang, K.Y., Biao, X.X.: Effect of passenger car curving on rail corrugation at a curved track. Wear 260, 619–633 (2006) [15] Igeland, A., Ilias, H.: Rail head corrugation growth predictions based on non-linear high frequency vehicle/track interaction. Wear 213, 90–97 (1997) [16] Anderson, C., Johansson, A.: Prediction of rail corrugation generated by threedimensional wheel-rail interaction. Wear 257, 423–434 (2007) [17] Jin, X.S., Biao, X.X., Wen, Z.F., Wang, K.Y.: Effect of sleeper pitch on rail corrugation at a tangent track in vehicle hunting. In: Proc., 7th International Conference Contact mechanics and Wear of Rail/Wheel Systems, Brisbane Australia, September 24–26, pp. 179–188 (2006) [18] Jin, X.S., Wen, Z.F., Zhang, W.H., Shen, Z.Y.: Numerical simulation of rail corrugation on curved track. Computers & Structures 83, 2052–2065 (2005) [19] Kalker, J.J.: Three-Dimensional Elastic Bodies in Rolling Contact. Kluwer Academic Publishers, The Netherlands (1990) [20] Bolton, P.J., Clayton, P., McEwan, I.J.: Rolling-sliding wear damage in rail and tyre steels. Wear 120, 145–165 (1987)

A Hybrid Model for Noise Generation from a Railway Wheel Due to Wheel/Rail Impact Xinbiao Xiao, Xuesong Jin*, and Xiaozhen Sheng State Key Laboratory of Traction Power, Southwest Jiaotong University, Chengdu 610031 China Tel.: +86-28-87634355; Fax: +86-28-87600868 [email protected]

Summary A hybrid model is developed for noise generation from a railway wheel due to wheel/rail impact at a rail joint. It consists of a coupled vehicle/track dynamic interaction model working in the time domain and a FE-BE vibro-acoustic model for the wheel working in the frequency domain. In the coupled vehicle/track interaction model, the vehicle is described as a multi-body system, the rail is idealised as a Timoshenko beam resting on discrete sleepers, and the sleepers are treated as Euler beams. The lateral, vertical, and torsional vibrations of the rail are all taken into account. Sleepers are assumed to move backward at the train speed to simulate the travelling of the vehicle along the track. Wheel/rail normal forces are calculated using the Hertzian contact theory and creep forces are determined using Shen’s nonlinear creep theory. The differential equations of motion of the vehicle/track system are solved by means of an explicit integration method, giving wheel/rail force time-histories. The wheel/rail force between a wheel and the rail is then transformed into the frequency domain and input to a FE model of the wheel to calculate its dynamic response. Sound radiated from the wheel is then calculated from the surface response of the wheel using the acoustic boundary element method. Results produced from this hybrid model demonstrate its suitability for predicting noise radiation from a railway wheel due to wheel/rail impact.

1 Introduction Wheel/rail impact noise is generated at locations with roughness of high amplitudes and short wavelengths, of rail joints, rail defects, and other discontinuities on the rail running surface. Wheel/rail impact noise is also generated when there are flats on the wheel rolling surface. Noise levels, both in passenger compartments and along the wayside, may be dominated by the impact. However, although wheel/rail impact dynamics has been studied over many years, there have been to date few published papers dedicated to the modelling of wheel/rail impact noise generation. Early work on wheel/rail interactions in response to wheel flats was carried out by Newton and Clark [1] at British Railways, including aspects of both prediction and measurement. Then, Refs.[2-4] developed models to investigate the characteristics of *

Corresponding author.

B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 278–284, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

A Hybrid Model for Noise Generation from a Railway Wheel

279

impact loads due to wheel/rail tread defects. Recently, Wu and Thompson [5] studied the effects of track non-linearity on wheel/rail impact. All the aforementioned studies were not directly driven by the issue of impact noise; nevertheless, they provide a basis for the work of modelling wheel/rail impact noise generation. Ver et al [6] carried out a detailed study on impact noise due to wheel and rail irregularities. They first established the concept of the critical speed above which wheel/rail separation occurs. They also developed analytical models to identify the major variables controlling the generation of impact noise for five different wheel and rail irregularities: smooth irregularity, wheel flat, level rail joint, step-up rail joint and step-down rail joint. Remington [7] extended this work and introduced an equivalent roughness spectrum for wheel flats and rail joints. Wu and Thompson developed a hybrid model to predict the noise generation from railway wheel flats [8] and rail joints [9]. Firstly, a time-domain model, which was a simplified dynamic wheel/track model but able to account for the non-linearity of the contact spring and the possibility of loss of wheel/rail contact, was used to calculate the wheel/rail impact force. Secondly, the wheel/rail interaction force was transformed into the frequency domain and then converted to an equivalent roughness spectrum. Finally, the radiated impact noise was predicted by inputting the equivalent roughness into the TWINS (TrackWheel Interaction Noise Software) model [10]. It is a fact that in previous work on wheel/rail impact noise, the coupling of the wheels of the vehicle and two rails of the track is always ignored. Numerical calculation of high frequency rail corrugation shows that the coupling influences between the two rails through the wheelset cross-admittances, and the leading and trailing wheelsets through the rail cross-admittances are great, and cannot be ignored in the corrugation analysis [11]. It is also true the sound radiation model is over simplified. This paper offers an effort to overcome those shortcomings. In the present paper, a coupled vehicle/track dynamic model is developed in the time domain to predict impact forces due to wheel/rail discontinuities. In the model a passenger car is considered and modelled as a multi-body system. The track is treated as a pair of Timoshenko beams which are discretely supported by sleepers. The sleepers are described as Euler beams, rather than a rigid mass as in many previous studies. The supports, including the rail pads and sleepers, are assumed to move backward to simulate the travelling of the vehicle along the track at the train speed. Wheel/rail impact forces are then transformed into the frequency domain to calculate dynamic responses of the wheels using the finite element method. Sound radiation from a vibrating wheel is calculated using the conventional (direct or indirect) acoustic boundary element method. Results, including the wheel/rail impact force and noise radiation from the wheel, are produced and analysed for a step-up rail joint. These results demonstrate the suitability of the hybrid model for predicting railway noise generated from wheel/rail impact.

2 The Vehicle/Track Model 2.1

The Vehicle Model

The vehicle considered in this paper is equipped with a pair of two-axle bogies with double suspension systems. The wheelset and the bogie are connected by the primary

280

X. Xiao, X. Jin, and X. Sheng

suspension, while the car body is supported on the bogie through the secondary suspension, see Fig. 1. Since the dynamic stiffness of the wheel is much higher (apart from at the natural frequencies of the wheel) than that of the rail, the wheelset is modelled as a rigid body when calculating wheel/rail forces. 2.2 The Track Model The track model is also shown in Fig. 1. This model was firstly used by Sato et al [12] and then by Zhai et al [13]. It is further improved here to account for high frequency vibration. Each rail is modelled as a Timoshenko beam hinged at the two ends. The vertical, lateral and torsional vibra-tions of the rails are all taken into account. For vertical vibration, each sleeper is modelled as a uni-form Euler beam with a free-free end condition while for lateral vibration, it is simplified as a lumped mass. 2.3 Coupling of the Vehicle and Track Models and Wheel/Rail Impact The vehicle and the track are coupled through the wheel/rail contact geometry and forces. The numerical method used here for the wheel/rail contact geometry calculation is the one detailed in Ref. [14]. Wheel/rail contact forces include a normal force and three tangential (creep) forces (or moments). The normal wheel/rail contact force is calculated using the Hertzian nonlinear contact theory with a unilateral restraint. The creep forces are determined by applying Shen’s formulae [15]. Different approaches have been used to describe vehicle/track interactions, as reviewed by Knothe [16]. The moving irregularity model has been widely used to deal with problems of vehicle/track dynamics [13] and wheel/rail noise [8 - 11]. However, this model is unable to simulate the effect on wheel/rail interactions of periodic variations in the track dynamic stiffness [17]. Having carefully examined the advantages and disadvantages of various vehicle/track interaction models, the approach proposed by Young et al [18] and then further refined by the current authors [19] is chosen to simulate the effect of the discrete sleeper supports on the vehicle and track dynamics as the vehicle runs at a constant speed v.

(a) Elevation

(b) Side elevation

Fig. 1. Coupled vehicle/track model

A Hybrid Model for Noise Generation from a Railway Wheel

281

In this paper, the rail is assumed to be perfect except for a rail joint. As a wheel passes over the joint, impact occurs and impact noise is generated. Due to the huge impact force between a wheel and the rail, strong non-linearity exists in the wheel/rail contact process. This will result in a noise spectrum which may be significantly different from what would be predicted using a completely linear model.

3 The Noise Radiation Model for a Railway Wheel 3.1 Vibration Modes of the Wheel Thompson [20] analysed the vibration of a railway wheel using the finite element method. In the present paper, a standard China passenger wheel with diameter of 0.915m is considered. Due to the model used for the rail, the coupled vehicle/track model is not appropriate to investigate vibrations at frequencies higher than 5000 Hz [16]. Therefore, natural frequencies for the wheel are predicted up to about 5000 Hz covering 57 modes, using the commercial software ANSYS with the block Lanczos method. A mode can be labelled by two numbers, n and m, where n (n = 0, 1, 2, 3, …) indicates the number of nodal diameters and m (m = 0, 1, 2, 3, …) the nodal circles [20]. 3.2 Forced Vibration of the Wheel The forced response is calculated using the mode superposition method. To do so, natural frequencies, mode shapes and modal damping ratios are required. The natural frequencies and mode shapes are produced from the wheel FE model, as detailed above. The modal damping ratios cannot be predicted from a FE analysis and they must be estimated from measurement such as the modal testing. Ref [20] shows the modal damping ratios range from 0.1‰ to 0.3‰. In this paper, a value of 0.2‰ has been used for all the modal damping ratios. 3.3 Noise Radiation from the Wheel The sound field generated from the surface vibration of the wheel is calculated using the direct boundary element method implemented in the software SYSNOISE. Having considered the frequency limitation of the coupled vehicle/track model and the computation time of the BE model, the maximum frequency for the BE model is set to be 3000 Hz.

4 Results and Discussion 4.1 Wheel/Rail Impact Force Due to a Step-Up Rail Joint Simulation of wheel/rail impact is carried out for a step-up rail joint with a height difference of 1 mm. In the simulation, the step-up rail joint is equivalently represented by a vertical wheel velocity impact [6]. The axle load of the passenger car is 150 kN corresponding to a static wheel load of 75 kN. Fig. 2 shows the normal wheel/rail impact force of the left wheel/rail of the leading wheelset when the wheel passes over the stepup joint at 80 km/h. During the impact, the wheel/rail contact force rises dramatically

282

X. Xiao, X. Jin, and X. Sheng

and, after about 0.5 ms, reaches the first peak P1, which is 394 kN and around 5 times of the static load. Loss of wheel/rail contact is seen for about 3 ms during which the wheel vibrates freely and radiates noise without energy flowing into the track.

Fig. 2. Wheel/rail impact force

Fig. 3. Spectrum of the impact force

Fig. 3 shows the frequency spectrum of the impact force. The time signal length is 125 ms, from which the spectrum is derived. Two peaks are present. The first peak at about 68 Hz corresponds to the unsprung mass-on-track resonance frequency. The second peak at about 848 Hz corresponds to the first pinned-pinned frequency of the track [16]. It can be seen that components at frequencies higher than 2 kHz are more than 20 dB lower than that at the pinned-pinned frequency. The frequency range of the impact force is well within the capacity of the track model. 4.2 Wheel Response to the Impact When the wheel passes through the rail joint, the wheel/rail contact point is within a small vicinity of the nominal radius of the wheel. To calculate the forced vibration of the wheel, the wheel/rail impact force may be approximated to be applied at a fixed point on the wheel, that is node A shown in Fig. 3. Since the vertical component of the impact force is dominant over other components, only the vertical component is taken into account. Forced responses of three nodes are shown in Fig. 3. These three nodes are located on the wheel tread, the tyre, and the web. As sound radiation is only related to velocity component normal to the wheel surface, only the radial displacement of node A and the axial displacements of nodes B and C are presented. The numbers of nodal diameters and circles of the modes at f1 (185 Hz), f2 (380 Hz) and f3 (1070 Hz) are (0,1), (0,2) and (radial,1). As can be seen in Fig. 4, due to different modes being excited, the location having the maximum normal velocity varies with frequency. It is particularly noticed that, for frequencies higher than 1500 Hz, the out-of-plane vibration of the web is prominent. 4.3 Noise Radiation from the Wheel From Fig.5 it can be seen that, the noisiest frequencies are fa (380 Hz, f2 in Fig. 4) and fb (1070 Hz, f3 in Fig. 4), although at these two frequencies the forced responses (described by the spatially averaged mean-square normal velocity of the wheel surface)

A Hybrid Model for Noise Generation from a Railway Wheel

283

may not be the strongest. This is because sound power also depends on the radiation efficiency.

Fig. 4. Forced responses of the wheel

Fig. 5. Sound power level of railway wheel

Lwc indicates the highest peak for frequencies above 1500 Hz, which occurs at the mode of 1950 Hz. The mode shape of this frequency is also shown in Fig. 5. The mode shape looks like a speaker cone of high radiation efficiency. It can be seen from Fig. 4 and 5 that there are many distinct very sharp, lightly damped peaks which correspond to the resonance frequencies of the wheel. This is because the impact force spectrum is directly used in a mode model of the wheel. In practice, these peaks will experience much more damping when the wheel is coupled to the rail, and not so sharp, as shown in Figs. 4 and 5. It is necessary to develop more reasonable procedures to treat the damping of wheel coupled with rail, for example, Wu and Thompson [8, 9] used an equivalent roughness as an input to wheel/rail system.

5 Conclusion A hybrid model is developed for noise generation from a railway wheel due to wheel/rail impact at a rail joint. Results are produced from this hybrid model for a vehicle running over a step-up rail joint of 1 mm height. The impact force fluctuates about a mean at the pinned-pinned frequency of the track. The mean is around 3 times the static load while the maximum impact force is as high as 5 times. Wheel/rail separation occurs for about 3 ms. Only sound power from a single wheel is presented, showing the sound power level can be as high as 110 dB. This, however, does not necessarily imply the total impact noise is mainly from the wheel. Contributions from the rail and sleepers will be investigated in future work.

Acknowledgements This project is supported by FANEDD (No. 2002048). The authors are grateful to thank Professor David Thompson (ISVR) and Professor Tianxing Wu (Shanghai Jiaotong University) for their valuable technical advice.

284

X. Xiao, X. Jin, and X. Sheng

References [1] Newton, S.G., Clark, R.A.: An investigation into the dynamic effects on the track of wheelflats on railway vehicles. Journal of Mechanical Engineering Science 21, 287–297 (1979) [2] Cai, Z.Q.: Modelling of rail track dynamics and wheel/rail interaction, PhD Thesis, Queen’s University (1992) [3] Nielsen, J.C.O., Igeland, A.: Vertical dynamic interaction between train and track – influence of wheel and track imperfections. Journal of Sound and Vibration 187, 825–839 (1995) [4] Dong, R.G.: Vertical dynamics of railway vehicle-track system, PhD Thesis, Concordia University (1994) [5] Wu, T.X., Thompson, D.J.: The effects of track non-linearity on wheel/rail impact. Proceedings of the Institution of Mechanical Engineers Part F: Journal of rail and rapid transit 218, 1–15 (2004) [6] Ver, I.L., Ventres, C.S., Myles, M.M.: Wheel/rail noise – part III: Impact noise generation by wheel and rail discontinuities. Journal of Sound and Vibration 46, 395–417 (1976) [7] Remington, P.J.: Wheel/rail squeal and impact noise: what do we know? What don’t we know? Where do we go from here? Journal of Sound and Vibration 116, 339–353 (1985) [8] Wu, T.X., Thompson, D.J.: A hybrid model for the noise generation due to railway wheel flats. Journal of Sound and Vibration 251, 115–139 (2002) [9] Wu, T.X., Thompson, D.J.: On the impact noise generation due to a wheel passing over rail joints. Journal of Sound and Vibration 267, 485–496 (2003) [10] Thompson, D.J., Hemsworth, B., Vincent, N.: Experimental validation of the TWINS prediction program for rolling noise. Part 1: Description of the model and method. Journal of Sound and Vibration 193, 123–135 (1996) [11] Jin, X., Wen, Z., Wang, K., Xiao, X.: Effect of passenger car curving on rail corrugation at a curved track. Wear 260, 619–633 (2006) [12] Sato, Y., Odaka, T., Takai, H.: Theoretical analysis on vibration of ballasted track (in Japanese). Railway Technical Research Report, 13–17 (1987) [13] Zhai, W.M.: The vertical model of vehicle-track system and its coupling dynamics (in Chinese). Journal of The China Railway Society 14, 21–29 (1992) [14] Jin, X.S., Wen, Z.F., Zhang, W.H., Shen, Z.Y.: Numerical simulation of rail corrugation on a curved track. Computers and Structures 83, 2052–2065 (2005) [15] Shen, Z.Y., Hedrick, J.K., Elkins, J.A.: A comparison of alternative creep-force models for rail vehicle dynamic analysis. In: Proceedings of the Eighth IAVSD Symposium, Cambridge, MA, pp. 591–605 (1984) [16] Knothe, K.L., Grassie, S.L.: Modeling of railway track and vehicle/track interaction at high frequencies. Vehicle System Dynamics 22, 209–262 (1993) [17] Sheng, X., Li, M., Jones, C.J.C., Thompson, D.J.: Using the Fourier series approach to study interactions between moving wheels and a periodically supported rail. Journal of Sound and Vibration 303, 873–984 (2007) [18] Young, T.H., Li, C.Y.: Vertical vibration analysis of vehicle/imperfect track systems. Vehicle System Dynamics 40, 329–349 (2003) [19] Xiao, X.B., Jin, X.S., Wen, Z.F.: Effect of disabled fastening systems and ballast on vehicle derailment. ASME Journal of Vibration and Acoustics 129, 217–229 (2007) [20] Thompson, D.J.: Wheel-rail noise generation, part II: Wheel vibration. Journal of Sound and Vibration 161, 401–419 (1993)

A Time Domain Model for Wheel/Rail Interaction Aiming to Include Non-linear Contact Stiffness and Tangential Friction A. Pieringer1, W. Kropp1, and J.C.O. Nielsen2 1

Division of Applied Acoustics / CHARMEC, Chalmers University of Technology, Sven Hultins gata 8a, SE-41296 Göteborg, Sweden Tel.: +46 31 772 2209; Fax: +46 31 772 2212 [email protected] 2 Department of Applied Mechanics / CHARMEC, Chalmers University of Technology, Sven Hultins gata 8a, SE-41296 Göteborg, Sweden

Summary A time domain model is presented for the dynamic interaction between a railway wheel and rail, which takes into account the non-linear processes in the contact zone and aims at predicting both normal and tangential contact forces. The model follows an approach that has been used successfully, for instance for the modelling of the interaction between road and tyre. Track and wheel are described as linear systems by the means of impulse response functions. The contact zone is modelled by non-linear contact springs with stiffnesses depending on the roughness of rail and wheel. Here, the method of the area of real contact is applied in order to obtain the required spring characteristics. For the tangential contact, a characteristic function for the friction coefficient is applied. In a first stage, the approach is demonstrated for the calculation of normal contact forces. For validation, the results from the model are compared with an existing time domain model that itself has been validated by field testing. Very good agreement is found for different types of roughness excitation.

1 Introduction The interaction between wheel and rail is considered as the dominating source for noise emission from railway operations over a wide speed range. On one hand, this interaction concerns contact forces in the direction normal to the surfaces, caused by the roughness on rail and wheel, giving rise to rolling noise. On the other hand, it considers tangential forces caused by the friction between rail and wheel seen as responsible for squealing noise. During the past decades, a considerable number of frequency domain (e.g. [1,2]) and time domain (e.g. [2-4]) models for the normal wheel/rail interaction have been published. In contrast to frequency domain models, time domain models can include non-linear contact models. Non-linearities cannot be neglected in cases where loss of contact is likely to occur [4,5]. Only a few models predicting the coupled normal and tangential interaction are available, see e.g. [6]. In general, time domain models require much more computational effort than frequency domain models. In this paper, a computationally efficient time domain model B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 285–291, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

286

A. Pieringer, W. Kropp, and J.C.O. Nielsen

that aims to include the coupled normal and longitudinal/lateral interaction is presented. Wheel and rail are represented by precalculated impulse response functions. The non-linear contact model allows for loss of contact. In order to obtain the required contact spring characteristics, the area of real contact of the rough surfaces is evaluated. The contact filtering effect [7] is hereby integrated into the model in a natural manner. The generality of the approach separating wheel, rail and contact in different modules, allows the inclusion of any wheel or rail model that can be represented by Green's functions. The concept described here has already been successfully applied in the area of tyre/road noise, see e.g. [8]. In the area of wheel/rail contact the utilisation of Green's functions goes back to Heckl's proposal for a railway simulation program [9].

2 Wheel/Rail Interaction Model 2.1 Track and Wheel Model The track model is a linear finite element model accounting for discrete supports (see [3]). Fig. 1(a) shows the magnitude of the track point receptance for two different excitation positions. For inclusion into the wheel/rail interaction model, the track has x to be represented by so-called moving Green's functions g% R0,v (t ) (Fig. 1(b)). The x function g% R0,v (t ) describes, for excitation of the rail (index R) at the position x0 at

time t0 = 0 , the displacement response of the rail at a point moving with train speed v away from the excitation, thus at the contact point between wheel and rail [2]. The wheel is modelled by a mass and a primary suspension and is represented by the wheel's Green's function g% W ( t ) . The vehicle components above the primary suspension are simplified to a static preload P .

(a)

(b)

Fig. 1. Response of the track. ⎯ excitation at midspan between two sleeper positions, − ⋅ − excitation above a sleeper position: (a) Magnitude of the point receptance; (b) Moving Green’s function for v = 100 km/h .

A Time Domain Model for Wheel/Rail Interaction

287

2.2 Normal Contact Model

In the interaction model, the wheel moves over the rail with constant speed v (Fig. 2(a)). The normal contact force, denoted Fn , excites the wheel upwards and the rail downwards in the contact point. The static preload P is applied by pressing the wheel onto the rail until the static contact force equals P . The suspension is then fixed in the corresponding position ξS = ξS ( P ) . The normal displacement of the wheel ξ W ( t ) is obtained by convoluting Fn ( t ) with g% W ( t )

ξ W ( t ) = -∫ Fn (τ ) g% W ( t - τ ) dτ + ξS . t

0

(1)

In a similar manner, the normal displacement of the rail ξ R ( t ) is calculated as

ξ R ( t ) = ∫ Fn (τ ) g% R,vτv ( t - τ ) dτ . t

(2)

0

(a)

(b)

Fig. 2. Dynamic wheel/track interaction model: (a) Interaction scheme; (b) Bedding model for the wheel/rail contact

For the calculation of Fn , a Winkler bedding consisting of independent springs is introduced between wheel and rail (see Fig. 2(b)). The combined roughness is contained in the variable r ( x ) (positive for an asperity on the rail). For the wheel positioned at x , the deflection Δζ ( x, x′ ) of all contact springs involved depends on the

wheel deflection ξ W ( x ) , the rail deflection ξ R ( x ) , the combined roughness

r ( x + x′ ) and the wheel profile yW ( x′ )

Δζ ( x, x′ ) = ξ W ( x ) − ξ R ( x) + r ( x + x′ ) − yW ( x′ ) .

(3)

The force Fn is obtained by integrating over the bedding that has a stiffness per unit length k ( x, x′ )

288

A. Pieringer, W. Kropp, and J.C.O. Nielsen

⎧ a′ ⎪ Fn ( x ) = ∫ k ( x, x′ ) Δζ ( x, x′) dx′ with k ( x, x′ ) = ⎨ − a′ ⎪⎩

1 E 2 1−ν 2

0

for Δζ ( x, x′ ) ≥ 0 for Δζ ( x, x′ ) < 0

, (4)

where E is the Young's modulus and ν the Poisson's ratio of rail and wheel. The integration domain has to be chosen long enough to include all potential points of contact. Assumptions about the wheel and rail radii of curvature are the same as in reference [7] and make it possible that the bedding correctly models Hertz contact for smooth surfaces. Considering the relation x = vt , the non-linear system of equations (1) to (4) is a complete description of the contact problem. 2.3 Modelling of the Non-linear Contact Stiffness Including Roughness

The previously described approach for the modelling of the contact stiffness using a bedding model already has a non-linear character, since the spring stiffness switches to zero for loss of contact. An alternative is to formulate the total contact stiffness k tot by one single spring with a non-linear characteristic that depends on the roughness of rail and wheel and the moduli of elasticity of both. Mathematically, this reads as

Fn ( x, Δζ 0 ) = ∫

Δζ 0 0

(

)

k tot x, Δζ 0 d Δζ 0 ,

(5)

where Δζ 0 is the compression of the spring. This can be visualised in terms of the process of pressing a wheel on the rail, which itself is assumed at rest. In the very first moments of contact, only a few asperities of the rough wheel and rail surface will touch each other and form junctions. Forcing the wheel closer to the rail, more and more junctions form and the stiffness increases due to an increasing area of real contact. There are different ways of obtaining the stiffness k tot from this process. One procedure could be based on the bedding model presented in section 2.2. The only difference to the previously presented procedure is that the resulting stiffness function for each wheel position x is calculated in advance. Numerically, it is of advantage to work with a resulting contact force as a function of the compression of the spring avoiding the integration in equation (5). This force can be calculated by evaluating the integral over the compressed bedding area a′

Fn ( x, Δζ 0 ) = ∫ k ( x, x′ ) Δζ ( x, x′ ) dx′ with Δζ ( x, x′ ) = Δζ 0 + r ( x + x′ ) − yW ( x′ ) , (6) − a′

where Δζ 0 is given by the position of the wheel. After calculating these func-

tions Fn ( x, Δζ 0 ) for each centre point of contact, they can be used in the contact algorithm to determine the force as a function of wheel and rail motion. The variable Δζ 0 is then the difference between both movements. Using a bedding model for calculating the resulting force is just one possibility. Alternatives can be found in the literature. Andersson [10] uses e.g. a third body in the form of an elastic layer between tyre and road and calculates the resulting non-linear stiffness as a function of the area of real contact. To follow this approach, however, would demand a detailed description of rail and wheel roughness with sufficient resolution.

A Time Domain Model for Wheel/Rail Interaction

289

2.4 Inclusion of Tangential Friction

The main motivation for the approach discussed in this paper is its potential to include longitudinal and lateral frictional forces in a straightforward manner. Although the implementation is beyond the scope of this paper, in the following it is briefly shown how to proceed. For simplicity, only one tangential direction t is considered. To include friction, the set of Green's functions has to be extended. The moving Green's x function g% R0,v (t ) describing the response ξ R,n of the rail in the normal direction due x to a unit force Fn in the normal direction is renamed to g% R0,nn (t ) . Furthermore, a

moving Green's function g% R0,tt (t ) representing the response ξ R,t in the tangential dix

rection due to a tangential force Ft is introduced. Normal and tangential forces will also lead to tangential and normal displacements, respectively, which are expressed x x by the Green's functions g% R0,nt (t ) and g% R0,tn (t ) . Using a similar set of functions for the wheel, one can establish a set of equations as vτ vτ ξ R,n ( t ) = ∫ Fn (τ ) g% R,nn ( t - τ ) dτ + ∫ Ft (τ ) g% R,tn ( t - τ ) dτ t

t

0 t

0

vτ vτ ξ&R,t ( t ) = ∫ Fn (τ ) g%& R,nt ( t - τ ) dτ + ∫ Ft (τ ) g%& R,tt ( t - τ ) dτ t

0

0

ξ W,n ( t ) = − ∫ Fn (τ ) g% W,nn ( t - τ ) dτ − ∫ Ft (τ ) g% W,tn ( t - τ ) dτ + ξS,n ξ&

t

t

0 t

0

(7)

t &% &% W,t ( t ) = − ∫ Fn (τ ) g W,nt ( t - τ ) dτ − ∫ Ft (τ ) g W,tt ( t - τ ) dτ , 0

0

where the dot denotes a time derivative. Furthermore, a relation between normal displacements and the normal force as well as a relation between the relative tangential velocity between wheel and rail and the tangential force is needed. The first relation can e.g. be taken from section 2.3 while for the latter the friction characteristic as a

(

) (

)

function of relative velocity Ft ( t ) = Fn x ( t ) , Δζ 0 μ Δu ( t ) can be applied. The

relative velocity is here defined as Δu ( t ) = ξ&W,t ( t ) − ξ&R ,t ( t ) . As a result, a nonlinear equation system is obtained which apart from its higher complexity is similar to that for normal contact forces only. Although the model is expected to work well and be numerically efficient, the critical point certainly is to obtain adequate input data,

(

)

such as the friction characteristic μ Δu ( t ) .

3 Validation of the Model against an Existing Time Domain Model For validation, simulation results of the normal interaction model (section 2.2) are compared with results from the train/track interaction model DIFF [3] that uses an extended state-space vector approach in conjunction with a complex modal superposition for the track. DIFF itself has been validated by field testing [11]. To facilitate the comparison, the wheel model is replaced by an unsprung mass without suspension,

290

A. Pieringer, W. Kropp, and J.C.O. Nielsen

the Winkler bedding (Fig. 2(b)) is substituted by one single contact spring modelling Hertz contact and allowing for loss of contact, and the contact filter effect is included in a preprocessing step by averaging the roughness over the contact patch length. The parameters applied are a Young's modulus of wheel and rail E = 210GPa , a Poisson's ratio of wheel and rail ν = 0.283 , a wheel and railhead radius of curvature R = 0.46 m , a static preload P = 65 kN and a train speed v = 100 km/h . Parameters of the track are given in [11]. For the calculation of the Green's functions representing the track, receptances in the range 0 to 4 kHz have been used. In order to include the same frequency content in DIFF, the lowest 261 modal pairs are accounted for in the modal synthesis performed for the track. (a)

(b)

Fig. 3. Discrete Fourier transform of the contact force for two different roughness excitations. Comparison between the proposed model (⎯) and DIFF ( − ⋅ −): (a) Rail with sinusoidal corrugation λ = 0.015m and A = 10 μm , perfect wheel; (b) Corrugated rail and moderately rough wheel. Data from measured third-octave band spectra.

Fig. 3 shows results of the comparison in the form of discrete Fourier transforms of the calculated normal contact force. The cases considered are a rail with sinusoidal corrugation and a roughness profile calculated from measured data. The measured roughness data was available in terms of a third-octave band spectrum measured on a corrugated rail at the test site Vretstorp on the line Göteborg-Stockholm and a mean of five third-octave band spectra measured on X2 trailer wheels [11]. From the combined wheel/rail roughness spectrum, a sample of an energy-equivalent roughness sequence was generated [12]. It should be noted that this sequence differs from the original roughness sequence at the test site and does not necessarily generate the same contact forces as the original roughness. For both cases, the agreement between the proposed model and DIFF is very good. The additional, periodically appearing, peaks (Fig. 3(a)) in the DIFF simulations correspond to the element passing frequency (and multiples) in the finite-element model of the track and occur due to the noncontinuous slope of the deflection over the element boundaries. The simulations show that the proposed model is also very time efficient. Given precalculated Green's functions, the

A Time Domain Model for Wheel/Rail Interaction

291

computation time for one simulation is typically less than 20 seconds on a PC with a Pentium 2.0 GHz processor.

4 Conclusions A numerical time-stepping model to predict the wheel/track dynamic interaction has been presented. The wheel is represented by Green's functions and the track by moving Green's functions. The model includes a non-linear contact spring, allows for loss of contact and accounts for parametric excitation due to space-dependent track stiffness. The normal interaction model has been validated by comparison with results from the dynamic interaction model DIFF [3]. Very good agreement was obtained for the considered excitation cases. One major advantage of the presented model is its high computational efficiency, which can be attributed to the fact that the Green's functions are precalculated and, by this means, wheel, rail and contact calculations are separated. Another important advantage of the model presented is its potential to include frictional forces in a relatively simple and straightforward manner.

References [1] Thompson, D.J.: Wheel-rail noise generation, part I: Introduction and interaction model. Journal of Sound and Vibration 161(3), 387–400 (1993) [2] Nordborg, A.: Wheel/rail noise generation due to nonlinear effects and parametric excitation. Journal of the Acoustical Society of America 111(4), 1772–1781 (2002) [3] Nielsen, J.C.O., Igeland, A.: Vertical dynamic interaction between train and track – influence of wheel and track imperfections. Journal of Sound of Vibration 187(5), 825–839 (1995) [4] Wu, T.X., Thompson, D.J.: Theoretical investigation of wheel/rail non-linear interaction due to roughness excitation. Vehicle System Dynamics 34, 261–282 (2000) [5] Wu, T.X., Thompson, D.J.: A hybrid model for the noise generation due to railway wheel flats. Journal of Sound and Vibration 251(1), 115–139 (2002) [6] Wu, T.X., Thompson, D.J.: Wheel/rail non-linear interactions with coupling between vertical and lateral directions. Vehicle System Dynamics 41(1), 27–49 (2004) [7] Ford, R.A.J., Thompson, D.J.: Simplified contact filters in wheel/rail noise prediction. Journal of Sound and Vibration 293, 807–818 (2006) [8] Wullens, F., Kropp, W.: A three dimensional contact model for tyre/road interaction in rolling conditions. Acta Acustica united with Acustica 90(4), 702–711 (2004) [9] Heckl, M.: Proposal for a railway simulation program. In: A Workshop on Rolling Noise Generation, Institut für Technische Akustik, Technische Universität Berlin, pp. 128–148 (1989) [10] Andersson, P.B.U.: Modelling non-linear contact stiffness in tyre/road contact. In: Proceedings of the 19th International Conference on Acoustics, Madrid, Spain (September 2007) [11] Nielsen, J.C.O.: High-frequency vertical wheel-rail contact forces – validation of a prediction model by field testing. In: Proceedings of the Seventh International Conference on Contact Mechanics and Wear of Rail/Wheel Systems, Brisbane, Australia (September 2006) [12] Hiensch, M., Nielsen, J.C.O., Verheijen, E.: Rail corrugation in The Netherlands – measurements and simulations. Wear 253, 140–149 (2002)

Optimization of Track Parameters, Considering Their Physical Dispersion, to Minimize Rail Corrugation O. Oyarzabal, J. Gómez, J. Santamaría, and E.G. Vadillo Mechanical Engineering Department, University of the Basque Country UPV-EHU Escuela Técnica Superior de Ingeniería. Alda Urquijo s/n 48013 Bilbao, Spain Tel.: +3494-601-4223; Fax: +3494-601-4215 [email protected]

Summary In this paper, the influence on corrugation and therefore on the noise levels, of the most significant track parameters has been examined. After this parametric study, the optimization of the track parameters to minimize the undulatory wear growth has been achieved. Finally, the influence of the dispersion of the track and contact parameters on corrugation growth has been studied. This work is based on the computer application RACING (RAil Corrugation INitiation and Growth) which has been developed by the authors to predict rail corrugation features.

1 Introduction Rail corrugation is a periodic undulatory wear that frequently appears on the rolling surface of the rail. In Fig. 1, a particular case of corrugation studied in the surroundings of Bilbao can be observed [1, 2]. This undulatory wear provokes, apart from high dynamic loads between wheel and rail and components degradation, high levels of noise and vibrations. In 1993 Grassie and Kalousek published an article that reviewed 42 references and compiled the work accomplished from 1970 to 1990 [3]. In this comprehensive article, the undulatory wear is classified in six different types of corrugation depending on the wear mechanism and wavelength fixing mechanism. Although all authors do not agree with this classification, the basic idea of distinguishing the two mechanisms still applies [4]. Corrugation continues to be an active field of research [5, 6]. The work presented in this paper is based on a linear model developed by the authors to explain short pitch corrugation [7, 8]. The formation of short pitch corrugation is analysed using a feedback process combining wheelset and track dynamics, contact mechanics and wear. The track and wheelset dynamics are introduced into the global model by using receptances. The track model comprised in RACING takes advantage of both Periodic Structure Theory and the Finite Strip Method [9]. The contact quasi-static forces and creepages are obtained through DINATREN, a theoretical model for bogie curving developed by the authors [10].

2 Parametric Study From the particular case of corrugation studied in Bilbao mentioned above [1, 2] and using the application RACING, a parametric study has been developed. This study B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 292–298, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Optimization of Track Parameters, Considering Their Physical Dispersion

293

allows the influence on corrugation of the 14 most significant track parameters to be evaluated. These parameters are: • • • •

Distance between sleepers, which can vary from 0.6 to 1 m. Half sleeper mass, with characteristic values between 90 and 140 kg. Pad stiffness and damping in longitudinal, lateral and vertical direction. The stiffness ranges from 5·107 to 109 N/m, while the damping factor can take values between 0.1 and 0.8. Ballast or boot stiffness and damping in the longitudinal, lateral and vertical direction. The stiffness is of the order of 108 N/m, while the damping factor can take values between 0.1 and 0.8.

Fig. 1. Corrugation observed on the rail

With the parametric study it has been concluded that the five most important parameters are: distance between sleepers, sleeper mass, pad vertical stiffness, pad lateral stiffness and ballast vertical stiffness. The influence in G(f) (the function that provides the predisposition to appearance of corrugation depending on frequency in Hz [11]) of the first and the third parameter is shown in Figs 2 and 3. A high value of G(f) at a certain frequency means a high tendency of corrugation growth at that frequency. In Fig. 2, the G(f) function at midspan is depicted for three values of pad stiffness in the vertical direction. As it can be observed, the G(f) function changes notably in the 200-350 Hz range of frequency, when the pad stiffness in the vertical direction is changed. The lower the pad stiffness, the lower the corrugation peak becomes in these frequencies. Above 800 Hz, more peaks appear in G(f), but due to the frequency range where usually experimentally observed corrugation takes place, these variations at high frequency are not considered in this study. Fig. 3 shows G(f) at midspan for different distances between sleepers. It can be observed that between 200 and 400 Hz the maximum peak of the G(f) function is

294

O. Oyarzabal et al.

different for each sleeper spacing, both in amplitude and in frequency. For the three values of distance between sleepers considered in the figure, it can be concluded that the smaller the distance between sleepers, the lower is the maximum peak of corrugation. Therefore, according to this tendency, the optimum distance between sleepers would be the smallest value, i.e. 0.6 m. However, this is not what it is observed in the optimization study as it will be shown below. 8

x 10

-5

Distance between sleepers 8

L=0.8 m L=1 m L= 0.6 m

4 2 0 -2 -4 0

-5

Pad stiffness in the vertical direction kyp=10e8 N/m kyp=6e8 N/m kyp=2e8 N/m

6

G Function

G Function

6

x 10

4 2 0 -2

200

400 600 frequency (Hz)

800

Fig. 2. G(f) function at midspan for different distances between sleepers (L)

-4 0

200

400 600 frequency (Hz)

800

Fig. 3. G(f) function at midspan for different pad vertical stiffnesses (kyp)

3 Optimization As a conclusion of the parametric study, from the first five most important parameters, the three which have the main influence have been selected to carry out the optimization study. The optimal combination of the selected parameters that gives the smallest value for the maximum peak of G(f) is now obtained. 3.1 Nelder-Mead Optimization The optimization has been performed using the Nelder-Mead simplex algorithm [12], which is a direct-search method that uses only function values (does not require derivatives) and handles non-smooth functions. The Nelder-Mead method is implemented in Matlab [13]. The computational time required to obtain each of the optima is 36 hours, in a standard Pentium IV with 2 Gbytes of RAM memory. 3.2 Genetic Algorithms The Nelder-Mead optimization has several disadvantages, mainly that it does not explore the whole solution region and that it may find local minima. For these reasons, in this paper Genetic Algorithms (GA) have been used to find other combinations of parameters not based on a specific initial combination. The GA are very

Optimization of Track Parameters, Considering Their Physical Dispersion

295

useful for the optimization of this problem, that has many local maxima and minima. The optimization with the genetic algorithms has proved to be more effective than with the Nelder-Mead method, as it can be seen in Fig. 4, where above 200 Hz the solution obtained using GA is better than the one obtained using the Nelder-Mead method. 3.3 Results from the Optimization The combination of the three main parameters that lead to the lowest maximum peak of G(f) is considered as the optimum. Table 1 shows the four best combinations of parameters obtained running the optimization process. The optimization study has shown that the tendencies observed in the parametric study are not valid when all the possible combinations of parameters are taken into account. Namely, the parametric study shows that the smallest distance between sleepers gives the minimum for G(f). However, as it can be observed in the second column of Table 1, none of the values is 0.6 m. Something similar occurs for the pad vertical stiffness: the parametric study suggested that the lowest stiffness would be the optimal one, but as it can be checked in the fourth column of Table 1, two of the optimum solutions are higher than 5·107 N/m (the lowest stiffness). Therefore, by means of the parametric study it is known which are the most important parameters, but in order to obtain the best parameter combination, the optimization study is necessary. 2

x 10

-6

Nelder-Mead vs. Genetic Algorithm

G Function

1 0 -1 Nelder-Mead Genetic Algorithm

-2 -3 0

100

200 300 Frequency (Hz)

400

Fig. 4. The G(f) function at midspan obtained using the Nelder-Mead method and using the Genetic Algorithms Table 1. Combination of parameters giving the lowest maximum peak of G(f)

Optimum solution 1 2 3 4

Distance between sleepers (m) 0.75 0.76 0.92 0.99

Half a sleeper mass (kg) 96.72 133.32 135.35 138.97

Pad vertical stiffness(N/m) 5×107 5×107 7.4×107 7.8×107

Maximum peak of G(f) function 4.03×10-6 3.32×10-6 3.17×10-6 3.13×10-6

296

O. Oyarzabal et al.

Besides, the other four optimum solutions are of the same order. Fig. 5 left shows the frequency spectrum of the best and the worst optima of G(f). It can be concluded from this graph that all the optima shown in Table 1 are very similar in terms of corrugation growth. However, the difference in G(f) between the initial data and the data that comes from the optimization is remarkable (figure 5 right). Therefore, the combination of parameters obtained from the optimization leads to an important corrugation decrease.

5

x 10

-6

8 6

G Fu nctio n

G F un ction

0 -5

-10 -15 -20 0

x 10

Best optimization Worst optimization 200 400 Frequency (Hz)

-5

Initial data Best optimization

4 2 0

-2

600

-4 0

200 400 Frequency (Hz)

600

Fig. 5. Left: G(f) at midspan for the best and the worst optimization. Right: G(f) at midspan for the initial parameters and for the best optimized parameters.

4 Track Parameter Dispersion and Contact Parameter Uncertainties Due to the tolerances that appear during the track assembly, there is a dispersion in the track parameters and therefore in the dynamic properties. In conventional railways variations in the span length of +/-5% are acceptable. In the sleeper mass it is necessary to distinguish between the monobloc and bibloc sleeper. Infrastructure administrations accept a tolerance in the weight of monobloc sleepers of +/- 5.5%, whereas in the bibloc sleeper the upper tolerance is 4.3% and the lower tolerance is around 3% [14]. The pad stiffness in the vertical direction can vary around the nominal value by +/-10%. These uncertainties of the track parameters lead to differences in the track receptances as can be observed in Fig. 6 for a 96 kg half bibloc sleeper. If the distance between sleepers (L) is increased by 5% and the pad vertical stiffness (kyp) by 10%, the track vertical and lateral receptances in midspan change significantly. Variations in train speeds give rise to uncertainties in the contact parameters of each wheel. There have been estimated changes of 10 % in the semi-axes of the contact ellipse, in the lateral and longitudinal creepages and in the contact normal forces.

Optimization of Track Parameters, Considering Their Physical Dispersion

297

This estimation is based on calculations made with the program DINATREN to obtain the contact values while negotiating a bend [10].

10

10

10

Msleeper=96

-7

10 Lateral receptance at midspan

Vertical receptance at midspan

10

-8

-9

L=0.8; kyp=5e8 L=0.84; kyp=5.5e8

-10

0

200

400 freq(Hz)

600

800

Msleeper=96

-6

L=0.8; kyp=5e8 L=0.84; kyp=5.5e8 10

10

10

10

-7

-8

-9

-10

0

200

400 freq(Hz)

600

800

Fig. 6. Changes in track receptances (vertical and lateral) due to a change of 5% in distance between sleepers (L) and of 10% in the pad vertical stiffness (kyp)

It has been shown that the track parameters have a considerable dispersion which influences significantly track receptances, and that there are also important uncertainties in the contact parameters. Therefore, the corrugation growth tendencies are affected. As a consequence, although there is always a theoretical optimum solution, small changes in the track and in the contact values, due to this dispersion, could cause an important peak to appear in the G(f) function, which would imply an increase in the tendency to corrugation growth.

5 Conclusions In this paper a computational application developed by the authors (RACING) has been used to study the corrugation growth. This application obtains track and wheelset receptances and considers corrugation as a complex feedback determined also by contact parameters. The tool developed has been used to carry out a parametric study in order to find the influence of the 14 main parameters. Among the track parameters, the following five have more influence on the corrugation: distance between sleepers, sleeper mass, pad vertical stiffness, pad lateral stiffness and ballast vertical stiffness. The optimization study has been developed focusing the attention on the first three parameters. From the optimization work, combinations of parameters are obtained which give an optimum G(f) function, this is, the lowest maximum peak of G(f). Optimization using Genetic Algorithms has proved to be more effective than using the NelderMead method. The improvement obtained with the optimization is very significant. It has been considered that the distance between sleepers as well as the sleeper mass and the pad vertical stiffness have a physical dispersion that cannot be neglected. Because of this dispersion, changes appear in the track receptances. This

298

O. Oyarzabal et al.

work has been continued to take into account that small changes in the track and in the contact values could cause changes to appear in the G(f) function.

Acknowledgements The authors are grateful to CEDEX for their support through contract PT2006-02419CCPM. The authors also acknowledge the financial help received from the Department of Education, Universities and Research of the Basque Government.

