Brittle Matrix Composites 9

matrix2009.qxd 9/19/2009 11:55 AM Page 1 BRITTLE MATRIX COMPOSITES 9 matrix2009.qxd 9/19/2009 11:55 AM Page 2 ...

Author: A.M. Brandt | V.C. Li | I.H. Marshall

93 downloads 1732 Views 27MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

matrix2009.qxd

9/19/2009

11:55 AM

Page 1

BRITTLE MATRIX COMPOSITES 9

matrix2009.qxd

9/19/2009

11:55 AM

Page 2

Proceedings of the Ninth International Symposium on Brittle Matrix Composites BMC9, held in Staszic Palace, Warsaw, Poland, 25-28 October 2009

Also published previously: Brittle Matrix Composites 1 Proceedings of the 1st International Symposium and EUROMECH Colloquium 204, Elsevier Science Publishers, 1985 Brittle Matrix Composites 2 Proceedings of the 2nd International Symposium, Elsevier Science Publishers, 1988 Brittle Matrix Composites 3 Proceedings of the 3rd International Symposium, Elsevier Science Publishers, 1991 Brittle Matrix Composites 4 Proceedings of the 4th International Symposium, Woodhead Publ. Ltd. (Cambridge) and Inst. of Economic Education (Warsaw), 1994 Brittle Matrix Composites 5 Proceedings of the 5th International Symposium, Woodhead Publ. Ltd. (Cambridge) and BIGRAF (Warsaw), 1997 Brittle Matrix Composites 6 Proceedings of the 6th International Symposium, Woodhead Publ. Ltd. (Cambridge) and BIGRAF (Warsaw), 2000 Brittle Matrix Composites 7 Proceedings of the 7th International Symposium, Woodhead Publ. Ltd. (Cambridge) and ZTUREK Research-Scientific Institute (Warsaw), 2003 Brittle Matrix Composites 8 Proceedings of the 8th International Symposium, Woodhead Publ. Ltd. (Cambridge) and ZTUREK Research-Scientific Institute (Warsaw), 2006

II

matrix2009.qxd

9/19/2009

11:55 AM

Page 3

INSTITUTE OF FUNDAMENTAL TECHNOLOGICAL RESEARCH POLISH ACADEMY OF SCIENCES

BRITTLE MATRIX COMPOSITES 9 Edited by

A.M. BRANDT Institute of Fundamental Technological Research Polish Academy of Sciences, Warsaw, Poland

J. OLEK Department of Civil and Environmental Engineering Purdue University, West Lafayette, USA

I.H. MARSHALL Monash University, Melbourne, Australia

Woodhead Publishing Ltd., Cambridge, and Institute of Fundamental Technological Research Warsaw 2009

III

matrix2009.qxd

9/19/2009

11:55 AM

Page 4

© 2009 Institute of Fundamental Technological Research, Warsaw WOODHEAD PUBLISHING LIMITED Abington Hall, Abington, Cambridge, England INSTITUTE OF FUNDAMENTAL TECHNOLOGICAL RESEARCH Warsaw, Poland

ISBN 978-83-89687-48-7 ISBN 978-1-84569-775-4

Conditions of sale: All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. British Library Cataloguing in Publication Data. A catalogue record for this book is available from the British Library.

The selection and presentation of material and the opinion expressed in this publication are the sole responsibility of the authors concerned.

No responsibility is assumed by the Publishers for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any method, products, instructions or ideas contained in the material herein.

Produced by:

ZTUREK Research-Scientific Institute 00-950 Warsaw 1, P.O. Box 674 http://www.zturek.pl, [email protected]

IV

matrix2009.qxd

9/19/2009

11:55 AM

Page 5

PREFACE These Proceedings include a collection of papers presented during the 9th international symposium on Brittle Matrix Composites (BMC-9), which was held October 25-28, 2009 in Warsaw, Poland. The series of BMC Symposia started as the EUROMECH 204 Colloquium on Brittle Matrix Composites, which was held in Jablonna, Poland in November 1985. This meeting was later renamed BMC 1. Three years later (September 1988), the Second Symposium (BMC 2) was organized in Cedzyna, also in Poland. This was followed by a series of triennial Symposia, each held in Staszic Palace in Warsaw, Poland. In each case, the Institute of Fundamental Technological Research (IFTR) of the Polish Academy of Sciences served as the host. From the 1985 BMC 1 to the present BMC 9, these Symposia have served both as effective forums for dissemination of knowledge and as catalysts for innovation. They have undoubtedly contributed to international scientific collaboration, both on institutional and personal levels. The objective of this series of symposia is to bring together researchers in the broad field of different composites that behave (or may behave under certain conditions) like brittle materials. It covers materials with cement, ceramic and polymer matrixes. While their intended applications can be quite different, these composites share many similar characteristics. Some examples illustrating the positive impact of one field on another include the mechanics of fiber/matrix debonding, fiber bridging behavior, application of non-conventional components, and multiple cracking processes. The BMC Symposia provide a forum which encourages and enhances crossdisciplinary collaboration and exchange of knowledge. Since the beginning, the BMC Symposia have created an environment where researchers from many countries of the world gather once every three years to share their latest research findings and to learn first-hand about the developments in composite research in various regions of the world. Brittle matrix composites are used in various fields (civil engineering, mechanical equipment and machinery, vehicles, etc.). In the last decades, their importance and diversity amongst engineered materials has continuously increased. Examples of materials covered in the accepted papers include the following: x aggregate-binder composites (concretes, fiber concretes, polymer concretes), x sintered materials (ceramics), and x other composites with brittle matrices. Various approaches to material engineering problems are presented in the papers, including: x mechanical properties, strength, toughness and rheology, x analysis of materials structure and microstructure, x various degradation effects, crack propagation and control, x test methods and new test results, x computation methods and manufacturing processes, x durability assessment of materials and structures, and x applications of new materials and their behavior in structures.

V

matrix2009.qxd

9/19/2009

11:55 AM

Page 6

The present volume of the BMC 9 Proceedings should prove useful to experienced researchers and to students entering this field as a helpful reference. The included papers have been selected on the basis of a two-stage peer review process. The volume contains 45 papers prepared by 116 authors from 22 countries. To a great extent this set of papers represents the latest advancements in the field of Brittle Matrix Composites, including works on applications of different kinds of fibers in cement matrix, various test techniques used for evaluation of concrete and ceramics, durability and repair of structures, analytical methods and other new problems in composites. In 2008 we lost two of our eminent colleagues who were involved in the organization of BMC Symposia since the beginning. Professor Janusz Kasperkiewicz (Poland) died in May and Professor Ian H. Marshall (Scotland, UK) passed away in November. They participated actively in the creation of both intellectual and organizational frameworks for this series of Symposia. We feel their absence very deeply. Particular thanks are due to the authors of the articles in this volume for their readiness to present the results of their outstanding investigations in the form of original contributions. Grateful thanks are extended to members of the International Advisory Board for their significant help in reviewing the papers and in solving problems encountered during the organizational stages. The efforts and creative attitude of the local Organizing Committee during preparations for this event is highly appreciated. Without their dedicated work it would not have been possible to publish this volume before the Symposium date. The support from the Institute of Fundamental Technological Research of the Polish Academy of Sciences was essential for organization of this Symposium, as well as previous Symposia. The scientific sponsorship of RILEM is acknowledged with gratitude. We sincerely hope that this volume, along with the eight previously published in this series, will contribute to the development of science and technology in the field of composite materials.

A.M. Brandt J. Olek I.H. Marshall

VI

matrix2009.qxd

9/19/2009

11:55 AM

Page 7

INTERNATIONAL ADVISORY PANEL Prof. R. Al-Mahaidi, Melbourne (Australia) Prof. A. Bentur, Haifa (Israel) Prof. J.J. Biernacki, Tennessee (USA) Prof. J. Kasperkiewicz, Warsaw (Poland) D.A. Lange, Urbana-Champaign (USA) S. Mindess, Vancouver (Canada) Z. Mróz, Warsaw (Poland) Y. Ohama, Koriyama (Japan) H.W. Reinhardt, Stuttgart (Germany) S.P. Shah, Evanston (USA) H. Stang, Lyngby (Denmark) P. Stroeven, Delft (The Netherlands) R.N. Swamy, Sheffield (UK) A. Vautrin, St. Etienne (France)

ORGANIZING COMMITTEE

Prof. J. Kasperkiewicz Prof. M.A. Glinicki Mrs. A. Gutweter Dr. D. JóĨwiak-NiedĨwiedzka Prof. M. Marks Mr. M. Sobczak Mrs. J. Tymkiewicz

VII

matrix2009.qxd

9/19/2009

11:55 AM

Page 8

matrix2009.qxd

9/19/2009

11:55 AM

Page 9

Table of contents Preface

KEYNOTE PAPER Retrofit of shear strength deficient RC beams with sprayed GFRP Andrew J. BOYD, Nemkumar BANTHIA and Sidney MINDESS, Canada …...………..1 Study on the flexural behaviour of reinforced fibrous concrete beams Rashid HAMEED, Anaclet TURATSINZE, Frédéric DUPRAT and Alain SELLIER, France ……………………………………………………………11 Round panel vs. beam tests toward a comprehensive and harmonic characterization of FRC materials Fausto MINELLI and Giovanni A. PLIZZARI, Italy …………………………………...23 The influence of steel fibres content and curing conditions on mechanical properties and deformability of reactive powder concrete at bending Tomasz ZDEB and Jacek ĝLIWIēSKI, Poland …………………………………...……33 Interface bond characteristics between wood fibres and a cement matrix Mercedes G. SIERRA BELTRAN and Erik SCHLANGEN, The Netherlands ……...…43 Shear strength and ductility of beams reinforced with synthetic macro-fibers Salah ALTOUBAT, Samer BARAKAT, Yazdanbakhsh ARDAVAN and Klaus-Alexander RIEDER, UAE,USA, Germany ……………………………….......53 The impact of amount and length of fibrillated polypropylene fibres on the properties of HPC exposed to high temperature Izabela HAGER and Tomasz TRACZ, Poland …………………………………………63 Assessment by law of mixtures approach of interfacial adhesion strength in cellulose-cement composites Conrado S. RODRIGUES, Piet STROEVEN and Khosrow GHAVAMI, Brazil, The Netherlands ……………………………………71 Tensile fatigue response of Sisal Fiber Reinforced Cement Composites Flavio de Andrade SILVA, Barzin MOBASHER and Romildo D. TOLEDO FILHO, Brazil, USA ……………………………………..…81 Investigation on the strength and flexural toughness of hybrid fibre reinforced concrete Surinder Pal SINGH, India …………………………………………………………...…91

IX

matrix2009.qxd

9/19/2009

11:55 AM

Page 10

Textile reinforced concrete - durability issues: changes of the bond and tensile strength due to ageing Till BÜTTNER, Jeanette ORLOWSKY and Michael RAUPACH, Germany ………...101 Impact study of textile reinforced cementitious materials: test method and preliminary results Johan Van ACKEREN, J. BLOM, D. KAKOGIANNIS, J. WASTIELS, D. Van HEMELRIJCK, S. PALANIVELU, W. Van PAEPEGEM, J. DEGRIECK and J. VANTOMME, Belgium ……………………………………...…111 The ductile behavior of HPFRCC in compression Alessandro P. FANTILLI, Hirozo MIHASHI, Paolo VALLINI and Bernardino CHIAIA, Italy, Japan …………………………………………………121 Investigation of random distribution of fibres in cement composites Tomasz PONIKIEWSKI, Poland ………………………………………………………131 Permeability of SFRCC based on fine aggregate after pre-load cycles Jacek KATZER, Poland ………………………………………………………………..139 KEYNOTE PAPER Statistical optimization of low slump ternary concrete mixtures with ground granulated blast furnace slag (GGBS) and high calcium fly ash for pavement applications Adam RUDY, Jan OLEK, Tommy NANTUNG and Richard M. NEWELL, USA ……149 Durability performance of Roller Compacted Concrete using fly ash Jong-Pil WON, Chang-Il JANG, Sang-Woo LEE and Wan-Young KIM Republic of Korea …………………………………………………………………...…161 Structure and properties of NaOH activated cement free binder (CFB) concretes Deepak RAVIKUMAR, Sulapha PEETHAMPARAN and Narayanan NEITHALATH, USA …………………………………………………169 Model of concrete carbonation as limitable process – experimental investigations of fluidal ash concrete Lech CZARNECKI and Piotr WOYCIECHOWSKI, Poland …………………………183 Modification of mineral binding matrixes carbon nanostructures Grigory I. YAKOVLEV, Grigory N. PERVUSHIN, Jadviga. KERIENE, Hans-Bertram FISCHER and Bernd MÖSER B., Russia, Lithuania, Germany ……….195 Mechanical-acoustic and structural study of degradation processes in corundum ceramics and aluminous porcelain Przemysáaw RANACHOWSKI, Zbigniew RANACHOWSKI and Feliks REJMUND, Poland ………………………………………………………...201

X

matrix2009.qxd

9/19/2009

11:55 AM

Page 11

Descriptive microstructure and fracture surface observations of fired volcanic ash C. LEONELLI, E. KAMSEU, U.C. MELO, A. CORRADI, and G.C. PELLACANI, Italy, Cameroon ……………………………………………...215 Dehydration, dehydroxylation, densification and deformation during sintering of geopolymers based on the K2O-Al2O3-SiO2 system E. KAMSEU, A. RIZZUTI, C. LEONELLI and D. PERERA, Italy, Australia ……….217 Multicriterial optimization of autoclaved aerated concrete properties and expenditure of energy resources Tatiana V. LYASHENKO, Vitaly A. VOZNESENSKY and Varvara P. GAVRILIUK, Ukraine ………………………………………………...219 Application of machine learning for prediction of concrete resistance to migration of chlorides Maria MARKS, Daria JÓĩWIAK-NIEDħWIEDZKA and MICHAà A. GLINICKI, Poland ………………………………………………….227 Carbon spheres as possible micro-reinforcement of cement-based composites Jan M. SKOWROēSKI and Agnieszka ĝLOSARCZYK, Poland ………………….…237 Properties of fiber reinforced cement composites with cenospheres from coal ash Waldemar PICHÓR, Poland …………………………………………………………..245 How to get reliable 3D information on concrete porosity? Piet STROEVEN, The Netherlands ……………………………………………………255 Analysis of the influence of the type, amount and way of introduction of anti-foaming admixture (AFA) on the properties of self-compacting concrete mix Beata àAħNIEWSKA-PIEKARCZYK, Poland ………………………………………265 Predicting the elastic moduli of enhanced porosity (pervious) concretes using reconstructed 3D material structures Milani S. SUMANASOORIYA, Omkar DEO and Narayanan NEITHALATH, USA ...275 KEYNOTE PAPER Enhanced durability of concrete by superabsorbent polymers Hans W. REINHARDT and Alexander ASSMANN, Germany ……………………….291 Volumetric stability of concrete using recycled concrete aggregates Yogini DESHPANDE, Jacob E. HILLER and Cory J. SHORKEY, USA ……………..301 Damage assessment in sections for durability purposes: two arguments not to opt for automation Piet STROEVEN, Huan HE and Martijn STROEVEN, The Netherlands ……………..313 The effects of laser cleaning process on geometrical microstructure of cementitious composites Agnieszka KLEMM, Poologanathan SANJEEVAN and Piotr KLEMM, UK, Poland ..323 XI

matrix2009.qxd

9/19/2009

11:55 AM

Page 12

Assessing the repeatability of the restrained ring test Aleksandra RADLINSKA and Jason WEISS, USA …………………………………...335 Performance evaluation of rapid-setting materials for concrete pavements/bridge decks repair: laboratory and field perspectives Prashant V. RAM and Jan OLEK, USA ……………………………………………….347 The effect of temperature on the rheological properties of self - compacting concrete Jacek GOàASZEWSKI and Grzegorz CYGAN, Poland …………………………..….359 Assessment of the rheological properties of cement mixes using electrical resistance Dominik LOGOē, Poland …………………………………………………………….369 Experimental techniques for multi-scale characterization of mechanical response in cement-based materials Joseph J. BIERNACKI, USA …………………………………………………………..379 Influence of hydrothermal curing on microstructure and mechanical properties of ultra-high performance concrete Patrick FONTANA, Christian LEHMANN and Urs MÜLLER, Germany ……………391 Patches in concrete: recent experimental discovery of a natural phenomenon - supporting evidence by DEM Piet STROEVEN and Huan HE, The Netherlands ……………………………………..399 Relationships between fractal dimension and the mechanical and structural parameters of basalt aggregate concretes Janusz KONKOL and Grzegorz PROKOPSKI, Poland ……………………………….409 T-stress values during fracture in wedge splitting test geometries: A numerical study Stanislav SEITL, Pavel HUTAě, Václav VESELÝ and ZbynČk KERŠNER, Czech Republic ………………………………………………419 The influence of aggregate size on the width of fracture process zone in concrete members Marta SàOWIK and Ewa BàAZIK-BOROWA, Poland ………………………………429 Effect of steel strip geometry on pull-out strength of aerated concrete Dariusz ALTERMAN, Juan VILCHES, Thomas NEITZERT and Hiroshi AKITA, Poland, New Zealand, Japan ……………………………………439

INDEX OF CONTRIBUTORS ……………………………………………..……………….…449 SUBJECT INDEX ……………………………………………………………………………...451

*

* XII

*

01-Retrofit of shear.qxd

9/19/2009

11:55 AM

Page 1

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

KEYNOTE PAPER

KEYNOTE PAPER RETROFIT OF SHEAR STRENGTH DEFICIENT RC BEAMS WITH SPRAYED GFRP Andrew J. BOYD1, Nemkumar BANTHIA2,Sidney MINDESS2 McGill University, Department of Civil Engineering, Montreal QC, Canada e-mail: [email protected] 2 Department of Civil Engineering, University of British Columbia, Vancouver, BC, Canada 1

ABSTRACT Worldwide, a great deal of research is currently being conducted concerning the use of fibre reinforced plastic wraps or laminates in the repair and strengthening of reinforced concrete members. Such techniques can be both effective and economical when compared to the existing practice of retrofitting with steel plates. A novel technique which further simplifies the application procedure is to apply the fibre using a spraying process. By spraying the fibres onto the member surface concurrently with a suitable matrix resin a two dimensional random distribution of discontinuous fibres is obtained, resulting in an FRP plate that exhibits isotropic in-plane behaviour. This work reports the application of a sprayed glass fibre reinforced plastic (GFRP) retrofit system on reinforced concrete beams. Using E-glass fibres embedded in a polyester matrix, a sprayed FRP system was used to strengthen beams intentionally designed to be deficient in shear capacity. This approach was shown to be capable of significantly increasing both the strength and stiffness of such concrete beams, while at the same time dramatically improving their energy absorption characteristics. When compared with fabric wrapped specimens, and with published literature on similar specimens retrofitted with either fabric or laminates, the sprayed GFRP material proved to be capable of producing results equivalent or better than any of the other techniques. The sprayed FRP approach also tended to improve the bond between the FRP and concrete surface, thus providing resistance to the primary mode of failure exhibited in FRP retrofit systems; failure at the bond line.

Keywords Glass fibre reinforced plastics (GFRP), retrofitting RC beams, sprayed FRP

INTRODUCTION One of the largest problems facing the construction industry today is the maintenance, repair and strengthening required by a large proportion of our transportation infrastructure. Throughout the world, transportation agencies are constantly faced with the need to repair and maintain deteriorating structures, a task which is made even more difficult by shrinking resources, budgets and manpower [1-3]. Many countries are also in the process of upgrading existing structures that are now considered inadequate due to building code revisions or changes to their original purpose (such as the need to carry heavier loads or higher traffic volumes) [2,4-11]. For example, nearly half of all bridges in the U.S.A. are currently considered to be structurally deficient [3,5]. Some form of external reinforcement is needed to repair and strengthen these deficient structures.

01-Retrofit of shear.qxd

2

9/19/2009

11:55 AM

Page 2

Andrew J. BOYD, Nemkumar BANTHIA and Sidney MINDESS

In recent years, a great deal of research has been conducted into the use of fibre reinforced plastics (FRPs) in place of steel plates for such strengthening procedures. Such reinforcement is usually performed using FRP fabrics or plates adhesively bonded to the structure [12-17]. There are a number of significant advantages to the use of FRP laminates to replace steel as external reinforcement, including excellent corrosion resistance, excellent fatigue resistance, low density coupled with very high stiffness and strength and a very low coefficient of thermal expansion in the fibre orientation [1,3,7,9,10,12,13,18-20]. The higher costs associated with these FRP materials is often offset by savings in labour when compared to the difficulty of steel plate jacketing [3,12,13]. Banthia et a1 [21] investigated the effectiveness of using a sprayed fibre reinforced plastic in the external strengthening of plain concrete beams. By spraying the fibres onto the member surface concurrently with a suitable matrix resin, a two dimensional random distribution of fibres was obtained. Using E-glass fibres embedded in a matrix consisting of a polyurethane/polyester blend, it was shown that significant increases in load carrying capacity, fracture toughness and fracture energy can be achieved with such a system. Comparisons of the obtained results for confined cylinders with those available from current literature showed that the sprayed glass FRP technique can produce more effective confinement than either carbon or ararnid FRP wrappings [22]. This paper discusses an investigation into the usage of a sprayed glass fibre reinforced plastic (GFRP) retrofit system on small-scale reinforced concrete beams.

EXPERIMENTAL PROGRAM Existing RC beams can be deficient in shear strength for several reasons, including shear deficient design procedures in older codes, increased service loads on the structure and corrosion of the shear stirrups (which are typically placed outside the flexural reinforcement and are thus protected by less cover concrete) [23-26]. The primary objective of this investigation was to determine whether such a reinforcing system would be effective for retrofitting members exhibiting a shear failure mode. To this end, a series of 22 reinforced concrete beams was cast, with each beam measuring 96 mm wide x 125 mm high x 1.0 m long. The longitudinal flexural reinforcement was over-designed in these members and shear reinforcement was intentionally omitted in order to induce a shear failure under third point loading. After a 28-day immersion curing period, the beams were separated into three groups: the control specimens, specimens to be damaged and later repaired with GFRP, and undamaged specimens to be strengthened with GFRP. Damaging, in this case, consisted of subjecting the specimens to third point loading until the first visible shear crack appeared. This corresponded to a centre point deflection of L/180 (5 mm). Following this preloading stage, the surfaces of both the damaged and undamaged specimens were sandblasted in an attempt to optimize the GFRP-concrete bond. The GFRP material (ıult = 28 MPa, E = 4.0 GPa, Vf § 8 %) was applied in a coating 6 mm thick. Four different retrofit schemes were investigated, as depicted in Figure 1, resulting in a total of nine different specimen groups. Figure 2 illustrates the spraying operation.

01-Retrofit of shear.qxd

9/19/2009

11:55 AM

Page 3

Retrofit of shear strength deficient RC beams with sprayed GFRP

3

Figure 1: GFRP retrofit schemes

Figure 2: Photograph of sprayed GFRP application process All of the retrofitted beams, as well as the control specimens, were tested to failure under third point loading. During testing, the applied load and the deflection at rnidspan were monitored and recorded. Two additional LVDTs were added over the supports to account for settlement at the supports and crushing during loading. Load vs. corrected deflection curves were plotted for all specimens, where the corrected deflection was taken to be the deflection at mid-span less the average rigid body settlement in the testing apparatus. From these data, values for peak load and fracture energy absorbed up to peak load were determined. Also noted during testing was the failure mode exhibited by each specimen.

RESULTS AND DISCUSSION Retrofitting the beams with the sprayed GFRP coating actually succeeded in changing the failure modes exhibited by the specimens. The resulting failure modes varied, depending upon the particular retrofit scheme used. The control specimens, as expected, exhibited a characteristic shear failure mode (Figure 3). The specimens coated on the sides only (Scheme A) or on the sides and the top (Scheme B) failed due to rebar debonding in the shear span, as shown in Figure 4.

01-Retrofit of shear.qxd

4

9/19/2009

11:55 AM

Page 4

Andrew J. BOYD, Nemkumar BANTHIA and Sidney MINDESS

Figure 3: Shear failure in control specimens (side of beam)

Figure 4: Rebar bond failure in Scheme A and B specimens (bottom of beam)

Figure 5: Typical rebar bond/shear in Scheme C specimens (top of beam)

01-Retrofit of shear.qxd

9/19/2009

11:55 AM

Page 5

Retrofit of shear strength deficient RC beams with sprayed GFRP

5

Figure 6: Typical flexural failure in Scheme D specimens (side of beam) Specimens coated on the two sides and the bottom (Scheme C) failed in a combination of rebar debonding followed later by the shear splitting shown in Figure 5. Also evident in this photo is some localized debonding of the GFRP on either side of the specimen. This debonding actually occurred during the large deflections after specimen failure and was thus not a contributing factor. Coating of all four sides actually succeeded in shifting the failure location from the end span to the centre span and the failure mode from shear to a typical flexural failure, as depicted in Figure 6. The values obtained for peak load and fracture energy are depicted graphically in Figures 7 and 8, respectively. Retrofitting the beams with Scheme A or B resulted in a significant increase in peak load over the control specimens, though there was very little difference between the two. Similarly, Schemes C and D produced an even larger increase in peak load, though again there was little difference between the groups. Though these increases in peak load are significant, the real difference in effectiveness among the four schemes is better illustrated by the energy absorbed by the members up to the peak load. As shown in Figure 8, all of the retrofit schemes resulted in large increases in fracture energy. Scheme A actually outperformed Scheme B in this case. This was due to the stiffening effect of the top plate in the Scheme B specimens, which reduced the deflection capability of the beam. This effect will be mentioned again later. Scheme C specimens, in turn, absorbed significantly more energy than those of Scheme A or B, and the beams from Scheme D far outperformed all of the other configurations. No consistent and significant difference was observed between damaged and undamaged specimens within the same scheme. It appears that the change in failure mode induced by the GFRP coating rendered the effect of the preloading shear damage insignificant. As these figures show, all of the retrofit schemes significantly increased both the peak load and the fracture energy absorbed by the specimens. Though Scheme D is by far the most effective, it is Scheme C that is of the most interest from a practical standpoint due to the inability, in most cases, to gain access to the upper surface of in situ reinforced concrete beams. The superior performance exhibited by the Scheme D specimens can be attributed to the higher degree of confinement achieved by coating all four sides of the element.

01-Retrofit of shear.qxd

6

9/19/2009

11:55 AM

Page 6

Andrew J. BOYD, Nemkumar BANTHIA and Sidney MINDESS

Figure 7: Peak load by retrofit scheme

Figure 8: Fracture energy by retrofit scheme Upon examining the load-deflection curves of the 22 beams, it was found that the shapes of the curves were very consistent among beams retrofitted with the same scheme. Figure 9 shows the typical behaviour, up to failure, for each of the four repair schemes. The curve representing the control specimens is exactly as expected for a shear failure. Coating beams on both sides as in Scheme A resulted in an increase in load carrying ability and deflection compared to the control specimens. The increased fracture energy for these

01-Retrofit of shear.qxd

9/19/2009

11:55 AM

Page 7

Retrofit of shear strength deficient RC beams with sprayed GFRP

7

specimens is a direct result of the increased load carrying ability. This is basically the same curve as for the control specimens, but extended to a higher load level.

Figure 9: Typical load-deflection curves by retrofit scheme. Considering the curve for those beams coated on both sides plus the top (Scheme B) we see that now we have a significant increase in member stiffness. This is why the Scheme B specimens exhibited inferior energy absorption values compared to those from Scheme A. Basically, these beams reached the same peak loads but underwent significantly lower deflections due to this increase in member stiffness. Beams retrofitted using Scheme C also exhibited the same increase in stiffness, though this time it was accompanied by very large increases in both load carrying ability and deflection, resulting in much high energy absorption capacity. The Scheme D beams exhibited an even more interesting phenomenon, the appearance of significant plastic deformation before reaching peak load. These were the specimens that failed in flexure in the centre span of the member. It is also apparent from this figure that the Scheme C specimens also underwent some plastic deformation. Actually, one of the Scheme C beams did fail in flexure as well. This would indicate that the GFRP thickness used was very close to the optimum thickness required to produce flexural failure even in Scheme C retrofitted members. Comparison with other retrofit schemes Tables 1 and 2 provide a comparison between the beams from this study and comparable beams in previously published literature. The values shown in the table represent the percent increases exhibited by the retrofitted members relative to their control specimens. Table 1 refers to retrofit Scheme A. Both the peak load and fracture energy increases produced by the sprayed GFRP are significantly larger than those previously reported for GFRP plate bonding. Though the thickness of the sprayed GFRP material is significantly greater than the GFRP plates used for this process, its much lower fibre content still results in a retrofit material with a lower total cost.

01-Retrofit of shear.qxd

9/19/2009

8

11:55 AM

Page 8

Andrew J. BOYD, Nemkumar BANTHIA and Sidney MINDESS

Table 1: Comparison with previously published results – Scheme A

Retrofit Type

Fibre Type

Peak Load Increase (%)

Fracture Energy Increase (%)

Sprayed GFRP

GlassA

23

Plate Bonding

72

300

o B

20

-9

o C

32 to 66

51 to 173

Glass (0/90 )

18

Plate Bonding

Glass (±45 )

Matrix: A Polyester/Polyurethane

B

Polyester

C

Epoxy

Examination of the increases exhibited by the Scheme C specimens (Table 2) shows that the sprayed GFRP coating can produce similar or better results even when used in a Ushaped application where the sides and bottom of the beam are coated. In this case, previous data were found not only for GFRP plate bonding but also for wrapping with fabrics composed of aramid, glass and carbon fibre. This superior behaviour is primarily due to an improved bond between the FRP and concrete when using the sprayed system. In the previously published literature most failures occurred at the bond, meaning that the FRP did not reach its full potential. The sprayed FRP, though still not able to reach its ultimate capacity except in one case, was able to shift the failure mode to the rebar bond. Table 2: Comparison with previously published results – Scheme C

Retrofit Type

Fibre Type

Peak Load Increase (%)

Fracture Energy Increase (%)

Sprayed GFRP

GlassA Glass (0/90o)B Glass (±45o)B Glass (0/90o)C Aramid (0/90o)D Glass (0/90o)D Carbon (±45o)D

160 128 to 193 131 to 179 45 83 88 125

723 513 to 1002 549 to 942 187 168 317 479

Fabric Wrapping Plate Bonding27 Fabric Wrapping28

Matrix: A Polyester/Polyurethane

B

Polyester with vinyl ester coupling agent

C

Polyester

D

Epoxy

It should be noted that the figures shown in Tables 1 and 2 for fracture energy from previously published literature are estimates. These values were not actually reported by the respective authors but were estimated from the load-deflection curves provided. Procedure comparison When viewed from an application viewpoint, it must be noted that the thickness of the sprayed FRP coating can easily be controlled during the spraying operation simply by adjusting the number of passes made. With the wrapping systems, on the other hand, an increase in thickness requires a corresponding increase in labour as successive layers of material must be added separately. Thus, not only is the spraying process itself much less labour intensive than wrapping, this labour savings increases dramatically as the required material thickness (i.e. required strength gain) increases. This benefit could potentially result

01-Retrofit of shear.qxd

9/19/2009

11:55 AM

Page 9

Retrofit of shear strength deficient RC beams with sprayed GFRP

9

in large savings in production costs when one considers large scale application of this material. The equipment used for spraying FRP is completely portable and thus very easily adapted to on-site applications. It requires very little support equipment; an air compressor is needed though a power supply is not. Additionally, operation of this equipment requires only a single worker and its simplicity ensures that operator training times are minimal.

CONCLUSIONS The retrofit of shear strength deficient reinforced concrete beams with sprayed GFRP appears to be a feasible method not only for improving the strength and stiffness of the section, but also for helping to produce a more ductile failure mode by increasing the member's energy absorption capability. Furthermore, its ability to produce improvements similar or better than those achieved with other FRP retrofit methods, combined with the less labour intensive application procedure, makes the sprayed GFRP process very desirable from an economic viewpoint. It is essential that more work be done to fully explore the benefits of this retrofit technology. REFERENCES 1. Chajes, M.J., Thomson, T.A., Januszka, T.F., Finch, W.W., Flexural strengthening of concrete beams using externally bonded composite materials. Construction & Building Materials 8(3), 1994, pp 191-201 2. Ehsani, M.R., Rehabilitation of the infrastructure with advanced composite materials. Repair and Rehabilitation of the Infrastructure of the Americas, NSF, Mayaguez, PR, 1994, pp 193-205 3. Meier, U., Carbon fibre-reinforced polymers: modern materials in bridge engineering. Structural Engineering Int. 1(12), 1992, pp 7-12 4. Chajes, M.J., Thomson, T.A., Tarantino, B., Reinforcement of concrete structures using externally bonded composite materials. Symp. Non-Metallic (FRP) Reinforcement for Concrete Structures, Ghent, Belgium, 1995, pp 501-508 5. Ehsani, M.R., Saadatrnanesh, H., Fiber composite plates for strengthening bridge beams. Composite Structures 15(4), 1990, pp 343-355 6. Meier, U., Winistijrfer, A., Retrofitting of structures through external bonding of CFRP sheets. Symp. Non-Metallic (FRP) Reinforcement for Concrete Structures, Ghent, Belgium, 1995, pp 466-472 7. Meier, U., Deuring, M., Meier, H., Schwegler, G., Strengthening of structures with advanced composites. Alternative Materials for the Reinforcement and Prestressing of Concrete, Blackie, Glasgow, U.K., 1993, pp 153-171 8. Nanni, A., Concrete repair with externally bonded FRP reinforcement. Concrete International 17(6), 1995, pp 22-26 9. Nanni, A., Noms, M.S., Bradford, N.M., Lateral confinement of concrete using FRP reinforcement. Symp. Fiber-Reinforced-Plastic Reinforcement for Concrete Structures, ACI SP- 138, 1993, pp 193-209 10. Norris, M.S., Nanni, A., Construction and repair of concrete members with FRP reinforcement. Construction Technology Update, National Research Council of Canada, 1993, Vol 3, pp 2-3 11. Triantafillou, T.C., Plevris, H., Strengthening of RC beams with epoxy-bonded fibercomposite materials. Materials and Structures 25, 1992, pp 201-211

01-Retrofit of shear.qxd

10

9/19/2009

11:55 AM

Page 10

Andrew J. BOYD, Nemkumar BANTHIA and Sidney MINDESS

12. Meier, U., Deuring, M., The application of fibre composites in bridge repair. Strasse und Verkehr 77(9), 1991, pp 534-535 13. Meier, U., Kaiser, H., Strengthening of structures with CFRP laminates. Conf. Advanced Composite Materials in Civil Engineering, Las Vegas, NV, USA, 1991, pp 224-232 14. Plevris, N., Triantafillou, T.C., Time-dependent behavior of RC members strengthened with FRP laminates. Structural Engineering 120(3), 1994, pp 1016-1042.Triantafillou, T.C., Plevris, N., Post-strengthening of RC beams with epoxy-bonded fibre composite materials. Conf. Advanced Composites Materials in Civil Engineering Structures, Las Vegas, NV, U.S.A., 1991, pp 245-256 15. Meier, U., Deuring, M., Meier, H., Schwegler, G., CFRP bonded sheets. FiberReinforced-Plastic (FRP) Reinforcement for Concrete Structures: Properties and Applications, A. Nanni, editor, Elsevier, New York, 1993, pp 435-434 16. Meier, U., Deuring, M., Meier, H., Schwegler, G., Strengthening of structures with CFRP laminates: research and applications in Switzerland. Conf. Advanced Composite Materials in Bridges and Structures, Sherbrooke, QC, Canada, 1992, pp 243-251 17. Berset, J.D., Strengthening of reinforced concrete beams for shear using FRP composites. M.Sc. Thesis, Dept of Civil Eng, MIT, Cambridge, MA, U.S.A., 1992 18. Triantafillou, T.C., Plevris, N., Reliability analysis of reinforced concrete beams strengthened with CFRP laminates. Symp. Non-Metallic (FRP) Reinforcement for Concrete Structures, Ghent, Belgium, 1995, pp 576-583 19. Triantafillou, T.C., Deskovic, N., Deuring, M., Strengthening of concrete structures with prestressed fiber reinforced plastic sheets. Structural Journal 89(3), 1992, pp 235-244 20. Banthia, N., Yan, C., Nandakumar, N., Sprayed fibre reinforced plastics (FRPs) for repair of concrete structures. Conf. Advanced Composite Materials in Bridges and Structures, CSCE, Montreal, QC, Canada, 1996, pp 537-545 21. Nandakumar, N., Repair and retrofit of damaged concrete structures using sprayed fibre reinforced plastics. Dept of Civil Eng, UBC, Vancouver, BC, Canada, 1996 22. Al-Sulaimani, G.J., Sharif, A., Basunbul, I.A., Baluch, M.H., Ghaleb, B.N., Shear repair for reinforced concrete by fiberglass plate bonding. Structural Journal 91(3), 1994, pp 458-464 23. Shehata, E., Morphy, R., Rizkalla, S., FRP as shear reinforcement for concrete structures. Pres. 1997 ACI Fall Convention, Atlanta, Georgia, U.S.A., Nov 9-14, 1997 24. Dorton, R.A., Development of Canadian Bridge Codes. Conf. Developments in Short and Medium Span Bridge Engineering, CSCE, Halifax, NS, Canada, 1994, pp 1-12 25. Bonneau, I., Massicotte, B., Strengthening of shear deficient reinforced concrete beams by external prestressing. Conf. Developments in Short and Medium Span Bridge Engineering, CSCE, Halifax, NS, Canada, 1994, pp 767-778 26. Boyd, A.J., Banthia, N., Comparison of sprayed GFRP and FRP wraps for retrofit of reinforced concrete beams. Dept of Civil Eng, UBC, Vancouver, BC, Canada, 1998. 27. Chajes, M.J., Januszka, T.F., Mertz, D.R., Thornson, T.A., Finch, W.W. Shear strengthening of reinforced concrete beams using externally applied composite fabrics. Structural Journal, 92(3), 1995, pp 295-303

02-Study on the flexural.qxd

9/19/2009

11:55 AM

Page 11

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

STUDY ON THE FLEXURAL BEHAVIOUR OF REINFORCED FIBROUS CONCRETE BEAMS Rashid HAMEED, Anaclet TURATSINZE, Frédéric DUPRAT, Alain SELLIER Université de Toulouse UPS –INSA; LMDC (Laboratoire Matériaux et Durabilité des Constructions), 135 avenue de Rangueil ; F- 31077 Toulouse Cedex 04, France e-mail: [email protected]

ABSTRACT In this study, behaviour of composite concrete with conventional reinforcement (i.e., steel bars) and randomly distributed metallic fiber reinforcement, and subjected to flexural loading has been examined. The objective was to investigate the effect of adding fiber in mono and hybrid form on the flexural properties of reinforced concrete. Two types of metallic fibers were used: amorphous metallic fibers (type I fibers) characterised as high performance non-slipping fibers with large specific surface area, and carbon steel hooked-end fibers (type II fibers) characterised as slipping fibers. Four types of reinforced concretes were prepared: one control (without fibers) and three fibrous. Out of three reinforced fibrous concretes, two contained single fiber and one contained fibers in hybrid form. The total volume fraction of fibers was 0.25% (20 kg/m3) and 0.5% (40 kg/m3) for single fiber and hybrid fiber concrete respectively. Three point bending test were performed according to European standards “NF EN 14651” on beam specimens of 150 x 150 mm cross section and length of 550 mm. The experimental results showed that the presence of metallic fibers appreciably reduces the crack opening and deflection and improves moment capacity of the reinforced concrete section. Moreover, as a result of positive synergetic interaction between fibers, reinforced fibrous concrete containing fibers in hybrid form exhibited more improved response at all loading stages in comparison of reinforced fibrous concrete containing fibers of one type only. Cracking and ultimate moment capacities of the beams were also determined analytically. Analytical results were found to be in good agreement with experimental results.

Keywords Reinforced fibrous concrete beam, metallic fibers, hybridisation, moment capacity, deflection, CMOD

INTRODUCTION Among different construction materials, concrete exhibits low tensile strength and is relatively brittle material. Its mechanical behaviour is critically influenced by crack propagation. Reinforcement of concrete with randomly distributed fibers can address some concerns related to concrete brittleness and poor resistance to cracking, Banthia and Sappakittipakorn [1]. Many of the properties of fiber reinforced concrete can be used to advantage in the concrete flexural members reinforced with conventional bar reinforcement, Swamy et al. [2]. The use of steel fibers along with longitudinal steel bars improves the yielding moment, ultimate moment and post-yield behaviour. Moreover, addition of fibers reduces the immediate deflection, long-term deflection and crack width of beam, Ashour et al. [3], Vandewalle [4]. Many kinds of fibers have been used in concrete since last six decades and up till now no single fiber reinforced concrete could exhibit perfect mechanical properties. In recent years, attention has been given to the hybridization of fibers in the

02-Study on the flexural.qxd

12

9/19/2009

11:55 AM

Page 12

Rashid HAMEED, Anaclet TURATSINZE, Frédéric DUPRAT and Alain SELLIER

concrete to get more enhanced response in term of mechanical properties; Banthia and Soleimani [5], Pons et al. [6], Hsie et al. [7], Ahmed and Maalej [8]. Hybridization means to mix two or more than two different fibers in a matrix. The basic purpose of using hybrid fiber is to control cracks at different size level, in different zones of concrete and different loading stages, Barragam et al. [9]. The hybridization of fibers in concrete can be done in different ways, such as by combining different lengths, diameters, geometry, modulus and tensile strengths of fibers, [8]. Most studied hybrid combinations of fibers include steelpolypropylene, steel-carbon and carbon-polypropylene. In the family of metallic fibers, there are many types of fibers according to their geometrical, mechanical and physical properties. Use of fibers from the same family in hybrid-fiber reinforced concrete is not so common. Therefore, in the present experimental study, flexural behaviour of the reinforced concrete containing two different metallic fibers in hybrid form has been investigated. The two metallic fibers used differ in their geometrical, physical and mechanical properties. The main interest of this investigation is to obtained experimental data on the flexural strength of reinforced concrete beams containing the metallic fibers in mono and hybrid forms at different loading stages. The experimental results are to be compared with analytical solution based on engineering practices in reinforced concrete calculation, in order to render fiber addition in concrete elements more attractive for practical applications.

EXPERIMENTAL PROGRAM Material The materials used for the non-fibrous (control) and fibrous concrete mixtures consisted of CEM I 52.5 R cement, the coarse aggregates (gravels) having size range of 4 to 10 mm, and the locally available natural river sand of maximum particle size of 4 mm. The additives are two types of metallic fibers shown in Fig.1. Type I fibers, 30 mm length, 1.6 mm width and 0.03 mm thickness (Tensile strength of 1400 MPa and modulus of elasticity of 140000 MPa), are made of amorphous metal (Fe, Cr)80 (P, C, Si)20. Type II fibers, 30 mm length and 0.5 mm diameter (Tensile strength of 1200 MPa and Modulus of elasticity of 210000 MPa), are carbon steel wires that have been cut into suitable lengths and are hooked at each end. They are usually adhered together in clips of certain number of wires (Fig. 1). When these clips enter in the mix, the adhesive is dissolved and individual fibers are distributed evenly throughout the mix. A super-plasticizer was used in order to improve the fresh properties of concrete in the presence of metallic fibers. The composition of the control concrete is given in table 1.

Type I fibers

Type II fibers

Figure 1- amorphous metal fibers (on left), carbon steel fibers (on right)

02-Study on the flexural.qxd

9/19/2009

11:55 AM

Page 13

13

Study on the flexural behaviour of reinforced fibrous concrete beams

Concrete Compositions A total of four reinforced concretes: one control (without fibers) and three containing fibers were investigated. Among the three reinforced fibrous concretes, two contained single fiber and one contained fibers in hybrid form. Fiber type and dosage for all the concrete mixtures are given in Table 2. Test Specimens A total of 8 notched beams (2 beams for each concrete mix) with cross section of 150 x 150 mm and total length of 550 mm were constructed. Each beam was reinforced with two 6 mm diameter steel bars (fy = 500 MPa) fulfilling the minimum requirement of Eurocode 2 for the tension steel. The cross section and reinforcement details of tested beam specimen are shown in Fig.2.

Table 1 – Mixture Proportion of the control concrete Cement (Kg/m3) 322

Sand (Kg/m3) 872

Gravel (Kg/m3) 967

Water (Kg/m3) 193

Super-Plasticizer (Kg/m3) 1.61

Table 2 – Fiber contents in different concrete mixtures Mixture No.

Mixture Type

M-0 M-1 M-2

Control single fiber Hybrid fiber

M-3

Type I

Type II

--0.25 ---

----0.25

Total quantity of fibers, Vf (%) --0.25 0.25

0.25

0.25

0.5

Volume fraction of fibers, %

43

LOAD

150

150

Compressive strength, MPa 42 44 42

10

2 - Ø= 6 mm

2 - Ø= 6 mm

500 550

Figure 2: Cross-section and reinforcement details of the test specimen Testing Procedure 28 days after the casting, three point bending tests were performed on reinforced non-fibrous (control) and fibrous concrete notched beams according to European Standards NF EN 14651 “Test method for metallic fibered concrete”. Although this standard is for fiber concrete without steel bars, the same procedure has been adopted for the fiber reinforced concrete with conventional steel reinforcing bars. All the tests were controlled by crack mouth opening

02-Study on the flexural.qxd

14

9/19/2009

11:55 AM

Page 14

Rashid HAMEED, Anaclet TURATSINZE, Frédéric DUPRAT and Alain SELLIER

displacement (CMOD) using LVDT. In each test, mid-span deflection of the beam was also measured using LVDT. CMOD and deflection measurement scheme is shown in Fig. 3.

Figure 3: Test setup and measurement scheme

TEST RESULTS The experimental results (all compositions) on cracking and ultimate moment (moment corresponding to peak load attained by each composite) are given in table 3 and CMODmoment and deflection-moment curves for each mix are shown in Fig.4. It can be observed that for a given CMOD or a given deflection, the load carrying capacity of the reinforced concrete beam with metallic fibers is significantly improved.

Table 3: Cracking and ultimate moments for all compositions Concrete Mix M-0 M-1 M-2 M-3

Sample # 1 2 1 2 1 2 1 2

Cracking moment, Mcr (kN-mm) 2379 2281 3216 3163 2333 2338 3653 3292

Ultimate moment* Mult (kN-mm) 5831 5669 5874 6113 6138 6228 7029 6974

* Maximum value in CMOD-Moment or Deflection-Moment curve Fig.5 shows the values of cracking moment for all concretes and strength effectives with the addition of metallic fibers. Compared to control reinforced beam (without fibers), the cracking moment is increased by 36.8%, 0.3% and 49.1% for M-1, M-2 and M-3 concretes respectively. This shows that cracking moment is increased significantly in the presence of type I fibers. On the other hand, type II fibers do not affect importantly. Fig.5 also shows the increase in ultimate moment of reinforced concrete with the addition of metallic fibers. Similar to cracking moment, M-3 exhibited maximum value of ultimate moment. Compared to control beam, the ultimate moment is increased by 4.1%, 8.5% and 21.7% for M-1, M-2 and M-3 concretes respectively.

02-Study on the flexural.qxd

9/19/2009

11:55 AM

Page 15

15

Study on the flexural behaviour of reinforced fibrous concrete beams

The evolution of CMOD and deflection with the increase of moment is shown in Fig.6. It can be observed that for a given moment, addition of metallic fibers significantly reduced the crack opening and deflection. Deflection value at ultimate moment was 2.8 times less with M-1 compared to M-0. Similarly deflection was 2.2 times less with M-2 and 3.9 times less with M-3 compared to M-0. Same observations were also made with crack opening. CMOD value at ultimate moment was 2.6 times less with M-1, 2.1 times less with M-2 and 2.7 times less with M-3 compared to M-0.

8000

8000

M-3

M-3 M-2

M-1

6000 ) m m N k( t n e m o M

6000

4000

) m m N k( t n e m o M

M-1 M-0 M-0

2000

M-1 M-2

4000

M-2 M-0 M-0

2000

M-1 M-2 M-3

M-3

0

0

1

2 3 CMOD (mm)

4

0

5

0

1

2 3 Deflection (mm)

4

5

Figure 4: CMOD-moment and Deflection-moment curves 8000

Cracking Moment

49,1

Strength Effectiveness (%)

Moment (kN-mm)

M-1 M-2

60

Ultimate moment (Moment at Peak) 7000 6000 5000 4000 3000 2000 1000 0

M-3

50 36,8 40 30

21,7

20 8,5 10

4,1

0,3

0

M-0

M-1

M-2

M-3

Cracking Moment

Ultimate Moment

Figure 5: Cracking and ultimate moment (experimental results) 1,8

M-0

1,6

M-0

1,6

M-1

1,4

M-1

1,4

M-2

1,2

M-2

1,2

M-3

1

M-3

Deflection (mm)

CMOD (mm)

2 1,8

1 0,8 0,6

0,8 0,6

0,4

0,4

0,2

0,2 0

0 0

1000

2000

3000

4000

Moment (kN-mm)

5000

6000

0

1000

2000

3000

4000

5000

6000

Moment (kN-mm)

Figure 6: Evolution of CMOD and deflection with moment (up to Mult attained by M-0)

02-Study on the flexural.qxd

16

9/19/2009

11:55 AM

Page 16

Rashid HAMEED, Anaclet TURATSINZE, Frédéric DUPRAT and Alain SELLIER

FLEXURAL CAPACITY OF BEAMS: ANALYTICAL APPROACH Commonly accepted mechanics and equilibrium conditions have been used to calculate the flexural capacity in term of cracking and ultimate moment of the beams containing randomly distributed metallic fibers and reinforcing steel bars. The tensile forces carried by the added fibers create an additional internal moment capacity, which is simply superposed to the moment capacity of reinforced concrete beam section. The analysis is based on following three assumptions: (1) plane section remains plane after bending; (2) the tensile forces balance the compressive forces, (3) the internal moment is equal to external applied bending moment Cracking Moment Simplified procedure proposed by Campione [10] is used here to calculate the cracking moment. Neglecting the effect of steel bar and referring to Fig.7. The cracking moment is calculated using eq. 1.

(a) cross-section

(b) strain distribution

(c) stress distribution.

Figure 7: Linear stress and strain distributions just before cracking

M cr

bh 2 u MOR 6

(1)

Where, MOR is modulus of rupture of fiber reinforced concrete. Additionally, a separate series of 3 point bending tests on prisms of cross section 100 x 100 mm and length of 500 mm was also carried out in order to determine the modulus of rupture of concrete compositions used in this study. For each composition, three samples were tested. These tests were performed according to European standard NF EN 14651 with the exception of specimen size. The results (average of three samples) are shown in Fig. 8. Ultimate moment The assumed and simplified stress distribution of reinforced non-fibrous and fibrous concretes at failure is shown in Figs. 9 and 10. In reinforced fibrous concrete, steel reinforcing bar, concrete matrix and randomly distributed fibers contributes to carry the post cracking tension. Referring to assumed and simplified stress distribution of reinforced fibrous concrete shown in Fig.10, where parabolic compressive stress zone is divided into two parts: rectangular and triangular, the ultimate moment is calculated using Eq. 2.

02-Study on the flexural.qxd

9/19/2009

11:55 AM

Page 17

17

Study on the flexural behaviour of reinforced fibrous concrete beams

Ts u z1 T f u z 2

MOR (MPa)

Mu

8 7 6 5 4 3 2 1 0

(2)

7,0

6,6

4,3

4,0

M-0

M-1

M-2

M-3

Figure 8: Modulus of rupture (MOR) In Eqs.2, Ts and Tf are tensile forces carried by steel bar and fibers respectively, z1 and z2 are lever arm distances. For reinforced fibrous concrete beam containing fibers in hybrid form, the ultimate moment capacity is calculated by simply adding the contribution of individual fibers in carrying tension and the expression is given in Eqs.3

Mu

Ts u z1 (T f 1 T f 2 ) u z 2

(3)

Tf1 and Tf2 are the tension carried by each fiber used in hybrid combination. Ts and Tf are calculated using Eqs.4 and 5. Ts T f 1, 2

As u f y

(4)

V t 1, 2 u b u (h c)

(5)

In Eqs.5, ıt is ultimate tensile strength of fiber reinforced concrete and is greatly influenced by the properties and contents of fibers.

(a) cross-section

(b) strain distr.

(c) assumed stress distr.

(d) simplified stress distr.

Figure 9: Stress and strain distributions in reinforced concrete

02-Study on the flexural.qxd

18

9/19/2009

11:55 AM

Page 18

Rashid HAMEED, Anaclet TURATSINZE, Frédéric DUPRAT and Alain SELLIER

(a) cross-section

(b) strain distr.

(c) assumed stress distr.

(d) simplified stress distr.