References [1] Vadillo, E.G., Tárrago, J.A., Gárate, G., Angulo, C.: Effect of sleeper distance on rail corrugation. Wear 217, 140–146 (1998) [2] Gómez, I., Vadillo, E.G.: An analytical approach to study a special case of booted sleeper track rail corrugation. Wear 251, 916–924 (2001) [3] Grassie, S.L., Kalousek, J.: Rail corrugation: Characteristic, causes and treatments. Proc. Inst. Mech. Eng. 207, 57–68 (1993) [4] Grassie, S.L.: Rail corrugation: Advances in measurement, understanding and treatment. Wear 258, 1224–1234 (2005) [5] Sheng, X., Thompson, D.J., Jones, C.J.C., Xie, G., Iwnicki, S.D., Allen, P., Hsu, S.S.: Simulations of roughness initiation and growth on railway rails. Journal of Sound and Vibration 293(3–5), 819–829 (2006) [6] Jin, X.S., Wen, Z.F., Wang, K.Y., Zhou, Z.R., Liu, Q.Y., Li, C.H.: Three-dimensional train–track model for study of rail corrugation. Journal of Sound and Vibration 293(3–5), 830–855 (2006) [7] Gómez, I., Vadillo, E.G.: A linear model to explain short pitch corrugation on rails. Wear 255, 1127–1142 (2003) [8] Gómez, J., Vadillo, E.G., Santamaría, J.: A comprehensive track model for the improvement of corrugation models. Journal of Sound and Vibration 293(3–5), 522–534 (2006) [9] Gry, L.: Dynamic modelling of railway track based on wave propagation. Journal of Sound and Vibration 195(3), 477–505 (1996) [10] Santamaría, J., Vadillo, E.G., Gómez, J.: A comprehensive method for the elastic calculation of the two-point wheel-rail contact. Vehicle System Dynamics 44(supl.), 240–250 (2006) [11] Frederick, C.O.: A rail corrugation theory. In: Proceedings of the International Symposium on the Contact Mechanics and Wear of Rail/Wheel systems, Kingston, Rhode Island, pp. 181–211 (1986) [12] Vanderplaats, G.N.: Numerical optimization techniques for engineering design. McGrawHill, New York (1984) [13] MATLAB 7.2. User’s manual, The Mathworks. Inc, Natick (2006) [14] Euskotren, Provision and acceptance of bibloc sleepers for STEDEF slab track (In Spanish) EV-8-EV-1-003 (November 2000)

Predicting the Effect of Temperature on the Performance of Elastomer-Based Rail Damping Devices N. Ahmad1, D.J. Thompson1, C.J.C. Jones1, and A.H. Muhr2 1

Institute of Sound and Vibration Research, University of Southampton, Highfield, Southampton, SO17 1BJ, UK Tel.: +44(0)2380592294; Fax: +44(0)2380593190 [email protected] 2 Tun Abdul Razak Research Centre, Brickendonbury, Hertford, SG13 8NL, UK

Summary Rail dampers have been developed in recent years, formed by an elastomeric material and embedded steel masses. The loss factor and stiffness of the elastomer are very important for the performance of the system but, unfortunately, both are sensitive to changes in the temperature. Although having a high loss factor gives good noise reduction, it also means greater variation of stiffness, and consequently tuning frequency, with temperature. To investigate the effect of the temperature on the performance of a generic rail damper, a Timoshenko beam model of the track is used, to which is added a single-frequency tuned absorber. The noise reduction at each frequency is found from the ratio of the decay rates of treated and untreated beams. This is introduced into a typical noise spectrum obtained using TWINS. Account is next taken of the physical link between the damping loss factor and the stiffness variation with temperature. By assuming a constant loss factor, the rate of change of stiffness with log frequency is established. Then, using the time-temperature superposition principle, this can be expressed in terms of a temperature-dependence. This is finally used in the prediction of decay rates and thereby noise reduction. The results allow the relative importance of a high loss factor or a temperature-independent stiffness to be assessed.

1 Introduction The rail is usually the greatest source of railway rolling noise in the frequency range 500 to 2000 Hz and often forms the most important contribution overall [1]. One method of control that has been shown to be effective is to use rail dampers on the track. In one application a reduction of 6 dB(A) was found in the rail component of noise [1]. This paper considers an important aspect in the optimisation of the design of a rail damper. The damper considered here is based on an elastomeric material which is used with steel masses to form a tuned mass-spring system. The loss factor and stiffness of the elastomer are very important for the performance of the system. Both are sensitive to changes in the temperature and this must be taken into account as the device is required to operate in a range of environmental temperatures between about –20°C and 40°C. Ideally the elastomer in the rail damper (or ‘absorber’) should have a relatively high loss factor and a stiffness which does not vary strongly with temperature but these are physically conflicting requirements. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 299–305, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

300

N. Ahmad et al.

In this paper a simple model of a rail damper attached to a railway track is introduced and used to study the influence of its stiffness and damping on the noise reduction. By assuming a constant loss factor an estimate is made of the frequencydependence of the stiffness. Then, by using the time-temperature superposition principle, this is expressed as a temperature-dependence of the stiffness. Finally this is used to estimate the noise reduction as a function of temperature. The aim is to determine the relative merits of high damping and a low variation in stiffness, in order to establish a practical target for a new material.

2 Predicted Decay Rate The noise from the track depends on the rate of decay of waves propagating along the rail [2]. The sound power radiated by the rail is proportional to 1/Δ, where Δ is the decay rate in dB/m, given by Δ = –8.686 Im(k), where k is the complex wavenumber of waves in a rail [3, 4]. The untreated track is modelled as a Timoshenko beam on an elastic foundation (rail pad), as shown in Fig. 1a. By introducing the mass-spring system (absorber) to the untreated beam, as shown in Fig. 1b, an estimate is obtained for the decay rate of

(a)

(b)

Fig. 1. Model of the track (a) untreated, (b) with rail damper Table 1. Parameters used for railway track including absorber

Rail

Pad Absorber

Cross-sectional area Second moment of area Young’s modulus for steel Density for steel Timoshenko shear coefficient for rail Poisson’s ratio Damping loss factor of rail Support stiffness per unit length Damping loss factor of support Stiffness per unit length Damping loss factor Mass per unit length

A I E

ρ κ ν ηr Sp

ηp Sa

ηa μa

7.68×10-3 m2 3.0×10-5m4 2.11×1011 Nm-2 7850 kgm-3 0.4 0.3 0.02 3.8×108 Nm-2 0.16 6.91×108 Nm-2 0.35 17.5 kg/m

Predicting the Effect of Temperature on the Performance

301

10

10

Sound power level, dB

Decay rate [dB/m]

110 1

0

10

2

3

10 Frequency [Hz]

10

4

100 90 80 70 60 50 1 10

(a)

2

3

10 10 Frequency, Hz

10

4

(b)

Fig. 2. (a) The decay rate of the untreated (− −) and treated (—) track. (b) The predicted vertical component of rail noise: —, untreated track (113.2 dB(A)); − −, treated track (108.4 dB(A)). The input parameters are as in Table 1.

the track with absorbers. The ratio of the decay rates of the treated and untreated beams is used to determine the effect of the damping device. Only vertical vibration is considered in the model; this is justified in Section 3. Fig. 2a shows the predicted decay rate with and without the absorber. The parameters used are listed in Table 1. The tuning frequency of the absorber is set to 1000 Hz. The absorber introduces a large peak between 500 and 3000 Hz with its maximum around the tuning frequency. This corresponds to the region where the rail noise component is dominant (see Section 3 below). Increasing the stiffness of the absorber increases its tuning frequency whereas increasing the loss factor of the absorber will increase the breadth but decrease the peak of the decay rate.

3 Predicted Rail Noise In order to estimate the effect on the radiated sound, a prediction is made using the TWINS model [5] of the sound power radiated from an undamped track. The sound power of the damped track is then calculated by modifying the component associated with the vertical rail vibration predicted by the TWINS model. This initial situation represents a typical modern track with relatively soft rail pads, as in Section 2, a train speed of 100 km/h and a roughness spectrum corresponding to wheels with cast-iron block tread brakes. The noise from the rail is composed of components due to vertical and lateral vibration. In this untreated situation the vertical component is found to be 7.5 dB greater than the lateral component. Even in the treated situation, if the vertical component is modified by the effect of the rail absorbers but the lateral component is left unaffected, the vertical component would still be greater than the lateral component. In practice, the absorber will also introduce some damping to the lateral direction. However, from these comparisons it can be concluded that the lateral component can be neglected in optimising the effect of the absorber because of the dominance of the vertical motion. The predicted noise reduction has therefore been calculated in terms of the reduction in the vertical component. The sound power levels of damped and undamped

N. Ahmad et al.

3.53 2. 5

-1

0.5

10

5. 5

1. 5 2

0

1.5 1

Loss Factor

4 4.5

10

5

302

1 0.5

10

-2

0

10

8

Stiffness, N/m 2

10

9

Fig. 3. Contour plot of reduction of rail vertical component of noise (in dB) for a range of loss factors and stiffnesses (corresponding to tuning frequency between 250 and 2000 Hz)

tracks are shown in Fig. 2b, based on the parameters in Table 1. From this frequencydependent effect, the absorber is found to give a reduction in overall A-weighted noise level of 4.8 dB for this component of rail noise. Fig. 3 shows the reduction in A-weighted noise level for a range of values of the loss factor and stiffness (corresponding to a range of tuning frequency of 250 to 2000 Hz). This shows that the maximum noise reduction is achieved for loss factors of 0.3 and above, and a stiffness of about 4.0×108 N/m2 (tuning frequency of about 800 Hz). As the loss factor is increased beyond 0.3, the noise reduction increases, but only slightly.

4 Temperature Dependence of Loss Factor and Stiffness The dynamic properties of the absorber will vary with temperature: the higher the loss factor, the bigger the change in the stiffness over a range of temperatures. If the stiffness varies too much then the absorber will be less effective at extremes of temperature. Conversely if the loss factor is too low the effect of the absorber is reduced. The balance between these two effects over the range of temperature required is now investigated. It is known that, for viscoelastic materials, the slope of the shear storage modulus G' with log frequency is related to the loss modulus G'' by [6]

dG′ 2G′′ ≈ π dx

(1)

where x = loge(f). The loss factor η can be written as η = G''/G'. Substituting this into Eq. (1) gives,

dG′ ⎛ 2η ⎞ = ⎜ ⎟ G′ dx ⎝ π ⎠

(2)

Although the loss factor is generally dependent on frequency and temperature, insight can be gained by assuming a constant value of η. The solution to this can be written as

Predicting the Effect of Temperature on the Performance

G′ ( f2 ) ⎛ f 2 ⎞ =⎜ ⎟ G ′ ( f1 ) ⎝ f1 ⎠

2η

303

π

(3)

for two frequencies f1 and f2. Next, this frequency-dependence has to be converted into a temperature-dependence. According to the principle of time-temperature superposition [7], it is possible, for so-called ‘thermo-rheologically simple materials’, to superimpose the curves of G' against log10 f by adding a temperature-dependent factor log10 α(T) called a shift factor, to log10 f. The Williams-Landel-Ferry (WLF) model [7] can be used to generate the relationship between f and T. This model is found to apply to a wide range of elastomeric materials [7, 8]. The WLF model predicts the shift factor as

log10 α (T ) =

−8.86(T − Ts ) 101.6 + T − Ts

(4)

where Ts is a reference temperature, applying to a given material. It can be estimated approximately as Tg+50, where Tg is the glass transition temperature of the material [8]. Using this with Eq. (3), the ratio of storage modulus at two temperatures T1 and T2 and for a given frequency is given by,

⎛ G′ (T1 ) ⎞ 2η ⎛ 8.86 (T2 − Ts ) 8.86 (T1 − Ts ) ⎞ log10 ⎜⎜ − ⎟⎟ ≈ ⎜ ⎟ ⎝ G′ ( T2 ) ⎠ π ⎝ 101.6 + T2 − Ts 101.6 + T1 − Ts ⎠

(5)

Ratio of G

It is now necessary to select a value of Ts. Taking, for example, a value of Tg of – 70°C, corresponding to butyl rubber, Ts will be about –20°C. Fig. 4 shows the variation of G' with temperature for various loss factors, for a value of Ts of –20°C. The slope of G' increases as the loss factor increases. Here, the value of G' is shown relative to the value at 10°C.

10

0

-20

0

20

40

Temperature (oC)

Fig. 4. The ratio of G plotted against temperature and normalised to result at 10°C for Ts of –20°C and various loss factors: —, η =0.05; ⋅⋅⋅⋅⋅⋅, η =0.1; − −, η =0.2; − · −, η =0.4; −□−, η =0.8

304

N. Ahmad et al.

5 Effect of Temperature Dependence on Noise Reduction The temperature variation of the stiffness can be introduced into the noise predictions, based on the results of the previous section. Dynamic properties at 10°C are used as nominal input parameters in the prediction of noise reduction. These are based on a tuning frequency of 800 Hz. The stiffness is then varied according to the ratio given in Fig. 4 as temperature varies and according to Eq. (3) as frequency varies. Fig. 5 shows the reduction of the vertical rail noise as a contour plot against temperature and damping loss factor. Low loss factors give less noise reduction but the results are less sensitive to the change of temperature than at high loss factors. A loss factor of 0.1 enables a noise reduction of 3.5 dB to be sustained across most of the temperature range. A loss factor of 0.3 allows a maximum reduction of 5.5 dB to be achieved and this remains above 4.5 dB between –10°C and 40°C. However, when the loss factor is increased to 1.0, although the maximum noise reduction is 6 dB, the results are much more strongly dependent on temperature due to the higher variation in stiffness. Generally, a loss factor around 0.25 to 0.4 gives the best results across the range of temperatures considered. The value of Ts assumed also affects the results. In general, it is not easy to find a material with a low Ts and moderately high η, but butyl, having Ts ≈ –20°C, comes reasonably close.

Loss Factor

6 6

3.5

4 5 4 .5 5.5

5

3 3.5 4 4. 5

5. 5

0. 5 1 1. 5 2 2.5 3

0

10

5.5

4. 5 -1

5 4.5

4

10

4 3.5

3.5

3

-20

0 20 o Temperature, C

40

Fig. 5. The reduction of vertical rail noise (in dB) at various temperatures and loss factors. The parameters used are as in Table 1 with Sa =4.36×108 N/m2 at 10°C, Ts = –20°C.

6 Conclusions In this paper, an attempt is made to investigate the optimum loss factor and stiffness of the rail damper elastomer, in terms of the reduction of noise that can be achieved. Knowing that the stiffness is very sensitive to temperature, an approximate technique has been adopted to estimate this effect. By assuming a constant loss factor, the variation in stiffness across the temperature range has been estimated assuming only a value for the temperature Ts as used in the WLF equation [6]. It is shown that the

Predicting the Effect of Temperature on the Performance

305

noise reduction can be maintained within 1 dB of the maximum effect in a range – 10°C to 40°C for a loss factor of about 0.3. With lower loss factors the result is less sensitive to temperature but the overall reduction obtained is smaller. With higher loss factors, although the same maximum reduction can be achieved this is much more sensitive to variations in temperature. While the assumption of a constant loss factor is rather artificial, and in practice the loss factor is not independent of the choice of Ts, this study serves to show how the interdependence of the loss factor and the rate of change of stiffness affect the performance of a rail damper.

References [1] Thompson, D.J., Jones, C.J.C., Waters, T.P., Farrington, D.: Tuned damping device for reducing noise from railway track. Applied Acoustics 68, 3–57 (2007) [2] Jones, C.J.C., Thompson, D.J., Diehl, R.J.: The use of decay rates to analyse the performance of railway track in rolling noise generation. Journal of Sound and Vibration 293, 485–495 (2006) [3] Wu, T.X., Thompson, D.J.: Analysis of lateral vibration behaviour of railway tracks at high frequencies using a continuously supported multiple beam model. Journal of the Acoustical Society of America 106, 1369–1376 (1999) [4] Thompson, D.J.: The theory of a continuous damped vibration absorber to reduce broadband wave propagation in beams. In: ISVR Technical Memorandum, vol. 986, University of Southampton (2007) [5] Thompson, D.J., Hemsworth, B., Vincent, N.: Experimental validation of the TWINS prediction program for rolling noise, part 1: Description of the model and method. Journal of Sound and Vibration 193, 123–135 (1996) [6] Schwarzl, F.R., Struik, L.C.E.: Analysis of relaxation measurement. Advance in Molecular Relaxation Processes 1, 201–255 (1968) [7] Ferry, J.D.: Viscoelastic Properties of Polymers, 2nd edn. Wiley, New York (1970) [8] Williams, M.L., Landel, R.F., Ferry, J.D.: The temperature dependence of relaxation mechanisms in amorphous polymers and other glass-forming liquids. Journal of the American Chemistry Society 77, 3701–3707 (1955)

Estimation of Sound Transmission through Extruded Panels Using a Coupled Waveguide Finite Element-Boundary Element Method C.M. Nilsson, A.N. Thite, C.J.C. Jones, and D.J. Thompson Institute of Sound and Vibration Research, University of Southampton, Southampton, SO17 1BJ, UK Tel.: +44 23 8059 3224; Fax: +44 23 8059 3190 [email protected]

Summary A coupled waveguide finite element and wavedomain boundary element method is presented. This numerical method is suitable for analysing systems with uniform properties along one direction, but with complex cross-sections. Subsequently the transmission loss through an extruded aluminium panel, of a type commonly used in railway carriages, is calculated and compared to measurements.

1 Introduction The calculation of sound transmission through complex panels has been of interest for many years. Recently, the railway rolling stock manufacturing industry has adopted a method using double-skinned extrusions of aluminium over the whole length of a carriage. These offer advantages in manufacturing, crashworthiness and lightness. However, they are poor acoustic insulators and expensive noise control measures must often be employed. Because of the shape and size of the structures, ordinary numerical methods of finite elements (FE) and boundary elements (BE) cannot be used efficiently for these panels. The industry uses modelling methods like Statistical Energy Analysis (SEA) and attention has been given to how this should be implemented for these panels, [1-4]. In general, SEA is applied at an early stage of the design process where details are lacking. In [1] commercial software was used to obtain airborne transmission loss (TL): here estimations disagreed with measurements, especially at higher frequencies. In [2] two models with varying details of information were used. The first, less detailed model, gave better agreement with measurements. At low frequencies both models underestimated the transmission loss. In [3] low frequency global modal behaviour is not considered and also there were differences at higher frequencies compared with measurements. It is suggested that the influence is due to choices of damping and radiation efficiency. In [4] estimation of modal density and radiation efficiency is addressed. Proper estimation of these parameters can improve SEA results significantly. However, there are outstanding issues of determination of the number of subsystems, interaction and simplifications etc. Applying SEA to extruded profiles is still an area with a lot of research interest. One shortcoming of SEA, for the application considered in particular, B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 306–312, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Estimation of Sound Transmission through Extruded Panels

307

is the lack of detail in the results. Also, there is a mid-frequency region that is of interest for airborne transmission that is not properly addressed by SEA. There is a need for a vibro-acoustic modelling approach that is efficient, provides insight into structural/acoustic behaviour and contains enough detail to allow parametric studies. A numerical method is presented in this paper that is based on waveguide FE and wave-domain BE theory. Waveguide FE models have been presented by several researchers, e.g. [5] where the method was employed for studying waves in thin walled-structures. In [6] car tyre vibrations were studied. Wave-domain BE for acoustics have been developed in [7] and used to calculate the sound field around barriers and in [8] for studying propagation of environmental noise across cuttings. Here a method for the coupling between plate strip waveguide FE and fluid wavedomain BE is presented. The method is applied to calculate the sound transmission loss through an extruded aluminium panel. The panel is assumed to be of finite width but infinite length. The latter assumption can explain discrepancies with measurements at low frequencies. High frequency results are found to be very sensitive to changes in damping loss factors, which is known to have a large effect close to cut-on frequencies [9]. This sensitivity offers a possible means to increase high frequency transmission losses in extruded panels.

2 Waveguide Finite Element Model The uniformity of cross-section in one direction allows wave domain solutions with respect to this direction. The derivation of the shell strip elements are given by a number of references, e.g. [5-6] and complex structures are analysed by assembling many strips. Fig. 1 shows a plate strip which has out-of-plane displacement w and inplane displacements u and v. Equations of motion are formed by using Hamilton’s principle, expressions for strain and kinetic energies and by assuming test functions and trial solutions with dependence on the y-direction. These equations include stiffness and mass matrices, but different orders of differentiation with respect to the xdirection. The equations of motion are then given by

{

k4

}

∂4 ∂2 ∂ + k + k 1 + k 0 − mω 2 U = F 2 4 2 ∂x ∂x ∂x

z, w Φ x, u a

y, v a Fig. 1. Plate strip element

(1)

308

C.M. Nilsson et al.

where k i are stiffness matrices, m is a mass matrix, F is a vector of forces on nodal lines, and U is a vector containing rotations about, and displacements along, the strip’s edges. Assembling the element matrices and assuming wave solutions e − iκ x , the set of equations of motion become,

{K (κ ) − Mω 2 } U = F

(2)

U is a vector of nodal displacement depending on both frequency and wavenumber and F is a force vector.

where

3 Waveguide Boundary Element Method For derivation of the BE formulation consider a fluid system. Applying Hamilton’s principle, using expressions for kinetic and potential energy in terms of velocity potential and taking Fourier transforms to convert to the wavenumber domain, the equation of motion can be written as, * ⎧⎪ * ∂Ψ ⎫⎪ ∂δ Ψ ρ ∫ δ Ψ {Δ 2D Ψ + ( k − κ ) Ψ} dA − ∫ ⎨δ Ψ −Ψ dΓ ⎬ = 0 (3) ∂n ∂n A Γ⎩ ⎪ ⎭⎪ *

2

2

∂2 ∂2 , n is the vector normal to the surface, A is the area of the + ∂y 2 ∂z 2 domain, Γ is the boundary and δ denotes the first variation. The velocity potential Ψ , is defined such that, in the spatial domain: v = −∇Ψ , p = iωΨ . Here v = f ( x, y, z , ω ) is the particle velocity of the fluid, p is the acoustic pressure, k = ω / c is the wavenumber of freely propagating waves, c is the velocity of sound

where

Δ 2D =

in the fluid and ρ is the fluid density. Compared with a two dimensional (2D) model, Eq. (3) differs only in that

k 2 − κ 2 is used instead of k 2 . Hence a 2D BE model

may be used by substituting the argument k by

α = k2 −κ 2

as long as

k 2 > κ 2 , which is the case here. The matrix equation for the boundary element method is written as,

GP - HV n = P field

(4)

P and V n are vectors containing the nodal pressures and normal velocities of the boundary element model respectively. P field is the incoming pressure field evalu-

where,

ated on the boundary nodes. H and G are, generally full and complex-valued, matrices.

Estimation of Sound Transmission through Extruded Panels

309

4 Boundary Conditions The mixed boundary conditions for a boundary element method, for nodes not on the BE/FE interface can be written as [10],

CA P + CB V n = cc

(5)

where CA and CB are diagonal matrices and c c is a vector corresponding to pressure sources and moving boundaries. The essential boundary condition that must be fulfilled for the wetted boundary between the fluid BE and the plate FE model is,

iω UC2 = I 0 Vn where

(6)

I 0 is a diagonal matrix such that an entry on the diagonal equals one if the

node is connected to a boundary element and zero otherwise. C2 is a matrix projecting the displacements of the FE model onto normal displacements of the BE model. For any node on the fluid BE model there is only one boundary condition, given as a line in either Eq. (5) or (6). These two equations can be added to form a fully determined system.

5 Waveguide FE-BE Approach: System Matrix Manipulating Eq. (2) and (4), and using boundary conditions of Eq. (5) and (6), the following system matrix is obtained,

⎧ ∂Ψ ⎫ ⎡ ⎤ ⎪ ∂n ⎪ ⎧P field ⎫ ⎪⎪ ⎪⎪ ⎪ ⎪ ⎢ 0 −iωC1 K (κ ) − ω 2 M ⎥ ⎨ Ψ ⎬ = ⎨ F e ⎬ ⎢ ⎥ iωC2 ⎢⎣I 0 − CB iωρ CA ⎥⎦ ⎪ U ⎪ ⎪ cc ⎪ ⎭ ⎪ ⎪ ⎩ ⎩⎪ ⎭⎪ H

iωG

0

(7)

where C1 is a matrix that projects the pressure of the fluid onto the structure. A similar matrix can be constructed when a second BE model is connected. The above equation can be solved for pressure distribution for given wavenumber, frequency and input pressure field and/or force combination.

6 Complex Extruded Panel An extruded aluminium panel is considered. The model is considered to be infinite in length, whereas the panel tested experimentally (see below) is 1.5 m along the extrusion. The boundaries in the model are simply supported. The panel skin on the receiver side is covered with a rubber mat, the material properties of which are not

310

C.M. Nilsson et al.

known precisely. In the model it is considered not to add to the stiffness although the mass and damping are assumed to increase.

7 Waveguide Approach Estimations The system model is made of one FE model for the panel and two BE models, one on each side of the panel. For each frequency 20 incident waves of unit pressure amplitude and random phase are imposed at uniformly distributed angles in the yz plane. For each of these there are 17 incident waves at uniformly distributed angles in the zx plane. The incident power for each angle is calculated and the radiated power for each wavenumber is estimated as

PtWFeBe =

1 pt vt dΓ 2 ∫Γ

(8)

where Γ is the boundary of the BE on the receiver side. The transmission loss can then be calculated and averaged over different incident angles. The calculations outlined above are much more efficient than those made for a corresponding full 3D FE/BE model for two reasons. Firstly, the number of degrees of freedom (DOF) in the current model, N, corresponds to a 2D FE/BE model. The computing time for large linear systems of equations increases according to a power of N greater than unity. So, splitting a 3D problem into a number of smaller 2D problems is very efficient. Secondly, the x-domain can be quite large, whereas the wavenumber domain is limited since the in-coming and out-going waves cannot travel faster than the speed of sound in air, i.e. κ ≤ ω c . Fig. 2a shows the dispersion relation for the extruded panel calculated from equation (2). The bold line is the shear wavespeed in the composite structure. The average transmission loss predictions in Fig. 2b show the influence of damping on sound transmission across the panel. From Fig. 2c and 2d, it can be concluded that waves corresponding to global composite bending modes are significantly influenced by damping as well as the local mode waves and there is an overall reduction in power radiated from the panel with increased damping.

8 Measurement of Transmission Loss The 1.0 m × 1.5 m panel was mounted in an aperture between a reverberation chamber and an anechoic chamber. The rubber layer side faced the anechoic room. The cross-section of the panel is shown in Fig. 3a. The panel was mounted in a wooden frame and the exposed edges sealed with silicone rubber. A diffuse field in the reverberation room was set-up by a set of loudspeakers that excite both lower and higher frequencies. The average sound pressure was measured in the reverberation room from which the incident sound power could be estimated. On the anechoic room side the sound pressure was measured at 2m radius on a hemisphere. Then, the average transmission loss was obtained in 1/3 octave bands. Fig. 3b shows a comparison of measured results and numerical results. Above 500 Hz the results agree very well.

Estimation of Sound Transmission through Extruded Panels

311

150

Transmission Loss [dB re 1]

Wavenumber [rad/m]

60

100

50

50

40

30

η=0.02 η=0.2

20

10

0 0

2000

4000 6000 Frequency [Hz]

8000

0 2 10

10000

3

4

10 Frequency [Hz]

(a)

10

(b)

160 160

140

Wavenumber [rad/m]

Wavenumber [rad/m]

140

120 100 80 60

120 100

40

60 40

20 0

80

20

1000

2000

3000

4000 5000 6000 Frequency [Hz]

7000

8000

0

9000

(c)

1000

2000

3000

4000 5000 6000 Frequency [Hz]

7000

8000

9000

(d)

Fig. 2. Transmission loss across of extruded panel for different damping values. (a) dispersion curves of extruded panel, (b) transmission loss, (c) power radiated for structural damping η is 0.02, and (d) power radiated for structural damping η is 0.2. (Lighter shading means more power.) 60

Transmission loss [dB re 1]

50

40

30

Numerical simulation Measurements

20

10

0 2 10

(a)

3

10 Frequency [Hz]

4

10

(b)

Fig. 3. Measurement results- a) extruded plate picture showing cross section and b) transmission loss- measured and coupled BE and waveguide FE results

312

C.M. Nilsson et al.

Between 200 Hz and 500 Hz there is a dip in the numerical results because of the onset of global modes across the waveguide. In the measured results a fundamental mode occurs much earlier, about 70 Hz, as the panel has finite length of 1.5 m along the waveguide direction.

9 Conclusions The transmission loss through a complex, extruded aluminium panel has been calculated with an efficient combination of the waveguide finite element and wavedomain boundary element approaches. Calculated results indicate good agreement with measurements as well as providing means that aid interpretation of the physics of the system. The system’s sensitivity to damping is shown by the model. This may offer a way to increase the transmission loss through panels used in railway carriages.

Acknowledgement This work was funded by the Engineering and Physical Sciences Research Council of the UK under grant number GR/S81254/01. The assistance of Bombardier Transportation is also gratefully acknowledged.

References [1] Geissler, P., Neumann, D.: Modelling extruded profiles for railway coaches using SEA. In: Proc. ASME DETC, Las Vegas (1999) [2] Bruhl, S., Faulhaber, P., Grunewald, M.: AutoSEA2 studies on generic light weight structures. In: Proc. of the first international AutoSEA users conference, San Diego, CA (2000) [3] Shaw, N.J.: The prediction of railway vehicle internal noise using statistical energy analysis techniques, MSc. Thesis, Heriot-Watt University, Edinburgh (1990) [4] Xie, G., Jones, C.J.C., Thompson, D.J.: A modelling approach for the vibro-acoustic behaviour of aluminium extrusions used in railway vehicles. Journal of Sound and Vibration 293, 921–932 (2006) [5] Gavric, L.: Finite element computation of dispersion properties of thin-walled waveguides. Journal of Sound and Vibration 173, 113–124 (1994) [6] Nilsson, C.-M.: PhD Thesis, Waveguide Finite Elements Applied on a Car Tyre. KTH, Stockholm (2004) [7] Duhamel, D.: Efficient Calculation Of The Three-Dimensional Sound Pressure Field Around A Noise Barrier. Journal of Sound and Vibration 197, 547–571 (1996) [8] Peplow, A.T.: Noise Propagation from a Cutting of Arbitrary Cross-section and Impedance. Journal of Sound and Vibration 223, 355–378 (1999) [9] Nilsson, C.-M., Finnveden, S.: Input power to waveguides calculated by a finite element method. Journal of Sound and Vibration (in press) [10] Wu, T.W.: Boundary Element Acoustics, Fundamentals and Computer Codes. WIT Press (2000)

Squeal Prediction for a Bogied Vehicle in a Curve Z.Y. Huang, D.J. Thompson, and C.J.C. Jones Institute of Sound and Vibration Research, University of Southampton Southampton, SO17 1BJ, England [email protected]

Summary Curve squeal is an intense high pitched noise generated by railway wheels when traversing tight curves. In this paper, a general squeal model is presented, which is based on the interrelationship between the wheel/rail contact forces and their responses, in the longitudinal, lateral, vertical and spin directions. Using the parameters of steadystate curving behaviour of a passenger vehicle for a range of curve radii, the general curve squeal model can predict the occurrence of squeal at all four wheel/rail contacts in a bogie. The effectiveness of conventional control methods for curve squeal, i.e. wheel damping treatments and friction modification, are investigated.

1 Introduction During the passage of a vehicle through a curve the wheels do not align with the rolling direction. Some wheels run with the flange rubbing against the rails and others are subject to large lateral micro-slippage (termed creepage) within the contact region. It is generally accepted that, under some conditions, when the creepage exceeds a certain value, the friction force falls with further increases of the relative velocity. This may make the wheel/rail system experience unstable vibration and consequently squeal may occur. Many theoretical models of curve squeal have adopted this mechanism [1-3], as does the model presented here. In this model, the wheel/rail dynamic properties and friction forces are considered for all the possible directions at the contact zone, including the longitudinal, lateral, vertical and spin directions. The selfexcited vibration between the wheel and rail contact forces and their responses can be established as a feedback loop. The step-by-step integration of the loop can show the limit-cycle responses and indicate the intensity of the squeal. In order to predict reliably the onset of curve squeal, a dedicated steady-state curving program, validated using a commercial vehicle dynamic package, is used to provide the steady-state curving behaviour of a typical UK passenger vehicle, the Class 158 DMU, in various tight curves. Once the vehicle behaviour has been determined, the curve squeal calculation is applied on all four wheel/rail contacts in a bogie. The effectiveness of traditional control methods for the curve squeal, i.e. wheel damping treatments and friction modification, are also investigated. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 313–319, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

314

Z.Y. Huang, D.J. Thompson, and C.J.C. Jones

2 General Squeal Model 2.1 Wheel/Rail Contact Dynamics The wheel/rail contact system is shown in Fig. 1. The reference frame is established at the contact patch, where there are four degrees of freedom: the longitudinal and lateral directions (with index 1 and 2) in the contact plane, the ‘vertical’ direction (with index 3) normal to the contact plane, and the spin (with index 6) about the normal direction. In the longitudinal, lateral and spin directions, the dynamic sliding velocities vsi exist. In the vertical direction, the approach of wheel and rail can be considered as the compression of the contact spring. For steady-state curving at a speed V0, the sliding velocities are normally evaluated in terms of creepages γi. The total creepages can be split into steady-state and dynamic components:

γ i ≡ γ i 0 + vis / V0 , i = 1, 2, 6

(1)

where the steady-state creepages γi0 are determined by the steady-state curving behaviour, and the dynamic sliding velocities vsi are determined by the wheel and rail responses at the contact. The longitudinal and lateral friction forces and the spin moment can be evaluated by the product of the normal contact force and corresponding velocity-dependent friction coefficients. The steady-state components of friction forces and moment are balanced by the vehicle suspension forces, which means only the dynamic components are directly related to squeal. 1 2

6 3

sliding surfaces

contact spring

reference frame Fig. 1. Wheel/rail contact system

2.2 Wheel and Rail Structural Dynamics A Class 158 wheel is modelled by FE analysis, which provides the modal parameters at the nominal wheel/rail contact point (at the centre of wheel tread). Using the modal parameters, the modal analysis method can be applied to establish a state-space model of the wheel. A track based on UIC 60 rail is modelled analytically to provide dynamic properties. These are expressed in state-space form by fitting rational polynomial models to the drive point responses. The longitudinal and lateral mobilities of wheel and rail are shown in Fig. 2. The modal parameters of the most flexible circumferential (longitudinal) and lateral wheel modes are listed in Tables 1 and 2, respectively, although a full set of modes is included in the calculations. The damping ratio of each wheel mode is assumed based

Squeal Prediction for a Bogied Vehicle in a Curve

315

on previous measurement results with rounded values [4]. If the wheel/rail contact position moves to the wheel flange, the contact plane will slope at a large contact angle to the ground plane; this will lead to significant changes to the mobilities at the contact plane, which are taken account in the model. -2

-2

10 -1 -1

Modulus (ms N )

-1 -1

Modulus (ms N )

10

-4

10

-6

10

-6

10

-8

-8

10

-4

10

1

2

10

3

10

10

10

1

10

Frequency (Hz)

2

3

10

10

Frequency (Hz)

(a)

(b)

Fig. 2. The mobilities of a Class 158 wheel and UIC 60 rail at contact, (a) longitudinal, (b) lateral. ⎯ wheel; − − − rail. Table 1. Circumferential flexible modes of Class 158 wheel below 5 kHz Mode (n,c) 1 (0,c) (1,c) (2,c)

Frequency (Hz) Damping ratio Modal mass (kg) Longitudinal modeshape 70 0.001 1 5.2×10-2 2136 0.01 1 4.3×10-2 4924 0.0001 1 9.8×10-2

Table 2. Main lateral flexible modes of Class 158 wheel below 5 kHz Mode (n,m) 2 axle (2,0) (3,0) (4,0) (5,0) (6,0)

Frequency (Hz) Damping ratio 82 0.01 418 0.0001 1102 0.0001 1976 0.0001 2950 0.0001 3977 0.0001

Modal mass (kg) 1 1 1 1 1 1

Lateral modeshape 6.1×10-2 1.1×10-1 1.1×10-1 1.2×10-1 1.2×10-1 1.1×10-1

2.3 Nonlinear Friction Force The three-dimensional steady-state rolling friction can be obtained from Kalker’s FASTSIM algorithm [5]. The falling slope of the friction curve in gross sliding has been described mainly on the basis of experimental findings. Accounting for the falling characteristics at large creepages, a heuristic approach is adopted with following factor which is applied to the friction coefficient: 1

n: number of nodal diameters on the wheel plane; c: circumferential mode.

2

m: number of nodal circles on the wheel plane.

316

Z.Y. Huang, D.J. Thompson, and C.J.C. Jones

τ (γ ) = 1 − λ e −κ / γ where

λ

is the falling coefficient, and

κ

(2)

is the saturation coefficient which deter-

mines the value of creepage where the saturation region starts. To ensure a suitable falling friction in the saturation zone,

κ = 0.005

is used. The coefficient

λ

can

only be obtained from related experiments. The lateral creep-force relation adopted in the simulation is shown in Fig. 3. There are four friction curves. Two curves are for lateral friction in terms of only lateral creepages but with different falling coefficients λ = 0.1 and 0.3. The other two lateral friction curves are influenced by longitudinal and/or spin creepages: one has a small longitudinal creepage γ10 = 0.003; the other has a large spin γ60 = 1.5 m-1 as well as γ10 = 0.003. Both curves have the same falling coefficient λ = 0.3. Either a longitudinal creepage or a large spin can change the friction characteristics in the lateral direction. In the simulation, the friction forces (and spin moment) are obtained by using FASTSIM at each step of the simulation.

Normalised friction

1

0.5

0

-0.5

-1 -0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

Creepage

Fig. 3. The lateral friction with respect to the lateral creepage, where the friction force is normalised by the Coulomb friction. ⎯ λ = 0.1; − − − λ = 0.3; ⋅⋅⋅⋅⋅⋅ λ = 0.3, with γ10 = 0.003; − ⋅ − ⋅ λ = 0.3, with γ60 = 1.5 m-1 and γ10 = 0.003.

2.4 Self-Excited Vibration Loop A general self-excited vibration model has been developed by combining the wheel/rail structural models and the wheel/rail rolling model in a loop, as shown in Fig. 4. If some small transient disturbances {f'} are introduced to the steady-state conditions, the wheel and rail will produce dynamic responses {vs}, which can give dynamic sliding velocities and vertical compression. In the contact area, the dynamic friction forces may be produced due to either the dynamic sliding velocities or the vertical fluctuating force. Consequently, the contact forces {f}, including friction forces and vertical fluctuating force, are updated and fed back to the wheel and rail system. The time-domain simulation is realised by the step-by-step integration from any small disturbance until the limit-cycle responses of wheel and rail are reached.

Squeal Prediction for a Bogied Vehicle in a Curve

disturbances

wheel model

{ f'} +

+ +

+

317

{vs}

rail model

{f}

wheel/rail contact dynamics

+

+

steady-state conditions

Fig. 4. The self-excited vibration loop of wheel/rail contact system

3 Simulation 3.1 Steady-State Curving The steady-state curving behaviour of a Class 158 vehicle are given in Table 3. Since the curve radius is sharp in each case, it is seen that the leading outer wheel/rail contact is at the flange, with a large spin as well as a large lateral creepage, while the leading inner contact has a large lateral creepage but no spin. Both wheels of the trailing wheelset may have large longitudinal creepages but very small lateral creepages, especially when the curve radius is very tight. Table 3. Steady-state curving behaviour of Class 158 vehicle, no cant deficiency, Coulomb coefficient μ0 = 0.3 Curve radius Curve speed Wheel position (m) (m/s) 150

6

200

8

300

12

Leading outer Leading inner Trailing outer Trailing inner Leading outer Leading inner Trailing outer Trailing inner Leading outer Leading inner Trailing outer Trailing inner

γ10 (%) 0.3 -0.3 -0.9 0.9 0.4 -0.4 -0.6 0.6 0.5 -0.5 -0.3 0.3

Creepage γ20 (%) γ60 (m-1) -1.9 -1.5 0.2 0.2 -1.6 -1.2 0.0 0.0 -1.0 -0.8 0.0 0.0

-1.5 0.0 -0.0 0.6 -1.4 0.0 -0.0 0.4 -1.3 0.0 -0.1 0.1

Normal force (kN)

Contact Angle (º)

64 59 62 61 65 59 57 66 67 58 62 61

39 -0.8 0.9 -16 37 -0.9 1.0 -8.6 34 -0.9 1.6 -3.3

3.2 Squeal Results The squeal results for the leading wheelset in a bogie, curving at 6 m/s in a curve of radius 150 m, are shown in Fig. 5. Both leading wheel/rail contacts under this curving behaviour can generate squeal in one of the lateral flexible modes of the wheel. The self-excited motion is almost a sinusoidal oscillation, with a little distortion at small sliding creepage. This distortion can produce high-frequency harmonics of the dominant frequency. The noise radiation of wheel is calculated using an engineering

318

Z.Y. Huang, D.J. Thompson, and C.J.C. Jones

(b.1)

0

(a.3)

0.997 0.998 Time(sec) (s) Time

0.999

-20 -40

(b.2)

-60 -80 -100 -120 1 10

2

10 Frequency (Hz) Frequency (Hz)

3

10

100 80 60

(b.3)

40 20

32

63

130

250

500

Frequency (Hz)

1000

2000

4000

0

-0.05 0.995

1

2 -2

0.996

(re1 1m dBdB(re m2ss-2/Hz) /Hz)

2 -2

SPL (dB(A) re 20 μPa)

(a.2)

dBdB(re /Hz) (re11 m m2ss-2/Hz)

-0.05 0.995

Creepage Creepage

0.05

SPL (dB(A) re 20 μPa)

(a.1)

Creepage Creepage

0.05

0.996

0.997 0.998 Time (sec) (s) Time

0.999

1

-20 -40 -60 -80 -100 -120 1 10

2

10 Frequency (Hz) Frequency (Hz)

3

10

100 80 60 40 20

32

63

130

250

500

1000

2000

4000

Frequency (Hz)

Fig. 5. Squeal prediction for outer (a) and inner (b) wheels in a leading wheelset of a bogie in the Class 158 vehicle, running at 6 m/s on the curve with 150 m curve radius: (1) lateral wheel and rail velocities at the contact position, normalised by the rolling velocity, ⎯ wheel, − − − rail, ⋅⋅⋅⋅⋅⋅ total sliding velocity; (2) spectra of lateral wheel and rail velocities at the contact; (3) sound pressure level of wheel squeal noise at 7.5 m from the source.

method introduced by Thompson and Jones [6]. Both wheels in the leading wheelset have a sound pressure level (SPL) more than 90 dB(A) at 7.5 m from the source. The trailing wheels are in unstable vibration at the low-frequency circumferential mode, the (0,c) mode at 70 Hz. However, because the dominant vibration mode is so low, the acoustic radiation coefficient of this mode is very small and squeal does not occur at the trailing wheel/rail contacts. The squeal prediction results for a range of curve radii are listed in Table 4, where the dominant vibration modes and SPL of unstable cases are indicated. In all the curves considered, the leading wheel/rail contacts with large lateral steady-state creepage are prone to squeal. However, the occurrence of squeal in the trailing wheel/rail contacts depends on which circumferential mode is the dominant one. If the high-frequency mode (2,c) is the dominant mode, squeal can occur. The squeal intensity mainly depends on the dominant mode, but the damping condition of wheel and the falling coefficient of friction also have some influence. Normally, when squeal occurs at the high-frequency modes, e.g. (3,0), (4,0) and (2,c) modes, it is stronger than when it occurs at the lower-frequency modes, e.g. (2,0) mode. Examples are given of the effectiveness of conventional squeal control methods. Friction modification is simulated by using λ = 0.1, and the damping treatment on the wheel is modelled by increasing the damping ratio of flexible modes by a factor 50. Both friction modification and wheel damping treatments may lessen and even eliminate the squeal. It is clear that the squeal at the leading inner wheel is most difficult to control. However, the squeal control methods may also alter the mode excited which may lead to increase in noise level. The damping treatment considered corresponds to a

Squeal Prediction for a Bogied Vehicle in a Curve

319

damping ratio of 0.5% in the lightly damped modes; greater damping may be achieved in practice; values above 1.3% were found to eliminate squeal in the present example. The results of the model have been found to be very sensitive to the modal amplitude and to a lesser extent to the modal damping of wheel. Although the squeal corresponds to wheel modes, if the track is assumed to be rigid the results change considerably. Table 4. Squeal prediction on four wheel/rail contacts in a bogie of the Class 158 vehicle Curve radius

Curving speed

(m)

(m/s)

150

6

200

8

300

12

Wheel/rail contact Leading outer Leading inner Trailing outer Trailing inner Leading outer Leading inner Trailing outer Trailing inner Leading outer Leading inner Trailing outer Trailing inner

falling friction λ = 0.3

falling friction λ = 0.1

Dominant mode

SPL (dB(A))

Dominant mode

SPL (dB(A))

(3,0) (4,0) (0,c) (0,c) (3,0) (4,0) (0,c) (0,c) (4,0) axle — —

91.9 90.2 54.3 53.1 92.4 91.5 44.7 44.4 90.3 66.9 — —

— (4,0) — — — (4,0) — — — axle — —

— 86.9 — — — 78.5 — — — 67.0 — —

falling friction λ = 0.3 wheel damping ×50 Dominant SPL mode (dB(A)) (2,0) (2,0) — — (2,0) (4,0) — — — axle — —

81.2 80.0 — — 80.9 87.7 — — — 67.1 — —

4 Conclusions In this paper, a general squeal model is established by integrating wheel/rail submodels into a feedback loop based on the wheel/rail contact dynamics. The timedomain squeal prediction using the parameters of curving behaviour shows that the squeal is prone to occur at the leading wheels of a bogie, mainly due to the large lateral creepage. The trailing wheels with large longitudinal creepage may squeal if the dominant mode of self-excited vibration is at a high-frequency mode. Both the wheel damping treatments and the friction modification can be used to control the squeal, but will not always be effective.

References [1] Rudd, M.J.: Wheel/rail noise – Part II: Wheel squeal. Journal of Sound and Vibration 46, 381–394 (1976) [2] Heckl, M.A., Abrahams, I.D.: Curve squeal of train wheels, Part 1: Mathematical model for its generation. Journal of Sound and Vibration 229, 669–693 (2000) [3] de Beer, F.G., Janssens, M.H.A., Kooijman, P.P.: Squeal noise of rail-bound vehicles influenced by lateral contact position. Journal of Sound and Vibration 267, 497–507 (2003) [4] Jones, C.J.C., Thompson, D.J.: Rolling noise generated by railway wheels with viscoelastic layers. Journal of Sound and Vibration 231, 779–790 (2000) [5] Kalker, J.J.: A fast algorithm for the simplified theory of rolling contact. Vehicle System Dynamics 11, 1–13 (1982) [6] Thompson, D.J., Jones, C.J.C.: Sound radiation from a vibrating railway wheel. Journal of Sound and Vibration 253, 401–419 (2002)

IMAGINE Rail Noise Sources – A Practical Methodology M. Beuving1, B. Hemsworth2, and R.R.K. Jones3 1

DeltaRail BV, Concordiastraat 67, Postbus 8125, 3503 RC, Utrecht, The Netherlands 2 16 Whistlestop Close, Mickleover, Derby, DE3 9DA, UK 3 DeltaRail UK, Hudson House, 2 Hudson Way, Pride Park, Derby, DE24 8HS, UK Tel.: +44 870 190 1244 [email protected]

Summary European Commission Directive 2002/49/EC requires that Common Assessment methods for the modelling of noise from road, rail, aviation and industry shall be established by the Commission. The research project HARMONOISE was funded under the European 5th Framework programme in order to commence this process. The research has been continued through the 6th Framework project IMAGINE to a point where the delivered suite of procedures, algorithms and databases may be implemented as required by the Commission. This paper describes the overall IMAGINE approach in general terms. It then focuses on the development of a practical methodology for the rail-specific elements of the model. These comprise a source term database of examples and default information for individual rail vehicles, and a process for combining this data into traffic source lines that interface with the IMAGINE propagation model.

1 Introduction European Commission Directive 2002/49/EC requires that Common Assessment methods for the modelling of noise from road, rail, aviation and industry shall be established by the Commission. As a first step toward developing a common method, the EC 5th Framework project HARMONOISE was initiated in August 2001. Its main objective was to develop harmonised, accurate and reliable methods for the assessment of environmental noise from roads and railways. The HARMONOISE philosophy was to separate source and propagation. It therefore developed source models for road and rail, together with a single propagation model for these sources that included the effect of distance, air absorption, ground characteristics, barriers and meteorological variables such as wind and temperature gradient. Following validation of the propagation model this project was completed in August 2004. The IMAGINE project commenced in November 2003, under the EC 6th Framework Programme, with the aim of extending the HARMONOISE source databases for road and rail and using the HARMONOISE methodology to develop prediction methods for aircraft and industrial noise sources. This required the setting up of new source models for aviation and industry, together with such modifications to the propagation models as were necessary to account for elevated sources (aircraft), sources with large dimensions (industry) and diffraction by vertical barriers (industry). B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 327–333, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

328

M. Beuving, B. Hemsworth, and R.R.K. Jones

The overall single objective of HARMONOISE and IMAGINE was therefore to formulate a model that provides a common assessment method and that eventually can be adopted for use for strategic mapping as required by the Directive. The railway-specific activity of IMAGINE was carried out by the partners DeltaRail (UK and Netherlands), BUTE (Hungary), Kilde Akustikk (Norway), Labein (Spain), SP (Sweden) and TNO (Netherlands).

2 Overview of IMAGINE The main technical objectives of IMAGINE are to provide: • • • •

• •

•

practical guidelines for data management and information technology for noise mapping; guidelines and examples showing how and when noise measurements can add to the credibility and reliability of assessed noise levels; a reliable, and widely accepted, method for the assessment of environmental noise levels from airports, which is compatible with the calculation methods for noise propagation developed within HARMONOISE; default databases for the source description of road noise, i.e. vehicle category and road surface type, for a typical fleet of European road traffic, and provide guidelines on how to deal with situations deviating from the defaults; guidelines and examples to assist in the creation of an efficient link between road traffic flow management and noise action planning; databases for the source description of rail noise, i.e. vehicle category and track type, for a representative sample of the European rail traffic fleet, as well as a default dataset, and provide guidelines on how to deal with atypical situations; a harmonised, accepted and reliable method for the assessment of environmental noise levels from industrial sites and plants, which is compatible with the methods for calculation of noise propagation developed in HARMONOISE.

All the technical deliverables were in place by the required completion date of December 2006. Deliverables may be viewed on the IMAGINE website www.imagineproject.org.

3 The Approach to Rail Noise Sources within HARMONOISE/IMAGINE As is indicated above, the philosophy behind HARMONOISE and IMAGINE has been to develop separate source term descriptions for road, aviation, industry and rail, all of which interface with a single propagation model. As railway noise emanates from a range of sub-sources, which are potentially located at different heights and which can exhibit different characteristics and relationships with speed, power settings etc, it was considered important to disaggregate the overall acoustic source into

IMAGINE Rail Noise Sources – A Practical Methodology

329

a number of separately-defined elements. This includes the track element as well as the vehicle element. Such a disaggregation thus allows propagation behaviour to be accounted for more accurately, and also allows for cost-effective optimisation of noise control by enabling the influence of each sub-source to be understood. In the case of the identification of the relative contributions of vehicle and track, this also provides an apportionment of the responsibility for environmental noise between train operators and infrastructure owners, which is of use in optimising noise control treatments. In this way, the separation of rail sources into such sub-sources is of great benefit when formulating Action Plans under Directive 2002/49/EC. The sub sources of interest are: • • • •

The vehicle element of rolling noise (including curve squeal and brake squeal); The track element of rolling noise (including the effects of elevated support structures); Traction noise (engine exhaust, engine carcass, fans, compressors); Aerodynamic noise at the pantograph and in the bogie area.

The interface between the rail noise source model and the harmonised propagation model comprises a set of acoustically homogeneous source lines at defined heights (the railway “traffic noise model”). The defined heights are as shown in Fig. 1.

4m - traction, aero 3m - traction 2m - traction

0.5m - rolling:wheel, aero, traction 0.0m - rolling:track Fig. 1. Source heights above rail head level, chosen for the IMAGINE method

4 The Nature and Format of Data Stored in the Rail Noise Sources Database In order to provide the rail-specific source term database required under the HARMONOISE/IMAGINE philosophy, in a form that would be immediately useable by Member States, a relational structure was used. The database is currently implemented

330

M. Beuving, B. Hemsworth, and R.R.K. Jones

on an Access2000 platform. It holds both “Example” and “Default” Data. The Example Data has been acquired by measurements in the field (either already in existence elsewhere, or acquired during IMAGINE). The Default Data has been developed by reviewing theoretical and experimental datasets, and deriving from these a set of indicative spectra for typical European scenarios. The Example Data can be used where the train and track types match exactly what is to be modelled or where it is judged that they are sufficiently similar to justify its use. The Default Data provides generic information that may be used in the absence of appropriate Example Data, but with an associated reduction in accuracy. Therefore, in the absence of appropriate Example Data, the modeller has two options: either to acquire new field data or to use the supplied Default Data. Because rolling noise is a function of the combined wheel and rail roughness at their interface, it is important, for an accurate description of this phenomenon, that the combined roughness be included as a causal parameter within the model. Therefore rolling noise is modelled using the structure developed within the EC projects “METARAIL” and “STAIRRS” [1], as shown in Fig 2. Here, the track contribution and the vehicle contribution are shown as being the result of two separate transfer functions between combined roughness at their interface, when the filter effect at the contact patch has been taken into account, and their respective levels of resultant sound emission. This approach provides a significantly greater level of accuracy for the prediction of rolling noise than that available within the majority of current environmental prediction models, which simply assume that the rail roughness is at a fixed, and comparatively low, level. The core data within the database is in the form of 1/3 octave spectra, with associated metadata. For rolling noise, these spectra comprise transfer functions between

Vehicle Wheel roughness rveh

Train speed V

Vehicle Transfer Function H*veh

Sound pressure Sound pressure or orsound sound power: power: Vehicle Vehicle

Track Transfer Function H*tr

Sound pressure Sound pressure or power: orsound sound power: Track Track

rtot Rail roughness rtr Track

Contact Filter Cf Excitation

Transfer from roughness to sound pressure or sound power

Fig. 2. The mechanism of rolling noise generation applied within IMAGINE

IMAGINE Rail Noise Sources – A Practical Methodology

331

“combined effective roughness” (“effective” because the contact filter has been taken into account) and sound pressure level, as shown in Fig. 2, for a range of vehicle types and track types. Rail roughness and wheel roughness spectra are also stored, as well as contact filter information. For traction noise, the enhancement of rolling noise during braking and due to track joints, brake squeal, and curve squeal, a set of algorithms has been developed and default values for relevant parameters have been provided. Similarly, a model for aerodynamic noise is provided, again with default parameters. Where sound spectra are stored, these are in terms of sound pressure level. The database is managed with an administration tool, which includes a data viewer and an export facility. The data viewer is shown in Fig. 3.

Fig. 3. The IMAGINE rail noise source data viewer

5 Data Acquisition As delivered, it is intended that the IMAGINE rail noise sources database will be immediately useable by virtue of the 233 example spectra and the 76 default spectra that have been included within it. However, it is also intended to be fully amenable to updating whenever new example data or default data becomes available. To ensure that new example data is in an appropriate format and to an adequate quality for inclusion, data acquisition is controlled by the provision of a measurement protocol and a data input sheet with an associated import tool. The measurement protocol follows, to a great extent, the good practice indicated by ISO 3095, but with recommended enhancements to ensure that the various noise

332

M. Beuving, B. Hemsworth, and R.R.K. Jones

sub-sources can be sufficiently characterised to allow the relevant algorithms to be applied. In the case of rolling noise, the procedure becomes more complex, in order to derive the relevant transfer functions between combined effective roughness and sound pressure contributions from vehicle and track. This roughness can be obtained by direct measurement of wheel and rail roughness, or by the use of default values within the database, which can then be combined with a contact filter, either as provided or via separate modelling. Alternatively, methods for deriving combined effective roughness from track vibration behaviour during the passage of trains may be applied. The separate contributions of the vehicle and track to pass-by rolling noise need to be quantified in order to calculate the transfer functions, and this may be achieved via several alternative options. Three tools are identified, namely the TNO PassByAnalysis (PBA) approach, which requires a low response railway vehicle to enable separation to be achieved, the SNCF Multiple In Single Out (MISO) approach, and the DeltaRail Vibro-Acoustic Track noise (VTN) package. Guidelines have been provided to database users on the sound and vibration measurements that are required to implement all of these options. In essence these measurements are similar to each other in their specifications, involving sound recording at the trackside and vibration measurements on the track. It is often the case, however, that it is not possible to gain access to the track to install accelerometers, and in fact much of the existing data available to railway acousticians has been acquired by the use of a single trackside microphone. In order for such data to be included within the database, a suggested default apportionment of the sound energy from vehicle and track is provided, with the caveat that this approach leads to reduced precision.

6 Data Input The Data Input Sheet has been created as an Excel workbook and is designed to be referred to from the earliest stage of planning a measurement exercise, as its input fields provide a comprehensive list of all possible parameters and metadata that might be required. Pop-up help text guides the user through the input process. The Database Administration Tool, written in Access2000, automatically transfers example data and default data from the Input Sheet to the database, and therefore the use of the Input Sheet throughout the measurement exercise is again important for efficient operation of the system. Retrospective population of the Input Sheet is also possible, and in fact this was required to be done when existing datasets were included within the database.

7 Data Export Data is currently exported in Excel format, but it is expected that in the eventual commercial implementation of the IMAGINE method as software packages, the whole process will operate within a GIS environment.

IMAGINE Rail Noise Sources – A Practical Methodology

333

8 Data Acquired by Measurement during IMAGINE and from Pre-existing Information In parallel with the formulation of the database, measurements were carried out during IMAGINE, either dedicated to the project, or donated by partners with the permission of their customers. The data thus acquired has been included as example data within the database. It comprises separated rolling noise from the Netherlands, aerodynamic noise from the UK, separated rolling noise from Sweden, separated rolling noise and traction noise from Hungary. In addition to this data, pre-existing separated rolling noise data from France and the UK, and traction noise data from the UK, has been included as example data. As a result of this, the database contains representative example data from typical railway administrations and for a wide range of vehicle and track types in the 233 spectra provided as the initial dataset.

9 Building the Traffic Noise Model Output from the database takes the form of sound pressure level values from each sub-source at a defined reception point. For each sub-source height, this sound pressure information is used, via supplied algorithms, to generate lines of sound power, in 1/3 octave terms, which directly interface with the propagation model.

10 Conclusions IMAGINE, together with its predecessor project HARMONOISE, has produced a robust and flexible rail noise sources database, with an extensive initial example dataset, and a default dataset to be used where appropriate examples are not available. The database can be used, via defined processes, to generate acoustic source lines to represent rail traffic flows, which interface directly with the harmonised propagation model. This provides a considerably improved accuracy for the railway noise modeller, and a resource for the European Commission that is immediately useable for mapping under Directive 2002/49/EC, and for more discrete predictions. The database is designed to be updated as new data emerges over time, with a defined measurement protocol to assist in the process, and is fully described in deliverable D12/D13 of the project [2]. As a result, future railway noise mapping in Europe can be carried out efficiently using this practical methodology, in parallel with the mapping of other sources via a single propagation model, with many of the previous assumptions and inaccuracies having been eliminated.

References [1] Hemsworth, B.: STAIRRS - Final Technical Report, STR40TR181203ERRI (December 2003) [2] Jones, R., Dittrich, M., van der Stap, P., Zhang, X., Block, J.: Rail noise database and manual for implementation, IMAGINE deliverable D12/D13 (February 2007)

Optimization of a Wheel Damper for Freight Wagons Using FEM Simulation W. Behr1 and S. Cervello2 1

Deutsche Bahn AG, DB Systemtechnik, TZF12, Voelckerstr. 5, D-80939 Muenchen, Germany Tel.: +49 (0)89 1308 7344; Fax: +49 (0)89 1308 2590 [email protected] 2 LucchiniSidermeccanica, Railway Product, Lovere Plant,Via G. Paglia 45, I-24065 Lovere (BG), Italy Tel.: +39 035 963483; Fax: +39 035 963488 [email protected]

Summary Within the EU-Project SILENCE a new approach was followed to develop a wheel absorber which is suitable for freight trains with block-brakes. Lucchini has developed an absorber made exclusively out of steel consisting of various plates connected to each other while sliding is possible. In order to evaluate the effect of the absorber, FEM simulations have been performed in addition to measurements in the laboratory. The results of the simulations of a wheel with and without the absorber device are given, including the modelling of the absorber using the FEM technique. Also a comparison between calculated and measured results is presented.

1 Introduction The growing railway traffic causes increasing sound emission. Therefore Deutsche Bahn AG makes every effort to reduce the sound emission, especially of K-blockbraked freight trains, even if the requirements of operation make it difficult to develop a new wheel design regarding the acoustical needs by meeting the given structural and thermal properties. A good possibility to reduce the sound radiation is to fix absorbers to the wheel. In contrary to disc-braked wheels used in conventional rolling stock, freight trains are mostly equipped with brakes acting on the running surface. So, absorbers for blockbraked wheels have not only to withstand high temperatures up to 450°C but they have to be functional despite such temperatures. Within the EU-Project SILENCE [1] LucchiniSidermeccanica developed in cooperation with Deutsche Bahn the absorber system “Hypno” which consists only in steel components in order to withstand the thermal loads. This paper presents the working mechanism of the damping device as well as the calculations performed using the Finite Element Method (FEM) tool Ansys to estimate the acoustical effects due to this absorber. Since measurements on the first prototype of this absorbing system showed good damping and noise reduction effects, calculations performed with the finite element method concentrate on the particular case that the absorbing plates are rigidly coupled which may describe a certain kind of dysfunction due to extreme weather B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 334–340, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Optimization of a Wheel Damper for Freight Wagons Using FEM Simulation

335

conditions in winter like ice, or rusty surfaces of the absorbing plates. That condition can be handled as a worst case state of the damping system on which the investigations are focused. First the damping system “Hypno” and its function will be presented. In the next section the finite element model of the wheel with mounted damping system will be described. The major part afterwards is focused on the calculations performed with the finite element method especially for the condition of rigidly coupled absorber plates. The possibility of calculating the damping effect of the device with working absorber plates gliding against each other will be discussed in the outlook.

2 The Damping System “Hypno” (prototype 1) The main parts of the damping system are two thin plates made out of steel resting on each other without a rigid connection so that sliding is possible. Due to this sliding, vibration energy can be transformed into friction energy. The absorbing plates cover the complete rear face of the wheel. Therefore they are slitted to allow heat transfer from the back of the wheel during braking. Each absorbing plate is fixed to a mounting ring. One large ring is arranged at the inner side of the rim, a second small ring is arranged at the hub. For the arrangement of the mounting rings there are grooves milled into the wheel body in which the rings are clamped. So sliding is possible between each mounting ring and the wheel body. But due to the prestress effect of the clamping the rings can be handled as nearly rigidly fixed, especially the small ring. The single parts of the damping device and a cross section view of the wheel with mounted absorbers are shown in Fig. 1. Also shown is the computer model described in section 3. The wheel BA004 manufactured by Radsatzfabrik Ilsenburg was used for the investigations of the damper device.

Fig. 1. The damping system “Hypno”: single parts (left), a cross section view (middle) and the computer model (right)

3 Description of the Finite Element Model The finite element model shown in Fig. 1 consists out of solid elements which are sorted in five groups: the wheel BA004, the large mounting ring, the small mounting

336

W. Behr and S. Cervello

ring, the large absorbing plate and the small absorbing plate. For these elements the material parameters for steel have been used, which are a density of 7700 kg/m3 and an elastic modulus of 214 GPa. The large absorbing plate is connected to the large mounting ring, the small absorbing plate is fixed to the small mounting ring. Between the two absorbing plates there are contact-/target elements which enable the sliding of the two plates against each other [2]. The energy loss due to friction damping is modelled with additional damping elements between the damping plates (‘combin’ elements [2]). The strength of this damping of the plates can be adjusted with the damping factor Cv1. In the same way there are contact-/target elements modelled between each mounting ring and the wheel body to allow sliding. To model the prestress effect of each ring and also its participating damping effect there are damping elements between both ends of each ring. Again the strength of that damping effect due to the mounting rings can be adjusted with the damping factor Cv2. Constraints are set to the hub which is realistic for the wheel connected to an axle.

4 FEM Calculations 4.1 Modal Analysis of the Wheel BA004 In a first step a modal analysis of the wheel BA004 without the damping device was performed. Each wheel radiates sound only at specific frequencies which can be obtained performing a modal analysis. Whether a specific modal frequency will be excited depends on the location and direction of excitation, the strength of the vibration amplitude (in case of excitation) depends on the exciting force itself which is frequency dependent due to the surface conditions of the wheel and the track. The resulting mode shapes for the frequency range 900 Hz – 2200 Hz are shown in Fig. 2. 01

02

03

04

05

06

07

08

936 Hz

1201 Hz

1528 Hz

1705 Hz

1745 Hz

1894 Hz

2123 Hz

2190 Hz

Fig. 2. Modal shapes of the wheel BA004 in the frequency range between 900 - 2200 Hz (arbitrary numbering)

4.2 Harmonic Analysis and Validation of the FE Model To determine the vibration behaviour of a wheel only a few different frequency transfer functions (FRF) are sufficient. Common excitation points are shown in Fig. 3. An excitation on the running surface in radial direction (e.g. at 2X or 3X) simulates the behaviour of running on a straight track while an excitation on the rim (e. g. at 1Y) simulates the behaviour running in curves. Transfer functions like 3X-1Y or 2X-1Y can give a first hint of the sound radiation even though only one point of the front of the wheel is used for analysis.

y

1y

x

2x 6x

3x 4x

FRF modulus [dB ref 1]

Optimization of a Wheel Damper for Freight Wagons Using FEM Simulation 10,00 5,00 0,00 -5,00 -10,00 -15,00 -20,00 -25,00 -30,00 -35,00 1100

337

BA004 measured frequency response function 2X - 2X BA004 calculated frequency response function, wheel without constraints: 2X - 2X

Frequency / Hz 1200

1300

1400

1500

1600

1700

1800

1900

2000

2100

2200

Fig. 3. Location of common excitation and analysis points used to obtain a set of FRF (left). Comparison of measured (dark) and calculated (light) FRF of the wheel BA004 without constraints excited on the running surface at location 2.

A set of transfer functions has been measured in the test lab of Lucchini where the wheel was mounted on rubber blocks so that no constraints influence the vibration behaviour of the wheel. Therefore the FE model without any constraints was used to calculate that transfer functions. The comparison between measured and calculated data show good agreement for all transfer functions. For example, the transfer functions 2X-2X is shown in Fig. 3. It can be seen that the natural frequencies agree well. The maxima levels which are relevant for sound radiation differ only by a few decibel which may be caused by the individual modal damping of each mode. The structural behaviour indicates that the FE model can be considered to be validated. 4.3 Determination of the Sound Power Performing a harmonic analysis gives the vibration velocity ν~ n at each point of the wheel surface. Averaging over all points n of the wheel surface A using Eq. (1) gives the sound power with P0 = 10-12 W, A0 = 1 m2, ν0 = 5•10-8 m/s, ρ = 1.225 kg/m3 and c = 340 m/s [3]. Thereby the radiation efficiency is set to unity which is a good approximation for the wheel for frequencies above 1 kHz.

~ ρcA0 〈ν 02 〉 〈ν~ 2 〉 P A LP [ dB] = 10 log = 10 log + 10 log + 10 log n2 P0 P0 A0 〈ν 0 〉 75

1

Sound Power / dB

5

2 3

65

6

7

4

(1)

8

55 45 35 25 15 500

Wheel BA004 without damping device, excitation at 2X, front side Wheel BA004 without damping device, excitation at 1Y, front side

Frequency / Hz 1000

1500

2000

2500

3000

Fig. 4. Sound Power of the wheel BA004 for excitation on the running surface at 2X (dark) and at the rim at 1Y (light)

338

W. Behr and S. Cervello

The radiated sound power of the wheel shown in Fig. 4 was calculated for excitation locations on the running surface and the rim. It can be seen that the vibration modes shown in figure 2 are excited differently in dependence of the radial or tangential excitation direction. It should be noted that an excitation force of 1 N was used for the harmonic analysis. Therefore the absolute values of the sound power displayed in Fig. 4 do not represent the real radiated sound power of a wheel, but they do enable a quantitative comparison of different set-ups [4]. 4.4 Simulation of the Absorbing System with Rigidly Coupled Absorber Plates In order to estimate the quantitative effect of the sound radiation of the wheel with damping system for rigidly coupled absorber plates compared with the wheel without damping device, the sound power has been calculated for both options. Fig. 5 shows that sound power of the front side of the wheel when excited on the running surface (at location 2) compared with the corresponding sound power of the undamped wheel. It can be seen that there is a damping effect for roughly all modes, even for the damping device with rigidly coupled absorber plates. 75

5

2

Sound Power / dB

7

6

8

3

65 1

55

4

45 35 25

Wheel BA004 without damping device, excitation at 2X, front side Wheel BA004, damped, with rigidly connected absorbing plates, front side

15 500

1000

1500

Frequency / Hz

2000

2500

3000

Fig. 5. Sound power of the wheel body (dark) compared with the wheel with rigidly coupled absorber plates (light)

1201 Hz

1291 Hz

1291 Hz

Fig. 6. Comparison of the mode shape of the undamped wheel (left) with the mode shape of the wheel with rigidly coupled absorber plates (middle. front view, right: back view) for the mode shape numbered 2

The total level of the sound power over the frequency range 500 – 3000 Hz is 83.9 dB for the undamped wheel. For the damped wheel with rigidly coupled absorber plates the corresponding sound power level is 82.6 dB. That reduction of the sound radiation may be caused by the stiffening of the wheel. Fig. 6 demonstrates the effect of the rigidly coupled absorber plates for the mode numbered 2 as an example. As the rigidly coupled absorber plates without the possibility to slide can be stated as the

Optimization of a Wheel Damper for Freight Wagons Using FEM Simulation

339

worst case of damping, the sound radiation of the wheel with damping system can be assumed at least 1.3 dB less. Consequently, for a working absorbing system more reduction of the sound radiation can be expected. But that calculated changes in sound radiation for a constant force excitation will possibly not produce any difference when ‘rolling damping’ is included. 4.5 Sound Radiation of the Damping System Itself The calculations described in section 4.4 demonstrate the effect of the damping system on the wheel when the absorber plates are rigidly coupled. Due to reduced vibration levels on the front of the wheel, the resulting sound radiation caused by the wheel will be less. Nevertheless the absorbing plates themselves also vibrate and therefore radiate sound. Like the wheel, the absorbing plates also have certain vibration modes which will be excited when the wheel is excited. That source on the back of the wheel may contribute to the noise emission from the train due to reflection from the wagon body and the track. To get an estimation of the maximum possible amount of that sound level caused by the absorbing plates, the sound power from the back of the wheel was calculated and compared with the sound power from the front. That comparison is shown in Fig. 7 for the worst case (rigidly coupled absorber plates) using equation (1). Since the radiation efficiency is smaller than 1 for thin plates, the sound radiated by the plates may be over-estimated (for frequencies < 2 kHz in the specific case due to the combined thickness of 6 mm of the plates). Fig. 7 shows that the sound power of the absorbing plates compared to the front side is more, equal or less depending on the mode. Also there are additional mode shapes of the absorbing plates which count only for the sound radiated backwards. The highest sound power backwards results at the mode numbered 2 which splits into the two frequencies 1292 Hz and 1417 Hz. The sum level of the sound power radiated from the front gives 82.6 dB and for the back side 87.2 dB (for the wheel with rigidly coupled absorbing plates). Therefore the sound radiation backwards due to the rigidly coupled absorber plates is nearly 5 dB more compared to the sound radiation to the front. Due to absorption and reflection from the wagon body and the track a reduction of that radiated sound can be exspected and so this effect can be assumed as not relevant within the calculated values. 75

Sound Power / dB

65 55 45 35 25 15 500

Wheel BA004, damped, with rigidly connected absorbing plates, back side (damping plates) Wheel BA004, damped, with rigidly connected absorbing plates, front side

1000

1500

Frequency / Hz

2000

2500

3000

Fig. 7. Sound power of the front (light) of the wheel with rigidly coupled absorber plates compared with the sound power of the back which are the absorber plates (dark)

340

W. Behr and S. Cervello

5 Outlook Simulations demonstrated in sections 4.4 and 4.5 have been focussed on calculations of the wheel with the damping device with rigidly coupled absorber plates in order to investigate the worst case of dysfunction of the absorber system. As described in section 3 the FE model includes damping elements between the absorber plates. First calculations with that model with a working absorber device show a damping effect and will allow an optimization of the specific design of the different parts of the device.

6 Conclusion The paper presents calculations of the effect of the damping device “Hypno prototype 1” developed by LucchiniSidermeccanica in cooperation with Deutsche Bahn AG when fixed at the wheel BA004. The finite element model of the wheel BA004 without the damping system could be validated due to a comparison with measurements performed. The investigations described focus on a dysfunction of the absorbing device caused by rigidly coupled absorbing plates in order to assume the worst case of absorber action. A sound radiating effect of the damping plates on the back of the wheel could be calculated which was slightly higher than the sound radiation from the front of the wheel. That fact seems not significant in the present preliminary state due to reflection and absorption at the back region of the wheel but may lead to further adjustment when optimizing the device. Running a test train to prove the sound reduction effect on the wheel body and to check the effect of the damping systems on the sound radiation is scheduled for autumn 2007 and will hopefully demonstrate a good effect of the damping system described. The results of these intended measurements will give the opportunity to enhance the present prototype of the damping device using further continuative FE simulations.

Acknowledgements The authors like to thank the project leader Bernd Asmussen for all the fruitful discussions. This work was financially supported by the European Union within the Integrated Project SILENCE.

References [1] Project homepage, http://www.silence-ip.org [2] Ansys Release 8.1 Documentation / element reference / element library, Ansys Inc., Canonsburg, USA (2004) [3] Beranek, L.L.: Noise and Vibration Control. McGraw-Hill, New York (1971) [4] Behr, W., et al.: Calculation of the sound radiation of a railway bridge. In: 5th International Conference on Computation of Shell & Spatial Structures, Salzburg (2005)

Types of Rail Roughness and the Selection of Vibration Isolation Measures H.E.M. Hunt Cambridge University Engineering Department, Trumpington Street, Cambridge, CB2 1PZ, UK Tel.: +44 1223 332600 [email protected]

Summary A procedure is outlined for quantifying the significance of various types of rail roughness for the purposes of predicting and controlling low-frequency ground vibration near railway lines. The effectiveness of resilient elements inserted beneath the rail is investigated. It is found that roughness mechanisms can be classified according to the effectiveness of added resilience. Vibration from a track with “Class A” roughness will be well controlled by resilient rail support; a track with “Class B” roughness cannot be controlled; and vibration from a track with “Class C” roughness will be increased. Examples of “Class A” roughness are those due to variations in trackbed roughness and in rail-support stiffness and these can be controlled through the insertion of additional under-rail resilience. The reduction in roughness level can easily result in a 10dB reduction in perceived roughness down to 5Hz for an urban metro system. This reduction is additional to any reduction due to the mass-on-spring effect of the unsprung mass on its resilient track. This not only explains the good insertion performance often measured of under-rail countermeasures, but also why there is often far less amplification at the track-system resonance than might otherwise be expected. In some cases the effect of adding resilience has less effect than in others. For example there is no beneficial effect found for the case of roughness due to a rail that is naturally bent (a condition built-in after leaving the rail-straightener in the rolling mill).

1 Introduction There are conventionally two sources of roughness considered when calculating ground vibration from railways [1]. The first is rail-surface roughness and in principle this can be measured by passing a profilometer of some kind along the running surface of the rail. The second source is wheel roughness and this too can be measured, usually more easily than can be the rail roughness. Rail roughness measurements are generally made to enable calculation of acoustic noise and when it comes to predicting ground vibration it is tempting to use the same measurement of rail roughness. For a given train speed and for a given sleeper spacing this sleeper-pass frequency is fixed and the vibration generated is tonal. For the train speed of 15ms-1 given above, and for a sleeper spacing of 600mm the sleeper-pass frequency is 25Hz. Another form of parametric excitation is the variation along its length of the rail flexural rigidity EI (Young’s modulus E and second-moment-of-area I) and one might imagine that a B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 341–347, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

342

H.E.M. Hunt

worn rail, or one that has been manufactured to poor tolerances may generate roughness of this kind as the train passes over it. Two additional effects that generate longwavelength roughness are firstly that the sleepers themselves may be supported on an uneven trackbed and secondly that the rail may have been manufactured or welded with some initial curvature. In summary, there are seven ways in which random roughness can be generated: 1. trackbed roughness; 2. bent rail; 3. variation of rail support stiffness; 4. variation of rail EI; 5. variation of sleeper spacing; 6. surface irregularity of the railhead; 7. irregularity of the wheel. This paper takes the unusual step of regarding the last two (items 6 and 7) as unimportant for the generation of low-frequency vibration as these can be controlled through suitable track and vehicle maintenance regimes. The first five effects are much more significant at low frequencies and can be quantified using random-process theory. First it is necessary to establish some of the basic properties for the deflection of a rail under the action of a loaded wheel.

2 Random Variations and Their Effect on Roughness In this section the first five sources of roughness are described mathematically. The objective is for each to produce a transfer function of input disturbance to output perceived roughness. The disturbances are in turn: 1. trackbed roughness; 2. bent rail; 3. variation of rail support stiffness; 4. variation of rail EI; 5. variation of sleeper spacing. 2.1 Trackbed Roughness Consider a uniform beam on a uniform Winkler foundation [2] but let the trackbed be described by its profile yf(x) as shown in Fig. 1. The springs k each see an extension y(x) – yf(x) so that the governing differential equation under the action of a point load P at position x = xP is

EI

d4y + k ( y − y f ) = Pδ ( x − xP ) dx 4

(1)

where δ(x) is the Dirac delta function. This can be rearranged as

d4y P +α4y = δ ( x − xP ) + α 4 y f 4 dx EI

where

α4=k/EI

(2)

Taking the Fourier transform of both sides using wavenumber γ, noting ∞

∫δ (x −

−∞

gives

xP )e − jγx dx = e − jγx P

P α4 − jγx P Y (γ ) = e Y (γ ) + 4 EI (γ 4 + α 4 ) γ +α4 f

(3)

The deflection under the load is found by putting x=xP into the reverse transform

y P ( xP ) =

1 P 2π EI

∞

1 1 dγ + 4 4 −∞ γ + α 2π ∫

α4 Y (γ )e jγx dγ 4 4 f −∞ γ + α ∞

∫

P

Types of Rail Roughness and the Selection of Vibration Isolation Measures

y

343

P EI

x k

yf

Fig. 1. Beam on a Winkler foundation with an uneven trackbed

1 π 2 dγ = means that the expression 4 −∞ γ + α 2α 3 P 1 P ∞ 1 dγ is equal to y Po = being the deflection under the ∫ 4 4 8EIβ 3 2π EI − ∞ γ + α ∞

The result for the integral

∫

4

load for the case of an uniform trackbed. The deflection under the load from equation (3) can now be written as

1 yP ( xP ) = yPo + 2π

α4 Y (γ )e jγx dγ ∫ 4 4 f −∞ γ + α ∞

P

(4)

Subtracting the constant “regular” deflection y Po gives the perceived roughness ~ yP ( xP ) = yP ( xP ) − y Po under the load moving on the uneven trackbed. So we have

1 ~ y P ( xP ) = 2π

α4 Y (γ )e jγx dγ ∫ 4 4 f −∞ γ + α ∞

P

(5)

The Fourier transform with γP corresponding to the load position xP is

Y (γ ) =

α4 Y (γ ) γ 4 +α4 f

(6)

So roughness perceived under the load is found given roughness of the tunnel floor. The transfer function of input trackbed roughness to output perceived roughness is

T1 (γ ) =

α4 γ 4 +α4

(7)

The transfer function T1 is used to evaluate insertion gain. 2.2 Bent Rail For a bent rail (Fig. 2) the mathematics is similar and it is easy to obtain the transfer function which transforms the input bent-rail roughness to the output perceived roughness is T2(γ).

344

H.E.M. Hunt

γ4 T2 (γ ) = 4 γ +α4 y

.

(8)

P EI

x k

Fig. 2. Beam on a Winkler foundation with an initially-bent rail

2.3 Variation of Rail-Support Stiffness The input roughness above appeared as a term on the right-hand side of the differential equations. The case of variation in rail support stiffness is more complicated as it is an instance of parametric excitation. The uniform beam on a Winkler foundation shown in Fig. 3 has foundation stiffness k(x) is a function of the position x along the beam. The differential equation (1) becomes

EI

d4y + k ( x) y = Pδ ( x − xP ) dx 4

(9)

where, as before, the load is applied at position x=xP. Now suppose that the foundation stiffness can be expressed as the sum of a constant mean stiffness k0 and a small variation εk1(x), i.e.

k ( x) = k 0 + εk1 ( x) = k 0 (1 + εwk ( x))

(10)

where ε is a small constant and where wk(x) represents non-dimensionally the variation of foundation stiffness k about its mean value k0. This means that (9) can be rearranged as

d4y P 4 + α 0 (1 + εwk ( x)) y = δ ( x − xP ) 4 EI dx y

where

α04 =

k EI

(11)

P EI

x k(x)

Fig. 3. Beam on a Winkler foundation with variable spring stiffness

Types of Rail Roughness and the Selection of Vibration Isolation Measures

345

It should be noted at this point that the product of wk(x) and y(x) on the left-hand side of (11) prohibits a simple solution by taking the Fourier transform as before. Instead we use a perturbation analysis, but space does not permit its complete exposition here. The result of some substantial mathematics is that the Fourier transform of the deflection under the load is.

Y1P (γ P ) = −

1 P π 2α 0 (γ P + 12α 0 ) Wk (γ P ) 2 2 4 4 2π EI (γ P + 2α 0 )(γ P + 16α 0 ) 2

2

(12)

and the transfer function T3(γ) is therefore

T3 (γ ) =

2 P α (γ 2 + 12α 2 ) 2 EI (γ 2 + 2α 2 )(γ 4 + 16α 4 )

(13)

T3(γ) is significantly different from T1(γ), important for computing insertion properties in Section 4. 2.4 Variation of Rail EI The procedure for dealing with random variation of EI along the rail, also parametric, is analogous to that used in Section 3.3 so that the transfer function T4(γ) is therefore

T4 ( γ ) =

2 P 1 γ 6 + 2γ 4α 2 + 14γ 2α 4 + 8α 6 4 EI α 3 ( γ 2 + 2α 2 )( γ 4 + 16α 4 )

(14)

2.5 Variation of Sleeper Spacing Variation of sleeper spacing has two effects. The first is to cause a spatial variation in the Winkler foundation stiffness and this effect can in principle be evaluated by using the results above. But a second effect is that parametric excitation at the sleeper-pass frequency ceases being tonal (at 25Hz and its harmonics for the example given in Section 1) and it becomes a broad-band process. A small amount of random variation in sleeper spacing will not affect the tonality significantly, but larger amounts of variation will cause the perceived roughness to become increasingly broad band. This effect is difficult to evaluate analytically and so the effect of sleeper spacing needs to be evaluated numerically. The transfer function T5 obtained from this numerical calculation is also shown in Fig. 4 for comparison, normalized by the static deflection under the load to produce a transfer function which is dimensionless. So longwavelength roughness is very likely to be due either to track-bed roughness, variations in rail-support stiffness, random sleeper spacing, or any combination of the three. Roughness at short wavelengths can be due to bent-rail effects.

3 Calculation of Insertion Gain The purpose of this paper is to evaluate the insertion-gain performance of resilient rail-support systems. The method used to evaluate the insertion gain performance is to suppose that rail roughness derives from one of the five mechanisms described above

346

H.E.M. Hunt

(equations 7, 8 13 and 14, with T5(γ) obtained numerically. The effect of adding additional foundation springs ka in series with the original foundation springs k is to produce a new foundation spring ki for the insertion condition. ki is found using the standard result for springs in series so the insertion stiffness ratio R is defined as

ki α i = = k α4 4

R=

1 k 1+ ka

=

ka k + ka

(15)

where it can be seen that R<1 for all insertions. For cases 1, 2 and 4 where the input (trackbed roughness, bent rail and EI variation) is the same before and after insertion, the Insertion Gain (IG) can be computed easily as IG = T(γ)after / T(γ)before

(16)

But the other cases are a bit more difficult. The five insertion gain results are tabulated in Table 1 (IGcase 5 is obtained numerically) with the limits evaluated for γ = 0 and γ→∞. They are also plotted in Fig 4 for the case of R=0.2 and α4 = 1.35 (so αi4 = 0.2×1.35) for a rail EI=10MNm2 supported at 600mm intervals with stiffness 20kN/mm at each baseplate. Insertion Gain (dB) for different categories of roughness 20

dB

10

0

-10

-20 T1 T2 T3 T4 T5

-30

-40

0

5

rough trackbed bent rail random stiffness random EI random sleeper-spacing 10

15

20 25 30 frequency (Hz)

35

40

45

50

Fig. 4. Insertion Gain performance for resilient rail support for a train speed of 15ms-1 (34mph) and an insertion stiffness ratio R=0.2 (i.e. track support stiffness is reduced to 20% of its original value)

It is clear both from Table 1 and from Figure 5 that since both IGcase 1 and IGcase 3 are around -14dB to -18dB in the frequency range of interest there is potential benefit to be gained by introducing resilience for the cases of trackbed roughness (T1) and random stiffness (T3). This is provided, of course, that one or other of these mechanisms are

Types of Rail Roughness and the Selection of Vibration Isolation Measures

347

responsible for the original roughness (which is likely). If, however, roughness is dominated by bent-rail effects (T2) then there is no additional track-smoothing benefit to be had by introducing resilience. This is because IGcase 2 is unity (0dB) for all frequencies of interest. Similarly, if roughness is dominated by random sleeper spacing effects (T5) for which IGcase 5 is around -4dB for all frequencies of interest then there is very little additional track-smoothing benefit to be had by introducing resilience. Table 1. Insertion Gain performance of resilient track support. Values in parentheses are evaluated for R = 0.2 for comparison with the asymptotic values in Figure 7. IG(γ) Case 1 CLASS A γ 4 +α 4 Trackbed roughness R 4 γ + Rα 4 Case 2 CLASS B γ 4 +α 4 Bent rail γ 4 + Rα 4 Case 3 CLASS A (γ 2 + 12 Rα 2 )(γ 2 + 2α 2 )(γ 4 + 16α 4 ) R5/ 4 Random stiffness (γ 2 + 12α 2 )(γ 2 + 2 Rα 2 )(γ 4 + 16 Rα 4 ) Case 4 CLASS C (γ 2 + 2α 2 )(γ 4 + 16α 4 )(γ 6 + 2γ 4 Rα 2 + 14γ 2 Rα 4 + 8 R Rα 6 ) R −3 / 4 Random variation (γ 2 + 2 Rα 2 )(γ 4 + 16 Rα 4 )(γ 6 + 2γ 4 α 2 + 14γ 2 α 4 + 8α 6 ) of EI Case 5 CLASS B Random sleeper spacing

Not available analytically

IG(γ=0)

IG(γ→∞)

1

R

(0dB)

(-14dB)

1/R

1

(+14dB)

(0dB)

R1/4

R5/4

(-3.4dB)

(-17.5dB)

R-3/4

R-3/4

(+10.5dB) (+10.5dB) (+10dB)

(-4dB)

In addition to the smoothing effect described above (the main purpose of this paper) there is the conventional benefit due to standard mass-on-spring vibrationisolation theory. The combined effect of smoothing and of mass-on-spring effects is simply additive (in dB).

4 Conclusions The principal conclusion is best seen in Table 1. The reduction in perceived roughness can be as much as 15dB even at very low frequencies (below 5Hz). But this benefit is only observed if the principal source of roughness is unevenness of the tunnel floor, or random variation in the support stiffness (i.e. effects below the rail). These are CLASS A roughness sources. If the roughness is caused by bent rail then there is no additional benefit gained. This is CLASS B. Random variation of EI is CLASS C and resilience beneath the rail increases perceived rail roughness. It is of the utmost importance to know which class of rail roughness is present in order to make accurate prediction of the effectiveness of under-rail resilience. It is not at all clear how this can be known since all conventional measures of rail roughness cannot distinguish between the various classes.

References [1] Remington, P.J.: Wheel/rail noise – Part I: Characterization of the wheel/rail dynamic system. Journal of Sound and Vibration 46, 359–379 (1976) [2] Timoshenko, S., Goodier, J.: Theory of Elasticity. McGraw-Hill, New York (1959)

Rail Roughness Monitoring in the Netherlands A.H.W.M. Kuijpers M+P – consulting engineers, P.O. Box 2094, 5260 CB, Vught, The Netherlands Tel.: +31 73 6589050; Fax: +31 73 6589051 [email protected]

Summary Rail grinding is not a time-invariant noise mitigation measure. By reducing the rail roughness, the rolling noise will decrease, but since the rail roughness is not invariant, neither is the noise reduction that is achieved. In general, the rail roughness will again increase after grinding and consequently the rolling noise emission will increase. After a certain amount of time, the rail needs to be ground again to achieve a certain desired average noise reduction. Regular monitoring of the rail roughness condition of the rail is necessary to keep track of the rail roughness and rolling noise emission changes over time. The rail roughness can measured with direct and indirect measurement methods. In the Netherlands, we have developed a monitoring method based on a combination of direct and indirect measurements. The direct measurements are done with commercially available instruments. For the indirect measurements we have developed a new system called ARRoW. In addition, a software program was developed that automatically combines the results from the direct and indirect measurements and delivers the roughness condition of a complete track. The new monitoring method has been successfully used on the new high speed line (HSL-Zuid) that connects Amsterdam to Brussels and Paris and for conventional rail in a recent research project in the Dutch IPG (innovation) program. The technical details and some of the practical issues involved in the application of these methods will be discussed.

1 Introduction Rail grinding is a noise mitigation measure that reduces the rolling noise of trains at the source. By reducing the rail roughness, the combined (wheel/rail) roughness is reduced. This decreases the excitation of rail and wheel vibrations, which leads to a lower sound radiation into the surroundings. In the Netherlands, rail grinding for acoustic purposes has long remained in an experimental phase. This situation remained despite the fact that noise reduction by rail grinding is well-understood from theory and utilizable in practice. A problem is that the measure is only effective when the rolling stock using the track has relatively low wheel roughness. In the Netherlands in general, this is not the case because a large proportion of the rolling stock is (still) equipped with cast iron block brakes. Therefore, rail grinding was never put into practice on a large scale. Recent developments in the Netherlands have renewed the interest for rail grinding. Firstly, a new high speed line has been constructed between Amsterdam and B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 348–354, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Rail Roughness Monitoring in the Netherlands

349

Belgian border, which connects to the European high speed network. On this line, rail grinding is viable and is indeed applied because this line is only used by modern (disc-braked) rolling stock with relatively low wheel roughness. Secondly, there is increasing attention to retrofit existing rolling stock having cast-iron block brakes with new block brake types (e.g. K, LL-blocks). This will eventually reduce the average wheel roughness on the existing conventional rolling stock. This will increase the possible effectiveness of rail grinding on the Dutch network. But rail grinding itself is only half of the story that comes with rail grinding as a noise mitigation measure. We need measurement and monitoring methods to be able to put the measure in operation on a large scale. These methods are necessary to i) assess the actual noise reduction that is achieved due to rail grinding at a certain track, and ii) to monitor the development of the rail roughness and hence the noise reduction over a period of time. The results of the monitoring are used to schedule grinding maintenance to achieve a certain average noise reduction over a period of time. In this article we will discuss the methods to measure and monitor the rail roughness, in the Netherlands and we will present the data processing framework to process the results of the monitoring measurements into a noise reduction that fits one-to-one with the current Dutch legislative framework and also with the future European noise impact calculation models [1]. As an illustration of the rail roughness monitoring methodology, we will use results from the monitoring program on the Dutch high-speed line (called “HSL-Zuid”). A general overview of the monitoring program can be found in [2]. We are currently investigating if the same methodology can be applied for conventional rail. This work is done in the framework of the Dutch IPG program.

2 Direct and Indirect Rail Roughness Measurement Principles The monitoring method used on the HSL-Zuid and proposed for Dutch conventional rail consists of a combination of direct and indirect roughness measurements. This combination of measurement methods is a compromise between accuracy of the measurements and practical considerations. In principle, direct measurement of the roughness profile would be sufficient for monitoring purposes because the direct measurement method delivers data that can be directly interpreted as a noise reduction. However, application of the direct method is not practicable for the HSL-Zuid (approx. 2 x 90 km track length) because many measurements are needed to obtain a representative roughness for the whole track. Moreover, these measurements need to be done while the track is out of service and the procedure is rather labour intensive. To overcome the impracticality of the direct method for large track lengths, indirect roughness measurements can be done. In Germany this method is used to monitor the specially monitored track with the Sound Monitoring Coach). The indirect measurements consist of rolling noise measurements in the vicinity of the wheel-rail contact area. The idea behind this method is the fact that, for a given (low) wheel roughness and track system, there is a direct relationship between the rail roughness change and the change in of rolling noise. Thus, by measuring the noise variation, we know the roughness variation. However, due to several practical constraints, we believe that the indirect method is not as accurate absolutely as the direct method and it

350

A.H.W.M. Kuijpers

delivers noise level variations which cannot be directly translated into absolute changes of rolling noise emission level. To overcome the impracticality of the direct method and the inaccuracy and relative nature of the indirect measurements, we combine the results of both methods to yield an accurate description of the rail roughness over large track lengths.