Figure 10: Stress and strain distributions in reinforced fibrous concrete Determination of z1, z2 In the following section, step by step procedure is given to calculate z1 and z2. In the following equations A1 is area of the rectangular compressive stress block, A2 is area of triangular compressive stress block shown in Fig.10.

A1

0.85 f c' u 0.80c

y1

(0.80 u c) / 2

A2 y2

0.68 f c' c

(7)

1 (0.85 f c' u 0.20c) 0.085 f c' c 2 1 (0.20 u c) 0.80c 3

A1 A2

(6)

0.765 f c' c

(8) (9) (10)

y c'

A1 y1 A2 y 2 ¦A

z1

d y c'

(12)

z2

(

hc ) (c y c' ) 2

(13)

0.452c

(11)

Depth of the neutral axis “c” is determined using Eqs.14 for single fiber concrete and using Eqs.15 for hybrid fiber concrete. Eqs.14 and 15 are obtained by equating compression and tension forces i.e., C = Ts+Tf. c

V t bh As f y ( single fiber) 0.765 f c' b V t b (V t 1 V t 2 )bh As f y ( hybrid fiber) c 0.765 f c' b (V t 1 V t 2 )b

(14) (15)

Determination of ıt About the contribution of matrix in the ultimate tensile strength of fiber reinforced concrete ‘ıt’, one can consider interlocking between aggregates but it vanishes quickly with the crack

02-Study on the flexural.qxd

9/19/2009

11:55 AM

Page 19

19

Study on the flexural behaviour of reinforced fibrous concrete beams

opening after the peak load. Moreover, this contribution is small compared to tensile strength contributed by the fibers, so contribution by the matrix is ignored. Therefore, ultimate behaviour of FRC is governed by properties of fibers, number of fibers, etc. According to Hsu and Hsu [11], number of fibers acting across the cross section is determined by the following expression:

Do u

N

Vf

(16)

Af

From the Eqs.16, NAf is the area of fibers per unit area of section and is equal to ĮoVf.. Af is area of fibers and Vf is volume fraction of fibers. In the analysis of reinforced fibrous concrete, each fiber is considered as small longitudinal reinforcement present through the whole length of section. The ultimate tensile strength of fiber reinforced concrete is calculated by using the following equation:

Vt

Do uV f uV f uDb

(17)

Where, Įo is orientation factor and is equal to 0.41, Oh [12], Įb is bond efficiency factor and its value varies from 1 to 1.2 depending upon fiber characteristics, ACI Committee 544 [13]. For straight fibers, the value of Įb is taken equal to 1, Henager and Doherty [14], but in this study, for type I straight fibers, taking into account high bond strength with matrix due to surface roughness, its value is taken equal to 1.2; the maximum value proposed by ACI Committee. For hooked-ends fibers (type II), Dancygier and Savir [15] proposed the value of Įb equal to 1.2, here in this study, the same value (i.e., Įb =1.2) has been used. ıf is tensile strength of fibers and the values of ıf for the fibers used in this study are given in section 2.1. Analytical values of cracking and ultimate moment capacities are given in table 4 along with ratios between experimental and analytical values.

Table 4 – Analytical and experimental moment capacities of fibrous composites Composition M-1 M-2 M-3

Cracking moment (kN-mm) (Mcr)anl (Mcr)exp 3188* 3234 2338* 2107 3475* 3430

( M cr ) anl ( M cr ) exp 1.014 0.901 0.987

Ultimate moment (kN-mm) (Mult)exp (Mult)anl 5988* 5225 6238* 5101 7000* 6861

( M ult ) anl ( M ult ) exp 0.872 0.817 0.980

* Value is averaged of two samples

DISCUSSION

The experimental results show that the resistance of reinforced concrete beam is increased with the addition of randomly distributed metallic fibers. Fiber action to arrest cracking at micro or macro level is strongly depended on one hand, on the properties of fiber (geometry, strength and stiffness) and on the other hand, on the bond between the fibers and concrete matrix. For the bond between fiber and matrix, matrix compactness also plays an important role.

02-Study on the flexural.qxd

20

9/19/2009

11:55 AM

Page 20

Rashid HAMEED, Anaclet TURATSINZE, Frédéric DUPRAT and Alain SELLIER

Between the two fibers used in this study: type I fibers develop good bond with the matrix because of surface roughness and large specific surface area. According to a study carried out by Turatsinze and Bascoul [16], micro-cracking occurs inside the matrix before the peak load in flexure, in this context, type I fibers act as soon as the first micro-cracks open and immediately restrain their propagation due to good bond with concrete matrix. By this way, they enhance the response prior and just after the peak load. With further increase of crack opening, the fibers contribute to carry tension along with steel bar and response of the composite is improved in term of load bearing capacity, reduced deflection and smaller crack opening (Figs. 4 and 6). As well as the failure of the type I fibers is concerned, when the stress in the fiber exceeds its tensile strength with further increase of crack opening, the fibers break instead of pulling out from the matrix and the postpeak residual load bearing capacity approaches towards a value equal to control concrete (Fig. 4). On the other hand, type II fibers develop poor bond with concrete matrix due to their smooth surface and small specific surface area. As a result of poor bond, at micro-cracking, these fibers slip from the concrete matrix and minor effect on the response of the composite in term of strength is observed. Since these fibers are hooked-end, sufficient anchorage in the matrix helps these fibers to restrain macro-cracks propagation by transferring the stress across the crack; as a result, post-crack response (toughening effect) of the concrete composite is significantly improved. With the increase of crack opening, these fibers are stretched and anchorage in the concrete matrix due to hooked-end is further improved and fibers play important role in carrying tension along with steel reinforcing bar at macro cracking level. After the yielding of steel bars, crack opening is increased significantly; as a result, fibers are pulled out (slipped) from the concrete matrix instead of breaking and their hooked-ends turn straight. Since the two metallic fibers used in this study provide reinforcement at different level: type I fibers at micro-cracking level and type II fibers at macro-cracking level, the best composite properties are obtained from the hybrid composite containing both fibers which exhibits the greatest flexural strength and toughness because of positive synergetic interaction. The theoretical equations developed here to determine the cracking and ultimate moment capacities taking into account the metallic fibers are a simple and more practical way and thus a useful tool for engineers to assess the bending capacity of reinforced fibrous concrete. The analytical values of cracking and ultimate moment obtained show a good agreement with experimental results. The ratios between analytical and experimental values of cracking and ultimate moment lie between 0.8 and 1.1 with different concretes used in this study.

CONCLUSIONS AND FUTURE WORK

Based on the test results and predicted values of cracking and ultimate moment capacities, the following conclusions are drawn: x Addition of randomly distributed metallic fibers causes an increase in the flexure strength and toughness of the reinforced concrete beams. x The level of improvement in the flexural response of the reinforced fibrous concrete beam is greatly depended on the physical and mechanical properties of metallic fibers. x A significant reduction in the crack width and deflection is guaranteed by the addition of metallic fibers in the reinforced concrete beam.

02-Study on the flexural.qxd

9/19/2009

11:55 AM

Page 21

Study on the flexural behaviour of reinforced fibrous concrete beams

21

x The matrix-fiber high bond due to rough surface and large specific surface area of amorphous metallic fibers (type I fibers) makes the fibers more effective in strengthening the composite. In opposite to that ductile behaviour and hooked-ends of carbon steel fibers (type II fibers) make the fibers more effective in toughening the composite. x Positive synergetic effect exits between the two metallic fibers used in this study. Hence the composite containing these fibers in a well chosen hybrid form can exhibit high performance in term of strengthening and toughening. x Simplified analytical approach is developed to calculate the ultimate flexural strength of reinforced fibrous concrete rectangular section taking into account the contribution of metallic fibers. Analytical results showed good agreement when compared to experimental results. For the future work, research will be carried on to find more realistic value of bond efficiency factor for high performance amorphous metallic fibers used in this study. REFERENCES

1.

Banthia, N., Sappakittipakorn, M., Toughness enhancement in steel fiber reinforced concrete through fiber hybridization, Cement and Concrete Research, 37, 2007, 13661372 2. Swamy, R. N., Sa’ad A. Al-Ta’an, Deformation and ultimate strength in flexure of reinforced concrete beams made with steel fiber concrete, ACI Journal, SeptemberOctober 1981, 395-404 3. Ashour, S.A., Mahmood, K., Wafa, F., Influence of steel fibers and compression reinforcement on deflection of high-strength concrete beams, ACI Structural Journal, V.94, No.6, November-December 1997 4. Vandewalle, L., Cracking behaviour of concrete beams reinforced with a combination of ordinary reinforcement and steel fibers, Material and Structures, Vol. 33, April 2000, pp 164-170 5. Banthia, N., Soleimani, S.M., Flexural response of hybrid fiber-reinforced cementitious composites, ACI Materials Journal, V.102, No.6, November-December 2005. 6. Pons, G., Mouret, M., Alcantrara, M., Granju, J.L., Mechanical behaviour of selfcompacting concrete with hybrid fiber reinforcement, Materials and Structures, 2007, 40, 201-210 7. Hsie, M., Tu, C., Song, P.S., Mechanical properties of polypropylene hybrid fiberreinforced concrete, Materials Science and Engineering, A 494, 2008, 153-157 8. Ahmed S.F.U., Maalej M., Tensile strain hardening behaviour of hybrid steelpolyethylene fiber reinforced cementitious composites, Construction and Building Materials, 23, 2009, 96-106 9. Barragam, B., Gettu, R., Zalochi, R.F., Martim, M.A., Agullo, L., A comparative study of the toughness of steel fiber reinforced concrete in tension, flexure and shear. Fiber Reinforced Concrete (FRC) BEFIB’ 2000, Proceedings of the Fifth International RILEM Symposium, pp- 441-450 10. Campione, G., Simplified flexural response of steel fiber-reinforced concrete beams, journal of Materials in Civil Engineering, Vol.20, No. 4, April 2008 11. Hsu, L.S., Hsu, T., Stress strain behaviour of steel-fiber high-strength concrete under compression, ACI Structural Journal, V.91, No.4, July-Aug. 1994, pp. 448-457 12. Oh, B.H., Flexural analysis of reinforced concrete beams containing steel fibers, Journal of structural engineering, Vol.118, No.10, October 1992

02-Study on the flexural.qxd

22

9/19/2009

11:55 AM

Page 22

Rashid HAMEED, Anaclet TURATSINZE, Frédéric DUPRAT and Alain SELLIER

13. ACI Committee 544.4R-88, Design consideration for steel fiber reinforced concrete 14. Henager, C.N., Doherty, T.J., Analysis of reinforced fibrous concrete beams, journal of structural division, ASCE, 102 (STT): 177-188 15. Dancygier, A.N., Savir, Z., Flexural behaviour of HSFRC with low reinforcement ratios, Engineering Structures, 28, 2006, 1503-1512 16. Turatsinze, A., Bascoul, A., Restrained crack widening in Mode I crack propagation for mortar and concrete, Advanced Cement Based Materials, 1996, 4, 77-92

03-Round panel.qxd

9/19/2009

11:55 AM

Page 23

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

ROUND PANEL VS. BEAM TESTS TOWARD A COMPREHENSIVE AND HARMONIC CHARACTERIZATION OF FRC MATERIALS Fausto MINELLI, Giovanni A. PLIZZARI DICATA - Department of Civil, Architectural, Environmental and Land Planning Engineering, University of Brescia. Via Branze, 43, 25123 Brescia, Italy, e-mail: [email protected]; [email protected].

ABSTRACT Standard test methods for determining the mechanical properties of Fibre Reinforced Concrete (FRC) are better defined if they reproduce the actual structural behavior. A comparison between different test typologies for characterizing FRC is reported and discussed in the present paper, with special emphasis on the different scatter that each test produces. Tests are performed on beams as well as panels. All specimens have the same concrete mechanical properties and fibre content. Aim of the investigation is to critically discuss advantages and disadvantages of each testing procedure, focusing on the applicability of the method and on the reliability of results toward a consistent characterization of the structural behavior. A new geometry for the panel test is herein proposed and discussed in order to make the panel easier to place and handle, avoiding one of the major drawbacks which limits an extensive utilization of the panel tests. Suitable correlations among the different fracture and energy parameters defined in the assumed standards are reported, resulting very useful for a harmonization of the available standards. Moreover, an analytical study is finally carried out focusing on the determination of the constitutive V-w law of FRC materials from panel tests. Results from the analytical study are compared with the experimental results and critically discussed.

Keywords Fibre-Reinforced Concrete, standard test methods, cracking, toughness

INTRODUCTION Fibre Reinforced Concrete (FRC) is gaining an increasing interest among the concrete community for the reduced construction time and labor costs. For this reason, many structural elements are now reinforced with steel fibres as partial or total substitution of conventional reinforcement (rebars or welded mesh, [1]). Besides cost issues, quality matters are of paramount importance for a construction and FRC also fulfill these requirements since fibres allow for more distributed cracks with a smaller opening that enhances durability. New construction materials requires standards for measuring their mechanical properties and building codes or guidelines for structural design [2-3]. As far as FRC is concerned, design guidelines are already available in some countries [4-6] and work is in progress for including them in the coming fib Model Code. In these guidelines, structural design is usually based on design values of the material parameters that are normally determined by dividing the characteristic values by a partial safety factor (JM). Mechanical properties of FRC are traditionally determined from beam tests that are usually based on a three [3] or four point bending schemes [7]. Early experiences with the low

03-Round panel.qxd

9/19/2009

11:55 AM

Page 24

24

Fausto MINELLI and Giovanni A. PLIZZARI

volume fractions of fibres that are nowadays used in practice (Vf < 0.8 - 1.0%) evidence that the characteristic values determined from beam tests [3,7] are quite low because of the high scatter present in the beam-test results; it should be observed that the latter scatter is not related to the material itself but it is mainly due to the small fracture areas of the beams (ranging from 160 to 180 cm2). Such a scatter becomes particularly high when low contents (25-50 kg/m3) of macro steel fibres (length ranging between 30 and 60 mm) are used, Sorelli et al. [8]. It is commonly accepted that FRCs with a low volume fraction of fibres are particularly suitable for structures with a high degree of redundancy where stress redistribution may occur. Because of this redistribution, large fracture areas are involved (with a high number of fibres crossing them) and, therefore, structural behaviour is mainly governed by the mean value of the material properties; furthermore, because of the large fracture areas, the scatter of experimental results from structural tests is remarkably lower than that obtained from the beam tests. A typical example is shown in Fig. 1, which exhibits a set of curves obtained from a standard (bending) test on notched specimens (Fig. 1a) and from structural tests on full scale slabs on grade made of the same material (Fig. 1b); the different scatter between material and structural tests is clearly evident (Cominoli, [9]). Load vs. Central deflection 350

4

300 3

250 Load [kN]

Nominal Stress s N [MPa]

5

2

200 Slab P22 50/0,75 Vf=0,38 %

150

Slab P21 50/0,75 Vf=0,38 %

100

1

Slab P20 50/0,75 Vf=0,38 %

50 0 0.0

0.5

1.0

1.5

2.0

CTODm [mm]

2.5

3.0

3.5

0 0

1

2

3

4

5

6

Central deflection [mm]

(a) (b) Fig. 1: Experimental results from bending tests on a notched beam (a) and from a full-scale slab on grade (b) made of the same FRC (di Prisco et al. [6]). In order to obtain a more realistic value of the scatter from FRC material tests, specimens with larger fracture areas are needed; this suggests the use of larger beams or different specimens like slabs where stress redistribution may also occur. A Round Determinate Panel (RDP) test was proposed by ASTM [10]; it is a statically determinate test (a round slab: I = 800 mm; thickness = 75 mm, with three supports at 120 degrees) where the crack pattern is predictable and the post-cracking material properties can be better determined. However, handling and placing such a specimen is quite complicated due to the large size and, consequently, the high weight. In addition, standard servo-controlled loading machines may not fit with the geometry of the panel, which is too big for most of them. The need to have a specimen easier both to handle and to test brought the Authors to come up with a proposal of a smaller round panel having a diameter of 600 mm and thickness of 60 mm. The present paper focuses on the comparison of different tests for FRC materials tested in the last few years at the University of Brescia. A comparison between beam and panel tests, a discussion on the smaller round panel herein proposed as well as the correlations between the different fracture properties obtained from different Standards are here presented.

03-Round panel.qxd

9/19/2009

11:55 AM

Page 25

Round panel vs. beam tests toward a comprehensive and harmonic characterization of FRC materials

25

Last but not least, an analytical study is finally carried out focusing on the determination of the constitutive V-w law of FRC materials from panel tests. Results from the analytical study are compared with the experimental results and critically discussed.

MATERIALS, SET UP, EXPERIMENTAL RESULTS AND DISCUSSION Among more than 100 comparative experiments carried out on beams and panels, the following discussion will deal with a number of tests conducted on members containing either 20 or 30 kg/m3 of hooked-end steel fibres having a length of 50 mm and a diameter of 1 mm (the aspect ratio L/Iis 50); fibres have a circular cross section and tensile strength of 1100 MPa. Besides the FRC specimens, plain concrete beams and panels were also made. Beams according to UNI 11039 [7] and EN 14651 [3] as well as classical round panles (called round panels large in the following, RPL) according to ASTM (2004) and round panel small (RPS, radius 600 mm; thickness 60 mm) were tested. All geometrical characteristics and further details are reported in Minelli and Plizzari [11-12]. In order to study the behavior of all specimens up to failure, including any possible unstable branch after cracking, a displacement controlled testing method was adopted. The equipment shown in Fig. 2 (a) was adopted for all tests except for RPL, whose dimensions, as already mentioned, required to utilize a different equipment. In the first case, an INSTRON 1274 machine was used (having a closed loop and a maximum load of 500 kN). In the second case, the displacement was imposed by adopting an electro-mechanical screw jack (having a maximum load of 500 kN and a stroke of 300 mm) placed into a steel frame. Fig. 2 (b) shows the steel supporting system suitably designed for performing tests on RPS with INSTRON 1274 machine.

(a)

(b)

Fig. 2: Set up for performing panel tests (a), and steel supporting system (b) A CMOD (Crack Mouth Opening Displacement)-controlled procedure was adopted for both notched beams (UNI and CEN prescribe a notch with a different length) and round panels small, whereas a displacement-controlled test (screw control with the electromechanical jack) was performed in the case of the large panel tests, resulting in an instability of the test immediately after the peak of the concrete matrix [13]. Fig. 3 shows the Nominal Stress (according to a linear stress distribution in the cracked section) versus the CTOD for both CEN and UNI beam tests as well as the load versus the central vertical displacement of both large and small round panels made of the identical material. As expected, the post-peak behavior is similar for the two types of beam tests, which are characterized by a large scatter of experimental responses. Referring to Fig. 3c and Fig.

03-Round panel.qxd

9/19/2009

11:55 AM

Page 26

26

Fausto MINELLI and Giovanni A. PLIZZARI

3d, the dispersion of results is definitely much smaller than in the corresponding beam tests, both for large and small round panels. It is worth underlining that, in order to characterize the post-peak behavior of panels, three LVDTs were placed to monitor the development of the crack widths, at a distance of 120 mm from the center of the panel. Crack widths greater than 20 mm were measured with a post-cracking load higher than one third of the peak load. Such an instrumentation is not required by the ASTM Standard [10], which states that one should only calculate the energy absorption (defined by the vertical load and the vertical displacement). However, monitoring the crack widths supported the analytical studied, which will follow in next paragraph. 7

Comparison UNI Beams Steel fibres

7

20 kg/m³

20 kg/m³

30 kg/m³

5

Comparison RILEM Beams Steel fibres

6

Nominal Stress s N [MPa]

Nominal Stress s N [MPa]

6

4 3 2 1

5

30 kg/m³

4 3 2 1

0

0 0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0

0.5

1

1.5

2 2.5 CTODm [mm]

(a)

CTODm [mm] 35

3

3.5

4

4.5

(b)

25

Comparison LARGE ROUND PANEL Steel fibres

30

Comparison SMALL ROUND PANEL Steel fibres 20

20

Load [KN]

Load [KN]

25 20 kg/m³ 30 kg/m³ 15

15 20 kg/m³ 30 kg/m³ 10

10 5 5 0

0 0

10

20

30

40

50

0

10

20

(c)

Displacement [mm]

40

30

(d)

Displacement [mm]

Fig. 3: Comparison of UNI, CEN and round panels tests, both for FRC with 20 and 30 kg/m. Comparison coefficient of variation Steel Fibers 0.5

Comparison coefficient of variation Steel Fibers FF1

Properties determined at SLS

Properties determined at ULS 20 kg/m³

0.41

0.4

0.5

0.46

20 kg/m³

30 kg/m³

30 kg/m³ 0.4

0.37 0.32

0.3 0.3

0.24

0.23 0.23

0.2

0.16

0.15

0.09

0.1

0.23 0.19

0.2

0.14 0.08

0.05

0.0

0.08

0.1

0.0 UNI

CEN

RPL

RPS

UNI

CEN

RPL

RPS

Fig. 4: Coefficient of variation calculated for all parameters required by the standards considered in the present experimental campaign The coefficient of variation, as an indicator of the test result scatter, was calculated for all properties and indexes defined in the different standards (Fig. 4), both for quantities which refer to the serviceability limit states (SLS) and ultimate limit states (ULS). Once again, a significant lower coefficient of variation can be outlined for panel tests in comparison with beam tests, according to the experimental results aforementioned.

03-Round panel.qxd

9/19/2009

11:55 AM

Page 27

Round panel vs. beam tests toward a comprehensive and harmonic characterization of FRC materials

27

ANALYTICAL MODEL Based on the broad experimentation, an analytical model was then undertaken to the aim of: - determining local stress-crack opening cohesive law from round panels; - defining suitable equivalent post-cracking strength directly from panel tests, as it usually occurs from beam tests. First of all, according to the thin plate theory, appropriate general relations, in the elastic range, were determined describing the maximum tangential stress (Vt) and the displacement at the load point (Fig. 5). An elastic finite element analysis was performed with the following results: PD V t 0.001816 2 (1) t

P D3 (2) E t3 where P is the external load applied at the slab center, D is the diameter of the panel measured form the supports, t is the panel thickness, E and Q are the Young Modulus and the Poisson ratio of concrete, respectively. The elastic analysis was also able to predict the distribution of radial and tangential stresses along the line of crack formation, as depicted in Fig. 5. Note that the elastic solution is an easy-to-use tool that allows a simplified calculation of the equivalent post-cracking strength being feasible for practical and design applications, in the same way as it was done for beam tests.

K RPS

0.000429

1.2

Comparison distribution r(r)/ r,max - t(r)/ t,max, z = 60 mm SMALL ROUND PANEL 1.0

] a P M [ 0.8

t

x

s t(r) Serie1

r

x a m t,

s r(r) Serie1

/ t 0.6 d n a

y

z

x a ,m r

0.4 r

s s e tr 0.2 S

0.0 0

50

100

150

200

250

300

r [mm]

Fig. 5: Elastic radial and tangential stresses along the line of crack formation By using the above equations for the determination of the local tangential maximum tensile stress at the load point, it is possible to come up with a V-w cohesive law and compare it with those determined from beam tests, being w the experimental measured crack widths. As an alternative, without direct measurement of the crack width, a kinematic approach based on the yield line theory could be adopted to calculate the crack width, as done by Bernard [14] and Tran et al. [15]. In the yield line analysis of a round panel, the governing mode of failure is taken to comprise three symmetrical yield lines-cracks emanating from the center of the face opposite to the point load and running radially to the edge while bisecting each sector between adjacent pivot supports. The crack width was experimentally measured at a distance from the point load of 120 mm (point Q in Fig. 6). Based on geometry consideration, it can be written that:

03-Round panel.qxd

9/19/2009

11:55 AM

Page 28

28

Fausto MINELLI and Giovanni A. PLIZZARI

BC

BQ

K RPS

GQ

(3)

where CQ =120 mm, according to the experiments. From trigonometry one can derive that: 2r

BQ

K RPS

GQ

§K · BQ ¨ RPS ¸ © 2r ¹

GQ

(4)

with r = 275 mm. Hinge Line

Zero Displacement

B

Yeld Line

Q C Central Load Point

P

Support Pivot

Pivot Lines

Fig. 6: Yield Line Approach for the determination of crack width The rotation at the support is therefore equal to:

GQ

D P PQ

(5)

By substituting PQ and Gq one can obtain: BQ §K · BQ ¨ RPS ¸ D P 3 © 2r ¹ The rotation at the support P therefore results:

(6)

3 K RPS (7) 2 r The rotation in Q, at the crack location (point of experimental measurement) can be derived as:

DP

M RPS

2 DP

3

K RPS

r The corresponding bottom crack opening) is therefore evaluated as:

(8)

§ 3 K RPS · ¸ 2t tg ¨¨ (9) ¸ © 2 r ¹ where r is the radius (r = 275 mm) and t is the panel thickness (t = 60 mm). It is worthy noticing that the crack opening obtained using Eq. 9 is constant along the radius, for a given value of panel displacement. However, in the actual case, the crack width will increase from the external surface to the load point. wm

§M · 2t tg ¨ RPS ¸ © 2 ¹

03-Round panel.qxd

9/19/2009

11:55 AM

Page 29

Round panel vs. beam tests toward a comprehensive and harmonic characterization of FRC materials

29

Fig. 7 plots the comparison between experimental and analytical crack width, which was derived from regression of experiments as follows:

wm

0.67 K RPS 0.32

(10)

The experimental straight line is bilinear, with a first constant line with zero crack representing the uncracked phase, in which the vertical displacement increases prior to cracking (up to a value of around K0 = 0.5 - 0.6 mm), and a second linear branch, showing a quite good fitting of the experiments (R2 = 0.99). The difference between the two curves (with the experimental showing higher cracks for the same vertical displacement) is probably due to the fact that a little elastic deformation is always measured by the LVDTs in the experiments (150 mm gauge length) and, moreover, their location is at 120 mm from the load point, while the yield line theory would match better with a measurement at r/2 from the load point (i.e., 137.5 mm). This procedure allows now the definition of a constitutive cohesive V- w law based on round panel experiments.

Fig. 7: Crack width at bottom surface of a small round panel As far as the beam tests are concerned, from a linear regression of experimental results, an analytical relation between the crack tip opening displacement CTOD and the vertical displacement at midspan was derived as follows: CTODm 1.03 K B (11)

MB

From a kinematic model shown in Fig. 8, it was obtained that: 4 K B l

(12)

Load Points

Fig. 8: Kinematic model for 4 point bending tests UNI beams The Italian Standard (UNI 11039) defines two different equivalent post-cracking strengths, feq(0-0.6) and feq(0.6-3), which represent the toughness given to the matrix by fibres in a conventional serviceability (crack width from 0 to 0.6 mm) and ultimate limit states (crack widths from 0.6 to 3 mm).

03-Round panel.qxd

9/19/2009

11:55 AM

Page 30

30

Fausto MINELLI and Giovanni A. PLIZZARI

Using the two above mentioned equations, it is possible to correlate crack ranges with vertical displacements KB and rotations MB of beams. By imposing that the rotation of round panels be equal to that of beams (Eqs. 8 and 12) according to the yield line theory, one can first determine the vertical displacement at the point load KRPS , and then the crack width (Eq. 10) corresponding to the rotation given. The crack width values then define two ranges that allow the definition of two equivalent postcracking strengths, relevant for serviceability and ultimate limit sates, respectively. The results of this procedure are reported in Table 2. Table 1: Midspan displacements and rotations corresponding to the CTODm defined by the UNI standard 11039 (2003). CTODm [mm] 0.6 3

KB [mm] 0.58 2.91

MB 0.052 0.0259

Table 2: Vertical displacement and crack width values corresponding to the same yield line rotation between panels and beams. MRPS=MB

KRPS

0.0052 0.0259

[mm] 0.82 4.11

wm [mm] 0.55 2.75

It is now possible to define two crack width ranges and to determine the corresponding equivalent post-cracking strengths, which are: x feq,1 from w0 to w0 + w1; x feq,2 from w0 + w1 to w0 + w2. here w0 - crack width at first cracking, w1 = 0.55 mm and w2 = 2.75 mm (Table 2). The two equivalent post-cracking strengths can therefore be defined as: k2 D U1 (13) f eq1 w1 t2 f eq 2

k2 D U2 w2 w1 t2

(14)

in which: x feq,1 is the equivalent post-cracking strength relevant for serviceability limit states; x feq,2 is the equivalent post-cracking strength relevant for serviceability limit states; x U1 is the area under the P-w experimental graph calculated in the crack range from w0 to w0 + w1 (Fig. 9); x U2 is the area under the P-w experimental graph calculated in the crack range from w0 + w1 to w0 + w2 (Fig. 9); x k2 = 0.001816 mm-1 is the constant already calculated. This approach can therefore be used to determine average values, standard deviations and characteristic values of the toughness parameters defined, taking strong advantage from the lower scatter of panel tests over beams (see Conti and Flelli for details [16]). Fig. 10 depicts a V- w curve with the indication of feq1 and feq2. Moreover, these two values allows the definition of a constitutive simplified bilinear law according to CNR DT 204 [4]. A similar approach can be derived starting from the load-displacement curve of panels, neglecting the crack width measurements (often not provided). In such circumstances, one can define suitable displacement ranges for the definition of feq1 and feq2. The displacement ranges, calculated using Eq. 8 and reported in Table 2, are incorporated into the definition of the two equivalent post-cracking strengths as:

03-Round panel.qxd

9/19/2009

11:55 AM

Page 31

31

Round panel vs. beam tests toward a comprehensive and harmonic characterization of FRC materials

f eq1

k 2 D E1 K1 t2

(15)

f eq 2

k2 D E2 t 2 K 2 K1

(16)

Load C arico

where all coefficients were already defined.

U

U

1

w

w

+

w

1

2

w

+

w

2

w

Fig. 9: Load-crack width curve for small round panels depicting areas U1 and U2

Fig. 10: Typical V-w curve for a small round panel showing the equivalent postcracking strengths

CONCLUDING REMARKS

A comparative study between beam and panel tests was discussed in the present paper. The experimental high scatter generally present in beam tests is definitely caused by the small geometry and fracture area involved in the tests. It does not represent the real structures where much larger fracture areas are involved and, consequently, a lower dispersion occurs. Panel tests are therefore more suitable for representing the actual behaviour of FRC materials. A number of improvements in the test procedures, such as close-loop control and crack width measurements were adopted allowing the characterization of FRC with a much lower dispersion of experimental results. The proposal of carrying out panel tests on smaller specimens brought to positive results, as such geometry does not affect the low scatter of the standard ASTM panel and, moreover, it allows for an easier placing and handling (lower weight and smaller geometry that fits with many servo-controlled testing machines). The correlation among fracture parameters found are in general very promising and allow engineers not to repeat different tests in different countries but to analytically calculate the fracture parameters of different tests from performing only one or few test typologies. These correlations, that were somehow expected since we are always referring to fracture properties of the same material, are important from practical point of view as they can give immediate indications related to the fracture properties without performing expensive tests. Given the analytical study undertaken, the round panel test can be considered as a complete test for the characterization of FRC: suitable range of crack widths are defined and the corresponding equivalent (or residual) post-cracking strengths are derived (from V-w plots) following the same procedure as done for beam tests. The panel test is relatively easy to perform, not expensive, it might require only one vertical displacement transducer, and, last but not least, it is characterized by a very low scatter of results.

03-Round panel.qxd

32

9/19/2009

11:55 AM

Page 32

Fausto MINELLI and Giovanni A. PLIZZARI

ACKNOWLEDGEMENTS

The Authors would like to deeply thank Engs. Yuri Marinoni, Raoul Martin Milla, Nicola Vezzoli, Moira Conti and Giovanni Flelli for their valuable assistance in performing the experiments and in data processing. REFERENCES

1. di Prisco, M., Failla, C., Plizzari, G.A. and Toniolo, G., Italian guidelines on SFRC, in: Fibre Reinforced Concrete: from theory to practice, di Prisco, M. and Plizzari, G.A. Eds., Proceedings of the International Workshop on advances in Fibre Reinforced Concretes, Bergamo (Italy), September 24-25, pp. 39-72, 2004 2. Vandewalle, L., Test and design methods for steel fibre reinforced concrete proposed by RILEM TC 162-TDF, Proceeding of the International Workshop “Fibre Reinforced Concrete. From theory to practice”, Bergamo, September 24-25, pp. 3-12, 2004 3. EN 14651, Test method for metallic fibred concrete. Measuring the flexural tensile strength (limit of proportionality (LOP), residual), CEN, 2005, pp.15 4. CNR DT 204, Guidelines for the Design, Construction and Production Control of Fibre Reinforced Concrete Structures, National Research Council of Italy, 2006 5. RILEM Final Recommendation TC-162-TDF, Test and Design Methods for Steel Fibre Reinforced Concrete; VHDesign Method, Materials and Structures, Vol 36, pp. 560-567, 2003 6. di Prisco, M., Felicetti, R. and Plizzari, G.A. Eds., Proceedings of the 6th RILEM Symposium on Fibre Reinforced Concretes, BEFIB 2004, RILEM PRO 39, 1514 pp, 2004 7. UNI 11039, Steel Fibre Reinforced Concrete - Part I: Definitions, Classification Specification and Conformity - Part II: Test Method for Measuring First Crack Strength and Ductility Indexes, Italian Board for Standardization, 2003 8. Sorelli, L., Meda, A. and Plizzari, G.A., Bending and uni-axial tensile tests on concrete reinforced with hybrid steel fibres, ASCE Materials Journal, Vol.15(5), pp. 519-527, 2005 9. Cominoli L., Studio sul calcestruzzo fibrorinforzato per applicazioni industriali: dalle proprietà del materiale al comportamento strutturale (in Italian), Ph.D. Thesis, University of Brescia (ISBN 978-88-96225-10-3), pp. 431, 2007 10. ASTM C1550-04, Standard test method for flexural toughness of fibre reinforced concrete (using centrally loaded round panel), pp. 9, 2004 11. Minelli, F. and Plizzari, G.A., Round Panel Tests for Characterizing Steel Fibre Reinforced Concrete; Proceedings of the Fib Symposium Keep Concrete Attractive, Budapest, Hungary, 23-25 May 2005 (Edited by György L. Balázs and Adorján Borosnyói, Published by the Publishing Company of Budapest University of Technologi and Economics), Vol.1, pp. 310-315, 2005 12. Minelli, F. and Plizzari, G.A., Fibre Reinforced Concrete characterization: Round Panel vs. beam test toward a harmonization, Proc. of the 3rd Central European Congress on Concrete Engineering CCC 2007 – Visegrád, Hungary, 17-18 September 2007. G.L. Balázs and S.G. Nehme (eds), Publishing Company of Budapest University of Technology and Economics, pp. 213-220, 2007 13. Marinoni, Y., Minelli, F., Plizzari, G.A., and Vezzoli, N., Prove di frattura su calcestruzzi fibrorinforzati in accordo con le principali normative internazionali (in Italian), Technical Report, University of Brescia, Department DICATA, 2007 14. Bernard, E.S., Correlations in the behavior of fibre reinforced shotcrete beam and panel specimens, Material and Structures, Vol. 35, pp. 156–164, 2002 15. Tran, V.N.G., Bernard, E.S. and Beasley, A.J., Constitutive Modeling of Fibre Reinforced Shotcrete Panels, Journal of Engineering Mechanics, Vol. 131, No. 5, pp. 512-521, 2005 16. Conti, M. and Flelli, G., Caratterizzazione del calcestruzzo fibrorinforzato attraverso prove su piastre sottili (in Italian), M.Sc. Thesis, University of Brescia, March 2009, 2009

04-The influence.qxd

9/19/2009

11:55 AM

Page 33

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

THE INFLUENCE OF STEEL FIBRE CONTENT AND CURING CONDITIONS ON MECHANICAL PROPERTIES AND DEFORMABILITY OF REACTIVE POWDER CONCRETE AT BENDING Tomasz ZDEB, Jacek ĝLIWIēSKI Cracow University of Technology, Faculty of Civil Engineering Warszawska 24, 31-155 Kraków, Poland e-mail: [email protected] ; [email protected]

ABSTRACT This paper refers to cementitious material belonging to Reactive Powder Concretes (RPC) and containing variable content of steel fibres. The test results include mechanical properties of this composite, cured in three different conditions: in water, steam-cured and in autoclave. Depending on curing conditions and fibres volume the compressive strength ranged from 200 to 315 MPa and the tensile strength at bending varied from 10 to 30 MPa. Moreover three-point bending test results are presented as load-deflection curves. On the basis of the afore mentioned measurements the following coefficients were calculated: work of fracture WF and toughness indices I5, I10 and I20. It was observed that both volume of steel fibres and curing conditions influence on flexural behaviour of RPC.

Keywords Reactive powder concrete, ultra high performance cementitious composites, steel fibres, fibre volume ratio, curing conditions, strength, fracture energy, fracture toughness index

INTRODUCTION According to Collepardi’s definition of reactive powder concretes (RPC), dispersed reinforcement is a standard component, [1]. Amongst the concretes the most recognized materials are the ones containing steel fibres. Fundamental aims for steel fibres application are: enhancing of mechanical properties, reduction of the shrinkage and inducing ductility of loaded material. The mean length of steel fibres in RPC ranges from 6 to 15 mm and diameter from 0.16 to 0.18 mm, while the volumetric content varies from 2.0 to 2.5% [2]. Characteristics of compressive and tension strength of RPC composites with constant amount of fibres are described among the others in papers [3-7]. However, it is difficult to find some information on influence of variable content of fibres with constant geometrical parameters on mechanical properties of RPC.

04-The influence.qxd

34

9/19/2009

11:55 AM

Page 34

Tomasz ZDEB and Jacek ŒLIWIÑSKI

MATERIALS AND SPECIMEN PREPARATION Materials In order to produce RPC test specimens the following components were used: portland cement CEM I 52.5R, silica fume, ground quartz 0/0.2 mm, quartz sand 0/0.5 mm, ordinary acrylic-based superplasticizer and straight and smooth steel fibres. Characteristics of applied materials are presented in Tables 1÷4. Table 1. Properties, chemical and phase composition of cement CEM I 52.5R Properties Initial setting time [min] 130 Final setting time [min] 220 Specific surface [cm2/g] 4100 Compressive strength after 2 days [MPa] 34.5 Compressive strength after 28 days [MPa] 70.8 Chemical composition [% w.] CaO – 65.58; SiO2 – 22.98; Al2O3 – 4.41; Fe2O3 – 2.10; SO3 – 3.32; MgO – 1.06; Na2OE – 0.51; Cl- - 0.009 Phase composition [% w.] C3S – 59.09; C2S – 17.97; C3A – 8.12; C4AF – 6.38 Table 2. Particle size distribution and chemical composition of ground quartz and sand Ground Sand Properties quartz Dmax [m] 200 500 D50 [m] 16 110 Specific surface BET [m2/g] 0.80 0.04 Density [g/cm3] 2.65 Polimorfic modification ȕ-quartz SiO2 [%] 99.0 98.5 Al2O3 [%] 0.30 0.80 Fe2O3 [%] 0.05 0.03 Table 3. Properties and chemical composition of silica fume Properties Specific surface [m2/g] 22.4 Density [g/cm3] 2.23 Chemical composition [% w.] SiO2 – 94.06; Al2O3 – 0.74; Fe2O3 – 0.78; CaO – 0.06; MgO – 0.49; Na2Oe – 1.43; SO3 – 0.63; LOI – 0.74 Table 4. Properties of steel fibres Length [mm] 6 Diameter [m] 175 Modulus of elasticity [GPa] 210 Tensile strength [MPa] 2200

04-The influence.qxd

9/19/2009

11:55 AM

Page 35

The influence of steel fibres content and curing conditions on mechanical properties and deformability of reactive powder concrete at bending

35

Fig. 1. Applied steel fibres (l = 6 mm; ĳ = 0.16 mm) Composition of RPC matrix Composition of examined RPC was developed in the other research project. Mass fractions of ingredients are presented in Table 5. The maximum volume fraction of steel fibres (Vfmax = 4% vol.) was defined by obtaining proper workability of concrete mix without changing its composition and assumption that inclusion is homogeneously dispersed in whole composite. Table 5. Composition of RPC matrix by weight ratio Cement CEM I 52,5R 1.00 Silica fume 0.20 Grounded quartz 0/0,20 mm 0.34 Quartz sand 0/0,50 mm 0.81 Superplasticizer 0.02 Water 0.24 Above mentioned mixture includes the following amount of steel fibres: Vf = 0.5; 1.0; 2.0; 3.0 and 4.0 % vol. (i.e. 39, 78, 155, 233 and 310 kg/m3). Specimens and curing conditions The mixtures without and with fibres were molded and densified in gravitational manner. Beams dimensions were as follow: 40x40x160 mm. Preliminary setting of concretes lasted for 6 or 24 hours at +20oC, during this time evaporation of water was prevented. After demoulding the specimens were cured under three different conditions: - curing in water (W) at +20oC after 24 hours of preliminary setting, - steam curing (S) at +90oC after 6 hours of preliminary setting, according to cycle presented on Fig. 2, - autoclaving (A) at +250oC and under pressure of 40 bars after 24 hours of preliminary setting, according to cycle presented on Fig. 2.

04-The influence.qxd

36

9/19/2009

11:55 AM

Page 36

Tomasz ZDEB and Jacek ŒLIWIÑSKI

Fig. 2 Temperature versus time for the two curing conditions

PROCEDURES AND TEST RESULTS Compressive and tensile strength at bending The tensile strength of each material was obtained by testing 6 beams (40x40x160 mm) in three-point bending test. 12 cubes (40x40x40 mm) were cut from the broken beams and tested for compressive strength. Regardless of steel fibres volume, the cross section confirmed that each specimen has homogeneous fibres dispersion (see Fig. 3).

Fig. 3 Cross sections of tested beams with homogeneously dispersed steel fibres Table 6 presents average values of compressive and tensile strength at bending. In case of curing in water (W) , specimens were examined after 28 days of setting. Steam cured (S) and autoclaved (A) specimens were tested after heat treatment was completed and test specimens were cooled down. Table 6 Compressive and tensile strength of RPC with and without steel fibres, cured in three different conditions Variable volume fraction Without Curing of fibres [% vol.] fibres Features conditions [MPa] 0.5 1.0 2.0 3.0 4.0 W 194 202 201 208 211 217 Compressive S 212 219 215 224 228 235 strength [MPa] A 268 282 281 297 308 315 W 10.6 12.1 13.2 16.0 19.3 22.7 Tensile strength at S 14.3 14.5 15.6 17.2 22.7 23.0 bending [MPa] A 18.6 21.2 23.7 25.1 26.1 26.8

04-The influence.qxd

9/19/2009

11:55 AM

Page 37

The influence of steel fibres content and curing conditions on mechanical properties and deformability of reactive powder concrete at bending

37

Deformability at bending In the study, the three-point bending test for specimens with dimensions 40x40x160 mm was used. The distance between supports was equal to 100 mm. Loading of the specimen was determined by force gain. Dimensions of the specimens and the rate of loading were not consistent with ASTM C1018-97 [8]. Therefore, the following analysis of calculated parameters has only comparative character. Fig. 4 presents representative load-deflection curves that were recorded during studies.

Fig. 4 Representative load-deflection curves recorded in the bending test of RPC with variable steel fibres content; (a) setting in water, (b) steam curing, (c) autoclaving (x - the 1st crack) On the basis of the load-deflection curves for each RPC test specimen containing variable steel fibres, work of fracture WF and fracture toughness indices I5, I10 and I20 were calculated [8]. Calculated toughness indices I5, I10 and I20, showed that first cracking point (marked as x on Fig. 4) was defined as the point where nonlinearity in the load-deflection

04-The influence.qxd

38

9/19/2009

11:55 AM

Page 38

Tomasz ZDEB and Jacek ŒLIWIÑSKI

curve becomes evident. Values of work of fracture, toughness indices and their variability versus fibres content are presented on Fig. 5.

Fig. 5 Influence of curing conditions and fibre content on work of fracture WF (a) and fracture toughness indices I5 , I10 , I20, (b, c, d) respectively Microstructure of RPC The influence of curing conditions and fibre content on compressive and tensile strength as well as deformability at bending can be explained by microstructure formation of tested materials. There are at least three factors that permit to obtain ultra-high performance cementitious composites: - very compacted microstructure of CSH phase (Fig. 6), - very good adhesion of CSH phase to mineral inclusions (grains of grounded quartz and quartz sand) and also to steel fibres (Fig. 7 and 8), - fill in voids in microstructure of materials caused by crystallization of xonotlite and tobermorite during autoclaving (Fig. 9). Fig. 6 and 7 show microstructure formation of the composites cured in three different hydrothermal environments. Very compacted microstructure of CSH phase and its excellent adhesion to grains of cement (light inclusion) as well as to quartz grains (dark inclusion) can be observed in each case of curing conditions. Steel fibres with tightly sheathed products of cement hydration are presented on Fig. 7 and 8. Each voids of autoclaved RPC (pores, microcracking etc.) are completely or partly filled with crystals of xonotlite and tobermorite (Fig. 9).

04-The influence.qxd

9/19/2009

11:55 AM

Page 39

The influence of steel fibres content and curing conditions on mechanical properties and deformability of reactive powder concrete at bending

Fig. 6 CSH phase (a) curing in water, (b) steam curing, (c) autoclaving, SEM, magnification 10 000x (photo: B. Trybalska)

Fig. 7 Contact zone of steel fibre and CSH phase in RPC, (a) curing in water, (b) steam curing, (c) autoclaving, SEM, magnification 2 000x (photo: B. Trybalska)

Fig. 8 Steel fibre in RPC, (a) curing in water, (b) autoclaving, SEM, magnification 500x (photo: B. Trybalska)

Fig. 9 Partly filled micropore with xonotlite and tobermorite crystals, (a) magnification 2 000x, (b) magnification 10 000x (photo: B. Trybalska)

39

04-The influence.qxd

9/19/2009

11:55 AM

Page 40

Tomasz ZDEB and Jacek ŒLIWIÑSKI

40

DISCUSSION Compressive strength Depending on curing conditions the compressive strength of RPC without fibres varies from 194 to 268 MPa. The lowest value is ascribed to material setting in water, while the highest to autoclaved one. After obtained test results, presented in Table 6, it could be affirmed that steam curing causes increase of compressive strength about 10%, while autoclaving about 40% in comparison to material setting in water. Increase of compressive strength of RPC with variable fibre volume are similar for steam curing as well as autoclaving. The values of compressive strength of RPC with fibres varies from 202 MPa (Vf=0.5% vol. setting in water) to 315 MPa (Vf=4.0% vol. autoclaving). Linear progression of compressive strength versus steel fibres volume fraction can be observed for all studied curing conditions, [10]. Moreover it is worth to notice that regardless of curing conditions, the influence of steel fibres volume on increase of compressive strength is similar. Only in case of autoclaved RPC containing 2 to 4% vol. of fibres a little deviation in plus can be observed. Tensile strength at bending Depending on curing conditions the flexural strength of RPC without fibres ranges from 11 to 19 MPa. As well as in case of compressive strength, above mentioned extreme values are ascribed to setting in water and autoclaving respectively. The steam curing contributes to 35% flexural strength enhancement, while autoclaving to 75% in comparison to material setting in water. The tensile strength of RPC with variable volume of fibres varies from 12 MPa (Vf=0.5% vol. setting in water) to 27 MPa (Vf=4.0% vol. autoclaving). The influence of fibres volume on discussed strength is greater than in case of compressive strength. The highest, more than double, increase of tensile strength was observed for RPC cured in water and containing 4 % vol. of fibres. In case of steam cured and autoclaved materials with the same volume of fibres the increase of tensile strength was equal to 60 and 45%, respectively. Like in case of compressive strength, the tensile strength for all curing conditions increases in linear rate versus volume fraction of fibres, [10]. Opposite to compressive strength test results (Table 6), the influence of hydrothermal treatment on tensile strength of RPC versus volume fraction of steel fibres is less distinct. Deformability at bending Work of fracture enhances, with increase of steel fibres volume. It is worth to notice that regardless of fibre volume in all considered cases, work of fracture is the highest for autoclaved materials (Fig. 5a). This phenomenon could be explained by two following mechanisms: the first is strengthening of mineral matrix of the composite as a result of hydrothermal treatment at 250oC and the second is the increase of adhesion between matrix and fibres. Materials that were steamed and cured in water show nearly identical values of fracture toughness indices for whole range of fibres dosage (Fig. 5b,c,d). The maximum value of indices I5, I10 and I20 can be observed in composites containing 3% vol. of fibres. Further increase of fibre volume (Vf = 4% vol.) leads to decrease of all determined toughness indices. This effect is ascribed to steamed and cured in water materials, while the autoclaved composites do not reveal such a tendency. The similar tendency was observed in case of mortars with addition of carbon fibres. Fracture toughness indices increase until 1% of fibre volume and their further growth causes decrease of their values, [12].

04-The influence.qxd

9/19/2009

11:55 AM

Page 41

The influence of steel fibres content and curing conditions on mechanical properties and deformability of reactive powder concrete at bending

41

This can be attributed to increase of strength of mineral matrix. Calculated fracture toughness indices I5 and I10 suit to results presented by Katz et al. [13]. The authors showed 6.4 and 11.4 values for indices I5 and I10, respectively. Their specimens had the same dimensions, were cured in natural conditions and contained as much as 9% vol. of steel fibres (length 6 mm, diameter 0.16 mm). In turn, autoclaved RPC require relatively high energy to trigger off the first crack in mineral matrix, what entails decrease of all fracture toughness indices up to 3% vol. content of steel fibres. CONCLUSIONS The presented test results allow to propose following conclusions: 1. It is possible to receive ultra-high performance concrete using regular components and generally traditional curing conditions. In case of material without fibres and cured in natural conditions its obtained compressive strength is about 200 MPa and tensile strength in bending 11 MPa. Application of steam curing allows to increase both strengths up to 212 and 14 MPa, respectively, while autoclaving even to 268 MPa and 18 MPa. 2. Addition of steel fibres allows to further increase both compressive and tensile strength. The highest values reached 315 MPa for compressive strength and 27 MPa for tensile strength. These parameters were obtained for autoclaved composite, containing 4% vol. (310 kg/m3) of steel fibres. 3. On the basis of presented test results of compressive and tensile strength (Table 6), it can be observed that content of examined steel fibres (l = 6mm, ĳ = 0.16 mm) should not exceed 3% of volume fraction. Further increase of fibre dosage does not bring significant advantages. 4. Content of steel fibres and curing conditions strongly influence deformability of reactive powder concrete in bending test. In order to obtain optimum ductility of tested composites, addition of steel fibres should not exceed 3% vol. It is similar phenomenon which was observed during determination of compressive and tensile strength. 5. Taking into consideration presented test results; it seems to be purposeful to continue research in order to determine impact of hybrid fibres, with variable length on mechanical properties of RPC. REFERENCES 1. Collepardi S., Coppola L., Troli R., Collepardi M., Mechanical Properties of Modified Reactive Powder Concrete, American Concrete Institute, 173, 1997, 1 – 22 2. Richard P., Cheyrezy M., Composition of Reactive Powder Concrete, Cement and Concrete Research, vol. 25 Nr 7, 1995, 1501 – 1511 3. Blais P., Couture M., Precast, prestressed pedestrian bridge – World’s first reactive Powder Concrete Structure, PCI Journal IX-X 1999, 60-71 4. Acker P., Behloul M., Ductal technology: a large spectrum of properties, a wide range of application, International Symposium on Ultra High Performance Concrete, Kessel Germany, 2004, 11 – 23 5. Herold G., Müller H. S., Measurment of porosity of ultra high strength fibre reinforced concrete, International Symposium on Ultra High Performance Concrete, Kessel Germany, 2004, 685-694 6. Staquet S., Espion B., Influence of Cement and Silica Fume Type on Compressive Strength of Reactive Powder Concrete, 6th International Symposium on High Strength /High Performance Concrete, Leipzig, June 2002, 1421-1436 7. Orgass M., Klug Y., Fibre Reinforced Ultra-High Strength Concrete, International

04-The influence.qxd

42

9/19/2009

11:55 AM

Page 42

Tomasz ZDEB and Jacek ŒLIWIÑSKI

Symposium on Ultra High Performance Concrete, Kessel Germany, 2004, 637-647 8. ASTM C1018-97 Standard test method for flexural toughness and first-crack strength of fiber-reinforced concrete (using beam with third-point loading), Withdrawn 2006 9. Dong Joo Kim, Naaman, A.E., El-Tawil, S., Comparative flexural behaviour of four fiber reinforced cementitious composites, Cement & Concrete Composites, 30, 2008, 917-928 10. Zdeb, T., ĝliwiĔski, J., Effect of curing conditions and steel fiber addition on the compressive and flexural strength of reactive powder concrete (in Polish), InĪynieria i Budownictwo, 12, 2008, 693-697 11. Cherezy M., Maret V., Frouin L., Microstructural analysis of RPC (Reactive Powder Concrete), Cement and Concrete Research 25, 1995, 1491 – 1500 12. Brandt A. M., Cement-based composites, materials, mechanical properties and performance, Routledge, Taylor and Francis Group, London and New York, 2009 13. Katz A., Bentur A., Dancygier A., Yankelevsky D., Sherman D., Ductility of high performance cementitious composites, Concrete Science and Engineering. A Tribute to Arnon Bentur, International RILEM Symposium, Evanston, IL, USA, March 2004, 117-127

05-Interface bond.qxd

9/19/2009

11:56 AM

Page 43

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

INTERFACE BOND CHARACTERISTICS BETWEEN WOOD FIBRES AND A CEMENT MATRIX M. Guadalupe SIERRA BELTRAN1, a, Erik SCHLANGEN1,b Microlab, M & E, Faculty of Civil Engineering and Geosciences Delft University of Technology, P.O. Box 5048, 2600 GA Delft, The Netherlands a [email protected], [email protected] 1

ABSTRACT When a brittle cement matrix is reinforced with fibres the composite’s toughness improves and a ductile behaviour could be developed. Wood fibres are currently being studied as a less energy-intensive alternative to synthetic fibres that have been successfully applied in reinforcing cementitious materials. In order to maximize the potential of wood fibres as reinforcement the fibre-matrix interface is being studied. The bonding will depend on: physical and chemical properties of the bundles, water/cement ratio of the matrix and its admixtures, curing conditions and age. One common way to determine important interface parameters such as chemical bond, frictional bond and slip-behaviour, either softening or hardening, is from a single fibre pull-out test. Therefore pull-out tests have been conducted using different matrices and softwood fibres. The effects of air-curing and water-curing are evaluated, as well as the influence of fibre embedment lengths. Beside pull-out tests the fibrematrix interface is also evaluated through Environmental Scan Electron microscope ESEM, CT-scan images and optical microscopes. The results from this study should allow the researchers to tailor the interface properties in order to achieve the desired composite behaviour.