3 Combining Direct and Indirect Rail Roughness Measurements The concept behind the combination of direct and indirect measurement method is that the indirect measurement results which are relative by nature can be made absolute. This is done by “calibrating” the indirect results on reference sections with additional direct roughness measurements. Or looking from the perspective of direct measurements, the direct measurement results on the reference sections are “smeared out” over the whole track by means of the indirect measurements results. In the calculation methods for noise impact studies (e.g. [1][3]), there is a direct spectral relationship between combined wheel and rail roughness on the one hand, and noise emission change on the other hand:

ΔL p ,grinding-average,i = ( Lr ,track,grinding,i ⊕ Lr ,vehicle,i ) − ( Lr ,track,average,i ⊕ Lr ,vehicle,i ), (1) with ΔL p ,grinding-average as the noise emission change between ground and average track, Lr ,vehicle as the wheel roughness level, Lr ,track as the rail roughness level, ⊕ denoting energetic summation, and i denoting a certain frequency band. Since roughness is usually known as a function of wavelength λ and noise level difference is expressed as a function of frequency f , a spectral transformation f i = v / λi is made, which depends on vehicle speed v . With equation (1), the direct measurement results, expressed as

Lr ,track,grinding,i

can be translated into a change of the rolling noise level, given an assumed wheel and average rail network roughness. In the Netherlands, these assumptions are taken from the national noise impact calculation method. With the indirect measurement method, we directly measure the rolling noise level, but not the noise level difference between ground and average track. However, since the variation of rolling noise results from rail roughness variation, we can still use equation (1) to express this variation. Assuming we know the noise spectrum difference between sections A and B of the same track, we can express this difference mathematically as

⎡⎣ ΔL p,A-B,i ⎤⎦ = ( Lr ,track,section A,i ⊕ Lr ,vehicle,i ) − ( Lr ,track,section B,i ⊕ Lr ,vehicle,i ) . indirect (2) We can then substitute this expression in equation (2) to yield

ΔL p ,section A-average,i = ⎡⎣ L p ,A,i − Lp ,B,i ⎤⎦ + ⎡⎣ ΔL p ,section B-average,i ⎤⎦ . indirect direct

(3)

Rail Roughness Monitoring in the Netherlands

351

This equation gives a direct relationship between measured quantities and the noise level change on section A compared to average rail roughness. In practice, we use do not use just one but several reference sections to scale measured noise level difference spectra. This means that, instead of eqn. (3), we use a transformation based on the least square fit between the indirectly measured noise level differences and the noise level differences computed with roughness spectra obtained with the direct measurements at the reference sections. Mathematically this can be expressed as:

ΔL p ,section A-average,i = ⎡⎣ L p ,A,i ⎤⎦ − H reference,i , indirect

(4)

with H reference,i as the average transformation function between computed grinding effect and measured noise level. If we have M reference sections (indicated with index m) then the average transformation function is defined as

H reference,i =

1 M

∑ ( ⎡⎣ Lp,reference m,i ⎤⎦indirect − ⎡⎣ΔLp,reference m-average,i ⎤⎦ direct ) . M

m =1

(5)

4 Measurement and Analysis Systems 4.1 Direct Roughness Measurement Systems Direct roughness measurements have been done with the Müller-BBM 1200e, and its successor, and the ØDS TRM02 measurement devices. All measurements have been analysed according to the prEN 15610 standard [4]. Both systems are able to deliver results accurate enough to determine the noise reduction by grinding: although the systems deliver a different roughness spectrum, we found that the resulting Aweighted noise reduction is comparable within 0.2 dB. However, the systems are very different with respect to robustness, ergonomics and user-friendliness. A full comparison is beyond the scope of this article. 4.2 ARRoW Measurement System The data acquisition part of this framework is called the ARRoW system. The system measures rolling noise, position and speed onboard a measurement vehicle. The system consists of 4 removable microphones (see Fig. 1) combined with a GPS receiver. The microphones are placed close to the wheel/rail interface to measure directly the rolling noise avoiding interfering reflections. The GPS receiver is used for position and speed information. A data acquisition system (Müller-BBM-VAS PAK mk II) is used to simultaneously register the (spectral) noise information and speed and position information. The system is completely self-supporting with respect to power. For the ARRoW system we have to measure the sound at all four wheels of the measurement bogie. This quadruples the requirements for the data acquisition system at the benefit of i) being able to distinguish between left and right side rail of the track, and ii) introducing redundancy in the measurement chain which increases the robustness of the whole system.

352

A.H.W.M. Kuijpers

Fig. 1. ARRoW system on Thalys bogie (left) and on BAM rail measurement vehicle (right)

4.3 Data Processing and Analysis System For the purpose of automatically processing and analyzing the measurement data, we have developed a dedicated software program that automatically couples the noise spectra, speed, and position information to the roughness spectra obtained with the direct measurements. As an end result, the program produces a complete overview of the roughness condition of the track as a function of the chainage. The analysis steps that are implemented in the software for the coupling of indirect and direct measurement results are described next. 1.

2. 3.

4.

5. 6.

Noise spectra for all four microphones are available as a function of time with a sampling interval of 0.05 s. This sampling interval results in a spatial resolution of about 2 m when measuring at a maximum speed of 160 km/h. Position and speed information is available as a function of time with a sampling interval of 1 s. Noise spectra and position/speed information are coupled to obtain the noise spectra as a function of chainage and speed alongside the track. For this step, the software requires a translation table from geographical coordinates to track chainage. To take speed variation during the noise measurement into account, the noise spectra are scaled to a nominal measurement speed (vmeas) with a (frequency-dependent) logarithmic scaling according to the Dutch noise impact calculation model. The noise spectra are averaged over a evaluation length (e.g. 20 m) to remove transient effects. The noise level reductions

⎡⎣ ΔL p ,reference-average,i ⎤⎦ are computed for all direct

reference sections, based on the average roughness-wavelength spectra for the reference sections. The wavelength to frequency transformation is based on the nominal measurement speed vmeas. 7. 8.

Compute the average transformation functions H reference,i using eqn. (5) for each microphone. Translate the measured noise spectra for each microphone to noise reductions due to grinding applying eqn. (4). This gives the noise reduction spectrum due to grinding averaged over each small evaluation length interval.

Rail Roughness Monitoring in the Netherlands

353

9.

If the nominal measurement speed is different from the target track speed, a frequency shift is applied to obtain the noise reduction spectrum at the target speed. This is in general required for high-speed lines, where we cannot measure at the target track speed. 10. A further averaging over left and right side microphones and over a certain chainage section length may be required as a final step. For the HSL-Zuid line e.g. the noise reduction effect due to grinding is evaluated over 1 km sections. This evaluation length and averaging process has been agreed upon with the environmental authorities. A typical end result of this procedure is given in Fig. 2. For this particular case, the noise reduction is in compliance with the environmental requirements if the noise immission coefficient Cb,c is below zero, which is the case over the whole track length in this example.

Fig. 2. A-weighted noise immission coefficient Cb,c due to rail grinding measured at the HSLZuid track in the Netherlands, taken from [5]. The green line represents the small section average noise reduction, the black line represents the noise immission coefficient averaged over 1 km sections. The red circles represent the noise reductions computed from the direct rail roughness measurements.

5 Practical Experiences At the moment we have applied the monitoring procedure described above a number of times on the HSL-Zuid line with different ARROW measurement setups (both with

354

A.H.W.M. Kuijpers

Thalys and designated measurement vehicle measurements). During these measurements and the subsequent analyses we encountered a number of issues that need to be addressed to improve the applicability of the monitoring method for other (conventional) tracks: •

•

•

At high speed lines, the nominal measurement speed is different from the target track speed. This means that the wheels of the measurement vehicle might not be running on the running band that the high speed rolling stock uses. This issue arises especially in narrow high speed curves. We plan to use cameras to check if the wheel rolls over the correct running band. Another issue arises when we measure at a speed different from the target speed. Then we need a frequency shift for the noise reduction spectrum based on the quotient of the measurement and target speed. However, if the measurement speed is much lower than the target speed (which occurs in practice on highspeed lines) then the low-frequency part of the measured noise reduction spectrum determines the high speed noise reduction. This part can be distorted by interference from noise sources other than rolling noise, such as aerodynamic or traction noise which normally lie outside the frequency region of interest. This means that the measurement speed should not differ much from the target speed or that interferences need to be minimized. The rolling noise might change due to factors other than rail roughness, for instance change of superstructure type, reflections from platform sidewalls etc. If these are encountered in our measurements, we need to correct the measured rolling noise first before they are used in the monitoring analysis. We are still studying on methods how to do this.

Despite these issues, we believe that the current methodology is a sound basis to further develop a monitoring method to assess the noise reduction due to rail grinding.

Acknowledgements This work is a result from our work for the Infraspeed consortium on the Dutch high speed line “HSL-Zuid” and our work in the “monitoring acoustic grinding” project in the framework the Dutch IPG program (“Innovation Program Noise”). We gratefully acknowledge the contribution of our partners in these projects, especially Infraspeed and BAM rail.

References [1] Imagine WP 6 consortium, Rail noise database and manual for implementation, Image deliverable D12/D13, doc. id. IMA6TR-061015-AEATUK10 (February 2007) [2] Kuijpers, A.H.W.M., Bekooy, M.E., Schaffner, J.C.: For noise reduction and preventative maintenance - Innovative grinding programme for Dutch high-speed rail line. Railway Gazette International, 345–348 (June 2006) [3] Dutch ministry of VROM, Reken- en Meetvoorschrift Geluidhinder 2006 – Bijlage 4 (December 2006) [4] ISO, prEN 15610 Railway applications - Noise emission - Rail roughness measurement related to rolling noise generation (2006) [5] Kuijpers, A.H.W.M.: Roughness monitoring for the HSL-Zuid with the ARRoW system: South section - April 2006 (initial state), report M+P.ISP.06.01.5 (July 2006)

Rail Roughness Level Assessment Based on High-Frequency Wheel–Rail Contact Force Measurements Jens C.O. Nielsen CHARMEC/Department of Applied Mechanics, Chalmers University of Technology, SE-412 96 Gothenburg, Sweden Tel.: +46 31 772 1500; Fax: +46 31 772 3827 [email protected]

Summary A strategy for the assessment of rail surface quality in terms of a limit value for the mean rail roughness level in the wavelength interval 3 – 8 cm is outlined. The assessment may be used for planning of rail grinding intervals if used together with a system for regular monitoring of rail roughness levels. One such system could be based on instrumented wheelset technology that measures vertical wheel–rail contact forces on a disc-braked passenger wheel in the frequency range up to about 2 kHz. The root mean square value of the vertical wheel–rail contact force after band-pass filtering with cut-off wavelengths 3 cm and 8 cm is shown to be an efficient indicator for detection of track sections with short-pitch rail corrugation.

1 Introduction Small amplitude undulations (irregularities, roughness, waviness) with wavelengths in the order of 1 – 10 cm on the running surfaces of wheels and rails induce highfrequency vertical wheel–rail contact forces. Consequences of such broad-band excitation are vibrations and rolling noise. In severe cases, it may lead to further degradation of wheels and rails in the form of sub-surface initiated rolling contact fatigue (RCF) [1, 2]. Vertical wheel–rail contact forces have been measured on a high-speed (200 km/h) X2 train operating on the line Stockholm – Gothenburg. The contact forces were determined using a trailer wheelset instrumented with strain gauges on the wheel discs [3]. It was observed that significant contributions to the contact forces occurred in the frequency range 500 – 1350 Hz [4]. Track sections leading to high force magnitudes were identified. Based on subsequent measurements with the Corrugation Analysis Trolley (CAT), it was confirmed that the rails at these sections were corrugated with dominating wavelengths in the interval 3 – 8 cm and corresponding roughness levels in the order of 20 dB (re 1 μm). Tread braking with cast-iron brake blocks are known to generate wheel corrugation with similar wavelengths and amplitudes [5]. The objective of the present paper is to outline a strategy for the assessment of rail quality in terms of a limit mean roughness level for short-pitch rail irregularities in the interval 3 – 8 cm. The criterion may be used for planning of rail grinding intervals if used together with a system for regular monitoring of roughness levels. One such B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 355 –362, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

356

J.C.O. Nielsen

efficient system could be an instrumented trailer wheelset on an X2 train that regularly is operating on the main lines of the railway network.

2 Field Tests – Rail and Wheel Roughness In connection with the X2 test campaign, several sites along the railway line generating exceptionally high vertical wheel–rail contact forces were identified. In general, the origin to increased magnitudes of dynamic contact forces is vertical rail irregularities, such as short-pitch corrugation, welds, crossings or insulating joints. To determine the magnitudes of the irregularities at these locations, the accelerometer-based system Corrugation Analysis Trolley (CAT) [6] was used. A survey on roughness measurements is given in [7]. Roughness level Lr is defined by

Lkr where rref = 1 μm and

⎛ ~ r = 20 log10 ⎜⎜ k ⎝ rref

⎞ ⎟⎟ [dB re 1 μm], ⎠

(1)

~ rk [m] is the root-mean square value of the roughness profile

30

30

20

20

Roughness Level [dB re 1 mum]

Roughness Level [dB re 1 mum]

r(x) evaluated in one-third octave band k with centre wavelength λk . Rail roughness level spectra from three sites, Vretstorp, Södertälje and Töreboda, are shown in Fig. 1. It is observed that the spectra have a common local maximum in the wavelength interval 3 – 8 cm with levels in the order of 20 dB re 1 μm indicating severe short-pitch corrugation. The energy mean of the three spectra was calculated and it is

10

0

-10

Södertälje 1993 Vretstorp 1 1993 Töreboda 1994 Corrugated rail ISO 3095 Järna

-20

-30 31.5

16

8 4 2 Wavelength [cm]

10

0

-10

-20

1

0.5

-30 31.5

X2 trailer wheel Freight wheel ISO 3095 16

8 4 2 Wavelength [cm]

1

0.5

Fig. 1. (Left) rail roughness level spectra based on CAT measurements on selected track sections of the line Stockholm – Gothenburg. Measurements were performed in March 2003 by Banverket. The years when the rails, according to available database information, were installed or ground are reported in the legend. “Corrugated rail” is the energy mean of the three spectra from Södertälje, Vretstorp and Töreboda. (Right) wheel roughness level spectra [8].

Rail Roughness Level Assessment

357

denoted “Corrugated rail”, see Fig. 1. In the following, it is used as a reference for a poor quality rail surface. All three sites are on tangent track with UIC 900A rails, nominal sleeper distance 0.65 m, resilient rail pads and concrete monobloc sleepers (250 kg) on ballast. Roughness on five X2 trailer wheels and 14 freight wheels (that were used in the freight bogie design G66) was measured by Johansson [8]. Three probes in mechanical contact with the wheel tread measured the deviation from the mean radius with sampling distance 0.5 mm and amplitude resolution 0.06 μm. All wheels had a minimum travelled distance of 100 000 km. The evaluated wheel roughness level spectra (energy means) are shown in Fig. 1. It is observed that the X2 trailer wheels are very smooth compared to the corrugated rails, whereas the freight wheels are rough due to tread braking with cast iron brake blocks.

3 Field Tests – Vertical Wheel–Rail Contact Force Wheel–rail contact forces were measured on an X2 trailer bogie operating on the line Stockholm – Gothenburg in October 2002 [3]. Measurement wheels instrumented with strain gauge bridges on each side of the wheel disc were employed. The wheels were calibrated for vertical (Q) and lateral (Y) static wheel–rail contact forces. The instrumented wheelset technology and the assessment of high-frequency contact forces are discussed in [3,4]. 1

PSD of vertical contact force [kN2/Hz]

10

0

10

-1

10

-2

10

-3

10

-4

10

0

250

500

750

1000

1250

1500

1750

2000

Frequency [Hz] Fig. 2. Power spectral density of measured vertical wheel–rail contact force. X2 trailer bogie on track in Vretstorp. Train speed 197 km/h. Measurement performed by Interfleet Technology Sweden.

358

J.C.O. Nielsen

Based on measurements in Vretstorp at train speed 197 km/h, the Power Spectral Density (PSD) spectrum of the vertical wheel–rail contact force is shown in Fig. 2. It is observed that contributions to the contact force are significant in the frequency range 500 – 1350 Hz. These contributions are caused by rail corrugation with wavelengths 4 – 8 cm. The distinct peaks in the spectrum (see peaks at 157 Hz and 314 Hz) are explained by the assessment procedure that derives the Q force by alternatively using information from the strain gauge bridge on either side of the wheel disc. These latter peaks are thus not caused by the dynamic train–track interaction and should be ignored.

4 Rail Roughness Spectra Used in a Parameter Study Based on the shape of the “Corrugated rail” spectrum in Fig. 1, several other artificial spectra were generated by adding or subtracting multiples of 3 dB from the original spectrum in the wavelength interval 4 – 8 cm, see Fig. 3. For longer and shorter wavelengths, roughness levels were increased or reduced between 0 and 2.5 dB to maintain the same shape of the new spectra. The alternative roughness level spectra can be regarded as spectra corresponding to rails after different time intervals with rail corrugation growth and no rail grinding. For example, the shape and levels of the new spectrum “- 6 dB” are similar to the corresponding features of the roughness spectrum that was measured at test site Järna, see Fig. 1. The extent of the time intervals between the spectra is different for different tracks depending on several parameters such as type of track and traffic load. Many track sections will show no significant growth of rail roughness even after many years of traffic, whereas roughness levels on other track sections may increase in the order of 2 – 3 dB per year [9]. A quantification of rail roughness/corrugation in the wavelength interval 3 – 8 cm is suggested by taking the mean square of rail roughness levels in the five 1/3 octave bands with centre wavelengths 3.16, 4.0, 5.0, 6.3 and 8.0 cm as 5 Lr,i / 10 2 2 1 ~ ~ rmean, = r . 3-8 cm ref ∑10 5 i =1

The corresponding mean roughness level

(2)

Lr, 3-8 cm is obtained as, cf. Eq. (1),

rmean, 3-8 cm ⎛~ Lr, 3-8 cm = 20 log10 ⎜⎜ rref ⎝

⎞ ⎟ [dB re 1 μm]. ⎟ ⎠

(3)

The mean roughness levels for the “Corrugated rail” and the ISO 3095 [10] spectra are 17.7 dB and 4.0 dB (re 1 μm), respectively, see the legend in Fig. 3. The mean roughness levels for the new spectra are obtained by adding or subtracting the corresponding multiples of 3 dB from the mean roughness level of the “Corrugated rail”.

Rail Roughness Level Assessment

359

30

Roughness Level [dB re 1 mum]

20

10

0

-10

Corrugated rail: 17.7 dB + 3 dB - 3 dB - 6 dB - 9 dB ISO 3095: 4.0 dB

-20

-30 31.5

16

8 4 2 Wavelength [cm]

1

0.5

Fig. 3. Rail roughness level spectra used in parameter study

5 Identification of Track Sections with Rail Corrugation A rail surface quality acceptance criterion based on a maximum allowed mean rail roughness level in the wavelength interval 3 – 8 cm needs to be determined by the responsible railway administration. For example, the criterion could be based on evaluations of the influence of roughness level on rolling noise and rolling contact fatigue impact. Different acceptance criteria can be used for different track sections depending on for example population density, type of traffic and train speed. If the mean rail roughness level for a given track section exceeds the acceptance criterion, grinding of the rails needs to be performed. The Corrugation Analysis Trolley can be used to monitor rail roughness growth on selected track sections at low cost. However, the use of measurement wheel technology on an X2 trailer wheelset is a more efficient tool to regularly monitor large portions of the main lines in the railway network as normal operating speeds of an X2 train are in the interval 150 – 200 km/h. In addition, a closing of the track to perform the measurement is not required. It is important that the measurement wheels are discbraked and that wheel roughness levels have been confirmed to be negligible when compared to roughness levels for rough rails. Interfleet Technology [11] has suggested using the root mean square value (rms) of the measured vertical wheel–rail contact force as a detector for severe rail irregularities. In particular, to identify track sections with short-pitch rail corrugation, the rms

~ Q 3-8 cm of the contact force after

360

J.C.O. Nielsen

16

Mean rail roughness level [dB re 1μm]

20

18

18 12

14

16 14

10

8

12 6

10 8 4

6 4

160

180

200 220 Train speed [km/h]

240

260

Fig. 4. Contour plot illustrating the calculated influence of train speed and mean rail roughness level on root mean square value of the vertical wheel–rail contact force. Contact forces have been band-pass filtered with cut-off wavelengths at 3 cm and 8 cm. X2 train with disc-braked trailer wheelsets. Track with UIC60 rails, resilient rail pads and concrete monobloc sleepers on ballast.

band-pass filtering with cut-off wavelengths at 3 cm and 8 cm has been proposed. Before calculating the rms, the measured time history of the contact force is partitioned into segments corresponding to 25 m of track. The rms-values of the band-pass filtered vertical wheel–rail contact force have been calculated in the computer program DIFF [12] for different combinations of train speed and mean rail roughness level. Wheel roughness for an X2 trailer wheel according to Fig. 1 has been assumed. A validation of DIFF was performed in Ref. [4]. A contour plot of the calculated rms-values is shown in Fig. 4. An increase in rms-values with increasing mean rail roughness level is observed. The increase in rms-values with increasing train speed is not as significant. Based on calculated results in the train speed interval 150 – 200 km/h, a logarithmic response surface function [13] for the mean rail roughness level was derived by a least-square solution as

Lr, 3-8 cm

~ ⎛ Q3-8 cm ⎛ v ⎞ = 14.5 − 13.4 log10 ⎜ ⎟ + 20.4 log10 ⎜ ⎜ 10 200 ⎝ ⎠ ⎝ 150 ≤ v ≤ 200 km/h

For a given train speed v [km/h] and a measured rms-value

⎞ ⎟ [dB re 1 μm], ⎟ ⎠ (4)

~ Q 3-8 cm [kN], the re-

sponse surface function can be used to determine the mean rail roughness level. As an

Rail Roughness Level Assessment

example,

361

~ Q 3-8 cm = 14 kN measured at 200 km/h indicates that the rails have similar

corrugation as the test sites in Södertälje, Vretstorp and Töreboda ( Lr, 3-8 cm = 17.5 dB re 1 μm).

6 Concluding Remarks The influence of mean rail roughness level and train speed on rolling noise and rolling contact fatigue impact will be discussed in [14].

Acknowledgements The work was performed at the Department of Applied Mechanics, Chalmers University of Technology in Göteborg, Sweden. It forms part of the activities within the Centre of Excellence CHARMEC (CHAlmers Railway MEChanics), see www.charmec.chalmers.se. The measurements were performed by Interfleet Technology Sweden and Banverket. Discussions with Per Gullers and Lars Andersson of Interfleet Technology Sweden are gratefully acknowledged.

References [1] Nielsen, J.C.O., Ekberg, A., Lundén, R.: Influence of short-pitch wheel/rail corrugation on rolling contact fatigue of railway wheels. Proc Instn Mech Engrs Part F – Journal of Rail and Rapid Transit 219, 177–187 (2005) [2] Ekberg, A., Kabo, E., Nielsen, J.C.O., Lundén, R.: Subsurface initiated rolling contact fatigue of railway wheels as generated by rail corrugation. International Journal of Solids and Structures 44, 7975–7987 (2007) [3] Gullers, P., Andersson, L., Lundén, R.: High frequency wheel–rail contact forces – field measurements and influence of track irregularities. In: 7th International Conference on Contact Mechanics and Wear of Rail/Wheel Systems (CM 2006), Brisbane, Australia, pp. 137–143 (2007) [4] Nielsen, J.C.O.: High frequency wheel–rail contact forces – validation of a prediction model by field testing. In: 7th International Conference on Contact Mechanics and Wear of Rail/Wheel Systems (CM 2006), Brisbane, Australia, pp. 41–48 (2007) [5] Nielsen, J.C.O., Lundén, R., Johansson, A., Vernersson, T.: Train–track interaction and mechanisms of irregular wear on wheel and rail surfaces. Vehicle System Dynamics 40(1-3), 3–54 (2003) [6] Grassie, S.L., Saxon, M.J., Smith, J.D.: Measurement of longitudinal rail irregularities and criteria for acceptable grinding. Journal of Sound and Vibration 227(5), 949–964 (1999) [7] Verheijen, E.: A survey on roughness measurements. Journal of Sound and Vibration 293(3-5), 784–794 (2006) [8] Johansson, A.: Out-of-round railway wheels – assessment of wheel tread irregularities in train traffic. Journal of Sound and Vibration 293(3-5), 795–806 (2006)

362

J.C.O. Nielsen

[9] Hiensch, M., Nielsen, J.C.O., Verheijen, E.: Rail corrugation in The Netherlands – measurements and simulations. Wear 253, 140–149 (2002) [10] CEN/TC256, Railway applications – acoustics – measurements of noise emitted by railway vehicles. Preliminary European norm prEN 3095, p. 35 (September 2003) [11] Gullers, P., Andersson, L.: Interfleet Technology Sweden, Personal communication (2006) [12] Nielsen, J.C.O., Igeland, A.: Vertical dynamic interaction between train and track – influence of wheel and track imperfections. Journal of Sound and Vibration 187(5), 825–839 (1995) [13] Montgomery, D.C.: Design and analysis of experiments, p. 649. Wiley & sons, New York (1991) [14] Nielsen, J.C.O., Ekberg, A., Kabo, E.: Acceptance criterion for limit rail roughness levels based on assessments of rolling noise and rolling contact fatigue (preliminary title). Department of Applied Mechanics, Chalmers University of Technology, Gothenburg, Sweden. International publication (to be submitted)

Testing the New Acoustic Rail Roughness Measurement Standard Chris Jones1, Pascal Fodiman2, Fabien Létourneaux3, and Briony Croft1 1

Institute of Sound and Vibration Research, University of Southampton, Highfield, Southampton UK Tel.: +44 (0) 23 8059 3224; Fax: +44 23 8059 3190 [email protected] 2 SNCF - Direction Générale Déléguée Infrastructure, 34 rue du Commandant Mouchotte, 75699 PARIS, France Tel.: +33 1 53 25 32 69; Fax: +33 1 53 25 32 26 [email protected] 3 SNCF - Agence D’Essai Ferroviaire, 21, avenue du Président Allende, Vitry-Sur-Seine, France Tel.: +33 1 47 18 82 32; Fax: +33 1 47 18 82 30 [email protected]

Summary Railway rolling noise arises from the combined roughness of the wheel and rail running surfaces. The rail roughness has therefore become an important parameter in the assessment of pass-by noise from trains. This paper describes a ‘road test’ of a new draft standard for rail roughness measurements being prepared for use in the Technical Specifications for Interoperability (TSI’s) in Europe which has been commissioned by CEN from TC 256 Working Group 3. Because of the timing of the ‘road test’, this paper outlines the purposes and aims of the project with some preliminary results and conclusions shown. It should not be taken to convey the final considered view of the working group. However, the general consistency of the measurements is shown as well as the variance. Once the results have been analysed and considered in more detail, they may be used to inform some clarifications or modifications of the standard.

1 Introduction Noise emission levels of the railway system are considered to be an important basic parameter to be controlled by the Technical Specifications for Interoperability (TSI’s). These have been adopted in Europe as vehicle acceptance test criteria for both high speed [1] and conventional vehicles [2] on the Trans European Network (TEN). They state noise emission limits when the vehicle is static and during pass-by at various speeds. For most of the practical range of vehicle speed, rolling noise is the dominant source, traction noise usually only being important at low speed and aerodynamic noise gaining roughly equal level with rolling noise only towards the highest operational speeds. The new TSI’s improve upon previous acceptance standards in that they reflect the understanding of rolling noise that has been reached through many years of rolling B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 363–369, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

364

C. Jones et al.

noise research [3]. In particular they recognise the important role of the track in determining noise levels. This is the case (i) because the track is very often the dominant source of rolling noise rather than the wheels of the vehicle, and (ii) the roughness of the rail combines with that of the wheel to excite noise radiation in both the wheels and the track. Of course, it is the sum of noise contributions from the wheels and the track that is measured against the acceptance criteria of the TSI’s, but the track’s influence is controlled by two conditions placed upon the ‘reference track’ on which the test takes place. The first constraint limits the ‘track noise transfer function’ (i.e. the spectrum of sound power per unit roughness) by specifying a minimum spectrum of track decay rates [4]. The second limits the reference track roughness so that its contribution to the combined roughness is as low as reasonably possible for an operational track. Although both noise TSI’s are already in force, the TSI working groups have requested two new European standards to be developed by CEN Technical Committee 256, Working Group 3, responsible for railway noise applications [5]. One of these relates to the measurement of the track decay rates [6]. The other relates to the measurement of the rail head roughness [7]. A project to ‘road test’ the provisions of the new roughness standard is the subject of this paper. This project is taking place currently during the time that the draft standard is out for comment by national bodies.

2 Main Provisions of the Standard For the method of pass-by noise measurement, the TSI’s refer to prEN ISO 3095:2001 although the final version has now been published and is not significantly different [8]. This standard itself already sets a limit spectrum for the track on which acceptance tests are made and prescribes a method for its measurement. The limit spectrum set in EN ISO 3095 is not used in the TSI’s, rather a tighter limit is set from within the TSI’s according to what was found possible by the associated NOEMIE project [9]. It found, for high speed trains (above 200 km/h), that a minimum wavelength range up to 0.25 m is required. 2.1 Longitudinal Position of Measurement Records and Sample Length EN ISO 3095 specifies a set of six positions for 1 or 1.2 m records of the rail-head profile. These are fixed with respect to ‘the microphone position’. This leads occasionally to the measurement of rail-head defects, welds etc. Such large localised irregularities are not appropriate to include in the roughness spectrum since they create forces and noise that are not linear with their depth (the contact geometry, and therefore the contact stiffness, changes radically). They also strongly distort the mean of the six sample records leading to both an overestimate of the level and uncertainty in the true operational roughness level. This has been a problem many times in the past and specifically at one of the test sites in the NOEMIE project. In the new standard, the choice of location of the measurement records is made by the measurers and they are advised not to include such irregularities. Moreover, the new standard envisages that a certain track section is to be characterised rather than assuming a microphone position. To keep the variance in the estimated spectrum at 0.25 m wavelength consistent with that at 0.1 m in EN ISO 3095, the new standard requires there to be a 15 m sample length in total.

Testing the New Acoustic Rail Roughness Measurement Standard

365

2.2 Lateral Position of the Measurements on the Rail Head EN ISO 3095 requires that the ‘running band’ on the rail head be identified (as ‘clearly visible’) and 1 or 3 lines of roughness measurement record be taken depending on its width. The new standard refers to a ‘reference surface’ that must be defined by the measurer. Noise measurements will then be valid as long as the wheel-rail contact remains within the reference surface. Its identification from the running band is an important subject in the new standard. Three different criteria depending on the situation and the purpose of the measurements are offered: (1) the running band is visible and is known to be a product of the rolling stock for which the roughness measurement is to be used, (2) the contact position can be measured for the specific rolling stock at the time of roughness measurement, (3) the contact position can be predicted from the geometry of rail and wheel transverse test section. 2.3 Processing The data must be processed to remove some unwanted ‘pits and spikes’ and produce a one-third octave level roughness spectrum. EN ISO 3095 does not prescribe how the processing is done although it recognises that large differences can result. The processing is much more tightly controlled in the new standard. To remove the effects of dust or grains of dirt on the railhead, an algorithm is included that removes ‘spikes’, i.e. very short (much shorter than the wheel-rail contact patch), sharp, upward deviations. This recognises that such features would be crushed or strongly deformed in the contact not leading to significant relative displacement between wheel and rail. A second algorithm, ‘curvature processing’ is specified to deal with short downward features found by the small probe of the instrument that would not affect a much larger radius wheel. The new standard specifies alternative spectrum analysis methods, (i) Hanning window, discrete Fourier transform and averaging in one-third octave bands, or (ii) digital one-third octave filtering.

3 The ‘Road Test’ The purpose of the road test is to check that the standard can be interpreted consistently and leads to a consistent estimate of roughness spectrum when used by different measurers with different instruments. Many of the instructions of the new standard have not been practiced by measurers before and so these are also being tested for practicability and effectiveness. The exercise is not concerned with testing instruments or measurement technology. The standard specifies minimum performance criteria but otherwise is designed to be as inclusive as possible with regard to technology; otherwise standardisation could stifle innovation. 3.1 Test Sites Two sites were used. The first was on the main ring of the Siemens Test Track Centre at Wildenrath in northern Germany. The rail-head had been ground about 6 months before the test using a special ‘acoustic grinding’ with longitudinal grinding action. Fig. 1(a) shows a typical sample of the rail head at this site. There were very few

366

C. Jones et al.

(b)

(a)

Fig. 1. (a) Rail head at Wildenrath, (b) at Loiriol

significant defects of the rail head within the 100 m ‘reference section’ of track. However, an interesting consideration arises; the site is used for testing rolling stock with (mainly new) 1 in 20 and 1 in 40 coned wheel profiles. This has resulted in two clear separate (narrow) running bands. The measurers were directed to consider the more recent brighter band of the two. The line speed is 120 km/h. The second site was at Loriol on a conventional-speed service line in southern France. It is mostly trafficked by freight trains with some regional multiple units, locomotive-hauled passenger stock and a few TGV’s. Fig. 1(b) shows a sample of the rail head typical of the Loriol test section. Here the running band was wider and less distinct than at Wildenrath. In these circumstances the teams were guided to test the contact position of the passenger stock in deciding the position of the reference surface. A method used by one team is illustrated in Fig. 1(b). 3.2 Teams and Instruments The measurement teams and specific device types shall not be identified here. At the Wildenrath site, five teams took part using four 1.2-metre fixed straight-edge instruments of two different types. These move linear voltage displacement transducers (LVDT) along the instrument to measure the displacement profile of the rail head. The fifth team used a trolley that moved along an accelerometer along the rail head, then deriving the roughness by integration. At the Loriol site, seven teams took part with eight instruments. Three separate types of instrument measured 1.2 m records using LDVTs. Two types of instrument measured continuously over the whole 100 m using an accelerometer and one measured the whole section with LDVTs. All teams that took part in the road test measured at Loriol. 3.3 The Test Procedure The teams measured separately so that decisions on the positioning of the reference surface and measurement records were made independently. Each team was shown the same 100 m test section of track and asked to characterize its roughness in the wavelength range up to 0.25 m. Apart from the very few directions already indicated in Section 2.1, they then interpreted the standard freely. A second measurement was

Testing the New Acoustic Rail Roughness Measurement Standard

367

then made by each team along the same single 15 m line specified by the test coordinator. This was done so that it would be possible to differentiate between effects of different instruments and positioning of measurement samples.

4 Observations Made during the Tests The test coordinator observed the practice of each team in response to the standard. Given the differences in the running band of the two sites and their relationship to the rolling stock, the first two techniques used for identifying the reference surface (Section 1.2) worked well and led to closely similar positions of the reference surface. One team measured the rail head profile and calculated a theoretical (static-geometry) contact position for an unworn wheel. This was then used with the method illustrated in Fig. 1(b) to decide on the reference surface location. Different teams had different practices in cleaning the rail head before measurement. The teams using short record instruments used solvent and rags. It was clearly not practicable for the long-record measuring teams to follow this practice. At Loriol the rail head was regularly ‘cleaned’ apart from easily-moved dust or moisture, by the running trains. The cleaning practice may have been more significant at Wildenrath where, at the start of the test, there were a lot of bird droppings on the rail head. Apart from removal of some gross matter, the one long-record measurement at Wildenrath was made without cleaning this from the rail. For the 1.2 m-record instruments, most teams took the strategy of scattering the 13 to 16 records locations (approximately) evenly over the 100 m with a few locations moved a little to avoid geometrical features. One team placed them in a pattern strongly weighted towards the mid-point of the 100 m. The trolley instruments measured the whole 100 m with extra length on the ends so that start-up and stopping transients could be discarded from the record afterwards. These differences in practice are allowable within the freedoms of the standard except that the standard does demand that ‘moisture and contamination should be removed from the rail-head before measurement’. No further stipulation on this is given. 20 15

10

Roughness(dB re 1um)

10

Roughness (um)

5 0 -5

-10

0

-10

-15

-20

-20

-25 -30 0

0.2

0.4

0.6 0.8 Distance (m)

(a)

1

1.2

1.4

0.25

0.125

0.063

0.0315

0.016

0.008

0.004

1/3 Oct Band Centre Wavelength (m)

(b)

Fig. 2. (a) Typical 1.2 m record of rail head profile, ──, before and, ──, after spikes removal and curvature processing. (b) Roughness spectra: ──, from each of 15 1.2 m records; ──, mean; - - -, mean ± one standard deviation.

368

C. Jones et al.

10

Roughness (dB re 1um)

Roughness (dB re 1um)

10

0

-10

-10

-20

-20 0.25

0

0.125 0.063 0.0315 0.016 0.008 1/3 Oct Centre Band Wavelength (m)

0.004

0.25

0.125 0.063 0.0315 0.016 0.008 1/3 Oct Centre Band Wavelength (m)

(a)

0.004

(b)

Fig. 3. (a) Comparison of spectra (a) from each of the three different types of 1.2 m record measuring instruments (──, as mean of Fig 2(b), ·····, mean ± one standard deviation, - - - and ·-·- other 1.2 m instruments), (b) from a 1.2 m record instrument and a continuous measurement instrument (- - -, whole 100m record; -·-·-, 16 × 1.2 m segments of the 100 m measurement)

5 Example Results from Loriol Within the present paper, it is only possible to show a small subset of the measurements. They are limited to those from Loriol. A simple processing code that is believed to comply with the new standard has been developed and has been made available to all teams to use and to comment on. This code has been used for the results presented here. Fig. 2(a) shows a 1.2 m record with the pits and spikes that are removed in the first stages of processing. Only a few spikes are removed but a more general modification of pits is made by the curvature processing. Fig. 2(b) shows the ensemble and average of the spectra of 13, 1.2 m records. Though 15 m of record has been used the standard deviation is still about ±4 dB in the 0.25 m wavelength band and about ±1.5 dB in the 0.00315 m band. (The dB value conforms approximately to Gauss distribution statistics.) This measurement shows that the Loriol site was well within the TSI limit, also plotted. Mean spectra produced by each of three different 1.2 m instruments are shown in Fig. 3(a). The result from Fig. 2(b) is shown with its standard deviation. Bearing this in mind, the three spectra are consistent. The analysis of an unedited 100 m record (Fig. 3(b)) produces a significantly higher spectrum. This is because of the localized geometrical features that have not been avoided. When 16, 1.2 m segments are arbitrarily taken from this and analysed in the same way as the short records the result is consistent apart from a peak in the 0.02 m band and again in the 0.00315 m band. These may be a function of localised minor corrugation in one of the records (again, that may have been avoided by the other instruments). Neither peak is present in all measurements by that instrument.

6 Conclusions All the measurements of a road test of the new CEN standard for roughness measurement have been carried out and some initial analysis. The provisions for choosing the

Testing the New Acoustic Rail Roughness Measurement Standard

369

reference surface have been shown to work well and, within the variance, associated with the 15 m of record length for the wavelength range, the measurements generally agree, even for the long record instruments. A lack in comparability with the short record instrument seems to depend more on the difficulties of cleaning and avoiding localised rail-head defects in the data than on the technology used. Complete results and discussion will emerge later for consideration in finalising the standard.

Acknowledgements Various organisations took part in the test at their own cost. Siemens AG Transportation Systems and the SNCF provided the tests sites and facilities. The independent coordination has been funded by the EC via the European Committee for Standardization (CEN).

References [1] Commission communication in the framework of the implementation of the Council Directive 96/48/EC of July 1996 on the interoperability of the trans-European high-speed rail system. Official Journal of the EU (8/2/2006) [2] Commission decision of 23, concerning the technical specification for interoperability relating to the subsystem ‘rolling stock - noise’ of the conventional rail system. Official Journal of the EU (30/12/2006) (Dec. 2005) [3] Thompson, D.J., Jones, C.J.C.: A review of the modelling of wheel/rail noise generation. Journal of Sound and Vibration 231, 519–536 (2000) [4] Jones, C.J.C., Thompson, D.J., Diehl, R.J.: The use of decay rates to analyse the performance of railway track in rolling noise generation. Journal of Sound and Vibration 293, 485–495 (2006) [5] Fodiman, P., Létourneaux, F., Jones, C.J.C.: European standardisation related to railway noise emission: New standards to assess the reference track performances, Forum Acusticum, Budapest (2005) [6] Railway applications – Noise emission – Characterization of the dynamic properties of track sections for pass by noise measurements. pr EN 15461, 2006, Comité Eurpéen de Normalisation (CEN) [7] Railway applications – Noise emission – Rail roughness measurement related to rolling noise generation, pr EN 15610, 2006, Comité Eurpéen de Normalisation (CEN) [8] Railway applications – acoustics – measurements of noise emitted by railbound vehicles, EN ISO 3095: 2005 [9] Fodiman, P., Staiger, M.: Improvement of the noise Technical Specifications for Interoperability: the input of the NOEMIE project. Journal of Sound and Vibration 293, 475–484 (2006)

Practical Implementations and Benefits of Highly Accurate Rail Roughness Measurements S. Lutzenberger and P. Holm Müller-BBM GmbH, Robert-Koch-Strasse 11, 82152 Planegg near München, Germany Tel.: +49 89 85602-251; Fax: ++49 89 85602-111 [email protected]

Summary Smooth rails are an important factor to reduce railway noise. An evaluation of the rail surface roughness can be done by means of roughness measurement devices. This is, however, a challenging task as measurements in the sub μm domain have to be carried out under harsh conditions. This paper deals with requirements of rail roughness measurement systems and how these requirements can be fulfilled. Numerous measurement campaigns show that these measurements are difficult to carry out under the varying environmental conditions. These requirements will be explained in detail. An important task is to resolve a maximum accuracy of relative measurement devices. Contributions to the measurement error are discussed in detail. A calibration procedure is shown, that allows for minimising deterministic errors. Stochastic errors, however, are inevitable. These errors, can be minimised by an optimised design of the device. It will be shown how the inherent noise level of a device can be determined. Noise levels less than –20 dB re1μm are possible. Highly accurate measurement profiles in a wavelength range from 3 mm to 1.2 m are shown.

1 Introduction Rolling noise is the predominant railway noise source in the speed range from approximately 60 km/h up to 250 km/h. As most trains, especially when passing urban areas, travel at foresaid speeds, rolling noise can be seen to be the dominant noise source. Rolling noise results from the unevenness of the wheel and the rail. This causes wheels and rails to vibrate and to radiate noise. Due to intensive research activities [1] and [2], this mechanism is well understood today and has lead to various simulation models to predict railway rolling noise [3], [4] and [5]. The characterization of the rail roughness requires specific measurement devices. Such devices determine the acoustical relevant roughness of the rail by sensors that scan the surface of the rail directly (displacement transducers/accelerometers) or indirectly by determining the combined roughness of wheel and rail e.g. by acceleration sensors at the axles of a railway car [6]. It is obvious that direct measurements give more accurate rail roughness values than indirect methods. The main purposes of direct roughness measurements are B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 370–377, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Practical Implementations and Benefits

371

1. Type testing of railway vehicles (EN ISO 3095 [7], EN ISO 3881 [8] and TSI [9] and [10]). 2. Quality tests for the so called acoustical grinding of rails. 3. Rail research purposes e.g. to monitor the short pitch corrugation. 4. Calibration of new roughness measurement methods for tasks that require less precision. 5. Evaluation of parameters for acoustical simulation schemes. Different programs exist for this task like the RMR noise prediction method developed in the "HARMONOISE" project, TWINS (developed for ERRI, [4] and [5]) or RIM (by MüllerBBM and the Deutsche Bahn AG, [3]).

2 Requirements to Roughness Measurement Devices The first version of the roughness measurement device RM1200E was developed in the late 1980s and since then was frequently used by various railway companies. In numerous measurement campaigns a lot of valuable experience could be gained in the field of roughness measurements that contributed to the construction of the new device mbbmRM1200 (introduced 2005). In general the problem is to measure reliably in the sub μm domain under challenging environmental conditions. Some practical requirements are discussed here in detail: 1.

2.

3. 4. 5.

6.

7. 8.

When measuring during train operation, the number of necessary adjustments should be a minimum. For the same reasons, a high measurement speed is required. Environmental conditions are challenging. The device should be equipped with a weather protection (for rainy days), a temperature compensation, a daylight suitable screen (for hot days) and a lighting of the stylus (for night measurements or measurements in tunnels). The device should be capable to cope with all types of rails (e.g. UIC60 or embedded rails). For research purposes measurements over the whole width of the rail head are necessary. Due to the short time spans between two consecutive trains, fast measuring cycles should be well supported by the system. The device itself should store all measurement parameters automatically and check the user settings as comprehensive as possible (QM system). Interfering environmental conditions and different disturbances occasionally cause invalid measurement results. It is recommendable to have the measurement signal shown in real time on the screen and to have a cross check of the position of the sensor e.g. by a laser light to interpret the measurement results on site directly after each measurement. In order to reduce stick-slip effects, the sensor should not touch the rail directly. A stylus should allow for using different curvature radius and for replacement if worn out. Laser pointing at the sensor position allow for a cross check of the mirror. This is of special importance for measurements in the dark but also for day time measurements in order to verify that the measurements are carried out in the centre of the mirror.

372

S. Lutzenberger and P. Holm

9.

A good manageability requires the independency of external power supply, an automatic calibration, no external measurement controller, a wireless connected external control PC and the fast and easy transportability between two measurement sections. 10. Regardless of specific accuracy requirements the device should have a maximum inherent noise level of al least –20 dB and the variation of the inherent noise should be small. Normative requirements are given in the standards EN ISO 3095:2005 [7], EN ISO 3381:2005 [8], TSI conventional rail [9], TSI high-speed rail system [10] and prEN 15610:2006 [11].

3 Accuracy of the Device prEN 15610 requires a verification of the measurement system. This task is difficult due to the lack of a calibration surface with a defined roughness that can be used for these purposes. In this section, the concept of relative displacement measurements as shown in Fig. 1 is discussed in detail. A stylus measures the difference between the rail and a reference surface.

guidance

reference surface drive mechanism stylus rail

Fig. 1. Principle of relative displacement measurements

Different sources lead to errors in the measured data (Fig. 2). As the reference surface is not perfectly even, deviations ΔzRS are inevitable due to static deflections or production tolerances. Additional deviations occur due to inaccuracies in the guidance ΔzG and the sampling process ΔzS. ΔzG,d + ΔzG,s reference surface rail

ΔzRS,d + ΔzRS,T ΔzS r

sensor deflection discretised signal

Fig. 2. Error accumulation in the relative displacement measurements

The particular meaning of the major contributions used is given by ΔzRS,d permanent variations of the reference surface (deterministic) ΔzRS,T variations of the reference surface due to temperature changes (temperature dependent) ΔzG,d variations in the guidance (deterministic, repeatable)

Practical Implementations and Benefits

373

ΔzG,s irregular variations in the guidance and the reference surface (stochastic, not repeatable) ΔzS inaccuracies in the sensor sampling process, due to the discretisation and inaccuracies resulting from the measurement length (included in the discretised sensor signal) r surface roughness ΔzSR inaccuracies in the sampling process when the stylus moves along a rough rail ΔzA inaccuracies in the data processing, e.g. pit-spike removal, and modifications that are necessary in order to determine the roughness that is seen by the wheel As various factors influence the accuracy of the device it is not sufficient to consider single aspects. The whole concept of the device and its constructive details have to be taken into account. In a general approach, the major contributions to the measurement signal s can be characterised by

s = r + Δz RS ,d + Δz RS ,T + ΔzG ,d + ΔzG ,s + Δz S + Δz SR + Δz A

(1)

together with the corresponding measurement error e which is given by

e = Δz RS ,d + Δz RS ,T + ΔzG ,d + ΔzG ,s + Δz S + Δz SR + Δz A

(2)

The deterministic contributions ΔzRS,d and ΔzG,d can be reduced significantly by means of a calibration procedure (section 5). Changes in the temperature cause a curvature of the device and, therefore, variations of the reference surface ΔzRS,T. This problem can be solved by calibrating the device after each significant change in the temperature (which means that the device has to be recalibrated on site) or preferably by an internal temperature compensation. Irregular variations in the guidance ΔzG,s are inevitable e.g. as each movement causes vibrations. It is therefore evident to minimize ΔzG,s by an optimized design of the device. These variations are the main contributions to the inherent noise level. ΔzG,s can be minimized for example by extremely smooth sliding guides, a brushless motor design and blanked out gear meshing frequencies. Inaccuracies ΔzS depend on inaccuracies of the roughness signal acquisition and processing system itself. Possible errors are, e.g., errors in the discretisation level of the stylus or aliasing effects in the discretisation. The accuracy of the sensor signal itself can be tested by a gauge. The inaccuracies ΔzSR occur if the sensor motion along a rough rail causes stick-slip effects of the stylus on the surface or the lift of the stylus due to a too high velocity. The error ΔzA depends of the data post processing procedure. In the new standard prEN 15610 different aspects concerning the data acquisition and data analysis are specified.

4 Calibration, Inherent Noise and Repeatability A calibration procedure can minimize the errors significantly and allows for the determination of the error level of the device. A surface with a very low roughness level („calibration stone“) is required for this task. A calibration stone with the lowest roughness available (less than –40 dB in the whole frequency range under consideration)

374

S. Lutzenberger and P. Holm

is used here. Due to the extreme evenness of this surface, the error ΔzSR can be neglected. Later modifications of the roughness profile will not be considered here, therefore ΔzA = 0. Moreover, a temperature compensation is assumed to be included in the system, therefore ΔzRS,T can be neglected. The resulting error can be described by

e = Δz RS ,d + ΔzG ,d + ΔzG ,s + Δz S

(3)

The total deterministic error c (“calibration curve”) of the device

c = Δz RS ,d + ΔzG ,d ≅

1 n ∑e m i =1

(4)

can be determined by averaging several measurements m (m = number of calibration runs) at a location x on the calibration stone. The stochastic contributions ΔzG,s and the contribution ΔzS are reduced by averaging. This only holds as long as ΔzG,s is small compared to the other contributions (high signal to noise ratio). ΔzS can be assumed to be small as it is just the discretisation error. n measurements in several lanes over the whole width of the calibration stone can be used to rate the error level ΔzG,s + ΔzS if the calibration curve is subtracted from each measurement record ri.

Δz G , s + Δz S ≅

1 n ∑ ( ri − c ) n i=1

(5)

Different lanes over the width of the calibration stone have to be measured as the error varies with the lateral position of the sensor. The inherent noise level ΔzG,s can be estimated to

ΔzG ,s ≅

1 n ∑ (ri − c) n i =1

(6)

A statistical evaluation of the error shows some scattering over different measurements. This scattering should be very small compared to the rail roughness in order to have a high repeatability of the measurements. An example is shown in Fig. 3. Different lanes along the calibration stone were measured and the calibration curve was subtracted from each single measurement.

Envelope of 5 measurements Envelope of 20 measurements Average of 20 measurements over the width (“error level”) EN ISO 3095 TSI

Wavelength [cm ] 32

16

8

4

2

1

0.5 0.25 0.125 0.06

Fig. 3. EN ISO 3095 limit (thick solid line), “error level” and scatter band of mbbmRM1200

Practical Implementations and Benefits

375

Rounded to 0.5 μm smooth signal Rounded to 1.0 μm smooth signal Rounded to 2.0 μm smooth signal

Wavelength [cm]

Fig. 4. Roughness measurement results, dependent on signal discretisation

The error level was determined by averaging 20 measurements. The scattering is about 2 dB for small wavelength and about 5 dB for long wavelength. The maximum error of all measurements is about –10 dB lower than the smoothest limit curve for small wavelengths and about than –20 dB for long wavelengths. As the scattering is small, the measurements are highly repeatable. Furthermore, occurring errors are much lower the roughness limit curves. Another study was carried out to examine the influence of the discretisation of the sensor. As an example, Fig. 4 shows the effect of sensor discretisation levels on measured roughness levels on the calibration stone. In order to study the discretisation level, the calibration curve was not subtracted. It can be seen that for such a rail the results already converge with a discretisation level of 0.5 μm. The remaining error at a discretisation level of 0.5 μm is small compared to ΔzG,s. Another remaining influence factor is the measurement length of the device. Relative displacement measurements are, in general, limited to a certain length (e.g. 1.2 m) in contrast to continuous acceleration measurements. This disadvantage can be compensated in a simple manner by measuring successive sections with an overlap and then combining the records. The combining process can be done automatically. The Smooth signal, overlapping 3.2 m Rough signal, overlapping 5.2 m

Wavelength [cm] 256 128

64

32

16

8

4

2

1

0.5

0.25

Fig. 5. Example of combined measurement records

376

S. Lutzenberger and P. Holm

use of an on-board inclinometer enhances this procedure. Fig. 5 shows section wise roughness measurements on two different rails. The measurements were combined and evaluated for long wavelengths. Overlapping measurements with a total length of 5.2 m (3.2 m) allow for evaluating wavelength of about 0.5 m (0.3 m) reliable. The relevant wavelength up to 0.2 m for high speed trains [11] can be measured very accurately, by combining two consecutive measurements of a device with a length of 1.2 m. One single measurement is sufficient to measure the required wavelengths for conventional trains accurately.

5 Conclusion Rail roughness measurements are required for different tasks e.g. like qualifying track sections for type testing of railway vehicles or controlling the rail roughness after acoustical grinding. To these aims, roughness levels have to be measured in the sub μm domain under adverse environmental conditions. In particular measurements under running traffic are difficult to conduct. Roughness measurement devices should therefore be suitable for all climatic conditions and they should allow for fast and reliable measurements during running traffic. The devices should be capable to measure different types of rails, e.g. vignol rails or embedded rails. For the interpretation of the measurement data it is important to display the signal in real time on the screen and to have a cross check to the actual position of the sensor. Moreover, a real-time visual control helps to detect disturbances that may occur in the measurements. Different sources leading to errors in the measured data were discussed and it was shown how a calibration procedure can reduce these errors significantly. Furthermore it was shown how the error level of a device can be determined by testing the sensor with a gauge and by executing several measurements on a calibration stone. The error levels, however, can be minimised by a sophisticated constructive design of the device itself. Moreover an internal temperature compensation of the device is required. It was shown that by combining these principles very low error levels can be achieved. One single measurement with a device with a length of 1.2 m allows for evaluating the roughness profile for conventional trains. Two consecutive measurements provide the roughness profile for high speed trains. Very accurate results can be obtained by relative roughness measurements.

References [1] Thomson, D.J.: On the relationship between wheel and rail surface roughness and rolling noise. Journal of Sound and Vibration 1, 149–160 (1996) [2] Remington, P.J.: Wheel/rail noise – part 1: characterisation of the wheel /rail dynamic system. Journal of Sound and Vibration 46, 359–379 (1976) [3] Diehl, R., Hölzl, G.: prediction of wheel/rail noise and vibration — validation of RIM. In: Proceedings Euronoise 1998 (1998) [4] Thompson, D.J., Hemsworth, B., Vincent, N.: Experimental validation of the TWINS prediction program for rolling noise, Part 1: Description of the model and method. Journal of Sound and Vibration 193(1), 122–135 (1996)

Practical Implementations and Benefits

377

[5] Thompson, D.J., Fodiman, P., Mahe, H.: Experimental validation of the TWINS program for rolling noise, Part 2: Results. Journal of Sound and Vibration 193(1), 137–147 (1996) [6] Verheijen, E.: A survey on roughness measurements. Journal of Sound and Vibration 293, 784–794 (2006) [7] EN ISO 3095: Railway applications - Acoustics - Measurement of noise emitted by railbound vehicles (2005) [8] EN ISO 3381:2005: Railway applications - Acoustics - Measurement of noise inside railbound vehicles (2005) [9] Commission Decision concerning the technical specification for interoperability relating to the subsystem rolling stock — noise of the trans-European conventional rail system, Annex, Technical specification for interoperability relating to the subsystem ‘rolling stock — noise’ of the trans-European conventional rail (2005) [10] Commission decision concerning the technical specification for the interoperability relating to the rolling stock subsystem of the trans-European high-speed rail system referred to in Article 6(1) of Directive 96/48/EC, Annex, Technical specification for interoperability relating to the rolling stock subsystem (2002) [11] pr EN 15610, Bahnanwendungen – Geräuschemission – Messung der Schienenrauheit im Hinblick auf die Entstehung von Rollgeräusch (Dezember 2006)

New Rail Dampers at the Railway Link Roosendaal-Vlissingen Tested within the Dutch Innovation Program E. van Haaren1 and G.A. van Keulen2 1

DHV B.V., P.O. Box 1132, Amersfoort, The Netherlands Tel.: +31 33 468 2986; Mob.: +31 6 5021 3469 [email protected] 2 Delta Rail B.V., P.O. Box 8125, 3503 RC, Utrecht, The Netherlands Tel.: +31 30 3005100 [email protected]

Summary The upgrade of the railway link Roosendaal-Vlissingen will result in more freight traffic. Hence, sufficient measures are necessary to prevent more noise and vibration annoyance. Source mitigation plays an important role in the realisation of this upgrade of the existing infrastructure. Besides silent freight trains, rail dampers will be applied near urban areas to limit the volume of noise barriers. A number of candidate rail dampers has been selected and tested to be checked whether a 2 or 3 dB(A) reduction can be reached. Measurements have been carried out within the Dutch Innovation Program Railway Noise (InnovatieProgramma Geluid: IPG). Tests include the dampers by manufacturers Schrey & Veit and Alom (James Walker). Additionally, a prefab damper by Corus will be part of the program. The tests have been carried out at the railway link Roosendaal-Vlissingen, near Krabbendijke, according to the National (draft) legislation. Results show that a 2 to 3 dB reduction can be reached for one of the new types of dampers. As demonstrated previously for the Corus rail damper, the reduction is strongly dependent on the rail roughness. The paper presents results of the Schrey & Veit and Alom (James Walker) dampers applied at the track near Krabbendijke. A comparison of the Schrey & Veit dampers with the previous published results of the Corus damper shows that the dependence on rail roughness is roughly similar for different types of dampers. However, spectral results vary strongly. Based on findings of these measurements the final prescribed measurement method in the new noise legislation, valid by January 2007, has been modified.

1 Railway Link Roosendaal-Vlissingen The Schrey & Veit and Alom (James Walker) dampers have been applied at a site near Krabbendijke, which is part of the railway link Roosendaal-Vlissingen. Since spatial planning includes a new urban area near the track, noise measures have to be applied. In case the rail dampers are approved, they will remain in place and be part of the total package of noise measures. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 378–383, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

New Rail Dampers at the Railway Link Roosendaal-Vlissingen

379

In total 3 types of dampers have been tested: 1. 2. 3.

Schrey & Veit (S&V High Speed Line): 1 section; Schrey & Veit (S&V modified) 2 sections; James Walker/Alom (JW Tiflex: @10u8): 2 sections.

The type of track is ballasted track with UIC54 on concrete sleepers. The track has been grinded about a year prior to the measurements.

Ref

S&V JW gem.

S&V S&V HSL gem.

JW

Fig. 1. Test site near Krabbendijke: each test section is 100 meter long. Noise emission is measured at 7.5 meter by 3 microphones at each section, according to ISO3095.

2 Applied Method to Determine the Damping In the Netherlands the regulation of railway noise is set by the Noise abatement act (Wet geluidhinder) and more specific about the noise emission characteristics in appendix IV [1]. The Technical regulation prescribes the way to determine the classification of railway vehicles into noise emission classes. Procedure C of this regulation prescribes the method to determine the noise emission of new types of track construction as well as relevant change to an existing type of track. This procedure is applied to rail dampers. Procedure C prescribes to perform noise emission measurements at the test track as well as to a representative reference track, based on ISO 3095. The principle of this procedure is based on a relative emission of the test track compared to the reference track. It is assumed that the variation due to the railway vehicle can be neglected. Therefore it is required that the wheel roughness of the railway vehicles is dominant. Hence, the railway vehicles used will have to have wheels with high roughness. For the track it is recommended that the roughness at the reference track as well as the test track is below the ISO curve. The procedure supplies an emission correction term (spectrum) for the track that is independent to the train velocity and the type of train. Therefore the result should be representative for all types of railway vehicles.

380

E. van Haaren and G.A. van Keulen

The measurements in Krabbendijke have been carried out based on a preliminary version of procedure C. Based on the understanding of the results in Krabbendijke, some additional prescriptions to the procedure have been made: • • •

the “noise reduction” should be corrected for differences in rail roughness of the reference and test sections; the “noise reduction” should be representative for “average” rail roughness; the “noise reduction” should be averaged based on measurements for at least 2 types of trains.

These additional prescriptions have been used in the determination of the emission correction term (noise reduction) described in this paper.

3 Results of the Measurements 3.1 Rail Roughness The results of the rail roughness measurements are shown in figure 2. The figure also shows the ISO3095 limit as well as the average rail roughness in the Netherlands. Reference

10,0

Rail roughness [dB (re 1 µm)]

S&V---gemod 2 JW-----Tiflex 2

5,0

S&V HSL S&V---gemod 1 JW-----Tiflex 1

0,0

ISO3095-limit NL average -5,0

-10,0

-15,0

-20,0 10

5

2,5

1,25

0,63

0,315

Wavelength (cm)

Fig. 2. The rail roughness lines representing the average rail roughness for both rails at each test section as shown in figure 1

The rail roughness is well below the average rail roughness in the Netherlands, 10 to 5 dB for decreasing wavelengths, except at 2.5 cm. Some average wave lengths exceed the ISO limit. However, the exceptional limits of a maximum of 6 dB for a single band and of a maximum of 3 dB for three neighbouring bands are met. Therefore all average roughness levels meet the ISO limit.

New Rail Dampers at the Railway Link Roosendaal-Vlissingen

381

3.2 Noise Reduction Results Pass by measurements have been carried out for 4 passenger trains and 5 freight trains. Measurements have been averaged for both types of trains individually. Additionally, noise measurements results have been averaged for three microphones at each test sections. The noise emission has been corrected for differences in rail 4,0 3,0

Reduction [dB]

2,0 1,0 0,0 -1,0 JW tiflex 2

-2,0

JW tiflex 1 -3,0

S&V (HSL)

-4,0

S&V gemod 1

-5,0

S&V gemod 2

-6,0 31,5

63

125

250

500

1000 2000 4000 8000

LAeq LAeq

Frequency [Hz] Fig. 3. The noise reduction for each damper section, measured at freight trains 4,0 3,0

Reduction [dB]

2,0 1,0 0,0 -1,0 JW tiflex 2

-2,0

JW tiflex 1 -3,0

S&V (HSL) S&V gemod 1

-4,0

S&V gemod 2

-5,0 -6,0 31,5

63

125

250

500

1000 2000 4000 8000

LAeq LAeq

Frequency [Hz] Fig. 4. The noise reduction for each damper section, measured at passenger trains

382

E. van Haaren and G.A. van Keulen

roughness between the damper sections and the reference section. In this correction typical wheel roughness levels are assumed for the types of trains measured as described in the emission characterisation document [1]. The noise reductions figures show a similar trend for all three types of rail dampers. At low frequencies an increase in noise emission can be observed. At higher frequency the damper increases to a maximum around 1 kHz. At 500 Hz a local minimum of the damper effect is found. The highest damper effect is found in case of the S&V (HSL) damper at passenger trains (2.6 dB). The damper effect is 0.6 to 1.3 dB in the case of freight trains, while the passenger trains measurements show reduction of 1.5 to 2.6 dB. On average the damper effect is 1 dB higher at passenger trains. Table 1 presents the average results for each type of damper. Table 1. Average measured noise reductions

Noise reduction [dB] Schrey & Veit (HSL) Schrey & Veit (mod.) James Walker/Alom

Passenger train 2.6 2.0 1.9

Freight train

Average

1.0 0.8 1.1

1.8 1.4 1.5

3.3 Damping Effect Versus Rail Roughness The noise reduction presented in the previous section does not reach the minimum noise reduction requirement of 2 dB set by ProRail. However, results gained for the Corus rail dampers in former experiments carried out under the IPG program, have shown a strong dependency on rail roughness. A higher roughness level results in higher noise reduction [3]. Since the rail roughness levels at the Krabbendijke test site are low, a higher noise reduction can be expected at higher rail roughness levels. Therefore additional measurement results at high rail roughness levels are required to determine a representative noise reduction of the rail dampers. For the S&V HSL damper type results at higher roughness are available as a result of tests carried out at the High Speed Line near Rotterdam. Figure 5 shows the noise reduction of the standard S&V HSL damper as well as the Corus rail damper, measured at different rail roughness levels. The overall trend in the noise reducing effect versus the rail roughness level is quite similar, although the reducing effect of the Corus damper is about 0.6 dB higher. The overall rail roughness dependency of the damper effect is similar; but the relations at specific frequency bands vary strongly. Frequency band specific interpolation has been used to determine the representative noise reduction effect of the S&V damper at the average rail roughness in the Netherlands. The noise reduction effect for the S&V damper has been determined as Cbb value of 2.6 dB, according to the emission characterization document. The Alom rail damper has not yet been tested at high rail roughness. Therefore no representative noise reduction is available. Additional tests will be carried out later this year.

New Rail Dampers at the Railway Link Roosendaal-Vlissingen

383

Noise reduction [dB]

6 5 4 3

S&V

2

Corus 1 0 0

5

10

15

20

Rail roughness LλCA Fig. 5. Noise reduction in relation to rail roughness

4 Conclusions New tests with rail dampers have been carried out at the railway link RoosendaalVlissingen near Krabbendijke. Since the measurements were carried out at low rail roughness the noise reductions were lower than expected. As demonstrated previously in case of the Corus rail damper, the reduction of this damper is strongly dependent on the rail roughness. Combination with results at higher rail roughness results in a representative reduction of 2.6 dB for the S&V rail damper. Based on findings of these measurements prescribed measurement method in the new noise legislation, has been modified. The modifications include the following prescriptions: • • •

The “noise reduction” should be corrected for differences in rail roughness of the reference and test sections; The “noise reduction” should be representative for “average” rail roughness; The “noise reduction” should be averaged based on measurements for at least 2 types of trains.

References [1] Appendix to the Calculation and measurement guideline Noise abatement act (in Dutch: Bijlage IV Behorende bij hoofdstuk 4 Spoorweg van het Reken- en meetvoorschrift geluidhinder 2006), CROW/Ministry of Environment and Spatial Planning (January 2007) [2] Technical regulation measurement.methods for emission of railway transport 2006 (in Dutch: Technische Regeling Emissiemeetmethoden Railverkeer 2006), CROW/Ministry of Environment and Spatial Planning (December 21, 2006) [3] Janssen, G., van Tol, P.F.: Rail grinding and damping, alternatives to barriers along railway tracks (in Dutch), IPG-projects 2.2.1 and 2.2.2. IPG report ProRail (May 2005)

Theoretical Study on Noise Reduction of Rail Component by Use of Rail Absorber T.X. Wu State Key Laboratory of Mechanical System and Vibration, Shanghai Jiao Tong University 800 Dong Chuan Road, Shanghai 200240, P.R. China Tel.: +86 21 34206332 Ext. 819; Fax: +86 21 34206006 [email protected]

Summary Compound track model with the rail absorber is developed for investigation of the performance of the absorber on reduction of the rail component of rolling noise. The noise radiated from the track with the absorber due to the roughness is predicted. Different types and installation positions of the rail absorber are studied for their effectiveness in terms of the vibration receptance, decay rate, overall response and sound power radiated. The predictions show that the rail radiated noise can be reduced by about 6 dB(A) by use of the rail absorber. The most effective installation position for the discrete absorber is at mid-spans of sleeper bay. As the rail vibration amplitude is not large at sleepers, they are not appropriate places for the discrete absorber to dissipate vibration energy.

1 Introduction The predominant source of noise from railway is associated with the rolling of the wheel on the rail. The roughness on the wheel and rail tread forms the excitation and causes vibration of the wheel and track. When vibration propagates in the wheel and rail, the structure radiates noise. The sound radiated by the rail dominates in the middle frequency range 500-1500 Hz. The main parameter with the strongest influence on the amount of noise radiated by the rail is the rate of decay of vibration along the rail, usually expressed in dB/m [1]. The rail pad stiffness is of particular importance for the decay rate and the track component of noise. Theoretical studies have shown that stiff pads increase the decay rate and thus decrease the rail noise [2]. However, soft pads have become commonly used to reduce track forces and thus damage to the track components. Alternative ways adding damping to the rail vibration have been developed. A promising means to increase the decay rate of vibration along the rail is using a rail damper. A reduction of 6 dB in track component of rolling noise was achieved in the tests by using a two-layer tuned absorber system attached continuously to the rail on each side [3]. The main benefit of such a rail absorber/damper is to allow soft rail pads to be used for reducing dynamic loads to the track components, without any increase in noise levels. A so-called “double tuned rail damper” was developed to suppress the pinned–pinned resonance, which is mounted between sleepers on the rail B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 384–391, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Theoretical Study on Noise Reduction of Rail Component by Use of Rail Absorber

385

[4]. Measurements showed that the decay rate of rail vibration can effectively be improved by the rail damper in the frequency range 500–2500 Hz. In this study the effectiveness of the rail absorber/damper on noise reduction is studied through detailed investigations into the dynamic properties of the track and sound radiation from the track with the continuous and discrete rail absorber installed at different positions. The results show that the rail radiated noise can be reduced by about 6 dB(A) by use of the rail absorber. The most effective installation position for the discrete absorber is at mid-spans of sleeper bay because the vibration amplitude there is large and the pinned-pinned resonance can be suppressed.

2 Modelling of Track: Continuous Rail Absorber The rail absorber studied consists of the steel mass bars or blocks and elastomeric material layers. The elastomeric layers are glued to the lower part of the rail web and the upper surface of the rail foot, as shown in Fig. 1. The absorber is attached to each side of the web and works as a vibration system of two degrees of freedom, provided that the length of the mass bar is short. The parameters of the absorber should be selected such that at the second resonance the bottom mass of the absorber, m1, oscillates with large amplitude, whereas the vibration amplitude of the upper mass, m2, is small compared with m1, and even smaller than the rail vibration. In this way the rail vibration energy can effectively be dissipated by the elastomeric layers. The absorber’s second resonance frequency is designed to be that at which the rail vibration and radiation reach maximum. z

Feiωt m2

x

k3

−∞ m2 m1

Fig. 1. Rail absorber cross-section

rail

k2 m1 k1 + ∞ kp ms kb

Fig. 2. Track model for vertical vibration with rail absorber

To have the required resonance frequency, theoretically, different sets of parameter can be chosen for the mass and stiffness of the absorber. However, they may leads to different effects on energy reduction of rail vibration. A larger active mass of the absorber results in higher decay rate of rail vibration. However, it is confined by the geometrical shape and size of the rail cross-section, where the absorber is installed. The damping of the elastomeric material is important for the rail vibration attenuation and, appropriate damping loss factor should be chosen for the absorber. Fig. 2 shows the track model with the rail absorber for vertical vibration. The rail is modelled as an infinite Timoshenko beam on the discrete supports composed by the

386

T.X. Wu

rail pad, sleeper and ballast. The upper parts above the rail represent the continuous absorber that is modelled as the multiple spring-mass layers along the rail, where k1 represents the stiffness of the elastomeric layer between m1 and the rail, k2 represents the stiffness of the elastomeric connection between m1 and m2, and k3 is the elastomeric connection between m2 and the rail. k1, k2 and k3 are all complex with loss factor ηa. The two absorbers are apart on both sides of the rail web and can be designed to have different natural frequencies, but they are devised here to have the same parameters for simplicity. Treating the absorber attached to the rail as a continuous dynamic stiffness composed of the multiple spring-mass layers, the equations of motion in the frequency domain for the rail is given by N

− ρ Aω 2u + GAκ (φ ′ − u′′) + ka u + ∑ K s uδ ( z − zn ) = Fδ ( z ) ,

(1)

n =1

− ρ I ω 2φ + GAκ (φ − u′) − EI φ ′′ = 0 ,

(2)

where u is the vertical amplitude of rail vibration, F is the amplitude of the harmonic load to the rail at z = 0, N is the number of the rail support considered in the model, zn represents the sleeper position. ka is the dynamic stiffness of the multiple spring-mass layers, determined by the absorber parameters k1, k2, k3, m1 and m2. Ks is the dynamic stiffness of the discrete support of the rail, determined by Ms, Kp and Kb, the sleeper mass, pad and ballast complex stiffness respectively. Table 1. Parameters of the track Bending stiffness EI

Shear stiffness GA

Density ρA

Shear parameter κ

6.42×106 (Nm2)

5.92×105 (N)

60.4 (kg/m)

0.4

0.6 (m)

Pad stiffness Kp

Loss factor ηp

Ballast stiffness Kb

Loss factor ηb

Sleeper mass Ms

200 (MN/m)

0.2

50 (MN/m)

1.0

162 (kg)

Sleeper spacing d

The parameters used in calculations of the track dynamics with the absorber are listed in Table 1. The parameters for the absorber are: m1 = 14kg/m, m2 = 6kg/m, k1 = 260MN/m2, k2 = 12MN/m2,

k3 = k2/4,

ηa = 0.25.

By using the above parameters for the rail absorber its first and second resonance frequencies are calculated to be at about 250 Hz and 700 Hz respectively. The rail displacement amplitude u can be solved using the concept of cross receptance of the rail with the absorber. The cross receptance between two points is defined as the displacement response at one point caused by a unit excitation at the other point. After the rail vibration displacement being calculated, the displacements of the absorber masses and sleepers can be obtained. The calculated dynamic properties of the discretely supported track with the continuous absorber are shown in Fig. 3 for the harmonic load applied at mid-span in terms of the point receptance and decay rate of rail vibration along the track. Comparisons are made in Fig. 3 to the track without the rail absorber. The pinned-pinned resonance appears at about 1 kHz for the track without the absorber, and is effectively suppressed by use of the rail absorber. The decay rate of rail vibration is estimated

Theoretical Study on Noise Reduction of Rail Component by Use of Rail Absorber

387

20

10

10

-8

Decay rate (dB/m)

Receptance (m/N)

10

-9

15

10

5

-10

2

10

0

3

10 Frequency (Hz)

10

2

3

10 Frequency (Hz)

Fig. 3. Receptance and decay rate of rail vibration, excitation at mid-span: — with rail absorber; ⋅⋅⋅⋅⋅⋅ without absorber

over ten sleeper spans from the excitation point, and can be seen to be moderate at low frequencies and to reach a peak of 17 dB/m at about 230 Hz. At high frequencies above 800 Hz the decay rate is generally lower than 1 dB/m if no rail absorber is used. Using the rail absorber, the decay rate of rail vibration is significantly increased in the frequency range 600-1000 Hz, with a local peak over 10 dB/m around the second resonance frequency of the absorber. This is helpful to reduce rolling noise when the rail component dominates in that frequency band.

3 Modelling of Track: Discrete Rail Absorber The absorber can be devised to be either continuous or discrete attached to the rail. Two places of installation are considered for the discrete absorber, at mid-spans of sleeper bay and above sleepers. The model of the discrete track with the discrete absorber is developed in the base of an infinite Timoshenko beam for the rail, with both the rail supports and the absorbers being replaced by the discrete forces to the rail. The equation of motion for the infinite rail is given by M

N

m =1

n =1

− ρ Aω 2u + GAκ (φ ′ − u′′) + ∑ K a uδ ( z − zm ) + ∑ K s uδ ( z − zn ) = F δ ( z ) ,

− ρ I ω 2φ + GAκ (φ − u′) − EI φ ′′ = 0 ,

(3) (4)

where M and N are the number of the discrete absorbers and supports considered, respectively, zm is the installation position of the absorber that is either at mid-spans or above sleepers, zn is the sleeper position, Ka is the dynamic stiffness of the discrete absorber and Ks is the dynamic stiffness of the discrete rail support. The parameters of the discrete absorber, M1, M2, K1, K2, and K3, are obtained by multiplying the parameters of the continuous absorber, m1, m2, k1, k2, and k3, respectively, by the sleeper span length d. Using the concept of cross receptance of the rail (infinite Timoshenko beam), the rail vibration displacement amplitude u can be solved.

388

T.X. Wu

Fig. 4 shows the calculation results for the track using the discrete rail absorber arranged at sleepers and mid-spans. The decay rate for the absorber at sleepers can be seen to be lower by about 3 dB/m at the second resonance of the absorber, compared with the absorber at mid-spans, although above the pinned-pinned resonance it is higher by about 3 dB/m. The receptance peak at the pinned-pinned resonance can be seen from Fig. 4 not to be suppressed by the discrete absorber at sleepers. This is because at the pinned-pinned resonance the rail vibration nodes are at sleepers, where the vibration amplitude is very low, and the rail absorber there can not function well as an energy consumer. These results indicate that it is not a right place at sleeper for the discrete absorber. 20

10

10

-8

Decay rate (dB/m)

Receptance (m/N)

10

-9

15

10

5

-10

10

2

0

3

10 Frequency (Hz)

10

2

3

10 Frequency (Hz)

Fig. 4. Vibration receptance and decay rate of the rail with discrete absorber, excitation at midspan: — absorber at sleepers; ⋅⋅⋅⋅⋅⋅ absorber at mid-spans

4 Wheel/Track Interaction Suitable models for the prediction of structural response and sound radiation of wheel and track are available within the TWINS models [5]. For a random roughness excitation R(ω), considering the interaction only in the vertical direction, the wheel/rail contact force F(ω) is given by

F (ω ) = −

R(ω ) , α (ω ) + α C (ω ) + α R (ω ) W

(5)

where αW, αC and αR are the receptance of the wheel, contact spring and rail (with the absorber) at the contact point respectively. The contact stiffness kC is chosen to be 1.14MN/mm, and αC=1/kC. The wheel model used is composed of a mass Mw (all the un-sprung mass) and a modal spring KM representing the average stiffness of high frequency modes. The wheel receptance αW is given by

α W = 1 K M − 1 ( M wω 2 ) .

(6)

Theoretical Study on Noise Reduction of Rail Component by Use of Rail Absorber

389

8 Roughness spectrum (um)

The parameters are chosen from a UIC 920 mm standard freight wheel with Mw = 600 kg and KM = 4.39 MN/mm. Roughness is present on both the rail and the wheel contact surface and forms a relative displacement excitation that can be regarded as a broadband random process. Fig. 5 shows a typical roughness spectrum. This corresponds to the roughness of a wheel with cast-iron block brakes on a smooth rail [6], and the frequencies correspond to a train speed of 100 km/h.

6

4

2

0

2

10

3

10 Frequency (Hz)

Fig. 5. Roughness spectrum

5 Noise Radiation from Track The vibration response of the track including the rail, absorber and sleeper to the roughness excitation can be calculated using the track models developed. As an intermediate step to calculating the noise from the track, the overall response of the track is obtained as the sum of the squared vibration amplitude over the length of the track. It is evaluated by Σ|v|2Δz, where v is the amplitude of vibration velocity and Δz is the length of a small track section, chosen to be Δz = d/4. The radiation ratios of the rail and sleeper are taken from the models included in the TWINS software [5]. Although the absorber is now attached to the rail and its radiation ratio may be altered to some extent, the rail radiation ratio from TWINS is still used for the sound power calculation. As the vibration velocity of the upper mass of the absorber is different from the rail and, the overall response of rail vibration, Σ|v|2Δz, consists of the contributions from both the absorber mass and the rail, when calculating Σ|v|2Δz for the rail with the continuous absorber, a sixth contribution is from the upper mass and five-sixths from the rail, whereas for the discrete absorber a twelfth contribution is from the upper mass and eleven-twelfths from the rail. rail: 110

10

10

10

SPL (dB re 10–12 W)

Σ|v|2Δz (m3/s2)

10

— 107.2dB(A)

-3

-4

-5

⋅⋅⋅⋅⋅⋅113.7dB(A)

100

sleeper:

90

– – 90.9dB(A)

80

⋅⋅⋅⋅⋅⋅ 93.3dB(A)

70 60

-6 2

10

3

10 Frequency (Hz)

10

2

3

10 Frequency (Hz)

Fig. 6. Σ|v|2Δz and sound power level of the track with and without continuous absorber at speed 100 km/h, excitation at mid-span: — rail; – ⋅ – upper mass of absorber; – – sleeper; ⋅⋅⋅⋅⋅⋅ for the corresponding results without rail absorber

390

T.X. Wu

Fig. 6 shows the radiated sound power spectra from the rail and sleepers in onethird octave bands due to the roughness excitation from Fig. 5 for the track with and without the continuous absorber. Also shown in Fig. 6 is the overall vibration response of the track given by Σ|v|2Δz. It can be seen from Fig. 6 that the sound power radiated from the track is related to its overall vibration response Σ|v|2Δz. The rail absorber works well as an energy consuming facility in the frequency region 500-1200 Hz, where the rail component of rolling noise dominates and reaches maximum. The vibration response of the upper mass of the absorber can be seen to be much lower at high frequencies, but higher at low frequencies, compared with the rail response. Theoretically, use of the continuous rail absorber can reduce the rail component of rolling noise by about 6.0-6.5 dB(A) and the sleeper component by about 2.4 dB(A). In field measurements [3] a reduction by 6.0 to 6.2 dB(A) for the track noise at a speed of 100 km/h is achieved using the rail absorber with 17.5 kg/m active mass. The theoretical predictions are consistent with the field measurements.

10

10

10

-3

SPL (dB re 10–12 W)

Σ|v|2Δz (m3/s2)

10

rail:

110

(a)

-4

-5

— 109.0dB(A) ⋅⋅⋅⋅⋅⋅107.6dB(A)

100

sleeper:

90

– – 91.1dB(A)

80

⋅⋅⋅⋅⋅⋅ 90.6dB(A)

70 60

-6 2

10

3

10 Frequency (Hz)

10

2

3

10 Frequency (Hz)

Fig. 7. Σ|v|2Δz and sound power level of the track with discrete absorber installed at sleepers, excitation at mid-span: — rail; – ⋅ – upper mass of absorber; – – sleeper; ⋅⋅⋅⋅⋅⋅ for the corresponding results with absorber at mid-spans

Fig. 7 shows the results from the track using the discrete rail absorber. Since the rail vibration amplitude at sleepers is relatively low, the absorber there can not dissipate much energy of the rail vibration, and thus the pinned-pinned resonance cannot effectively be suppressed. As a result, the rail radiated sound is higher by 1.4 dB(A) than that with the absorber installed at mid-spans, where the vibration amplitude is high and the absorber works effectively at the pinned-pinned resonance.

6 Conclusions Predictions show that the rail radiated noise can be reduced by about 6 dB(A) using the rail absorber. The effectiveness of the continuous and discrete absorber is quite similar. The most effective installation position for the discrete absorber is at midspans of sleeper bay as the vibration response there is higher and the pinned-pinned resonance can be damped.

Theoretical Study on Noise Reduction of Rail Component by Use of Rail Absorber

391

References [1] Jones, C.J.C., Thompson, D.J., Diehl, R.J.: The use of decay rates to analyse the performance of railway track in rolling noise generation. Journal of Sound and Vibration 293, 485–495 (2006) [2] Vincent, N., Bouvet, P., Thompson, D.J., Gautier, P.E.: Theoretical optimization of track components to reduce rolling noise. Journal of Sound and Vibration 193, 161–171 (1996) [3] Thompson, D.J., Jones, C.J.C., Waters, T.P., Farrington, D.: A tuned damping device for reducing noise from railway track. Applied Acoustics 68, 43–57 (2007) [4] Maes, J., Sol, H.: A double tuned rail damper-increased damping at the two first pinnedpinned frequencies. Journal of Sound and Vibration 267, 721–737 (2003) [5] Thompson, D.J., Janssens, M.H.A.: TWINS: Track-wheel interaction noise software, theoretical manual, version 2.4. TNO report TPD-HAG-RPT-93-0214, TNO Institute of Applied Physics (1997) [6] Dings, P.C., Dittrich, M.G.: Roughness on Dutch railway wheels and rails. Journal of Sound and Vibration 193, 103–112 (1996)

Reducing Wheel-Rail Interaction Forces and Roughness Growth by Application of Rail Dampers* B.E. Croft, C.J.C. Jones, and D.J. Thompson Institute of Sound and Vibration Research, University of Southampton, Southampton SO17 1BJ, UK Tel.: +44 23 8059 2936; Fax: +44 23 8059 3190 [email protected]

Summary High roughness growth occurs in situations with stiff vertical structural dynamics of the track. In particular the antiresonance above a sleeper at the pinned-pinned frequency has been identified as a wavelength fixing mechanism for short pitch corrugation. Rail damping devices (developed to reduce the noise from railway tracks) change the dynamic response of the rail, shifting the pinned-pinned frequency and smoothing the track receptance. Here, a simple time-stepping model is applied to calculate the interaction forces between wheel and rail for a track with and without rail dampers. The calculations show that rail dampers reduce dynamic interaction forces and shift the force spectrum to longer wavelengths. The interaction forces are used to predict the roughness growth after many wheel passages. Track without rail dampers is predicted to develop corrugation at the wavelength corresponding to the pinnedpinned frequency. With rail dampers the corrugation growth is reduced and shifted to a longer wavelength where its significance is diminished.

1 Introduction Railway rolling noise is induced by the roughness of the wheels and rails. Theoretical models such as TWINS exist to predict the noise resulting from a given roughness input, and these allow different wheel and track designs to be compared. However these models do not account for changes in roughness over time or assess the effect that a change in design may have on the propensity of the system to develop roughness. For a given roughness level, the noise from the track can be reduced by increasing the damping of the track and by acoustic optimisation of track components. Over time, the noise level is dependent on both the noise for a given roughness and the rate of roughness growth. Therefore parameters which influence the rate of degradation of the railhead must be assessed. The addition of rail dampers (a system of tuned absorbers) affects the dynamic response of the track structure to the time-varying wheel-rail interaction. Although it is known that rail damping reduces noise [1], the resulting change in the track dynamics may also affect the roughness or corrugation growth rate (‘corrugation’ is a periodic wear that appears as a peak in the more general ‘roughness’ spectrum). *

Awarded Best Paper of IWRN9.

B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 392–398, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Reducing Wheel-Rail Interaction Forces and Roughness Growth

393

The first step towards predicting the change in roughness over time is to model the system to calculate the dynamic interaction forces. These forces are input into a wear model to predict the change in roughness over time. Commonly, frictional abrasive wear is assumed [2,3,4,5], with the amount of material removed being proportional to the work done by the tangential stress in the slip zone of the contact patch. Alternative wear mechanisms including plastic deformation [6] and ratchetting wear [7] have been proposed, and it is likely that multiple wear mechanisms exist, or that different wear mechanisms dominate at different sites. Despite the uncertainty over the mechanism of roughness growth, the wavelength fixing mechanism of short pitch corrugations is well understood [8]. It is generally accepted that the wavelength of short pitch corrugations is determined by resonances in the coupled train-track system, in particular the pinned-pinned resonance of the rail, other vertical antiresonances in the track, and system resonances caused by wave reflections between the wheels. Hempelmann and Knothe [3] developed a linear model for the prediction of short pitch corrugation, obtaining an exponential growth law for corrugation formation. They compared frequency dependent corrugation growth rates at certain positions in a sleeper bay with the track receptance. Their results predict higher corrugation growth rates in situations with stiff vertical structural dynamics of the track. This occurs where the rail has an antiresonance. Hiensch et al. [4] and Nielsen [2] have suggested remedying high roughness growth by increasing the track receptance above the sleepers at the pinned-pinned frequency, by adding damping to the rails. The idea is to smooth the minima in the track receptance and hence achieve lower interaction forces. The present work demonstrates that rail dampers developed to control noise radiation from the track also have an effect on the track dynamics, resulting in reduced interaction forces and hence in reduced roughness or corrugation growth. The work is being carried out in parallel with measurement tests near Gersthofen in Germany as part of the EU project ‘Silence’.

2 Model Description and Input Parameters 2.1 Finite Element Model of Track and Vehicle A finite element model of 60 sleeper bays of UIC60 track has been set up. This uses four Timoshenko beam elements per sleeper bay. Half the track only is considered, i.e. a single rail on half sleepers. Track parameters are chosen to match the Silence project test site. The half-sleepers are modelled as 147 kg masses equally spaced at 0.6 m intervals, while rail pads and ballast are modelled as spring – damper sets. The pad stiffness is 2x108 N/m, the pad damping 5x103 Ns/m, the ballast stiffness 5x107 N/m and the ballast damping 1x105 Ns/m. The track ends are constrained in displacement and rotation. The degrees of freedom of the model allow displacement in the vertical direction and rotation in the vertical plane. Lateral effects are not included. This model is based on that developed by Nielsen et al., [2, 9]. The vehicle is modelled by two uncoupled wheel masses, each linked to the rail by a Hertzian contact spring of stiffness 1.4x109 N/m. An external static force represents the sprung vehicle mass. The wheelbase and wheel loads vary with each train type. Addition of rail dampers is achieved by adding additional elements to the finite element model. The dampers are modelled with a single lumped mass and a tuning

394

B.E. Croft, C.J.C. Jones, and D.J. Thompson

frequency of 800 Hz. To represent the properties of the elastomeric material, a combination of Maxwell elements in parallel with a spring is used as shown in Fig. 1. In practice, implementing Maxwell elements in a finite element code results in a singularity in the absence of any mass at the node between each spring and damping element. To circumvent this difficulty, small additional masses (with natural frequencies well above the range of interest) are included at each of the intermediate nodes.

Fig. 1. Representation of a rail damper with a single lumped mass on spring and damper elements

The equivalent frequency dependent loss factor and stiffness of the damper modelled in this way are shown in Fig. 2. In practice the stiffness of the elastomer also increases with frequency, consequently the model can be made (as here) a better representation of stiffness than would be achieved by using a single frequencyindependent value. 10

Stiffness [N/m]

Loss Factor

10 0.4 0.2 0 (a)

9

10

8

2

10

3

10 Frequency [Hz]

4

10

10 (b)

2

10

3

10 Frequency [Hz]

4

10

Fig. 2. Equivalent frequency dependent loss factor and stiffness of finite element model of rail dampers tuned to a natural frequency of 800 Hz: (a) loss factor, (b) stiffness. ⎯⎯ as modelled; ⋅⋅⋅⋅⋅⋅ nominal value.

Global mass, stiffness and damping matrices are assembled and the modes of vibration of the track system are calculated using standard finite element methods. The frequency range of interest is limited so to reduce calculation times only modes up to 3000 Hz are included. The track receptance over a sleeper and at mid-span are calculated from a modal analysis based on the finite element model and are shown in Fig. 3 for the track with and without rail dampers. The finite element model parameters were tuned to those of the test site by comparing calculated decay rates with measured track decay rates as shown in Fig. 4. From Fig. 4(a) it can be seen that the decay rate characteristic of the track model without rail dampers is a good match with the measurements. The rise in decay rate in the measurement towards 4 kHz is due to interaction between the rail pad and the first mode of cross-sectional deformation of the rail. This is not represented in the model.

Reducing Wheel-Rail Interaction Forces and Roughness Growth

395

Receptance m/N

Receptance m/N

The effect of the rail dampers on the track decay rate is clear in Fig. 4(b) but it is evident that the simple model of the damper used here does not fully represent the effect of the actual rail damper at all frequencies. This is to be expected as each rail damper is made up of two beams (with two tuning frequencies) rather than a single lumped mass as modelled. Consequently the model with the rail dampers does not have a uniformly high decay rate as measured. Nevertheless the effect of the dampers on the track dynamics around the pinned-pinned frequency is well represented.

-8

10

-10

10

2

-10

10

3

10

(a)

-8

10

10

2

3

10

(b)

Frequency [Hz]

10 Frequency [Hz]

Decay rate dB/m

Decay rate dB/m

Fig. 3. Predicted change in track vertical receptance with the addition of rail dampers: (a) above sleeper, (b) mid-span. ⎯⎯ without rail dampers; ⋅⋅⋅⋅⋅⋅ with rail dampers.

1

10

0

10

2

(a)

10

3

10 Frequency [Hz]

1

10

0

10

(b)

2

10

3

10 Frequency [Hz]

Fig. 4. Predicted and measured change in track vertical decay rate one third octave spectrum with the addition of rail dampers: (a) without rail dampers, (b) with rail dampers. ⎯⎯ from model; ⋅⋅⋅⋅⋅⋅ from measurements at Gersthofen.

2.2 Calculation of Wheel-Track Interaction Force The modal forms of the equations of motion for the track, vehicle and their interaction are solved as a state space system using a standard time-stepping routine with variable step size in similar fashion to the technique used by Nielsen and Igeland [9]. An initial low-level roughness profile has been generated from a sum of sine functions with random phase to match the measured roughness spectrum. The interaction forces between the wheels and the rail are determined as the wheels move along the model of the track, taking account of the filtering effect of the contact patch [10]. The traffic at the Gersthofen site is mixed. Three train types are modelled here: freight, regional and ICE trains, using typical parameters for each. The wheel radius for all vehicles is 0.46 m.

396

B.E. Croft, C.J.C. Jones, and D.J. Thompson

Table 1. Train parameters and wavelengths corresponding to the pinned-pinned frequencies Parameter Train wheel velocity Unsprung wheel mass

Freight Train

Regional Train

ICE Train

29.44 m/s

37.78 m/s

43.06 m/s

488.5 kg

702.5 kg

782 kg

100 x103 N

60 x103 N

60 x103 N

1.8 m

2.5 m

2.5 m

Pinned-pinned wavelength without rail dampers

0.027 m

0.034 m

0.039 m

Pinned-pinned wavelength with rail dampers

0.043 m

0.055 m

0.062 m

Static load on wheel Wheel spacing

70

60

60

50 40 30 20

(a)

-1

50 40 30 20

-2

10 10 Wavelength [m]

(b)

70 Force [dB re 1N]

70 Force [dB re 1N]

Force [dB re 1N]

The interaction force spectra are calculated for each of the two wheels in the vehicle model for each train type. Results shown in Fig.5 are for the first wheel only although small differences are present between the two wheels. Different wavelengths are dominant for each of the train types due to their differing average speeds. Wavelength is shown rather than frequency to allow easier comparison with the predictions of roughness growth. Fig. 5 shows the effect of the dampers on the interaction force. The track receptance (Fig. 3) shows a shift in the pinned-pinned frequency corresponding to the shift in the force spectrum to longer wavelengths with the dampers applied. In each case, the dynamic interaction force in the one-third octave wavelength band corresponding to the pinned-pinned frequency is reduced. A reduction in the peak dynamic interaction force of 2-3 dB is observed for all train types (this peak does not necessarily occur at the pinned-pinned wavelength). For all cases the interaction forces at short wavelengths are low, due to the filtering effect of the contact patch.

60 50 40 30 20

-1

-2

10 10 Wavelength [m]

(c)

-1

-2

10 10 Wavelength [m]

Fig. 5. Predicted change in wheel-track dynamic vertical interaction force with the addition of rail dampers: (a) typical freight train, (b) typical regional train, (c) typical ICE train. ⎯⎯ without rail dampers; ⋅⋅⋅⋅⋅⋅ with rail dampers.