Keywords Wood fibre, cement matrix, pull-out tests, bonding

INTRODUCTION In order to turn a brittle cement matrix into a ductile composite different types of man-made energy intensive fibres are currently in use: stiff fibres such as glass or steel and flexible fibres as polyvinyl alcohol, polypropylene and polyethylene fibres. However, in this study natural fibres are proposed. They are more easily available worldwide as well as cheaper and friendlier to the environment compared to the artificial fibres. Natural fibres in general demand less energy to be produced than synthetic fibres do. However, the properties of the wood fibres are not as constant as those of artificial fibres: they shrink and swell in the presence of water and they have a lower tensile strength. From the different types of wood, softwood fibres have been chosen for this study because they are longer and more uniform than hardwood fibres. Additionally, this paper proposes the use of bundles of wood fibres instead of single fibres. Bundles will achieve slightly lower tensile strength than single fibres but bundles have the advantage over single fibres that they have a rough surface which may promote a better mechanical bond with the cement matrix. In order to maximize the potential of wood bundles as reinforcement the fibre-matrix interface is studied. The bonding will depend on: physical and chemical properties of the bundles, water/cement ratio of the matrix and its admixtures, curing conditions and age. Pullout tests have been conducted using different matrices and softwood bundles to determine important interface parameters such as chemical bond, frictional bond and slip-behaviour. Six

05-Interface bond.qxd

9/19/2009

44

11:56 AM

Page 44

Mercedes G. SIERRA BELTRAN and Erik SCHLANGEN

different mix designs and three softwood types (pine, larch and spruce) have been used. Some of the mix designs are high strength mortars and all include by-products admixtures. The effects of air-curing and water-curing are evaluated, as well as the influence of fibre embedment lengths. During the study two different tests setups have been used. Beside pullout tests the fibre-matrix interface is also evaluated using optical microscopes, Environmental Scan Electron microscope ESEM and CT-Scan images. The pullout tests results and the ESEM and CT-Scan images are finally presented and discussed.

EXPERIMENTAL PROGRAM Materials Wood. Softwood fibres from spruce, larch and pine are used in this study. Fibres from softwood have a more simple form without vessels compared to hardwood fibres and the fibres are also longer (3 to 5 mm average) and thicker (up to 45 Pm). Softwood was found by Blankenhorn et al. [1] to be less inhibiting to cement setting than hardwood is. For larch and spruce and pine two different pulping procedures have been used. To obtain bundles of fibres out of lumber pieces of larch and spruce blocks of 1x1x2 cm have been cooked following a neutral sulphite semi-chemical (NSSC) pulping procedure as described by Walker [2] after which the bundles were manually taken apart. The bundles from pine on the other hand were prepared out of a veneer sheet which was first cut into pieces of about 2x3 cm and then cut into bundles using a microtome. Due to the difference in pulping procedure the pine bundles have a uniform rectangular section while the spruce and larch bundles cross sections vary in dimensions and shape. All bundles were cut at length 10 mm. The bundles have a tensile strength of 700 to 1000 MPa, and Modulus of elasticity of 25 to 40 GPa according to Sierra and Schlangen [3]. From now on these bundles will be referred as fibres. Cement matrix. Different cement matrices, of which the proportions per weight are shown in Table 1, were tested. The materials used are cement type CEM I 52.5 N, pulverized fly ash, blast furnace slag, limestone powder (Zhou et al [4]) and sand of which the size ranges from 0.125 – 0.25 mm. The last column in the table shows the water-cementitious material ratio. Table 1. Compression strength (MPa) and mix proportions (weight percentage) Fly Mix Compr. Sand BFS Lime. SP Water Cement Ash ID strength M1 1 1.2 0.8 0.013 0.6 67 M2 1 2.8 0.030 1 NC M3 1 0.8 1.2 0.013 0.6 58 M4 1 1.2 0.8 0.025 0.81 NC M5 1 1.2 1.5 0.023 0.98 48 M6 1 1.2 2 0.018 1.09 41

w/cm 0.27 0.26 0.27 0.27 0.27 0.26

NM Not measured

Fibre-matrix interface To evaluate the fibre-matrix interface properties single wood fibre pullout tests are performed. In these tests a fibre with a controlled embedment length is pulled out from a block of matrix while the load vs. displacement relation is recorded (Li et al. [5]).

05-Interface bond.qxd

9/19/2009

11:56 AM

Page 45

45

Interface bond characteristics between wood fibres and a cement matrix

Two sets of specimens were prepared. In the first set the fibre embedment length and the matrix length were the same while the second set was prepared in such way that the fibre was completely embedded in a cement matrix and the fibre embedment length and the cement matrix were independent from each other. The embedment lengths for both sets were chosen from 1 to 5 mm. A shorter length will ensure full debonding, while a longer length will lead to a higher pullout stress. The maximum embedment length will be half the total fibre length (5 mm in this case). Both sets of samples were prepared by hand lay up of fibres in the centre of the cement matrix along the specimen length (Fig. 1). In the first set each specimen was cut 28 days after casting to dimensions 6 mm thick, 8 mm wide, and 1 to 5 mm long. In the second set each specimen was cut to dimensions 12 mm thick, 10 mm wide and 5 or 10 mm long. All the specimens were demolded after 1 day. Some samples were thereafter kept continuously in water and the other samples were air cured at 20% RH and about 22 °C for at least 28 days.

a

b

Casting direction

Casting direction Cutting line Cutting line 10 mm

8 mm 6 mm

Fibre 1 to 5 mm

12 mm

1 to 5 mm

Fibre

5 or 10 mm

Fig. 1 Pullout specimens: (a) same matrix-fibre embedment length, (b) fibre fully embed Microstructure analysis The fibre-matrix interface microstructure was analyzed by optical microscope (Leica M26), electronic scanning electron microscope (ESEM, XL30 FEI) and computer tomography (CTScan Phoenix). Also an EDAX energy dispersive x-ray analyzer system (EDS) was used for determining phase chemistry of the interface. For the ESEM and XDS analyses tested and untested samples were cut and polished along their width showing the fibre cross section as well as the interface. Different microstructure features were studied, such as the alignment of the fibre inside the cement matrix, the matrix penetration into the fibre or the gaps between fibre and matrix and the chemical elements present in the interface. Once the microstructure of the different composites was characterized it was correlated with the mechanical properties of the samples. Pullout tests The two sets of specimens were tested with two different setups for the pullout tests. All tests were conducted at speed of 0.002 mm/s. For the first set of specimens the pullout tests were carried out in a MTS 810 testing machine (Fig. 2a) in which the displacement of the test is given by the displacement of the actuator (±80 mm stroke) and the results data interpretation and calculation of the interface parameters will be done according to Lin et al. [6]. A 2-N load cell is used to measure the pullout forces. The cement matrix sample was glued to a specimen mount plate and the free side of the fibre was glued to a fibre mounting plate.

05-Interface bond.qxd

9/19/2009

11:56 AM

46

Page 46

Mercedes G. SIERRA BELTRAN and Erik SCHLANGEN

Fig. 2 Pullout tests set up: (a) same matrix-fibre embedment length, (b) fibre fully embeded The second set of pullout specimens were tested in a micro tension-compression testing device (developed by Kammrath & Weiss) (Fig. 2b) where the pullout forces were measured with a 500-N load cell. The displacement of the test is given by the displacement of the actuator (±6 mm). The cement matrix sample in one side and the free fibre end on the other were glued to two steel non-rotating loading plates with a two component epoxy resin. In both tests a free fibre length of 1 mm was left between the cement matrix and the plate. The force and displacement were recorded during the tests.

Pullout load

Slip hardening

Pmax

2

Pfr Constant friction 1 Slip softening Displacement Debonding

Fibre slippage

3

le embedment length

Fig. 3 Schematic description of fibre pullout behaviour A schematic pullout load-displacement curve is shown in Fig. 3. As described by Redon et al [7] there are three zones that can be clearly distinguished. Initially a stable debonding process takes place between the fibre and the matrix, zone 1, until reaching a maximum pullout resistance by the fibre Pmax. Then, at zone 2, the load decreases from Pmax to Pfr pullout frictional load. If this drop of load is sudden the chemical bond between the fibre and the matrix is assumed to be broken. At this point fibres with a homogeneous cross section such as polyvinyl alcohol (PVA), polyethylene or steel, are fully debonded. For natural fibres with a variable cross section and irregular fibre surface this may not be the case. At the slippage zone (zone 3), the fibre load is resisted by frictional forces as mentioned by Redon et al [7]. Depending on the relative hardness of the fibre and the matrix, and the curvature of the fibre

05-Interface bond.qxd

9/19/2009

11:56 AM

Page 47

47

Interface bond characteristics between wood fibres and a cement matrix

in the cement matrix the fibre undergoes sliding with slip hardening, constant friction or slip softening effect. RESULTS Fibre type When comparing the overall results for the three types of wood, spruce and larch fibres show higher maximum pullout stress compared to the pine fibres. Fig. 4 shows the results for pullout tests of 1 mm thick air cured samples of pine and spruce. Here the spruce fibres embedded in M4 and M6 have higher pullout stress than pine fibres.

Pullout stress (MPa)

Pine

0,8

Spruce

0,6 0,4 0,2

Pullout stress (Mpa)

0,8

1

0,6 0,4 0,2 0

0

0

M1

M2

M4

M6

Fig. 4 Pullout stress of air cured samples (1 mm thick)

Fig. 5 Pullout curves of pine (M6, air cured)

Air cured Water cured

2 1,5 1 0,5 0

Pullout Stress (MPa)

Pullout stress (MPa)

3 2,5

1,6 Air cured

1,2

Water cured

0,8 0,4 0 0

M1

M2

M3

M4

M5

M6

1

0,5 Displacement (mm)

0,5 1 Displacement (mm)

1,5

Fig. 6 Pullout stress of spruce fibres (1 mm thick) Fig 7 Pullout curves of spruce fibres and M6 Regarding the pullout behaviour, in 54% of the pine fibres specimens the fibre broke during the test. The fibres that undergo sudden rupture have a higher ultimate stress than the ones that were pulled out due to a very strong interface bond. 60% of the cases where the fibre was pulled out have an interface bond mainly governed by friction. This behaviour can be seen in Fig. 5 where several load-displacement pullout curves recorded for pine fibres embedded in M6 are presented. In these curves there is no sudden drop from the maximum stress Pmax to maximum friction stress Pfr. The larch fibres show mixed results. In 30% of the cases the fibre broke, in 30% the pullout behaviour is mainly frictional and in 40% there is a noticeable chemical bond. The spruce fibres on the other hand have a strong chemical bond with most of the cement matrices tested (66%) and only 13% of the spruce fibres composites undergo fibre rupture during the pullout test.

05-Interface bond.qxd

9/19/2009

11:56 AM

48

Page 48

Mercedes G. SIERRA BELTRAN and Erik SCHLANGEN

Cement matrix In fibre reinforced composites the pullout rather than rupture will favour the development of ductility (Lin and Li [8]). A higher amount of energy will be released during pullout if the fibre undergoes sliding with slip hardening or constant friction (Fig. 3). Pine samples with matrices M4 and M6 show these effects. Specimens tested with pine and 1 mm thick samples of M1 and M4 have the higher stress values (Fig. 4). Spruce fibres on the other hand have the best pullout behaviour with matrices M1 and M4. Stress wise M3 has the higher stress values for both air and water cured spruce samples (Fig. 6). Larch casted in M3 matrix shows good pullout behaviour. From EDX measurements there are very small differences in the atomic components recorded in the fibre-matrix interface (less than 5 um from the fibre) and the components recorded at 250 Pm distance, or in bulk matrix. Curing condition For both air and water cured pine samples tested, the main pullout behaviour is fibre rupture with a higher percentage within water cured samples. Air cured specimens show, beside fibre rupture, distinctive non chemical bond behaviour. In order to understand this behaviour both tested and untested water and air cured samples for all the fibres were analyzed with the ESEM. The samples were prepared by impregnating them with epoxy but not following the common procedure in which the samples are first put in a vacuum chamber. The samples were not vacuum to avoid further volume changes in the fibres. The fibres swell with the water present in the cementitious mixture. In this initial moment cement paste next to the fibre hydrates creating the thin layer of reacted paste shown in Fig. 8. Later as the cement hydration continues and if the curing conditions do not provide additional water (as with air curing) the fibre shrinks and breaks the young cement paste leaving a gap that was filled with epoxy when preparing the sample for the ESEM images. When the fibre shrinks the chemical bond with the cement matrix is lost. From the samples observed the gap width range between 3 and 30 Pm. In images of tested fibres and videos of pullout tests done in the ESEM patches of cement matrix covering the fibres have been observed. Those patches are these thin layers of reacted paste. ESEM images of water cured specimens reveal a sound and intimate contact between the fibre and cement matrix. 30

Cement matrix

Thin cement layer

25

Load (N)

Epoxy

20 15 10 5

Fibre

0

0

Fig. 8 ESEM image of pine fibre-M3 interface

0.5 Displacement (mm)

1

Fig. 9 Pullout curves of pine fibres, M2, water cured

Pine fibres embedded in M2 matrix have a higher pullout stress when water cured. Two load-displacement pullout curves recorded from these samples are presented in Fig 9 showing

05-Interface bond.qxd

9/19/2009

11:56 AM

Page 49

49

Interface bond characteristics between wood fibres and a cement matrix

different behaviour: one fibre broke during testing and the other fibre was fully pulled out with a strain hardening effect. As mentioned before for spruce fibres, air and water cured, the main pullout behaviour is slip softening. Within the water cured samples the percentage of slip softening was even higher than for air cured samples. Air cured samples of spruce embedded in M4 show mainly a non-chemical bond while water cured samples show chemical bond behaviour. With this mix the pullout stress was about the same for both curing conditions. 1 mm thick samples of M3 and M6 that were water cured have higher pullout stress than the air cured samples as shown in Fig. 6. Two curves with different behaviour due to different curing conditions can be seen in Fig. 7. The air cured sample shows a non chemical bond while the water cured sample has a chemical bond assumed as the first peak in the curve and later with an increase in stress that corresponds to slip hardening. Larch fibres tend to break when pulled out of an air cured specimen. On the other hand most water cured samples have constant friction effect (Fig. 10).

12 Air cured

12

Water cured

8 4 0 0

1 Displacement (mm)

2

Pullout stress (MPa)

Pullout Stress (MPa)

16

10 8 6 4 2 0 0

1

Area (mm2)

2

3

Fig. 10 Pullout curves of larch fibres and M3 Fig. 11 Pullout stress vs. embedment area for larch fibres, M3, water cured Interface geometry The wood fibre is a natural mini-composite by itself with naturally bonded filaments. The shape of the fibres made partially or completely mechanically is not uniform. The fibres have a rough surface with short filaments coming out which may help the adherence and mechanical bond resulting in an effective reinforcement for a cement matrix composite as long as the fibres do not shrink. The mechanical bond is also increased by the fact that the fibres are not straight. The fibres curve naturally even inside the cement matrix. Images produced by a CT scan of single fibre embedded in a cement matrix reveal the importance of the curving angle in which the fibres are inside the matrix and its influence in the behaviour during the pullout. For example: A pine fibre in matrix M1 with 5 mm embedment length that showed a constant friction effect when pullout has an angle of 3.6° while a fibre that broke during the test had an angle of 11.6° (Fig. 12). In Fig. 12a it is noticeable that the embedment end has slipped inside the fibre tunnel before rupture. Larch fibres composites with pullout areas between 1 and 2.5 mm2 have the best behaviour when tested and the highest pullout stress. The pullout area is calculated as the perimeter of the fibre times the embedment length. For larch samples in M3 and water cured an optimal area of 1.7 mm2 is calculated (Fig. 11). Regarding the angle of curvature the CT-scan showed that a larch fibre that broke when tested had an angle of 27° while for a sample that was fully pulled out with a non-chemical bond the angle was 15°. From images produced for spruce fibres samples in M2 matrix the

05-Interface bond.qxd

50

9/19/2009

11:56 AM

Page 50

Mercedes G. SIERRA BELTRAN and Erik SCHLANGEN

angle of a fibre that shows slip softening behaviour is smaller (6.9°) than for a fibre that shows slip hardening (8.2°). Spruce fibres embedded in M3 and water cured have a clear stress- area relation, where the smaller the pullout area, the higher the stress will be. For the same fibre and matrix type, embedment length of 1 to 2 mm, but air cured the results were too scattered. There is a tendency to lower pullout stress with bigger area but it is not so clear.

a

b

Fig. 12 CT-Scan of pine and M1 air cured samples: (a) fibre rupture, (b) constant friction

SUMMARY AND CONCLUSIONS The interface behaviour between softwood fibres and cement matrices depends on the fibre type, the cement matrix composition, the curing conditions and fibre geometry, as shown in this study. Among this factors it is necessary to highlight the lost of interface bond due to the volume changes of the fibres in the presence of water. Therefore fibres should be treated in order to reduce these changes and keep the chemical and mechanical bond between wood fibres and cementitious matrices. On the other hand the natural curving from the fibres increases the mechanical bond. Currently the wood fibres are being treated successfully to reduce the shrinkage and swelling and therefore promote a constant and sound interface bond. Future work also includes the design of new cement matrix with lower strength since this will promote better pullout behaviour. Depending on the type of fibres certain matrices with different by products produce the highest stress or the better pullout behaviour. In order to further study the interface IR-spectroscopy to investigate a possible organic bond between the wood fibres and the cement matrix ACKNOWLEDGEMENTS Financial support for this research from the National Council of Science and Technology (CONACYT), Mexico (Grant # 206108) and from INTRON is gratefully acknowledged. Furthermore the authors thank Prof. V. C. Li and Prof. K. Van Breugel for valuable discussions on the subject. REFERENCES 1. Blankenhorn, P.R., Blankenhorn, B.D., Silsbee, M.R., DiCola, M., Effects of fiber surface treatments on mechanical properties of wood fiber-cement composites. Cement and Concrete Research, 31, 2001, 1049-1055

05-Interface bond.qxd

9/19/2009

11:56 AM

Page 51

Interface bond characteristics between wood fibres and a cement matrix

51

2. Walker, J.C.F. Pulp and paper manufacture. In: “Primary wood processing principles and practice”. Springer, The Netherlands 2006, 477-534 3. Sierra Beltran, M.G., Schlangen, E., Wood fibre reinforced cement matrix: a micromechanical approach. Key Engineering Materials, 385-387, 2008, 445-448 4. Zhou, J., Qian, S., Sierra Beltran, M.G., Ye, G., van Breugel, K., Development of Engineered cementitious composites with limestone powder and blast furnace slag. To be published in: Journal of Materials and Structures, RILEM, 2009 5. Li, V.C., Wu, C., Wang, S., Ogawa, A., Saito, T., Interface Tailoring for Strain-hardening PVA-ECC. ACI Materials Journal, 99, 2002, 463-472 6. Lin, Z., Kanda, T. and Li, V.C., On Interface Property Characterization and Performance of Fiber Reinforced Cementitious Composites. Journal of Concrete Science and Engineering, 1, 1999, 173-184 7. Redon, C., Li, V.C., Wu, C., Hoshiro, H., Saito, T., Ogawa, A., Measuring and modifying interface properties of PVA fibers in ECC matrix. Journal of Materials in Civil Engineering, 13, 2001, 399-406 8. Lin, Z., Li, V.C., Crack bridging in fiber reinforced cementitious composites with sliphardening interfaces. Journal of the Mechanics and Physics of Solids, 45, 1997, 763-787

06-Shear strength.qxd

9/19/2009

11:58 AM

Page 53

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

SHEAR STRENGTH AND DUCTILITY OF BEAMS REINFORCED WITH SYNTHETIC MACRO-FIBERS Salah ALTOUBAT 1, Samer BARAKAT2, Yazdanbakhsh ARDAVAN 3 and Klaus-Alexander RIEDER 4 1 Assistant Prof., University of Sharjah, Sharjah, UAE, [email protected] 2 Associate Prof., University of Sharjah, Sharjah, UAE 3 PhD Student, Texas A&M University, College Station, TX, USA 4 Senior Principal Scientist, Grace Construction Products, Luegde, Germany

ABSTRACT The application for synthetic macro fibers has been extended beyond shrinkage and cracking control to include some structural applications such as composite metal deck and industrial floors. One issue that has remained a matter of concern for engineers and designers is the shear behavior of the concrete when macro synthetic fibers are incorporated in the concrete mix. This paper is part of a recent research project that showed macro synthetic fibers to significantly improve the shear strength and ductility of reinforced concrete beams. The paper presents results from testing of 14 large beams under center-point monotonic loading. The beams were tested with a shear span to depth ratio of 2.3 and 3.5. The beam cross-section was 390 × 230 mm. Macro synthetic fibers were added at three volume fractions of 0.5 %, 0.75 %, and 1.0 %, which is equivalent to 4.6, 6.9, and 9.2 kg/m3. The results showed that the addition of macro synthetic fibers significantly improved the shear strength and ductility of the RC beams and modified the cracking and failure behavior. The ultimate shear strength of slender and short beams was increased up to 30% compared to the control RC beams.

Keywords beams; concrete; shear behavior; shear strength; macro synthetic fibers

INTRODUCTION Research over the past three decades has clearly established the potential use of fiber reinforcement for enhancing the shear capacity of reinforced concrete beams [1, 2, 3, 4, 5]. The effect of fiber reinforcement on shear strength of concrete beams is attributed to the improved post-cracking behavior of the concrete and the associated improvements in the aggregate interlock and dowel action of flexural reinforcement. Consequently, shear strength of the concrete beam, ultimate shear capacity and ductility of FRC beams are improved. Several research studies which involved small and large scale testing of FRC beams have confirmed this theory [6, 7, 8, 9, 10]. The majority of the past published research on shear of FRC beams has focused exclusively on steel fiber reinforced concrete (SFRC). The results showed that steel fibers can be used to boost the shear capacity of concrete and to improve the shear crack distribution in concrete structural members. A large database of test results for shear strength of steel fiber reinforced concrete (SFRC) beams obtained by many investigators was compiled by ParraMontesinos, [11]. Based on this, the list of beams exempt from the minimum shear reinforcement requirement in Section 11.5.6.1 of ACI 318-08 now includes beams that are

06-Shear strength.qxd

54

9/19/2009

11:58 AM

Page 54

Salah ALTOUBAT, Samer BARAKAT, Yazdanbakhsh ARDAVAN and Klaus-Alexander RIEDER

constructed with SFRC. This addition to ACI 318 represents the first permitted structural use of SFRC in the ACI building Code. Unlike in the case of SFRC, there are only few studies reporting results on shear with synthetic fibers [4, 12, 13, 14]. The limited research on shear behavior with synthetic fibers is attributed perhaps to the little increase in toughness and associated structural performance of concrete when low-modulus micro synthetic fibers, typically made of polypropylene, are added to concrete. In recent years however, increasing efforts have been devoted towards the development of new generation of macro synthetic fibers which impart significant toughness and ductility to concrete comparable to commonly used steel fibers. Accordingly, the application of macro synthetic fibers in the concrete industry has extended beyond shrinkage and thermal cracking control to structural applications, and therefore it becomes necessary to study the shear behavior of concrete beams with macro synthetic fibers. This study is part of a comprehensive experimental program conducted at the University of Sharjah, which focused on the shear behavior of beams reinforced with a new type of macro synthetic fiber [15, 16]. This fiber has a higher modulus of elasticity compared to regular polypropylene fibers and an optimized geometry to enhance the bond between the fiber and the concrete matrix, which leads to an increase in the toughness properties of concrete. Results obtained from testing 14 large scale beams under shear loading are summarized in this paper. The beams were reinforced with longitudinal flexural reinforcement and were designed to fail in shear under monotonic center-point loading in a simply supported configuration. The macro synthetic fibers were added to the concrete at three different volume fractions, namely 0.50%, 0.75% and 1.0% in addition to the control beams without fiber reinforcement. The beam cross section was 390 mm × 230 mm. Loaddeflection measurements and the shear capacity of the beams are presented and discussed in this paper.

EXPERIMENTAL PROGRAM Large beams testing Fourteen large-scale concrete beams were designed, instrumented and tested in displacement control mode under a monotonic three-point loading system in a simply supported configuration. The main experimental parameters in this test series were shear-span to depth ratio a/d, and the macro synthetic fiber content Vf. Two shear span to depth ratios were considered, namely 3.5 and 2.3 to reflect the behavior of slender and short beams, respectively. The cross-section of the beams was 390 mm × 230 mm. All beams were reinforced with three longitudinal reinforcing bars with a diameter of 32 mm, which corresponds to a steel ratio of 0.0318. The average cylinder compressive strength of the concrete used in this study was around 41 MPa (6000 psi). Macro synthetic fibers were added at volume fractions of 0.50%, 0.75% and 1.0%. Table 1 presents details of the beams. The beams were labeled to indicate the type of beam (short or slender), and the fiber content. The letter ‘L’ denotes long or slender (a/d =3.5) while ‘Sh’ means short (a/d =2.3). For example the beam labeled as L-0.50 stands for a slender beam with a fiber volume fraction of 0.50 %.

06-Shear strength.qxd

9/19/2009

11:58 AM

Page 55

55

Shear strength and ductility of beams reinforced with synthetic macro-fibers

Table 1 – Details of the large scale beams tested in the study Test Series L-0.0 L-0.50 L-0.75 L-1.0 Sh-0.0 Sh-0.50 Sh-0.75

No. of beams 2 2 2 2 2 2 2

h, mm 390 390 390 390 390 390 390

d, mm 330 330 330 330 330 330 330

b, mm 230 230 230 230 230 230 230

Length, Span, m m 2.7 2.31 2.7 2.31 2.7 2.31 2.7 2.31 1.9 1.5 1.9 1.5 1.9 1.5

a/d

ȡ, %

Vf , %

3.5 3.5 3.5 3.5 2.3 2.3 2.3

3.18 3.18 3.18 3.18 3.18 3.18 3.18

0.00 0.50 0.75 1.00 0.00 0.50 0.75

Identical pairs (duplicate) of slender and short beams were cast to enhance the reliability of the test results. The large beams and the companion lab specimens, including the 100 mm by 200 mm cylinders for compressive strength measurement and the 150 mm by 150 mm by 550 mm beams for flexural toughness measurement were all cast in one day in a modern precast plant. Materials The mix proportions and properties of the concrete used for casting the beams are presented in Table 2. The final water to cement ratio was around 0.47. The coarse to fine aggregate ratio was targeted to be close to 50:50 in order to maintain workability and to have sufficient paste volume for coating the fibers. The coarse aggregate used in this concrete mix was a Gabro gravel with a maximum aggregate size of 20 mm and a specific gravity of 3.1. The fine aggregate constituents were natural washed sand with a specific gravity of 2.6 and dune sand with specific gravity of 2.63. The 28-day compressive strength values measured according to ASTM C39-05 and flexural strength values measured according to ASTM C78-08 are summarized in Table 2. Typical values of the equivalent flexural strength, fe,3, measured according to JSCE-1984 [17] up to 3 mm deflection are also reported. The equivalent flexural strength, fe,3, was developed in Japan and can be obtained from bending of beams under third150 toughness value defined in the point loading. The fe,3 value is directly proportional to the T150 ASTM C1609-07 [18] flexural beam test using a 150 x 150 x 550 mm beam. The fe,3 value as 150 well as the T150 value are directly proportional to the area under the load-deflection curve up to a central beam deflection of 3 mm (for a support span of 450 mm). The main components of the macro synthetic fibers used in this study are polypropylene and polyethylene. This “macro” synthetic fiber’s mechanical properties and geometry are significantly different from traditional “micro” synthetic fibers, which are used to control plastic shrinkage cracking. The fiber’s nominal length is 40 mm and has an aspect ratio of 90 and a specific gravity of approximately 0.92. The fiber has a rectangular cross section with an average width of 1.4 mm and average thickness of 0.105 mm. The average tensile strength of the fiber is 620 MPa with a modulus of elasticity of 9500 MPa measured according to EN 14889-2. The fibers were added at volume fractions of 0.50 %, 0.75 % and 1.0%, which correspond to 4.6, 6.9 and 9.2 kg/m3, respectively.

06-Shear strength.qxd

56

9/19/2009

11:58 AM

Page 56

Salah ALTOUBAT, Samer BARAKAT, Yazdanbakhsh ARDAVAN and Klaus-Alexander RIEDER

Table 2 - Mix proportions and average properties of concrete

Vf = 0.0%

Vf = 0.50%

Vf = 0.75%

Vf = 1.0%

Coarse Aggregate, kg/m3 943 943 Fine Aggregate, kg/m3 942 942 Cement, kg/m3 380 380 Water, kg/m3 195 195 Superplasticizer Daracem 205, kg/m3 2.15 4.55 Water to Cement Ratio 0.474 0.474 Slump, mm 100 100 Cylinder Compressive Strength, MPa 42 42 Flexural Strength, MPa 5.8 5.8 Equivalent Flexural Strength (fe,3), MPa 0.15 2 1 kg/m3 = 1.686 lbs/cyd; 25.4 mm = 1 inch; 1 MPa = 145 psi

943 942 380 195 6.70 0.474 90 42 5.9 2.8

943 942 380 195 7.45 0.474 80 35* 5.8 3.1

Materials

* This mix exhibited signs of segregation due to high contents of fiber and superplasticizer

Testing and measurements The test setup consisted of a simply supported loading configuration with roller supports to prevent restraint to axial elongation. The beams were loaded at midspan (center point) and tested in displacement control mode using a hydraulic actuator with a capacity of 500 kN (Fig. 1). The loading rate was 0.48 mm/min to allow for accurate observation and mapping of the cracks development.

Fig. 1 – Test setup and the arrangement of LVDTs The parameters measured in the test program were beam deflections, strain in the concrete at different locations, strain in the reinforcing flexural steel bars, and the applied load. The cracking patterns were observed visually and marked during testing at different load levels. The load levels corresponding to formation of first diagonal shear crack as well as the ultimate shear capacity were determined for the beams using the load deflection and strain measurement data with the aid of visual observation. The formation of the first diagonal shear

06-Shear strength.qxd

9/19/2009

11:58 AM

Page 57

57

Shear strength and ductility of beams reinforced with synthetic macro-fibers

crack was associated with a sudden reduction in load carrying capacity of the RC beam in the load-deflection curve. The maximum load carried by the RC beam before failure was used to calculate the ultimate shear capacity of the beam. Further details of the experimental plan and test results can be found in Altoubat et al. (2009) [16].

EXPERIMENTAL RESULTS AND DISCUSSION Load-deflection results The load versus mid-span deflection curves of the tested slender and short beams are presented in Fig 2 and 3, respectively. The load versus deflection response of all beams were similar up to the load at which the first diagonal shear crack was formed in the control beams. The first diagonal crack was associated with a sudden load reduction in the load deflection curve. The control beams failed upon the formation of the first diagonal crack, and thus it marked their ultimate capacity. The addition of macro synthetic fibers increased the first diagonal crack load relative to the control beams as can be seen from the results shown in Fig. 2 and 3. Furthermore, the fiber concrete beams did not fail when the first diagonal shear crack was formed and the post cracking behavior was dependent on the fiber dosage. The beams with 0.50% of macro synthetic fibers exhibited little load reduction after the first diagonal cracking and then continued to resist similar or higher loads until another shear crack formed at a much higher deflection before failure. The beams with higher dosage of fibers (0.75 % and 1.0 %) exhibited no reduction in load after the first diagonal crack and continued to resist greater deflection and more shear cracks were developed before failure. The plateau in the load deflection curve in Fig. 2 for the beams with 0.75 % and 1.0 % of the fibers reflect the multiple shear cracks developed in the beams before failure and showed the improvement in ductility.

400

Control Vf = 0.50% Vf = 0.75% Vf = 1.0%

350

Load (kN)

300 250 200 150 100 50 0 0

1

2

3

4

5

6

7

8

9

10

11

12

Deflection (mm)

Fig. 2 – Load versus deflection curves of the slender beams

06-Shear strength.qxd

58

9/19/2009

11:58 AM

Page 58

Salah ALTOUBAT, Samer BARAKAT, Yazdanbakhsh ARDAVAN and Klaus-Alexander RIEDER

400 350

Control Vf = 0.50% Vf = 0.75%

Load (kN)

300 250 200 150 100 50 0 0

1

2

3

4

5

6

7

8

9

10

11

12

Deflection (mm)

Fig. 3 – Load versus deflection curves of the short beams Table 3 presents a summary of the loads corresponding to the formation of first diagonal shear crack and ultimate load capacity of the tested beams. The table also includes the percent increase of the load carrying capacity of the SNFRC relative to the corresponding control RC beams. The results show that the addition of macro-synthetic fibers to the concrete beams significantly increased the first diagonal cracking load and the ultimate load. Table 3 Loads at first diagonal crack and ultimate capacity of the beams Beam

a/d

First diagonal crack load, kN

Ultimate load, kN

Increase in ultimate load, %

L2-0.0 L2-0.5 L2-0.75 L2-1.0 Sh2-0.0 Sh2-0.50 Sh2-0.75 1 kN = 224.809 lbf

3.5 3.5 3.5 3.5 2.3 2.3 2.3

230 236 266 260 270 285 308

230 265 279 303 270 318 339

14 20 30 18 26

Table 4 presents a summary of the average deflection resisted by the RC beams at ultimate load. The macro-synthetic fiber increased the deflection at ultimate above the control beam by 63% to 100% for slender beams and by 103% to 138% for short beams as shown in Table 4. The deflection results showed the significant improvement in the ductility of the beams due to the addition of macro synthetic fibers. Table 4 - Average values of beam deflections at ultimate load Beams Slender beams Short beams 25.4 mm = 1 inch

Control

Vf = 0.50%

Defl., mm 3.7 1.7

Defl., Increase, mm % 6.0 63 3.5 103

Vf = 0.75% Defl., mm 7.1 4.1

Vf = 1.0%

Increase, Defl., Increase, % mm % 93 7.3 100 138 -

06-Shear strength.qxd

9/19/2009

11:58 AM

Page 59

59

Shear strength and ductility of beams reinforced with synthetic macro-fibers

Cracking and failure Cracking of the beams were carefully monitored and mapped during testing. Fig. 4 presents pictures of cracking pattern and sequence for slender and short beams. Flexural cracking in the slender beams started at mid-span and spread out to the shear span, where the flexural cracks - with increasing load - began to incline as flexural shear cracks (diagonal crack). The number of flexural cracks and the inclination of the diagonal crack that lead to failure characterize the cracking pattern of the tested RC beams. The control RC beams failed with a single and steep diagonal shear crack as can be seen in Fig. 4a. The SNFRC beams developed multiple flexural and diagonal shear cracks before failure occurred as can be seen in Fig. 4b and c. The creation of multiple flexural and shear cracks in the SNFRC beams contributed to the increase of the ductility as reflected in the load deflection curves, particularly for the RC beams reinforced with 0.75% and 1.0% of fibers. Furthermore, the primary diagonal crack that leads to failure of the SNFRC beams (Fig. 4b and c) was flatter relative to that of the control beam (Fig. 4a), and extended further toward the support. According to Fenwick and Paulay [19], appreciable arch action develops when the diagonal crack extends to the support, thereby separating the tension and compression zones of the shear span and allowing the relatively large translational displacement associated with arch action to occur. The cracking patterns of the SNFRC beams suggest that macrosynthetic fibers improve the arch action in slender RC beams and thus lead to an increase of the shear strength. Fig. 4d - e show the cracking pattern and sequence for short beams. The short control RC beams shown in Fig. 4d developed web-shear cracks that led to a sudden and brittle shear failure. Conversely, the SNFRC short beams developed flexural shear cracks before failure as can be seen in Fig. 4e and f. The addition of macro synthetic fibers changed the mode of failure of short beams from web-shear cracking to flexural shear cracking. This can be attributed to the fact that macro-synthetic fibers increased the shear strength of the beam to a level that was sufficient to mobilize flexural cracking prior to shear failure.

Slender

(a)

0.50% SNFRC beam Slender

Control RC beam

Control RC beam Short

Short

0.50% SNFRC beam (e)

(b)

0.75% SNFRC beam

0.75% SNFRC beam Slender

(d)

(c)

Short

(f)

Fig. 4 – Cracking sequence and patterns of the tested beams. Comparison with previous studies Few studies [12-14] that report experimental results on the shear capacity of slender concrete beams with different synthetic fibers are compiled and compared to the test results obtained in this study. The properties of the synthetic fibers used in those studies are listed in Table 5.

06-Shear strength.qxd

60

9/19/2009

11:58 AM

Page 60

Salah ALTOUBAT, Samer BARAKAT, Yazdanbakhsh ARDAVAN and Klaus-Alexander RIEDER

The previous published results showed that macro synthetic fibers enhanced the shear strength of the concrete beams. The improvement in shear strength of the SNFRC beams above the control beams due to the addition of macro synthetic fibers are compiled for the different studies in Fig. 5. The results presented in this study showed that macro synthetic fiber reinforced concrete increased the first diagonal cracking strength and the ultimate shear strength relative to the control RC beams as summarized in Table 3. The increase in shear strength above the reference beams ranged between 14% and 30% for slender SNFRC beams depending on the dosage of fibers and by more than 18% to 26% for short beams. This significant improvement in the ultimate shear strength is consistent with published results obtained by these previous investigators as shown in Fig. 5. Table 5 – Properties of synthetic fibers used in previous studies

Researcher

Noghabai[12]

Elastic Tensile Aspect strength, Modulus, ratio MPa MPa 275 2600

Specific Density, g/cc 0.90

Polyolefin

Fiber length, mm 50

Polyolefin

25

-

275

2600

0.90

54

360

375

3500

0.90

50

85

540

9500

0.92

50

63

400

-

0.91

40

90

640

9500

0.91

Fiber material

Polypropylene Majdzadeh et Polypropylene/ al.[13] Polyethylene Greenough et Polypropylene al.[14] Polypropylene/ Current study Polyethylene

% increase above control

50

40

30

20 Noghabai [12] Majdzadeh et al [13] Greenough et al [14] Current study

10

0 0

0.25

0.5

0.75 1 1.25 1.5 Fiber volume fraction

1.75

2

Fig. 5 – Increase in shear strength of slender SNFRC in previous studies

06-Shear strength.qxd

9/19/2009

11:58 AM

Page 61

Shear strength and ductility of beams reinforced with synthetic macro-fibers

61

SUMMARY AND CONCLUSIONS The results obtained from large scale testing of 14 longitudinally reinforced concrete beams with macro synthetic fibers were presented and discussed. Slender and short beams with shear span to depth ratios (a/d) of 3.5 and 2.3 were tested. The length of the beams was varied between 1.9 m and 2.7 m, and the macro synthetic fibers were added at volume fractions of 0.50%, 0.75% and 1.0%. Global deflection of the RC beams, load at first diagonal cracking and ultimate load were measured during the test. The cracking pattern was also monitored during the shear tests at different stages of loading. The experimental results showed that the addition of macro synthetic fibers significantly improved the shear strength and ductility of the RC beams and modified the cracking and failure behavior of the RC beams. Macro synthetic fibers at 0.50, 0.75, and 1.0 percent volume fraction increased the ultimate shear strength of slender beams by 14%, 23%, and 30%, respectively compared to the control beams. Similarly, macro synthetic fibers at 0.50% and 0.75% increased the ultimate shear strength of short beams by at least 18% and 26%, respectively. The load versus deflection curves showed that the SNFRC beams are more ductile compared to the control RC beams. The control beams failed abruptly upon the formation of the first diagonal crack whereas the SNFRC beams continued to resist higher load and deflection after the formation of the first diagonal crack. The addition of macro synthetic fibers at 0.50%, 0.75%, and 1.0% increased the deflection at maximum load for slender beams by 63%, 93% and 100% compared to the control beams. Similarly, macro synthetic fibers at 0.50% and 0.75% increased the deflection at maximum load for short beams by 103% and 138% relative to the control RC beams. This increase in the deflection capacity of the SNFRC beams demonstrates the significant improvement in ductility that macro-synthetic fibers could impart to the concrete beams.

ACKNOWLEDGEMENTS Support to this project was provided in part by W.R. Grace, Cambridge, MA, USA; Juma Almajid Company in Dubai; and by the Hazard and Risk Management Research Group at the University of Sharjah. The authors would like to acknowledge support provided for this project. REFERENCES 1. Batson, G.B., E. Jenkins, and R. Spatney, “Steel Fibers as Shear Reinforcement in Beams”, ACI Journal, Proceedings, vol. 69, No. 10, 1972, p. 640-644 2. Calixto, J.M., L.V. Filho, and C.M., “Goncalvez, “Shear behaviour of reinforced concrete beams with the addition of short steel fibers”, in 3rd International Conference on HighPerformance Concrete: Performance and Quality of Concrete Structures, Recife, Pernambuco, Brazil, eds. American Concrete Institute, Farmington Hills, Mich., 2002 3. Dupont, D. and L. Vandewalle, “Shear Capacity of Concrete Beams Containing Longitudinal Reinforcement and Steel Fibers”, Innovation in Fiber Reinforced Concrete for Value, SP-216, N. Banthia, M. Criswell, P. Tatnall, K. Folliard, eds., American Concrete Institute, Farmington Hills, Mich., 2003, pp. 79-94 4. Li, V., R. Ward, and A.M. Hamza, “Steel and Synthetic Fibers as Shear Reinforcement”, ACI Materials Journal, vol. 89, No. 5, 1992, pp. 499-508 5. Swamy, R.N. and H.M. Bahia, “Effectiveness of Steel Fibers as Shear Reinforcement”, ACI Concrete International, vol. 7 N. 3, 1985, pp. 35-40

06-Shear strength.qxd

62

9/19/2009

11:58 AM

Page 62

Salah ALTOUBAT, Samer BARAKAT, Yazdanbakhsh ARDAVAN and Klaus-Alexander RIEDER

6. Sharma, A.K., “Shear Strength of Steel Fiber Reinforced Concrete Beams”, ACI Journal, Proceedings, vol. 83, No. 4, 1986, pp. 624-628 7. Narayanan, R. and I.Y.S. Darwish, “Use of Steel fibers as Shear reinforcement”, ACI Structural Journal, vol. 84, No. 3, 1987, pp. 216-227 8. Mirsayah, A.A. and N. Banthia, “Shear Strength of Steel Fiber Reinforced Concrete”, ACI Materials Journal, vol. 99, No. 5, 1987, pp. 473-479 9. Mansur, M.A., K.C.G. Ong, and P. Paramsivam, “Shear Strength of Fibrous Concrete Beams without Stirrups”, Journal of Structural Engineering, vol. 121, No. 9, 1986, pp. 2066-2079 10. Kwak, Y.K., et al., “Shear Strength of Steel Fiber Reinforced Concrete Beams without Stirrups”, ACI Structural Journal, vol. 99, No. 4, 2002, pp. 530-538 11. Parra-Montesinos, G.J., “Shear Strength of Beams with Deformed Steel Fibers”, ACI Concrete International, vol. 28, No. 11, 2006, pp. 57-67 12. Noghabai, K., “Beams of Fibrous Concrete in Shear and Bending: Experiment and Mode”, Journal of Structural Engineering, vol. 126, No. 2, 2000, pp. 243-251 13. Majdzadeh, F., M. Soleimani, and N. Banthia, “Shear Strength of Reinforced Concrete Beams with a Fiber Concrete Matrix”, Canadian Journal of Civil Engineering, vol. 33, 2006, pp. 726-734 14. Greenough, T., Nehdi M., “Shear Behavior of Fiber-Reinforced Self-Consolidating Concrete Slender Beams”, ACI Materials Journal, vol. 105, No. 5, 2008, pp. 468-477 15. Yazdanbakhsh, A., “Shear Behavior of Synthetic Fiber Reinforced Concrete Beams”, Master Thesis in the Department of Civil and Environmental Engineering. University of Sharjah: Sharjah, UAE, 2008, pp. 125 16. Altoubat, S. A., Yazdanbakhsh, A., Rieder, K.-A.,” Shear Behavior of Macro-Synthetic Fiber Reinforced Concrete Beams without Stirrups” ACI Materials Journal, vol. 106, No. 6, July-Auguts-2009 17. JSCE-SF4, ”Methods of Tests for Flexural Strength and Flexural Toughness of Steel Fiber Reinforced Concrete”, Japan Society of Civil Engineers, Concrete Library International, No. 3, Part III-2, 1984, pp. 58-61 18. ASTM C 1609-07, “Standard Test Method for Flexural Performance of Fiber-Reinforced Concrete (Using Beam With Third-Point Loading),” ASTM International, 2007, West Conshohocken, Pa. 19. Fenwick, R.C. and T. Paulay, “Mechanisms of Shear Resistance of Concrete Beams”, Journal of Structural Division, ASCE, Proceedings, vol. 94, ST10, 1968, pp. 2325-2350

07-The impact.qxd

9/19/2009

11:58 AM

Page 63

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

THE IMPACT OF AMOUNT AND LENGTH OF FIBRILLATED POLYPROPYLENE FIBRES ON THE PROPERTIES OF HPC EXPOSED TO HIGH TEMPERATURE Izabela HAGER, Tomasz TRACZ Institute of Building Materials and Structures Cracow University of Technology Warszawska 24, 31-155 Kraków, Poland e-mail: [email protected]; [email protected]

ABSTRACT This paper presents the results of research on high performance concretes (HPC) modified with polypropylene fibres (PP fibres). The scope of the research was the measurement of the residual transport properties of heated and recooled concretes: gas permeability and surface water absorbability. Seven types of concrete modified with fibrillated PP fibres were tested. Three lengths: 6, 12 and 19 mm and three amounts of fibres: 0, 0.9 and 1.8 kg/m3 were used. The research programme was designed to determine which length of fibres, used in which minimum amount, will, after the fibres melt, permit the development of a connected network and pathway for gases and liquids.

Keywords High performance concretes, high temperature, spalling, PP fibres, gas permeability

INTRODUCTION The behaviour of high performance concrete (HPC) exposed to high temperature can prove a major limitation on the use of such concrete in the construction industry. HPC exposed to high temperature can be prone to explosive spalling, which could result in the exposure of steel reinforcement. This would create a vulnerability in the load-bearing capacity and stability of the structural element. According to various sources (Bazant [1], Hertz [2] , Kalifa et al. [3]), the explosive behaviour of concrete is a combination of two phenomena occurring in parallel in heated concrete: a thermo-mechanical effect and a hydro-thermal effect. The temperature increases the pressure of the gas and liquid contained in the material’s pores, accompanied by rapid moisture vaporization within the surface area of the heated concrete (hydro-thermal effect). Additionally, the difference of thermal strains in layers of concrete with different temperature (thermo-mechanical effect) becomes apparent in the heated material, and the accumulated energy is released violently, resulting in so-called explosive spalling. These two effects together cause unfavourable stresses in the concrete. In situations where stress exceeds the ultimate tensile strength of the concrete, explosive spalling can occur. As numerous tests have shown (Kalifa et al. [4], Nishida et al. [5], Hoff et al. [6] ) polypropylene fibres improve the stability of high performance concrete exposed to high temperature. At temperature near to 170°C the fibres melt. The melted polypropylene is partly absorbed by the cement matrix [4],

07-The impact.qxd

64

9/19/2009

11:58 AM

Page 64

Izabela HAGER and Tomasz TRACZ

creating a network of open pores, which increases permeability, and consequently reduces the internal pressure in the heated concrete. However, the exact mechanism by which PP fibres work is still not fully explained. In research conducted by Kalifa et al. [4], an increase was observed in the micro-cracking of heated concretes with PP fibres compared to concretes without fibres. This suggested that the presence of fibres in the cement matrix may be considered a discontinuity that favours the initiation and development of micro cracks during heating, contributing to an increase in concrete permeability. Moreover, it has been pointed out that when melting, polypropylene undergoes a transition from the crystalline to the amorphous phase (polypropylene density decreases from 910 kg/m3 to 850 kg/m3), which causes its volume to increase by approximately 7%, Pasquini [7]. The results presented here are a continuation of the research carried out by the authors and described in Hager and Tracz [8]. In that study, the properties of two types of concrete were compared: concrete without fibres and concrete with an addition of PP fibres amounting to 1.8 kg/m3. The results obtained suggested that the addition of PP fibres was effective and had the desired impact causing a considerable increase in concrete permeability after melting. The increase in the transport properties of the concrete was confirmed by measurements of its surface absorbability, which demonstrated that the capillary porosity of concrete with fibres when heated to 160÷200°C was higher than that of concrete without fibres. The current research, the results of which are presented in this paper, is an attempt to establish the optimum amount and length of PP fibres. From the technical point of view, the use of polypropylene fibres in quantities amounting to 0.1÷0.2% of concrete volume is an effective way of limiting the occurrence of spalling in HPC; however, there are no reports that would clearly indicate the optimum fibre length. According to the percolation model presented by Bentz [9], given a fixed amount of fibre, long fibres should be more efficient. The objective of this research was to assess the impact of the temperature to which concrete is heated on the concrete’s properties affecting its ability to allow liquid and gas transport. The research included the determination of such properties as gas permeability, surface water absorbability and microporosity for several types of concrete mortars before and after heating. These properties are the most relevant to the quantitative assessment of the effectiveness of polypropylene fibres as an addition limiting the occurrence of spalling in HPC. Moreover, the impact of heating temperature on changes in the compressive strength of samples heated to a temperature of 600oC was also assessed.

MATERIALS AND SAMPLE PREPARATION The study was carried out using concretes made from basalt aggregate and CEM I 52.5R cement. The quantitative and qualitative composition of all high-performance concretes examined was the same, with the exception of the amounts and lengths of the polypropylene fibres used. Fibrillated polypropylene fibres with lengths of 6, 12 and 19 mm, and in amounts of 0, 0.9 and 1.8 kg/m3 were used for the purposes of the research. According to the figures quoted by the manufacturer, the melting temperature for the fibres used was 163°C, and their flash point was 360°C. The fibrillated fibres used had a density of 0.91 g/cm3. Cylindrical samples with a diameter of 150 mm and 50 mm high were used for gas permeability and surface water absorbability measurements. These samples were sawn out from 150/300 mm standard cylinders. Compressive strength was determined using 100 mm cubic samples. All samples were formed and cured in compliance with the PN-EN 12390-2 standard, and the tests were conducted after 90 days of curing. Detailed compositions of the concretes that were prepared together with their designations are shown in Table 1.