2.3 Wear Model to Predict Rail Roughness Growth The wear is calculated over the middle ten sleeper bays of the model. The force at each wheel as they pass over these sleeper bays is extracted from the results of the time-stepping calculation and used in conjunction with the initial rail profile in those sleeper bays to predict the wear of the rail after a large number of wheel passages. Wear is assumed to be proportional to the work done by the tangential stress in the slip zone of the contact patch, with a wear constant of 2.5x10-9 kg/Nm [2]. The wear model is two dimensional as in [5], simplifying the contact patch to a rectangle of constant width. The longitudinal creepage is 0.001. There is no variation in forces or

Reducing Wheel-Rail Interaction Forces and Roughness Growth

397

(a)

0 -5 -10 0.25 0.125 0.063 0.0315 0.016 1/3 Oct Wavelength (m)

(b)

5 0 -5 -10 0.25 0.125 0.063 0.0315 0.016 1/3 Oct Wavelength (m)

Roughness (dB re 1um)

5

Roughness (dB re 1um)

Roughness (dB re 1um)

roughness growth across the rail head. In each time interval the length of the Hertzian contact patch is determined, and this contact patch is divided into a stick zone and a slip zone. The frictional work and hence the wear are calculated in the slip zone, and this wear accumulates over many wheel passages to form a revised rail profile. As the wear from a single wheel passage is very small it is assumed that any resulting change in interaction forces is also very small. The wear after each passage may then be multiplied to simulate thousands of wheel passages, without recalculating the contact forces [2]. In this work a factor of 1x105 is used.

(c)

5 0 -5 -10 0.25 0.125 0.063 0.0315 0.016 1/3 Oct Wavelength (m)

Fig. 6. Predicted roughness spectrum with the addition of rail dampers compared to the roughness developed for an undamped track: (a) typical freight train, (b) typical regional train, (c) typical ICE train. ⎯⎯ initial roughness level; − − − final roughness level without rail dampers; ⋅⋅⋅⋅⋅⋅ final roughness level with rail dampers; − ⋅ − ⋅ TSI+ limit spectrum.

The initial roughness level is very low. The model predicts that after 1x105 wheel passages, roughness of wavelengths between 0.01 m and 0.1 m will grow significantly. Shorter wavelengths are not expected to grow. Significant roughness growth at longer wavelengths is not indicated by the model. Each train type typically runs at a different speed, and the model results show that the slower freight trains encourage roughness growth at shorter wavelengths than the faster ICE trains. For the freight and ICE cases without rail dampers there is a particular wavelength at which a peak in the roughness spectrum develops, showing a tendency to corrugation which corresponds to the pinned-pinned frequency. This effect is not so clear for the regional train case which displays a broader peak of interaction force and hence roughness growth between 0.02 m and 0.05 m. With the rail dampers applied, the predicted roughness spectrum in all cases displays a shift to longer wavelengths reflecting the shift in the pinned-pinned frequency. At the original pinned-pinned frequency of the track, the roughness is not predicted to grow significantly. It can be seen that the peak roughness level is reduced only slightly by the dampers. This is partly because longer wavelength roughness tends to have higher amplitudes anyway. However, the shift in the corrugation wavelength when the rail dampers are applied increases the difference between the predicted roughness level and the TSI+ limit spectrum.

3 Conclusions Rail dampers are shown to shift the pinned-pinned frequency of the track and to smooth the peaks and troughs in the track receptance. This work predicts a decrease in

398

B.E. Croft, C.J.C. Jones, and D.J. Thompson

interaction forces around the pinned-pinned frequency and a shift in the interaction force spectrum to lower frequencies. The resulting roughness growth predicted after many wheel passages using a simple model supports the theory that corrugation wavelength is fixed by the pinned-pinned frequency, and suggests that a reduction in roughness growth compared to initial levels (or at least a shift in a corrugation wavelength to a longer wavelength) may be achieved by the application of rail dampers.

Acknowledgements This work has been undertaken as part of the EU project ‘Silence’. The authors appreciate the assistance provided by DBAG and Corus.

References [1] Thompson, D.J., Jones, C.J.C., Waters, T.P., Farrington, D.: A tuned damping device for reducing noise from railway track. Applied Acoustics 68, 43–57 (2007) [2] Nielsen, J.C.O.: Numerical prediction of rail roughness growth on tangent railway tracks. Journal of Sound and Vibration 267, 537–548 (2003) [3] Hempelmann, K., Knothe, K.: An extended linear model for the prediction of short pitch corrugation. Wear 191, 161–169 (1996) [4] Hiensch, M., Nielsen, J.C.O., Verheijen, E.: Rail corrugation in the netherlands - measurements and simulations. Wear 253, 140–149 (2002) [5] Sheng, X., Thompson, D.J., Jones, C.J.C., Xie, G., Iwnicki, S.D., Allen, P., Hsu, S.S.: Simulations of roughness initiation and growth on railway rails. Journal of Sound and Vibration 293, 819–829 (2006) [6] Bohmer, A., Klimpel, T.: Plastic deformation of corrugated rails - a numerical approach using material data of rail steel. Wear 253, 150–161 (2002) [7] Franklin, F.J., Chung, T., Kapoor, A.: Ratcheting and fatigue-led wear in rail-wheel contact. Fatigue and Fracture of Engineering Materials and Structures 26, 949–955 (2003) [8] Grassie, S.L.: Rail corrugation: Advances in measurement, understanding and treatment. Wear 258, 1224–1234 (2005) [9] Nielsen, J.C.O., Igeland, A.: Vertical dynamic interaction between train and track - influence of wheel and track imperfections. Journal of Sound and Vibration 187, 825–839 (1995) [10] Ford, R.A.J., Thompson, D.J.: Simplified contact filters in wheel/rail noise prediction. Journal of Sound and Vibration 293, 807–818 (2006)

Mitigation of Wheel Squeal and Flanging Noise on the Australian Rail Network D. Anderson1 and N. Wheatley2 1

Rail Corporation NSW, Level 4, 18 Lee Street, Chippendale, 2008 Sydney NSW Australia [email protected] 2 Queensland Rail, 305 Edward St, GPO Box 1429, Brisbane 4001, Queensland, Australia [email protected]

Summary Standard practice in Australia is to differentiate between the two curve squeal effects, namely tonal “wheel squeal” and the more broad-band (and often intermittent) metalon-metal rubbing noise referred to as “flanging”. This paper presents results from recent curve squeal investigations and mitigation trials on the Australian rail network, covering curves varying in radius (under 200m to over 400m), carrying both passenger and freight traffic on standard and narrow gauge ballast track with either timber or concrete sleepers. Treatments include gauge face lubrication; top of rail friction modification and the maintenance and modification of track and rolling stock. While the results are generally consistent with other research, observations suggest that curve squeal may be governed by two different effects. At some sites, curve squeal appears to be controlled by prevailing friction conditions and treatment of these sites with friction modification can be very effective. But at sites where poor curving performance of certain bogies is the controlling factor, friction modification is ineffective and maintenance of individual items of rolling stock is the only cost effective option. Most curves show a combination of both of these effects.

1 Introduction The prevalence of wheel squeal noise issues has increased over the last 20 years. This is due to heightened community awareness and expectation, coupled with growth in rail transport and residential populations near rail operations. Changes in the design, maintenance and operation of rolling stock and track may also be a contributing factor. In common with rail networks overseas [1], substantial effort has been expended in Australia to attempt to understand and tackle the issue and it remains a high priority. However, in contrast to some research [2], the focus has been on monitoring the effects and evaluating treatment options, rather than laboratory analysis or theoretical modelling. Some aspects of the Australian work in this area may therefore complement that by other researchers, including: • •

The extensive use of wayside angle-of-attack (AoA) detection systems for the specific purpose of studying wheel squeal [3, 4] The development of trackside applicators for top-of-rail friction modification [5, 6]

B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 399–405, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

400

• • • •

D. Anderson and N. Wheatley

The trial of a wide range of mitigation techniques on the Brisbane rail network in Queensland [6] The development of a sophisticated wayside noise detection device to accurately identify axles that generate curve squeal in South Australia [7] The development of techniques to allow unattended monitoring of noise from a very large number of trains, over extensive periods (usually several weeks), and Work on detection and differentiation between squeal and flanging noise characteristics.

2 Differentiation between Squeal and Flanging Noise For some years, Australian practice has been to differentiate between tonal “wheel squeal” and the more broad-band metal-on-metal rubbing noise termed “flanging”. Although sharing some features, there are often differences in community response as well as possible differences in the underlying causes and treatments. Affected members of the community usually express more concern about tonal squeal noise than flanging noise [8, 9], probably due to the “piercing” tonal nature of squeal and the perception that is caused by a fault with the rail system. Rail engineers consider “flanging noise” to arise due to wheel flange contact at the gauge face/gauge corner of the rail. Whether this is true remains to be seen [10], but it is clear that it is gaining recognition as a possible separate effect from squeal [2, 11, 12]. Researchers have discussed the frequency characteristics of the two effects [7, 12, 13], although not all agree on the frequency ranges. Jiang and Dwight [13], suggest the two effects can be differentiated by tonality and report flanging noise as covering the frequency range 1 to 10kHz. Eadie et al [12], suggest that the two effects cover different frequency ranges (<5kHz for squeal, >5kHz for flanging). The table below summarises the differences between flanging and wheel squeal with likely causes. In the remainder of this paper, the terms “wheel squeal” and “flanging” will be used to differentiate between the effects, while “curve squeal” will be used to describe the combination. Characteristic Time Frequency Description Sometimes also termed Normally arising from Suspected cause Dependent on speed or tractive effort? Dependent on curving performance?

Curve squeal effect Wheel squeal Often sustained Tonal High pitched scream “Squeal due to lateral creepage” [11] Low rail Lateral creep (which efficiently excites wheel bending vibration modes)

Flanging Sometimes intermittent Usually broad band, high frequency Hiss, “tisk-tisk”, “ching-ching”, metal on metal sliding / screeching, grating noise “Flange contact” noise [11], “flange rubbing” noise [2] High rail Longitudinal / vertical sliding at wheel flange / gauge face contact

Generally not

Yes [16]

Yes, particularly Angle of Attack

Yes, particularly lateral position

More work is required to clarify the extent to which the treatment of the two effects may differ. Conventional wisdom in the rail industry is that effective gauge face (grease) lubrication is sufficient to mitigate flanging noise [14], although this has been

Mitigation of Wheel Squeal and Flanging Noise on the Australian Rail Network

401

questioned in recent studies [15]. However, what is clear is that squeal noise is not strongly influenced by gauge face lubrication (except when lubricant contaminates the top of the rail).

3 Overview of Australian Experience with Curve Squeal 3.1 New South Wales (NSW) Wheel squeal problems in NSW were first studied in detail at Beecroft (Sydney) in the late 1980’s and at Wollstonecraft (Sydney) in the early 1990’s [17]. Squeal noise levels over 100 dBA were noted at surrounding residential properties (25 to 50 m from the track). Initial efforts failed to establish the cause of the problem, or identify an effective solution, so an in-depth study was carried out in 1996 [3], including measuring angle-of-attack (AoA) in tandem with noise monitoring at Wollstonecraft. This curve is approximately 200m in radius and carries passenger rail traffic. The study found that approximately 5% of axles caused squeal (translating to over 30% of trains) while 0.5% of axles generated flanging noise (not considered further). Average AoA was 11mrad for leading axles and 0.6mrad for trailing axles, but there was no correlation between squeal and AoA or lateral position. The contact band on the high rail was located near the centre of the rail, rather than being approximately 20mm to the gauge side to promote rolling radius differential, but grinding an asymmetrical rail profile (to promote improved steering) did not reduce noise. Squeal effects varied significantly during the study, particularly with changes in humidity. Application of a friction modifier to the top of the low rail eliminated the squeal. Contrary to the supplier’s previous experience, the effectiveness of the high positive friction product disappeared after a few passing axles or trains. An automatic trackside applicator was developed to deliver friction modifier to the running surface of the rail prior to each train [5]. Trackside applicators have been in service at Wollstonecraft since 1997, applying a friction modifier to the top of the rail. Squeal noise complaints are now infrequent here, but complaints continue to increase at other sites, even though friction modifier applicators are installed at many of these locations. Further investigations [19] indicate freight trains are the main squeal noise source in NSW and that friction modifier is only partially effective at mitigation (Section 4). 3.2 Queensland Powell [6] and Nelson [8] describe the extensive investigations, trials and treatments carried out to combat wheel squeal on the Queensland Rail network in Brisbane between 1997 and 2000. Squeal noise levels of over 110dBA had been measured at affected residential locations. Application of friction modifier to the top of the rail was successful. Trials included application in liquid form by hand and both solid and liquid form by hi-rail vehicle. Ongoing application involved the development of purpose-designed trackside applicators. Issues encountered during the trials included rain washing friction modifier from the rail and wind blowing the dried product during hot dry weather. Water

402

D. Anderson and N. Wheatley

spray is also successfully used in a yard location, although it causes some issues with weed growth, corrosion and ballast degradation. Rail grinding asymmetric rail profiles to improve steering did not provide benefit. Poor steering by certain freight wagons was traced to ineffective centre-bowl lubrication; improved lubrication eradicated squeal from these wagons. Investigations of the influence of wheel and rail hardness were not conclusive, nor were trials involving altering gauge. Fin and ring dampers fitted to some passenger rolling stock have been successful in reducing squeal. Overall, Queensland Rail developed an effective strategy to manage wheel squeal and current work aims at optimising these techniques to ensure cost-effectiveness (section 4). 3.3 South Australia (Adelaide Hills) The Australian interstate freight network traverses the Adelaide Hills in South Australia via numerous 200 m radius curves on relatively steep grades. It is also an expanding residential area. The Australian Rail Track Corporation (ARTC) is manager of this network and has been addressing wheel squeal and flanging noise since 2000 [4],[7]. Trials with a top of rail friction modifier were not successful. Detailed studies in 2000 and 2003 were carried out using AoA detection in tandem with an acoustic detection system (using a microphone array with horizontal resolution to pin point each noisy wheel). These studies identified that less than 4% of the passing axles accounted for the majority of the noise issues and that, of these axles, nearly 90% were also displaying irregular AoA or inter-axle misalignment. This led to the installation of a permanent acoustic detection system (termed “RailSqad” [4]) in October 2005 (Section 4).

4 Recent Developments 4.1 Introduction An Australian research project [13] has developed algorithms for detecting various types of wheel/rail noise, including wheel squeal and flanging. It was developed for on-train monitoring to identify track without effective lubrication, but the same algorithms are now processing wayside noise recordings from simple, low cost, portable battery-powered monitoring systems. This has allowed monitoring for extended periods for minimal cost, which is of particular benefit to studying curve squeal because events can vary randomly between trains, times of day, and meteorological conditions. Numerous studies have now exploited this approach [18], allowing the analysis of thousands of trains at curve sites. 4.2 New South Wales A 320 m radius curve was upgraded from 53 kg/m rail on timber sleepers to 60 kg/m head hardened rail on concrete sleepers. Freight trains were found to have generated wheel squeal on timber sleepers, but not at “severe” levels, whereas wheel squeal increased substantially after the track upgrade, with approximately 40% of freight trains generating “severe” levels.

Mitigation of Wheel Squeal and Flanging Noise on the Australian Rail Network

403

A temporary speed restriction (from 75 km/h to 40 km/h) on a 315 m radius curve (with 60 kg/m head hardened rail on concrete sleepers) gave a reduction in the number of squeal events, but this was countered by the increased duration of remaining events. At the same curve, extended monitoring was carried out over a 4-month period, spanning the installation, upgrade and duplication of a track-side applicator for top of rail friction modifier. 47% of freight trains generated “moderate” or “severe” squeal with no friction modifier, reducing to 35% with friction modifier, 29% with an improved applicator design, and 24% following the installation of an additional applicator at the mid-point of the curve. Monitoring was carried out at a complaint location approximately 25 m from a 240 m radius curve with a 1 in 33 grade. Following the installation of a top of rail applicator, squeal events above 100 dBA were eliminated in coal traffic and reduced by over 50% in container traffic. Flanging noise also reduced considerably, with events above 100 dBA eliminated for all traffic. In 2006 an extensive monitoring program over nearly six months involved five simultaneous recording locations spanning four tight reverse curves (300 to 560 m radius). Two track-side applicators for top of rail friction modifier had been installed at this location in 2001 and the tests were carried out during the installation of two additional units, intended to improve the coverage of product over the 2 km section. Results showed negligible difference in wheel squeal following the installation of additional applicators and gave a very mixed picture when analysed in terms of particular types of freight train. Some categories showed no change, others showed improvement, while others performed worse. Monitoring at a complaint location 33 m from the 300 m radius curve showed that around 20% of freight trains continue to generate noise levels exceeding 105 dBA. 4.3 South Australia The permanent RailSqad installation in the Adelaide Hills has now been operating for over a year. The majority of squeal and flanging noise events are in the region of 95 to 100 dBA, measured at 6m from the low rail; however, some events have exceeded 115 dBA. The system allows detailed analysis of a very large data-set and trends established to date include: • • • • •

A very weak correlation between speed and squeal or flanging noise level, A strong correlation with travel direction, with more squeal and flanging noise occurring in the uphill direction (1:45 grade , bi-directional track), Leading axles of leading bogies consistently have a higher tendency to generate squeal or flanging noise. Certain types of wagon have a higher tendency to generate squeal or flanging noise. Temperature and humidity have almost no influence on the occurrence of the noisiest events, but a strong influence on the occurrence of moderate squeal and flanging. Consistent with experience at Wollstonecraft in NSW and Keperra in Brisbane, moderate squeal noise becomes more prevalent with increasing humidity.

Some wagons identified as “repeat offenders” have been inspected at the workshop and often found with defects that may explain poor curving performance. Most commonly, repairs and adjustments have been made to constant-contact side-bearers,

404

D. Anderson and N. Wheatley

brake rigging, mismatched side frames, worn wedge blocks and broken bogie springs. The process of corrective maintenance of “repeat offenders” is in its infancy but, to date, some classes of rolling stock have started to show improved noise performance at the monitoring site. 4.4 Queensland Trials are underway in Brisbane to assess noise reduction effectiveness of lubrication and friction modification treatments, and to allow refinement of product quantities to optimise costs. At a 212 m radius dual gauge bi-directional curve, the track-side applicators for top of rail friction modifier were temporarily switched off. After 3 days, visual inspection showed that gauge-face lubricant no longer gave good coverage of the gauge corner region, suggesting that top of rail friction modification was providing a positive effect on the coverage of gauge face lubrication. Noise monitoring showed an increase in the incidence of flanging noise. It is not clear whether this is a direct result of the change in top of rail conditions, or an indirect effect due to the corresponding reduction in lubricant at the gauge corner. However, the results do confirm that the top of rail treatment is effective and further trials are now proposed by switching gauge face lubrication off and varying friction modifier application rates.

5 Discussion and Recommendations Classification of squeal and flanging noise as separate effects can be helpful, both in terms of understanding the impact on affected communities and in studying the underlying causes and effectiveness of treatments against curve squeal. Australian experience highlights that it can be necessary to obtain a very large number of curve noise measurements in order to obtain meaningful information on the effects. At some locations, measurements during daytime hours only, for example, can give a completely misleading picture. Australian experience also suggests that curve squeal effects may follow one of two patterns (both of which may be occurring simultaneously at some sites). Some effects appear to be controlled almost exclusively by top of rail friction conditions, and can be very effectively treated by friction management. Wollstonecraft in NSW is an example, where random variation in the effects thwarted initial attempts to understand the issue and, somewhat contrary to expectations, irregular angle of attack data showed almost no correlation with the small percentage of axles that caused squeal. Other effects appear to be almost entirely controlled by the curving performance of a very small proportion of passing axles. Changes in friction (due to weather or the application of friction modifier) may only have a mild influence on squeal, if any. In this scenario, the only cost effective means of treatment may be the detection and treatment of the particular rolling stock items that generate the noise.

Acknowledgements The authors would like to acknowledge the invaluable input and collaboration from Mike Sowden in Adelaide regarding wheel squeal generally, and recent experience with the issue in the Adelaide Hills (South Australia) in particular.

Mitigation of Wheel Squeal and Flanging Noise on the Australian Rail Network

405

The authors are also grateful to Jiandong Jiang of the University of Wollongong for his frequent refinements and improvements to the “TrackSide Noise” processing algorithm used for much of this work; to Sav Shimada of RailCorp for her tireless pursuit of robust analysis of hundreds of freight train noise records; and to Sherrie Meshki of QR for her diligent management of noise monitoring in Brisbane.

References [1] Working Group on Railway Noise, Position Paper on the European Strategies and priorities for railway noise abatement, European Commission, Version 19403 [2] Thompson, D.J., et al.: Project A3 – Railway noise: curve squeal, roughness growth, friction and wear – start up study, Rail Research UK (June 2003) [3] Kerr, M., Kalousek, J., Elliot, G., Mau, F., Anderson, D.: Squeal Appeal: Addressing Noise at the Wheel / Rail Interface. In: Conference on Railway Engineering, Rockhampton (1998) [4] Cowley, A., Kopke, U., Lindqvist, P., Hamilton, R., Rennsion, D., Southern, C., Sowden, M.: Operating the RailSqad wheel-rail noise monitoring system. In: 7th International Conference on Contact Mechanics and Wear of Rail/Wheel Systems (2006) [5] Kerr, M., Lak, A.: Wheel squeal problems solved: a trackside solution. In: RTSAconference (1999) [6] Powell, J.: Wheel squeal noise control at Queensland Rail. ARM wheel/rail interface seminar, Chicago (2001) [7] Kopke, U., Rennison, D., Southern, C.: RailSqad: A wheel/rail noise emission monitoring system. In: 14th International Wheelset Conference (2004) [8] Nelson, J.: A review of wheel/rail noise ameliorative techniques used by Queensland Rail, QR commissioned report (February 2000) [9] Anderson, D.: Personal communication with affected residents at Teralba, NSW (2004) [10] Nelson, J.: Wheel/rail noise control manual, TCRP report 23 (sponsored by US FTA) (1997) [11] Müller, B., Oertli, J.: Combating curve squeal: Monitoring existing applications. Journal of Sound and Vibration 293 (2006), (Proceedings of 8th International Workshop on Railway Noise 2004) [12] Eadie, D., Santoro, M.: Railway noise and the effect of top of rail liquid friction modifiers: changes in sound and vibration spectral distributions. In: 6th Int’l Conf. on Contact Mechanics and Wear of Rail/Wheel Systems (2003) [13] Jiang, J., Dwight, R.: Determining Wheel-Rail Wear Conditions Using Wheel-Rail Noise. In: 7th International Conference on Contact Mechanics and Wear of Rail/Wheel Systems (2006) [14] Larke, R.: Survey of Wheel / Rail Lubrication Practices, RSSB Engineering Research Program (2003) [15] Anderson, D.: personal communication with J Jiang (2007) [16] Anderson, D.: personal communication with S Grassie (2006) [17] Anderson, D.: Wheel Squeal Measurement, Management and Mitigation on the NSW Rail Network, AAS (2004) [18] Fogarty, B., Anderson, D.: Top-Of-Rail Friction Modification for Wheel Squeal Mitigation, Conference on Rail Engineering, Melbourne (2006)

How to Avoid Squeal Noise on Railways State of the Art and Practical Experience S. Bühler and B. Thallemer PROSE Ltd., Zürcherstr. 41, CH - 8400 Winterthur Tel.: +41 (0) 52 262 74 00; Fax: +41 (0) 52 262 74 01 [email protected], [email protected]

Summary Curve squeal noise is noise generated in narrow curves only, producing a sound pressure level more than 10 dB(A) higher than normal rolling noise and which is dominated by pure tones. Squeal noise can occur when the wheel cannot take a position in the track channel with sufficient radius difference to allow the wheel set to roll freely through the curve. Operational requirements such as a minimum friction requirement have to be taken into account too by evaluating measures against squeal noise. Their weight depends on the operating scheme of the railway system. Squeal noise can be inhibited by measures at vehicle level as well as by way side measures. Experience with such measures are presented and provide support to railway operators (infrastructure and rolling stock) in the decision how to tackle squeal noise effectively and which parameters have to be focussed on, to ensure a successful squeal noise reduction.

1 Introduction Different types of curve squeal noise are presented: how it is generated, on which steps in the generation process measures could intervene, how operational requirements build selection criteria for measures and reports on realised measures and first experiences with such applied measures are presented. Operators decisions which strategy shall be used in their particular case can be supported as it is shown, which practical approach fits to which set of operational requirements.

2 Definition of Curve Squeal Noise Curve squeal noise is noise generated in narrow curves only, producing a sound pressure level which is more than 10 dB(A) higher than normal rolling noise on straight track under the same conditions and which is dominated by pure tones [1].

3 How Is Curve Squeal Noise Generated 3.1 Running Surface Stick Slip Induced Curve Squeal Noise The leading wheel set of a bogie wants to run straight forward. As the outer wheel is led by form fit between wheel flange and outer rail, the wheel set is shifted into the B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 406–411, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

How to Avoid Squeal Noise on Railways State

407

direction to the curve centre. By this, the bend in the inner wheel increases. As soon as the restoring force in the wheel is higher than the friction force, the wheel swings back in a tension free position. This repetitive process stimulates the wheel with forces containing frequencies of a wide band. Some eigenfrequencies grow and generate noise dominated by pure tones [1, 2, 3, 5]. The second wheel set of a vehicle or a bogie behaves generally vice-versa [2]. Due to the generation process of this type of squeal noise, the noise is mostly generated by the wheel which vibrates in its eigenmodes. Nearly no noise is radiated by the rail. This type of squeal noise is called "surface curve squeal noise". 3.2 Wheel Flange Contact Curve Squeal Noise The wheel set is guided through the curve by the form fit of the wheel flange of the outer wheel with the outer rail. As the wheel set wants to run straight forward, an angle of attack between wheel and rail occurs and a second contact point in front of the position, where the wheel rolls on the rail, occurs. At this contact point, the wheel flange is grinding downwards at the rail and excites wheel and rail which both then radiate noise. This type of squeal noise is called "flange curve squeal noise".

4 Operational Requirements to Measures to Tackle Curve Squeal Noise The operation conditions of railway systems are widely spread. Main operating requirements are given by the existing infrastructure, especially the minimum installed curve radii. A minimal friction is required to transfer brake forces and traction forces. On systems with locomotive hauled trains, the latter is generally dominant. Life cycle costs of measures and environmental requirements form further requirements. To assess the applicability of measures to eliminate curve squeal noise, a railway system which represents the envelope of all requirements, was evaluated. This system was found with the Montreux-Oberland-Bahnen (MOB) in Switzerland [9], which operates a metre gauge network between Montreux at lake of Geneva and Zweisimmen in the highlands of Bern, a region with a wide variety of weather conditions. Gradients up to 65 ‰, curve radii down to 75 m exists and long curves (covering angles larger than 180°) exist. The line is operated with electrical multiple units and locomotive hauled trains. It is assumed, that measures which work under the conditions of MOB, are able to be applied successfully in all kind of wheel rail systems.

5 Approaches to Tackle Curve Squeal Noise To avoid the surface curve squeal noise, design solutions can be used which allow the wheels to roll tension free through the curve. Squeal noise is avoided in curves of all radii by applying independent wheels instead of wheel sets. Due to the missing centring mechanism of independent wheels on straight track sections, the wear at the wheels is higher than with wheel sets [7]. With radial steering wheel sets, the minimal curve radius, which a wheel set can roll trough tension free, can be reduced. Different rail profiles at inner and outer rail have the same effect; with a special profile at the

408

S. Bühler and B. Thallemer

inner rail, the contact point is moved far away from the wheel flange [8]. The durability of such measures is not yet known. By friction reduction at the contact patch between the inner wheel and the inner rail, the amplitude of the wheel excitation can be reduced. Smaller stick slip amplitudes are not able to excite the eigenmodes of the wheel any more. Systems for surface or flange lubrication can be either installed on the vehicle or way side [4, 6] and have an impact to both types of curve squeal noise. Additional damping of the wheel may inhibit the excitation of eigenmodes. This measure mainly reduces surface curve squeal noise. To avoid flange curve squeal noise, designs can be used, which allow the wheels to roll through the curve in parallel to the rail. With additional damping of the wheel and rail, the excitation of flange curve squeal noise can be reduced. Rail dampers generally reduce only this type of squeal noise as nearly no damping is applied by rail dampers to the wheel due to the very local contact patch. As wheel and rail radiation contribute to flange curve squeal noise, rail dampers can reduce this noise partially by reducing the rail contribution. Measures which reduce the radiation of the wheel or rail and measures which influence the propagation path are not discussed as their reduction effect is too limited.

6 Experience with Realised Measures to Tackle Curve Squeal Noise PROSE performed several consulting, engineering and measuring projects on behalf of railway operators, vehicle manufacturers, rail fastening suppliers and the Swiss Agency for Environment like wheel and wheel set guiding concepts (bogie and running gear design), validation measurements of vehicles with wheel noise absorbers [10], of rail running surface conditioning systems [9] and of rail noise absorbers [10].

Fig. 1. Different types of wheel noise absorbers. Left: retrofit solution at passenger coach of MOB. Right: solution at a driven wheel set of MOB.

Bogie and running gear design has to find an optimised solution within a huge number of requirements. A lot of them, like the ones to wheel guiding, are contradicting. Running stability at high speeds requires stiff guiding whereas low noise, forces and wear in curves requires soft guiding. On the measurements on MOB less squealing was observed with bogies which allow radial steering [9]. PROSE had the opportunity to perform several measurements to validate the benefit of wheel noise absorbers which was performed with microphones in a bogie with

How to Avoid Squeal Noise on Railways State

409

absorbers and in one without absorbers. In one case with dominating frequencies between 2 and 8 kHz, wheels without wheel noise absorber were quiet in curves with radii larger than 170 m and squealed systematically at curves with smaller radii [10]. In another case, with a dominating frequency at 700 Hz, no improvement by wheel noise absorbers was observed [10]. At the measurements on MOB less squealing was observed at vehicles which were equipped with wheel noise absorbers [9]. LpAeq, 5s [dB(A)]

LpAeq, 5s [dB(A)]

Curvature (1/radius) [1/km] positive: left curve

Curvature ( 1/radius) [1/km] positive: left curve

Fig. 2. Sound pressure level (each point for 5 sec) at the middle of the bogie in relation to the curvature [1/km]. Right, no wheel noise absorbers: curve squealing occurs at curves with radius smaller than 170m (curvature larger than 6 1/km). Left, with wheel noise absorbers: no squealing was observed [9].

On the network of MOB, two different systems for rail running surface conditioning were evaluated and taken into service on long curves. One at the curve Fontanivent with radii down to 75 m, a length of 284m and a gradient of 50 ‰ and one at the curve Sonzier with a radius of 81 m, a length of 315 m and a gradient of 65 ‰. For the validation measurements, only vehicles with bogie types which tend to squeal and vehicles without wheel noise absorbers were taken into account. Both systems were not able to eliminate curve squealing completely. The system at Fontanivent showed a reduction of only 2.2 dB(A) and the one at Sonzier one of 5.0 dB(A). Both improvements are far below the expectations [9]. In the cases when PROSE had the opportunity to perform measurements to assess rail dampers, no improvement related to the same situation without dampers could be identified. Compared with the processes of the generation of surface curve squeal noise, this is no surprise.

410

S. Bühler and B. Thallemer

Fig. 3. Lower end of the curve Sonzier on the line Montreux - Zweisimmen of MOB

7 Conclusions and Proposed Process to Realise Measures to Tackle Squeal Noise To tackle a squeal noise problem efficiently, it is important to determine which kind of curve squeal noise is dominant. The boundary conditions of network, rolling stock and operation (radius, gradient, friction requirements, status of rolling stock, ...) have to be assessed. This information allows to identify appropriate measures.

Acknowledgements We appreciate to have been given the opportunity to perform projects on behalf of the Swiss Agency for the Environment which led to some of the results presented here.

References [1] Thompson, D.J., Jones, C.J.C., Monk-Steel, A.D., Allen, P.D., Hsu, S.S., Iwnicki, S.D.: Railway Noise: Curve Squeal Noise, Roughness Growth, Friction and Wear. UIC Research Project C242.1 [2] Weber, H.H.: Prof. Heumanns Arbeiten auf dem Gebiet der Spurführung im Zeichen der heutigen Rad/Schiene-Technik. ZEV-Glasers Annalen 102, Nr. 7/8, pp. 201–213 (July/August 1978) [3] Hecht, M.: Kurvenkreischen – Ursachen und Gegenmassnahmen. Schweizer EisenbahnRevue 3/1995, pp. 103–108 [4] Int. Railway Journal: Controlling Wheel Squeal On Rapid Transit Systems, 10.04, pp. 30–31 [5] Nelson, J.T.: Transit Cooperative Research Program Report 23: Wheel/Rail Noise Control Manual. Transportation Research Board, pp. 54–65. National Academy Press, Washington D.C (1997) [6] Rudd, M.J.: Wheel/Rail Noise, Part II: Wheel Squeal. Journal of Sound and Vibration 46(3), 381–394 (1976)

How to Avoid Squeal Noise on Railways State

411

[7] Harsy, G.: Fahrwerke mit gesteuerten Radsätzen, ZEV + DET Glasers Annalen 119, Nr 9/10, pp. 335–349 (September/October, 1995) [8] Müller, R.: Die Problematik der Berührgeometrie Rad/Schiene, Schienenfahrzeugtagung 8./9. Berichte und Informationen 2/96 Hochschule für Technik und Wirtschaft Dresden (February 1996) [9] Thallemer, B., Raubold, J., Widmer, M.: Vermeiden von Kurvenkreischen bei Eisenbahnen, Bericht Phase 1 und 2, PROSE Report 2-116, 24.1.2005, Bericht Phase 3, PROSE Report 2-169, 21.6.2005, Bericht Phase 4, PROSE Report 2-240, 18.12.2006 [10] PROSE Reports on behalf of customers, confidential

Noise Reduction Measures at Freight Train Locomotives “Blue Tiger” C. Czolbe and M. Hecht Technische Universität Berlin FG Schienenfahrzeuge, Sekr. SG 14, Salzufer 17-19, D-10587 Berlin, Germany Tel.: +49 30 31425195; Fax: +49 30 31422529 [email protected]

Summary In this project acoustical measures at three modern diesel locomotives are presented. Due to additional rebuilding of vehicles to a small amount the noise limitations of the TSI-Noise 2006 also in rolling stock can be kept. The project is realised with the railway operator HVLE Havelländische Eisenbahn and is promoted as a contribution to the conversion of the Directive 2002/49/EU within the scope of the environmental innovation program of the federal ministry for the environment. The BlueTiger is a six-axle freight train locomotive with 126 t heavy weight, supplied by a 2.5 MW diesel power unit and moved by electric traction. For the classification in categories of the sound sources the sound intensity was measured along the standing and onto the moving vehicle. Mainly fans and auxiliary aggregates are responsible for the excess of the noise limit values. The diesel engine in the middle of vehicle is sound insulated by lockable doors. The high-frequency sound parts of traction fan and electro dynamic brake can be reduced by absorption and insulation.

1 Introduction Noise reduction in freight transport gets up to high relevant theme according to the EU frame work environmental noise in urban areas. Large diesel locomotives for freight transport are often used on side rails or in lines with steep gradients because of their flexibility and autonomy energy supply. The main part of radiated noise of freight trains running with higher speeds (60 up to 120 km/h) comes from the rolling noise. On the opposite at particular slow approach locomotives are the most significant sound sources. Railways and logistics in Europe become more and more interoperable, therefore new technical standards were set. The new ambitious limits of noise radiation of rail vehicles are valid for the new manufactured stock since June 2006. Most of the older locomotives of the rolling stock don’t pass the TSI Noise limits. Due to the long life cycle of modern locomotives up to 30 years noise reduction into the environment will dramaticly delay. This Project is to present effective and affordable rebuilding measures on heavy diesel locomotives to fit the international acoustic standards of the TSI Noise. Advantages B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 412–418, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Noise Reduction Measures at Freight Train Locomotives “Blue Tiger”

413

to the railway company are more flexibility at international (interoperable) transport offers and less trouble with residents along the line. cooling area

diesel engine

cooler fans, transformer

4150

e.-brake resistor fan

compressor 23410

Fig. 1. BlueTiger diesel locomotive areas of measures after retrofitting

2 Preliminary Investigations and Initial Situation 2.1 Measurements According to ISO 3095 To acquire the initial noise situation in the relevant operating states the TU Berlin department of rail vehicles did measurements of pass by travel and standstill. Special enterprises are done in particular slow approach and starts under load, this is the relevant operating condition in urban areas near by lines with steep gradients (Fig. 2). The noise limit values of the TSI-Noise at these velocities are exceeded only through the compressor for the air supply of the braking system, which frequently is active in case of switching moves and tracks with high radient.

Fig. 2. Pass by noise level over time (measured & calculated) double traction freight train at 30 km/h

2.1.1 Standstill Noise The noise limit value of 75 dB(A) LAeq at standstill are exceeded due to the working piston compressor up to 2.5 dB (Fig. 3). Standstill is defined as an operating condition without auxillary equipment, but on heavy rail transport on steeply sloping sections the compressor permanently runs to provide the brake pressure for the freight train. Thus we decided to involve this special operating state into our investigation to compare standstill limit values.

414

C. Czolbe and M. Hecht

+

current state current state + comp. TSI Noise Limitation retrofitted state + comp

70 dB 77 dB 75 dB 70 dB

Fig. 3. Measurements at standstill of loudest measuring point (current state) with 7.5 m distance and muffler

2.1.2 Pass by Noise Constant Speed During pass by on constant speed the noise limit excesses are clearly caused by the working auxiliary equipment (Fig. 4). Rolling noise at these speeds takes a minor part without relevant acoustical influence. Item Unit Train Loc. front Loc. rear

t s

v LAeq LAFmax TEL km/h dB(A) dB(A) dB(A) 22 40 83,0 89,5 83,5 2 40 87,3 88,8 2 40 87,8 89,5

Fig. 4. Measurement pass by level of a freight train with two Blue Tigers

2.1.3 Pass by Noise at Start-Up Start-up measurements of a locomotive have to be performed under customary load, either to use another locomotive with breaking state or a freight train of adequate mass. This task according to TSI-Noise is not easy to be realised because both methods produce additional noise. In this case an empty freight train with a second rear locomotive without traction was measured. The whole wheel set of the train was free from flat spots and the environmental noise was disregarded. The start-up noise under load (780 t train) mostly stays below the limit value 89 dB(A) LAFmax (Fig. 5), while the rotational speed of the engine and the power output is maximal. L AFmax Time 10:19:59 12:18:30 15:21:23

7.5Li 1.2h 7.5Re 1.2h 7.5Li 3.5h 7.5Re 3.5h 25Li 3.5h 7.5Li 1.2h dB(A) M2 M3 Service M1 M4 M5 M6 Start-up1 88,0 86,7 87,2 79,3 78,5 89,4 Start-up2 85,0 85,5 87,4 85,4 78,7 84,3 Start-up3 84,5 85,0 87,7 85,0 77,7 82,9

Fig. 5. Loaded start-ups according to TSI (780t freight train)

2.2 Sound Intensity Investigation During a run of a loaded freight train (1500 t) on a track with steep gradients all auxillary equipment were active at least once a time. This chance was used for measurements

Noise Reduction Measures at Freight Train Locomotives “Blue Tiger”

415

traction cooling fan 92-99 dB(A)

traction cooling fan 95-99 dB(A)

electro dynamic brake resistor fan 102-108 dB(A)

motor 82 dB(A)

air compressor on 92 dB(A)

Fig. 6. Sidewise intensity level of separate aggregates while load driving

of sound intensity levels of relevant aggregates onto the running vehicle as it’s shown below (Fig. 6). 2.2.1 Classification of Sound Sources According to this, the auxiliary equipment sound sound sources can be classified in two categories, the low frequent compressor intake noise with harmonic components at 160 Hz, 316 Hz (Fig. 7) and the high frequent wide-banded fan noise (Fig. 9). So the measures similarly differs, mufflers against the intake noise and absorber louvres for the cooler vents. The intensity levels keep in dependency of the regular condition and the number of revolutions of the fans within an interval but they differ essentially in height and spectral component.

3 Mufflers for the Compressor’s Intake Noise 3.1 Spectral Components In Fig. 7 third octave spectrums while standstill are shown. The pressure spectrum on the left contents the loudest measuring point (Nr.18 Fig. 3) with a working compressor in a distance of 7.5m from the middle of track. Wherein the spectral components at 160 Hz with the number of revolutions based ground frequency and the 400 Hz band with overtone parts specify the level. In fact of the close location of driver cap to the compressor the engineer is exposured to an high noise impact too. sound intensity level IAeq near by compressor 1m

sound pressure level LAeq at loudest point 7.5m

Iaeq Komp an Originalzustand Iaeq Komp an mit SD Iaeq Komp aus

80 77,1

noise limit value of TSI 75 dB(A)

75

110

74,4

105 100

Kompressor 2005

95

Kompressor mit SD 2007

90

dB(A)

65 60

85 80 75 70

55

65 60

50

55

Terzband in Hz

Fig. 7. Standstill noise 1/3 octave spectrum current state and reduced state with muffler

L

A

8000

10000

6300

4000

3150

5000

2500

1250

1600

2000

800

630

1000

500

400

315

250

200

160

125

A

8000

6300

5000

4000

3150

2500

2000

1600

1250

800

630

1000

315

500

400

160

250

200

80

125

63

f [Hz]

100

50

45 100

LAeq dB(A)

70

416

C. Czolbe and M. Hecht

3.2 Noise Reduction The muffler’s construction between the filter and cylinder intake can sufficiently reduce the dominant low frequent bands. So the limit value of 75 dB(A) in standstill was already kept with running compressor. Sound level reduction between 10-15 dB can be achieved in the third octave bands between 100 and 200 Hz which follows the trend of the included absorbability of this muffler (Fig. 8). The higher frequency noise parts (Fig. 7 left diagramm) during the compressor’s working do not develop in fact of the characteristic of intake noise and the additional noise of the main fan. A decrease of sided radiation is expected because the absorber louvres will be installed and their insulation effect increases by high frequencies.

Compressor intake

Air r te Fil

Fig. 8. Damping loss of the Cowl-Muffler

4 Absorber Louvres and the Cooler Vents The locomotive’s cooler exits are constructed with screwed or locked grids. That means on an acoustic view the auxiliary equipment is located in free field conditions. An absorbing cover panel for this vents has to let flow through the necessary air for the cooling process of this locomotive. The used louvres (Fig. 10) are made of blade sheet metal which is filled with absorbing material and arranged with vent sheet in inside direction. 4.1 Spectral Components The spectral components of the intensity investigation (Fig. 6) of cooling vents are shown below in Fig. 9. Electric brake resistor fan

Traction cooling fan Iaeq

Laeq

110 105 100 95

dB(A)

90 85 80 75 70 65 60 55

L

A

8000

6300

10000

5000

4000

3150

2500

2000

1600

800

1250

1000

630

500

400

315

250

200

160

125

100

50

Terzband in Hz

Fig. 9. Spectral components IAeq (filled) LAeq (transp.) of tractio100n cooling fan and electric brake resistor fan

Noise Reduction Measures at Freight Train Locomotives “Blue Tiger”

417

4.2 Damping Loss of Slimshield Absorber Special difficulties concerning the absorbing louvre construction appeared because of the changing high flow rate of the aggregates as well as big head and cross winds developed by high speed. The maximum air flow of the main cooling fan with 2.5m diameter amounts 70 m³/s, the outcome of this are maximum flow rate velocities reach 5 – 8 m/s for the sided doors at the current status. To ensure the cooling effect and not to push the absorbing material out of the blades the louvres were produced with 50 % minimum passage area by coming off norm dimensions. The damping loss of norm elements according to the manufacturer’s catalogue [2] is shown in Fig. 10. But this would be the “best case” so we correct our calculated values to a 3 dB less per 1/3 octave band because of higher ranges of passage.

D in dB

16 14 12 10 8 6 4 2 0 63

125

250

500 1000 2000 4000 8000 f in Hz

Fig. 10. Damping loss IAC slimshield 102mm, old grid elements vs. new absorber slim shields

4.3 Construction Details and Durability In the current state doors and grids (Fig. 11 left picture) cover several diagonal crossing beams inside the cooling area frame. This fact has to be realised at the construction of slimshild absorber elements and diagonal notch sections have to be installed. Another purpose of the project is to collect experiences of this construction concerning the stress endurance limit in rail vehicle building especially by recurrent vibrations on a locomotive. This should be reached by special blade bondings and covers and by proving with these vehicles. The refitting of three locomotives is partial supported by the BMU [3] within a pilot project.

Fig. 11. Cooling section current state (left) and after rebuilding with noise shields (right)

418

C. Czolbe and M. Hecht

5 Final Acoustical State 5.1 Noise Levels The final situation of noise radiation after passing the acoustical investigations under testing conditions is in brief at Fig. 12 below. The measures are working and the BlueTiger loc now is TSI-Noise compatible. Operation state

Standstill Standstill + comp. Starting passby v80 km/h

TSI Noise limits dB(A) 75 75 89 85

LAeq LAFmax initial state dB(A) dB(A) 70 77 89 ?

LAeq LAFmax retrofitted state dB(A) dB(A) 69 70 88 83

Fig. 12. Noise values before and after retrofitting BlueTiger Loc

The operating condition which is relevant for the transport on steep tracks is constructed and can not be extrapolated to the requested noise limit of the TSI at 80 km/h. The TSI is quoted as an effective standard but the included limits are no longer state of the art because of the conciliation process for years. The limit values for vehicles are too lenient (in this case too high dB limits) and do not represent all operating conditions. An effective noise reduction in railway traffic can only be reached in 10 years by executing such measures o the rolling stock.

References [1] TSI Noise 2006/66/EC, http://www.era.europa.eu [2] IAC, Industrial Acoustics Company, http://www.iac-gmbh.de [3] BMU, Bundesministerium für Umwelt, http://www.bmu.de/foerderprogramme/pilotprojekte_inland/doc/38466.php

Noise Reduction at Urban Hot-Spots by Vehicle Noise Control U. Orrenius, S. Leth, and A. Frid Bombardier Transportation, Specialist Engineering Mainline & CoC Acoustics, Östra Ringvägen 2, SE – 721 73 Västerås, Sweden Tel.: +46 21317000 [email protected]

Summary In the present paper the potential for reducing noise exposure at “urban hot-spots” due to rail traffic is investigated. In particular vehicle based control measures are discussed including various measures against curve and brake squeal, intelligent fan speed regulation taking advantage of the thermal inertia of the systems to be cooled, choice of fan types and innovative energy storage devices. It is shown that dedicated measures on the vehicles can lead to significantly reduced emissions at locations where people are mostly disturbed. For future decision making on urban noise control measures it is essential that the calculation models used must incorporate in sufficient detail the source mechanisms that are mostly relevant from a disturbance perspective.

1 Introduction The Environmental Noise Directive (END) imposes strategic noise maps to be created for large urban areas, major roads, railways and airports. The member states shall “ensure that no later than 30 June 2007 strategic noise maps have been made and, where relevant, approved by the competent authorities, for all agglomerations with more than 250 000 inhabitants” [1]. Special focus will be on so called “urban hotspots”, locations where the acoustic emissions affect a lot of people. Once the strategic noise maps are in place, action plans will have to be defined, containing a number of measures to manage noise and reduce it where necessary. To define action plans for noise reduction around a major railway requires, as a start, ranking of the sources that are creating the hot-spot. Freight wagons with cast iron brakes are by far the most important railway noise problem in Europe. It is wellknown that a replacement of cast iron brake blocks with sinter or composite materials will greatly reduce the freight wagon noise. Technical solutions exist and the rate of implementation is primarily a matter of funding. Consequently, this will not be discussed further in this paper. For rail traffic in general, measures to reduce rolling noise will have to be taken on the infrastructure, using rail dampers as well as rail grinding and wheel re-profiling schemes. Also other sources on the vehicle will have to be controlled. For mass transit vehicles motor cooling fans can typically contribute to half of the total vehicle sound power at full speed on a low noise track. At acoustic hot-spots also such sources must B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 419–425, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

420

U. Orrenius, S. Leth, and A. Frid

be controlled to achieve the reduction needed. New strategies for optimizing noise control measures, allowing different levels of noise at different locations along a line are therefore of interest. In addition, for rail vehicles recent TSI legislation [2] imposes greater focus on noise emissions at standstill and acceleration where noise from propulsion and cooling equipment are dominating. For stationary and accelerating vehicles, the operating conditions of cooling equipment are crucial and a systematic approach to thermal management can reduce ventilation noise exposure considerably.

2 Noise Emissions from Urban Rail Vehicles 2.1 Significance of Rail Noise When comparing exposure from different modes of transportation, road traffic noise generally dominates the noise maps. Rail, air and industrial noise are of secondary importance to the number of people exposed. However, certain urban hot-spots are solely due to rail traffic. With rail traffic speeds increasing the noise generation will also increase if measures are not taken, both on the track and on the vehicles.

Rail Hot-Spot

Fig. 1. Noise maps of a central part of Stockholm. The left and right map show calculated equivalent sound levels due to road and rail traffic respectively.

In Figure 1 noise maps from central Stockholm are displayed [3], as determined from applying the Nordic noise calculation model. To the right the exposure to rail noise is shown, with a rail hot-spot indicated. 2.2

Rail Noise Sources in Urban Areas

Rolling noise has received much attention the last decades, being the dominating source for most rail vehicles at full speed. However, in an urban context many other source mechanisms are also important in an annoyance perspective. For mass transit vehicles, like metros, trams and many regional trains, traction motor cooling fans are very important sources. Other ventilation systems such as diesel motor blowers and

Noise Reduction at Urban Hot-Spots by Vehicle Noise Control

421

air-conditioning fans are significant at and around platforms. In addition, tonal noises, such as curve and brake squeal as well as electro-magnetic sound from transformers and propulsion systems, are unfortunately both rather frequent and highly disturbing.

3 Noise Control Scenarios 3.1 Traction Motor Cooling Traction motor cooling for mass-transit vehicles is often provided by a fan mounted directly to the motor shaft. The main reason is that this is a robust and fairly inexpensive design. One important side effect is the noise generation. With the fan sound power increasing with ~50 log(v), where v is the vehicle speed, the cooling system noise is typically in the same order of magnitude as the rolling noise at full speed. If rolling noise is reduced, e.g. by rail damping systems or acoustic rail grinding programs [4], the cooling systems often dominates the total. Alternative more silent cooling systems, e.g. water cooled motors or separate roof mounted fans, are possible but at a higher expense and also with technical constraints. One option for reducing noise exposure is to apply a slip clutch system that mechanically disengages the fan at the hot-spots [5]. Such systems are standard on modern truck diesel engines. In view of the fairly large thermal inertia of the motor, the reduced cooling can be compensated by either pre- or post cooling at a location where community noise is not critical, e.g. in a tunnel. This means that the noise emission can be systematically adapted along the track. Certain intelligence of the system is needed in such that any reduction of fan speed is determined also with respect to future cooling needs. The rail hot-spot indicated to the right of Figure 1 is solely caused by the traffic from the A32 high speed light rail vehicle. This vehicle is driven by self ventilated motors. It is estimated that the external noise can be reduced by 3-4 dB at the vehicle top-speed at 70 km/h by temporarily slipping the fan to a lower speed, subject to that rail and wheel roughness are kept at low levels. 3.2 Cooling and Noise Management at and Around Platforms At urban areas, noise from trains at low speeds and at standstill, for example in or around platform areas, are important for how people are affected. At these situations, the dominating sources will typically be noise from diesel engines, cooling systems and auxiliary equipment. In this section, a few examples will be used to demonstrate how intelligent cooling control strategies and innovative energy saving devices can be used to considerably reduce such noise. 3.2.1 Intelligent Cooling Control Strategies The recently launched diesel-electric locomotive TRAXX P160 DE is a good example of how a systematic effort to optimise the operation of cooling fans has lead to excellent low noise performance [6]. The measured noise levels are well below those specified in the contract and in legislation.

422

U. Orrenius, S. Leth, and A. Frid traction motor fan

brake resistance

traction system: motor + gear

air inlet

exhaust

compressor brake resistance

cooling tower

traction motor fan

traction system: motor + gear

Fig. 2. TRAXX P160 DE locomotive with important noise sources indicated

This locomotive is the latest member of the TRAXX product platform family which also includes a variety of electric locos. By using modern frequency controlled motors, the fans of traction motor and braking resistor can be adjusted to operate at a speed corresponding to the exact cooling needs. The fan of the diesel engine coolers is driven by a hydrostatic motor and can be adjusted as well. The flexibility of the vehicle level control system allows, in principle, any control scheme to be implemented. For instance, so-called “super cooling” is activated at some distance before the train approaches a station. This means that pre-cooling is provided at locations where a noise increase is less harmful. When coming in to the station the cooling can then be brought down substantially and the thermal inertia of the components cooled can be used as a buffer until normal cooling is again needed. For the TRAXX loco, there is also a “Low noise button” at the driver’s desk that will turn the loco in a quiet mode at standstill. Of course, the temperature monitoring of the critical systems will override this if necessary. The combination of individually controlled motors with the use of information from GPS systems and advanced control software offers many possibilities to create control strategies. For instance, a “drive style manager” which currently displays to the driver a recommended speed to minimise the energy consumption for the route could easily be adjusted to include also noise parameters. 3.2.2 Energy Storage Systems As a logical extension to the use of intelligent control strategies as described above, the use of energy storage systems can be applied to further reduce the need for noisy fans and engines operation at urban hot-spots. Although the main purpose of such systems (super capacitors, batteries, fly wheels) is to reduce energy consumption, there is a clear potential benefit in terms of acoustics as well. The principle is to store energy during braking and re-use it when accelerating or during other shorter periods. Bombardier has developed an Energy Saver concept based on super capacitor technology [7]. It has been in successful passenger operation since 2003 on a prototype tram in Mannheim. It is offered for diesel multiple units and trams. The acoustic benefit is particularly interesting for diesel vehicles as the engine and its cooling

Noise Reduction at Urban Hot-Spots by Vehicle Noise Control

423

systems can hence be turned off when the train is in the station area. Moreover, coasting operation with the engine more or less switched off can be extended while passing sensitive areas close to housing and thereby reduce the noise annoyance experienced in urban areas. 3.3 Curve and Brake Squeal Abatement Curve and brake squeal are particularly annoying noises due to their tonal characters, even if they do not explicitly show up on present noise maps. Squeal is however included in, for example, the Harmonise/ Imagine source model for railway noise [8] and the German Schall03 2006 model [9].

400

without absorber with absorber 10 dB(A)

800

1.6k Frequency (Hz)

3.2k

Fig. 3. Effect of plate absorbers

6.3k

Sound Pressure Level dB(A)

Sound Pressure Level dB(A)

3.3.1 Curve Squeal A lot of R&D work has been devoted to curve squeal in special research programs for example in the Netherlands [9], UK [11], France [12], by UIC [13] and at the moment in Germany within the framework of Leiser Verkehr [14]. The phenomenon is difficult to predict since a number of different mechanisms are involved and the occurrence is in practice also weather dependent. However, a large majority of the problems with curve squeal can be solved by applying one or a combination of two of the following methods: radial steering, lubrication/friction control on track or wheel and wheel damping. In the very few cases when these methods are not enough, special forms of active control with a secondary feedback system could in the future potentially be applied as suggested in [15]. Radial steering bogies and wheel dampers are provided on the C20 Stockholm Metro cars, practically eliminating the occurrence of curve squeal. Many Bombardier vehicles of today are also equipped with various systems for flange lubrication to reduce wear. These systems will in many cases eliminate the squeal. Tests switching off/on the flange lubrication system show for a K4500 tram a 20 dB (A) reduction of the squeal noise level in the 2 kHz band [16]. The vehicle lubrication system can be controlled in several different ways; time, distance, position-in-track system or detection on the bogie if a curve is approaching. Dedicated track based friction control systems can also be very efficient [17]. For Bombardier Light Rail Vehicles the wheels are prepared so that wheel dampers can be fitted in case this is desired by the customer. For the FLEXITY Outlook vehicle for Linz a 9 dBA reduction was measured when plate absorbers were mounted, see Figure 3, and for K4000 in Nottingham and Cologne the tram squeal was practically eliminated by introducing block absorbers, see Figure 4 [16].

400

without absorber with absorber

10 dB(A)

800

1.6k Frequency (Hz)

3.2k

Fig. 4. Effect of block absorbers

6.3k

424

U. Orrenius, S. Leth, and A. Frid

3.3.2 Disc Brake Squeal Also brake squeal is an intermittent phenomenon that is difficult to accurately predict. Repeated tests and a statistical approach are required for evaluation [18]. For brake squeal there are no off-the- shelf methods that will eliminate the squeal in all cases. In present Bombardier vehicle projects, challenging requirements during braking from 30 to 0 km/h are included. The solutions are developed at Bombardier in close cooperation with brake supplier and the customer. Different organic pads are systematically tested for finding the best technical solution combining high braking performance with low noise.

4 Concluding Remarks A significant potential for reducing noise exposure at urban hot-spots is anticipated when appropriate control measures are taken. This potential is particularly evident when the impact of vehicle noise emissions is accounted for on a minimal disturbance basis. For example, curve squeal can be reduced to a minimum by applying a combination of known methods. Brake squeal can be considerably reduced with optimised brake pads and practically eliminated for cases when stable methods for electric braking down to 0 km/h is in place for wider implementation. The annoyance from ventilation systems, both at full speed and in and around platforms, can be significantly reduced by applying thermal and acoustic management schemes in which the thermal inertia of the system to be cooled is accounted for so that fan operation at sensitive areas can be kept at a minimum. The concept of efficient thermal and acoustic management can be further enhanced by utilizing various energy storage systems, like super capacitors. The use of strategic noise maps can be very effective when defining noise control programs. However, it is the authors’ view that the calculation schemes applied when producing future noise maps must incorporate the source mechanisms that are mostly relevant from a disturbance perspective. If this is not the case there is a great risk that noise control measures to reduce hot-spots appearing on the maps will be misdirected with a poor efficiency in annoyance reduction per Euro spent as a result. For example, brake and curve squeal are highly disturbing source mechanisms and should be accounted for in sufficient detail in both noise maps and directives, e.g. as described in reference [8].

Acknowledgements The support from colleagues throughout the Bombardier Acoustics Network in the preparation of the present article is gratefully acknowledged. In particular the authors would like to thank Martin Ognar, Bert Stegemann, Erik Thoss and Urs Treichler.

References [1] Directive 2002/49/EC of 25 June 2002 relating to the assessment and management of environmental noise. Official Journal of the EC, L 189/12 [2] Technical specification for interoperability – Subsystem conventional rail rolling stock – Scope Noise – 23/12/2005. Official Journal of the EC (08.02.2006)

Noise Reduction at Urban Hot-Spots by Vehicle Noise Control

425

[3] http://www.map.stockholm.se/kartago/kartago_fr_buller.html [4] Asmussen, B., et al.: Status and perspectives of the “Specially Monitored Track”. Journal of Sound and Vibration 293(3-5) (2006) [5] Orrenius, U.: Feasibility study of clutch system for traction motor fans, NoV-SE-2007002, Bombardier (InMAR) (2007) [6] Thoss, E., et al.: Optimierung der Schallemission von Schienenfahrzeugen mit “nicht” akustischen Massnahmen. In: DAGA (2007) [7] Steiner, M., Scholten, J.: Energy Storage on board railway vehicles. In: European Conference on Power Electronics and Applications, Dresden (2005) [8] Dittrich, M.: The Imagine Source Model for Railway Noise Prediction. Acta Acustica 93, 185–2002 (2007) [9] Moehler, U., et al.: The new German prediction model for railway noise Schall 03. In: Proceedings Euronoise 2006, Tampere, Finland (2006) [10] de Beer, F.G., et al.: Curve squeal of rail bound vehicles (part1-3). In: Proceedings of Internoise 2000, Nice, France (2000) [11] Thompson, D.J., et al.: A Theoretical Model for Curve Squeal. ISVR TM 904 (February 2000) [12] Vincent, N., et al.: Curve squeal of urban rolling stock. Journal of Sound and Vibration 293(3-5) (June 2006) [13] Muller, B., Oertli, J.: Curve squeal of urban rolling stock. Journal of Sound and Vibration 293(3-5) (June 2006) [14] http://www.leiserverkehr.de/web/projekte/kurvengeraeusche/projektdetails.html [15] Heckl, M.: Curve Squeal of Train Wheels, Part 3: Active Control. Journal of Sound and Vibration 229(3) (January 2000) [16] Ognar, M.: Investigations on curve squeal, internal report Bombardier Transportation (2006) [17] Eadie, D.T., Satoro, M.: Top-of-rail friction control for curve noise mitigation and corrugation rate reduction. Journal of Sound and Vibration 293(3-5) (June 2006) [18] Beier, M., et al.: Acoustical Investigations of Disc Brake Squeal. In: Proceedings Euronoise 2006, Tampere, Finland (2006)

Directivity of Railway Rolling Noise Xuetao Zhang SP Technical Research Institute of Sweden, Box 857, SE-501 15 Borås, Sweden Tel.: +46 10 5165021; Fax: +46 33 138381 [email protected]

Summary Directivity is an important parameter to describe a sound source. For railway rolling noise this becomes a complicated issue because its vertical and horizontal directivities depend on many factors. In this paper a complete procedure to determine the vertical and horizontal directivities of railway rolling noise in one-third octave bands has been proposed. The procedure consists of two parts: (1) to define the vertical and horizontal directivities both for track and wheel radiations; (2) to determine the effective vertical and horizontal directivities for the total railway rolling noise. For practical reasons it is better to include the shielding effect of the car-body, also the low barriers when they are present alongside a track. Since this shielding effect dominates, the inclusive vertical directivity of railway rolling noise can vary with train types, and will also be different when influenced by low barriers (such as viaducts). The horizontal directivity of railway rolling noise is determined by the horizontal directivities of the track and wheel radiations, and weighted by the relative importance of the two sub-sources which is described by the track and vehicle transfer functions. In general, the horizontal directivity of railway rolling noise varies with different vehicle-track combinations.

1 Introduction A traffic noise source is usually specified in terms of the noise sound power including speed dependence, the representative source height(s) and the directivity. For railway rolling noise the noise sound power is described using the roughness and transfer functions (and a train speed); the representative source heights are the railhead for track radiation and 0.5m above the railhead for wheel radiation, according to the Harmonoise proposal [1]. The directivity of railway rolling noise is only roughly described because of the complicated nature of the noise generation process. More investigations on the directivity are required. Railway rolling noise consists of two sub-sources: track radiation and wheel radiation. Some authors also suggest that the car-body vibration can be important at low frequencies [2]. In this paper only track and wheel radiations will be considered because (1) the total power of the car-body radiation is negligible compared with the other two sources; (2) the directivity of the car-body radiation can be estimated to be similar to a monopole. In this paper the recent study on the directivity of railway rolling noise is presented. The directivity of railway wheel radiation will be discussed first in section 2. The B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 426–432, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Directivity of Railway Rolling Noise

427

discussion on the directivity of railway track radiation is followed in section 3. Section 4 and section 5 will respectively discuss how to construct the vertical and horizontal directivities of railway rolling noise using the corresponding directivities of the wheel radiation and track radiation. The equivalent horizontal directivity of an incoherent line source such as train pass-by noise is discussed in section 6. Conclusions will be given in section 7.

2 The Directivity of Railway Wheel Radiation A systematic investigation on railway wheel radiation was reported by Remington in 1976 [3]. A wheel set consisting of two curved-web wheels with a diameter of 0.76m was used for the investigation. Concerning directivity, Remington concluded that the railway wheel radiation is omni-directional within 5 dB for both radial and axial excitations, in one third octave bands. The measurement radius was 3.1m. Ten Wolde and van Ruiten reported different results on the directivity of railway wheel radiation in 1982 [4]. They used also curved-web wheels with a diameter of 0.95m and measured sound intensity at 2m distance from the wheel centre. “The wheels do not show uniform directivity for all frequencies and both ways of excitation.” How the results differed from the Remington’s was not described in detail in the paper. During the Harmonoise project, aiming at reaching a suitable description on the directivity SP made some investigations on the issue [5]. A SJ 57H freight car wheelset consisting of two curved-web wheels with a diameter of 0.92m was measured at a test rig also in SP’s semi-anechoic chamber, with a measurement radius of 2.45m. The SP’s results showed that (1) the one third octave band results contained strong interference effects and it was difficult to make a conclusion on the directivity; (2) the Aweighted total sound pressure level presented a directivity pattern similar to the Remington’s, for both radial and axial excitations. In Ref. [6], the sound radiation characteristics of a railway wheel were investigated using boundary element calculations. Different mode shapes of a vibrating railway wheel can be approximated by multipoles. The boundary element calculations did not predict the directivity of wheel radiation. Based on some TNO measurements, where a measurement radius of 1.5m was used and the wheel (0.92m in diameter) was excited radially on the tyre at the natural frequencies, the authors proposed that the directivity of wheel radiation “is based on a dipole distribution for axial motion and a monopole (omni-directional) distribution for radial motion”, for wheels with a straight web. The TNO data also showed that “results for wheels with a curved web resembled the radial mode results.” It is interesting that only the Remington’s data showed a clear directivity pattern in one third octave bands. The SP one third octave band data [5] and the TNO narrow band data [6] all contained strong interference effects. The differences concerned with the measurements can be found in two aspects: (1) the wheels used by Remington are smaller (the diameter is 0.76m for Remington’s compared with 0.92m for SP’s and for TNO’s); (2) the measurement distances with respect to the wheel radius are different (the ratio is 4.08 for Remington, 2.66 for SP and 1.63 for TNO). Some simple calculations made by the author show that it might be better to have this ratio between 4 and 6.

428

X. Zhang

The SP data in A-weighted total sound pressure level support Remington’s conclusion on the directivity of wheel radiation. According to the description in [6], for wheels with a curved web the radial mode results, i.e. the monopole directivity, will be assumed. Thus, it seems that one can reach the conclusion that the radiation from railway wheels with a curved web will be only slightly directional: it is quite close to a monopole. Railway wheels with a straight web need more investigations, because the TNO data [6] were only made for some wheel natural frequencies. One third octave band data are required, and these data shall better be measured at a distance not less than 4 times the wheel diameter. At this time it is not clear for the radiation of railway wheels with a straight web what kind of modes (radial or axial) will dominate in one third octave bands, which will determine if the directivity of the radiation is like a monopole or a dipole. Thus, based on the discussions above, for railway wheels with a curved web the vertical and horizontal directivities of wheel radiation in one third octave bands can be estimated as [5]

ΔLwheel (ϕ ) = 10 lg[0.4 + 0.6 cos(ϕ )]

(1)

This function gives 4 dB level differences between 0 and 90 degrees.

ψ

Source r v Horizontal plane

Receiver

ϕ

Fig. 1. Angles used in this article

3 The Directivity of Railway Track Radiation The vertical directivity of railway track radiation is not important [3, 4]. In Ref. [5] this vertical directivity is estimated as

[

]

ΔLVrail (ψ ) = 10 lg 0.4 + 0.6 cos 2 (ψ )

(2)

Since there is only a small difference between Eq. (1) and Eq. (2), it is then proposed that using Eq. (1) to replace Eq. (2) to describe the vertical directivity of railway track radiation in order to gain some convenience in engineering without losing accuracy. In general, an open track has small decay rate. When a track is excited the effective vibrating length is quite long and the vibration will be attenuated gradually from the excitation point. There is no effective method to directly measure the horizontal directivity of railway track radiation. The estimation of the horizontal directivity of railway track radiation is based on the physical features of track vibration.

Directivity of Railway Rolling Noise

429

It is assumed the dipole directivity for track radiation above the decoupling frequency of the track, is because the effective vibrating length of a track is much larger than the rail height irrespective of track decay rate (even in the case where rail dampers are applied). Below the decoupling frequency sleeper vibration dominates which is assumed to have the monopole directivity. In Ref. [7] sound radiation from a vibrating rail is investigated using a twodimensional model. A free track is modelled as a coherent beam. A measurement distance of 3m from the rail is used for the calculations. In the paper the directivity of the rail radiation is described in terms of the angle of inclination of the intensity vector to o

the normal of the rail. Thus, for a point dipole the angle will be 26.5 , and for a point o

monopole the angle will be 45 . Ref [7] showed that, for a free track with soft pads all wave types except the lateral bending wave will have directivities more directional than a point dipole (i.e., these radiations will direct more energy close to the normal of the rail than a point dipole). One may simply assume, for engineering applications, that the horizontal directivity of the radiation from a short track section is roughly like a dipole. A railway track is excited at multiple axle positions simultaneously and incoherently, during a train passing-by. The coherent beam model may not be suitable for train pass-by cases, at least when long distances are concerned. If not measuring the sound power of rolling noise, a measurement distance will usually not be less than 25m. Thus, the power of railway track radiation can be located at the axle positions of the passing-by train. And, the total power of the track radiation at all the axle positions can then be added incoherently, together with the horizontal directivity considered. This is the way people handle the cases in engineering applications.

4 The Vertical Directivity of Railway Rolling Noise There are two options to handle the vertical directivity of railway rolling noise: 1. Considering only the vertical directivities of wheel and track radiations. The shielding effect of car-body and/or low barriers is left to a sound propagation module to handle. The advantage of this choice is that the vertical directivity of railway rolling noise is as described by Eq. (1), for all train-track types. The disadvantage of this choice is that it does not help in engineering. It is both very time-consuming and difficult to calculate accurate screening effect in the near field of the sound sources. 2. The alternative is to take the shielding effect into account in a source module. The advantage of this choice is that this inclusive vertical directivity of railway rolling noise can be determined with accuracy. But, this option requires that, for each different train-track combination, the inclusive vertical directivity of railway rolling noise has to be determined based on measurements. It is impossible to work out a universal vertical directivity function for general applications. In the Harmonoise project and Imagine project option 1 has been chosen. Fortunately, for present applications, the vertical directivity of railway rolling noise is important only under a few special situations such as above a tunnel opening or neartrack high buildings. The vertical directivity of railway rolling noise usually has a

430

X. Zhang

small effect (<1 dB) for vertical angle positions under 40 degrees relative to the horizontal plane, for traditional wagon types. But, modern trains can have special bogieskirt designs which can have strong shielding effect even for those small vertical angles. And, when low barriers are present alongside a track such as a viaduct, the inclusive vertical directivity of railway rolling noise can become very different from Eq. (1), as reported in [8, 9]. Taking option 2 is expected in future engineering applications. Much measurement effort is then required for the purpose. A database will be built up for accumulating the data of the inclusive vertical directivity of railway rolling noise, for different train types and tracks with low barriers. An example of such inclusive vertical directivity is given in [5].

5 The Horizontal Directivity of Railway Rolling Noise The horizontal directivity of railway rolling noise is determined by the combined effect of the horizontal directivities of wheel and track radiations, weighted by their relative importance in sound power. Shielding on wheels could cause some extra effect on the horizontal directivity but usually not. Applying dampers on rail can significantly reduce the effective vibrating length of an excited track and reduce the track contribution to the total sound power of railway rolling noise. Similarly, when wheel dampers are used the importance of the wheel radiation will be reduced. Thus, the horizontal directivity of railway rolling noise is determined as follows: 1. 2.

for wheel radiation the directivity is given by Eq. (1); for track radiation it is assigned the dipole directivity above the decoupling frequency of the track and the monopole directivity while below the decoupling frequency; weighted by the relative importance between the wheel and track radiations, which is determined by the vehicle transfer function and track transfer function.

3.

The vehicle transfer function and track transfer function can be obtained either using the TWINS calculations or using the indirect roughness method [2]. The formulation to determine the horizontal directivity of railway rolling noise is given below:

⎧ Dwheel ⊕ {LH ,tr ( f ) − LH ,veh ( f )}, ⎪ ΔL (ϕ , f ) = ⎨ ⎪D ⎩ wheel ⊕ {Ddipole + LH ,tr − LH ,veh },

f ≤ f de

R H

where

(3)

f > f de

[

]

Dwheel = 10 lg[0.4 + 0.6 cos(ϕ )] ; Ddipole = 10 lg 0.01 + 0.99 cos 2 (ϕ ) ;

LH ,veh ( f ) is the vehicle transfer function; LH ,tr ( f ) the track transfer function; f de denotes the de-coupling frequency of the track, railway rolling noise, and ⊕ energy summation.

ΔLRH the horizontal directivity of

Directivity of Railway Rolling Noise

431

6 The Equivalent Horizontal Directivity of an Incoherent Line Source In section 5 what has been discussed is the horizontal directivity of railway rolling noise at one axle position. For train pass-by applications one has to handle the equivalent horizontal directivity of a line of incoherent point sources. 500

8

400

6

Monopole Wheel noise Rolling noise Dipole Source line

300 200

Monopole Wheel noise Rolling noise Dipole Source line

4

2

100

0

0 -500

0

500

-100

-50

(a)

0

50

100

(b)

Fig. 2. The SPL contours for a line of incoherent point sources with different horizontal directivity assigned (using the Nord 2000 propagation module together with a regular terrain profile for a railway track). The source line is 154m long and the measurement distance 3 times the source length (a) or 7.5m (b). The vertical coordinate is the distance from the source line, the horizontal coordinate the distance from the source centre, both in meters. The horizontal directivity for wheel radiation is given by Eq. (1). The horizontal directivity for rolling noise at each wagon or bogie position is given by, just as an example, ΔLRH (ϕ ) = 10 lg 0.15 + 0.85 cos 2 (ϕ ) .

[

]

As has been shown in Ref. [10], the effective horizontal directivity of a line of stationary, or, slowly-moving, incoherent point sources is calculated using Eq. (4), with (a) ΔLW φ ij being the horizontal directivity assigned to the i -th wagon or bogie

( )

which has an open angle of

δφij

to the receiver and (b)

φj

the j-th angle position of

the line source which is defined as the-source-middle-to-receiver line relative to the normal of the source line,

⎛ N ΔL (φ ) / 10 δφij ΔLW (φ j ) = 10 lg⎜⎜ ∑10 W ij Δφtotal ⎝ i =1

⎞ ⎟⎟ , ⎠

N

Δφtotal = ∑ δφij

(4)

i =1

7 Conclusions The radiation of railway wheels is not a dipole source, at least for wheels with a curved web. Its directivity is described by Eq. (1). For wheels with a straight web more investigations are needed.

432

X. Zhang

The vertical directivity of railway track radiation can also be described by Eq. (1) without losing accuracy. Concerning the horizontal directivity the radiation of a railway track is assumed to be a dipole source above the decoupling frequency of the track and a monopole source while below this decoupling frequency. The useful vertical directivity of railway rolling noise should include the shielding effect of the car-body, and the shielding effect of low barriers when they are relevant. These shielding effects dominate thus the inclusive vertical directivity can be determined with accuracy only by measurements. The horizontal directivity of railway rolling noise at an axle position is determined by the horizontal directivities of the wheel radiation and track radiation where the relevant track radiation has been equivalently concentrated at the axle position, and weighted by the relative importance between the two sub-sources which is described by the vehicle transfer function and track transfer function. The equivalent horizontal directivity of a line of incoherent sources like train pass-by noise will be determined in the procedure described in section 6. Two calculation examples of the equivalent horizontal directivity of a line of incoherent point sources are provided. In figure 2-a it is shown that at a long distance the equivalent directivity of a line of incoherent point sources will be the same as that assigned to each point source (when each point source is assigned the same directivity). In figure 2-b it is shown that, at a position close to the line source, directivity will have negligible effect on the measured sound pressure level. This result has two meanings: (1) one can measure the sound power level of train pass-by noise at a short distance with its directivity neglected; (2) a short measurement distance is not suitable to measure the directivity of a line source. To measure the horizontal directivity of a train pass-by noise a distance about the train length is proposed [10].

References [1] Talotte, C., et al.: Railway source models for integration in the new European noise prediction method proposed in Harmonoise. Journal of Sound and Vibration 293, 975–985 (2006) [2] Dittrich, M.: The Imagine source model for railway noise prediction. Acta Acustica united with Acustica 93(2), 185–200 (2007) [3] Remington, P.J.: Wheel/Rail Noise – Part I: Characterization of The Wheel/Rail Dynamic System. Journal of Sound and Vibration 46(3), 359–379 (1976) [4] Ten Wolde, T., van Ruiten, C.J.M.: Sources and mechanisms of wheel/rail noise: state-ofart and recent research. Journal of Sound and Vibration 87(2), 147–160 (1983) [5] Zhang, X., Jonasson, H.: Directivity of Railway Noise Sources. Journal of Sound and Vibration 293, 995–1006 (2006) [6] Thompson, D.J., Jones, C.J.C.: Sound radiation from a vibrating railway wheel. Journal of Sound and Vibration 253(2), 401–419 (2002) [7] Thompson, D.J., Jones, C.J.C., Turner, N.: Investigation into the validity of twodimensional models for sound radiation from waves in rails. Journal of Acoustical Society of America 113(4), 1965–1974 (2003) [8] Heng, C.C.: Vertical directivity of train noise. Applied Acoustics 51(2), 157–168 (1997) [9] Chew, C.H.: Vertical directivity pattern of train noise. Applied Acoustics 55(3), 243–250 (1998) [10] Zhang, X.: To determine the horizontal directivity of a train pass-by, in03_627, The Proceedings for Inter-noise 2003, Jeju, Korea (August 25-28, 2003)

Complex Eigenvalue Analysis of Railway Curve Squeal G.X. Chen, J.B. Xiao, Q.Y. Liu, and Z.R. Zhou Tribology Research Institute, National Traction Power Laboratory, Southwest Jiaotong University, Chengdu, 610031, China Tel.: +86-28-87600971; Fax: +86-28-87600971 [email protected]

Summary A finite element complex eigenvalue analysis of curve squeal is carried out. Two models for the wheel tread/rail top contact and the wheel flange root/rail gauge corner contact are established. In the both models, the contact between the wheel and rail is simulated with a spring. The friction force is considered as the contact spring force multiplied by a coefficient of friction. The rail is supported by vertical and lateral springs at each sleeper. The simulation result shows that the coefficient of friction, contact point positions and stiffness of the rail support spring all have distinct influences on the curve squeal occurrence. Appropriate stiffness of the rail support spring can suppress or eliminate curve squeal.

1 Introduction When a train negotiates a tight curve, it often emits an intense unpleasant noise which is termed squeal. In 1976, Rudd proposed an excitation mechanism of curve squeal related to stick-slip phenomena in the lateral direction on the wheel/rail interface [1]. Since then, almost all investigators accepted Rudd’s theory [2-6]. Several remedies were reported to be capable of suppressing curve squeal to some extent [4]. However, there is no method to completely eliminate the squealing noise up to now. The difficulty lies in the random occurrence of curve squeal and that the mechanism of curve squeal generation is partially unknown. It was reported that in a small scale model test curve squeal can still occur without negative friction-velocity slope [7]. Moreover, the stick-slip mechanism can not be used to satisfactorily explain why squeal does not occur until the coefficient of friction arrives at a larger value, why the tangential vibration and normal vibration of the curve squeal system are always coupled dynamically, and why the curve squeal more easily occurs with the inner wheel of a trailing wheelset [8]. All these problems suggest that there is still much work needed to be done for obtaining a complete solution to curve squeal. In the present paper, the finite element complex eigenvalue analysis is applied to study the curve squeal. It is a main method used to predict and suppress squeal in the design stage at the present time [9]. The object of the present work is to extend understanding of the generation mechanism of curve squeal and to explore more solutions to curve squeal. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 433–439, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

434

G.X. Chen et al.

2 Modeling of Curve Squeal 2.1 Interaction between the Wheel and Rail Generally, wheels are always kept in contact with rails when a train travels on a track. Dynamic simulations of a vehicle negotiating a curve and field measurements all demonstrate that on curved tracks the leading wheelsets of both the front and rear trucks have positive attack angles and the trailing wheelsets have positive or negative attack angles mainly depending on the running speed, as shown in Fig. 1 [10]. For a wheelset with a positive attack angle, the lateral creep forces have defi- nite directions as shown in Fig. 2. Dynamic simulation results also show that when a vehicle negoti- ates a severe curved track, the lateral creep force of the leading wheelset becomes saturated general- ly, that is, equal to the normal force multiplied by the friction coefficient. If a wheelset is one with- out driving power or braking power, it has a small longitudinal

Fig. 1. Wheelsets on a curved track

(a)

(b)

Fig. 2. The positions of contact points and directions of lateral creep forces, F and N denote lateral creep force and normal force, respectively. Subscript l and r represent outer wheel/rail and inner wheel/rail. (a) Wheel flange root/rail gauge corner contact for the outer wheel and rail, (b) wheel tread/rail top contact for the inner wheel and rail.

Complex Eigenvalue Analysis of Railway Curve Squeal

435

creep force. In curved tracks, the contact point between the outer wheel of a wheelset and rail will shift to the flange root for wheel and the gauge corner for rail, while the contact point between the inner wheel and rail is roughly kept in the vicinity of the center of rail top for rail and the tread for wheel, as shown in Fig. 2. 2.2 Finite Element Modeling of Curve Squeal The motion equation of the wheel and rail system without structural damping and friction at the contact interface is written as follows:

&& + Ku = 0 Mu

(1)

where M is a matrix of mass, K is a matrix of stiffness and u is a vector of node displacements. Since both M and K are symmetrical matrices, the real parts of eigenvalues of Equation (1) are always equal to zero. In the presence of friction, K becomes an asymmetrical matrix. In the case, one or some of the real parts of eigenvalues of Equation (1) are probably more than zero, which are the criterion used to identify unstable vibration of systems. In the present paper, the lateral creep force is assumed to reach saturation, that is, equal to the normal force multiplied by the friction coefficient. Lateral creep forces on the wheel and rail contact surfaces are proportional to the normal contact forces, which in turn may vary with dynamic response. They are given as follows:

Fw = − Fr = μN

(2)

where Fw and Fr are the lateral creep force acting on the wheel and one acting on the rail, respectively. μ is the coefficient of friction. N is the normal compression force variable. The normal force may be simulated with a contact spring. It is given as follows:

N = − k × (uwy − ury )

(3)

where k is the stiffness of contact spring, u wy and u ry are the displacements of wheel and rail, respectively at the consistent contact node in the normal direction,. Combining Equation (2) and Equation (3), we obtain the following matrix relationship:

⎧ Fw ⎫ ⎡− 1 1 ⎤ ⎧u wy ⎫ ⎨ ⎬ ⎨ ⎬ = μk ⎢ − 1⎥⎦ ⎩ury ⎭ ⎩ Fr ⎭ ⎣1

(4)

All lateral creep forces at contact nodes may be written:

Ff = K f u

(5)

where F f is a vector of lateral creep forces and K f is a friction stiffness matrix. Taking the friction forces at all contact nodes into account, Equation (1) may be rewritten as follows:

&& + (K − K f )u = 0 Mu

(6)

In the present paper, a finite element tool NASTRAN is applied. It provides the ability to conveniently generate stiffness matrix K − K f and to solve complex eigenvalues.

436

G.X. Chen et al.

(a)

(b)

Fig. 3. Models of curve squeal, (a) finite element model for the outer wheel/rail contact as shown in Fig. 2a, (b) finite element model for the inner wheel/rail contact as shown in Fig. 2 b

From Fig. 2, it is seen that there are two distinct different contact conditions for left and right rails. Therefore, two contact conditions for left and right rails are modeled respectively as shown in Fig. 3. In the models, the rail is 5000 mm in length. The distance between two sleepers is 580 mm. The width of each sleeper is 170 mm. The rail is supported by a lateral spring and a vertical one at every sleepers location. The wheel of freight cars is 840 mm in nominal diameter. The rail is a type of 60 kg/m. The maximum wheel load is about 120 kN. The radii of Hertz’s contact ellipse are calculated as about a=9.94 mm (along the longitudinal direction of track) and b=4.4 mm. These two values are very small in respect to the wheel size. Therefore, the contact between the wheel and rail may be considered as a point contact in the finite element modeling of curve squeal. Fig. 3 shows two meshed models corresponding to the outer wheel/rail contact and the inner wheel/rail contact, respectively. 2.3 Nominal Parameter Values of Curve Squeal System The stiffness of the contact spring between the wheel and rail is set to 1.2×109 N/m. The density of the wheel and rail materials is 7800 kg/m3. The Youg’s modulus of the wheel and rail materials is 2.1×1011 N/m. The lateral stiffness kl of rail support spring is 50×106 N/m. The vertical stiffness kv is 100×106 N/m.

3 Results 3.1 Influence of the Friction Coefficient on Curve Squeal The real parts of eigenvalues of Eq. (6) are generally equal to zero. If there is a positive real part, there must be a conjugated negative real part. That suggests that two specific

Complex Eigenvalue Analysis of Railway Curve Squeal

437

adjacent frequencies merge. Therefore, one can identify that there is a positive real part according to the mergence of the adjacent frequencies [9, 11]. Fig. 4 shows the variation of the mode frequencies with the friction coefficient. It indicates all unstable frequencies and their adjacent frequencies for the prescribed parameters. From Fig. 4, it is seen that there are two unstable frequencies. One is about 3248 Hz and occurs when the coefficient of friction is equal to 0.5 or more. Another is about 4354 Hz and arises when the coefficient of friction is equal to 0.1 or more. It is also seen that when the vibration becomes unstable two specific adjacent frequencies merge to create an unstable frequency.

(a)

(b)

Fig. 4. Variation of the frequencies with the friction coefficient using the model of Fig. 3b, (a) at the vicinity of 3246 Hz, (b) at the vicinity of 4355 Hz

3.2 Influence of the Different Wheel/Rail Contacts on Curve Squeal Fig. 5 shows the variation of mode frequencies with the coefficient of friction for the outer wheel/rail contact. From Fig. 5a, it is seen that there is a chattering noise whose frequency is about 388.4 Hz when the coefficient of friction is equal to 0.25 or more.

(a)

(b)

Fig. 5. Variation of the frequencies with the friction coefficient using the model of Fig. 3a, (a) at the vicinity of 390 Hz, (b) at the vicinity of 4050 Hz

438

G.X. Chen et al.

From Fig. 5b, it is found that there is a squeal when the coefficient of friction is equal to 0.4 or more because there is a mergence of adjacent frequencies. Comparing Fig. 5b with Fig. 4b, it can be found that the curve squeal more easily occurs under the condition of inner wheel/rail contact than under the condition of outer wheel/rail contact. That result is consistent with what was found in tests [8]. It explains why curve squeal more easily occurs under the inner wheel/rail contact. 3.3 Effect of the Stiffness of the Rail Support Spring on Curve Squeal Fig. 6 shows the variation of mode frequencies with the coefficient of friction for the inner wheel/rail contact when the stiffness of the rail support spring is changed to 1.5 times of the nominal values. From Fig. 6, it is seen that there is an unstable squeal when the coefficient of friction is equal to 0.4 or more. Comparing Fig. 6 with Fig. 4, it is found that the stiffness of the rail support spring has a distinct influence on the occurrence of curve squeal. When the stiffness of the rail support spring is increased to 1.5 times of the nominal value, the coefficient of friction needed to induce curve squeal is increased from 0.1 to 0.4. That suggests that the propensity of curve squeal occurrence is clearly suppressed. The result indicates that changing the stiffness of the rail support spring can also suppress and eliminate curve squeal in addition to the ring damping solution [4].

Fig. 6. Variation of the frequencies with the friction coefficient, when kv=150 MN/m and kl=75 MN/m using the model of Fig. 3b

4 Conclusions 1.

The coefficient of friction between the wheel and rail has an important effect on the occurrence of curve squeal. The propensity of the curve squeal occurrence is increased with increasing coefficient of friction.

Complex Eigenvalue Analysis of Railway Curve Squeal

2.

3.

439

The position of wheel/rail contact point has a clear influence on the occurrence of curve squeal. The curve squeal more easily occurs under the wheel tread/rail top contact than under the wheel flange root/rail gauge corner contact. The stiffness of the rail support spring has a significant effect on the occurrence of curve squeal. Changing the stiffness of the rail support spring may suppress or eliminate curve squeal.

Acknowledgement The authors gratefully thank the financial supports from National Natural Science Foundation of China (no. 50521503 and no. 50675183).

References [1] Rudd, M.J.: Wheel/rail noise, part II: wheel squeal. Journal of Sound and Vibration 46(3), 381–394 (1976) [2] Heckl, M.A., Abrahams, I.D.: Curve squeal of train wheels, part 1: mathematical model for its generation. Journal of Sound and Vibration 229(3), 669–693 (2000) [3] Chiello, O., Ayasse, J.B., Vincent, N., Koch, J.R.: Curve squeal of urban rolling stock-Part 3: Theoretical model. Journal of Sound and Vibration 293, 710–727 (2006) [4] Brunel, J.F., Dufrenoy, P., Nait, M., Munoz, J.L., Demilly, F.: Transient models for curve squeal noise. Journal of Sound and Vibration 293, 758–765 (2006) [5] de Beer, F.G., Janssens, M.H.A., Kooijman, P.P.: Squeal noise of rail-bound vehicles influenced by lateral contact position. Journal of Sound and Vibration 267(3), 497–507 (2003) [6] Van Ruiten, C.J.M.: Mechanism of squeal noise generated by trams. Journal of Sound and Vibration 120(2), 245–253 (1988) [7] Koch, J.R., Vincent, N., Chollet, H., Chiello, O.: Curve squeal of urban rolling stock-Part 2: Parametric study on a 1/4 scale test rig. Journal of Sound and Vibration 293, 701–709 (2006) [8] Vincent, N., Koch, J.R., Chollet, H., Guerder, J.Y.: Curve squeal of urban rolling stock-Part 1: State-of- the-art and field measurements. Journal of Sound and Vibration 293, 710–727 (2006) [9] Ouyang, H., Nack, W., Yuan, Y., Chen, F.: Numerical analysis of automotive disc brake squeal: a review. Int. J. Vehicle Noise Vib. 1(3/4), 207–231 (2005) [10] Guangxiong, C., Xincan, J.: Influence of periodic irregularities on wheel climb derailment safety of a freight car running on a transition curve. In: Proceedings of the 2000 ASME/ IEEE Joint Railroad Conference, pp. 19–29 (2000) [11] Huang, J.C., Krousgrill, C.M., Bajaj, A.K.: Modeling of automotive drum brakes for squeal and parameter sensitivity analysis. Journal of Sound and Vibration 289(1-2), 245–263 (2006)

Wave Propagation in Railway Tracks at High Frequencies J. Ryue1, D.J. Thompson1, P.R. White1, and D.R. Thompson2 1

Institute of Sound and Vibration Research, University of Southampton, Southampton, S017 1BJ, UK Tel.: +44 (0)23 8059 4930; Fax: +44 (0)23 8059 3190 [email protected] 2 Balfour Beatty Rail Technologies, Midland House, Nelson Street, Derby, DE1 2SA, UK

Summary In terms of the long range rail inspection, rail vibration which can propagate over long distances along rails may be a useful tool to detect rail defects. In order to understand long range wave propagation in railway tracks, it is required to identify how far vibration can travel along a rail. To answer this question, the attenuation characteristics of all propagating waves should be determined. In this work decay rates of propagating waves are investigated for frequencies up to 80 kHz. The Wavenumber Finite Element (WFE) method is used to represent a track which has a rail on a continuous foundation. Different damping loss factors are introduced in this model for the damping in the rail and in the foundation. By this simulation, the efficient wave types and their decay rates are predicted. These are presented in terms of what is measurable on various regions of the rail cross-section. In order to validate the simulation results, an experiment was performed on an operational railway track. The measured results are presented for comparison with the simulated ones and good agreement between them is found.

1 Introduction At low frequencies a railway track can be modelled as a simple beam on an elastic support, but at high frequencies, above about 1.5 kHz, deformation of the crosssection occurs and multiple wave types are sustained [1]. For the purpose of long range rail inspection, it was suggested in literature [2,3] that frequencies up to 80 kHz need to be considered. It was also identified that several tens of different wave types can occur in this higher frequency range making the dispersion curves very complicated. In terms of the numerical approach to track modelling, the wavenumber finite element method (WFEM) has been used by several authors, particularly for high frequency analysis [3-6]. However, most of the previous work was confined to obtaining dispersion relations and group velocities in undamped free rails and did not present what type of waves are most efficient for propagating further. Most recently, using this method, Bartoli et al. [7] presented decay rates of each wave propagating along a damped free rail by accounting for material damping up to 50 kHz. However, since they disregarded the contribution of the rail pad, the predicted decay rates seem to be unrealistic from a practical point of view. B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 440–446, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Wave Propagation in Railway Tracks at High Frequencies

441

As an experimental method for the high frequency region, Lanza di Scalea et al. [8] applied an impulse excitation to the rail and measured the direct and echo signals reflected from the opposite end of a 7.2 m long rail section. Comparing the direct and reflected signals, they extracted the frequency-dependent attenuation up to 50 kHz for a free rail. Rose et al. [2] measured a decay rate of about 0.56 dB/m at around 60 kHz at the top of the rail head. However, detailed information on the wave types effective for long range propagation and contribution of the rail foundation on decay rates in the high frequency region is lacking in these previous works. This study aims to investigate the propagation of waves in a railway track up to 80 kHz to answer a question, how far along a rail can vibration travel? To find an answer for this question, each propagating wave's attenuation characteristics are required. As a numerical method, the WFE method is employed to predict decay rates. Then a single quantity is introduced to determine the measurable waves at different positions on the rail surface using the simulated results. Finally, in order to validate the simulated results, an experiment is conducted on an operational track. The results are compared with the simulated ones.

2 Wavenumber Finite Element Analysis of a Railway Track 2.1 Track Model In this work, UIC60 rail is modelled. Since the UIC60 rail has a symmetric crosssection, only half of the width is included as shown in Fig. 1. This has an additional advantage of separating the possible waves into two groups that are uncoupled from each other. That is, a symmetric boundary condition that constrains the deformation of the mid-plane in the y direction gives the vertical and symmetric longitudinal modes; an antisymmetric boundary condition in which x and z directions are constrained in the mid-plane gives the lateral and torsional modes as well as the antisymmetric longitudinal modes. Only the rail pad is included in this model as a continuous foundation.