07-The impact.qxd

9/19/2009

11:58 AM

Page 65

The impact of amount and length of fibrillated polypropylene fibres on the properties of HPC exposed to high temperature

65

Cement CEM I 52,5

kg/m3

490

Water

kg/m3

145

Sand 0/2 mm

kg/m3

611

Basalt aggregates 2/8 mm

3

kg/m

712

Basalt aggregates 8/16 mm Fibrilated PP fibres: length 6 mm length 12 mm length 19 mm

kg/m3

712

Plasticizer

% mc

0.9

Superplasticizer

% mc

1.0 ÷ 2.0

kg/m3 kg/m3 kg/m3

-

B100/1.8/19

B100/0.9/19

B100/1.8/12

B100/1.8/6

Type of component

B100/0.9/6

Unit

B100

Concrete labels

B100/0.9/12

Table 1 Mix designs of HPC with and without PP fibres

0.9 1.8 - 0.9 1.8 - 0.9 1.8

Complementary data Water-cement ratio

-

0.30

Cement paste content

dm3/m3

§ 310

Mortar content

dm3/m3

514

MEASUREMENT RESULTS AND ANALYSIS Concrete samples were heated in a Nabertherm LH30/13 furnace at a constant rate of 1°C/min. After the target temperature had been reached, the samples were heated for a further two hours in order to stabilise the temperature over their entire cross-section. The next stage was the free cooling of samples in the furnace to room temperature. Therefore all the properties established are residual ones. This procedure was applied to all the samples examined. Permeability Gas (nitrogen) permeability was determined using the RILEM-Cembureau method for samples with a diameter of 150 mm and a height of 50 mm cut from 150/300 mm standard cylinders. Permeability measurements started with the determination of initial permeability for samples dried to constant mass at a temperature of 105oC, according to the guidelines for the method applied described in RILEM Technical Recommendation [10]. Subsequently, the samples were heated to temperature ranging from 140 to 200oC, and their permeability was measured following cooling. The measurement results presented in Figure 1 show the ratio of permeability of concrete after heating to its permeability before heating (initial permeability). Each measurement series comprised three samples.

07-The impact.qxd

66

9/19/2009

11:58 AM

Page 66

Izabela HAGER and Tomasz TRACZ

Fig. 1 Relative change in average HPC permeability with and without polypropylene fibres added depending on heating temperature. An increase in heating temperature causes residual permeability to rise. For concrete without fibres (B100) and concrete with a small addition of short fibres (B100/0.9/6), this increase is relatively small. At 160oC, a significant increase in residual permeability starts to takes place, and at the same time considerable differences with respect to its value start to emerge. An addition of 12 mm and 19 mm long fibres had a very large impact on the increase of concrete permeability after heating. An addition of 0.9 kg/m3 of 12 mm long fibres (B100/0.9/12) caused a 12-fold increase in permeability after heating to 200oC compared to the initial value. When the amount of these fibres was doubled to 1.8 kg/m3, a 45-fold increase in permeability was recorded. When analysing measurement results, the immense impact of 12 mm fibre dosage was observed. For 19 mm long fibres, even higher permeability was recorded, but the impact of dosage was not as significant as for 12 mm long ones. Surface Water Absorbability Changes in surface water absorbability in g/cm2 over time were measured for cylindrical samples heated to 140, 160, 180 and 200oC following permeability measurements. The increase in sample mass was recorded at intervals of 60 seconds and with an accuracy of ±0.01 g. During measurements, one sample surface was in permanent contact with water. Irrespective of the amount of water absorbed by the samples tested, the water level remained constant during the measurement, i.e. for 72 hours. The measurement set-up is shown in Fig. 2. Selected measurement results are presented in two sections. In section one, test results depend on the quantity of fibres used, and in section two, test results depend on their length. In Figure 3, changes in surface water absorbability in g/cm2 following heating for concrete samples without fibre additions and for concrete samples with 0.9 and 1.8 kg/m3 of 12 mm long fibres added are shown.

07-The impact.qxd

9/19/2009

11:58 AM

Page 67

The impact of amount and length of fibrillated polypropylene fibres on the properties of HPC exposed to high temperature

67

Fig. 2 Set-up for measuring water absorbability. The curves on Figure 3 demonstrate significant variation in the magnitude of the increase in surface water absorbability of the concretes tested as a result of heating. Concrete without fibres (B100) exhibits an increase in the value measured during the initial heating phase for a heating temperature of 200oC. A similar trend can be observed for concretes that contain 0.9 kg/m3 of 12 mm long fibres. For a larger amount of fibres (1.8 kg/m3), surface water absorbability is already starting to increase in temperature exceeding 140oC. This is reflected in the results of the residual permeability measurements presented above. Heating the B100/1.8/12 concrete to 160oC caused residual permeability to increase more than 25fold, while for the B100/0.9/12 concrete, only a six-fold increase in the value measured was observed.

Fig. 3 Changes in the surface water absorbability of HPC without fibres (B100) and with fibres (B100/0.9/12 and B100/1.8/12) caused by heating to 140, 160, 180 and 200oC

Figure 4 presents the results of measurements concerning the impact of the length of the PP fibres on the magnitude of surface water absorbability changes as a result of heating.

07-The impact.qxd

68

9/19/2009

11:58 AM

Page 68

Izabela HAGER and Tomasz TRACZ

Fig. 4 Changes in the surface water absorbability of HPC with 1.8 kg/m3 of 6, 12 and 19 mm long fibres caused by heating to 140, 160, 180 and 200oC. As the tests carried out have shown, it is not only the quantity of PP fibres added that has an impact on the surface water absorbability of heated concretes, but also their length. In general, as the length of the fibres used increases, changes in concrete surface water absorbability become more pronounced. It is probable that 19 mm long fibres added to concretes heated to 160oC or higher create a continuous network of interconnected capillary pores, which affects the magnitude of surface water absorbability increase during the first few hours of soaking. Porosity In order to explain the changes observed in properties affecting the ability of heated concrete to transport gas and water, tests were carried out to determine the microporosity distribution in mortars extracted from the HPCs tested. Concretes with additions of 0.9 kg/m3 of 6 mm long fibres and 1.8 kg/m3 of 19 mm long fibres were selected for testing. Measurements concerned the microporosity of concretes before heating and after heating to 180oC. Measurement results are presented in Figure 5.

Fig. 5 Distributions of porosity in mortar samples determined using the mercury porosimetry method

07-The impact.qxd

9/19/2009

11:58 AM

Page 69

The impact of amount and length of fibrillated polypropylene fibres on the properties of HPC exposed to high temperature

69

Before heating, very similar porosity distributions were observed for HPC mortars with different additions of PP fibres. After heating to 180C, the overall porosity of B100/0.9/6 concrete mortar rose from 0.058 to 0.068 cm3/cm3, while for the B100/1.8/19 concrete, the increase was from 0.061 to 0.101 cm3/cm3. The change observed does not result from the additional porosity caused by the melting of polypropylene fibres. It is most probably the result of the increase in the micro-cracking of heated concretes with fibres described by Kalifa et al. [4]. Residual Compressive Strength Compressive strength and its change as a function of temperature were determined using 100 mm cubic samples. Two samples of each concrete type were heated to 120, 160, 200, 250, 400 and 600°C, and their compressive strengths were tested after cooling. Residual strength and compressive strength results for unheated samples are presented in Figure 6.a. In Figure 6.b, relative changes in residual compressive strength are shown.

a)

b)

Fig. 6 Impact of heating temperature on absolute and relative changes in residual compressive strength On the basis of the Figures shown, it may be concluded that concrete strength initially declines as a result of heating to 120oC, and then, at 250oC, the strength is partially restored. Further heating leads to a steady decrease in strength caused, inter alia, by the dehydration of the CSH gel, portlandite decomposition and the destruction of the contact zone due to the different thermal expansion coefficients of cement paste and aggregate. On the basis of the compressive strength test results shown above, it may be concluded that this property does not vary considerably for the concretes tested. The results obtained for unheated and heated concrete samples without fibres do not differ significantly from the results obtained for concretes with 0.9 kg/m3 and 1.8 kg/m3 of PP fibres added.

07-The impact.qxd

9/19/2009

11:58 AM

Page 70

70

Izabela HAGER and Tomasz TRACZ

CONCLUSIONS The following conclusions may be stated on the basis of the measurements presented above: x Both the amount and length of polypropylene fibres added has an impact on the magnitude of changes in residual permeability caused by HPC heating. x The greatest increase in the residual permeability of heated concretes can be observed in samples in which 1.8 kg/m3 of 12 or 19 mm long fibres have been added. x Permeability measurements using the RILEM-Cembureau method are very useful when evaluating the effectiveness of adding polypropylene fibres which, after melting, enables water vapour to be evacuated from heated concrete and its pressure to be reduced. x The addition of polypropylene fibres to HPCs heated to temperature ranging from 160 to 200oC affects the capillary porosity of the cement matrix, which is confirmed by measurements of changes in surface water absorbability over time. x Microporosity measurements indicate a significant increase in the quantity of pores with diameters of less than 1 Pm in concrete with fibres (B100/1.8/19) heated to 180oC. The change observed does not stem from the additional porosity caused by melting polypropylene fibres, but is rather the result of an increase in the micro-cracking of heated concrete with PP fibres. x Compressive strength determined as a function of temperature indicates that, for the concretes tested, the presence of fibres and their length do not affect this property significantly. x The research results presented confirm the need for further tests aimed at confirming the beneficial effect of PP fibres on reducing spalling in real HPC elements during fires. REFERENCES 1.

Bazant Z. P.: Analysis of pore pressure, thermal stresses and fracture in rapidly heated concrete, Int. Workshop on fire performance of high strength concrete, February 1997, NIST Maryland, pp. 13-14, 2. Hertz K.: Limits of Spalling of Fire Exposed Concrete, Fire Safety Journal, Vol. 38, 2003, pp 103-116, 3. Kalifa P., Menneteau F.D., Quenard D.: Spalling and pore pressure in HPC at high temperature, Cement and Concrete Research, August 2000, pp 1915-1927, 4. Kalifa P., Chéné G., Gallé C.: High-temperature behaviour of HPC with polypropylene fibers: From spalling to microstructure, Cement and Concrete Research, Vol. 31, Issue 10, October 2001, pp. 1487-1499, 5. Nishida A., Yamazaki N., Inoue H., Schneider U., Diederichs U.: Study on the properties of high strength concrete with short polypropylene fibre for spalling resistance, Concrete Under Severe Conditions Environment and Loading, vol. 2. 1995, 6. Hoff A., Bilodeau A. and Malhotra V.M.: Elevated Temperature, Effects on HSC, Residual Strength, Concrete International, 2000, pp. 41-47 7. Pasquini, N. Polypropylene handbook, 2nd Edition, Hanser Publishers, Munich, 2005. 8. Hager I., Tracz T.: Influence of high temperature on selected properties of high performance concrete modified with the addition of polypropylene fibres, Cement-LimeConcrete, January / February 2009, nr 1, p. 3 9. Bentz, D.P.: Fibers, percolation, and spalling of high performance concrete, ACI Journal 97, 3, 2000, pp. 351-359 10. RILEM Technical Recommendation: Tests for gas permeability of concrete, TC 116PCD: Permeability of concrete as criterion of its durability, Materials and Structures, vol. 32, April 1999, 174-179

08-Assessment by law.qxd

9/19/2009

11:58 AM

Page 71

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

ASSESSMENT BY LAW OF MIXTURES APPROACH OF INTERFACIAL ADHESION STRENGTH IN CELLULOSE-CEMENT COMPOSITES Conrado S. RODRIGUES1, Piet STROEVEN2, Khosrow GHAVAMI3 Associate Professor, Department of Civil Engineering, Federal Centre for Technological Education of Minas Gerais, CEFET-MG, Brazil, [email protected] 2 Full Professor, Department of Civil Engineering and Geosciences, TU Delft The Netherlands, [email protected] 3 Full Professor, Department of Civil Engineering, Pontifical Catholic University, PUC-Rio Brasil, [email protected] 1

ABSTRACT Cellulosic-pulp-reinforced cement sheets are produced in large scale for more than 25 years. This followed studies that had focused on the fibre and matrix characteristics, and on their composite behaviour, solving to a limited extent the problems arising from fibre-matrix incompatibility and deterioration due to alkali attack. In spite of that, much has yet to be done in terms of providing a material with properties competitive with those of asbestos cement, mainly considering their long term performances. Current research on cellulose-pulp reinforced cement composites concentrates on evaluating the deterioration mechanisms, the treatment methods and the modelling of the effects of these mechanisms on the physical, mechanical and micro-structural aspects of the material. Interfacial adhesion governs the mechanical behaviour and many physical properties of composites. Therefore, it is a key issue when focusing on durability, since the deterioration mechanisms and the treatment methods exert direct influences on the fibre-matrix interface. Direct assessment of adhesion strength by pull-out tests on fibres generally shorter than 2 mm in cellulose-cement composites presents practical difficulties. Also, direct application of the equations derived from the composite’s micromechanics to this type of cement composites, as produced by the vacuum-dewatering process, and containing high volume fractions of curly fibres would be problematic. Fortunately, a significant amount of data is available in the literature regarding the physical and mechanical behaviour of cellulose-cement composites of a variety of compositions and produced under similar conditions. This allows establishing mutual correlations. This is used as reference framework for the design of the cellulose-cement compositions tested by the authors. The outcomes reveal this approach a valuable tool in comparative studies on the mechanical behaviour of cellulose-cement composites dealing with effects of treatment methods, deterioration mechanisms and processing parameters.

Keywords Cellulose-cement composites, interfacial adhesion, mechanical behaviour, semi empirical formulation

INTRODUCTION With the need to replace asbestos fibres in cement composites, cellulosic reinforcement has always being considered a viable alternative. Since the 70´s there has been a significant amount of data in the literature regarding the raw materials, production processes and properties of cellulose-pulp-reinforcing cement composites. Mostly, the many variables investigated are considered on the basis of their effects on the mechanical behaviour of the composites; more specifically, on the bending tests results. In spite of the amount of experimental data collected in three decades, there is not an analytical approach describing the

08-Assessment by law.qxd

72

9/19/2009

11:58 AM

Page 72

Conrado S. RODRIGUES, Piet STROEVEN and Khosrow GHAVAMI

behaviour of these composites, as for other materials produced commercially, such as de semi empirical formulations for asbestos-cement and for steel fibre reinforced concrete [1, 2]. The law of mixtures is the most usual approach for a preliminary evaluation of the mechanical behaviour of composites. For many types of composites, the law of mixtures and its variations result in a good approximation for the mechanical parameters [3], being a key tool for the design, materials selection and processing control in the composites’ industry. In the case of high performance fibre-cement sheets, with a high reinforcement content and produced by the vacuum-dewatering process, the application of the law of mixtures presents practical difficulties resulting from the characteristics of these composites, including: a) the high reinforcement contents used prevent uniform fibre distribution in the matrix, favouring the formation of fibre bundles; b) the cellulosic pulp fibres are not uniform in length and the length distribution is depending on the source of the pulp; c) due to their low modulus, the fibres can curve during composite production, reducing their efficiency as reinforcement; and d) the small length of the fibres seriously complicate carrying out pull-out tests, impairing the direct assessment of the interfacial adhesion between the fibres and the matrix that would allow experimental basis for assessment of the efficiency factors [4]. Nevertheless, the law of mixtures concept has yielded good approximation with experimental data for asbestos reinforced cement composites, despite facing similar difficulties [2]. The use of cellulosic pulp as reinforcement results in fibre-matrix interaction that is quite different from that observed for the asbestos-cement, requiring different parameters to be adopted for the law of mixtures equations. Thus, based on data from the literature and the results of experiments carried out by the authors, empirical equations and the formulation from the law of mixtures were investigated, considering cellulosic pulpreinforced cement composites.

METHODOLOGY Among the vast literature related to the mechanical performance of fibre cements with cellulosic pulp reinforcements, articles were selected based on common characteristics that enabled comparative studies of the properties reported. In this sense, data were considered related to composites produced by the vacuum-dewatering process, whose bending tests followed the same methodology (such as the moisture content of the specimens). The papers had to present results on the modulus of elasticity (MOE), the bending strength (Vu), the apparent porosity (p) and also data regarding the reinforcement characterization, at least average fibres’ length. These requisites greatly limited the number of studies considered. The composites produced by the authors are based on kraft bamboo pulp reinforcement. Also, a modified matrix is considered by blending the cement with different amounts of rice husk ash. Characterization of the raw materials, the composite production and the experiments carried out have been reported in a previous study [5].

RESULTS Empirical relationships based on relevant data in the literature It was intended to compare empirical relationships between fibre content, on the one side, and stiffness, strength and porosity, on the other side, with estimates from the law of mixtures approach. In the first, fibre content is described in terms of binder mass, whereas in the law of mixtures the fibre and matrix fractions are described in terms of volume. Therefore, a correlation had to be established among the reinforcement percentages in mass and volume.

08-Assessment by law.qxd

9/19/2009

11:58 AM

Page 73

Assessment by law of mixtures approach of interfacial adhesion strength in cellulose-cement composites

73

An approximate expression was developed assuming that: a) the cellulosic pulp density was 1500 kg/m3, a value reported for a kraft Pinus radiatta pulp [2]; b) square 120mm fibrecement sheets produced by the vacuum-dewatering process employed a fixed 120g of blended cement; and c) the sheets with reinforcement rates ranging from 2 to 14% had an average thickness of 6.5mm. Therefore, the sheet volume was around 93,600 mm3. From these considerations, the relationship between the fibre volume fraction, Vf (in relation to the volume of the fibre-cement sheets), and the fraction in mass, Mf (related to the blended cement weight) can be expressed by Vf

120 M f 1.5 x93.6

0.86 M f

(1)

Data in the literature reporting the physical and mechanical properties of cellulosicpulp-reinforced fibre-cements were used to verify the dependence of these properties on reinforcement content. For this purpose, composites reinforced with bamboo pulp submitted to conventional moist curing [6, 7] and accelerated autoclave curing [8] were considered, in addition to studies with banana [9] and Pinus radiatta pulps [10], both in fibre-cements submitted to conventional curing. Figs. 1 and 2 show the dependence of MOE and porosity on fibre content. It was observed that MOE decreased with fibre content due to the additional porosity incorporated into the composites with the cellulosic reinforcement [1, 2]. Furthermore, the correlations were similar for the different composites considered, even considering that these composites were produced by using different infrastructures. Fig. 2 shows that the matrix of the composites submitted to autoclave curing was more porous than that of composites submitted to normal curing, but the increasing rate of porosity with Vf was similar to the other composites considered.

Figure 1: Empirical relations of MOE with Vf

Figure 2: Empirical relations of porosity with Vf

According to the linear regression of the experimental data in Figs.1 and 2 dealing with composites with normal curing, the unreinforced matrix yields values for MOE and p of about 17 GPa and 27%, respectively. The dependence of MOE and porosity on Vf in these materials can be approximately expressed by p Vf 0.27 MOE

110V f 17

(2) (3)

08-Assessment by law.qxd

9/19/2009

74

11:58 AM

Page 74

Conrado S. RODRIGUES, Piet STROEVEN and Khosrow GHAVAMI

This allows correlating MOE and porosity by MOE

110 p 0.27 17

110 p 46.7

(4)

Fig. 3 shows the dependence of bending strength, Vu on Vf for different composites not to follow an approximate common equation as found for MOE. While MOE was a function principally of porosity of the matrix and interface, Vu depends also on the adherence conditions, and, consequently, on particular characteristics of matrix and reinforcement. We are confronted here with differences in structure-sensitivity between MOE and Vu [11].

Figure 3: Empirical relationships of Vu versus Vf Mechanical parameters of fibre-cement due to law of mixtures approach MOE (Young’s Modulus) The basic formulation of the law of mixtures for short fibres defines efficiency factors related to length (K1) and orientation distribution (K2) of the fibres as expressed by [1] Ec

Em 1 Vf Ș1Ș2 Ef Vf

(5)

in which Ec, Em and Ef are the Young´s Modulus of composites, matrix and fibres, respectively; and Vf is the volume fraction of fibres. In the case of cellulosic-pulp-reinforced fibre-cements there are practical difficulties in defining K1 and K2. This is due to: the small fibre length, the lack of dimensional uniformity and their flexibility and curling during production. Furthermore, Em was considered constant for different fibre volumetric fractions, despite Fig. 2 and eq. (4) showing that with the increase in the reinforcement contents there was an increase in the incorporated porosity (mainly at the interface), resulting in reduced stiffness of the matrix and interface. In agreement with Aveston et al. [2, 12] it is assumed that the pull-out resistance of ductile fibres in cementitious matrices is not depending on the angle of inclination. Thus, the efficiency factor is solely governed by the number of fibres that cross the fractured surface. For the case of random fibres in the plane this leads to K2 = 2/S [1, 4, 12]. To consider the correlation between the composite’s porosity and the reinforcement rate, the empirical relationship between p and Vf of eq. (2) was combined with the law of mixtures expression in eq. (5). Thus, the matrix volume fraction, Vm, is given by (1-Vf - p(Vf)), instead of by (1-Vf). Also, Em = 17 GPa was derived from the empirical relationship, eq. (3), while Ef = 40 GPa was taken from the literature [13, 14]. The MOE of cellulosic pulp-reinforced fibrecement is then described by

08-Assessment by law.qxd

9/19/2009

11:58 AM

Page 75

Assessment by law of mixtures approach of interfacial adhesion strength in cellulose-cement composites

MOE

Em (1 Vf p (Vf )) K Ef Vf ? MOE

MOE 12.41 8.4Vf

Em (1 Vf (Vf 0.27)) K Ef Vf ?

75

(6)

An increase in Vf from 2 to 14% would result according to eq. (6) in a decline in MOE from 12.2 GPa to 11.2 GPa. This is not reflected by the experimental data used in Fig. 1 and in eq. (3), however.

Vu (Bending strength) Application of the law of mixtures concept for estimating tensile and bending strength of asbestos cement composites yielded good correlation with experimental results, [2]. The contribution of the matrix was neglected when considering tensile performance; fibre strength (Vf) was determined assuming a bonding strength of W 2.4 MPa in accordance with a regression analysis of experimental data [2, 15]. The matrix is taken into consideration for estimating Vu; good correlation with experimental data is observed for asbestos-cement, assuming an aspect ratio of fibres lf / d f = 160 and bonding strength of 0.83 MPa, [2]. The expression for the law of mixtures estimate the bending strength for cellulosic-pulpreinforced fibre-cements V u V m (1 V f ) KV f V f

(7)

requires the definition of the fibre and matrix strength, Vf and Vm, respectively, and the efficiency factor, K. The linear regression of experimental results for different composites reported in the literature (Fig. 3) revealed the plain matrix bending strength (Vm) to range from 9.6 to 12.5 MPa; an approximate value of 10 MPa was adopted. Further, an efficiency factor K= 2/S was used [1, 4, 12]. In this case, the 2D distribution of the fibres was the tendency observed during the production of fibrecements following the vacuum-dewatering method. Regarding the fibre strength, Vf = 500 MPa was obtained by estimation from composite data on Pinus radiatta pulp available in the literature [2, 13]. As a consequence of these considerations, Fig. 4 shows good correlation between the Vu dependence on Vf as given by eq. (7) and the experimental data for the composites with this particular type of reinforcement (Pinus radiatta, in Fig. 3).

Figure 4: Comparison between bending strength by law of mixtures and experimental data However, for the other composites the same approach did not yield good correlation. This is mainly due to the high Vf adopted for Pinus radiatta kraft pulp, which may not occur in the other fibres. The correlation between Vu and Vf became more compatible to the other

08-Assessment by law.qxd

76

9/19/2009

11:58 AM

Page 76

Conrado S. RODRIGUES, Piet STROEVEN and Khosrow GHAVAMI

composites when Vf = 200 MPa was adopted. In this case, the distribution of Vu with Vf was close to that observed in composites with cellulosic pulp submitted to normal curing [6]. For cement composites reinforced by cellulosic pulps, in which the fibres contribution is governed by pull-out, the equilibrium for a single fibre is expressed by [2] Vf

§ lf 2W ¨ ¨d © f

· ¸ ¸ ¹

(8)

In the present case, eq. (8) yields an adhesion strength of W = 1.33 MPa when Vf = 200 MPa. When efficiency is reduced by a factor 2/S, this leads to a value of KW = 0.85 MPa. The values of fibre length (lf =1.5mm) and diameter (df = 20Pm) measured by authors were employed in this case. Hence, aspect ratio lf / df = 75. The experimental dependence of bending strength on fibre length was considered to verify the goodness of fit of this average value of adherence tension for other cellulosic pulp composites. For this purpose, data from the literature were used where the mechanical properties of composites submitted to conventional curing were reported together with the characteristics of the reinforcement. The correlation of Vu with lf shown in Fig. 5 considered composites with reinforcements of bamboo [6], Pinus radiatta [13], banana [16], sisal [16, 17], pineapple [18], eucalyptus [16] and recycled paper [19], and referred to an 8% reinforcement rate (on weight basis). Using the general equation of the law of mixtures (eq. (7)) in which Vf was given by eq. (8) and KW = 0.85 MPa, and assuming that the fibres of the different reinforcements had an average 20m diameter, the following expression was derived Vu

0.93V m 5.84l f

(9)

Fig. 4 shows that the approximation represented by eq. (9), adopting Vm=10 MPa, had a correlated properly with the experimental data points, and was close to their linear regression line. Thus, the adhesion strength calculated from a fibre tensile strength estimated at 200 MPa, was shown to be a good approximation for fibre cements reinforced by cellulosic pulp with 1 to 3.5mm long fibres, even considering that each experimental point in Fig. 5 represents a different composite regarding the type of reinforcement, matrix and composites’ production infrastructure.

Figure 5: Vu versus lf - experimental results and approximation by law of mixtures concept Adhesion strength of cellulosic-pulp-reinforced cement due to law of mixtures approach Table 1 shows the mechanical properties obtained by bending tests carried out on plain cement pastes and fibre-cements reinforced by bamboo pulp with a constant 8% fibre content

08-Assessment by law.qxd

9/19/2009

11:58 AM

Page 77

Assessment by law of mixtures approach of interfacial adhesion strength in cellulose-cement composites

77

(in relation to dry cement weight) [5]. In these composites, the matrix was modified by partial substitution of Portland cement by rice husk ash with high (RHA-I) and low (RHA-II) carbon content. The nomenclature of the different materials considered in Table 1 is as follows: CP15-II refers to the composite (CP) with 15% cement replacement with RHA-II. Further, the results concern materials tested under two moisture regimes: in equilibrium (stored at 23°C, 50% UR before the tests) and saturated, indicated by indices “eq” and “sat”, respectively. Table 1. Mechanical parameters of cellulosic pulp reinforced fibre cements Composites M0 M15-I M30-I CP0 CP15-I CP30-I CP15-II CP30-II

Vu

Vi(eq)

Vi(sat)

KW

KW1

MPa 12.93 14.72 13.32 15.53 14.61 11.86 15.31 16.51

MPa

MPa

MPa

MPa

12.85 11.90 9.51 13.95 15.78

7.86 9.26 9.51 6.75 6.66

0.36 0.35 0.30 0.24 0.19

0.54 0.33 0.07 0.75 0.94

Figs. 3 and 4 show that adoption of an adhesion strength of KW = 0.85 MPa for fibrecements reinforced by 8% bamboo pulp submitted to conventional curing enabled the description of an approximate correlation between the bending strength and lf. However, this approach did not ascertain how modifications in the matrix and interface (whether by different compositions, or by deterioration mechanisms) affect the interfacial adherence strength. The law of mixtures equations was again used to verify the influence of incorporating rice husk ash on the interfacial adherence of bamboo pulp reinforced fibre cement. In this case, the adhesion strengthKW was determined from experimental data of Vu pertaining to different fibre-cements. Thus, considering lf / d f = 75 and a reinforcement content on weight basis of 8% (Vf = 0.86 x 0.08), the adhesion strength can be obtained by Vu

V m 1 0.08 x0.86 2KW 75 x0.08 x0.86 ? KW

V u 0.93V m 10.32

(10)

Analysis of the composites of which the mechanical properties of corresponding plain pastes were determined (CP0, CP15-I and CP30-I), demonstrated that only CP0 revealed bending strength significantly higher than that of the corresponding plain matrix (M0) resulting in a KW adhesion strength of the order of 0.34 MPa (much lower than the 0.85 MPa adopted as approximation in eq. (9) and in Fig. 5). The non-significant differences between the Vu values of composites with RHA-I and of the corresponding plain pastes, M15-1 and M30-I, reflect very low adhesion strength to be obtained by eq. (10), in conflict with the real interactions at the interface of these composites. Since it was observed that the Vu was significantly higher than the limit of proportionality (Vi) in these composites [20], it can be concluded that there was energy absorption by fibres and interface after the matrix cracking. Thus, interfacial adhesion in the composites of different constitutions can be assessed more realistically when identifying Vi as the as the matrix resistance of the composites (Vm). The interfacial adhesion resulted from the difference of the bending stresses Vu - Vi, obtained from specimens in equilibrium condition, is related to the frictional (mechanical) adherence KW2, acting during the pull-out of the fibres after the matrix cracking. Thus, KW2 did not consider the fibre-matrix adhesion during the elastic stage of the loading, where the chemical interaction KW1 prevailed, [21]. High KW1 levels indicate composites with enhanced

08-Assessment by law.qxd

9/19/2009

78

11:58 AM

Page 78

Conrado S. RODRIGUES, Piet STROEVEN and Khosrow GHAVAMI

chemical interaction between fibres and matrix, leading to a higher resistance to matrix cracking. Therefore, the bending strength of the composites was related to the set of interfacial adhesion strength values, KW and KW. Considering the Vi of composites tested under the aforementioned equilibrium and saturated conditions, it could be detected that saturation decreases the chemical adhesion between fibres and matrix, resulting in Vi significantly lower in the saturated state (Table 1). When it is assumed that the resistance to matrix cracking is much less sensitive to saturation than the interactions at the interface, this would point towards chemical adherence among the fibres and the matrix being responsible for the relatively high Vi values manifested by the composites under equilibrium condition. Thus, the difference between the Vi values obtained in the equilibrium and saturated states, respectively, can be used to calculate the chemical adherence (pre-cracking). So, KW 2 KW 1

V u ( eq ) 0.93V i ( eq )

(11)

10.32

V i ( eq ) 0.93V i ( sat ) 10.32

Table 1 and Fig. 6 show the adhesion strength calculated for the fibre-cements studied. These parameters contribute to mechanical characterization, and enable the assessment of the energy absorption mechanisms acting in the elastic phase of the loading and after the matrix cracking. These results refer to fibre-cements subjected to standard moisture curing.

Figure 6: Interfacial adhesion strength in the bamboo cellulosic pulp-reinforced fibre-cements

In the composites submitted to conventional curing it was observed that KW2 presented values ranging from 0.19 to 0.36 MPa and KW1 from 0.07 to 0.94 MPa. The analysis of these results corroborated previous conclusions regarding the influence of rice husk ash (RHA) on the mechanical performance of bamboo cellulosic pulp-reinforced fibre-cements [20]. RHA-I did not favour the chemical interaction among fibres and matrix, so, did not permit extra energy absorption through the pull-out mechanism. Contrary, composites with RHA-II presented higher chemical adhesion, with less energy absorption in the form of pull-out in the post cracking phase. CONCLUSIONS Based on experimental data of the mechanical performance of cellulosic pulp-reinforced fibre-cements reported in the literature, a study was carried out addressing the suitability of the law of mixtures concept in describing these properties.

08-Assessment by law.qxd

9/19/2009

11:58 AM

Page 79

Assessment by law of mixtures approach of interfacial adhesion strength in cellulose-cement composites

79

Fibre-cement stiffness was shown to be not very sensitive to variations in the fibre and matrix characteristics. Thus, it was observed that the different fibre-cements obey an approximate common expression relating stiffness to fibre volume fraction. However, the law of mixtures concept cannot satisfactorily represent the empirical stiffness results, as dependence on Vf is a function mainly of the porosity associated with modifications in fibre reinforcement, which is difficult to describe in terms of isolated matrix and fibre characteristics. Bending strength (Vu) is a function of specific reinforcement and matrix characteristics. There is not a common approximate expression that reflects the dependence of Vu on Vf for experimental data of different composites. However, when different composites with similar reinforcement rates were considered, the law of mixtures concept provided a relationship between Vu and average fibre length (lf) that correlated well with the experimental data. Also, this expression of Vu versus lf considers an approximate value for the adhesion strength which varies only as a function of the fibre length, regardless the different fibres, matrix and composites’ production infrastructure considered. Because this expression satisfactorily represents empirical data in the literature, this approach is instrumental in assessing the effects of modification in the fibrecement matrix on the interfacial adhesion strength. Hence, the mechanical parameters of bending tests of bamboo-pulp-reinforced fibrecements allowed evaluation of the interfacial adhesion strength values by law of mixtures concept. This finally made it possible to establish the structural effects of partial substitution of cement by rice husk ash.

ACKNOWLEDGMENTS The authors thank the public research agencies FAPEMIG, CAPES and CNPq, that supported this study financially; and ABMTENC, that encourages and brings together researchers of unconventional materials and technologies. REFERENCES 1. Hannant, D.J. Fibre cements and fibre concretes, John Wiley & Sons, 1978 2. Bentur, A., Mindess, S. Fibre reinforced cementitious composites, 2nd ed. Taylor & Francis, 2007 3. Jones, R.M. Mechanics of Composite Materials, McGraw-Hill Company, 1975 4. Stroeven, P., Stereological principles of spatial modeling applied to steel fiber reinforced concrete in tension. Mat. J. A.C.I., 2009 (accepted for publication) 5. Rodrigues, C. S., Ghavami, K., Stroeven, P. Porosity and water permeability of rice husk ash-blended cement composites reinforced with bamboo pulp, J. Mat. Sci. 41, 2006, 69256937 6. Coutts, R.S.P., Tobias, B.C. Air-cured bamboo pulp reinforced cement, J. Mat. Sci., Letters, 13, 1994, 283-285 7. Anjos, M.A.S., Ghavami, K., Barbosa, N.P. Cement based composites reinforced by bamboo pulp. Part I- determination of the optimum reinforcement rate. Brazilian J. Agricul. Environ. Engn., 7, 2003, 139-145 (in Portuguese) 8. Coutts, R.S.P., Ni, Y. Autoclaved bamboo pulp fibre reinforced cement. Cem. Concr. Comp., 17, 1995, 99-106 9. Zhu, W.H., Tobias, B.C, Coutts, R.S.P., Langfors, G. Air-cured banana-fibre-reinforced cement composites. Cem. Concr. Comp., 16, 1994, 3-8 10. Coutts, R.S.P., Warden, P.G. Air-cured, wood pulp, fibre cement composites. J. Mat. Sci.

08-Assessment by law.qxd

80

9/19/2009

11:58 AM

Page 80

Conrado S. RODRIGUES, Piet STROEVEN and Khosrow GHAVAMI

Letters, 4, 1985, 7-119 11. Stroeven, P., Hu, J., Stochastic heterogeneity as fundamental basis for the design and evaluation of experiments. Cem. Concr. Comp., 30, 2008, 506-514 12. Aveston, J., Mercer, R.A., Sillwood, J.M. Fibre reinforced cements – scientific foundations for specifications. In: “Composites, standards, testing and design”. Proc. National Physical Laboratory Conference, UK, 1974, 499-508 13. Coutts, R.S.P. Wood fibre reinforced cement composites. In: “Concrete technology and design, volume 5: Natural fibre reinforced cement and concrete”, Swamy, R.N. ed., Blackie and Son Ltd, Glasgow, 1988, 1-62 14. Fördös, Z., Tram, B. Natural fibres as reinforcement in cement-based composites. In: Swamy, R.N., Wagstaffe, R.L., Oakley, D.R., eds. Rilem 3rd International Symposium on Developments in Fibre Reinforced Cement and Concrete - Vol. 1, Rilem Symposium FRC 86, 1986, pp.6 15. Akers, S.A.S., Garrett, G.G. Fibre-matrix interface effects in asbestos-cement composites. J. Mat. Sci., 18, 1983, 2200-2208 16. Savastano Jr., H., Warden, P.G., Coutts, R.S.P. In: “Performance of low-cost vegetable fibre-cement composites under weathering”, Proc. CIB World Building Congress. Wellington-New Zealand, April 1st–6th 2001, 1-11, www.irbdirekt.de/daten/iconda/CIB3084.pdf

17. Savastano Jr., H., Warden, P.G., Coutts, R.S.P. Mechanically pulped sisal as reinforcement in cementitious matrices. Cem. Concr. Comp., 25, 2003, 311-319 18. Coutts, R.S.P., Warden, P.G. Air-cured abaca reinforced cement composites. Int. J. Cem. Comp. Lightw. Concr., 9, 1987, 69-73 19. Coutts, R.S.P. Wastepaper fibres in cement products. Int. J. Cem. Comp. Lightw. Concr., 11, 1989, 143-147 20. Rodrigues, C.S., Ghavami, K. The influence of rice husk ash on physical and mechanical properties of high performance cellulose-cement composites. In: Proc. First Colloquium on Non-Conventional Materials Brazil-Colombia, Cali - Colombia, May 23rd - 26th 2005, 1-11, (available on CD-Rom) 21. Coutts, R.S.P., Kightly, P. Bonding in wood fibre-cement composites. J. Mat. Sci. 19, 1984, 3355-3359

09-Tensile fatigue.qxd

9/19/2009

11:58 AM

Page 81

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

TENSILE FATIGUE RESPONSE OF SISAL FIBER REINFORCED CEMENT COMPOSITES Flávio de Andrade SILVA1, Barzin MOBASHER2 and Romildo D. TOLEDO FILHO3 1 Institute of Construction Materials, Technical University of Dresden, 01062 Dresden, Germany, e-mail: [email protected] 2 Department of Civil and Environmental Engineering Arizona State University, Tempe, AZ 85287-8706, USA, e-mail: [email protected] 3 Department of Civil Engineering Alberto Luiz Coimbra Institute – Graduate School and Research in Engineering (COPPE) Federal University of Rio de Janeiro (UFRJ) P.O. Box 68506, 21945 – 970, Rio de Janeiro, RJ, Brazil, e-mail: [email protected]

ABSTRACT Tensile fatigue behavior of composite materials are of significant interest since in structural applications they are often subjected to cyclic loading. However, very few results for fiber reinforced concrete under cyclic load have been reported. In the present research sustainable cement composites were produced by partially replacing 50% of Portland cement with calcined clays and using natural sisal fibers as reinforcement. These composites presented ultimate monotonic tensile stress of 12 MPa and strain at failure of up to 1.5%. Tension-tension fatigue tests were performed with maximum stress levels ranging from 4 to 9.6 MPa at a frequency of 2 Hz. These tests were carried out up to 106 cycles or until the composite failure, whichever occurred first. It was found that up to 6 MPa the composites were able to survive 106 cycles. Composites that survived 106 cycles were re-tested under monotonic load to establish its residual strength. Optical fluorescent microscopy was used to observe the cracking mechanisms after fatigue tests.

Keywords Cementitious composites, sisal fiber, fatigue.

INTRODUCTION Sisal fiber is one of the most widely used natural fibers in yarns, ropes, twines, cords, rugs, carpets, mattresses, mats, and handcrafted articles. During the past two decades sisal fibers have also been used as reinforcement in cement and polymer based composites [1-7]. The sisal reinforcement can be used as short randomly distributed fibers, long oriented fibers, or as a fiber fabric. The sisal fiber is extracted from the sisal plant leaf by a mechanical process called decortication. The sisal leaf consists of a sandwich structure composed of approximately 4% fiber, 1% cuticle, 8% dry matter, and 87% water [8]. Figure 1 shows a picture of the sisal plant, its leaf and the fiber distribution inside the leaf.

09-Tensile fatigue.qxd

82

9/19/2009

11:58 AM

Page 82

Flavio de Andrade SILVA, Barzin MOBASHER and Romildo D. TOLEDO FILHO

10 mm

(a)

Arch fiber

(b)

Structructural 200 m fibers

(c)

Figure 1. The sisal: (a) its plant, (b) cross section view and (c) optical microscopy of region selected in (b). Within the leaf, there are three basic types of fibers: Structural, arch, and xylem fibers. The structural fibers give the sisal leaf its stiffness and are found in the periphery of the leaf. The equivalent diameter of the structural fibers is around 200 m and the cross-section is rarely circular and usually has a “horse-shoe” shape. The structural fibers are of great importance commercially because they almost never split during the process of extraction. The arch fibers grow in association with the conducting tissues of the plant (see Figure 2a) and are usually found in the middle of the leaf (see Figure 1c). These fibers run from base to tip of the plant and have good mechanical strength [9]. The xylem fibers grow opposite to the arch fibers and are connected to them through the conducting tissues (see Figure 2a). According to Nutman [9], they are composed of thin walled cells, and are invariably broken up and lost during the process of fiber extraction.

ML

Lumen Conducting tissue

200 m

10 m

(a) (b) Figure 2. Sisal fiber microstrucutre (a) arch fiber showing its conducting tissue and (b) Details of the fiber-cells showing the lumens and middle lamellae (LM). Every fiber contains numerous elongated individual fibers, or fiber-cells, which are about 6 to 30 m in diameter (see Figure 2b). Each fiber-cell is made up of four main parts, namely the primary wall, the thick secondary wall, the tertiary wall, and the lumen. The fiber-cells are linked together by means of the middle lamellae (ML), which consist of hemicellulose and lignin (see Figure 2b). Long aligned sisal fiber reinforced cement composites have been investigated under static tension and bending tests in previous studies [10-13]. Continuous sisal fiber reinforced cement based composites are a new class of sustainable construction materials with superior tensile strength and ductility. These materials are strong enough to be used as load bearing

09-Tensile fatigue.qxd

9/19/2009

11:58 AM

Page 83

Tensile fatigue response of Sisal Fiber Reinforced Cement Composites

83

structural members, therefore, they can be used in different types of applications such as structural panels, impact & blast resistance, repair and retrofit, earthquake remediation, and strengthening of unreinforced masonry walls. Although high ultimate tensile strength (UTS) and modulus of rupture (MOR) was obtained no information regarding to its fatigue resistance is known. Fatigue can be defined as a permanent, localized, and progressive structural change that occurs in a material subjected to cyclic strains [14]. Those strains are created by loads smaller than the UTS of the material in a static test. In practice, fatigue affects all of the known engineering materials. Many structures are often subject to repetitive cyclic loads. Examples of such cyclic loads include machine vibration, sea waves, wind action and automobile traffic [15]. Fatigue failure occurs when a concrete structure fails at less than design load after being exposed to a large number of stress cycles [16]. The exposure to repeated loading results in a steady decrease in the stiffness of the structure, which may eventually lead to fatigue failure [15]. According to Hsu the interest in the fatigue of concrete started with the development of concrete railroad bridges which were exposed to millions of cycles during their entire life [17]. There is still a lack in understanding of the fatigue behavior of concrete materials and this is even more pronounced for fiber reinforced concrete (FRC). Although natural fibers such as cotton [18], wood pulp [19] and sisal [20] have been tested under fatigue load, cement composites reinforced with those fibers have not yet been investigated. Most of the fatigue tests in concrete or FRC have been performed under bending loads [21-26]. Naaman [23] found that fiber reinforced concrete mixtures containing 2% of hooked steel fibers can sustain bending fatigue stresses more than twice of the plain concrete. This type of FRC (pre-cracked specimens) presented average fatigue lives of the order of 10 cycles for loads ranging between 10 and 90% of their static strength, 8000 cycles for loads ranging between 10 and 80%, and more than 2.7 x l06 cycles for loads ranging between 10 and 70%. Parant et al. [22] tested multi scale steel fiber cement composite (MSCC) under bending fatigue and observed that below a loading ratio of 0.88 (maximum fatigue stress ranging from 35.9 to 40.8 MPa for a MOR of 61.5 MPa), specimens do not fail by fatigue before 2 million cycles. Just a few works in the literature are related to fatigue in uniaxial compression [27-28] and tension loads [29-31]. In this paper we investigate the stress versus cycles fatigue behavior of sisal fiber reinforced composites. The composites were subjected to tensile fatigue load with maximum stresses ranging from 4 to 9.6 MPa at a frequency of 2 Hz. These stress levels represents approximately 30% and 80% of the UTS, respectively. Monotonic tensile tests were performed in previous works [12]. The fatigue tests were stopped either at 106 cycles or complete failure of the composite, whichever occurred first. Composites that survived 106 cycles were tested under monotonic tension to establish its residual strength. Optical fluorescent microscopy was performed to investigate the composites microstructure after fatigue tests.

EXPERIMENTAL PROCEDURE Materials and Processing Continuous sisal fibers were obtained from an agricultural farm located in the city of Valente, state of Bahia – Brazil. Their mechanical properties of bulk fibers defined in terms of Young´s modulus and tensile strength of 18 GPa and 400 MPa, respectively were reported by Silva et al. [32]. The Wollastonite fiber (JG class), obtained from Energyarc, were used as a micro-reinforcement in the composite production.

09-Tensile fatigue.qxd

84

9/19/2009

11:58 AM

Page 84

Flavio de Andrade SILVA, Barzin MOBASHER and Romildo D. TOLEDO FILHO

The matrix was produced using the Portland cement CPII F-32, Metakaolin (MK) from Metacaulim do Brasil Industria e Comércio LTDA, calcined waste crushed clay brick (CWCCB) from an industry located in Itaborai – RJ, Brazil, burned at 850 oC, river sand with maximum diameter of 1.18 mm and density of 2.67g/cm3 and a naphthalene superplasticizer Fosroc Reax Conplast SP 430 with content of solids of 44%. The mortar matrix used in this study presented a mix design 1:1:0.4 (cementitious material:sand:water by weight). The Portland cement was replaced by 30 % of MK and 20 % of CWCCB following recommendations of a previous work [20] to increase the durability of the composite system. The matrix was produced using a bench-mounted mechanical mixer of 20 liters capacity. The cementitious materials were dry mixed during 30 seconds (for homogenization) with the subsequent addition of sand and then a volume fraction of 5% wollastonite. The powder material was mixed for more 30 seconds when the superplasticizer diluted in water were slowly poured in the running mixer and then mixed for 3 minutes. The production of the laminates was achieved by placing the mortar mix in a steel mould one layer at a time, followed by one layer of unidirectional aligned fibers (up to 5 layers) and vibration resulting in a sisal fiber volume fraction of 10% (see Figure 3). The vibrating table was used at a frequency of 65 Hz. After casting the composites were compressed at 3 MPa during 5 minutes. The specimens were covered in their molds for 24 hours and after this time they were demolded and fog cured for 28 days in a cure chamber with 100% RH and 23±1 oC.

Figure 3. Molding procedure for long aligned sisal fiber reinforced composites. Fatigue tests The composites were tested under tensile fatigue loading at a stress ratio (R ratio = Vmin/Vmax) of 0.2 and frequency of 2 Hz. Testing was conducted on a MTS 810 testing system under force control. The experiment was conducted in samples with a 300 mm gage length at five different stress levels: 4, 4.8, 6, 7.2, and 9.6 MPa which presented mean static UTS of 12 MPa [12]. Three samples were tested for each stress level and they presented exactly the same geometry (400 mm x 50 mm x 12 mm – length x width x thickness) than the ones used in the authors previous works [12]. The tests were stopped after 106 cycles or after failure, whichever occurred first. Specimens that survived 106 cycles were tested under monotonic tensile load using the same testing system (MTS 810). The monotonic tensile test was controlled by the cross-head displacement at a rate of 0.1 mm/min. Microstructural investigation The microstructural analysis was performed using a Nikon Elipse TE300 Inverted Video Microscope. Samples that were tested under fatigue up to 106 cycles were pre-loaded to 0.4% strain and then glued with epoxy so the cracks remained open. These samples were then embedded in a polymer mixed with fluorescent dye in order to be analysed using the Nikon

09-Tensile fatigue.qxd

9/19/2009

11:58 AM

Page 85

Tensile fatigue response of Sisal Fiber Reinforced Cement Composites

85

Inverted microscope. This microscope is outfitted with two cameras; a low light level CCD camera (Quantix) and also a standard CCD camera that allows realtime image capture. A fluorescent light (including both Phase and DIC contrast enhancement capabilities) was used. Images were captured and processed by Inovision’s hardware/software interface on a SGI O2 R5000 computer.

DISCUSSION AND ANALYSIS Figure 4 shows the stress versus cycles to failure behavior of the sisal reinforced cement composite tested at various maximum fatigue stresses (4 MPa-9.8 MPa). It can be seen that the composites can survive 106 cycles up to 6 MPa, which represents 50% of its UTS. The sisal fiber composite average UTS and tensile Young´s modulus was determined in the author´s previous work [12]. The average UTS was 12 MPa and the Young´s modulus computed from both cross head displacement and strain gage data was 9.5 GPa and 34 GPa, respectively. The fatigue maximum stress of 6 MPa can be considered a threshold limit where composites may present fatigue failure at cycles close to 106. Beyond 6 MPa all the composites failed below reaching 103 cycles. It was observed that for high fatigue stress levels (i.e.>6 MPa) all the cracks are formed at the first cycles. The number of cracks (12) were the same as the ones observed in monotonic tensile tests. After the crack formation, cracks started to widen. The cycles at these high stress levels caused a degradation process in the fibermatrix interface which increased the rate of cracking opening leading the composite to complete failure at low cycles (i.e. < 103). 1

12

8

4

Maximum Fatigue Stress / UTS

Maximum Fatigue Stress, MPa

16

0 1x101 1x102 1x103 1x104 1x105 1x106 1x107 Cycles to Failure

0.8 0.6 0.4 0.2 0 1x101 1x102 1x103 1x104 1x105 1x106 1x107 Cycles to Failure

(a) (b) Figure 4. Stress versus cycles fatigue curve for composites subjected to maximum stress levels ranging from 4 to 9.8 MPa at constant R ratio of 0.2. Fatigue runout was taken at 106 cycles. The maximum stress was normalized by the ultimate tensile stress (UTS) of the composites in (b). Composites that survived 106 cycles were tested under monotonic tensile load and the results are presented in Figure 5a. When comparing the UTS from monotonic to post-fatigue tensile tests a slight decrease was observed. Nevertheless, this decrease lies in the standard variation range of the monotonic tests and no significant variation among the post-fatigue UTS for

09-Tensile fatigue.qxd

86

9/19/2009

11:58 AM

Page 86

Flavio de Andrade SILVA, Barzin MOBASHER and Romildo D. TOLEDO FILHO

different stress levels was observed. Stiffness degradation was observed when calculating the modulus for the post-fatigue tensile tests. Figure 5b shows that samples subjected to a maximum fatigue stress level of 4 MPa presented higher modulus as the one calculated for the monotonic tensile tests. Above 4 MPa the modulus started to decrease from approximately 11.5 to 2 GPa. It is important to mention that these moduli were calculated from cross-head displacement data. First crack strength was also computed for post-fatigue tensile tests. The composites presented increased first crack strength for higher maximum fatigue stress levels (see Figure 5b). 14

8

8

4 Stress Level = 4 MPa Stress Level = 4.8 MPa Stress Level = 6 MPa Reference

0

0

0.004 0.008 0.012 0.016 Strain, mm/mm

(a)

0.02

7

10 8

6

6 4

5

2 0

First Crack Strength, MPa

12

Initial Modulus, GPa

Monotonic Tensile Stress, MPa

12

Modulus First crack strength

4

0

2

4

6

8

10

Fatigue Maximum Stress, MPa

(b)

Figure 5. Monotonic tensile behavior of composites that have survived 106 cycles: (a) stressstrain curves of composites subjected to maximum fatigue stresses of 4, 4.8 and 6 MPa and (b) effect of the cycles on modulus of elasticity and first crack strength. To understand the evolution of damage, stress-strain hysteresis measurements were conducted at various stress levels. These are shown in Figure 6. These plots were obtained from composites that survived 106 cycles at maximum stress levels of 4, 4.8 and 6 MPa. The Young’s modulus was computed from the linear unloading portion of the cycle for several cycles (see Table 1). For the maximum stress level of 4 MPa the thickness of the individual hysteresis loops, a measure of inelastic damage or energy during a given cycle, was not significant and did not change with cycles. No stiffness degradation was observed for this stress level. The dynamic modulus computed from cross head displacement data at a fatigue stress of 4 MPa was 16.81 GPa whereas the one calculated from the monotonic tests was 9.5 GPa. The dynamic modulus computed from strain gage measurements was only computed for the maximum fatigue stress of 4 MPa. It was found an increase from 34 to 46.61 GPa when comparing to monotonic tests. When increasing the maximum fatigue stress to 4.8 and 6 MPa a different behavior in the hysteresis loops was observed. An increase in the thickness of the hysteresis loops as a function of cycles was observed for both levels. This behavior can be explained due to the formation of several cracks in the first cycles and posterior widening of these cracks. Stiffness degradation as a function of number of cycles was observed for fatigue stresses equal to 4.8 MPa and above. Nevertheless, when computing the dynamic modulus at the first cycle a decrease was only observed for maximum fatigue stresses of 6 MPa and above. It was

09-Tensile fatigue.qxd

9/19/2009

11:58 AM

Page 87

87

Tensile fatigue response of Sisal Fiber Reinforced Cement Composites

observed a change of shape in the hysteresis loops above 103 cycles for maximum fatigue tensile stresses of 4.8 and 6 MPa. The hysteresis loop changes to an “s” shape due to a continuous degradation process in the fiber-matrix interface. 6

6

106

4

Fatigue Stress, MPa

Fatigue Stress, MPa

1

1

2

0

0

105

3.105

106

4

2

0

0.0005 0.001 0.0015 0.002 0.0025 Strain, mm/mm

103

0

0.0005 0.001 0.0015 0.002 0.0025 Strain, mm/mm

(a)

(b)

Fatigue Stress, MPa

8

1

6

102 103

104

5.104 105

0.004

0.006

106

4

2

0

0

0.002

0.008

0.01

Strain, mm/mm

(c) Figure 6. Hysteresis stress-strain behavior of composites subjected to 106 cycles. Maximum stress levels of (a) 4 MPa, (b) 4.8 MPa and (c) 6 MPa. The labels over the hysteresis loops refer to the number of cycles. Samples that survived 106 cycles at a stress level of 6 MPa were investigated using fluorescent optical microscopy. Figure 7 shows the capacity of the fibers to arrest and bridge the cracks formed during fatigue cycles in lateral cross section views. This behavior attests the high efficiency in the fiber matrix bond adhesion of the composite system even when subjected to 106 cycles at a maximum stress of 6 MPa. Two ranges of crack widths were observed in the micrographs: (i) from 1 to 20 Pm and (ii) from 150 to 200 Pm at a

09-Tensile fatigue.qxd

9/19/2009

88

11:58 AM

Page 88

Flavio de Andrade SILVA, Barzin MOBASHER and Romildo D. TOLEDO FILHO

deformation of 0.4%. The use of fluorescent dye possibilities the visualization of micro cracks with width less than 20 Pm that were not observed in conventional optical microscopy. The enhancement in contrast was also achieved with the fluorescent microscopy.