Fig. 1. The cross-sectional FE model of a rail on foundation for WFE analysis. The shaded elements represent the rail pad.

442

J. Ryue et al.

The stiffness of the single rail pad was chosen initially to be 150 MN/m in the vertical direction and 20 MN/m in the lateral and longitudinal directions. At high frequencies, however, the stiffness of the rail pad usually becomes much higher than the static or low frequency stiffness. From a one-dimensional foundation model, this ‘dynamic stiffness’ is expected to be strongly frequency dependent. For simplicity a constant value 10 times larger than the nominal values given above is used. This is expected to be appropriate for frequencies around 20 kHz. 2.2 Predicted Decay Rates In the WFE method, the equation of motion for an undamped structure is expressed as

~ {K − ω 2 M}W = 0 , 2

2

10

10

1

1

10

Decay Rate (dB/m)

Decay Rate (dB/m)

10

0

10

−1

10

−2

0

0

10

−1

10

10

(1)

−2

10

20

30

40

50

60

70

80

10

0

10

20

30

40

50

60

70

80

Frequency (kHz)

Frequency (kHz)

(a)

(b)

Fig. 2. The decay rates predicted by using the measured structural damping for (a) the vertical and symmetric longitudinal waves, (b) the lateral, torsional and antisymmetric longitudinal waves

where K = K 2 (− jκ ) 2 + K 1 (− jκ ) + K 0 and K 2 , K 1 and K 0 are stiffness ma-

~

trices, M is the mass matrix of the cross-section, and W contains the displacements of the cross-section which define the deformation shapes of waves [9, 10]. The track model shown in Fig. 1 has two damping components: the damping in the rail, η r , and in the foundation, η p . The stiffness matrix of this track model can be separated as

K d = (1 + iη r )K r + (1 + iη p )K p ,

(2)

where K d indicates the complex stiffness matrix, K r and K p denote the stiffness matrices for the rail and foundation, respectively. Then a damped wavenumber, κ d , can be expressed as

Wave Propagation in Railway Tracks at High Frequencies

κd =κ +

∂κ d ∂η r

ηr + η r =0

∂κ d ∂η p

ηp .

443

(3)

η p =0

For this model, the damped wavenumber is obtained as

~ ~ W H (K rη r + K pη p ) W , κd =κ − j ~ ~ W H K ′W

(4)

where K = K r + K p . Finally the decay rates of waves propagating along the track model can be evaluated using Δ = -8.686 Im( κ ). In this track model, damping of the rail pad, η p , was chosen as 0.2 which corresponds to typical values [11,12]. In general, the damping of the rail, η r , is much smaller than that of the rail pad. For example, the material damping loss factor of steel is about 0.0002. It was found from the simulation that at frequencies above 20 kHz the lower limits of the decay rates are directly associated only with the structural damping of the rail. Therefore, structural damping of the rail is a principal factor in determining long range wave propagation. In order to improve the simulated decay rates, structural damping loss factors were measured for several rail samples. For the rest of the work reported here, the frequency-dependent damping loss factor approximated from the measured results is used for η r to improve the accuracy of the simulated decay rates. The decay rates predicted are illustrated in Fig.2. From Fig.2, it can be seen that the minimum decay rates occur at frequencies between 20 kHz and 40 kHz. 2.3 Prediction of Measurable Waves on Rail Surface There are several tens of different waves propagating below 80 kHz in the track model used here. Since dynamic responses are usually measured normal to the rail surface, the energy distributions around the rail surface can provide useful information on which waves are measurable in a specific region on the rail surface. For this purpose, three separate regions were specified on the rail surface. These are the top and side of the rail head and the middle of the web. Then the normalized energies for each region and each direction were predicted using

⎛ 1 Q j ,z = ⎜ ⎜ nj ⎝ ~

∑ nj

~ 2⎞ Wz , j ⎟ ⎟ ⎠

⎛1 ⎜ ⎜N ⎝

∑W ~

s

N

2

⎞ ⎟, ⎟ ⎠

(5)

~

where Ws denotes displacements at nodes on the rail surface, Wz, j is displacement in the z direction (and similarly for the y direction) at nodes belonging to the region j, n j is the number of nodes in the region j and N is the total number of nodes on the rail surface. Therefore, this quantity should depict implicitly which waves are measurable in region j among all the waves in the system. The measurable decay rates predicted using Eq.(5) will be presented later, comparing them with the measured ones.

444

J. Ryue et al.

3 Decay Rate Measurement on an Operational Track In order to validate the simulated decay rates, a field measurement was performed on the Up Slow track of the British West Coast Main Line in August 2006. Three accelerometers were used to detect the waves at different points on the rail cross-section: the field side of the rail head, the underside of the rail head and the middle of the rail web. The underside of the rail head was shown in numerical simulations to give very similar responses to the top of the rail head and this was validated from a test track measurement. All accelerometers were mounted at a position midway between sleepers using glue. Decay rates can be obtained from the level difference between two train positions by dividing it by the train's running distance between them. The averaged decay rates obtained from the signals for all trains measured are shown in Fig. 3, together with the previous simulated ones. In the simulated graphs, all curves shown in Fig. 2 are present but the strength of the line represents the level of the normalized energy at each frequency according to Eq.(5). The darker curves are the more detectable on the corresponding rail surface. The measured decay rates below 10 kHz were disregarded in this comparison because the waves decay too rapidly in this frequency band, leaving 1

1

10

Decay Rates (dB/m)

Decay Rates (dB/m)

10

0

10

−1

10

−2

10

0

0

10

−1

10

−2

10

20

30

40

50

60

70

80

10

0

10

20

30

Frequency (kHz)

40

50

60

70

80

Frequency (kHz)

(a)

(b) 1

Decay Rates (dB/m)

10

0

10

−1

10

−2

10

0

10

20

30

40

50

60

70

80

Frequency (kHz)

(c) Fig. 3. Decay rates obtained from the field test, presented with the previous simulated results. (a) At the side of the rail head, (b) at the underside of the rail head, (c) at the middle of the web.

Wave Propagation in Railway Tracks at High Frequencies

445

just the background noise. The field test results agree very well with the simulated ones for all frequencies greater than 10 kHz. The slight difference between the measured and simulated results between 10 kHz and 20 kHz in Fig. 3(b) may be associated with the rail geometry or the stiffness of the rail pad. The exact types of the rail and rail pads in the operational track are not known. Nevertheless, the measured decay rates show extremely good agreement with the predicted ones particularly between 20 kHz and 50 kHz. In addition, from the simulated decay rate diagrams, the deformation shapes of dominantly measurable waves are obtained as illustrated in Fig. 4 at around 25 kHz. These deformation shapes verify that the respective types of these waves are the vertical bending wave localized in the rail head, the lateral bending wave of the rail head with a global deformation of the web and the 1st order web bending wave.

(a)

(b)

(c)

Fig. 4. The simulated deformation shapes of the dominantly measurable waves (a) shown in Fig. 3(a), (b) shown in Fig. 3(b), (c) shown in Fig. 3(c)

4 Conclusions In this work, wave propagation along the railway track was examined in the frequency region up to 80 kHz by means of WFE analysis. The simulations were validated experimentally on the basis of decay rates, obtained on an operational track. Excellent agreement between predicted and measured decay rates was obtained. Finally, the question of how far along a rail can vibration travel? could be answered clearly from the simulated and experimental results. • • •

A localized head bending wave in the vertical direction travels efficiently through the rail head. For this wave, the minimum decay rate of about 0.04 dB/m occurs between 22 and 40 kHz. At the side of the rail head, the primary propagating wave is a lateral bending wave which has global deformation including the web and rail head. The minimum decay rate of this wave is about 0.04 dB/m and occurs between 22 and 35 kHz. The 1st order web bending wave propagates dominantly through the web, having the minimum decay rate of about 0.05 dB/m between 20 and 40 kHz.

446

J. Ryue et al.

Conclusively, if a 50 dB level reduction is assumed detectable in the rail vibration, the maximum propagating distances will be about 1.2 km at the rail head and about 1.0 km at the web. It has to be noted that these results may depend on the track structures and their material properties, although rail geometry does not vary greatly in many situations. Also it was found from the field test that service trains are very effective in exciting rail vibration even at high frequencies.

Acknowledgements This work was supported by Balfour Beatty Rail Technologies. The authors are grateful to Barny Daley of Network Rail for permission to access the operational track for this research.

References [1] Thompson, D.J.: Wheel-rail noise generation, part III: Rail vibration. Journal of Sound and Vibration 161, 421–446 (1993) [2] Rose, J.L., Avioli, M.J., Song, W.-J.: Application and potential of guided wave rail inspection. Insight 44, 353–358 (2002) [3] Rose, J.L., Avioli, M.J., Mudge, P., Sanderson, R.: Guided wave inspection potential of defects in rail. In: Proceedings of Railway Engineering 2002, 5th International Conference and Exhibition, London, UK (2002) [4] Guided Ultrasonics Ltd, Long range screening of rail using guided waves. In: Third Vehicle/Track Interaction Course, Emmanuel College, Cambridge, UK (2001) [5] Wilcox, P., Evans, M., Alleyne, D., Pavlakovic, B., Vine, K., Cawley, P., Lowe, M.: Long range inspection of rail using guided waves. In: Proceedings of Railway Engineering 2002, 5th International Conference and Exhibition, London, UK (2002) [6] Hayashi, T., Song, W.-J., Rose, J.L.: Guided wave dispersion curves for a bar with an arbitrary cross-section, a rod and rail example. Ultrasonics 41, 175–183 (2003) [7] Bartoli, I., Marzani, A., Lanza di Scalea, F., Viola, E.: Modeling wave propagation in damped waveguides of arbitrary cross-section. Journal of Sound and Vibration 295, 685– 707 (2006) [8] Lanza di Scalea, F., McNamara, J.: Measuring high-frequency wave propagation in railroad tracks by joint time-frequency analysis. Journal of Sound and Vibration 273, 637– 651 (2004) [9] Nilsson, C.-M.: Waveguide finite elements applied on a car tyre. Ph.D. Thesis, KTH, Stockholm (2004) [10] Finnveden, S.: Evaluation of modal density and group velocity by a finite element method. Journal of Sound and Vibration 273, 51–75 (2004) [11] Thompson, D.J., Vincent, N.: Track dynamic behaviour at high frequencies. part 1: Theoretical models and laboratory measurements. Vehicle System Dynamics Supplement 24, 86–99 (1995) [12] Vincent, N., Thompson, D.J.: Track dynamic behaviour at high frequencies. part 2: Experimental results and comparisons with theory. Vehicle System Dynamics Supplement 24, 100–114 (1995)

Stability and Transient Analysis in the Modelling of Railway Disc Brake Squeal X. Lorang1 and O. Chiello2 1

SNCF, Innovative & Research Department, PSF, 45 rue de Londres, 75379, Paris, France Tel.: +33 (0)1 53 42 92 28 [email protected] 2 INRETS, 25 av. F. Mitterrand, 69675 Bron cedex, France Tel.: +33 (0)4 72 14 24 05; Fax: +33 (0)4 72 37 68 37 [email protected]

Summary The paper deals with friction induced vibrations and especially with railway disc brake squeal. The first part of the paper is devoted to the strategy used to model the general problem of self-excited vibrations of a rotating disc in frictional contact with two pads. Unilateral contact conditions with Coulomb friction and constant friction coefficient are considered. In order to predict the occurrence of self-excited vibrations, a classical stability analysis is performed, which consists on computing the complex modes associated to the linearised problem. A common interpretation of the stability analysis is that frequencies of unstable complex modes correspond to squeal frequencies. To check this assumption, the behaviour of the solution far from the sliding equilibrium is determined by using a non linear transient analysis. Moreover, an expansion of the transient solution on the complex modes provided by the stability analysis helps us to highlight the role of the unstable modes. The second part of the paper focuses on the application of the approach to the TGV disc brake system. Results on complex modes are presented and compared with measurements. The interest of both analyses is discussed.

1 Formulation of the Problem The mechanism of the simplified disc brake system represented on figure 1 is considered. The rotation speed Ω of the disc is assumed to be constant and sufficiently small so that the gyroscopic terms and the volume forces induced by rotation may be neglected. An “eulerian” description is adopted. Unilateral contact with Coulomb friction conditions are taken into account. By using a finite element method, the nonlinear dynamics problem may be written in a discrete form as follows (see details in reference [3]):

[M ]{U&&}+ [C ]{U& }+ [K ]{U } = {F } + [Pn ]T {rn } + [Pt ]T {rt } {rn } = Projℜ ({rn } − α n ([Pn ]{U } − {g 0 })) {rt } = ProjC ({rt } − α t ([Pt ]{U& }+ {V })) −

B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 447–453, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

(1)

448

X. Lorang and O. Chiello

where [M], [C] and [K] denote the mass, Rayleigh damping and stiffness matrices whereas {U} and {F} represent the vectors of nodal displacements and external nodal forces. In addition, {rn} and {rt} denote the vectors of normal and tangential reactions forces at the contact nodes whereas [Pn] and [Pt] are projection matrices on the normal and tangential relative displacements between the disc and the pads at the contact nodes. C is the Coulomb cone and coefficients αn and αt must be positive. Finally, {g0} is the vector of initial gaps whereas {V} is the vectors of imposed sliding velocities due to the disc rotation speed Ω at the contact nodes.

2 Stability and Transient Analysis 2.1 Linear Stability Analysis System (1) is a set of non-linear differential equations characterised by a sliding equilibrium. Considering small regular perturbations that does not break the contact (bilateral contact), the frictional forces may be linearised and the evolution of the perturbations {U*} verifies (see [2]):

([M ] + f [M ]){U&& }+ ([C ] + f [C ] + f [C ]){U& }+ ([K ] + f [K ]){U }= [P ] {r } [P ]{U }= 0 *

f

*

f

e

T

*

f

n

* n

(2)

*

n

where [Mf], [Cf], [Kf] and [Ce] are non symmetrical matrices provided by the linearisation of the frictional forces, f is the friction coefficient and {rn*} denote the vector of perturbed normal reactions forces at the contact nodes. By eliminating the bilateral contact constraints, a non symmetrical linear system of equations is obtained and the stability of the equilibrium may then be deduced from a complex eigenvalue analysis of the system, providing complex modes and complex eigenvalue λi . A complex mode i is unstable if Re(λi)>0, which may happen since the system is non symmetric. A modal growth rate ζi=Re(λi)/Im(λi) may also be defined (physically equivalent to a negative modal damping). 2.2 Non Linear Transient Analysis In addition to the stability analysis, an implicit numerical resolution of the system of equations (1) may be performed. A time discretisation method is used here, resting on a former work of M. Jean and J.J. Moreau (see [3]). The θ-method allows one to avoid numerical problems at the time of an impact. Indeed, the instability of the sliding equilibrium may lead to strongly non-linear events like a separation followed by a shock, as well as stick/slip transitions. In order to introduce an inelastic shock law, one uses the modified version of the θ-method (see [4]). 2.3 Relations between Stability and Transient Analyses In order to understand the role of the unstable modes in the non linear part of the transient behaviour, it is proposed to expand the numerical transient solution on the basis of the complex modes. Taking into account the non symmetrical character of the matrices involved in the problem (2), it is necessary to introduce the complex modes

Stability and Transient Analysis in the Modelling of Railway Disc Brake Squeal

449

provided by transposed linearised problem. Indeed, the generalisation of the usual orthogonality conditions in the case of symmetrical matrices leads to bi-orthogonality conditions, and given by:

∀i ≠ j

{Li }T [A]{Φ j }= 0

(3)

where {φj} are the complex modes of the direct problem (2) and {Li} the complex modes of the transposed problem of (2). [A] is the mass matrix in state space variable (see [5]). This bi-orthogonality property allows one to compute the evolution of the complex amplitude of the mode j, βj(t), from the transient perturbation written in statespace variables {α*(t)}:

{L } [A]{α (t )} (t ) = {L } [A]{Φ } T

βj

*

j

(4)

T

j

j

with {α*(t)} provided by numerical resolution of system (1). The contribution of the mode j to the perturbed displacement and velocity fields can then computed from βj(t). Finally, the variation of the total perturbed energy of a mode j can be computed from these contributions.

3 Application to a Simplified Disc Brake System In this part, the simplified disc brake system of figure 1 is considered. The geometric and physical characteristics of the structures are given in table 1. The other parameters are f=0.35, Ω=2.5 rad/s and δ=3.33×10-6 m. A stability analysis is performed in the [0 15kHz] frequency range. Among the 100 computed complex modes, three modes are found unstable, called M1 (8583 Hz, ζ=0.05 %), M2 (9288 Hz, ζ=0.13 %) and M3 (10130 Hz, ζ=0.17 %). The transient solution is also calculated with θ = 0.5 Table 1. Physical and geometric characteristics

Young’s modulus E Poisson’s ratio ν Density ρ Damping param. α Damping param. β Ext. diameter Int. diameter Thickness

Disc 2.02×1011 Pa 0.29 7850 kg.m−3 0 s−1 7.510−9 s 0.6 m 0.2 m 0.04 m

Pads 4.5×109 Pa 0.3 5250 kg.m−3 108 s−1 1.4410−6 s 0.16 m

-t Pad P

z

Disc D

δ

Pads Disc

0.04 m δ

Fig. 1. Simplified brake system

disc

450

X. Lorang and O. Chiello

and time step Δt = 5×10-6 s. In order to study the influence of the initial conditions on the transient solution, 4 cases are considered (A to D) for which the initial contributions of the unstable modes are different. The contributions of the unstable modes to the total perturbed energy ETM1(t), ETM2(t) and ETM3(t) are represented on the figure 2 for the 4 cases. The different figures show that the stabilised solution is not dependent on the initial conditions. It may be observed that this stabilised solution is made up of 2 of the 3 unstable modes (M1 and M2). These computations show that the stability analysis is able to predict the prone-squeal modes but that a transient analysis is necessary to predict which modes remain in the stabilised solution. It also highlights the possible coexistence of several unstable modes in the self-sustained vibrations.

Case A

Case B

Case C

Case D

Fig. 2. Evolution of modal contribution to total perturbed energy ET(t) [J] for the different cases as a function of time [s]. ETM1(t), ◊ ETM2(t), O ETM3(t).

4 Stability Analysis of a TGV Brake System In this part, a finite element model of the TGV brake system (cf. fig 3) is considered. First, the modes of the disc in free conditions have been calculated and have been classified according to the direction of the dominant deformations. In particular, inplane circumferential modes Cn-m, in plane radial modes Rn-m and out-of-plane axial modes An-m may be distinguished where n and m denote respectively the number of nodal circles and nodal diameters.

Stability and Transient Analysis in the Modelling of Railway Disc Brake Squeal

451

A stability analysis has been performed. More than 800 complex modes have been calculated to reach an upper limit frequency of about 14 kHz. The corresponding growth factors are represented on figure 5. Two kinds of modes may be distinguished: the pad modes, for which the disc vibrations are very small, and the disc modes for which the vibrations of the disc are dominating. The corresponding mode shapes and frequencies of the disc modes are close to the modes of the disc in free conditions but rotate along the disc. Most of these modes are axial modes without nodal circles and with one nodal circle (see fig 3). Another mode is rather an in-plane mode (C0-2) but with some axial components.

Mode C0-2 : 6684 Hz - ζ=0.02%

Mode A1-6 : 9494 Hz - ζ=0.12% Mode A0-10 : 10040 Hz - ζ=0.23%

Fig. 3. Some unstable modes of the TGV brake F.E. model

Fig. 4. Experimental power Spectrum of the normal velocity [dB] (below, ref 1 [m/s]) and acoustic pressure [dB] (above, ref 20 [μPa])

452

X. Lorang and O. Chiello

Some experimental results have been obtained from a TGV in braking operation at about 10 km/h. The acoustic pressure at one centimetre from the disc and the axial vibratory velocity at a point on the disc surface have been measured during braking. Figure 4 shows that the disc is responsible for the emitted noise and that the vibrations are composed of 8 high frequencies from 5000 to 15000 Hz. Except the axial modes with one nodal circle, all the unstable modes are close to the experimental squeal frequencies.

Fig. 5. Growth rates ζ [%] of complex modes as a function of frequency [Hz] (O: disc mode ×: Pad modes)

5 Conclusion In this paper, the modelling strategy of disc brake squeal has been studied. In order to investigate the relations between stability and transient classical analyses, a new method has been proposed, which consists on expanding the transient vibratory field on the complex modes provided by the stability analysis. This method has been tested on a simplified disc brake model for various initial conditions. Results have shown that the stabilised solution is made up of two coexisting unstable modes and that this solution is the same for different initial conditions. A stability study has also been performed on a finite element TGV brake system and compared with vibration measurements. It has been found that the frequencies of the unstable disc modes correspond to most of the vibration frequencies.

Stability and Transient Analysis in the Modelling of Railway Disc Brake Squeal

453

References [1] Moirot, F., Nguyen, Q.S.: Brake squeal: a problem of flutter instability of the steady sliding solution? Arch. Mech. 52, 645–661 (2000) [2] Moirot, F.: Etude de la stabilité d’un équilibre en présence de frottement de coulomb. PhD Thesis, Ecole polytechnique, Palaiseau, France (1998) [3] Jean, M.: The non-smooth contact dynamics method. Comput. Methods Appl. Mech. Eng. 177, 235–257 (1999) [4] Vola, D., Pratt, E., Jean, M., Raous, M.: Consistent time discretization for a dynamical frictional contact problem and complementarity techniques. REEF 7 (1998) [5] Balmès, E.: Structural Dynamics Toolbox (2006), http://www.sdtools.com

Large-Scale Fatigue Test of Stone Wool Based Anti-vibration Mats K.B. Gatzwiller RockDelta a/s (ROCKWOOL a/s), Hovedgaden 584, 2640 Hedehusene, Denmark Tel.: +45 46 56 50 20; Fax: +45 46 56 50 80 [email protected]

Summary As the basis for an anti-vibration mat in a resiliently supported track, stone wool exhibit outstanding resistance to mechanical stress fatigue with virtually unchanged functional performance even after decades of use. In-situ, this has for instance been verified by an exhaustive 1996 Norwegian field study of the long-term functional performance and material characteristics of stone wool based anti-vibration mats that had been installed in a tunnel track section beneath the Oslo Cathedral in 1978. A recent large-scale laboratory test at the Danish Technical University – with control and final conclusion of the test by the Technical University of Munich, Chair and Institute for Road, Railway and Airfield Construction – have confirmed the durability of stone wool fibres when used in track anti-vibration applications. Based upon this laboratory test – that spanned more than 15 months and included a total of 100 million dynamic load cycles – this paper will present the important findings and discuss the results and various practical implications with special reference to the findings from the 1996 Norwegian field study.

1 Introduction As an effective means by which to mitigate the propagation of ground-borne vibrations, stone wool based anti-vibration mats have been employed successfully in track applications for more than three decades and remain among the popular materials for highefficiency, sheet-type, full-contact, resilient mats. This popularity probably stems from the fact that stone wool based track anti-vibration product lines, amongst other things, are designed to have a prolonged service life with constant functional performance and a long-term constant low-frequency dynamic stiffness even at high amplitudes. The following sections will draw attention to some of the most important historical aspects of qualifying and quantifying this prolonged service life, featuring, as it shall be demonstrated, a remarkable fatigue resistance against repeated dynamic loading.

2 Stone Wool Anti-vibration Mats: Important In-Situ Experiences 2.1 Long-Term Practical Experiences: Results from the 1996 Gardermobanan Research Project In 1996, as part of a preliminary investigation to determine the optimum method for vibration isolation for the “Gardermobanan” (the railway connection between the B. Schulte-Werning et al. (Eds.): Noise and Vibration Mitigation, NNFM 99, pp. 454–460, 2008. © Springer-Verlag Berlin Heidelberg 2008 springerlink.com

Large-Scale Fatigue Test of Stone Wool Based Anti-vibration Mats

455

Oslo city centre and the airport), NSB decided to conduct extensive research into the “total” behaviour of stone wool, taking into account several key factors, including: total cost structure, vibration isolation efficiency, spring and damping characteristics, environmental friendliness and – not least – the experiences that were gained from the 1978 Oslo Cathedral tunnel track. 2.1.1 Evaluation of Stone Wool for the 1996 Gardamobanan Project in Norway Related to the experiences gained during the first 18 years of operation (1978 to 1996), Gardermobanan A/S conducted – in cooperation with Rockwool A/S, Brekke & Strand Akustikk a/s, ViaNova and Norges Byggforskningsinstitutt – an exhaustive research into the behaviour of Rockwool stone wool when used as anti-vibration mats for the isolation of the Oslo Cathedral tunnel track [1]. 2.1.2 Oslo Cathedral Tunnel Track: Assessment of the Vibration Isolation Effectiveness The company Brekke & Strand Akustikk conducted the in-situ transmissibility based measurements used to provide a swift assessment of the vibration isolation effectiveness [2]. One piezoelectric accelerometer was used as a trigger device (triggering the measurement whenever a train would pass) and two piezoelectric accelerometers were used for the measurement task: one positioned at a track section without stone wool based vibration isolation and the other positioned at a track section with stone wool based vibration isolation. 2.1.3 Oslo Cathedral Tunnel Track: Level of Measured Ground-Borne Vibration Isolation It was concluded that the stone wool based mats gave significant isolation against structure-borne vibrations in the order of 8 to 12 dB in the frequency range from approximately 30 to 125 Hz. The stone wool mats are clearly seen on Fig. 1 as part of the track structure. 2.1.4 Oslo Cathedral Tunnel Track: Notes on Water, Frost, Fungi and Bacteria The mats were in outstanding condition. As expected, signs of neither fungi nor bacteria could be found. However, the excavated stone wool mat samples were remarkably wet when taken from the ground. A subsequent measurement revealed that the water absorption was varying between 0,2% and 52,9% (volume). It is also important to notice that the solution prescribed in 1978 was based upon a design where the stone wool mats would be laying directly on the concrete trough, and further, would be without any type of surface protection. This had obviously meant that the ballast stones had been free to mechanically impact the stone wool mats. 2.1.5 Oslo Cathedral Tunnel Track: NSB Maintenance Experiences over 18 Years Naturally, an important part of the project was to investigate whether or not the 18 years usage of the stone wool anti-vibration mats had had any impact on the quality, maintenance and total cost of ownership of the track. To that end, several important issues were noted; (1) The maintenance requirements for the isolated Oslo Cathedral tunnel track section, when compared to other non-isolated tracks, had shown no need for special maintenance care; (2) NSB informed that no “setting” of the ballast could

456

K.B. Gatzwiller

Fig. 1. The stone wool based anti-vibration mats are clearly seen here as part of the track structure. Note the forming of ballast degradation and sand deposits that – owing to the effective stone wool filter functioning – had no apparent effect on the functional performance of the mat. Furthermore, the mats were found to be in a very wet condition – up to approximately 50% water absorption was found. This, however, was not found to have an adverse effect on the level of vibration isolation. Extensive in-situ tests have confirmed this insensitivity to water [3].

be found with the Cathedral tunnel track section, and finally; (3) Maintenance during the 18 years period was reported by NSB to have been ballast tamping a total of three times – same three times as it had been for the non-isolated track sections.

3 Stone Wool Anti-vibration Mats: Important Laboratory Testing Experiences 3.1 Large-Scale Fatigue Test Designed to Determine Fatigue-Life Limit Early fatigue-life tests based upon the German Railways norm DB BN 918 071-1 clause 2.6 revealed practically no change in dynamic stiffness of the test specimen. This naturally led to the question of how many dynamic load cycles a stone wool based anti-vibration mat solution could actually sustain without changes to its stiffness properties? To that end, RockDelta turned to the Technical University of Munich, Chair and Institute for Road, Railway and Airfield Construction for a discussion on how a large-scale fatigue life test could be devised. 3.1.1 Test Setup at the Technical University of Denmark Based upon these discussions, a test scheme was carefully developed that ultimately would see the involvement of the following universities and test institutes; (1) The Technical University of Denmark, Laboratory of Building Physics and Design – housing the elaborate test setup and conducting the repeated dynamic loadings according

Large-Scale Fatigue Test of Stone Wool Based Anti-vibration Mats

457

to the test scheme devised [4]; (2) Ingemansson Technology AB of Sweden – performing the essential dynamic stiffness measurements [5] at test onset, during testing and at the end of the test, and finally; (3) The Technical University of Munich, Chair and Institute for Road, Railway and Airfield Construction – supervising the test and performing the important final test conclusion [6]. 3.1.2 Overall Test Design As shown on Fig. 2 (left side), the test setup centred around an Amsler press connected to an Amsler pulsator. An HBM pressure transducer measured the oil pressure applied to the press hence providing for a measure of the force applied to the test specimen, while a strain gauge was used to accurately measure displacement. Spanning a continuous test period of more than 15 months, and as shown on Fig. 5, the applied single frequency sinusoidal loading pattern was as follows; (a) for the first 10 million cycles the sinusoidal load fluctuated between 15 and 30 kN/m2; (b) for the next 2,5 million load cycles the sinusoidal load fluctuated between 15 and 40 kN/m2, and finally; (c) for the remaining 87,5 million cycles the applied sinusoidal load fluctuated again between 15 and 30 kN/m2. 3.1.3 Test Specimen Stiffness Measurements at Ingemansson Technology AB After a predefined number of load cycles – see Fig. 3 for details – the four test specimens were sent to Ingemansson Technology AB of Sweden who then conducted the dynamic stiffness measurements according to ISO 10846-2:1997 and dynamic bed modulus measurements according to German Railways norm DB BN 918 071-1 clause 2.4. Following each stiffness measurement, the test specimens were returned for continued repeated dynamic loading. 3.1.4 Dynamic Transfer Stiffness According to ISO 10846-2:1997 As shown in Fig.2 (right side), each RockXolid®50 test specimen was placed between two stiff metal plates, the upper plate driven by a servo-hydraulic actuator, the lower plate mounted on a load cell. These components are parts of a test rig which conforms to the ISO standard. A digital signal analyser and a control unit were used for the generation and measurement of a stepped sine wave in the range from 10 to 200 Hz. Constant velocity of 5 mm/s RMS was achieved through a control circuit which included a piezoelectric accelerometer carefully mounted on the upper plate and an integrating charge amplifier connected to the signal analyser. A static preload of 1 T/m2 and 3 T/m2 respectively was applied to the test specimen via the actuator. The results may be observed from Fig. 4 (left side). 3.1.5 Low Frequency Dynamic Bed Modulus According to DB BN 918 071-1 Clause 2.4 Using a similar setup as that described in the foregoing section – although using a displacement transducer instead of the piezoelectric accelerometer – a stationary sine force (1620 - 8100 N corresponding to a pressure of 0.02 - 0.10 N/mm2) was applied to the test specimen in accordance with the requirements of German Railways norm DB BN 918 071-1 clause 2.4. The results may be observed from Fig. 4 (right side).

458

K.B. Gatzwiller

3.1.6 Notes on Peak Loads and Measurement Uncertainty During the measurements at Ingemansson Technology AB, all four test specimens were exposed to approximately 3200 load cycles with a peak load reaching 100 kN/m2. A measurement uncertainty analysis (encompassing both Type A and Type B uncertainties) was established that included; (a) instrumentation uncertainties; (b) calibration uncertainties; (c) read-out uncertainties, and (d) temperature variation uncertainties. A careful analysis resulted in an overall uncertainty value of 8% (assuming normal distribution and 95% confidence level).

Fig. 2. Left side: A close-up view of the fatigue-life test setup. The four RockXolid® 50 test specimens – positioned side by side – can be seen at the bottom of the test trough. Each test specimen was of the following dimensions (length x width x height): 270 mm x 300 mm x 50 mm. Right side: Details of the dynamic stiffness (ref. ISO 10846-2) and low frequency dynamic bed modulus (ref. German Railways norm DB BN 918 071-1) test setup at Ingemansson Technology AB, showing one of the four RockXolid® 50 test specimen, force transducer, acceleration response transducer and the MTS servo hydraulic actuator. During the 15 months test period, i.e. until the 100 million dynamic load cycles were reached, a total of eight dynamic stiffness measurements and eight low frequency dynamic bed modulus measurements were conducted on each of the four test specimens at predefined load cycle intervals. Number of applied load cycles

Applied sinusoidal load

Dynamic Stiffness according to ISO 10846-2 (20-250 Hz). Time of measurement

Dynamic Bed Modulus according to DB BN 918 071 (1 – 20 Hz). Time of measurement

December 2002 January 2003 February 2003 March 2003 April 2003 July 2003 December 2003 March 2004

December 2002 January 2003 February 2003 March 2003 April 2003 July 2003 December 2003 March 2004

[kN/m2]

0 ( virgin measurement) 5 x 106 10 x 106 12,5 x 106 22,5 x 106 35 x 106 62,5 x 106 100 x 106

15 - 30 15 – 30 15 – 30 15 – 30 15 – 30 15 – 30 15 - 30

Fig. 3. The stiffness measurements conducted at Ingemansson Technology AB were according to ISO 10846-2 and German Railways norm DB BN 918 071-1 clause 2.4. This figure shows the number of predefined load cycles and the associated time of stiffness measurements spanning a period from December 2002 until March 2004.

Large-Scale Fatigue Test of Stone Wool Based Anti-vibration Mats

459

Fig. 4. Left side: Dynamic stiffness values according to ISO 10846-2 of one of the four RockXolid® 50 test specimens. With a single frequency, sinusoidal load fluctuating between 15 and up to 40 kN/m2, the non-stop durability test lasted for more than 15 months and reached 100 million load cycles. As can be seen from the figure, dynamic stiffness values (20 Hz, 50 Hz and 100 Hz with a preload of 30 kN/m2 shown as examples) at onset of the test and at the end of the test, show virtually zero change. Right side: Low frequency dynamic bed modulus measurements (3 Hz and 20 Hz) according to German Railways norm DB BN 918 071-1 clause 2.4 of one of the four RockXolid® 50 test specimens. The observations are in perfect accordance with those outlined under the left side of the figure, i.e. the dynamic bed modulus at onset of the test and at the end of the test show virtually no change.

Fig. 5. Left side: The force range (fluctuating between 15 kN/m2 and up to 40 kN/m2) can be seen as function of the number of load cycles up to the 100 million which marked the closing of the test. For 2,5 million load cycles, the maximum load reached 40 kN/m2 and for 97,5 million load cycles the maximum load reached 30 kN/m2. In all cases, the minimum (static) load applied never went below 15 kN/m2. Right side: The deflection of the test specimen (in mm) as function of the 100 million load cycles.

4 Conclusion By means of a large-scale laboratory fatigue-life test spanning more than 15 months, it has been demonstrated that a stone wool based anti-vibration solution, known under the brand name RockXolid® 50, exhibits virtually unchanged dynamic stiffness – according to both ISO 10846-2:1997 and German Railways norm DB BN 918 071-1 clause 2.4 – after being exposed to a total of 100 million repeated dynamic load cycles with a single frequency, sinusoidal load fluctuating between 15 and up to 40 kN/m2.

460

K.B. Gatzwiller

The result of this fatigue-life test was contrasted with an elaborate stone wool based anti-vibration mat evaluation report from 1996 by the Norwegian national railways. This report included; total cost structure, vibration isolation efficiency, fatigue life, spring and damping characteristics, environmental friendliness and the experiences gained from ballasted tunnel track section, built in 1978, beneath the Oslo Cathedral that had been equipped with stone wool based anti-vibration mats.

References [1] Rapport til NSB Gardermobanan A/S: Utvikling av metode for reduksjon av strukturstøy fra jernbane, delrapp. 2 [2] Using an FFT analyzer, transmissibility measurements (magnitude and phase difference between two accelerometers) was conducted via a Frequency Response Function setup. These in-situ transmissibility based measurements were used to provide an assessment into the vibration isolation effectiveness; however, such transmissibility measurements can obviously not be regarded as a true Insertion Loss measurement [3] Ingemansson Technology AB Test Report 13-00286-05042700: RockXolid® Wet Condition Tests [4] Traberg, S.: The Danish Technical University, BYG-test Report, 10th of March, BYGDTU SR-25338: Fatigue Test of RockDelta Vibration Isolation, Application for Mass Spring Systems in Railway Tracks (2004) [5] Svensson, J., Fredö, C.: Ingemansson Technology AB, report 13-00352-04030500, -03-05: RockXolid® 50 – Dynamic stiffness tests during 100 million cycles of ageing (2004) [6] Leykauf, Dr. Ing. G.: Technische Universität München, Lehrstuhl und Prüfamt für Bau von Landverkehrswegen, Forschungsbericht Nr 2167, 29.10.2004: Untersuchungen zur elastischen Steifigkeit und zum Frost/Tau-Verhalten einer RockXolid® – Mineralfasermatte

Author Index

Adedipe, A. 78 Ahmad, N. 299 Anderson, D. 101, 399 Asmussen, B. 229 Atkinson, K. 243 Auersch, L. 122, 129 Augusztinovicz, F. 115 Behr, W. 334 Bekaert, J. 19 Benton, D. 229 Beuving, M. 327 Bewes, O.G. 172 Bongini, E. 320 Bracciali, A. 257 B¨ uhler, S. 406 Carels, P. 193 Carman, R. 215 Cervello, S. 257, 334 Chen, G.X. 433 Chiello, O. 158, 447 Cobbing, C. 179 Cordier, J.F. 56 Cox, S.J. 78, 172, 201 Croft, Briony 363, 392 Czolbe, C. 412 Deeg, P. 40 Degen, K.G. 48 Degrande, G. 19, 108, 115, 136 Douarche, N. 56 Eisenlauer, M.

40

Farrington, D. 229 Fiala, P. 115 Fodiman, P. 363 Fran¸cois, S. 136 Frid, A. 419 Fukuda, T. 9 Garburg, R. 150 Gatti, P. 257

Gatzwiller, K.B. 454 Gautier, P.E. 71, 320 Gerbig, Ch. 48 Glickman, G. 215 G´ omez, J. 292 Gupta, S. 108, 115, 136 Habault, D. 320 Hanson, C.E. 26 Harrisson, M. 143 Hecht, M. 412 Hemsworth, B. 327 Herrmann, A. 86 Hieke, M. 40 Holm, P. 370 Horiuchi, M. 63 Huang, H. 201 Huang, Z.Y. 313 Hunt, H.E.M. 136, 341 Hussein, M.F.M. 136 Ichikawa, K. Iida, M. 9

1

Jaquet, T. 150 Jiang, J. 201 Jin, X. 271, 278 Jones, C.J.C. 172, 179, 208, 299, 306, 313, 363, 392 Jones, R.R.K. 327 Kalivoda, M.T. 250 Kaltenbach, H.-J. 40 Kanda, H. 1 Kelly, R. 193 Kitson, P. 229 Koh, H.I. 222 K¨ ostli, K.P. 208 Kropp, W. 285 Kuijpers, A.H.W.M. 348 Kurita, T. 63 Kurze, U.J. 186 Kwon, H.B. 222

462

Author Index

Le Hou´edec, D. 158 Leth, S. 419 L´etourneaux, F. 56, 71, 363 Li, Z.G. 94 Liepert, M. 186 Liu, L. 201 Liu, Q.Y. 433 Liu, Q. 271 Lombaert, G. 19, 108 Lorang, X. 447 Lutzenberger, S. 370 Maeda, T. 9 Maldonado, M. 158 Matt´ei, P.O. 320 Miyachi, T. 9 Moehler, U. 186 Molla, S. 320 Muhr, A.H. 299 M¨ uller, R. 264 Nagakura, K. 33 Nelson, J.T. 143 Nielsen, J.C.O. 285, 355 Nilsson, C.M. 306

Schaeffler, M. 215 Schevenels, M. 136 Sheng, X. 278 Stiebel, D. 229 Sun, J. 201 Takaishi, T. 33 Talbot, J.P. 136 Thallemer, B. 406 Thite, A.N. 306 Thompson, D.J. 208, 299, 306, 313, 392, 440 Thompson, D.R. 440 Tielkes, Th. 40 Toyooka, M. 33 Tsuda, H. 1 Vadillo, E.G. 292 van den Brink, J.W. 165 van Haaren, E. 378 van Keulen, G.A. 378 Verheijen, E. 165

Oertli, J. 236 Onnich, H. 48, 186 Ophalffens, K. 193 Orrenius, U. 419 Oyarzabal, O. 292 Ozawa, S. 9

Wagner, H.-G. 86 Wakabayashi, Y. 63 Wang, A. 78, 172, 201 Weber, C. 101, 243 Wen, Z. 271 Wheatley, N. 399 White, P.R. 440 Wu, T.X. 94, 384

Park, J.H. 222 Pettersson, M. 143 Pieringer, A. 285 Pinto, P. 193 Poisson, F. 56, 71, 320

Xiao, J.B. 433 Xiao, X. 278

Reyes, C. 215 Rikse, L. 136 Roovers, M.S. 165 R¨ ucker, W. 129 Ryue, J. 440 Sagawa, A. 33 Santamar´ıa, J. 292

Yamada, H. 63 Yamazaki, N. 33 Yano, H. 33 Yoshida, S. 1 You, W.H. 222 Zhang, X. 426 Zhou, Z.R. 433 Zhou, Z. 271

Notes on Numerical Fluid Mechanics and Multidisciplinary Design

Available Volumes Volume 99: Burkhard Schulte-Werning, David Thompson, Pierre-Etienne Gautier, Carl Hanson, Brian Hemsworth, James Nelson, Tatsuo Maeda, Paul de Vos (eds.): Noise and Vibration Mitigation for Rail Transportation Systems - Proceedings of the 9th International Workshop on Railway Noise, Munich, Germany, 4–8 September 2007. ISBN 978-3-540-74892-2 Volume 98: Ali Gülhan (ed.): RESPACE – Key Technologies for Reusable Space Systems - Results of a Virtual Institute Programme of the German Helmholtz-Association, 2003–2007. ISBN 978-3-540-77818-9 Volume 97: Shia-Hui Peng, Werner Haase (eds.): Advances in Hybrid RANS-LES Modelling - Papers contributed to the 2007 Symposium of Hybrid RANS-LES Methods, Corfu, Greece, 17–18 June 2007. ISBN 978-3-540-77813-4 Volume 96: C. Tropea, S. Jakirlic, H.-J. Heinemann, R. Henke, H. Hönlinger (eds.): New Results in Numerical and Experimental Fluid Mechanics VI - Contributions to the 15th STAB/DGLR Symposium Darmstadt, Germany, 2006. ISBN 978-3-540-74458-0 Volume 95: R. King (ed.): Active Flow Control - Papers contributed to the Conference “Active Flow Control 2006”, Berlin, Germany, September 27 to 29, 2006. ISBN 978-3-540-71438-5 Volume 94: W. Haase, B. Aupoix, U. Bunge, D. Schwamborn (eds.): FLOMANIA - A European Initiative on Flow Physics Modelling - Results of the European-Union funded project 2002 - 2004. ISBN 978-3-54028786-5 Volume 93: Yu. Shokin, M. Resch, N. Danaev, M. Orunkhanov, N. Shokina (eds.): Advances in High Performance Computing and Computational Sciences - The Ith Khazakh-German Advanced Research Workshop, Almaty, Kazakhstan, September 25 to October 1, 2005. ISBN 978-3-540-33864-2 Volume 92: H.J. Rath, C. Holze, H.-J. Heinemann, R. Henke, H. Hönlinger (eds.): New Results in Numerical and Experimental Fluid Mechanics V - Contributions to the 14th STAB/DGLR Symposium Bremen, Germany 2004. ISBN 978-3-540-33286-2 Volume 91: E. Krause, Yu. Shokin, M. Resch, N. Shokina (eds.): Computational Science and High Performance Computing II - The 2nd Russian-German Advanced Research Workshop, Stuttgart, Germany, March 14 to 16, 2005. ISBN 978-3-540-31767-8 Volume 87: Ch. Breitsamter, B. Laschka, H.-J. Heinemann, R. Hilbig (eds.): New Results in Numerical and Experimental Fluid Mechanics IV. ISBN 978-3-540-20258-5 Volume 86: S. Wagner, M. Kloker, U. Rist (eds.): Recent Results in Laminar-Turbulent Transition - Selected numerical and experimental contributions from the DFG priority programme ’Transition’ in Germany. ISBN 978-3-540-40490-3 Volume 85: N.G. Barton, J. Periaux (eds.): Coupling of Fluids, Structures and Waves in Aeronautics - Proceedings of a French-Australian Workshop in Melbourne, Australia 3-6 December 2001. ISBN 978-3-54040222-0 Volume 83: L. Davidson, D. Cokljat, J. Fröhlich, M.A. Leschziner, C. Mellen, W. Rodi (eds.): LESFOIL: Large Eddy Simulation of Flow around a High Lift Airfoil - Results of the Project LESFOIL supported by the European Union 1998 - 2001. ISBN 978-3-540-00533-9 Volume 82: E.H. Hirschel (ed.): Numerical Flow Simulation III - CNRS-DFG Collaborative Research Programme, Results 2000-2002. ISBN 978-3-540-44130-4

Volume 81: W. Haase, V. Selmin, B. Winzell (eds.): Progress in Computational Flow Structure Interaction Results of the Project UNSI, supported by the European Union 1998-2000. ISBN 978-3-540-43902-8 Volume 80: E. Stanewsky, J. Delery, J. Fulker, P. de Matteis (eds.): Drag Reduction by Shock and Boundary Layer Control - Results of the Project EUROSHOCK II, supported by the European Union 1996-1999. ISBN 978-3-540-43317-0 Volume 79: B. Schulte-Werning, R. Gregoire, A. Malfatti, G. Matschke (eds.): TRANSAERO - A European Initiative on Transient Aerodynamics for Railway System Optimisation. ISBN 978-3-540-43316-3 Volume 78: M. Hafez, K. Morinishi, J. Periaux (eds.): Computational Fluid Dynamics for the 21st Century. Proceedings of a Symposium Honoring Prof. Satofuka on the Occasion of his 60th Birthday, Kyoto, Japan, 15-17 July 2000. ISBN 978-3-540-42053-8 Volume 77: S. Wagner, U. Rist, H.-J. Heinemann, R. Hilbig (eds.): New Results in Numerical and Experimental Fluid Mechanics III. Contributions to the 12th STAB/DGLR Symposium, Stuttgart, Germany 2000. ISBN 978-3-540-42696-7 Volume 76: P. Thiede (ed.): Aerodynamic Drag Reduction Technologies. Proceedings of the CEAS/DragNet European Drag Reduction Conference, 19-21 June 2000, Potsdam, Germany. ISBN 978-3-540-41911-2 Volume 75: E.H. Hirschel (ed.): Numerical Flow Simulation II. CNRS-DFG Collaborative Research Programme, Results 1998-2000. ISBN 978-3-540-41608-1 Volume 66: E.H. Hirschel (ed.): Numerical Flow Simulation I. CNRS-DFG Collaborative Research Programme. Results 1996-1998. ISBN 978-3-540-41540-4