Sand CONCLUSIONS

Crack Fiber

200 Pm

Figure 7. Fluorescent optical micrograph showing the crack bridging mechanisms on a composite subjected to 106 cycles at maximum fatigue level of 6 MPa. Note the higher contrast in the fluorescent optical microscopy that allows the visualization of small cracks (< 20Pm).

Table1. Summary of fatigue and post-fatigue monotonic tensile tests. Results in parenthesis were computed from strain gage measurements Max. Fatigue Stress (MPa)

Post-Fatigue Monotonic Tensile Test First Young´s UTS Crack Modulus (MPa) Strength (GPa) (MPa)

4

10.39

5.22

11.22

4.8 6 7.2 9.8

9.51 10.72 -

6.87 7.59 -

3.47 2.26 -

Dynamic Modulus (GPa)

Cycle 1 16.81 (46.61) 16.30 8.70 3.11 2.62

Cycle 102

Cycle 103

Cycle 104

Cycle 105

-

-

-

-

15.52 5.60 2.08 -

10.2 3.21 -

9.92 2.55 -

6.31 2.12 -

Cycle 106 16.69 (44.37) 5.13 1.96 -

CONCLUSION Long aligned sisal fiber reinforced cement composites were tested under tensile fatigue loading and the main findings are described below: x Composites did not fatigue up to 106 cycles when subject to maximum stress level below 6 MPa. Above this stress the composites presented fatigue below 103 cycles.

09-Tensile fatigue.qxd

9/19/2009

11:58 AM

Page 89

Tensile fatigue response of Sisal Fiber Reinforced Cement Composites

89

x Composites that survived 106 cycles and were tested under monotonic tension did not present significant reduction in UTS but presented decrease in Young´s modulus. The first crack strength increased when increasing the fatigue stress levels. x From the hysteresis stress-strain curves it was noticed no signs of degradation for maximum stress level of 4 MPa. For maximum stress levels of 4.8 and 6 MPa there was noticed an increase in the hysteresis area and decrease in the Young´s modulus as a function of number of fatigue cycles. x The use of fluorescent optical microscopy resulted in high contrast images. The sisal fibers were able to arrest and bridge the cracks even when subjected to 106 cycles at 6 MPa of maximum stress. REFERENCES 1. Silva, F.A., Toledo Filho, R.D. Sisal Fiber Reinforcement of Durable Thin Walled Structures – A New Perspective. In: Proceedings of International Workshop on Cement Based Materials & Civil Infrastructure, Karachi, Pakistan, 2007, 575-586 2. Oksman, K., Wallstrom, L., Toledo Filho, R.D. Morphology and mechanical properties of unidirectional sisal-epoxy composites. Journal of Applied Polymer Science, 84, 2002, 2358-2365 3. Gram HE. Durability of natural fibers in concrete. Swedish Cement and Concrete Research Institute, Research Fo. 1983, p. 225 4. Bisanda, E.T.N., Ansell, M.P. Properties of sisal-CNSL composites. Journal of Materials Science, 27, 1992, 1690-1700 5. Toledo Filho, R.D., Ghavami, K., England, G.L., Scrivener, K. Development of vegetable fiber-mortar composites of improved durability, Cement and Concrete Composites, 25, 2003, 185-196 6. Toledo Filho, R.D., England, G.L., Scrivener, K., Ghavami, K. Durability of alkalisensitive sisal and coconut fibers in cement mortar composites, Cement and concrete composites, 22, 2000, 127-143 7. Towo, A.N., Ansell, M.P. Fatigue of sisal fiber reinforced composites: constant life diagrams and hysteresis loop capture. Composites Science and Technology, 68, 2008, 915-924 8. Murherjee, P.S., Satyanarayana, K.G. Structure properties of some vegetable fibers, part 1. Sisal fibre. Journal of Materials Science,19, 1984, 925-34 9. Nutman, FJ. Agave fibres Pt. I. Morphology, histology, length and fineness; frading problems. Empire Journal of Experimental Agriculture, 5, 1936, 75-95 10. Toledo Filho, R.D., Silva, F.A., Fairbairn, E.M.R., Melo Filho, J.A. Durability of compression molded sisal fiber reinforced mortar laminates. Construction and Building Materials, 68, 2009, 3438-3443 11. Silva, F.A., Melo Filho, J.A., Toledo Filho, R.D., Fairbairn, E.M.R. Mechanical behavior and durability of compression moulded sisal fiber cement mortar laminates (SFCML). In: 1st International RILEM Conference on Textile Reinforced Concrete (ICTRC), Aachen, Proceedings, 2006, 171-180 12. Silva, F.A., Mobasher, B., Toledo Filho, R.D. Cracking mechanisms in durable sisal reinforced cement composites. Cement and Concrete Composites 2009, (in press) 13. Silva, F.A., Melo Filho, J.A., Toledo Filho, R.D., Fairbairn, E.M.R. Effect of Reinforcement Ratio on the mechanical response of compression molded sisal fiber textile reinforced concrete. In: High Performance Fiber Reinforced Cement Composites (HPFRCC5), Mainz, 2007, 175-182

09-Tensile fatigue.qxd

90

9/19/2009

11:58 AM

Page 90

Flavio de Andrade SILVA, Barzin MOBASHER and Romildo D. TOLEDO FILHO

14. Revuelta, D. and Miravete, A. Fatigue damage in composite materials. International Applied Mechanics, 38, 2002, 121-134 15. Lee, M.K. and Barr, B.I.G. An overview of the fatigue behavior of plain and fibre reinforced concrete. Cement and Concrete Composites, 26, 2004, 299-305 16. Li, H., Zhang M-H., Ou J-p. Flexural fatigue performance of concrete containing nanoparticles for pavement. International Journal of Fatigue, 29, 2007, 1292-1301 17. Hsu, T.C. Fatigue of Plain Concrete. ACI Journal, 78, 1981, 292-304 18. Hkarlk, J.W.S. and Sparrow, J.T. Tensile Fatigue of Cotton Fibers. Textile Research Journal, 49, 1979, 242-243 19. Hamad, W.Y. Some microrheological aspects of wood-pulp fibres subjected to fatigue loading. Cellulose, 4, 1997, 51-56 20. Silva, F.A., Chawla, N., Toledo Filho, R.D. An experimental investigation of the fatigue behavior of sisal fibers. Materials Science and Engineering A, 2009, in Press 21. Herández-Olivares, F., Barluenga, G., Parga-Landa, B., Bollati, M., Witoszek, B. Fatigue behavior of recycled tyre ruber-filled concrete and its implications in the design of rigid pavements. Construction and Building Materials, 21, 2007, 1918-1927 22. Parant, E., Rossi, P. and Boulay, C. Fatigue behavior of a multi-scale cement composite. Cement and Concrete Research, 37, 2007, 264-269 23. Naaman, A.E. and Hammoud, H. Fatigue characteristics of high performance fiberreinforced concrete. Cement and Concrete Composites, 20, 1998, 353-363 24. Johnston, C.D., Zemp, R.W. Flexural fatigue performance of steel fiber reinforced concrete-influence of fiber content, aspect ratio and type. ACI Materials Journal, 88, 1991, 374-383 25. Zhang, J., Stang, H. Fatigue performance in flexure of fiber reinforced concrete. ACI Materials Journal, 95, 1998, 58-67 26. Ramakrishnan, V., Lokvik, B.J. Flexural fatigue strength of fiber reinforced concretes. In: Reinhardt H.W., Naaman AE, editors. High Performance Fiber Reinforced Cement Composites: Proceedings of the International RILEM/ACI workshop. London: E&FN SPON:1992, 271-287. 27. Gao, L. ,Hsu, T.C.C. Fatigue of concrete under uniaxial compression cyclic loading. ACI Materials Journal 95,1988, 575-581 28. Rafeez, A.S., Gupta, A., Krishnamoorthy, S. Influence of steel fibers in fatigue resistance of concrete in direct compression. Journal of Materials in Civil Engineering, 12, 2000, 172-179 29. Xianhong, M. and Yupu, S. Residual tensile strength of plain concrete under tensile fatigue loading. Journal of Wuhan University of Technology-Mater. Sci. Ed., 22,564-568. 30. Zhang, J., Stang, H., Li, V.C. Experimental study on crack bridging in FRC under uniaxial fatigue tension. Journal of Materials in Civil Engineering, 12, 2000, 66–73 31. Saito, M. Characteristics of microcracking in concrete under static and repeated tensile loading. Cement and Concrete Research, 17, 1987, 211–8 32. Silva, F.A., Chawla, N., and Toledo Filho, R.D. Tensile behavior of high performance natural (sisal) fibers. Composites Science and Technology, 68, 2008, 3438-3443

10-Investigation on.qxd

9/19/2009

11:58 AM

Page 91

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

INVESTIGATION ON THE STRENGTH AND FLEXURAL TOUGHNESS OF HYBRID FIBRE REINFORCED CONCRETE Surinder Pal SINGH Department of Civil Engineering Dr B R Ambedkar National Institute of Technology Jalandhar – 144 011, India, e-mail: [email protected]

ABSTRACT The paper presents results of an experimental investigation carried out to study the compressive strength, flexural strength and flexural toughness of steel fibre reinforced concrete (SFRC) containing hybrid fibres. An experimental programme was planned in which approximately seventy five beam specimens of size 100 mm x 100 mm x 500 mm were tested under four point static flexural loading. In addition, cube specimens of size 150 mm x 150 mm x 150 mm were also tested to obtain the compressive strength of SFRC. The specimen incorporated three different volume fractions i.e. 0.5%, 1.0% and 1.5% of corrugated steel fibres with each volume fraction containing steel fibres of two different sizes and in different mix proportions by weight. The complete load-deflection curves under static flexural loads were obtained using the displacement control mode of the Actuator system. The flexural toughness indices were calculated using procedure laid down in ASTM C-1018, JCI method and ASTM 1609/C 1609 M. The compressive strength results indicate that with the type of fibres used in this investigation, the best performance is given by concrete containing 100% short fibres at a volume fraction of 1.5% whereas, the best performance in terms of static flexural strength and toughness parameters is obtained with concrete containing 100% long fibres at a fibre volume fraction of 1.5%. However, concrete containing 100% short fibres at a fibre volume fraction of 1.5% gave the best results in terms of first crack load and first crack toughness.

Keywords Fibre reinforced concrete; flexural toughness INTRODUCTION The addition of fibres to concrete considerably improves its structural characteristics such as static flexural strength, impact strength, tensile strength, ductility and flexural toughness. The degree of improvement depends upon many factors such as size, type, aspect ratio and volume fraction of fibres. Many researchers have conducted investigations to study different characteristics of fibre reinforced concrete in the past. The mechanical properties of concrete and mortar reinforced with randomly distributed smooth steel fibres were investigated by Shah and Rangan (1). Hughes and Fattuhi (2) investigated the effect of fibre type (polypropylene, steel), fibre shape (straight, duoform, crimped, hooked), specimen size (cube, prismatic) and age on the compressive stress-

10-Investigation on.qxd

9/19/2009

11:58 AM

Page 92

92

Surinder Pal SINGH

strains curves of steel fibre reinforced concrete. The factors influencing flexural strength measurement of SFRC such as specimen span, width, depth and mode of loading was investigated by Johnston (3). The test results of Nagarkar et al. (4) indicated that the compressive, split tensile and flexural strength increase by addition of fibres, but the workability of concrete was reduced. Nakagawa et al. (5) reported the results of an investigation on the mechanical properties of concrete reinforced with carbon, aramid and high strength vinylon fibres. Toutaji and Bayasi (6) investigated the effect of manufacturing techniques on the mechanical properties of steel fibre reinforced concrete. Ramanalingam et al. (7) presented the results of an experimental investigation on fibre reinforced mortar incorporating different combinations of fibres to exhibit strain-hardening under flexural loading. Recently some investigators i.e. Sivakumar and Santhanam (8), Banthia and Sappakittipakorn (9), Hsie et al. (10) and Ding et al. (11) attempted to study the flexural toughness of fibre reinforced concrete using fibre hybridization. Different combinations of fibres such as metallic-metallic, metallic-nonmetallic and nonmetallic-nonmetallic were employed for different volume fractions of fibres. It has been reported that there has been improvement in the properties and toughness of fibre reinforced concrete through fibre hybridization. However, it is expected that this improvement will depend upon the type of the fibres used. Different types of steel fibres are commercially available in India. This investigation was, therefore, planned to study the flexural toughness of fibre reinforced concrete containing different combinations of metallic-metallic fibres. It is proposed to use the standard test specimen used in India for static flexural tests i.e. of size 100 x 100 x 500 mm and to obtain the toughness parameters using ASTM 1018 C, JCI Method and ASTM 1609/C 1609 M. This investigation forms a part of a large project work being undertaken in this direction in which different combinations and types of metallic and non-metallic fibres are proposed to be used.

EXPERIMENTAL PROGRAM The concrete mix used for casting the test specimens is shown in Table-1. Portland Pozzolana Cement, crushed stone coarse aggregates having maximum size 10 mm and river sand were used. The specimens incorporated two different types of corrugated steel fibres i.e. rectangular (38 mm long, 2 mm wide and 0.6 mm thick) and round (25 mm long and 1 mm in diameter) by weight of the longer and shorter fibres in mix proportions of 100% - 0%, 75% - 25%, 50% - 50%, 25% - 75% and 0% - 100% at each of the fibre volume fractions of 0.5%, 1.0% and 1.5%. The specimens used for compressive strength tests and static flexural strength tests were 150 x 150 x 150 mm cubes and 100 mm x 100 mm x 500 mm beams respectively. Table-2 presents the details of various fibre concrete mixes used in this investigation. Table 1: Concrete Mix Proportion. Water/Cement Ratio

Sand/Cement Ratio

Coarse Aggregate/Cement Ratio

0.46

1.52

1.88

The compressive strength tests were conducted on a 2000 kN Universal Testing Machine.

10-Investigation on.qxd

9/19/2009

11:58 AM

Page 93

93

Investigation on the strength and flexural toughness of hybrid fibre reinforced concrete

The maximum compressive load was recorded as that load at which the specimen failed to take any further increase in load. The compressive strength was calculated by dividing the maximum compressive load by the cross sectional area of the cube specimen on which the load was applied. Table 2: Steel Fibrous Concrete Mixes. Fibre Type

Fibre Mix Proportion by Weight (%) Vf = 0.5%

Vf = 1.0%

38 mm 100 75 50 25 0 100 75 50 25 0 100 Long* 25 mm 0 25 50 75 100 0 25 50 75 100 0 Long** * Rectangular, corrugated, 38 mm long, 2 mm wide and 0.6 mm thick. ** Round, corrugated, 25 mm long and 1 mm in diameter.

Vf = 1.5% 75

50

25

0

25

50

75

100

The static flexural strength tests were conducted on a 100 kN servo-controlled actuator and the specimens were loaded at third points. The specimens were turned on their side with respect to the position as cast before placing these on the support system so as to have a plain and uniform surface in contact with the supporting and loading rollers. Complete load-deflection curve was obtained for each specimen tested. A minimum of four samples were tested for each combination of fibres. The static flexural strength and toughness parameters were evaluated as per the standards mentioned in preceding section.

RESULTS AND DISCUSSION Compressive Strength Test Results The results of the compressive strength tests conducted on fibre reinforced concrete specimens with different fibre volume fractions, each fibre volume fraction containing different combinations of mixed steel fibres, are presented in Table 3 and Fig. 1. It can be seen that for a constant mixed aspect ratio of fibres, there is an increase in compressive strength of concrete as the percentage of fibres is increased. The results show that in general, there is an increase in compressive strength varying from 1% to 32%, on addition of fibres to the concrete mix. The results indicate that there is an increase in the compressive strength with respect to that of plain concrete varying from 1% to 26%, 3% to 29% and 4% to 32% for mixes having 0.5%, 1.0% and 1.5% volume fraction of fibres respectively. However, for a particular mix, there is an optimum volume fraction of fibres that will give the maximum compressive strength. In the case of mixes tested in this investigation, the optimum percentage of fibres was 1.5%, when using 100% short fibres and the maximum increase in compressive strength in this case was 32%. The maximum increase in compressive strength for all the other mixes was observed at a fibre volume fraction of 1.0%. In general, it may be concluded that on increasing the percentage of short fibres in the mix, the compressive strength increases. Chen and Carson (12) also reported that for a given volume fraction of fibres, the mixes with shorter fibres had higher compressive strength compared to those with the longer fibres.

10-Investigation on.qxd

9/19/2009

11:58 AM

Page 94

94

Surinder Pal SINGH

Table 3: Compressive Strength Test Results. Fibre Mix Proportion by Weight

Fibre Volume Fraction (%)

28 Days Compressive Strength (MPa)

% Increase

38 mm Long 25 mm Long Fibres* Fibres** 38.75 0 0 0.0 39.15 100 0 0.5 41.10 75 25 0.5 41.92 50 50 0.5 45.87 25 75 0.5 48.81 0 100 0.5 40.01 100 0 1.0 41.39 75 25 1.0 44.45 50 50 1.0 46.83 25 75 1.0 50.10 0 100 1.0 40.43 100 0 1.5 42.56 75 25 1.5 45.99 50 50 1.5 47.82 25 75 1.5 51.05 0 100 1.5 * Rectangular, corrugated, 38 mm long, 2 mm wide and 0.6 mm thick. ** Round, corrugated, 25 mm long and 1 mm in diameter.

Compressive Strength (MPa)

Fibre Mix Combination 55

100% 38mm Long + 0% 25mm Long Fibres

50

75% 38mm Long + 25% 25mm Long Fibres

45

50% 38mm Long + 50% 25mm Long Fibres 25% 38mm Long + 75% 25mm Long Fibres

40

0% 38mm Long + 100% 25mm Long Fibres 35 0

0.5

1

1.5

2

%age of Fibres

Figure 1: Compressive strength of concrete with mixed fibres at different fibre volume fractions.

0 1 6 8 18 26 3 7 15 21 29 4 10 19 26 32

10-Investigation on.qxd

9/19/2009

11:59 AM

Page 95

95

Investigation on the strength and flexural toughness of hybrid fibre reinforced concrete

Flexural Strength Test Results Flexural strength results for fibrous concrete mixes with mixed aspect ratio of fibres at different fibre volume fractions are shown in Table 4 and Fig. 2. The load-deflection curves obtained in this investigation for concrete with different volume fractions of fibres, each volume fraction containing different combinations of fibres are presented in Fig. 3. Figure 3 (a) presents the load-deflection curves for all the fibre mix combinations for fibre volume fraction of 0.5% whereas, Fig. 3 (b) and 3 (c) present the curves for concrete with fibre volume fraction of 1.0% and 1.5% respectively. It can be seen from Table 4 that the maximum increase in flexural strength with respect to plain concrete, taken as the average of the samples tested for a particular mix combination of fibres, varied from 32% to 36%, 39% to 43% and 50% to 59% for concrete mixes with 0.5%, 1.0% and 1.5% volume fraction of fibres respectively. The maximum increase in static flexural strength of 59% was obtained for concrete with 100% long fibres at a fibre volume fraction of 1.5%. Table 4: Flexural Strength Test Results Fibre Mix Proportion by Fibre Deflection Flexural % Weight Volume at Peak Strength Increase Fraction Load 38 mm 25 mm (MPa) (%) (mm) Long Long Fibres* Fibres** 0 0 0.0 4.65 0 0.502 100 0 0.5 6.33 36 0.677 75 25 0.5 6.17 33 0.628 50 50 0.5 6.26 35 0.602 25 75 0.5 6.12 32 0.619 0 100 0.5 6.19 33 0.652 100 0 1.0 6.57 41 1.106 75 25 1.0 6.58 42 0.678 50 50 1.0 6.53 40 0.693 25 75 1.0 6.48 39 0.736 0 100 1.0 6.65 43 0.702 100 0 1.5 7.38 59 1.235 75 25 1.5 6.97 50 0.953 50 50 1.5 6.99 50 0.830 25 75 1.5 7.06 52 0.701 0 100 1.5 7.15 54 0.820 * Rectangular, corrugated, 38 mm long, 2 mm wide and 0.6 mm thick. ** Round, corrugated, 25 mm long and 1 mm in diameter.

First Crack Load (kN)

First Crack Deflection (mm)

10.35 12.87 12.53 12.80 12.31 13.05 13.59 13.61 13.26 13.39 13.61 14.17 15.19 14.85 15.67 15.87

0.502 0.568 0.528 0.602 0.552 0.577 0.515 0.525 0.522 0.520 0.552 0.528 0.652 0.530 0.567 0.665

10-Investigation on.qxd

9/19/2009

11:59 AM

Page 96

96

Surinder Pal SINGH

Fibre Mix Combination 100% 38mm Long + 0% 25mm Long Fibres

Flexural Strength (MPa)

7.5

75% 38mm Long + 25% 25mm Long Fibres 7

50% 38mm Long + 50% 25mm Long Fibres 25% 38mm Long + 75% 25mm Long Fibres

6.5

0% 38mm Long + 100% 25mm Long Fibres 6 0

0.5

1

1.5

2

%age of fibres

Figure 2: Flexural strength of concrete with mixed fibres at different fibre volume fractions.

The average centre-point deflection corresponding to peak load for concrete with different mixed aspect ratios and different volume fractions of fibres are also listed in Table 4. The increase in average centre point deflection corresponding to peak load with respect to plain concrete was observed to vary between 20% to 35%, 36% to 120% and 40% to 146% for concrete mixes containing 0.5%, 1.0% and 1.5% fibre volume fractions respectively. The maximum increase in peak load deflection of 35%, 120% and 146% with respect to that of plain concrete was observed in fibrous concrete containing 0.5%, 1.0% and 1.5% fibre volume fractions respectively for fibre mix containing 100% long fibres. It can also be observed from Table 4 that there is an increase in first crack load over that of plain concrete, taken as average of the samples tested for a particular mix combination of fibres, of the order of 4% to 10%, 12% to 15% and 20% to 34% for concrete mixes having 0.5%, 1.0% and 1.5% volume fractions of fibres respectively. The maximum increase in first crack load of 10%, 15% and 34% with respect to plain concrete was observed with 100% short fibres at fibre volume fractions of 0.5%, 1.0% and 1.5% respectively. The maximum increase in first crack deflection of 20% to 33% with respect to that of plain concrete was observed for fibrous concrete specimens with fibre volume fractions of 0.5% to 1.5% respectively. Flexural toughness Un-reinforced concrete is a brittle material, with little ability to resist pronounced tensile stress and strain. As reported by Mohammadi (13), discontinuous fibres are added to concrete to improve post-cracking behaviour i.e. to improve its energy absorption capacity and to provide improved resistance to cracking. For fibre reinforced concrete, the term toughness is used to characterize such energy absorption. Hence flexural toughness is an important parameter in assessing the influence of fibres on the post-peak behaviour of fibre reinforced concrete.

10-Investigation on.qxd

9/19/2009

11:59 AM

Page 97

97

Investigation on the strength and flexural toughness of hybrid fibre reinforced concrete

15

Fibre Mix Combination

12

Load (kN)

100% 38mm Long + 0% 25mm Long Fibres 75% 38mm Long + 25% 25mm Long Fibres

9

50% 38mm Long + 50% 25mm Long Fibres 25% 38mm Long + 75% 25mm Long Fibres

6

0% 38mm Long + 100% 25mm Long Fibres 3

(a) 0 0

1

2

3

4

5

6

Deflection (mm) 16

Fibre Mix Combination 100% 38mm Long + 0% 25mm Long Fibres 75% 38mm Long + 25% 25mm Long Fibres

Load (kN)

12

50% 38mm Long + 50% 25mm Long Fibres 25% 38mm Long + 75% 25mm Long Fibres

8

0% 38mm Long + 100% 25mm Long Fibres

4

(b) 0 0

1

2

3

4

5

6

7

Deflection (mm)

Fibre Mix Combination

18

100% 38mm Long + 0% 25mm Long Fibres 75% 38mm Long + 25% 25mm Long Fibres

15

Load (kN)

50% 38mm Long + 50% 25mm Long Fibres 12

25% 38mm Long + 75% 25mm Long Fibres 0% 38mm Long + 100% 25mm Long Fibres

9 6 3

(c)

0 0

1

2

3

4

5

6

7

Deflection (mm)

Figure 3: Load-deflection curves for concrete with mixed fibres, (a) Vf = 0.5%; (b) Vf = 1.0%; (c) Vf = 1.5%.

10-Investigation on.qxd

98

9/19/2009

11:59 AM

Page 98

Surinder Pal SINGH

Over the years, a number of test methods have been developed in an attempt to measure or quantify the toughness and performance of fibre reinforced concrete. In this investigation, flexural toughness was determined from the load-deflections curves using the procedure laid down in different standards. The ASTM C 1018 (14) though withdrawn now a days, was used to quantify the toughness parameters of fibre reinforced concrete. Different indices such as TF.C., I5, I10 and I20, were calculated and are presented in this paper. The JCI Method (15) was used to calculate the toughness TJCI. The flexural performance parameters such as residual loads and strengths and specimen toughness were calculated from the load-deflection curves using ASTM C 1609/C 1609 M (16). The symbols used for various toughness parameters are the same as given in the above mentioned standards. Due to lack of space, only limited results are reported here as far as toughness parameters are concerned. The results of the toughness parameters using different methods for concrete mixes with different mixed aspect ratio of fibres at each of the three volume fractions of steel fibres as obtained from the corresponding load deflection curves are presented in Table 5. The toughness index for plain concrete is taken as 1.0 because plain concrete flexural test specimens fail immediately after the formation of first crack. From the perusal of the test results, it is evident that in general, the TJCI, the indices I5, I10 and I20 and the parameters f600 and f150 increase with increasing fibre content and increasing percentage of long fibres in the concrete mix. The indices I5 and I10 are relatively less sensitive to the use of mixed aspect ratio of fibres. Further, comparison of toughness indices shows that the mixes with long fibres have higher values of the indices than those with short fibres. The first crack toughness (TFC) increases with increasing fibre volume fraction and increasing content of shorter fibres in the mix. The maximum value of TFC were obtained for a mix with 100% short fibres, for all the three volume fractions tested in this investigation. The toughness indices were found to be sensitive to the volume fraction and percentage of the longer fibres in the concrete mix. Higher values of the toughness indices were obtained at higher fibre volume fractions and at higher percentage of the longer fibres in the concrete mix, in contrast to the trends in the first crack toughness values. The mixed fibres are used in concrete with the understanding that relatively short fibres are more effective in arresting micro-cracks whereas, the longer fibres are effective in arresting macro-cracks in concrete. When a specimen is subjected, for example, to flexural loads, a number of micro-cracks appear initially which are preceded by one or two major cracks that lead to the failure of the specimen. As observed in this investigation, the ultimate load increases with the increase in aspect ratio of the fibres in the concrete mix, hence it can be concluded that the longer fibres are offering more resistance to the pull out of fibres because of their better bond characteristics being longer in length. The residual strengths f600 and f150 also present more or less same trends as presented by toughness parameters obtained by other methods. The residual strength f600 increases with increase in the percentage of long fibres in concrete mix except at fibre volume fraction of 1.5% where some conflicting results are obtained. However, the residual strength f150 increases with increase in the number of long fibres in concrete for all the three volume fractions of fibres tested in this investigation.

10-Investigation on.qxd

9/19/2009

11:59 AM

Page 99

99

Investigation on the strength and flexural toughness of hybrid fibre reinforced concrete

Table 5: Flexural Toughness/Performance Parameters for Fibrous Concrete. Fibre Mix Proportion by Weight

Fibre Volume Fraction (%)

Flexural Toughness/Performance Parameters

JCI ASTM C – 1018 Method T(JCI) TF.C. I10 I5 (kN-mm) (kN-mm) 100 0 0.5 19.06 4.05 3.20 4.76 75 25 0.5 17.77 3.95 3.07 4.61 50 50 0.5 15.90 4.02 2.77 4.14 25 75 0.5 14.25 3.75 2.61 3.78 0 100 0.5 11.61 4.08 2.27 2.91 100 0 1.0 30.81 4.32 4.02 6.90 75 25 1.0 27.23 4.29 3.53 5.97 50 50 1.0 25.60 4.24 3.36 5.60 25 75 1.0 21.19 4.19 3.04 4.85 0 100 1.0 20.12 4.23 2.70 4.41 100 0 1.5 36.18 4.68 4.72 8.15 75 25 1.5 35.19 5.00 4.71 7.85 50 50 1.5 34.23 4.72 4.70 7.78 25 75 1.5 33.80 4.64 4.70 7.67 0 100 1.5 30.72 4.86 4.48 6.76 * Rectangular, corrugated, 38 mm long, 2 mm wide and 0.6 mm thick. ** Round, corrugated, 25 mm long and 1 mm in diameter. 38 mm Long Fibres*

25 mm Long Fibres**

I20 6.32 --6.13 4.74 --9.91 8.35 7.55 6.83 5.19 11.07 10.72 10.19 9.94 8.83

ASTM C – 1609/C 1609 M f150 f600 (MPa) (MPa) 4.12 1.57 3.75 1.44 4.01 1.31 3.11 1.24 2.56 0.63 5.97 3.21 5.74 2.96 5.29 2.18 4.86 1.73 4.34 1.28 6.49 4.13 6.79 3.79 6.56 3.73 6.61 3.57 6.63 3.26

CONCLUSION Properties, such as compressive strength, flexural strength and flexural toughness of fibre reinforced concrete containing mixed steel fibres have been investigated. Complete loaddeflection curves were obtained under static flexural loading and flexural toughness indices/parameters were obtained as per procedure laid down in ASTM C 1018, JCI Method and ASTM C 1609/C 1609 M. A maximum increase in compressive strength of the order of 32% over plain concrete was observed in case of concrete containing 100% shorter fibres at a fibre volume fraction of 1.5%. In case of static flexural strength tests, a maximum increase in static flexural strength of the order of 59%, centre point deflection corresponding to peak load of the order of 146% and toughness parameters were obtained for fibrous concrete with 100% long fibres at a fibre volume fraction of 1.5%. The maximum increase in first crack load in flexure equal to 34% and first crack toughness were obtained in a concrete mix with 100% short fibres at a fibre volume fraction of 1.5%. Keeping in view the size and type of fibres used, it can be observed from the results reported in this investigation that no single combination of fibres can be adjudged as the optimum combination for the all the properties tested in this investigation.

10-Investigation on.qxd

9/19/2009

11:59 AM

Page 100

100

Surinder Pal SINGH

REFERENCES 1. 2. 3. 4.

5.

6. 7.

8.

9.

10. 11.

12. 13. 14. 15. 16.

Shah, S. P., Rangan, B.V., Fibre reinforced concrete properties. ACI Journal, Vol. 68, No. 2, February, 1971, pp. 126-135. Hughes, B.P., Fattuhi, N.I., Stress-strain curves for fibre reinforced concrete in compression. Cement and Concrete Research, 7, 2, 1977, 173-183. Johnston, C.D., Steel fibre reinforced and plain concrete: factors influencing flexural strength measurement. ACI Journal, 79, 2, 1982, 131-138. Nagarkar, P.K., Tambe, S.K., Pazare, D.G., Study of fibre reinforced concrete. Proceedings of International Symposium of Fibre Reinforced Concrete, December 16-19, Madras, India, 1987, 2.130-2.138. Nakagawa, P.K., Akihama, S., Suenaga, T., Mechanical properties of various types of fibres reinforced concretes. Fibre Reinforced Cement and Concretes: Recent Developments, Ed. by R N Swamy and B. Barr, University of Wales, College of Cardiff, UK, 1989, 523-532. Toutanji, H., Bayasi, Z., Effects of manufacturing techniques on the flexural behaviour of steel fibre-reinforced concrete. Cement and Concrete Research, 28, 1, 1998, 115-124. Ramanalingan, N., Paramasivam, P., Mansur, M.A., Madlej, M., Flexural behaviour of hybrid fibre reinforced cement composites containing high volume fly ash. 7th Canmet/ACI International Conference on Fly Ash, Silica Fume, Slag and Natural Pozzolans in Concrete, SP-199, Vol. 1, Ed. V.M. Malhotra, 2001, 147-161. Sivakumar, A., Santhanam, M., Mechanical properties of high strength concrete reinforced with metallic and non-metallic fibres”, Cement and Concrete Composites, 29, 2007, 603-608. Banthia, N., Sappakittipakorn, M., Toughness enhancement in steel fibre reinforced concrete through fibre hybridization. Cement and Concrete Research, 37, 2007, 1366-1372. Hsie, M., Tu, C., Song P.S., Mechanical properties of polypropylene hybrid fibre reinforced concrete. Material Science and Engineering A, 494, 2008, 153-157. Ding, Y., Jhang, Y., Thomas, A., The investigation on strength and flexural toughness of fibre cocktail reinforced self compacting high performance concrete. Construction and Building Materials, 23, 2009, 448-452. Chen, W.F., Carson, J.L., Stress-strain properties of random wire reinforced concrete”, ACI Journal, 68, 12, 1971, 933-936. Mohammadi, Y., Behaviour of steel fibre reinforced concrete in flexural fatigue. Ph.D. Thesis, IIT Roorkee, 2002, pp 304. ASTM C 1018, Standard test method for flexural toughness and first crack strength of fibre reinforced concrete. Annual Book of ASTM Standards, 04.02, 1994, 509-516. JCI-SF4, Methods of tests for flexural strength and flexural toughness of fibre reinforced concrete. 45-56. ASTM C 1609/C 1609 M, Standard test method for flexural performance of fibre reinforced concrete. ASTM, 2007, 1-9.

11-Textile reinforced.qxd

9/19/2009

11:59 AM

Page 101

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

TEXTILE REINFORCED CONCRETE – DURABILITY ISSUES: CHANGES OF THE BOND AND TENSILE STRENGTH DUE TO AGEING Till BÜTTNER1, Jeanette ORLOWSKY, Michael RAUPACH Institute of Building Materials Research, RWTH Aachen University, Germany 1 e-mail: [email protected] ABSTRACT Textile reinforced concrete (TRC) is one promising possibility to produce thin-walled high load bearing structural elements. Besides the maximum load bearing capacity, the durability under various conditions is one major issue for future applications. Usually AR-(Alkali-Resistant)-Glass is used as reinforcement. Even if it is AR-Glass, the high alkaline environment in concrete leads to a loss of the tensile strength. The paper deals with two major issues of the durability of TRC. On one hand the behavior of the bond due to ageing and on the other hand the loss of strength due to glass corrosion is discussed. Also the paper presents possible ways of predicting the long-term loss in strength based on the short-term laboratory ageing. Besides this, possibilities of improving the durability by the use of polymer modified concrete and polymer impregnated reinforcement are shown. In general, the results show that the bond of TRC is not affected by ageing and a polymeric modification of the reinforcement and also of the concrete can lead to a significant improvement of the durability – that means the loss of tensile strength due to the alkalinity is reduced.

Keywords Textile reinforced concrete (TXC), AR glass fibres, bond INTRODUCTION Textile reinforced concrete (TRC) represents an interesting new construction material, offering several additional advantages compared to steel or fibre reinforced concrete. These advantages dominate in those fields of applications where thin-walled, structural elements with a high load carrying capacity are required. In textile reinforced concrete, textiles made of alkali resistant (AR) glass fibers are applied in order to carry the forces arising by tensile loads. Previous investigations showed that the mechanical properties of the glass reinforcement have not been fully exploited [1]. The outer filaments, which are in direct contact to the concrete, only carry the tensile loads and the inner filaments are not connected to the concrete. Therefore these filaments cannot be activated for any load transmission. One possibility to improve the load bearing behavior is the use of polymeric impregnations which “glue” the inner and outer filaments together and the load is transferred to all the filaments, [1]. Even the reinforcement is called alkali resistant glass, there is a strength loss due to the alkalinity of the concrete, [2]. So it is obvious that the amount of alkali ions, which are able to attack the reinforcement, have to be reduced. On one hand the amount of alkali ions reaching the reinforcement is reduced by impregnating the reinforcement and on the other hand polymer modified concretes should reduce the water uptake of the concrete matrix and thus

11-Textile reinforced.qxd

9/19/2009

11:59 AM

102

Page 102

Till BÜTTNER, Jeanette ORLOWSKY and Michael RAUPACH

lead also to a reduction of the amount of alkali ions attacking the reinforcement. This chemical attack is eventually not the only factor, which leads to a reduction of the load bearing capacity of TRC – also the bond between the fiber bundles and the matrix material might alter due to ageing. Even if no loss of strength is measured, the stress-strain behavior and crack openings (both important characteristics in serviceability limit state design for constructions, according to the European standards) might be influenced. The paper considers various aspects concerning the durability of TRC: x Influences of the bond evolution and effecting the durability of TRC without any polymers added x Influences of the polymeric modifications of TRC in respect to the loss of strength.

THEORETICAL BACKGROUND The tensile stress-strain behavior of brittle matrix composites - including TRC composites has already been studied in previous publications [3] and is only discussed briefly here. Initially the TRC composite shows linear elastic behavior and the fibers and matrix share the applied load. At a certain stress, a first crack appears in the TRC specimen, leading to a sudden extra strain: at the vicinity of the crack face, the cracked matrix will show elastic unloading and redistribution of stresses leads to a sudden increase of the internal fiber stress (and thus also strain) at the vicinity of the crack. Simultaneously, the interface between matrix and fibers becomes debonded for a certain length along the fiber. At the vicinity of each crack face there is only frictional stress transfer between fibers and matrix in this debonded interface. When the load on the composite is further increased the composite stiffness is slightly lower. Upon further loading, extra cracks appear in the matrix, each time leading to extra deformations and loss of stiffness. The matrix progressively cracks in multiple places, until the crack spacing (distance between neighboring cracks) is such that, at every crack, the gradual stress transfer at the fiber matrix interface is interrupted by the stress relief from the neighboring crack before the strength of the matrix is exceeded. In order to determine a link between the interface stresses and the global measured behavior of the composite, a type of cohesive bond-slip relation should be proposed. In this case, expression of equilibrium of forces at every cross-section of the specimens leads to following relationship between composite stress Vc and composite strain Hc [4, 5] as long as full multiple cracking did not occur yet:

Vc §

DG · ¸ ; [6] ¨1 E c1 © x ¹

Hc here: x

D

Vf* Ec1

G

measured average crack spacing (distance between cracks) at Vc relative contribution of matrix stiffness to fiber stiffness (EmVm/EfVf*) effective fiber volume fraction (amount of fibers contributing to load transfer) linear elastic composite stiffness [N/mm2] debonding length, which can be calculated as follows:

G here:

Am Em T

(1)

Am E mV c ; [6] E c1T

effective matrix cross-sectional area [mm2] matrix stiffness [N/mm2] bond shear flow [N/mm]

(2)

11-Textile reinforced.qxd

9/19/2009

11:59 AM

Page 103

Textile reinforced concrete - durability issues: changes of the bond and tensile strength due to ageing

103

The normalized bond shear flow T is introduced to represent the global transfer of stresses between fibers and matrix, once debonding occurs. It corresponds to the integrated effect of the bond shear stress between the fibers and matrix. Since the unit reinforcing element is a strand rather than a discrete filament it is difficult to estimate its true contact perimeter with the matrix. The total stress transfer from the fiber bundle to the matrix, which might thus include non-homogeneous matrix-fiber and fiber-fiber stress transfer, is represented globally in this paper by a bond shear flow T in each section - perpendicular to the load - of the composite. It should be mentioned that equation (1) is only valid as long as the last crack in the matrix did not appear yet. Once the last matrix crack does appear, equation (3) should be used:

Hc

§

1 Dx ·¸ ; [6] * E V 4 G Ec1 ¸¹ © f f

V c ¨¨

(3)

MATERIALS Rovings and textiles Due to the low price mainly AR-glass reinforcement for TRC is used. The structure of the reinforcement made out of AR-glass rovings, carbon rovings and aramid rovings is similar. They all consist of some hundreds to some thousands of filaments with diameters of a few Pm. Theses rovings (yarns) are processed usually into two- or three-dimensional textiles like the one depicted in figure 1.

Figure 1: (a) AR-glass roving 2400 tex, (b) glass fibre textile MAG 07-03, (c) glass fibre textile MAG 07-03 impregnated with an epoxy The rovings used in this paper consist out of approximately 1560 filaments with an average diameter of 27 Pm, corresponding to a linear density of 2400 tex (1tex = 1 g/km). Table 1 gives an overview of the mainly used reinforcement materials for the current research. As previously mentioned one of the main intentions of concrete modification and roving impregnation with polymers is the improvement of the durability. Therefore different types of polymers, primarily polymer dispersions and epoxies, were used for roving impregnation in previous investigations. These investigations showed that the best results can be achieved by using epoxies for roving impregnation [6], which is the main reason why the investigations with epoxies are carried on. The results obtained with three different epoxies are presented: EP STF STD is a cold hardening epoxy system; EP 3.1.2 is also a cold hardening but waterborne system; EP STD PRE1 is a Prepreg (used for PREimPREGnated textile structures) system which was cured at 120 °C for 2 hours. The epoxy resins are hardened by addition of an amine hardener according to the stoichiometric ratio. The detailed mechanical and chemical properties of these epoxy resins can be found in [1].

11-Textile reinforced.qxd

9/19/2009

11:59 AM

104

Page 104

Till BÜTTNER, Jeanette ORLOWSKY and Michael RAUPACH

Table 1: Characteristics of the reinforcement including standard deviation Filament Roving VET-FName ARGVET-RO-ARG-2400-1-05 2400-1-05 STD Polymeric EP 3.1.2 EP STF STD EP PRE1 Impregnation Titer tex 1.55 2400 2400 2400 2400 Tensile 890 r 52 1541 ± 64 1885 ± 72 1700 ± 96 N/m 1691 r 275 strength m² 53340 51330 60400 63200 56200 E-Modulus *: in 0° direction (direction of the applied load)

Concrete Within the scope of the SFB 532 (Collaborative Research Center 532 at Aachen University) a special micro grained concrete denominated as PZ 0899-01 has been developed, which serves complex requirements for the application for textile concrete, [7]. The epoxy modified micro grained concrete mixture (called EP 3.1.2-10%) is based on this mixture PZ 0899-01. The epoxy content of the concrete mixture is set to 10 Mass-% with regard to total binder content. The volume, which is filled by the added polymer, is taken into account as proportionately to the total masses of the other components; the ratios of the other materials are kept constant. In order to achieve a constant water/binder ratio of 0.4 in the concrete mixture the water content of the epoxy is taken into account. The compositions of PZ 0899-01 and the epoxy modified concrete are shown in table 2. The pH value of both concretes is about 13.5. All specimens were cured at 23 °C and 99 % RH for duration of 28 days. After this the accelerated ageing starts.

Table 2: Composition of fine grained concrete with and without polymer modification, [7] Cement Fly ash Silica Epoxy Sand Quartz Super(CEM I fume (fa) 3.1.2 < 0.6 mm flour plastiziser Name 52.5 R) (sf) kg/m³ PZ 0899-01 EP-3.1.2-10 %

490 458

175 164

35 33

65

714 667

499 467

%Ȉc+fa+sf 1.5 1.5

EXPERIMENTAL PROGRAM Test methods to characterize the bond between AR-glass and Matrix The bond behavior of the filaments in combination with the previously mentioned concrete was published before [5, 6], so this paper focuses on the roving-matrix interphase, which is relevant for the transition of the research work to building applications. For all test series the previous described cohesive interface model was used to determine the evaluation of the bond from global load-displacement curves, [4, 5, 9-15]. These global load-displacement curves were derived by using a tensile test combined with visual crack recording. The geometry of the test specimens is identical to the specimen, depicted in figure 3 (specimen called “TSP-specimen”), but there is a small but nevertheless important difference. If one is interested in the tensile strength of the fibers as a function of

11-Textile reinforced.qxd

9/19/2009

11:59 AM

Page 105

Textile reinforced concrete - durability issues: changes of the bond and tensile strength due to ageing

105

ageing, the reinforcement can best be glued with epoxy resin to the ends of the specimens, so that the tensile load is applied to the fibers directly (as in figure 3). Since in this part of the investigations however the behavior of the specimen should be similar to the behavior of a building component and the main focus is put on the interaction between the matrix and the textile, the load is not transferred directly to the fibers any more, but should be introduced from matrix to fibers and vice-versa. Three layers of textile made out of VET-RO-ARG 24001-05 rovings (called MAG 07-03) are used, which corresponds to reinforcement degree of 2.67% vol. - the specimens were 10 mm thick. The tensile load is applied via rounded off steel elements. The TSP-Tests are carried out at a displacement rate of 0.5 mm/min. During the tensile tests pictures are taken at each load step of 0.5 kN to detect the cracks at the surface of the composite specimen (see Figure 2). Three series of specimens were tested. On the first series, no accelerated ageing has been applied. The second series was tested after 14 days of accelerated ageing (the specimens were kept under water at 50°C) and the third series after 28 days of accelerated ageing. Before the specimens are tested, they are dried under ambient conditions for several days until their weight stabilizes.

Figure 2: Test setup used for characterizing the bond between reinforcement and matrix TSP-Test for investigation of the durability Other than described before, the aim while investigating the durability of TRC is to achieve a roving failure within the specimen. As mentioned above before this roving failure is achieved by gluing the rovings, which are sticking out of the specimen, to the end of the specimen. So the individual filaments are fixed at each end of the sample. During the loading process, filaments are unable to slip towards the centre of the sample from the sample edges. Therefore the rovings can reach their maximum tensile strength. Due to this sample preparation the TSP-Test allows conclusions concerning the changes in the stress-elongation behavior, the cracking image and the maximum roving tensile strength after climatic stress. The amount of reinforcement depends on the type of roving – e.g. using 2400 tex rovings 8 single rovings are required; if the AR-glass is polymer modified only 6 single rovings are used (2400 tex). As described previously the specimen are tested in a displacement-controlled test with a velocity of 0.5 mm/min. During testing the change of length is measured on two sides over 250 mm using electrical gauges.

11-Textile reinforced.qxd

9/19/2009

106

11:59 AM

Page 106

Till BÜTTNER, Jeanette ORLOWSKY and Michael RAUPACH

The test results are then used to compute the loss of strength according to the following formula:

(a)

(b)

Figure 3: Testing of a TSP specimen; (a) Top view of the specimen; (b) Cross-section

'f l , t with

'fl,t ft

1

ft ft 0

(4)

degree of strength loss at time t tensile strength of the specimen at time t

The computed loss of strength due to accelerated ageing can then be used to calibrate a model developed in the SFB 532, which allows predicting the long term loss of strength due to outdoor weathering, [2, 6]. At the moment the model is functional for non polymer-modified materials as it can bee seen in figure 4. It can be seen that the predicted loss of strength and the measured loss of strength of specimens, which were stored outdoors over a time of four years show a good alignment.

Figure 4: Outdoor weathering: measured and predicted strength loss after four years of outdoor weathering

11-Textile reinforced.qxd

9/19/2009

11:59 AM

Page 107

Textile reinforced concrete - durability issues: changes of the bond and tensile strength due to ageing

107

RESULTS Bond behavior between textile reinforcement and matrix due to ageing Figure 5 shows the composite strain and the distance between cracks as a function of the applied composite stress. For each series, only one stress-strain curve is shown so that the diagram is not overloaded, but the difference between several specimens in one series was relatively small. The points indicating the average distance between neighboring cracks are the average values of the three tested specimens per series. From these measurements of the strain Hc and crack distance x as function of stress Vc and with help of equations (1), (2) and (3) the normalized bond shear flow T can be calculated at several stress levels per tested specimen. These results are depicted in figure 6. Figure 5 and figure 6 show that, although a distinct loss of composite failure stress due to ageing was measured, no evolution of T with accelerated ageing was noticed. This does not mean that there are no effects of embrittlement at all, since the effect was only measured after the standard curing and hydration products will be mainly deposited during curing. One can however say that the studied standard cured specimens will not suffer further noticeable degradation due to embrittlement during the service life of a structural element made of TRC.

Figure 5: Strain and distance between crack as a function of tensile stress (TRC specimen 2.67% vol fibres); cured at 23 °C and 99 % RH for 28 d before ageing

11-Textile reinforced.qxd

108

9/19/2009

11:59 AM

Page 108

Till BÜTTNER, Jeanette ORLOWSKY and Michael RAUPACH

Figure 6:Normalized bond shear flow T as function of ageing,TRC specimen 2,67% vol fibres Durability of polymer modified textile reinforced concrete The presented results were obtained by using the previously described TSP-specimens with the reinforcement glued to the end of the specimens. After the curing (28 d at 99 % RH) the specimens were stored under water at 50 °C in order to age them accelerated. This accelerated aging represents one possible set of parameters. In order to predict the long-term loss of strength due to outdoor weathering with the presented model, the ageing as a function of humidity and temperature has to be investigated. It is not possible to predict the long-term loss of strength just on the basis of a single parameter ageing. The detailed way is described in [2, 6]. The influence of different polymeric modifications of the reinforcements can be seen in figure 7 left, where the test results are displayed – the lines represent regression curves calculated with the durability model. In all investigations the reference is represented by a non-polymer modified concrete with AR-glass reinforcement, which is supposed to be the harshest environment for AR-glass reinforcement. It can be seen that after 200 days of accelerated ageing (50 °C and water storage) the loss of strength of the non-impregnated textile reinforcement in a non modified concrete is approximately 60 %. The use of a polymeric impregnation leads to a noticeable reduction of the loss of strength in accelerated ageing. The cold-hardening resin (EP STF STD) leads to higher losses than the Prepreg-Resin. At the moment it is assumed that the diffusion of alkaline ions is mainly effected by the different structures of the two resins. The actual research is dealing mainly with this phenomenon. In figure 7 right the results of a polymer modified concrete in combination with a plain AR-glass reinforcement and a polymer modified reinforcement are shown –the identical polymer was used for both modifications. It can be seen, that the loss of strength is reduced by the use of a polymer modified concrete. Investigations, which were carry out in parallel, show that the capillary water absorption is significantly reduced by the PCC. Actual research deals with the question how big does the reduction of the capillary water uptake has to be in order to reduce the loss of strength effectively and what the influence of the storage conditions (especially storage under water) on the behavior of the specimens is. By combining the PCC with a polymer modified reinforcement (identical polymer) the loss of strength can

11-Textile reinforced.qxd

9/19/2009

11:59 AM

Page 109

Textile reinforced concrete - durability issues: changes of the bond and tensile strength due to ageing

109

be reduced to nearly zero. This behavior could only be detected with this combination of polymers. The reasons are not clarified yet.

Figure 7: Left: strength loss of TSP-specimens using two different polymer modified reinforcements; right: strength loss of TSP-specimens using polymer modified reinforcement as well as polymer modified concrete

CONCLUSIONS

Two aspects of the durability of TRC are discussed in this paper. Results dealing with the evolution of the bond between matrix and reinforcement as well as the loss of strength of the AR-glass reinforcement due to glass corrosion are presented. The measurement of the stress-strain behavior was used in combination with the crack counting technique to evaluate the bond behavior under accelerated ageing for full TRC composite specimens. These investigations showed no loss of bond strength, although a distinct loss of the composite failure strength was detected. This loss of strength was investigated for various polymer modified materials. The durability of impregnated AR-glass textiles embedded in micro grained concrete is improved compared to the non-impregnated AR-glass textiles. But also a polymeric modification of the matrix can reduce the loss of strength of the AR-glass reinforcement. Actual research deals with the reasons for the improvement of the durability caused by polymeric systems [1]. As mentioned above, it is also important to mention that the obtained values of the strength loss in accelerated ageing conditions can not be correlated directly with any strength loss due to outdoor weathering – in order to predict a long term strength loss, calculations with the durability model have to be carried out. ACKOWLEDGEMENTS

The investigations at the ibac, RWTH Aachen University, are part of the Collaborative Research Center 532 ”Textile reinforced concrete – Basics for the development of a new technology” and sponsored by the Deutsche Forschungsgemeinschaft (DFG). The support is

11-Textile reinforced.qxd

110

9/19/2009

11:59 AM

Page 110

Till BÜTTNER, Jeanette ORLOWSKY and Michael RAUPACH

gratefully acknowledged. Furthermore the close collaboration with Dr. ir. H. Cuypers (VUB – Brussels) is gratefully acknowledged. REFERENCES

1. Keil, A., Raupach, M.: Improvement of the load-bearing capacity of textile reinforced concrete by the use of polymers, in: Proceedings of Ohama Symposium, 12th International Congress on Polymers in Concrete ICPIC'07, September 26-28, 2007 Chuncheon, Korea, 873-881 2. Orlowsky, J.: Zur Dauerhaftigkeit von AR-Glasbewehrung in Textilbeton (Durability of TRC; PhD Thesis). Berlin: Beuth. - In: Schriftenreihe des Deutschen Ausschusses für Stahlbeton, 2005, No. 558 3. Orlowsky, J., Raupach, M.: Modelling the loss in strength of AR-glass fibres in textilereinforced concrete. In: Materials and Structures (RILEM) 39, 2006, No. 6, 635-643 4. Cuypers H., Wastiels J.: A stochastic matrix cracking theory for glass-fibre reinforced cementitious composites. Materials and Structures, vol. 39 (292), 2006, 777-786 5. Cuypers H., Wastiels J., Van Itterbeeck P., De Bolster E., Orlowsky J. and Raupach M.: Durability of glass fibre reinforced composites experimental methods and results, Composites Part A: Applied Science and Manufacturing, vol. 37 (2), 2006, 207-215 6. Büttner, T., Orlowsky, J., Raupach, M.: Investigation of the durability of textile reinforced concrete - test equipment and modelling the long-term behaviour: [ISBN 9782-35158-046-2] Bagneux: RILEM, 2007. - RILEM Proceedings PRO 53. - In: High Performance Fiber Reinforced Cement Composites (HPFRCC5), Proceedings of the Fifth International RILEM Workshop, Mainz, July 10-13, 2007, Reinhardt, H.W. and Naaman, A.E. eds., 333-341 7. Brameshuber, W.; Brockmann, T.; Hegger J.; Molter, M.: Untersuchungen zum textilbewehrten Beton (Investigations of TRC). Beton 52, 2002, Nr.9, 424-429 8. Orlowsky J., Raupach M., Cuypers H., Wastiels J.: Durability modelling of glass fibre reinforcement in cementitious environment. Materials and Structures, vol. 38 (276), 2005, 155-162 9. Banholzer B., Brameshuber W., Jung W.: Analytical simulation of pull-out-tests - the direct problem. Cement and Concrete Composites, 27 (1), 2005, 93-101 10. Banholzer B., Brameshuber W., Jung W.: Analytical evaluation of pull-out-tests - the inverse problem. Cement and Concrete Composites 28, 2006, 564-571 11. Pryce A. W., Smith P. A.: Matrix cracking in unidirectional ceramic composites under quasi-static and cyclic loading. Acta Metall. Mater., No. 41., 1993, 1269-1281 12. He M. Y., Wu B.-X., Evans A. G., Hutschinson .J W.: Inelastic strains due to matrix cracking in unidirectional fiber-reinforced composites, Mechan. Mater., vol. 18, 1994, 213 13. Curtin W. A.: Multiple matrix cracking in brittle matrix composites, Acta Metal. Mater., No. 41, 1993, 1369 14. Curtin W. A.: Stochastic damage evolution and failure in fibre-reinforced composites, Advances in Applied Mechanics (36), 1999, 163-253 15. Naaman A. E., Namur G. G., Alwan J. M., Najm. H. S.: Fiber pull-out and bond slip, I: analytical study. II: experimental validation. Structural engineering 117 (9), 1991, 27692800

12-Impact study.qxd

9/19/2009

11:59 AM

Page 111

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

IMPACT STUDY OF TEXTILE REINFORCED CEMENTITIOUS MATERIALS: TEST METHOD AND PRELIMINARY RESULTS

J. Van Ackeren, J. Blom, D. Kakogiannis, J. Wastiels, D. Van Hemelrijck, Vrije Universiteit Brussel, Department of Mechanics of Materials and Constructions, Pleinlaan 2, 1050 Brussels, Belgium, e-mail: [email protected] S. Palanivelu, W. Van Paepegem, J. Degrieck, Department of Materials Science and Engineering, Universiteit Gent, Sint-Pietersnieuwstraat 41, 9000 Gent, Belgium. J. Vantomme, Koninklijke Militaire School, Civil and Materials Engineering Department, Building G, Level 0, 8 Av. Hobbema B-1000, Brussels, Belgium.

ABSTRACT In building engineering, impact loading and other accidental loads are mostly taken into consideration by measures on the structural level rather than on the material level, the latter one is even seldom explored. Textile reinforced Cements (TRCs) is a composite material group which is still in full development. Its potential to dissipate a significant amount of energy under low velocity impact loading was already indicated in a preliminary study. In the present study, the low velocity impact behaviour of TRC laminates is investigated more closely and this by means of an instrumented drop weight test. This study has shown the ability of the used test method to identify damage mechanisms underlying the energy absorption under low velocity impact loading in TRC composite laminates.

Keywords Low-Velocity Impact, Textile Reinforced Cements (TRCs), energy profile, damage mechanisms

INTRODUCTION In the last few decades, standards concerning impact and accidental loading have become more demanding with regards to the performance of materials. In aerospace engineering and aviation the solution was found in the use of several traditional composite materials which are tested extensively under impact loading, but also in special developed (tailored) impact resistant composite materials (e.g. aramid fibre composites). However, in building engineering most of the developed solutions are outside of the material level and are based on design calculations including safety factors or by redistribution of forces within the overall structure (e.g. progressive collapse in buildings) [1]. Improving the impact resistance of building constructions by the development of new (tailored) materials has until this present day not been explored yet. Steel reinforced concrete is the main material used in building constructions, due to the synergy of two materials, one with good compressive behaviour and the other with good tensile behaviour. However the main disadvantage of steel reinforced concrete is its own weight. By

12-Impact study.qxd

112

9/19/2009

11:59 AM

Page 112

J. Van ACKEREN, J. BLOM, D. KAKOGIANNIS, J. WASTIELS, D. Van HEMELRIJCK...

introducing fibres into concrete, the steel rebars can be (partially) omitted. The fibres provide more ductility to the concrete mix and more energy can be absorbed under loading. Next to the previous advantages, the fibres also contribute to the shear reinforcement within structural elements. Textile Reinforced Concrete/Cements (TRC) is a group of composite materials which is still being developed. By using textile reinforcements (e.g. woven textiles or chopped strand mats) in stead of loose fibres and production techniques of the composites industries, much higher fibre volume fractions can be reached than is the case with steel Fibre Reinforced Concrete (FRC) (up to 20 vol%, FRC: 2%). This higher fibre volume fraction significantly improves the mechanical properties of the material (higher strength and higher toughness). As a consequence, TRC can be used in load bearing applications and steel rebars can be omitted. This also implies that a concrete cover for the protection of steel rebars is not needed, which leads to very thin (minimum 1 to 2 mm) and even curved load bearing structures. TRCs have been successfully used in lightweight load bearing structures, cladding panels and protective (chemical) elements. The TRC which is investigated in this work consists of an Inorganic Phosphate Cement (IPC) matrix reinforced with 2D-random chopped glass fibre strand mats with a density of approximately 300 g/m². IPC, commercially available under the name Vubonite®, was developed at the Vrije Universiteit Brussel and consists of a liquid component based on phosphoric acid solution containing inorganic metal oxides and a calcium silicate powder component. This material has the advantage of being liquid during manufacturing, allowing production techniques of composite industries, such as hand lay-up and pulltrusion to be applied. Since IPC exhibits a neutral pH after hardening, the chemical attack of the fibres, generally associated with the use of glass fibres in an alkaline environment such as Portland cement, is significantly reduced. In earlier work [2] the potential of textile reinforced IPC as an energy absorbing material was shown. However, within these tests the reproducibility of the tests itself was not well controlled and there was no instrumentation on the impact event itself, since the aim was to investigate the potential of a TRC under impact loading. This study therefore focuses on post-impact behaviour and measurements of the residual deformations and residual mechanical properties. Compared to the previous publication, a different test set-up is used to investigate the impact behaviour of textile reinforced IPC more thoroughly.

THEORETICAL BACKGROUND Methodology The methodology for performing the drop weight tests on TRC laminates and subsequent processing of the results are similar to D. Liu et al [3, 4] and C. Atas and O. Sayman [5]. These researchers presented a new technique for the characterization of impact properties and the correlation with damage mechanisms of composite laminates. A large series of laminates was tested at increasing impact energy levels, either by increasing the drop height of the impactor or by increasing its weight. For each laminate a force-deflection curve is drawn (see figure 1), showing an ascending stage up to peak load followed by a descending stage depending on the impact energy. When the impactor rebounds after the impact, a closed force deflection curve is obtained, while an open curve is typical for a test in which all impact energy is absorbed and no

12-Impact study.qxd

9/19/2009

11:59 AM

Page 113

Impact study of textile reinforced cementitious materials: test method and preliminary results

113

more rebound energy is available. When all of these curves are combined in one figure for one series of laminates, it is called a mastercurve (figure 1, left). For each test the absorbed energy obtained from the area under the load-deflection curve can be plotted against the impact energy, which is the kinetic energy of the impactor just before impact. The obtained graph is called the energy profile and is depicted in figure 1, right. From this graph it is clear that a composite material is most efficient if the absorbed energy is equal to the impact energy. Together with the mastercurve, the energy profile provides a lot of information on the damage occurring during impact.

Figure 1: left: force-deflection mastercurve, right: energy profile (figures from D.Liu [3]) Damage mechanisms and main static properties The main impact characteristics and properties of most composite laminates were found to reach critical values when perforation occurred [3]. Therefore perforation was assumed to be the most critical damage mechanism leading to maximum energy absorption. Liu et al define a perforation and a perforation threshold, at which the curve from the energy profile intersects with the equal energy line (impact energy = absorbed energy) (see figure 1, right). Also the effects of delamination and other damage mechanisms can be discovered by studying the energy profile together with the impact data. Since damage mechanisms within TRC composite laminates differ from other composite materials, the major damage mechanisms are briefly described in this part. Understanding the static behaviour of TRC can provide essential information on the occurring damage mechanisms during impact loading. However, local damage due to contact with the impactor and effects due to vibrations will be added to this global damage. In compression a TRC behaves linear elastic up to failure. In tension the material shows a highly nonlinear stress-strain behaviour which has been studied and modelled in literature [6-8]. A typical stress-strain curve obtained from a tensile test on a textile reinforced IPC laminate specimen from a generally exhibits three distinct stages: a first linear elastic stage where the fibre-matrix bond is globally adhesive and thus the elastic modulus can be obtained by the classical law of mixtures. Once the global composite stress exceeds the matrix strength (around 10MPa), multiple cracking will occur leading to a fine parallel crack pattern and a deflection of the stress-strain curve. In the vicinity of a crack, the fibre-matrix bond is reduced to a frictional shear stress. Due to matrix cracking and fibre matrix de-bonding within the multiple cracking stage, new surfaces are formed causing energy dissipation. When all cracks are formed, the material behaves linear again (with a lower slope) and only the fibres will carry extra load up to failure of the composite.

12-Impact study.qxd

114

9/19/2009

11:59 AM

Page 114

J. Van ACKEREN, J. BLOM, D. KAKOGIANNIS, J. WASTIELS, D. Van HEMELRIJCK...

In a previous study [2], it was indicated that the strength loss due to impact might be related to matrix-cracking, subsequent de-bonding and further deterioration of the interface. However local fibre breakage was also observed when the impact energy was high enough.

MATERIALS AND TEST SET-UP Test specimens The investigated material is a composite material with the above described IPC as matrix material and glass fibre 2D-random chopped strand mats with a density of approximately 300 g/m² and a fibre bundle length of 5 mm as reinforcement. Three series of plates with dimensions 340mm x 340mm are manufactured by hand lay-up. Subsequently 4 specimens of 25 mm thickness and a length of 260 mm are cut off together with a plate of 260mm x 260mm. The latter are the test specimens for the impact tests and the other specimens are tested in tension in order to determine the mechanical characteristics under static loading. All plate specimens were clamped on 4 edges in a reproducible way with a clamping system provided with bolds, which were screwed equally with a torque key. An overview of the test specimens for the impact tests is given in table 1. There are three different series of plates: a series of five specimens containing 4 layers of fibres, one of four specimens containing 8 layers and a last series of specimens with 12 layers of fibres. Within the table, the thicknesses and fibre content are shown together with the total weight of each plate. The laminates are tested at different drop heights resulting in multiple impact energies and velocities.

Table1: overview of the impact specimens with their properties and test input energy (J)

200,7

Impact velocity (m/s) 1,98

150,0

1,72

11,8

249,8

100,6

1,40

7,9

23,0

248,8

150,2

1,72

11,8

1,95

24,2

230,2

150,0

1,72

11,8

3,70

25,6

463,3

250,0

2,21

19,6

JVA-DW-8-2

3,71

25,4

461,3

350,2

2,62

27,5

JVA-DW-8-3

3,98

23,7

501,7

350,1

2,62

27,5

JVA-DW-8-4

3,85

24,6

471,6

350,1

2,62

27,5

JVA-DW-12-1

5,90

24,0

724,6

550,1

3,29

43,2

JVA-DW-12-2

6,06

23,4

749,8

599,9

3,43

47,1

JVA-DW-12-3

5,80

24,4

728,4

750,1

3,84

58,9

JVA-DW-12-4

5,86

24,2

712,0

750,1

3,84

58,9

thickness (mm)

Fibre volume fraction (%)

weight (g)

drop height (mm)

JVA-DW-4-1

2,00

23,7

244,6

JVA-DW-4-2

1,85

25,6

221,0

JVA-DW-4-3

2,05

23,0

JVA-DW-4-4

2,06

JVA-DW-4-5 JVA-DW-8-1

name

15,8

Experimental set-up The test set-up, which is used for the impact tests, was developed at the Universiteit Gent and can be found in [9]. The impactor is sliding along guidance rails that are supported horizontally against the wall. The impactor has a weight of 8 kg and a maximum drop height of 3m providing maximum impact energy of 235 J. The impactor is equipped with a dynamic load sensor with a range of 20 kN and an accelerometer from which the displacements can be derived by integrating

12-Impact study.qxd

9/19/2009

11:59 AM

Page 115

Impact study of textile reinforced cementitious materials: test method and preliminary results

115

the signal twice. Another measuring technique that is used to calculate the displacement of the impactor is Digital Image Correlation (DIC). For this technique a high speed camera with a recording rate of 5000 fps is used (Photron APX RS 250K). With an impact duration of 30ms, each impact event is recorded in approximately 150 images. This method was applied in an earlier study on the same set-up by S. Palanivelu et al. [9]. The results of these measurements were found to be in good correlation with the measurements on the system itself and with the data derived from the accelerometer. The same procedures for processing the data are applied in this paper. The images from the camera can also be used to follow the impact event itself (see figure 2) in order to obtain more information on the occurring damage(s).

Speckle pattern for DIC

Impactor Test plate Figure 2: impact event captured by high speed camera

COMPARISON OF THE MEASURING TECHNIQUES Data processing Data from the high speed camera is processed with the commercially available DIC software Vic2D. Since the results were good, the same speckle pattern and subset size were used as in [9]. The data of the different measuring systems (DIC, load sensor, accelerometer and linear measurement system (LIMES)) are processed so that comparisons can be made. The signal of the accelerometer can be integrated two times giving consecutively velocity against time and displacement against time. The displacement-time data from the DIC can be derived twice in order to obtain velocity and acceleration. The derivation of the displacement was filtered in order to avoid accumulated error. From the derived acceleration time history, the force-time history can be calculated through Newton’s first law. Comparison The displacements integrated from the acceleration signal of the accelerometer are compared with those originating from the LIMES system and DIC. Most of these results showed a good correlation (see figure 3a), however in a few cases the displacements derived from the DIC are shifted up or down a little while the shape of the curve stays the same (see figure 3b). This might be due to minor errors in the calibration of the DIC. Figure 4a shows the velocity-time curves, obtained from DIC and accelerometer, of a specimen which was impacted with a calculated

12-Impact study.qxd

116

9/19/2009

11:59 AM

Page 116

J. Van ACKEREN, J. BLOM, D. KAKOGIANNIS, J. WASTIELS, D. Van HEMELRIJCK...

initial velocity of 2,62 m/s. This initial velocity is also found on the curves and moreover the curves are almost identical. Figure 4b is a plot of load-time curves experimentally measured and derived from DIC. It can be concluded that the DIC technique is a good alternative for contact measuring techniques. 16

20

Displacement-time, JVA-DW-8-2

Displacement-time, JVA-DW-8-4

18

14

16

Displacement (mm)

Displacement (mm)

12 10

8 6

14 12 10 8 6

DIC_derived

4

2

accelerometer

4

measured

2

accelerometer

a) 0,005

0,01

0,015

measured

b)

0

0 0

DIC_derived

0

0,02

0,005

Time (s)

0,01

0,015

0,02

Time (s)

Figure 3: comparison of displacement-time curves from different measuring techniques 4500

3

Force-time, JVA-DW-8-1

Velocity-time, JVA-DW-8-2 2,5 3500

integrated velocity from data accelerometer

2

2500

DIC derived velocity

1

Force (N)

Velocity (m/s)

1,5

0,5

1500

500

0 0 -0,5

0,005

0,01

0,015

Time (s)

-500

-1

-2

0,01

0,02

0,03

Time (s) load cell

-1500

-1,5

0

a)

DIC derived

-2500

b)

Figure 4: comparison of: a) velocity profile, b) force-time history

RESULTS OF IMPACT TESTS Force-deflection curves The obtained force-deflection curves are presented in figure 5 and 6. None of these curves are closed curves which is the result of the non linear behaviour of the laminates. Matrix-cracking occurs at relatively low stresses, leading to permanent damage and thus residual deflections after impact. Generally, the peak force and deflections increase with increasing impact energy and the shape of the curves is similar for all tests.

12-Impact study.qxd

9/19/2009

11:59 AM

Page 117

117

Impact study of textile reinforced cementitious materials: test method and preliminary results

After comparison with images from the high speed camera and knowledge of the material under study, the force deflection curves can, according to this author, be divided into 4 stages (figure 6b). When the impactor hits the plate it is expected that the force will increase immediately. The first stage however shows a zigzag shaped flat signal. This might be due to the fact that the surface of the plate is never totally flat causing local matrix crushing. Another possible explanation for this phenomenon is that the impactor is not completely rigid, which leads to vibrations. After this first stage a second ascending stage is observed. Comparison with images originating from the high speed camera indicates that this stage represents the bending of the laminate up to peak force. After the peak force is reached, the force begins to drop again while the deflections still increase. This stage does not represent the rebounding of the impactor since the deflections keep increasing. On the high speed images it is clearly visible that the impactor is still going down while the plate is coming back up. This means that it is most likely that this stage is linked to the local perforation of the laminate by the impactor. Finally, the deflections also begin to decrease with decreasing force, which was identified as being the stage where the impactor rebounds from the plate. These results have to be confirmed in future work, but are very promising concerning damage identification. If it is possible to link each stage to one or more damage mechanisms, this makes it easier to optimize/tailor the behaviour of the plate. During the tests, the reproducibility was checked by performing identical tests on a limited number of plates, keeping the testing conditions (boundary conditions, drop height and drop weight) constant. The obtained force-deflection curves are very similar (identical) to each other, with exception of laminates JVA-DW-12-3 and -4 (The problem within one of these two tests is investigated). However, the curves seem to be shifted with respect to each other. This is due to differences in thicknesses of the plates resulting from the hand lay-up technique. Static bending laws imply that a thinner laminate will encounter more deflection which will be proportional with the third power of the thickness in bending and with only the thickness in shear. 4500

2000

8 layers of fibres

4 layers of fibres

1800

4000

JVA-DW-4-1_200,7mm

JVA-DW-8-1_250mm

JVA-DW-4-2_150mm

1600

3500

JVA-DW-8-2_350,2mm

JVA-DW-4-3_100,6mm

1200

JVA-DW-8-3_350,1mm

3000

JVA-DW-4-5_150,2mm JVA-DW-44_150mm_SimplySupported

force (N)

force (N)

1400

1000 800

JVA-DW-8-4_350,1mm

2500 2000 1500

600

1000

400

500

200

a)

b) 0

0 0

5

10

deflection (mm)

15

20

0

5

10

15

20

deflection (mm)

Figure 5: Force-displacement mastercurves for: a) plates with 4 layers of fibres, b) with 8 layers For the results of the plates JVA-DW-4-2_150mm and JVA-DW-4-5_150,2mm (figure 5a) the curves fit together if the deflections are normalised by multiplying them with the square

12-Impact study.qxd

118

9/19/2009

11:59 AM

Page 118

J. Van ACKEREN, J. BLOM, D. KAKOGIANNIS, J. WASTIELS, D. Van HEMELRIJCK...

of the thickness. This can be ascribed to a combination of shear and bending. To avoid other information to be lost, the normalization is not applied in the following results. Next to the reproducibility, the effect of boundary conditions was tested. In the series of laminates with 4 layers of fibres, one laminate was tested simply supported at all edges. The curve shows a clear difference in behaviour (figure 5a): the plate reacts more elastic and shows bigger deflections without leading to higher energy absorption (see energy profiles).

8000

2000

4 stages in F-d curve

12 layers of fibres 7000

1800

JVA-DW-12-1_550,1mm JVA-DW-12-2_599,9mm

1600

JVA-DW-12-3_750,1mm

1400

JVA-DW-12-4_750,1mm

1200

5000

force (N)

force (N)

6000

4000 3000

Stage 2: bending

1000

Stage 3: perforation

800

Stage 1

600 2000

400 1000

5

10

15

b)

0

0 0

Stage 4: rebound

200

a)

0

20

5

10

15

20

deflection (mm)

deflection (mm)

Figure 6: a) F-d mastercurve for plates with 12 layers, b) four stages in F-d curve Energy profile In the energy profiles depicted in figure 7, it is clear that the absorbed energy gets nearer to the equal energy line when the impact energy is increased. In the energy profile of the laminates with 8 layers the point most to the right actually consist of three points, which again shows the good reproducibility of the tests. To obtain a full energy profile, more tests are necessary. However, from these tests it can be concluded that the method of energy profile can be used for TRC laminates, but the interpretation of the curves needs to be done differently due to the difference in damage mechanisms.

35

25

4 layers

15

10

simply supported

5

25 20 15 10 5

a) 0

5

10

15

impact energy (J)

20

25

50 40 30 20 10

b)

0

0

12 layers

60

absorbed energy (J)

absorbed energy (J)

20

absorbed energy (J)

70

8 layers

30

c)

0 0

5

10

15

20

25

30

35

0

10

impact energy (J)

Figure 7: energy profiles for the three series of tests

20

30

40

50

impact energy (J)

60

70

12-Impact study.qxd

9/19/2009

11:59 AM

Page 119

Impact study of textile reinforced cementitious materials: test method and preliminary results

119

Post-impact tests After the impact tests, the TRC laminates are visually inspected and subsequently cut into strips which are subjected to post impact tensile tests. Post-impact tests were also done in the work of Cuypers and Van Ackeren [2], resulting in essential information on occurring damage mechanisms under impact loading. In this paper post-impact tensile tests are performed to check the residual mechanical properties of the impacted plates. Visual inspection consists of spreading an ink solution on the non impacted side of the plate to make the cracking pattern in the matrix visible. This technique was only successful on the laminates containing 4 layers of fibres, because some fibre bundles were to close to the surface of the plate in the other series preventing the ink to reveal the cracks. Figure 8 illustrates the different reaction of plates JVA-DW-4-1 and -3. The latter was loaded with an impact energy of 7,9J and a velocity of 1,4 m/s while the first plate was loaded with an impact energy of 15,8J at a speed of 1,98 m/s. These figures show that the plate JVA-DW-4-1 has more local damage since the cracking pattern is less dense near the edges than on plate JVA-DW-4-3. The post impact tensile tests were performed on 5 strips of 50mm x 260mm cut from the impacted plates. The specimens are numbered 4-1-L (left), 4-1-ML (middle left), 4-1-C (centre), 4-1-MR (middle right) and 4-1-R (right). The centre strip contains the point of impact and thus has a smaller net section in the middle. Tensile tests were performed on an Instron 5885 static test bench with a cross head speed of 1 mm/min. From these stress-strain curves it can be concluded that the impact has reduced the strength with more than 50% in the centre specimen. Also the stiffness in the third stage, which is the stiffness of the fibres multiplied by the fibre volume fraction, is reduced with more than 50%. However, in the other post impact specimens of the same plate the reduction in stiffness of the third stage and maximum stress and strain is rather low (maximum 20%). In the first stage, the stiffness is reduced within the whole plate. This is normal since the impact induced stresses in the plate causing multiple cracking [6]. These post-impact results together with the results from the impact tests prove that the energy is dissipated partly globally, but the major energy absorption is done by the local damage around the impact point.

Figure 8: Crack pattern and local penetration damage of plates JVA-DW-4-1 and-3

12-Impact study.qxd

120

9/19/2009

11:59 AM

Page 120

J. Van ACKEREN, J. BLOM, D. KAKOGIANNIS, J. WASTIELS, D. Van HEMELRIJCK...

CONCLUSIONS The presented work within this paper showed some promising results. The method of energy profile, where the absorbed energy and the main impact characteristics are linked with damage induced by a low velocity impact, seems to be a suitable method for TRC composite laminates. However interpretation of the results should be done differently due to the typical damage mechanisms in TRC composites, as multiple matrix cracking and fibre-matrix de-bonding. The test set-up which was used in this work has been proved to be reproducible and the data is valuable to obtain information on the damage induced by low velocity impact loading. Although the results are good, extensive testing with more detailed analysis of the individual F-d curves and the damage mechanisms should be performed in future work in order to obtain a full analysis of an individual force-displacement curve, the dynamic characteristics of all individual parts in the testing system should be known. To gain more insight in the damage mechanisms occurring in the test plates, the test set-up will be adapted so that two high speed camera’s can be placed under the test plates. With the use of the DIC technique the in- and out of plane displacements can then be monitored during the complete impact tests. If the four different stages in the F-d curve can indeed be separated, it would be interesting to compare the bending stage with static bending tests. This would initially be performed on beam specimens in order to make simplified comparisons. In parallel future work, it will be attempted to use numerical models to predict the behaviour of TRC laminates. The presented test results can be used for the validation of this model. ACKNOWLEDGEMENTS Research funded by a Ph.D grant of the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen).

REFERENCES 1. (UFC) 4-023-03, Design of buildings to resist progressive collapse, Department of Defense, USA, 2005 2. Cuypers, H., Van Ackeren, J., Belkassem, B., Wastiels, J., Impact resistance of load bearing sandwich elements with textile reinforced concrete faces, Conference: ECCM 13, “13th European Conference on Composite Materials”, Stockholm,02-05 June 2008 3. Liu, D., Raju, B.B., Dang, X., Impact Perforation Resistance of Laminated and Assembled Composite Plates, Int. J. of Impact Engineering, 24(6-7), 2000, 733-746 4. Liu, D., Richman, T., Li, G., Raju, B., Penetration and perforation thresholds of laminated glass/epoxy composite plates, Proc. of the 13th Int. Conf. on Composite Materials (ICCM13), 25-29, paper 1320, Beijing, China 5. Atas, C., Sayman, O., An overall view on impact response of woven fabric composite plates, J. of Composite Structures, 82, 2008, 336-345 6. Aveston, J., Cooper, G.A., Kelly, A., Single and multiple fracture, The properties of fibre composites, National Physical Laboratories Conf. Proc., IPC Science & Technology Press Ltd. London, 1971:15-24 7. Cuypers, H., Analysis and Design of Sandwich Panels with Brittle Matrix Composite Faces for Building Applications, PhD Thesis VUB, 2002, available at http://wwwtw.vub.ac.be/memc/web site/index.htm 8. Proc. of the 1st Int. RILEM Conf. on Textile Reinforced Concrete, September 6-7, RWTH Aachen University, Germany, J. Hegger, W. Brameshuber and N. Will, 2006, Pages: 418 9. Palanivelu, S. De Pauw, S. Van Paepegem, W., Degrieck, J., Van Ackeren, J., Kakogiannis, D., Wastiels, J., Van Hemelrijck, D., Vantomme, J., Validation of digital image correlation technique for impact loading applications, Conference: DYMAT2009: 9th Int. Conf. on the Mechanical and Physical Behaviour of Materials under Dynamic Loading, Brussels, 07-11 Sept. 2009

13-The ductile.qxd

9/19/2009

11:59 AM

Page 121

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

THE DUCTILE BEHAVIOR OF HPFRCC IN COMPRESSION Alessandro P. FANTILLI(1), Hirozo MIHASHI(2), Paolo VALLINI(1), Bernardino CHIAIA(1) (1)

Department of Structural and Geotechnical Engineering Politecnico di Torino, Corso Duca degli Abruzzi, 24 -10129 Torino, Italy e-mails: [email protected], [email protected], [email protected] (2)

Department of Architecture and Building Science Tohoku University, Aramaki Aoba 6-6-11-1209, Aoba-ku, Sendai 980-8579, Japan. e-mail: [email protected]

ABSTRACT The ductility of High Performance Fiber Reinforced Cementitious Composites (HPFRCC) can be developed both in tension and compression. This aspect is evidenced in the present paper by measuring the mechanical response of normal concrete (NC), self-compacting concrete (SC) and HPFRCC cylindrical specimens under uniaxial and triaxial compression. The post-peak behaviour of these specimens is defined by a non-dimensional function that relates the inelastic displacement and the relative stress during softening. Both for NC and SC, the increase of the fracture toughness with the confinement stress is observed. Conversely, HPFRCC shows, even in absence of confinement, practically the same ductility measured in normal and self-compacting concretes with a confining pressure. Thus, the presence of HPFRCC in compressed columns is itself sufficient to create a sort of active distributed confinement.

Keywords High performance concrete, self-compacting concrete, confining pressure, triaxial tests, fracture toughness.

INTRODUCTION Several reinforced concrete (RC) structures fail via concrete crushing in compressed zones. This is the case, for instance, of over-reinforced concrete beams, like in the four point bending tested by Mansur et al. [1]. When fiber-reinforced, the post-peak behaviour of such members is remarkably more ductile than that observed in beams having the same geometry, the same steel rebars, and the same bearing capacities, but made of normal concrete (NC) without fiber. Thus, when crushing occurs, the type of concrete rules both the mechanical response and the ductility of RC structures. The experimental campaign conducted by Khayat et al. [2] on highly confined RC columns, subject to concentric compression, also confirms the influence of the cement-based composites on the structural performances. More precisely, for a given cross-section, the load vs. average axial strain diagrams appear more ductile in the case of columns made of selfcompacting concrete (SC) than in NC columns. These experimental observations can be usefully applied to designing RC compressed columns in seismic regions.

13-The ductile.qxd

122

9/19/2009

11:59 AM

Page 122

Alessandro P. FANTILLI, Hirozo MIHASHI, Paolo VALLINI and Bernardino CHIAIA

Figure 1. The stress-strain relationship of compressed concrete with and without confinement [4].

According to Eurocode 8 [3], if a required ductility cannot be attained because concrete strains are larger than 0.35% , a compensation for the loss of resistance due to crushing can be achieved by means of an adequate confinement (§ 5.4.3.2.2). Such a confinement, usually provided by transversal steel reinforcement (i.e., stirrups), and indicated by the confining pressure V3 (Fig.1), allows designers to consider a more ductile stress strain (Vc-Hc) relationship in compression. For instance, Fig.1 shows the so-called parabola-rectangle diagrams proposed by Eurocode 2 [4] for confined and unconfined concretes (§3.1.9). Short steel fibers randomly dispersed in a cement-based matrix can generate confining pressures comparable with that of stirrups. The experimental campaign of Ganesan and Ramana Murthy [5], performed on short confined columns with and without fibers (Fig.2a), investigates on this aspect. As shown in Fig.2b, the applied load- average strain (P-Hcm) diagram of RC columns, made with ordinary concrete and a transversal reinforcement percentage equal to Us=1.6%, is more or less similar (in terms of strength and ductility) to that of fiber-reinforced (FRC) columns, made with a reduced quantity of stirrups (Us =0.6%) and FRC (volume fraction Vf = 1.5%, aspect ratio L/) = 70).

Figure 2. The columns tested by Ganesan and Ramana Murthy [5].

13-The ductile.qxd

9/19/2009

11:59 AM

Page 123

The ductile behavior of HPFRCC in compression

123

Although fiber-reinforcements have been introduced in order to increase the ductility of cement-based composites in tension, they can also provide a sort of confinement, and therefore higher ductility in compression. For this reason, in concretes capable of reaching high tensile strains at failure, like the High Performance Fiber Reinforced Cementitious Composites (HPFRCC) tailored by Mihashi and co-workers [6], higher compressive fracture toughness should be expected. To confirm such a conjecture, the post-peak responses of different cementitious composites under uniaxial and multi-axial compression are here investigated.

POST-PEAK RESPONSE OF CONCRETE UNDER COMPRESSION The stress-strain relationships of concrete and quasi-brittle materials in compression (Fig.3a) can be divided into two parts (Fig.3b). In the first part, when the stress is lower than the strength fc (and Hc < Hc1 ), the specimen can be considered undamaged. In the case of plain concrete, the ascending branch of Vc-Hc can be defined by the Sargin’s relationship proposed by CEB-FIP Model Code [7]. As soon as the peak stress is reached, localized damage develops and strain softening begins. In this stage, the progressive sliding of two blocks of the cement-based material is evident. In Fig.3a, the angle between the vertical axis of the specimen and the sliding surfaces is assumed to be D=18°. This value, as measured in many tests, can be also obtained through the Mohr-Coulomb failure criterion, if the tensile strength is assumed to be 1/10 of compression strength ( fct = 0.1 fc ). The inelastic displacement w of the specimen, and the consequent sliding s of the blocks along the sliding surface, are the parameters governing the average post-peak compressive strain Hc of the specimen (Fig. 3). Referring to the specimen depicted in Fig. 3a, post peak strains can be defined by the following equation [8]: Hc

H c , el

w H

H c1

'V c w Ec H

(1)

where, Hc1 = strain at compressive strength fc ; 'Vc = stress decrement after the peak; H = height of the specimen (see Fig. 1b).

Figure 3. The post-peak response of quasi-brittle materials in compression.

13-The ductile.qxd

9/19/2009

124

11:59 AM

Page 124

Alessandro P. FANTILLI, Hirozo MIHASHI, Paolo VALLINI and Bernardino CHIAIA

According to test measurements [8, 9], the post-peak slope of Vc-Hc increases in longer specimens (Fig.3b), due to the w/H ratio involved in the evaluation of Hc [Eq.(1)]. The stress decrement 'Vc can be defined as: 'V c

fc V c

f c >1 F w @

(2)

where, F(w) = non-dimensional function which relates the inelastic displacement w and the relative stress Vc / fc during softening (Fig.3c); fc = compressive strength (assumed to be positive). Substituting Eq.(2) into Eq.(1), it is possible to obtain a new equation for Hc : Hc

H c1

f c >1 F w @ w Ec H

for Hc > Hc1

(3)

Eq.(3), adopted for the post-peak stage of a generic cement-based material in compression, is based on the definition of F(w), which has to be considered as a material property [8-9]. In all cement-based composites, this function should be evaluated experimentally on cylindrical specimens, as performed by Jansen and Shah [9] for plain concrete (Fig.3c). Fig.4a shows the F(w) relationships proposed by Fantilli et al. [10]. It consists of two parabolas and a constant branch: F w

V

F w

V

F w

V

fc fc

fc

1 a w2 b w

for 0 d w d

§ b 2 ·¸ §¨ 4 a 2 2 4 a ·¸ ¨¨1 ¸ ¨ 2 w b w¸ © 4a ¹ © b ¹

for

0

for

b 2a

b b wd 2a a

w!

b a

(4a) (4b) (4c)

The parabolas are both defined by the same coefficients a, b and have the same extreme point at w = -0.5 b/a , whereas w = - b/a (i.e. twice the value at extreme point) is considered the maximum inelastic displacement corresponding to F(w) higher than zero.

Figure 4. The stress-inelastic displacement relationship proposed by Fantilli et al. [10].

13-The ductile.qxd

9/19/2009

11:59 AM

Page 125

The ductile behavior of HPFRCC in compression

125

In the case of the plain concrete specimens, the values a = 0.320 mm-2 and b = -1.12 mm-1 were obtained by means of the least square approximation of several tests [10]. As observed in Fig.4b, the curves defined by Eqs.(4) fall within the range of the data experimentally measured by Jansen and Shah [9]. In the case of multi-axial compression, stress-inelastic displacement relationships, which should reproduce the confined post-peak stage, cannot be found in the existing literature. As is well known, two types of confinement, namely passive and active, can be produced. In compressed columns, passive confinements provided by transversal reinforcement (i.e., stirrups, tubes, strips, spirals, etc.), are only activated by concrete displacements. Thus, to define quantitatively this contribution, it is necessary to know the stress-transversal displacement relationship of concrete. Active confinement is due to external stresses V3 applied by multi-axial compression tests on cubes in two or three directions, or by triaxial tests on cylinders (see the book by van Mier [8] for a review). Only a single campaign of triaxial tests, performed by Jamet et al. [11] on micro-concrete, is reported in the current literature. In that case, the applied confinement was relatively high (V3 >3 MPa), if compared to those produced by stirrups in ordinary RC columns. In accordance with Eurocode 2 [4] (§ 9.5.3), in columns under concentric compression, transverse reinforcement can develop about V3 = 1MPa [12]. Consequently, with the aim of analyzing the equivalent confing pressures produced by high-performance concretes, the comparison between the results of new triaxial tests on NC and SC cylinders and the mechanical response of HPFRCC under uniaxial compression are reported.

EXPERIMENTAL PROGRAM The post-peak behaviour of cement-based composites under multi-axial compression has been investigated at the Department of Structural and Geotechnical Engineering of Politecnico di Torino (Italy) by means of triaxial tests on concrete cylinders (Fig.5a). The experimental equipment, named HTPA (High Pressure Triaxial Apparatus) and described by Chiaia et al. [13], is generally used to test cylindrical specimens made of soft rocks. Each triaxial test consists of two stages. A specimen is initially loaded with a hydrostatic pressure ı3 (Fig.5b), then deviatoric loads P are applied along the longitudinal direction with a velocity of 0.037 mm per minute (Fig.5c). During the second stage of loading, the confining pressure V3 = const. is applied to the lateral surface, whereas the longitudinal nominal stress Vc becomes: Vc

V3

4P

S D2

(5)

where, P = applied deviatoric load; D = diameter of the cross-section. Through a couple of LVDT, local longitudinal displacements, and therefore nominal longitudinal strains Hc , are also measured (Fig.5a). Two confining pressures, namely ı3 = 0 MPa and ı3 = 1 MPa (reached in 10 minutes), are applied to the specimens. During the application of hydrostatic loads (Fig.5b), stress increments are electronically recorded every 10 seconds. Similarly, in the second stage, when ı3 = const. and P increases, the values of deviatoric load, the relative displacement between the specimen’s ends, and the longitudinal displacement along the lateral surface (taken by the LVDTs of Fig.5a) are measured.

13-The ductile.qxd

126

9/19/2009

11:59 AM

Page 126

Alessandro P. FANTILLI, Hirozo MIHASHI, Paolo VALLINI and Bernardino CHIAIA

Figure 5. The two stages of triaxial tests on cement-based cylinders.

Two types of self-consolidating concrete (SC_mix1 and SC_mix2) and a single ordinary concrete (NC) were tested. Their compositions are reported in Table 1. Specifically, the selfconsolidating concretes have the same unit weight, but different amounts of aggregates. With respect to SC_mix1, in a cube meter of SC_mix2 the content of carbonate filler was increased by 90 N and, contemporarily, the weight of coarse aggregate was reduced by the same quantity. Regarding the high performance fiber-reinforced cementitious concrete, two specimens were tested, under uniaxial compression (ı3 = 0), at the Department of Architecture and Building Science of Tohoku University (Japan). HPFRCC is a hybrid concrete, whose reinforcement consists of both polyethylene fibers and steel cords [6]. The composition of such innovative composite is shown in Table 2, where W/B = water-binder ratio, SF/B = silica fume-binder ratio, S/B = sand-binder ratio; SP/B = superplasticizer-binder ratio. In the same table, the volume fraction Vf of polyethylene fibers PE (diameter ĭ = 12ȝm and length L = 6mm) and of steel cords SC (diameter ĭ =0.33 mm and length L = 32mm) are also reported. Table 1. Composition and strength of NC, SC_mix1 and SC_mix2. NC 3

SC_mix 1 3

SC_mix 2

Component

N/m

N/m

N/m3

Water Superplasticizer (Addiment Compactcrete 39/T100) Superplasticizer (Addiment Compactcrete 39/T11) Cement (Buzzi Unicem II/A-LL 42.5 R) Carbonate filler (Nicem Carb VG1-2) Fine aggregate (0-4 mm) Coarse aggregate (6.3-12 mm) Cubic strength -MPa-

1770

1770

1770

-

44

44

14

-

-

2840

2450

2450

0 8830 6380 30.0

3240 8930 6380 31.1

3730 8930 5890 30.4

13-The ductile.qxd

9/19/2009

11:59 AM

Page 127

127

The ductile behavior of HPFRCC in compression

Table 2. Composition of HPFRCC [6]

W/B weight (%) 45

SF/B weight (%) 15

S/B weight (%) 45

SP/B weight (%) 0.9

PE Vf (%) 1

SC Vf (%) 1

Table 3. Mechanical and geometrical properties of the specimens tested in uniaxial and triaxial compression

Specimen NC0 NC1 SC0 SC1 SC0b SC1b HC0 HC0b

H (mm) 140 140 140 140 140 140 100 100

D (mm) 70 70 70 70 70 70 50 50

Type of concrete NC NC SCC mix 1 SCC mix 1 SCC mix 2 SCC mix 2 HPFRCC HPFRCC

V3 (MPa) 0 1 0 1 0 1 0 0

The specimens of each concrete mixture were cast simultaneously in polystyrene form, then cured for one week under identical laboratory conditions, and finally tested after one month. Three couples of cylinders, with H=140 mm and D=70 mm, were made of NC (NC0 and NC1), SC_mix1 (SC0 and SC1), and SC_mix2 (SC0b and SC1b). The two specimens of these couples were tested, respectively, at ı3 = 0 MPa and ı3 = 1 MPa. Two HPFRCC cylinders, with H=100 mm and D=50 mm, were tested in uniaxial compression. The properties of each specimen are reported in Table 3.

TEST RESULTS Figure 6 reports the stress-strain relationships obtained from the specimens made respectively with HPFRCC (Fig.6a), normal concrete (Fig.6b) and self-consolidating concrete (Fig.6c). In all the cases, after the peak stress fc , a remarkable strain softening branch can be observed in the Vc-Hc diagrams. The highest compressive strength is achieved in the high-performance concrete. However, HPFRCC does not behave in a brittle manner under compression, as usually occurs in high-strength concrete specimens [12]. As a matter of fact, the post peak response of high performance concrete appears more ductile than NC and SC.

Figure 6. The stress-strain relationships of HPFRCC, NC and SC.

13-The ductile.qxd

128

9/19/2009

11:59 AM

Page 128

Alessandro P. FANTILLI, Hirozo MIHASHI, Paolo VALLINI and Bernardino CHIAIA

Only when the confining pressure V3 increases, does the post-peak response of NC and SC appear more ductile. The higher the confinement, the higher the values of fc and Hc1 , which are reported, together with Young’s modulus Ec , in Table 4. By comparing all the post-peak branches reported in Fig.6, it seems that the ductility of SC and NC specimens in the presence of V3 =1 MPa is more or less the same of HPFRCC without any confinement. However, a direct comparison between the analyzed concretes is not possible in terms of nominal stress and strain, because specimens are different in length and have different nominal strengths. Table 4. Mechanical properties measured in the tests.

Specimen 0NC0 0NC1 0SC0 0SC0b 0SC1 0SC1b HC0 HC0b

fc (MPa) 19.4 30.5 20.1 23.2 36.4 32.0 48.0 41.5

Hc1 (%) 0.293 0.473 0.479 0.372 0.604 0.696 0.551 0.401

Ec (MPa) 24000 23000 17000 23000 19000 27000 16700 15100

Post-peak comparison in terms of F(w) A more accurate comparison between the post-peak responses of HPFRCC, NC and SC under compression can be conducted in terms of F(w) (Fig.7). In particular, for a given Hc ! Hc1, the decrease of compressive stress 'Vc = fc-Vc (and F = Vc / fc ) can be obtained through the Vc-Hc diagrams experimentally evaluated (Fig.6), whereas the corresponding w (Fig.3a) can be obtained from Eq.(3) (fc , Hc1 , Ec and H are known from the tests). The F(w) curves reported in Fig.7 are limited to w = 2mm, when compressive strains Hc are relatively high although, in some cases, stresses are higher than zero. However, in all the tests the relative stress F = Vc / fc decreases with w. The dashed curves reported in Fig.7 represent the behaviour of NC and SC as predicted by Eq.(4) in the case of zero confinement.

Figure 7. The post peak behaviour in terms of F(w)

13-The ductile.qxd

9/19/2009

11:59 AM

Page 129

The ductile behavior of HPFRCC in compression

129

Figure 8. The active confinement of HPFRCC.

As in the case of V3 = 0 the post-peak responses of the specimens NC0, SC0, SC0b are correctly predicted by Eq.(4), and all the tests can be considered consistent [10]. Both for NC and SC, Fig.7a and Fig7b, respectively show the increase of the compressive fracture toughness (within the range wa0-2 mm) with the confining pressure V3. However, this phenomenon is also evident in the case of HPFRCC, which can show, in absence of confinement, more or less the same F(w) obtained for NC and SC when V3 = 1MPa. Fig.8a shows the post-peak responses of the specimens HC0 and HC0b, which are closer to those of confined SC and NC (i.e., the range defined by NC1, SC1, SC1b), than to the theoretical F(w) obtained in absence of confinement [10] (the dashed line in Fig.8a). Within the observed range (wa0-2 mm), compressive facture toughness of different concretes can be objectively measured the by area AF under the function F(w): 2

AF

³0 F w dw

(6)

In fact, as F(w) is a relative stress normalized with respect to the compressive strength fc , a comparison between all the cement-based composites, under uniaxial and multi-axial compression, is possible. Higher values of AF are attained in concretes capable of maintaining high loads after failure (i.e., in the case of ductile materials). Obviously, the maximum ductility AF,max = 2mm is reached in the case of plastic behaviour [F(w) = 1= const.]. The areas AF computed by Eq.(6) for the tested specimens (Table 4) are also reported in the histogram of Fig.8b. In all cases, AF is between AF,max = 2 mm and the lower limit AF,min = 0.61 mm, corresponding to the normal and self-consolidating concretes without any confinement (Fig.8b). To be more precise, AF,min is obtained by substituting Eqs.(4) (with a = 0.320 mm-2 and b = -1.12 mm-1 ) into Eq.(6). At V3 = 1MPa, for the specimens made of SC and NC (i.e., NC1, SC1, SC1b) the values of AF range between 1.39 mm and 1.46 mm (Fig.8b), and do not differ substantially from those measured for HPFRCC (AF #1.31 mm) without confinement.

CONCLUSIONS From the results of an experimental campaign performed on NC, SC and HPFRCC cylinders under uniaxial and multi-axial compression, the following conclusion can be drawn:

13-The ductile.qxd

130

9/19/2009

11:59 AM

Page 130

Alessandro P. FANTILLI, Hirozo MIHASHI, Paolo VALLINI and Bernardino CHIAIA

x In normal and self-consolidating concrete, fracture toughness in compression increases in the presence of an active confinement. x HPFRCC specimens, which show strain hardening in tension, provide a very ductile behaviour in compression, despite the higher strength. x During the post-peak stage, the ductility of HPFRCC is comparable with that of NC or SC at 1MPa of confining pressure. x In compression, the performance of fiber-reinforced composites can be quantified by the distributed confining pressure generated by the fibers. The presence of an active confinement can improve the mechanical behaviour of concrete and, consequently, its durability. Thus, further researches should be developed in order to introduce new sustainability indexes, which take into account fracture toughness, both in tension and compression.

REFERENCES 1.

Mansur M.A., Chin M.S., Wee T.H., Flexural Behaviour of High-Strength Concrete beams. ACI Structural Journal, 94(6), 1997, 663-674. 2. Khayat K.H., Paultre P., Tremblay S., Structural Performance and In-Place Properties of Self-Consolidating Concrete Used for Casting Highly Reinforced Columns. ACI Materials Journal, 98(1), 2001, 371-378. 3. UNI EN 1998-1:2005. Eurocodice 8 – Design of structures for earthquake resistance Part 1: General rules, seismic actions and rules for buildings, pp.229 4. UNI EN 1992-1-1:2005. Eurocodice 2- Design of concrete structures- Part 1-1: General rules and rules for building, pp. 225. 5. Ganesan N., Ramana Murthy J. V., Strength and Behaviour of Confined Steel Fiber Reinforced Concrete Columns. ACI Materials Journal, 87(3), 1990, 221-227. 6. Kawamata A., Mihashi, H., Fukuyama, H., Properties of Hybrid Fiber Reinforced Cement-based Composites. Journal of Advanced Concrete Technology, 1(3), 2003, 283290. 7. CEB (Comite Euro-International du Beton), “CEB-FIP Model Code 1990”, bulletin d'information n°203-205, Thomas Telford, London, UK, 1993. 8. van Mier, J. G. M., Fracture Processes of Concrete: Assessment of Material Parameters for Fracture Models. CRC Press, 1996, pp. 448. 9. Jansen, D. C., Shah, S. P., “Effect of length on compressive strain softening of concrete”, ASCE Journal of Engineering Mechanics, 123(1), 1997, 25-35. 10. Fantilli A.P., Mihashi H., Vallini P., Post-Peak Behaviour of Cement-Based Materials in Compression. ACI Materials Journal, 104(5), 2007, 501-510. 11. Jamet P., Millard A., Nahas G., Triaxial behaviour of a micro-concrete complete stressstrain curves for confining pressures ranging from 0 to 100 MPa. International conference on concrete under multiaxial conditions, Toulouse 1984, 133-140. 12. Foster S. J., Liu J., Sheikh S. A., Cover Spalling in HSC Columns Loaded in Concentric Compression. ASCE Journal of Structural Engineering, 124(12), 1998, 1431-1437. 13. Chiaia B., Fantilli A. P., Vallini P., Post-peak response of confined SCC. In 3rd North American Conference on the Design and Use of Self-Consolidating Concrete (SCC), Chicago 2008.

14-Investigation of.qxd

9/19/2009

11:59 AM

Page 131

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

INVESTIGATION ON RANDOM DISTRIBUTION OF FIBRES IN CEMENT COMPOSITES Tomasz PONIKIEWSKI The Silesian University of Technology, Department of Building Processes Akademicka 5, 44-100 Gliwice, Poland, e-mail:[email protected]

ABSTRACT The problem of homogenous distribution of fibres in a cast element always occurs with application of concrete mixes modified by fibres. Past research has proven that the size of an element influences the direction of fibres in the concrete mix. If the depth of a given piece is not much greater than the fibres’ length, in the majority of cases they are positioned horizontally. The lack of regularity and directional reorientation of fibres during technological processes causes some difficulties, linked with expected random distribution in the concrete mass. The analysis of the impacted this randomness on workability and strength parameters of concrete is investigated for normal and self-compacting concretes. In this paper an analysis is presented in which cross-sections of concrete beams with steel fibres are closely examined.

Keywords steel fibres, rheology, self-compacting concrete, workability

INTRODUCTION The analysis is carried out using special software for based on images of polished microsections of fibre-concretes elaborated in Master Thesis [1]. The analysis of fibre distribution was performed on the matrix of a hardened specimen. A method of investigating fresh concrete mixes with added fibres is presently being developed. The main feature of application of steel, polypropylene and other fibres in concrete mixes has already been dealt with in earlier publications [2], [3]. The general propensity to improve the properties of hardened concrete, both normal and self-compacting, by increasing volume of fibres, leads at the same time to worsening its workability during the process of formation, causing, among the others, a clustering effect and formation of so called “balls”.

ASSUMPTIONS AND METHODOLOGY OF RESEARCH The software process ‘Quest for Fibre’ [1] for the analysis of micro-sections of fibreconcretes is based on experimental analysis of cross-sections of cores or other specimens cut from tested structural parts. By calculating the traces of fibres in cross-sections, credible information about the actual structure of the dispersed reinforcement in a given fibre-concrete piece can be obtained. An example of the distribution of points illustrating traces of fibres in a cross-section prepared for computer-aided analysis is in Fig. 1a from [4], whereas Fig. 1b shows the result achieved by the ‘Quest for Fibre’ software [1]

14-Investigation of.qxd

132

9/19/2009

11:59 AM

Page 132

Tomasz PONIKIEWSKI

Fig. 1 Graphic presentation of fibre marks on an image of a cross-section, prepared for computer-aided analysis: from [5] (a), and from original investigation (b). Within this project formulas were also presented for determining the quantity of fibres per unit of cross-section area in relation to the distribution of fibres. These formulas are [6]: for 1D distribution: N1D = 4Vt/d2 for 2D distribution: N2D = 8Vt/2 2d2 for 3D distribution: N3D = 2Vt/d2 where: Vt – percentage of fibre volume, d – fibre diameters. The values obtained from ideal structures and from those where the fibres were calculated on images of real cross-sections could be compared with each other, enabling to estimate the actual volume of dispersed reinforcement and the formed fibre structure, [4]. DESCRIPTION OF THE SOFTWARE FOR ANALYSIS OF IMAGES OF MICRO-SECTIONS In order to analyse effectively the image of concrete containing dispersed fibres, a special software was elaborated, whose accuracy depends entirely on the resolution of the image photography used. The task is to locate all cross-sections of fibres and determine their position. This computer program is written in Microsoft Visual BasicҘ (for 32-bit Windows Development, version 5.0). The process for sample preparation is presented in Fig. 2. By this way images of concrete samples are prepared, which in turn are subjected to computer analysis. The software processes a matrix whose dimensions are determined by photographic image resolution. Each cell of this matrix contains information on the colour identified at that point on the photo. The program operates step by step generating a virtual matrix (later referred to as the colour matrix), with values representing colours. The colours are documented in the format RGBҘ (Red Green Blue). This software simplifies the colours, replacing real colours by degrees of gray. At the same time, in a simple but effective way the contrast is improved, which is essential for the next phases of the program operation. Now the cells of this matrix contain numbers in a range 0 to 255, which represent colour in a scale of greyness.

14-Investigation of.qxd

9/19/2009

11:59 AM

Page 133

Investigation of random distribution of fibres in cement composites

133

Fig. 2. Phases of micro-section preparation and a specimen prepared for analysis [1] The next step of image investigation is elimination of cells with smaller values than those determined by the user. After this procedure the matrix of colours becomes a logical matrix, with cell values 0 or 1 only, in this way determining whether or not in a given spot theres a piece of detected fibre. Since the exchange of real colours to a gray scale inevitably implies some information loss, the next step is to fill in the fibre contours in order to get more rounded shapes. During this phase a new matrix is generated, describing the quantity of fibres and data for each. The data include: geometric centre, dimensions (presently by projecting only on axis OX and OY), and area. In the next step the program eliminates the fibres whose area is too small or too large for positive identification. This is justified by the fact that the software also recognises aggregate of colour shades close to white. Next, it shows all identified fibres on the background of the original graphics, where hand correction is possible. After correction is finalized, usually 3-5 fibres are eliminated from the scope of all detected, the program moves to the counting phase of counting. By rotating the position of each detected fibre 180 degrees around the centre of the system of coordinates, and projecting them on the axis OX, the extreme points for each fibre are found. A new geometric centre is calculated based on these new fibre dimensions, a precise localization in the sample. This software also determines the angle at which a given fibre is cut, based on the ratio of minimal and maximal radius. Unfortunately, determining this relation is to a large extent dependent on the quality of the initially inputted graphic image. The next phase is to display the results of the image analysis. This program provides three kinds of interpretation for fibre distribution: x a tabled display of fibre numbers with the angles at which they are cut, x a general distribution of fibres in a sample, x a detailed distribution of fibres in a sample, i.e. enabling one to determine how to crosssection the sample , into anywhere from 2 to 32 sections. This gives a division of a sample into 4 to 1024 fields, wherein the fibre quantities are counted. The last phase of work generates a report from the investigation. Namely, after investigation is completed on a few micro-sections of a given element, the description of fibre distribution in the whole element is possible, as the software has an appendix enabling such an analysis. After listing the cross-sections in an order corresponding with their real position,

14-Investigation of.qxd

9/19/2009

11:59 AM

Page 134

134

Tomasz PONIKIEWSKI

the program proceeds with calculations. The results in turn can then be analysed by the user. Presentation results falls into two categories: local and global analysis. The local analysis presents the results of a singular cross-section analysis. In a similar way as analysing the results from the main program, one can divide a sample into 2 to 32 parts. This division may be elected into fields, rows or columns. The user can select the form of result presentation, from text-only format to a colourful representation of fibre quantities in the tested field. The global analysis presents the results of the whole element, consisting of all earlier mentioned cross-sections. As in the local analysis, the user can select the degree of subdivision, as well as select the method of result presentation. METHOD OF THE COMPUTER-AIDED ANALYSIS The results, presented below are based on investigation of the whole concrete element (Fig.2), consisting of 9 sections. The full report generated by the computer program is very large, so only a local analysis of 1 micro-section is presented here. Fig. 3 shows a pictorial presentation of the results of a fibre count in a selected cross-section of concrete divided into 16 x 16 squares. On this basis we determine the real distribution of fibres per unit of area in the sample cross-section.

Mean value: 0.35 Standard deviation: 0.52 Number of fibres in a cross-section: 91 Real size of the mesh field: 0.93 cm

Minimum

Maksimum

Minimal amount of Maximal amount of fibres: 0 fibres: 3 Fig. 3. Example of cross-section for the division 16 x 16 and analysis of the number of fibres Fig. 4 is a pictorial presentation of the results of the fibre count in a selected cross-section of concrete with a division mesh 1 x 16 (rows). On this basis we can determine the summary distribution of fibres in horizontal cross-sections of the concrete sample, describing their local and global 3-dimensional concentration in a horizontal direction. Fig. 5 is the pictorial presentation of the results of the fibre count in a selected crosssection of concrete with a division mesh 1 x 16 (columns). On this basis we can determine the summary distribution of fibres in vertical cross-sections of the concrete sample, describing their local and global 3-dimensional concentration in a vertical direction. The application of global scale to a substantial extent decreases the transparency of result presentation, because depending on the degree of divisions, the number of fibres in a sampled field fluctuated from 0 to 53. The global results are obtained by generating diagrams illustrating the number of fibres in cross-sections and in the whole concrete structural element (Fig. 6).

14-Investigation of.qxd

9/19/2009

11:59 AM

Page 135

135

Investigation of random distribution of fibres in cement composites

Mean value: 5.68 Standard deviation: 2.80 Number of fibres in a cross-section: 91 Real size of the mesh field: 0.93 cm

Minimum

Maksimum

Minimal amount of Maximal amount of fibres: 1 fibres: 13 Fig. 4. Example of a cross-section for the division 16 x 16 and analysis of the number of fibres, data from the rows

Mean value: 5.68 Standard deviation: 3.64 Number of fibres in a cross-section: 91 Real size of the mesh field: 0.93 cm

Minimum

Maksimum

Minimal amount of Maximal amount of fibres: 1 fibres: 11 Fig. 5. Example of cross-section for the division 16 x 16 and analysis of the number of fibres, data from the columns

Fig. 6. Diagram illustrating the number of fibres in global analysis In Figs. 7 - 9 there are presented results for selected 3 out of 9 cross-sections of a tested fibreconcrete element with a division into 32 x 32 parts.

14-Investigation of.qxd

9/19/2009

11:59 AM

Page 136

136

Tomasz PONIKIEWSKI

Mean value: 0.46 Standard deviation: 0.47 Number of fibres in a cross-section: 478 Real height of the mesh field: 0.46 cm Real width of the mesh field: 0.46 cm

Minimum

Maksimum

Minimal amount of Maximal amount of fibres: 1 fibres: 4 Fig. 7. Cross-section 1 for the division 32 x 32 and analysis of the number of fibres

Mean value: 0.33 Standard deviation: 0.43 Number of fibres in a cross-section: 338 Real height of the mesh field: 0.46 cm Real width of the mesh field: 0.46 cm

Minimum

Maksimum

Minimal amount of Maximal amount of fibres: 1 fibres: 3 Fig. 8. Cross-section 2 for the division 32 x 32 and analysis of the number of fibres

Mean value: 0.32 Standard deviation: 0.36 Number of fibres in a cross-section: 330 Real height of the mesh field: 0.46 cm Real width of the mesh field: 0.46 cm

Minimum

Maksimum

Minimal amount of Maximal amount of fibres: 1 fibres: 3 Fig. 9. Cross-section 3 for the division 32 x 32 and analysis of the number of fibres On the basis of the presented selected simulations of fibre distribution in the tested specimen, some general tendencies can be concluded as far as the level of their dispersion and of the homogeneity of the modified concrete matrix. It is proved that the fibres are more

14-Investigation of.qxd

9/19/2009

11:59 AM

Page 137

137

Investigation of random distribution of fibres in cement composites

or less evenly distributed, regardless of the direction in which the specimen was cast. No concentrations of fibres were observed; either did the so called “balls” occur. One can observe the position of fibres, and the possible interference near the form walls. Based on the elaborated program for analysing fibres distribution, an initial investigation was carried out for mortar strength and how it depends on fibre quantity per unit. For investigated mortars modified by steel fibres of 6mm and 12 mm long, some initial results were obtained. A significant increase of compressive strength was detected in the case of mortars with 6 mm long fibres. It was only in this case that an increase of compression strength was observed together with an increase of dispersed reinforcement in the tested cross-sections. For mortars with 12 mm long fibres, the strength fc changes insignificantly, not proportional to the fibre quantity in the cross-sections. So, it would seem justified to apply smaller steel fibres in greater density rather than the 12 mm fibres. No influence of fibre quantity on the parameter fct,f1 was proven in tested mortars. However, it should be taken into consideration what type and size of structural members are produced, as well as which other processes which could have influence on the mechanical parameters of fibreconcrete (such as mixing method, casting procedures, nurturing of the concrete). 50 45

fc, fct,fl [MPa]

40 35 30 25 20 15 10 5 0 60

140

230

340

370

Number of fibres in 3 sections [pcs.] compressive strength fc for concrete with 6 mm fibres compressive strength fc for concrete with 12 mm fibres tensile strength at bending fct,fl for concrete with 6 mm fibres tensile strength at bending fct,fl for concrete with 12 mm fibres

Fig. 10. Influence of the number of fibres on mechanical properties in cross-section [1]. CONCLUSIONS Basing on own research, with application of the elaborated method and software for analysis fibre distribution in mortars and concretes, some initial results were obtained in the undertaken realm of investigation. 1. The results of the analysis of the images did not reveal any tendencies for fibres to adhere to the form walls, regardless of the type of fibres. The phenomenon called “the wall effect” is

14-Investigation of.qxd

9/19/2009

11:59 AM

Page 138

138

Tomasz PONIKIEWSKI

evidently not connected with the shape of the fibres; noticeably smaller was the quantity of steel fibres in direct proximity to the form walls, regardless of their shape and percentage in the total concrete mix. 2. The results of the image analysis revealed no effect of subsidence of fibres in the concrete mix, regardless of their mix percentage, considered in the investigation. The fact that majority of fibres were oriented parallel to the longer walls of the form, allows one to conclude that this is a result of the direction of casting and fibres flowing into the form. 3. Moreover, less regular orientation of fibres occurred in the micro-sections next to ends of the tested element, which reconfirms the relation between orientation of fibres and the direction of mix casting. Initially, the existence (or lack) of some relation between the quantity of fibres in concrete and its mechanical properties (strength) was proven. 4. The testing of tensile strength at bending fct,f1 confirmed the principle of sliding or breaking fibres from the matrix. A visible small increase of bending strength was noticed in the case of 6 mm long steel fibres. 5. Analysis of cross-sections from one sample points to vast differentiation of the reinforcement quantities in the tested micro-sections. The tests carried out on mortars with straight steel fibres, 6 mm and 12 mm long, point to a high degree of difficulty in designing a self-compacting mix, characterized by high formability and also good strength parameters. 6. The maximal, tested participation of fibres in the mix - 0,5%, is not causing any significant improvement of strength in comparison to 0,1%. Such small differences in the mechanical parameters of tested fibre-concretes may be the result of this kind of dispersed reinforcement, and not the wide range in the quantity of fibre reinforcement in mortar. The distribution and quantity of fibres in the analysed cross-sections are not providing any definite answer as to their influence on the strength parameters. It would seem justified to perform similar investigations with steel fibres of other kind (e.g. hooks) and with higher volume intensity (Vf >0,5%) in mortar with a higher number of analysed cross-sections. REFERENCES [1] Potysz P., The computer analysis of cross-section of fibre reinforced concrete; Quest for

[2]

[3]

[4]

[5] [6]

Fibre software (in Polish)Graduate dissertation, Supervisor: Tomasz Ponikiewski, Gliwice 2006 Barragán B., Zerbino R., Gettu R., Soriano M., de la Cruz C., Giaccio G., Bravo M.: Development and application of steel fibre reinforced self-compacting concrete, 6th RILEM Symposium on Fibre-Reinforced Concretes (FRC) – BEFIB 2004, Varenna, Italy, 457 – 466 Ponikiewski T., The rheological properties of fresh steel fibre reinforced self-compacting concrete, in: Proc. Int. Symp. `Brittle Matrix Composites 8`, A.M.Brandt, V.C.Li, I.H.Marshall, Warsaw 2006 Brandt A.M, Kasperkiewicz J, eds.: Diagnosis of Concretes and High Performance Concrete by Structural Analysis (in Polish), Institute of Fundamental Technological Research, Polish Academy of Sciences, Warsaw 2003 Brandt A.M, Cement-based Composites: Materials, Mechanical Properties and Performance, E & FN Spon, London 1995, pp. 470 Kasperkiewicz J, Internal structure and cracking process in brittle matrix composites (in Polish), Institute of Fundamental Technological Research, Presses of Polish Academy of Sciences, vol. 39, pp. 239, Warsaw 1983

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 139

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

PERMEABILITY OF SFRCC BASED ON FINE AGGREGATE AFTER PRE-LOAD CYCLES Jacek KATZER Technical University of Koszalin, Laboratory of Building Engineering ĝniadeckich 2, 75-456 Koszalin, Poland, e-mail: [email protected]

ABSTRACT This research concerns steel fibre reinforced cement composites (SFRCC) based on fine aggregate modified by up to 25% of silica fume and reinforced by up to 2% (by volume) of steel fibre. The composite was modified by replacing cement by silica fume and adding steel fibres. All examinations were carried out after 120 days of curing. The received results of the research on ultimate compressive strength, flexural strength and permeability are presented. There was noted a considerable improvement in compression strength and watertightness of the examined composite mixes. The subject of the analysis of the research results was to define the relation between flexural strength of the examined cement composites and effective spacing of fibres.

Keywords Cement composite, fine aggregate, silica fume, steel fibre, permeability

INTRODUCTION Cement composites are a large family of structural and civil engineering materials different with mechanical properties, used raw materials and applied technology of production. The versatility of applications of cement composites influences a wide range of their properties being designed and tested [1]. There is a growing need for new cement composites of high durability, resistance to weather conditions and characterized by both high static and dynamic strength [2]. This need causes an intensive development of cement composites modified by a wide range of admixtures or additions, including fibres of different types and origin. Steel fibre reinforced cement composites (SFRCC) belong to a group of these composites which are being developed very fast and because of their high flexural strength, dynamic strength, and fatigue strength are a very promising structural material of the future, [3, 4]. Static and quasi-static properties of SFRCC are already well-known and described, [5, 6, 7, 8]. There were also carried out some interesting experimental programmes concerning dynamic properties of such composites, [9, 10]. However, there is still a lack of information about the permeability of SFRCC, which directly influences its durability. The examination of permeability of cement composite specimens does not reflect permeability of cement composite in an actual structure. Each, even the most basic structure, works within constantly changing conditions, where loads appear and disappear in time. The influence of the changeability of static loads in time on cement composite parameters has already been well known, especially as far as ultimate compressive strength is concerned [11]. Bearing this in mind a research programme of SFRCC examined after a pre-load cycles was initiated in the Laboratory of Building Engineering at Technical University of Koszalin.

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 140

140

Jacek KATZER

RESEARCH PROGRAMME The research programme of SFRCC was prepared on the basis of local fine aggregate, raw silica fume, steel fibre, and a superplasticizer. In the local Pomeranian pit deposits, fractions from 0 to 4.0 mm constitute 97% by weight of all aggregate [12]. Approximately half of documented deposits are constituted by deposits hydroclassified during the exploitation. This sand is a by-product of hydroclassification of natural all-in-aggregate. During the process of hydroclassification, all-in-aggregate is divided into gravel and sand. There is a deficit of gravel in the region, therefore gravel received during hydroclassification of all-in-aggregate is constantly being sold. Fine aggregate received during the same process, due to its excessive amount, is stored on continuously growing waste heaps. Fine aggregate used in this research study was obtained from aggregate pit in SĊpólno Wielkie (Figure 1). This aggregate has a lower grain-size distribution, a smaller amount of stone dust and a higher content of minerals and crystal rocks than the pit sand obtained from the same mine. In the Figure 1 fine aggregate used during the research programme is shown. White grains are quartz and black grains are granite. This aggregate was thoroughly described in previous publications [5, 12]. The SFRCC matrix was characterized by water/cement (w/c) ratio equal to 0.5. The contents of the mix was as follows: fine aggregate = 1760 kg/m3, c = 400 kg/m3, w = 200 g/m3. Harnessing fine aggregate matrix to prepare fibre reinforced cement composite enables even spacing of fibres in the whole volume of the composite mix while eliminating problems related to proportioning with more than 1.5 % of fibres [1, 3, 6]. The hooked steel fibre of a length equal to 50 mm and circular cross-section, with aspect ratio l/d = 50 and breaking strength of 1100 MPa is presented in Figure 2. These fibres were used to compose SFRCC. All mixes were modified by a constant amount of plasticizer equal to 1.8% (by mass of cement). The superplasticizer was based on sulphonated naphthalene formaldehyde condensates and codified as SNF. A rotary drum mixer was utilized to make the mixtures. There were applied moisture curing and manufacturing techniques as described by Toutanji & Bayasi [13]. The first stage of the research programme covered the testing of ultimate compressive strength fc, flexural strength ff, and deflection db of a beam. The second stage covered the testing of permeability and permeability after a cycle of loads.

Figure 1. Fine aggregate obtained from aggregate pit in SĊpólno Wielkie

Figure 2. Magnified (125 times) ending of the steel fibre

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 141

141

Permeability of SFRCC based on fine aggregate after pre-load cycles

EXPERIMENTAL DESIGN Tests were carried out with the help of an experimental design and the so-called “central composite experimental design” presented in Figure 3 and summarised in Table 1 was used. The subject of the research was defined as complex material whose interior structure was unknown and unavailable for an observer, [14, 15, 16]. The observer knew the input parameters, which were the addition of steel fibre - Vf, and addition of silica fume – SF. The output parameters were properties of cement composite. The examination results were statistically processed, and values bearing the gross error were assessed on the basis of Grubbs criterion. The objectivity of the experiments was assured by the choice of the sequence of the realization of specific experiments from a table of random numbers. All calculations connected with specifying an equation regress factor and graphic interpretation of the received model was carried out with the help of “Statistica 8.0” computer programme [17].

Figure 3. The scheme of the arrangement of measuring points Table 1. Experimental design values of input factor Number of experiment 1 2 3 4 5 6 7 8 9,10,11

Number of realization 2 5 10 9 1 8 3 4 7,6,11

Input factor in code values x1 -1.00000 -1.00000 1.00000 1.00000 -1.68179 1.68179 0.00000 0.00000 0.00000

x2 -1.00000 1.00000 -1.00000 1.00000 0.00000 0.00000 -1.68179 1.68179 0.00000

in natural values SF % Vf % 5.0 0.4 5.0 1.6 20.0 0.4 20.0 1.6 0.0 1.0 25.0 1.0 12.5 0.0 12.5 2.0 12.5 1.0

Eleven cement composite mixes were modified by the silica fume (replacement of cement from 0% to 25%), addition of steel fibre (from 0% to 2%) and by admixture of superplasticizer. From each cement composite mix 24 specimens were made - 6 cube

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 142

142

Jacek KATZER

specimens 150 mm · 150 mm · 150 mm, 12 prism specimens 100 mm · 100 mm · 500 mm and 6 cylinder specimens I 150 mm, h = 500 mm. The research programme covered the examinations of composite properties after 120 days of curing because of high silica fume reactivity in time.

RESULTS OF THE RESEARCH The results of the ultimate compressive strength examination are shown in Figure 4.

Figure 4. Ultimate compressive strength fc

Figure 5. Flexural strength ff

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 143

Permeability of SFRCC based on fine aggregate after pre-load cycles

143

The ultimate compressive strength examination was carried out on cube specimens 150 mm · 150 mm · 150 mm. Both, silica fume and steel fibre influence the growth in ultimate compressive strength, and both of their influences accumulate. The growth of ultimate compressive strength appears mainly because of the addition of steel fibre. Above a certain optimum value, of the steel fibre addition (about 1.5%), ultimate compressive strength of the examined composites slightly decreases. To realize the flexural strength examination under four point loading, two concentrated loads were applied on the specimens in spacing 1/3 of the span. The results of flexural strength examinations, carried out on prism specimens (100 mm · 100 mm · 500 mm), are presented in Figure 6. The shape of the presented surface shows a big influence of the steel fibre addition on acquiring quasi-plastic features by the composite. SFRCC mix without any silica fume replacement is characterized by the highest flexural strength.

Figure 6. Deflection db of a prism specimen 100 mm · 100 mm · 500mm

Figure 7. Stress-strain VH curves for axial compression

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 144

144

Jacek KATZER

During the flexural strength examination, deflection was measured, at the imposed force F = 2 · 2.5 kN. The results of the deflection measurements are shown in Figure 7. The interdependence stress versus strain (VH for chosen SFRCC is presented in Figure 8. The examination was carried out on cylinder specimens (I 150 mm, h = 500 mm) with axial compression. The examination was finished after having achieved about 0.9 failure load. The watertightness of SFRCC specimens was tested after a cycle of pre-loads. The cycle consisted of loading a specimen eight times. When the 45 % of breaking stress was achieved the specimen was unloaded. The specimens were loaded with an increasing force at a uniform rate (the speed of growing of stress 0.5 MPa/s ± 0.1 MPa/s). In these conditions the achieved pulse frequency was 1/35 Hz. Control specimens were not subjected to any load. After the cycle of load, specimens were dried in temperature +60 °C. Drying the specimens was to eliminate the influence of absorption of moist from air while curing. Permeability of SFRCC was tested using an instrument that supplies water from below and can test simultaneously six specimens under a constant pressure from 0.2 MPa to 1.2 MPa. The test was carried out under constant pressure of 1.2 MPa within 72 hours. After this period of time specimens were split and the depth of water penetration was measured. The results were presented with the use of a parameter kv calculated according to the equation (1): kV = xmax /(2·Ȉ hi ti),

(1)

where: xmax h ti

– maximum depth of water penetration in meters – pressure of water in meters H2O – time of lasting of pressure in seconds.

This parameter reflects the conventional speed of leaking water through a specimen. The expression of watertightness with the use of parameter kv allows to compare watertightness of both specimens which were thoroughly or partially penetrated by water during the experiment. One may also compare the permeability of specimens using varied equipment and methods.

Figure 8. Permeability of SFRCC after a pre-load cycle

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 145

Permeability of SFRCC based on fine aggregate after pre-load cycles

145

Figure 9. Permeability of SFRCC not subjected to pre-load cycle Permeability of SFRCC subjected to a cycle of loads is shown in Figure 8. The described composites are characterized by permeability that increases at a uniform rate in relation to SF replacement and fibre additions and is close to a linear relation. The composite matrix was characterized by kV =340·10-12m/s. The most tight composite mix modified by the maximum additions of SF and Vf was characterized by kV =120·10-12m/s. Permeability of SFRCC which were not subjected to any pre-load is shown in Figure 9. Permeability of composite matrix was equal to 23010-12 m/s. Modifying the composite mix solely with steel fibre addition allowed to achieve permeability of 103·10-12 m/s. Modifying the composite mix solely with silica fume replacement allowed to achieve permeability of 39·10-12 m/s. It is visible in Figure 10 that the addition of fibre influences the permeability attained with silica fume (when its quantity exceeds 10%). With the increase of the amount of silica fume the permeability of the composite improves, especially in mixes of a high content of fibre. With the replacement of more than 15% of cement by silica fume all mixes are characterized by considerable permeability, which stabilizes at a level of 25-23·10-12 m/s.

ANALYSIS OF THE RESULTS Ultimate compressive strength (which was equal to 24 MPa for the composite matrix) was improved by over 100 % (to 50 MPa) for some composite mixes. All composites were made on the basis of CEM I 32.5 (400 kg/m3). It proves that there are yet other possibilities of achieving significantly higher strengths. For this property the combination of the SF replacement and fibre addition in the hardened composite is very effective. The results of flexural strength examinations disclose the effectiveness of steel fibre addition behaviour with this kind of loading. Plain composite matrix is characterized by flexural strength equal to 2.75 MPa. Composite reinforced by 2 % of steel fibre is characterized by flexural strength equal to 6 MPa. The addition of silica fume influences significantly flexural strength obtained thanks to the steel fibre itself. Despite this negative tendency composite mixes modified by maximum replacement (15 %) of silica fume and by

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 146

146

Jacek KATZER

addition of steel fibre from 1.5 % to 2 % are characterized by much higher flexural strength than the composite matrix. The measurement of deflection, (investigated during the flexural strength test described above), shows the acquirement of quasi-plastic properties by SFRCC. The deflection of composite with 25 % silica fume replacement and 2 % of steel fibre addition is five times smaller than the deflection of composites modified by silica fume replacement alone. Permeability of SFRCC without any pre-load reaches the level from kV = 2310-12 m/s to 23010-12 m/s. Composites which were subjected to a cycle of pre-load are characterized by permeability from kV = 12010-12m/s to 34010-12m/s. Assuming the permeability of composites with no pre-load as a point of reference it is possible to calculate some proportions. The permeability of the composite matrix with no preload to pre-loaded one is 1:5. The permeability of the composite modified with a maximum replacement of SF and Vf addition with no pre-load to pre-loaded one is 1:1.5. It is observable in the above proportions, how significant is the influence of the addition of steel fibre and load onto permeability of SFRCC. Cement composite “working” in a construction carries most frequently loads causing about half of the breaking stress and changing in time. Bearing this in mind the most important mechanical property of the SFRCC for the application in structural and civil engineering is permeability after a cycle of pre-loads. In such a case permeability of SFRCC modified only by silica fume replacement examined in a traditional way is very high. The same composite subjected to the cycle of pre-loads loses about 80% of its primary permeability and it fails to meet the primary requirements as far as permeability is concerned. One of the main problems connected with SFRCC is their precise designing as far as their flexural strength is concerned, which has a great significance in their appropriate and effective application. The use of the parameter specifying spacing of fibres in composite matrix creates a potential possibility of designing flexural strength of SFRCC. The relation between the number of fibres and their general volume in a composite matrix influences the distance between single fibres. There is a noticeable influence of fibres on elongation of composite matrix when the average distance between axis of fibres is smaller than 12 mm. In order to achieve a precise description of the phenomenon Mangat [18] defined a parameter describing apparent fibre spacing in a composite matrix marked with a symbol Se. Calculating Se involves, apart from the geometry of a fibre, such features as fibre-matrix bond strength and fibre ultimate strength. The parameter describing apparent fibre spacing in composite matrix was called ‘effective fibre spacing’. The relation between Se and diameter (d), length (l) and volume of fibres (Vf) of a circular cross-section characterized by fibrematrix bond strength W = 1MPa [4], is presented below. Se

87.4

d Vf l

(2)

For Vf = 1.016 % the parameter Se of the SFRCC specimens is equal to 12 mm. Together with the increase of the quantity of fibres, Se decreases and finally reaches 8.74 mm for Vf = 2 %. Assuming Se = 12 mm as a binding border of effectiveness of the steel fibre addition one may say that in the described composites the effective addition of steel fibre equals Vf = 1 %. The course of the curve describing flexural strength ff120 of SFRCC after 120 days of curing and the relation 1/Se characterizing the same composites are shown in Figure 10.

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 147

147

Permeability of SFRCC based on fine aggregate after pre-load cycles

Figure 10. The parameter

1 and flexural strength ff120 of the SFRCC as function of Vf Se

Both of the presented curves have similar shapes and after scaling the relation 1/Se for the analyzed composite matrix reflects well flexural strength ff120. Assuming that k1 is a nondimension factor determined empirically one could write: ff

k1

1 Se

(3)

In case of considered SFRCC k1 equals about 524. The employment of the parameter Se for expressing flexural strength of SFRCC requires vast experiments of fibre reinforced cement composites based on different matrixes and modified by different types of fibres. The analysis of flexural strength in reinforced cement composites presented above is an encouraging beginning to the introduction of relations more universal and possible for wider applications in engineering practice. CONCLUSIONS

The presented research on cement composites based on fine aggregate matrix modified by steel fibre and silica fume lead to the following conclusions: 1. The addition of both silica fume and steel fibre improves ultimate compressive strength of the examined composites. 2. Steel fibre improves flexural strength of the examined composites and it decreases deflection and linear strain. 3. Silica fume does not improve flexural strength of the examined composites 4. Silica fume does not decrease strain under axial load of the examined composites. 5. The relatively high permeability of the examined composites after the pre-load cycle is achieved due to the addition of steel fibres. 6. It is possible to obtain composites, on the basis of fine aggregate, which are characterized by very high chosen features.

15-Permeability of.qxd

9/19/2009

11:59 AM

Page 148

148

Jacek KATZER

7. “Central composite design” may be effectively employed for fine aggregate cement composites modified by steel fibre and silica fume. REFERENCES

1. Neville, A.M., Properties of Concrete, 4th Edition. Addison Wesley Longman, Harlow, Essex, 1995 2. Collepardi, M., The New Concrete, Grafiche Tintoretto, 2006 3. Johnston, C.D., Fiber reinforced cements and concretes, Gordon and Breach Science Publishers, Amsterdam, 2001 4. Maidl, B.R., Steel Fibre Reinforced Concrete, Ernest & Sohn, Berlin, 1995 5. Katzer, J., Influence of different types of reinforcement on strength of mortar beams, Proceedings, 9th International Conference “Modern Building Materials, Structures and Techniques”, May 16–18, 2007 Vilnius, Lithuania, 276-277 6. Kovacs, I., Balazs, G.L., Structural performance of steel fiber reinforced concrete, Budapest: University of technology and Economics, Budapest, 2004 7. Nawy, E.G., Fundamentals of high strength high performance concrete, Longman, UK, 1996 8. Zollo, R.F., Fiber-reinforced Concrete: an Overview after 30 Years of Development, Cement and Concrete Composites 19/97, 107-122 9. Katzer, J., Properties of Precast SFRCC Beams Under Harmonic Load, Science and Engineering of Composite Materials, Vol.15, No.2, 2008, 107-120 10. Katzer, J., Impact resistance of SFRM modified by varied superplasticizers, Proceedings, XV International Conference Mechanics of Composite Materials, 26-30 May 2008, Riga, Latvia, 122 11. Lenkiewicz, W. and Pidek, W., Compressive strength of concrete in debuting elements (in Polish), Proceedings, XVI Conference “Krynica 1970”, Poland, 97-101 12. Katzer, J., Kobaka, J., The assessment of fine aggregate pit deposits for concrete production, Kuwait Journal of Science and Engineering, Vol. 33, 2, 2006, 165-174 13. Toutanji, H., Bayasi, Z., Effects of manufacturing techniques on the flexural behaviour of steel fiber-reinforced concrete, Cement and Concrete Research, 28, 1998, 115-117 14. Bayramov, F., et al., Optimisation of steel fibre reinforced concretes by means of statistical response surface method, Cement and Concrete Composites, 26, 2004, 665-675 15. Mann, H.B., Analysis and Design of Experiments, Dover Publications, Inc., New York, USA, 1950 16. Schenck, H., Theories of engineering experimentation, Hemisphere Publishing Corporation, Washington DC, USA, 1979 17. Lee-Ing, T., Chung-Ho, W., STATISTICA V5.5 and Basic Statistic Analysis, TasngHai Publisher, Taiwan, 2002 18. Mangat, P.S., Tensile strength of steel fiber reinforced concrete, Cement and Concrete Research, vol.6, 1976, 245-252

16-Statistical optimization.qxd

9/19/2009

11:59 AM

Page 149

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

KEYNOTE PAPER

KEYNOTE PAPER

STATISTICAL OPTIMIZATION OF LOW SLUMP TERNARY CONCRETE MIXTURES WITH GROUND GRANULATED BLAST FURNACE SLAG (GGBS) AND HIGH CALCIUM FLY ASH FOR PAVEMENT APPLICATIONS 1

Adam RUDY1*, Jan OLEK1**, Tommy NANTUNG2 and Richard M. NEWELL3 Purdue University, School of Civil Engineering, 550 Stadium Mall Drive, West Lafayette, IN 47907, USA, e-mail: *[email protected], **[email protected] 2 Indiana Department of Transportation, Research Division, 1205 Montgomery Road, West Lafayette, IN 47906, USA, email: [email protected] 3 Berns Construction Co., Inc., 2081 North Catherwood Avenue, P.O. Box 19815, Indianapolis, IN 46219-9815

ABSTRACT This paper presents the results of optimization study of ternary concrete mixtures for use in pavements. The variables used in the study included the total volume of paste in the mixture (21 to 25%) and the level of cement replacement by combination of fly ash (10-20%) and ground granulated blast furnace slag (GGBFS) (18-30%). A total of 16 different concretes (including one control, cement-only mixture) were produced in the laboratory and tested for various fresh properties, compressive and flexural strength, scaling resistance, free shrinkage, and rate of water absorption. The optimal mixtures were found to contain, respectively, 15% of fly ash, 26% of GGBFS, and 21.5% of paste. Moreover, it was found that the fly ash content was the most influential variable in the ternary systems studied and that it could only be varied in a relatively narrow range (from 10 to 15.5%) in optimal ternary mixes with fixed amount of paste.

Keywords Fly ash, GGBFS, optimization, statistical methods, ternary mixtures

INTRODUCTION AND RESEARCH SIGNIFICANCE Driven by the desire to produce more durable and longer lasting materials, the concrete mixtures used in pavement applications are becoming more complex, often requiring use of multicomponent binders, chemical admixtures and reduced amount of water. In that sense, concrete paving mixtures can be viewed as highly “engineered”, high performance concretes (HPC) designed to achieve specific properties (Simon [1]). The process of proportions selection for such mixtures can be optimized by using statistical experimental design methods. These statistical techniques are rigorous in achieving desired concrete properties and mixture optimization for given constituents (Simon [1]). They are commonly used in industry to optimize products and processes (Derringer [2]), and have been applied in some research studies on improving properties of HPC (Nehdi [3]; Luciano et al. [4], Bajorski et al. [5]), SCC (Patel et al. [6], Khayat et al. [7,8]) or properties of cement based materials (Brandt and Marks [9, 10], Simone et al. [11], Lawler [12]). They have not, however, been applied in optimization of concrete mixture proportions for pavements.

16-Statistical optimization.qxd

150

9/19/2009

11:59 AM

Page 150

Adam RUDY, Jan OLEK, Tommy NANTUNG and Richard M. NEWELL

In order to simplify the optimization process for the system of mixtures used, only three parameters were selected as variables in the study. These parameters included paste content (as % of mixture volume) and both fly ash and GGBFS (ground granulated blast furnace slag) contents (as % of weight of total cementitious materials). These variables were allowed to vary within certain ranges, which were selected based on both technical literature and concrete pavements mixture specifications used by several state agencies.

MATERIALS The materials used in this study represented typical concrete constituents from pavement projects in the state of Indiana (USA) and included: ASTM C 150 Type I portland cement, ASTM C 989 Grade 120 slag cement (ground granulated blast furnace slag (GGBFS)) and ASTM C 618 Class C fly ash. The physical properties and chemical composition of all cementitious materials were well within ranges given by the respective specifications. Locally available natural siliceous sand (SpG = 2.66 and absorption 1.27%) meeting the gradation requirements of INDOT’s specification for #23 material [INDOT 2008] was used as fine aggregate. The coarse aggregate used was crushed limestone (SpG = 2.64 and absorption 1.30%) with maximum size of 19 mm (¾ in.) and gradation meeting the requirements of INDOT’s specification for #8 stone (INDOT [13]). Both fine and coarse gradations can be classified as well graded. For all concrete mixtures produced in this study the mass ratio of fine-to-total aggregate was constant and equal to 45%. This ratio was recommended by project advisors and was consistent with typical practices used in the State of Indiana (Thier [14]). Two chemical admixtures: vinsol-based air entraining agent (AEA) and modified glucose polymer-based normal range water reducing admixture (WR) (complying with ASTM C 494 Type A) were used to achieve target air content of 6.5% ± 1.0% and slump of 50 mm ± 25 mm (2 in. ± 1 in.).

EXPERIMENTAL DESIGN AND MIXTURE PROPORTIONS The test matrix of concrete mixtures used in this study was developed using three factorial Central Composite Design (CCD) statistical method. The CCD method is commonly used to develop experimental plans, results of which are later analyzed using Response Surface Methods of mixture optimization (Meyers [15]). A total of 15 different ternary concretes mixtures containing Type I portland cement and various percentages of GGBFS (ground granulated blast furnace slag) and Class C fly ash have been produced and tested during this research program. In addition to the ternary mixtures, one “cement only” mixture with standard INDOT’s composition was also tested to establish the baseline performance characteristics. The ranges for three experimental variables which have been used in the CCD method to develop a matrix of test mixtures are shown in Table 1. All test mixtures used in the study were designed using the following parameters: a) Constant water to binder ratio (w/b = 0.44). b) Constant fine to total aggregate ratio (FA/total aggregate = 0.45) by mass. c) Target slump of 50 mm with tolerance ± 25 mm (2 in. ± 1 in). d) Target air content of 6.5% with tolerance ±1.0%.

16-Statistical optimization.qxd

9/19/2009

11:59 AM

Page 151

Statistical optimization of low slump ternary concrete mixtures with ground granulated blast furnace...

151

Table 1 - Ranges of experimental variables used EXPERIMENTAL VARIABLES X2

X3 Paste Fly ash replacement GGBFS replacement (21-25%) by Ranges (10-20%) by mass (18-30%) by mass volume of mixture* Levels (%) 10, 12.5, 15, 17.5, 20 18, 21, 24, 27, 30 21, 22, 23, 24, 25 * Paste volume is the sum of absolute volumes of water and cementitious materials only (excluding entrapped and entrained air) X1

The detailed information regarding the composition of all mixtures used in this study, including the plain (cement only) mixture, is summarized in Table 2. The individual mixtures were labeled using the following pattern: % of fly ash, % of slag, _ % of paste volume. As an example, mixture with 15% of fly ash, 18% of slag and 23% of paste has been labeled as 15FA18SL_23. The only exception to the above pattern was plain concrete mixture which was labeled as CTRL.

Table 2 - Concrete mixture proportions for ternary system (Cement/FA/GGBFS) CONSTITUENT MATERIAL [kg/m3] MIX

Binder composition PC

CTRL

335

15FA24SL_21 12.5FA21SL_22 12.5FA27SL_22 17.5FA21SL_22 17.5FA27SL_22 10FA24SL_23 15FA18SL_23 15FA24SL_23 15FA30SL_23 20FA24SL_23 12.5FA21SL_24 12.5FA27SL_24 17.5FA21SL_24 17.5FA27SL_24 15FA24SL_25

166 189 172 174 157 196 199 181 163 166 206 187 190 171 197

1,2

FA

41 36 36 49 49 30 44 44 44 59 39 39 54 54 49

GGBFS

Total

Water

Aggregate Fine

Coarse

Control (plain) concrete mixtures 335 148 811 991 PC/FA/GGBFS (Ternary systems) 65 272 120 864 1055 60 285 125 852 1041 77 284 125 852 1041 60 284 125 852 1041 77 283 125 852 1041 71 297 131 840 1026 53 297 131 840 1026 71 297 131 840 1026 89 296 131 840 1026 71 296 130 840 1026 65 311 137 828 1012 84 310 136 828 1012 65 310 136 828 1012 84 309 136 828 1012 78 323 142 816 997

Total mass

Admixture dosages, ml/100kg of binder WR1 AEA2

2285

164

46

2311 189 172 174 157 196 199 181 163 166 206 187 190 171 197

558 460 493 447 460 427 394 361 361 250 263 230 197 197 33

46 49 53 53 56 53 53 56 59 61 62 62 59 66 82

The final quantities of water reducing (WR) and air-entraining (AE) admixtures were determined at the time of mixing and set at levels needed to meet the target values for slump and air contents

16-Statistical optimization.qxd

152

9/19/2009

11:59 AM

Page 152

Adam RUDY, Jan OLEK, Tommy NANTUNG and Richard M. NEWELL

SPECIMENS AND TEST METHODS The testing plan developed for all concrete mixtures produced during this study consisted of measurements of both fresh and hardened properties. The fresh concrete properties measured included the following: slump (AASHTO T 119), air content (AASHTO T 152), unit weight (AASHTO T 121) and VeBe time (EN 12350-3). The hardened concrete properties measured included the following: 7 and 56 days flexural strength (third-point loading using three 150x150x550 mm prisms) following AASHTO T 97 procedure, 7, 28, 90 days compressive strength using two 150x150x150 mm cubes sawed off from broken portions of the beams obtained during flexural test, Scaling resistance (modified ASTM C 672) using two 75x215x240 mm slabs (each with exposed area 0.046 m2). For each specimen the weight of material lost rather than the visual rating was recorded, Free drying shrinkage following ASTM C 157 using three 75x75x275 mm prisms, The rate of water absorption (by capillary suction) following ASTM C 1585 using two 100x200 mm concrete disks cut from the mid section of concrete cylinders The rate of water sorption (by ponding) following modified ASTM C 1585 using two 100x200 mm concrete disks cut from the top section of concrete cylinders.

FRESH CONCRETE PROPERTIES The average measured air content was 6.7% with standard deviation of 0.4%, whereas the average slump value was 57 mm with standard deviation of 11 mm. Based on these results, it can be concluded that all concrete mixtures satisfied the design requirements. In addition, the following observations were made during production of the ternary mixtures: x Despite formally satisfying the slump and fresh air content requirements, mixtures with 21 and 22% of paste were difficult to produce and required very high dosage of water reducing admixture (460 to 558 ml/100 kg of binder) x The dosage of water reducing admixture dropped linearly with the increase of the amount of paste in the mixtures x The VeBe time varied from 6 to 1.5 s, respectively, for mixtures with 21 to 25% of paste. The VeBe time decreased linearly with the increase in the amount of paste in the mixtures. x Paste and fly ash content were found to have the strongest influence on the VeBe time and dosage of water reducing admixture

HARDENED CONCRETE PROPERTIES Due to space limitation the detailed data analysis of hardened concrete properties obtained for ternary mixtures cannot be included in this paper. However, for reader’s convenience the average values of hardened properties are summarized in Table 3. The same Table also includes the estimated cost of all materials used to prepare the mixtures.

16-Statistical optimization.qxd

9/19/2009

11:59 AM

Page 153

Statistical optimization of low slump ternary concrete mixtures with ground granulated blast furnace...

153

Table 3 – Test results of hardened concrete properties

Mix label

15FA21SL_21 12.5FA21SL_22 12.5FA27SL_22 17.5FA21SL_22 17.5FA27SL_22 10FA24SL_23 15FA18SL_23 15FA24SL_23 15FA30SL_23 20FA24SL_23 12.5FA21SL_24 12.5FA27SL_24 17.5FA21SL_24 17.5FA27SL_24 15FA24SL_25 CTLR_565_25.4

Flexural strength [MPa]

Compressive strength [MPa]

7 days

56 days

7 days

28 days

90 days

4.6 5.2 4.9 4.7 4.6 5.3 4.8 5.2 5.2 4.9 5.0 5.2 5.1 5.1 5.3 5.4

6.2 6.2 6.2 6.7 6.4 6.3 6.0 6.4 6.3 6.3 6.4 6.0 6.4 6.2 6.5 5.6

34.5 33.2 35.7 36.5 29.1 37.9 31.8 35.4 32.0 36.5 32.9 34.5 31.6 32.2 29.9 38.1

48.8 40.3 49.2 51.4 43.1 50.2 42.4 47.5 45.4 45.7 43.6 46.6 42.3 46.1 41.9 44.7

51.6 44.4 51.8 56.9 44.6 51.7 51.0 54.7 48.3 56.8 48.2 46.4 46.8 54.4 45.5 49.9

Mass loss due to scaling [kg/m2]

Free shrinkage [%]

Rate of water sorpt./ absorpt. [10-4 mm/s0.5]

Cost [$/m3]

0.73 0.86 0.62 1.43 1.14 0.69 1.57 1.82 1.63 2.32 1.68 4.10 2.19 8.94 8.17 1.23

-0.0390 -0.0435 -0.0451 -0.0430 -0.0488 -0.0479 -0.0444 -0.0510 -0.0520 -0.0410 -0.0485 -0.0460 -0.0477 -0.0495 -0.0515 -0.0560

18/17 24/30 21/24 25/34 23/25.5 21/25 30/35 28/28 27/24 31/35 37/32 30/26.5 46/44 40/37 51/53 56/35

56.93 58.04 57.52 57.21 56.64 58.72 58.48 57.75 57.10 56.65 58.96 58.20 57.91 57.25 58.24 64.55

STATISTICAL OPTIMIZATION OF MIXTURE PROPORTIONS The performance characteristics chosen for the purposes of optimizing concrete proportions included the following: 7 and 56 days flexural strength, 7, 28 and 90 days compressive strength, scaling resistance (mass loss) after 50 cycles of freezing and thawing, free shrinkage after 448 days of drying, rate of water sorption, rate of water absorption, and cost of raw materials to produce each mixture. The objective of the optimization process used in this study was to select composition of a ternary mixture which will be most economical while, at the same time, capable to satisfy requirements of minimum or maximum value for all of the performance characteristics listed above. All of the above performance characteristics were treated in the experimental plan as the desired responses which depended on the combination of three independent variables: the total paste content in the mixture and the amount of both fly ash and GGBFS in the mixture. Excluded from the optimization process were the fresh properties of concrete as all mixtures were designed for nominally constant slump and air contents. The optimization process utilized in this research consisted of three main steps: Step 1- development of statistical models for prediction of selected responses (performance characteristics); Step 2 – selection of desirability functions and conversion of predicted responses to minimum (0) or maximum (1) desirability values; and Step 3 – combining of individual desirability values from step 2 into overall desirability function and using this function to select an optimum concrete proportions. Each of these steps, in turn, contained several sub-steps which are briefly summarized in the next section.

16-Statistical optimization.qxd

9/19/2009

154

11:59 AM

Page 154

Adam RUDY, Jan OLEK, Tommy NANTUNG and Richard M. NEWELL

Step 1 - Development of Statistical Models The process of models development was started by performing statistical analysis of the results. The analysis was initiated by investigation of the data correctness with respect to precision and bias included in the standard specifications. Next, the statistical software was utilized to perform multiple regression analysis. This multiple regression analysis involved running ANOVA evaluation for full quadratic model, checking t-statistic for each model coefficient and the F-values for the regression model. The ANOVA analysis helped to evaluate initial full quadratic model. If any of the model coefficients was found to be statistically not significant, the variable associated with this coefficient was removed and ANOVA analysis was repeated using partial quadratic or linear model. Finally, the validation of a model was performed by calculating case statistics (to identify outliers) and by examining the diagnostic plots such as Quantile-Quantile (QQ) plot and residual vs. predicted value plots. The detailed description of model development process for central composite design method is described by Simon (Simon [1]) and Lu (Lu [16]). Only the highest order models resulting from the verification process were used for mixtures optimization. Due to space limitation the details of these models are not presented in this publication but can be found elsewhere (Rudy [17]). Step 2 - Selection of Desirability Functions and Establishment of Desirability Values In Step 2, the desirability (objective) functions were selected for each of the performance responses. First, three types of desirability functions minimum, maximum and linearly decreasing (Simon [1]) were selected for optimization purposes. Next, the target performance values were assigned to each of the desirability functions. These target (critical) values were selected based on multiple literature sources, existing specifications and in-depth data analysis. Table 4 presents target values for all performance responses along with type of desirability function utilized in the optimization process.

Table 4 - Summary of target values for performance responses with assigned types of desirability functions Desirability function, di

Basis for selection

7d flex

L = 3.9 MPa (minimum value required for concrete pavements in section 500 of INDOT’s specifications) (INDOT [13])

56d flex 7d comp. 28d comp. 90d comp. Scaling

Shrinkage

L = 5.8 MPa (minimum value measured for reference mixture CTRL_515) L = 24.1 MPa (minimum design strength for concrete pavement mixtures in Michigan at 28 days by Specification 601) (MDOT [18]) L = 34.5 MPa (typical value of compressive strength for concrete pavements in Indiana) L = 37.4 9 MPa (arbitrary values considering 10% increase in strength from this at 28 days) U = 0.8 kg/m2 (the maximum amount of scaled material allowed for concrete barrier mixtures according to Ontario Provincial Standard Specification ) (OPS [19]) U = -515 İ (this value represents the long-term drying shrinkage of concrete mixtures for which it is expected that probability of cracking is only 28.6% (Radlinska [20])

Desirability function type Min. Min. Min. Min. Min.

Max. Max.

U = 23.4·10 mm/s0.5

Max.

Absorptivity U = 26.2·10 mm/s0.5

Max.

Sorptivity

3

Cost

U= 96.53 $/m L = 64.55 $/m3

Linear

16-Statistical optimization.qxd

9/19/2009

11:59 AM

Page 155

Statistical optimization of low slump ternary concrete mixtures with ground granulated blast furnace...

155

Rate of water ingress x10-4[mm/s0.5]

The symbols “L” and “U” used in Table 4 represent, respectively, lower and upper limits for desirability functions selected to optimize each response. The selection of critical values for water sorptivity and absorptivity, as well as the cost, requires further explanation and will be presented in the next paragraph. In case of water ingress data, the critical values were selected based on observed linear correlation between these measured properties and scaling data, as illustrated in Figure 1. Ternary mixtures

60

y = 3.8311x + 20.29 R² = 0.7629

50 40

y = 2.4996x + 24.199 R² = 0.5106

30

Sorptivity vs. scaling

20

Absorptivity vs. scaling 10

Linear (Sorptivity vs. scaling) Linear (Absorptivity vs. scaling)

0 0.000

1.000

2.000

3.000

4.000

5.000

6.000

7.000

8.000

9.000

Scaling [kg/m2]

Figure 1 - Correlation between scaling and the rate of water ingress for ternary mixtures. Using the value of 0.8 kg/m2 (this value is considered to be an acceptable limit for scaling by the Ontario Ministry of Transportation (OPS [19]) as an input into equations shown in Figure 1, the target (critical) values for sorptivity and absorptivity may be established. The resulting values are reported in Table 4. The cost was modeled by a linearly decreasing function, which required upper and lower limits. The upper cost limit used for ternary mixtures was the same as the one used for plain concrete with 335 kg/m3 of cement. The lower cost limit was established based on modified standard Iowa’s Class B paving mixture (IM 529 [21]). This modification allows for replacement of up to 40% of cement (20% by fly ash and 20% by GGBFS). Once the individual predicted responses had been converted into desirability values, the composite response (representing the influence of all variables studied) was calculated using Equation 1. D (d1 d 2 ... d n )1 / n (1) where, dn are the desirability values for individual performance responses, D is the overall desirability (geometric mean) and n is the number of performance responses utilized in optimization process. Finally the maximum value of the overall desirability (Dmax) was found and adequate plots were prepared. These plots are shown in Figures 2(a) through 2(f).

RESULTS OF NUMERICAL OPTIMIZATION FOR TERNARY SYSTEM

During numerical optimization of ternary systems, a total of 8924 different combinations of test variables (fly ash, GGBFS and paste content) were investigated by means of numerical simulations. Based on these simulations, the composition containing 15% of fly ash, 26% of GGBFS and 21.5% of paste resulted in the highest (maximum) desirability value of 0.975. From the practical perspective (i.e. need for adjustment due to variability in the properties of materials) utilizing the single combination of variables just because it resulted in highest desirability may be too restrictive. As such, it may be more advantageous to explore how the desirability changes around the highest value. This issue has been examined by

16-Statistical optimization.qxd

9/19/2009

156

11:59 AM

Page 156

Adam RUDY, Jan OLEK, Tommy NANTUNG and Richard M. NEWELL

creating the desirability surface for the pair of two independent variables, while keeping the third variable at its maximum (optimum) value (Derringer et al., [2]). This approach is shown in Figures 2 (a-f), where each pair of plots (a-b, c-d, e-f) represents, respectively, the surfacecontour plots for paste-GGBFS content prepared at max. value of fly ash (15%), surfacecontour plot for paste-fly ash content at max. value of GGBFS (26%), and surface-contour plot for GGBFS-fly ash at max. value of paste (21.5%). 25.0 24.5

0.0 0.2 0.4 0.6 0.8

1.0

0.8 0.6

Paste content [%]

0.0 0.2 0.4 0.6 0.8 1.0

1.0

Desirability

Point with higest desirability

24.0

1.2

0.4 0.2 0.0

23.5 23.0 22.5 22.0

25.0 24.5 24.0 23.5 Pa 23.0 ste 22.5 22.0 co nte 21.5 21.0 nt

[% ]

30 28

] [% nt nte co

21.5

26

24 22

20 18

S BF GG

21.0 18

20

22

24

26

28

30

GGBFS content [%]

b)

a) 25.0 24.5 1.2

0.8 0.6

Paste content [%]

Desirability

1.0

Point with higest desirability

0.0 0.2 0.4 0.6 0.8 1.0

24.0

0.0 0.2 0.4 0.6 0.8 1.0

0.4 0.2 0.0

23.5 23.0 22.5 22.0

25.0 24.5 24.0 23.5 Pa 23.0 ste 22.5 22.0 co 21.5 nte 21.0 nt

[% ]

20

21.5

18

] 16 t [% ten on hc as 14

12 10

Fly

21.0 10

12

14

16

18

20

Fly ash content [%]

c)

d) 30 0.0 0.2 0.4 0.6 0.8

28 1.2

Desirability

1.0 0.8

GGBFS content [%]

0.0 0.2 0.4 0.6 0.8 1.0

0.6 0.4 0.2

1.0

26

24

22

0.0 30

20

28

GG BF S

Point with higest desirability

20

18

26

co nte nt

] [% nt 14 nte co 12 h as Fly 16

24 22 20

[% ]

18

e)

10

18

10

12

14

16

18

20

Fly ash content [%]

f)

Figure 2 - Plots of desirability surfaces for ternary mixtures at optimum (15%) fly ash content (a-b), optimum (26%) GGBFS content (c-d), optimum (21.5%) paste content (e-f) As shown in these plots, the desirability surfaces (shown as the bright areas) appear to be flat. This indicates that any combination of variables contained within these areas does not significantly decrease the overall level of desirability. However, the allowable ranges within which the two variables can change without significantly decreasing desirability are not the same for different set of variables. For example, the data shown in Figures 2 (a) and 2(b) indicate that at optimum (15%) fly ash content, the area of acceptable desirability is very narrow. That practically eliminates possibility of adjusting mixture composition by changing slag or paste content while keeping the fly ash content at optimum. On the other hand, as shown in Figures 2 (c) and 2(d), when analyzed at optimum (26%) GGBFS content the

16-Statistical optimization.qxd

9/19/2009

11:59 AM

Page 157

157

Statistical optimization of low slump ternary concrete mixtures with ground granulated blast furnace...

mixture composition can be adjusted within much broader range without reducing overall desirability. While the desirability area presented in Figures 2(c) and 2(d) is broader than that presented in Figures 2(a) and 2(b), it will still be difficult to implement this approach in practice as one would have to change the volume of paste, which may have negative impact on workability. This is illustrated in Figure 3(a), where the paste-based desirability area is small. However, this area increases with an increase in paste content (Figures 3(b) through 3(d)). As mentioned earlier, mixtures with paste content below 22% were difficult to work with and to compact. It is therefore more practical to set a paste content in the range that gives relatively good workability (as indicated by broad desirability surface) in Figures 3(e) and 3(f) and adjust the fly ash and slag content.

24 22

24 22

26

22

20

20

18

18

18

14

16

18

10

20

12

14

16

18

12

14

16

18

10

20

12

14

16

Fly ash content [%]

Fly ash content [%]

Fly ash content [%]

a)

b)

c)

d)

30

GGBFS content [%]

26

30

28

24 22

26 24 22

26

22

20

20

18

18

18

14

16

18

20

10

12

14

16

18

20

28

24

20

18

20

30 0.0 0.2 0.4 0.6 0.8 1.0

28

0.0 0.2 0.4 0.6 0.8 1.0

GGBFS content [%]

0.0 0.2 0.4 0.6 0.8 1.0

28

12

22

Fly ash content [%]

30

10

24

18 10

20

0.0 0.2 0.4 0.6 0.8 1.0

26

20

GGBFS content [%]

12

28

24

20

10

GGBFS content [%]

26

0.0 0.2 0.4 0.6 0.8 1.0

28

0.0 0.2 0.4 0.6 0.8

GGBFS content [%]

GGBFS content [%]

GGBFS content [%]

26

28

GGBFS content [%]

0.0 0.2 0.4 0.6 0.8 1.0

28

30

30

30

30

0.0 0.2 0.4 0.6 0.8 1.0

26 24 22 20 18

10

12

14

16

18

20

10

12

14

16

Fly ash content [%]

Fly ash content [%]

Fly ash content [%]

Fly ash content [%]

e)

f)

g)

h)

18

20

Figure 3 - Changes of desirability surface for ternary mixtures at paste contents of: (a) 21.0%, (b) 21.25%, (c) 21.5%, (d) 21.75%, (e) 22.0%, (f) 22.25%, (g) 22.5%, and (h) 22.75% For mixtures with 22% of paste the amount of GGBFS allowable for making ternary systems with fly ash was broader than for mixtures with 21.5% of paste (which was determined as being optimum). More in-depth data analysis revealed that in case of ternary mixtures with 22% of paste the improvement in concrete workability helped to attain required scaling resistance and 7 days compressive strength for larger number of ternary mixtures. Based on these observations, it was concluded that the increase in GGBFS helped to increase scaling resistance and to gain sufficient compressive strength, but only if mixtures with 22% of paste are produced. The observations listed above led the authors to modify the desired optimum composition of ternary systems from that obtained by numerical optimization. The modified composition includes 15% of fly ash, 27% of GGBFS and 22% of paste. Although this modification reduced the desirability value from original (0.975) to 0.974 it is not expected that this adjustment will have negative influence on concrete properties and the price. At the same time, the proposed adjustment should also help to reduce potential workability problems.

16-Statistical optimization.qxd

158

9/19/2009

11:59 AM

Page 158

Adam RUDY, Jan OLEK, Tommy NANTUNG and Richard M. NEWELL

SUMMARY AND CONCLUSIONS

The main goal of this paper was to present the results of the optimization of ternary (PC+fly ash+GGBFS) systems for potential application in construction of concrete pavements. The research plan included test matrix of 15 statistically designed (using Central Composite Design method) concrete mixtures, each containing different quantities of fly ash, GGBFS and paste. In addition, one plain (cement only) mixture was produced in order to establish baseline criteria for optimization. The data for flexural and compressive strengths, scaling, free shrinkage, water sorptivity (absorptivity) and estimated cost were used as the performance responses in the numerical optimization process. The optimization process involved development of the predictive models for each performance response (using Response Surface Methodology), defining desirability functions with their threshold limits, and finally establishing maximum value of the overall (combined) desirability. The conclusions and recommendations resulted from this study can be summarized as follows: x The highest desirability value (0.975) was found for combination of 15% of fly ash, 26% of GGBFS and 21.5% of paste, x GGBFS content was found to be the most flexible variable with respect to changes in paste and fly ash contents, x For practical reasons, it was suggested to select a range of paste content that assures adequate workability and adjust fly ash and GGBFS contents in order to find the most desirable proportions x The broadest area of high desirability exists for various combinations of fly ash and GGBFS at paste content of 22% (slightly different from that indicated by the optimization process (21.5%)

ACKNOWLEDGMENTS

This research was supported by the Joint Transportation Research Program administered by the Indiana Department of Transportation and Purdue University. The content of this paper reflect the views of the authors, who are responsible for the facts and the accuracy of the data presented herein, and does not necessarily reflect the official views or policies of the Federal Highway Administration and the Indiana Department of Transportation, nor do the content constituents a standard, specification, or regulation. The authors also wish to express their gratitude to Matt Noelker for his help in preparation of concrete mixtures. REFERENCES

1. Simon, M. J., Concrete Mixture Optimization Using Statistical Methods: Final Report, Publication FWHA-RD-03-060. FHWA Office of Infrastructure Research and Development, McLean, VA, 2003, pp. 53 2. Derringer, G., Suich, R., Sumiultaneous Optimization of Several Responses Variables, Journal of Quality Technology, Vol. 12, No.4, October 1980, p. 214 3. Nehdi, M.L., Sumner, J., Optimization of ternary cementitious mortar blends using factorial experimental plans, Materials and Structures, Vol. 35, September-October 2002, 495-503 4. Luciano, J.J., Nmai, C.K., DelGado, J.R., A Novel Approach to Developing High-Strength Concrete, Concrete International, May, 1991, 25 -19

16-Statistical optimization.qxd

9/19/2009

11:59 AM

Page 159

Statistical optimization of low slump ternary concrete mixtures with ground granulated blast furnace...

159

5. Bajorski, P., Streeter, D.A., Perry, R.J., Applying Statistical Methods for Further improvements of High-Performance Concrete for New York State Bridge Decks, Transportation Research Record No. 1574, Transportation Research Board, Washington, DC, 1997, 71-79 6. Patel, R., Hossain, K.M.A., Shehata, M., Bouzoubaa, N., Lachemi, M., Development of Statistical models for Mixture Design of High-Volume Fly Ash Self-Consolidating Concrete, ACI Materials Journal, Vol. 101, No.4, July-August 2004, 294-302 7. Khayat, K.H., Ghezal, A., Hadriche, M.S., Factorial design models for proportioning selfconsolidating concrete, Materials and Structures, Vol. 32, November 1999, 679-686 8. Khayat, K.H., Ghezal, A., Hadriche, M.S., Utility of statistical models in proportioning self-consolidating concrete, Materials and Structures, Vol. 33, June 2000, 338-344 9. Brandt., A.M., Marks, M., Examples of the multicriteria optimization of cement-based composites, Composite Structures, Vol. 25, 1993, 51-60 10. Brandt., A.M., Marks, M., Optimization of the Material Structure and Composition of Cement based Composites, Cement and Concrete Composites, Vol. 18, 1996, 271-279 11. Simon, M. J., Lagergren, E.S., Snyder, K.A., Concrete Mixture Optimization Using Statistical Mixture Design Methods, Proceedings of the PCI/FWHA International Symposium on High Performance Concrete, New Orleans, Louisiana, October 20-22, 1997, 230-244 12. Lawler, J.S., Connolly, J.D., Krauss, P.D., Tracy, S.L., Janney, W., Ankenman, B., Supplementary Cementitious Materials to Enhance Durability of Concrete Bridge Decks, NCHRP Report for Project 18-08A, National Cooperative Highway Research Program, Decemeber, 2005, pp. 90 13. Indiana Department of Transportation (INDOT), Standard Specifications 2008 –Section 500, Indianapolis, Indiana, 2008 14. Thier, T., Examining the Time and Depth of Saw-Cutting Guidelines for Concrete Pavements, M.Sc. Thesis, Purdue University, 2004, pp.170 15. Meyers, R. H., Montgomery, D.C., Response Surface Methodology: Process and Product in Optimization Using Designed Experiments, 1st Ed.,Wiley, 1995, New York, pp. 824 16. Lu, A., Durability Design of High Performance Concrete and its Application in Bridge Decks, Ph.D. Thesis, Purdue University Graduate School, 2002, Purdue University, Indiana, pp. 286 17. Rudy, A., Development of Optimum Proportions for Concrete Paving Mixtures, Indiana DOT Report, Indiana, 2009, pp. 270 (in progress) 18. Michigan Department of Transportation (MDOT), Standard Specifications for Construction, http://mdotwas1.mdot.state.mi.us/public/specbook/, 2003 19. Ontario Provincial Standards for Roads and Public Works (OPS), Ontario Provincial Standard Specification: Section 1352: Material Specification for Precast Concrete Barriers, November 1989, p. 4 20. Radlinska, A.Z., Reliability-based Analysis of Early Cracking in Concrete, Ph.D. Thesis, Purdue University Graduate School, 2008, Purdue University, Indiana, pp. 190 21. IM 529, Portland Cement (PC) Concrete Proportions, Iowa Department of Transportation (IDOT), Office of Materials, 2008, 1-7

17-Durability.qxd

9/19/2009

11:59 AM

Page 161

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

DURABILITY PERFORMANCE OF ROLLER COMPACTED CONCRETE USING FLY ASH Jong-Pil WON1,*, Chang-Il JANG1, Sang-Woo LEE1, and Wan-Young KIM2 Civil & Environmental System Engineering, Konkuk University, Seoul, 143-701, Republic of Korea, e-mail: [email protected] 2 Dam Engineering Research Center, Korea Institute of Water and Environment, 305-730, Republic of Korea 1,*

ABSTRACT This study investigated the durability performance of roller-compacted concrete (RCC) with fly ash. The effects of long-term durability of RCC adding fly ash was studied. In this study, fly ash replaced 20, 30, 40 and 50% by mass of the cement. Laboratory tests of the chloride ion permeability, abrasion, drying shrinkage were conducted. The test results demonstrated that 30% fly ash replacement has excellent durability properties.

Keywords Durability, fly ash, Roller Compacted Concrete, shrinkage

INTRODUCTION Roller-compacted concrete (RCC) dams combine the advantages of earth-fill and concrete dams in construction. They have gained acceptance worldwide in a relatively short time owing to their low cost, which is derived in part from their rapid construction method (ACI, Nagataki et al., Hirose and Yanagida) [1,2,3]. RCC has recently emerged as an economically attractive material for dam construction, replacing the use of conventional concrete and even challenging the economics of earth-fill and rock-fill embankment dams (Logle, Hall and Houghton, Hansen and Reinhard, Choi et al.) [4,5,6,7]. Many studies of the methods of construction and mix-proportions of RCC dams have been performed (Reed et al., JCI, US Army Corps) [8,9,10]. The U.S. Army Corps of Engineers began a concerted effort to develop RCC for use in building concrete dams in the early 1970s (Hansen and Reinhard) [6]. The Corps built field test sections in Jackson, Mississippi, in 1972, and at the site of the Lost

17-Durability.qxd

9/19/2009

11:59 AM

162

Page 162

Jong-Pil WON, Chang-Il JANG, Sang-Woo LEE and Wan-Young KIM

Creek Dam in Oregon in 1973 (Hansen and Reinhard) [6]. The field tests confirmed the basic construction method and provided information on the material properties and the strength of the bonds between successive layers of RCC. Based on the data developed in these tests, a mix and a construction method were proposed. The mix and design method of the U.S. Army Corps of Engineers was directly applied to the mix for a RCC dam. In Japan, because of the seismic, hydrologic, and topographic problems associated with most dam sites in the country, researchers have taken a more conservative approach to RCC mix proportions. Their aim is for a product with the same quality and appearance as that of conventionally placed, mass concrete gravity dams. Various studies and experiments have led to a RCC method. Obtaining the mix for Japan required repeated experiments (JCI, US Army Corps) [9,10]. This method obtained quality concrete but required a long time. Although many studies have evaluated mix proportions and construction methods, few have analyzed the durability of RCC, which is very important. The various factors that can reduce the long-term performance of a RCC dam include abrasion, permeability of the construction joints. This study examined these durability properties so as to promote the practical application of fly ash in RCC construction. Therefore, the durability properties of RCC using increasing amounts of fly ash were evaluated.

MATERIALS AND TEST METHOD

Cement and aggregate ASTM Type II Portland cement was used. The chemical composition of the cement is shown in Table 1. The fine and coarse aggregates were equivalent to rock. Crushed coarse aggregate with a maximum size of 80 mm and crushed fine aggregate with a specific gravity of 2.61 were used. The grade distributions of the coarse and fine aggregates are listed in Table 2. Table 1 Chemical composition of Type II portland cement SiO2

Al2O3

Fe2O3

CaO

MgO

Na2O

K2O

SO3

Lg.Loss

23.3

3.9

3.6

62.4

2.9

0.11

0.82

1.9

1.0

17-Durability.qxd

9/19/2009

11:59 AM

Page 163

163

Durability performance of Roller Compacted Concrete using fly ash

Table 2 Grade distribution of coarse aggregate Size(mm)

Max. dimension(mm)

(%) 40~80

20~40

0~20

37%

32%

31%

80

Fly ash Fly ash was refined to satisfy the requirements of ASTM C 618. The physical and chemical characteristics of the fly ash are shown in Table 3. Table 3 Physical and chemical properties of fly ash Physical properties Specific gravity

Blaine fineness(cm2/g)

Absorption (%)

L.O.I (%)

2.14

3,400

0.13

3.28

Chemical composition (%) SiO2

Al2O3

Fe2O3

CaO

MgO

Na2O

K2O

TiO2

Mix proportion The basic mix proportions were determined for unit amount of water (95kg/m3), unit amount of cement (115kg/m3), and S/a (35%). The fly ash replacement by cement weight was evaluated based on the durability of the RCC mixture. The mix proportions are given in Table 4. Table 4 Mix proportions of RCC Type of mixture

Gmax (mm)

Vc (sec)

Air (%)

w/c

S/a

W

C 3

S 3

G 3

Fly ash 3

(%) (kg/m ) (kg/m ) (kg/m ) (kg/m ) (kg/m ) (kg/m3)

FA0

117

0

FA20

93.6

23.4

FA30

80

20±10

1.5± 1

0.79

SP

3

34

92

81.9

765

1468

35.1

FA40

70.2

46.8

FA50

58.5

58.5

0.33

17-Durability.qxd

9/19/2009

11:59 AM

164

Page 164

Jong-Pil WON, Chang-Il JANG, Sang-Woo LEE and Wan-Young KIM

Chloride ion permeability test The chloride ion penetration tests were conducted in accordance with ASTM C 1202-94. The dimensions of the test specimens were 150×50 mm. The specimens were tested at 7, 28, and 91 days of age. Abrasion test Abrasion tests were performed to evaluate the surface resistance of the RCC concrete. The tests were conducted according to ASTM C 944. Concrete specimens measuring 150×60 mm were cast. The specimens were tested at 7, 28, and 91 days of age. Drying shrinkage test The drying shrinkage of a RCC mixture is the volume reduction of the cement resulting from its hydration shrinkage after it starts to set. These tests were conducted in accordance with the JCI procedure (Tazawa) [11]. They were performed because the RCC mixture was applied to a large section. The drying shrinkage was measured over 28 days at 60% relative humidity and 23oC. Figure 1 shows the drying shrinkage test apparatus setup.

RESULTS AND DISCUSSIONS Chloride ion permeability The results of the chloride ion permeability tests of RCC mixed with fly ash replacement are shown in Fig. 2. The mixture without fly ash showed the lowest permeability initially, but the permeability of the mixture with fly ash gradually decreased, and was lowest after 91 days of curing. This is because fly ash is a pozzolan material that increases the long-term strength of the RCC mixture, thereby increasing the water tightness of the mixture.

17-Durability.qxd

9/19/2009

11:59 AM

Page 165

Durability performance of Roller Compacted Concrete using fly ash

(a) Drying shrinkage test frame

(b) Drying shrinkage test set-up

Fig.1 Test frame and set-up of drying shrinkage test for RCC

165

17-Durability.qxd

9/19/2009

11:59 AM

166

Page 166

Jong-Pil WON, Chang-Il JANG, Sang-Woo LEE and Wan-Young KIM

10000 9000 8000

Coulombs

7000 6000 5000 4000 3000 2000

0

1000

20

30

40

50

0 7

28

91

C uring age(days)

Fig.2 Chloride ion permeability with fly ash replacement ratio Abrasion The abrasion resistance test results are shown in Fig. 3. The RCC mixture had a relatively low abrasion resistance initially, but subsequently the abrasion resistance increased. The highest abrasion resistance was recorded after curing for 91 days. The RCC mixture with 30% fly ash replacement had a higher abrasion resistance than the other mixtures. 0.040

7days

28days

91days

2 Abraion loss(g/c )

0.035 0.030 0.025 0.020 0.015 0.010 0.005 0.000 0

20

30

40

50

Fly ash content(% )

Fig.3 Abrasion resistance with fly ash replacement ratio Drying shrinkage The characteristics of drying shrinkage are shown in Fig. 4. The RCC mixture using fly ash had better drying shrinkage control properties than did the RCC mixture without fly ash. Also, as the fly ash replacement increased, the drying shrinkage decreased. In particular, the drying

17-Durability.qxd

9/19/2009

11:59 AM

Page 167

167

Durability performance of Roller Compacted Concrete using fly ash

shrinkage of the RCC mixture with 30% fly ash replacement was less than that of the RCC mixture without fly ash.

-4 Drying shrinkage(ßĪ10 mm)

3.5

3

2.5

2

1.5

0

20

30

40

50

1

0.5

0 0

5

10

15

20

25

30

Tim e(day)

Fig.12 Drying shrinkage with fly ash replacement ratio

CONCLUSIONS The influence of different proportions of fly ash on the behavior of RCC was investigated. Based on the results presented in this paper, the following conclusions can be drawn. · The mixture without fly ash initially showed the lowest permeability, but the permeability of the mixtures with fly ash gradually decreased was lowest after 91days. · Initially, the RCC mixtures had relatively low abrasion resistance, but as their ages increased, their abrasion resistance increased. The RCC mixture with 30% fly ash replacement was more effective than the other mixtures. · As the amount of fly ash replacement increased, the drying shrinkage decreased. Also, the drying shrinkage of the RCC mixture with a 30% fly ash replacement was less than that of the RCC mixture without fly ash. REFERENCES 1. American Concrete Institute Committee 207, Roller Compacted Mass Concrete, ACI 207.5R-89, Detroit, Michigan, 1988

17-Durability.qxd

9/19/2009

11:59 AM

168

Page 168

Jong-Pil WON, Chang-Il JANG, Sang-Woo LEE and Wan-Young KIM

2. Nagataki, S., Yangida, T., and Okumur, T., Construction of Recent RCD-concrete Dam Projects in Japan, Proceedings of the Symposium sponsored by the Colorado Section and Construction Division of the American Society of Civil Engineers in conjunction with the ASCE Convention in Denver, Colorado, 1985, 62-70 3. Hirose, T. and Yanagida, T., Burst of Growth Demands Speed Economy, Concrete International, 1984, 14-18 4. Logle, C. V., Economic considerations in selection of a Roller Compacted Concrete Dam, Proceedings of the Symposium sponsored by the Colorado Section and Construction Division of the American Society of Civil Engineers in conjunction with the ASCE Convention in Denver, Colorado, 1985, 111-122 5. Hall, D. J. and Houghton, D. L., Roller Compacted Concrete Studies at Lost Creek Dam, U.S. Army Engineer District, Portland, USA, 1974 6. Hansen, K. D. and Reinhard, W. G., Roller-Compacted Concrete, McGraw-Hill, Inc., 1991 7. Choi, Y., Neighbors, J. D., and Reichler, J. D., Cold Weather Placement of RCC, J. of Materials in Civil Engineering, 2003, 118-124 8. Reed, P. A., Neeley, B. D., Green, M. L., and Amundson, C. T., Roller Compacted Concrete (RCC) Dams; Report 1, Laboratory Characterization of RCC Cores from Elk Creek Dam, Trail Oregon, Department of the army U.S. Army Corps of Engineers, Washington, DC, USA, 1998 9. Japan Cement Association, RCD Construction Method Committee, Report on Construction Example of RCD Dams, 1996 10. US Army Corps of Engineers, EM 1110-2-2006-Roller-Compacted Concrete, Department of the army, U.S. Army Corps of Engineers, Washington DC, USA, 2000 11. Tazawa, E., Autogenous Shrinkage of Concrete, E F&N Spon, 1999

18-Structure and.qxd

9/19/2009

11:59 AM

Page 169

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

STRUCTURE AND PROPERTIES OF NaOH ACTIVATED CEMENT FREE BINDER (CFB) CONCRETES Deepak RAVIKUMAR, Sulapha PEETHAMPARAN, Narayanan NEITHALATH* Department of Civil and Environmental Engineering, Clarkson University, Potsdam NY 13699 *

Corresponding Author; E-mail: [email protected]

ABSTRACT Increasing emphasis on sustainability of the built environment has resulted in attempts to drastically reduce the cement consumption in concrete and replace it with waste/recycled materials. This study reports the development of concretes with cement free binders (CFB) and the evaluation of their properties. A Class F fly ash and a ground granulated blast furnace slag (GGBFS) are used as the binding materials. The activating agent used in this study is sodium hydroxide (NaOH), at concentrations ranging from 6 M to 10 M. The optimal temperature and curing duration required to achieve desirable compressive strengths of CFB concretes are reported. The influence of the binding material and the concentration of the activator on the compressive strength and porosity of the CFB concretes are studied. The compressive strength of CFB concretes with fly ash as the binding material increases with increase of activating solution concentration but for CFB concretes with GGFBS as the binder, activation with 8 M NaOH is seen to result in the highest compressive strength. The strength-porosity relationship of CFB concretes shows an exponential trend, similar to that of conventional cement based materials. Microstructure and the phase composition of the reaction products are also discussed.

Keywords: Fly ash, ground granulated blast furnace slag (GGBFS), strength, porosity, sodium hydroxide (NaOH)

INTRODUCTION The production of Portland cement consumes tremendous amounts of natural materials and energy, and emits significant amounts of CO2 into the atmosphere. Traditionally, reduction in cement consumption has been attained by the use of industrial by products such as fly ash and ground granulated blast furnace slag (GGBFS) as supplementary cementing materials. However, with increasing emphasis on sustainability worldwide, researchers have tried to use industrial byproducts such as fly ash and GGBFS as sole binding materials in concrete instead of partial replacement of ordinary Portland cement. Geopolymer concretes, also known as alkali activated concretes, inorganic polymer concretes, or cement free binder (CFB) concretes is a consequence of this approach. Industrial by products that are rich in silica and alumina could be activated by alkalis to produce cement free binder (CFB) concretes. The alumina and silica from these materials, when subjected to highly alkaline conditions, undergoes polymerization to form alkali alumino-silicate 1

18-Structure and.qxd

170

9/19/2009

11:59 AM

Page 170

Deepak RAVIKUMAR, Sulapha PEETHAMPARAN and Narayanan NEITHALATH

gel (Davidovits [1, 2], Mozgawa and Deja [3]). This reaction of alumina and silica with the alkali is sometimes termed as geopolymerization. Sodium hydroxide and sodium silicate are the more commonly used alkaline activating agents (Hardjito and Rangan [4]) even though few studies have also been carried out with potassium hydroxide or sodium carbonate as the activator (Kong et al. [5], Shi [6]). The formation of alkali aluminosilicate gel, the reaction product of alkali activation of fly ash or GGBFS, requires heat. Temperatures ranging from 40oC to 90oC have been reported in order to produce CFB concretes with appreciable mechanical properties (Chinaprasirt et al. [7], Provis and Deventer [8], Panias et al. [9], Hardjito and Rangan [4]). The binder and activating agent compositions are the most important parameters that influence the properties of CFB concretes (Criado et al. [10], van Jaarsveld [11]). Previous studies have shown that the amount of SiO2 and Al2O3 present in the binder plays a significant role in the properties of CFB concretes (Silva et al [12], Duxon et al [13]). The presence of calcium oxides in the source material also influences the properties of CFB concretes such as the early age strength and setting times (Temuujin et al. [14]). Curing temperature and duration also are important factors influencing the properties of CFB concretes since the alkaline activation is an endothermic reaction (Duxon et al. [13]). This paper focuses on understanding the influence of the binder material (fly ash or GGBFS), concentration of the alkaline activator (NaOH solution), and the activator-to-binder ratio on the mechanical (compressive strength), physical (porosities) and microstructural (morphology and phase composition) properties of CFB concretes. The compressive strengths and porosities of CFB concretes made using the above mentioned source materials, NaOH concentrations of 6, 8, or 10 M, and activator-to-binder ratios of 0.40, 0.50, and 0.60 are evaluated. Microstructure of the reaction products are also studied with Scanning electron microscopy and X-Ray diffraction patterns.

EXPERIMENTAL PROGRAM Materials Two industrial by products, Class F fly ash conforming to ASTM C 618 and Ground granulated blast furnace slag (GGBFS) Type 100 conforming to ASTM C 989 were used in this research to produce CFB concretes. The chemical composition of the fly ash and GGBFS are given in Table 1. Both binders are rich in silica and alumina, which are the basic components required to form the strength imparting binder phase in CFB concretes. The silica-to-alumina (SiO2/Al2O3) ratios were found to be approximately 2.25 and 3.41 for fly ash and GGBFS respectively. It is also seen that GGBFS has a high CaO content of 40% while the CaO in fly ash was negligible (1.5%). X-ray diffraction patterns of the source materials are shown in Figure 1 and the particle morphologies obtained using scanning electron microscopy (SEM) are shown in Figure 2. The XRD patterns show that GGBFS is amorphous suggesting better reactivity as compared to fly ash which contains crystalline phases of silica and alumina. From the micrographs it is evident that the fly ash particles are spherical, and GGBFS particles are irregular. Analytic reagent-grade sodium hydroxide crystals were used to prepare the alkaline activating solutions of varying molar concentrations.

2

18-Structure and.qxd

9/19/2009

11:59 AM

Page 171

Structure and properties of NaOH activated cement free binder (CFB) concretes

Table 1: Chemical composition and physical characteristics of the materials used

SiO2

Fly Ash (%) 59.6

Al2O3

26.5

10.5

Fe2O3

5.3

0.67

Chemical composition

GGBFS (%) 36.0

CaO

1.5

39.8

MgO

-

7.93

Na2O

-

0.27

K2O

-

0.08

SO3

0.10

2.11

Loss on Ignition (LOI) 2.39 Fineness (% passing, 80 (45 ȝm) Sieve Size)

3.00 -

Figure 1: X-ray pattern of Fly ash and GGBFS powder (Q: Quartz, M: Mullite)

3

171

18-Structure and.qxd

172

9/19/2009

11:59 AM

Page 172

Deepak RAVIKUMAR, Sulapha PEETHAMPARAN and Narayanan NEITHALATH

(a)

(b)

Figure 2: Scanning electron micrographs of: a) Fly ash, b) GGBFS Mixing Procedure Prior to the preparation of the concrete mixture, NaOH solutions of the desired concentrations were prepared and allowed to stay at room temperature for two hours. The mixing procedure involved initial mixing of the binding material and the aggregates for two minutes. NaOH solution of the desired concentration was then added, and mixed for a further four minutes for CFB concretes with fly ash and two minutes for CFB concretes with GGBFS to obtain a uniform mixture. The reason for using only two minutes of mixing for the GGBFS mixtures was the faster rate of setting of NaOH activated GGBFS. Concrete was filled in cubes of 50 mm x 50 mm x 50 mm size and vibrated using a vibration table until proper compaction was achieved. GGBFS mixtures were vibrated for 2 minutes as opposed to fly ash mixtures which needed only 1 minute of vibration to achieve adequate compaction. For the paste specimens to be used for microstructural analysis, samples were prepared by mixing the activating solution and the binder in a blender. The pastes were then poured into circular discs of 100 mm diameter and 12.5 mm depth. Both the paste and concrete specimens were removed from the molds after 24 hours and subjected to heat curing in laboratory oven at either 60oC or 75oC for 24 hours or 48 hours. Determination of Optimal Binder Contents and Mixture Proportions In order to obtain the binder content (by volume) that provides the highest compressive strength, CFB concretes were proportioned using fly ash and GGBFS contents ranging from 15% to 30% by volume of concrete. For this study, NaOH solution of 8 M concentration (prepared by dissolving 320 g of NaOH in 1 liter of water) was used as the activating agent at an activator-tobinder mass ratio of 0.40. The specimens were heat cured at 75o C for 48 hours. Figure 3 shows the compressive strengths of the CFB concretes as a function of either fly ash or GGBFS contents. It can be seen that 18% fly ash content by volume and 25% GGBFS content by volume of the concrete mixture respectively provided the maximum compressive strengths in this study. These binder volumes were used to proportion concretes for further experiments. The mixture proportions of CFB concretes containing fly ash or GGBFS as the sole binder, activated by NaOH solution of different concentrations, and proportioned with different activator-to-binder ratios are shown in Table 2. A constant coarse aggregate to fine aggregate ratio of 1.13 was used for all the mixtures. The total aggregate volume in the CFB concretes ranged from 65 to 70 % for CFB concrete with fly ash as binder and 55 to 60% for CFB concretes with GGBFS as binder. 4

18-Structure and.qxd

9/19/2009

11:59 AM

Page 173

173

Structure and properties of NaOH activated cement free binder (CFB) concretes

Compressive Strength (MPa)

40 CFB concrete with Fly ash CFB concrete with GGBFS

35

30

25 % (GGBFS)

25

20

18 % (FA)

15 12

16

20

24

28

32

Binder content by volume (%)

Figure 3: Optimum binder content for CFB concretes Table 2: Mixture proportions of the CFB concretes used in this study

Mix No.

Binder type

Binder NaOH Activator to content Concentration Binder ratio (kg/m3)

1

Coarse aggregate content (kg/m3)

Fine aggregate content (kg/m3)

Alkaline activator content (kg/m3)

0.40

441

931

818

176

0.50

424

923

811

212

3

0.60

438

915

804

262

4

0.40

409

932

819

163

2

5

6M

Fly Ash

8M

0.50

422

924

813

211

6

0.60

435

917

806

261

7

0.40

405

934

821

162

8

10 M

9 10 11

6M

0.50

417

927

815

208

0.60

429

920

809

257

0.40

548

796

699

219

0.50

567

785

690

283

12

0.60

585

774

680

351

13

0.40

545

797

701

218 281

14

GGBFS

8M

0.50

563

787

692

15

0.60

581

776

682

348

16

0.40

540

801

704

216

0.50

557

791

695

278

0.60

573

781

690

344

17 18

10 M

5

18-Structure and.qxd

174

9/19/2009

11:59 AM

Page 174

Deepak RAVIKUMAR, Sulapha PEETHAMPARAN and Narayanan NEITHALATH

Determination of Compressive Strength and porosities After the heat curing duration, the concretes were allowed to cool down and return to ambient temperatures. The compressive strengths of the cubes were then determined in accordance with ASTM C 109. No moist curing was adopted for any of the mixtures after the heat curing procedure. The porosities of the CFB concretes were measured in accordance with procedure described in RILEM –CPC 11.3 [15]. The 50 mm size concrete cubes were cut into two equal halves after the heat curing duration. After measuring the initial mass, the specimens were vacuum dried for 3 hours and subsequently saturated under vacuum for 1 hour. The specimens were kept submerged in water for a further 18 hours after which they were removed from water, surface dried and weighed. The increase in the mass of the specimen, represented as percentage of initial volume of the specimen, is reported as the porosity. Microstructural analysis To understand the morphology and the composition of the reaction product formed, microstructural analysis was performed using Scanning electron microscopy (SEM) coupled with Energy dispersive X-ray analysis (EDX), and X-ray diffraction (XRD). The XRD patterns of the powdered paste samples were recorded using a Bruker DX-8 diffractometer using CuKĮ radiation with a wavelength of 1.504 Ao. The tests were carried out in the 2ș range of 5-40o. Scanning electron micrographs were obtained to observe the morphology of the reaction products. Elemental composition of the reaction products were analyzed using EDX.

RESULTS AND DISCUSSIONS Influence of Curing temperature and curing duration on compressive strengths of CFB concretes activated with NaOH Heat curing was required to achieve adequate mechanical properties for CFB concretes activated with NaOH. In order to obtain the curing temperature and duration that results in maximum compressive strength, CFB concretes proportioned with an activator-to-binder ratio of 0.40, containing the optimal fly ash and GGBFS contents (18% and 25% by volume respectively), and activated using 8 M NaOH solution were subjected to temperatures of 60oC and 75oC for 12, 24 or 48 hours. Figures 4(a) and (b) show the compressive strengths of CFB concretes made using fly ash and GGBFS as the binding material, as functions of both curing temperature and curing duration. As observed from Figure 4(a), curing at 60oC did not produce appreciable compressive strengths for CFB concretes made with fly ash as the binder, irrespective of the curing duration. Increasing the curing temperature to 75oC resulted in significant improvements in compressive strength. This points to the fact that a curing temperature of 75oC is necessary to produce the sodium aluminosilicate gel necessary to impart strength to CFB concretes containing fly ash. From Figure 4(b), it can be seen that, for CFB concretes containing GGBFS, heat curing for 48 hours at 60oC produces appreciable strengths, even though they are lower than those heat cured at 75oC. The difference in compressive strengths between the specimens heat cured at 60oC or 75oC is not as distinct as that of CFB concretes made using fly ash, especially at longer curing durations. This is indicative of the influence of processing conditions on the amount and structure of the binding product (alkali aluminosilicate gel) in CFB concretes made using fly ash or GGBFS as the binder. While the presence of calcium oxide in GGBFS would have resulted in the formation of calcium bearing strength imparting components also at lower temperatures, the 6

18-Structure and.qxd

9/19/2009

11:59 AM

Page 175

175

Structure and properties of NaOH activated cement free binder (CFB) concretes

50

Compressive strength(MPa)

Compressive strength(MPa)

strength development of CFB concretes with fly ash is believed to depend on the formation of sodium alumino silicate gel alone. Based on these observations, the results reported further in this paper are for specimens heat cured at 75oC for 48 hours, since this combination of curing temperature and duration provided the highest compressive strengths.

40 30 20 10 0

48 24

75

Cur i ng

12

60

tem per atur e (o C)

g rin Cu

io rat Du

) urs Ho n(

50 40 30 20 10 0 75

Cur i ng

(a)

u (Ho on ati 48

24 12

60

tem per atur e (o C)

g rin Cu

r Du

rs)

(b)

Figure 4: Influence of curing temperatures and durations on the compressive strengths of CFB concretes with: a) Fly ash, and b) GGBFS as binder

Influence of activator-to-binder ratio and NaOH molar concentration on compressive strengths of CFB concretes The influence of activator-to-binder ratio on the compressive strengths of the CFB concretes containing fly ash and GGBFS are shown in Figures 5 (a) and 5 (b) respectively. Compressive strength decreases with increase in activator-to-binder ratio even though a higher activator-tobinder ratio effectively contains a higher amount of the activating alkali. This shows that the influence of porosity (created as a result of increased amounts of the solution) was more prominent than the effect of increased NaOH concentration on the compressive strength. The activation of GGBFS with NaOH resulted in concretes of higher compressive strengths than those produced by the activation of fly ash. The higher reactivity and self cementing property of GGBFS could partly be the reason for the higher compressive strength of CFB concretes made using GGBFS. It needs to be also remembered that a higher volume of GGBFS could be incorporated into the CFB concrete mixtures for optimal compressive strength, thereby contributing to a portion of the increase strength. As could be observed from Figure 3, this study focused on obtaining maximum compressive strengths using fly ash and GGBFS as the binder, and not on obtaining identical strengths.

7

18-Structure and.qxd

176

9/19/2009

11:59 AM

Page 176

Deepak RAVIKUMAR, Sulapha PEETHAMPARAN and Narayanan NEITHALATH

50 6 M NaOH 8 M NaOH 10 M NaOH

40

30

20

10

Compressive strength (MPa)

Compressive strength (MPa)

50

0

6 M NaOH 8 M NaOH 10 M NaOH

40

30

20

10

0 0.4

0.5

0.6

0.4

Activator to binder ratio

0.5

0.6

Activator to binder ratio

(a) (b) Figure 5: Compressive strength of CFB concretes with a) Fly ash b) GGBFS as binder For CFB concretes made with fly ash as the binder, the compressive strengths obtained by activation using 8 M and 10 M NaOH were quite similar, suggesting that the use of 8 M NaOH resulted in necessary activation of the fly ash. CFB concretes with GGBFS showed the maximum compressive strength at 8 M NaOH concentration. There was a reduction in compressive strength of CFB concretes with GGBFS when activated with 10 M NaOH and the compressive strengths were similar to the concretes produced by activation using 6 M NaOH. The reason for this trend was due to the high reactivity of GGBFS which made the mixture non workable at high concentration of alkalis (10 M NaOH), thus disrupting proper finishing and compaction operations of the concrete. The reason for the early stiffening of the mixture could also be due to the increased amounts of carbonation in mixtures containing higher amounts of alkali and Calcium oxide, thus increasing the chances of formation of calcite at very early times that produces pseudo setting. A similar observation has been reported by Astutiningsih and Liu [16]. The lower difference in strengths between CFB concretes made with GGBFS activated by either 6 M or 8 M NaOH as compared to that of CFB concretes with fly ash points to the lesser influence of activator concentration (within the ranges studied) when GGBFS is used as the binder. The combined influence of activator concentration and activator-to-binder ratio on the compressive strength is represented in Figures 6(a) and (b) for CFB concretes made with fly ash and GGBFS respectively. These plots help in arriving at the activator-to-binder ratios or the NaOH concentrations required to obtain a particular compressive strength when heat cured at 75oC for 48 hours. It is observed that the compressive strength of CFB concretes with fly ash increases with an increase in activating solution concentration and a decrease in activator-tobinder ratio. For CFB concretes with GGBFS, NaOH activation using 8 M NaOH solution with an activator-to-binder ratio of 0.40 provided the maximum compressive strength, because of the reason explained earlier. Controlling premature stiffening of the 10 M NaOH activated GGBFS mixture through the use of set retarding admixtures might be beneficial in obtaining higher compressive strengths, but such an approach was not adopted in this paper.

8

18-Structure and.qxd

9/19/2009

11:59 AM

Page 177

177

Structure and properties of NaOH activated cement free binder (CFB) concretes

0.60

0.56 0.54 15

20

0.52

25

25

0.50 0.48

25

15

20

0.46 0.44 0.42

30 25

15

30

20

25

0.58

20

10

Activator to Binder ratio

Activator to Binder ratio

0.60

20 15

0.58

25

0.56 0.54

25 30

0.52 0.50

25

30 30

0.48 30 0.46

35

35

30

0.44 35

0.42

40

40

35

0.40

0.40 6.0

6.5

7.0

7.5

8.0

8.5

9.0

9.5

6.0

10.0

6.5

7.0

7.5

8.0

8.5

9.0

9.5

10.0

NaOH molar concentration

NaOH molar concentration

(a)

(b)

Figure 6: Influence of NaOH concentration and activator to binder ratio on compressive strength of CFB concretes with: a) Fly ash, and b) GGBFS as the binder The influence of activator-to-binder ratio and the activator concentration on the porosities of CFB concretes is discussed in this section. Figure 7 shows the porosity values for CFB concretes made using fly ash or GGBFS as the binder, and activated with different concentrations of NaOH at various activator-to-binder ratios. The porosities were determined using a vacuum saturation method as explained in previous section. From Figures 7(a) and (b), it can be observed that the porosity of CFB concretes increases with increase in the activator-to-binder ratio. From Figure 7(a) it can be seen that the porosities of CFB concrete with fly ash activated using 6 M NaOH exhibits the highest porosity values. The porosities of CFB concretes with fly ash activated using 8 M NaOH and 10 M NaOH were found to be similar, which is in line with their compressive strength results. This shows that when fly ash is used as the binding material, increasing the activator concentration to 10 M NaOH does not result in increased amounts of the alkali alumino silicate gel that fills space and imparts strength to the CFB concretes. Figure 7(b) shows the porosity of CFB concrete made using GGBFS as the binder. The porosity is found to be lowest for the CFB concrete activated with 8 M NaOH solution and highest for the CFB concrete activated with 10 M NaOH solution. As discussed earlier, CFB concretes with GGBFS activated with 10 M NaOH resulted in an unworkable mixture. It is postulated here that if the workability of these concretes were increased using a high range water reducing admixture, and the faster setting of CFB mixture with GGBFS controlled using a retarding agent, the porosity of these mixtures could have been lower. However, such an attempt was not carried out here because all the other mixtures did not contain any chemical admixtures, and the influence of chemical admixtures in combination with a high concentration of NaOH is not well studied.

9

18-Structure and.qxd

178

9/19/2009

11:59 AM

Page 178

Deepak RAVIKUMAR, Sulapha PEETHAMPARAN and Narayanan NEITHALATH

18

18 6M NaOH 8M NaOH 10M NaOH

6M NaOH 8M NaOH 10M NaOH

16

14

Porosity (%)

Porosity (%)

16

12 10 8

14 12 10 8

6

6 0.4

0.5

0.6

0.4

Activator to binder ratio

0.5

0.6

Activator to binder ratio

(a) (b) Figure 7: Porosities of CFB concretes with a) Fly ash b) GGBFS as binder The relationship between compressive strengths and porosities for all the CFB concretes (made using either fly ash or GGBFS as the binder, and with varying activator concentrations and activator-to-binder ratios) is shown in Figure 8. The porosity values ranged from 7 to 16% which is slightly higher than the porosities of typical conventional concretes. The relationship showed an exponential trend similar to that of conventional concretes. The strength-porosity relationship showed a better correlation for CFB concretes with compressive strength greater than 20 MPa.

Compressive strength,fc (MPa)

50 Fly ash CFB concrete GGBFS CFB concrete

fc= 131.0 e -0.14 I R² = 0.625

40

30

20

10

0 6

8

10

12

14

16

18

Porosity,) (%)

Figure 8: Compressive strength-porosity relationship for CFB concretes Microstructure of CFB pastes and reaction products Scanning electron microscopy (SEM) was used to observe the morphology of the reaction products in CFB concretes, while XRD was used to identify the phases present. Figure 9 (a) shows SEM micrograph of CFB pastes with fly ash as the binder and Figure 9 (b) shows the EDX spectrum corresponding to this image. Figure 9 (a) shows the presence of many unreacted 10

18-Structure and.qxd

9/19/2009

11:59 AM

Page 179

Structure and properties of NaOH activated cement free binder (CFB) concretes

179

particles of fly ash. It is also seen that the reaction product forms a shell around the fly ash particle. The alkali activator attacks the fly ash particle, resulting in dissolution of the alumina and silica phases of the fly ash. These phases then react with the alkalis from the activator to form the gel that condenses on the surface of the fly ash particles forming the shell as observed in Figure 9(a). A similar mechanism has been proposed for alkali activation of fly ash in Fernandez-Jimenez and Criado [17]. Figure 10 (a) shows the micrograph of CFB paste with GGBFS as the binder and Figure 10 (b) shows the EDX spectrum corresponding to this image. Uniform distribution of the reaction product around the starting material is not observed in this case. The EDX spectra for both the pastes suggest that the reaction product formed is a form of sodium aluminosilicate. The EDX spectrum for CFB paste with GGBFS show increased amount of calcium in the reaction product, confirming that some type of calcium alumino silicate might also have formed that can influence the material properties.

(a) (b) Figure 9: a) SEM micrograph of CFB paste with fly ash, activated by 8 M NaOH (activator-tobinder ratio of 0.4), and b) EDX analysis of this paste

(a) (b) Figure 10: a) a) SEM micrograph of CFB paste with GGBFS, activated by 8 M NaOH (activatorto- binder ratio of 0.4), and b) EDX analysis of this paste Figure 11 shows the XRD patterns of CFB pastes with fly ash or GGBFS as the binder. The CFB paste with fly ash is seen to be much more crystalline, owing to the presence of crystalline phases such as quartz and mullite in the fly ash which were shown in Figure 1 also. The presence of crystalline hydroxy sodalite peaks at 2ș of 12o and 24o shows the presence 11

18-Structure and.qxd

180

9/19/2009

11:59 AM

Page 180

Deepak RAVIKUMAR, Sulapha PEETHAMPARAN and Narayanan NEITHALATH

sodium aluminosilicate crystals. The formation of crystalline sodium aluminosilicate is generally seen when the source material is activated using alkali hydroxides without the addition of any external source of silicate activator (Criado et al. 2007 [10]) which is the case in this study. Sodium alumino silicate may also be present in other amorphous forms, but cannot be detected using XRD. The XRD pattern for the CFB paste with GGBFS as the binder is mostly amorphous. At a 2ș of 29o, a calcite peak is observed, that suggests carbonation. As explained earlier, it is possible that carbonation can occur during the mixing of concrete containing very high amounts of alkalis to a binder that contains significant amounts of calcium oxide (Astutiningsih and Liu [15]), which might result in premature stiffening. No sodalite peaks are observed for the CFB paste with GGBFS, indicating that the reaction products formed are different for both the starting materials even though it is a form of sodium aluminosilicate.

Figure 11: XRD pattern of Fly ash and GGBFS activated with 8 M NaOH (Q: Quartz, M: Mullite, N: Hydroxy Sodalite - Na6(Si6Al6O24).8H20, C:Calcite)

CONCLUSIONS This study shows the effective use of fly ash or GGBFS alone as starting materials to prepare CFB concretes with good mechanical properties. Compressive strengths up to 45 MPa was achieved. The optimum binder content that provided the maximum compressive strength was 18 % by volume of concrete for mixtures using fly ash as the binder and 25% by volume of concrete for mixtures using GGBFS as the binder. Heat curing was required for the activation of the binder with sodium hydroxide, and a curing temperature of 75oC and a curing duration of 48 hours resulted in maximum compressive strengths of CFB concretes. CFB concretes with fly ash showed lower compressive strengths when heat cured at 60oC irrespective of curing duration whereas CFB concretes with GGBFS showed reasonable compressive strengths even at lower curing temperatures. CFB concretes with GGBFS as the binder resulted in higher compressive strengths and lower porosities compared to CFB concretes with fly ash as the binder. The difference between compressive strengths of CFB concretes with fly ash as the binder and activated with 8 M or 10 M NaOH was not very significant. For CFB concrete with GGBFS as the binder, activated using 8 M NaOH resulted in the highest compressive strength. Increasing the activator concentration to 10 M NaOH resulted in an unworkable mixture possibly due to the 12

18-Structure and.qxd

9/19/2009

11:59 AM

Page 181

Structure and properties of NaOH activated cement free binder (CFB) concretes

181

reaction between the activator and calcium oxide present in GGBFS. Porosity and compressive strength of CFB concretes were found to be exponentially related, similar to that of conventional concretes. Microstructural investigations showed that the alkali aluminosilicate gel was formed as a shell around the fly ash particles whereas a more homogeneous distributed reaction product was found in CFB paste with GGBFS. The EDX spectra along with the XRD pattern showed that sodium aluminosilicate was formed in CFB pastes with fly ash as the binder. The CFB pastes with GGBFS showed no crystalline peaks, indicating the formation of more amorphous reaction products. REFERENCES 1. Davidovits, J., Geopolymers: Inorganic Polymeric New Materials. Journal of Thermal Analysis,37,1991,1633-1656 2. Davidovits, J., Geopolymer chemistry and sustainable development. The Poly (sialate) terminology: a very useful and simple model for the promotion and understanding of greenchemistry. In: Proceedings of the World Congress Geopolymer Saint Quentin, J. Davidovits eds. France, 28 June–1 July.2005, pp. 9-15. 3. Mozgawa, W., Deja, J., Spectroscopic studies of alkaline activated slag Geopolymers. Journal of Molecular Structure, doi:10.1016/j.molstruct.2008.12.026, 2009. 4. Hardjito, D., Rangan, B. V., Development and properties of low-calcium fly ash-based geopolymer concrete. Curtin University of Technology Research Report, Perth, Australia, 2006. 5. Kong, D.L.Y., Sanjayan, J.G., Damage behavior of geopolymer composites exposed to elevated temperatures. Cement & Concrete Composites, 30, 2008, 986–991 6. Shi, C., Strength, Pore structure and permeability of alkali-activated slag mortars, Cement and Concrete Research,26,1996,1789-1799 7. Chindaprasirt, P., Chareerat, T., Sirivivatnanon, V., Workability and strength of coarse high calcium fly ash geopolymer. Cement & Concrete Composites, 29, 2007, 224-229 8. Provis, J.L., van Deventer, J.S.J., Geopolymerisation kinetics. 2. Reaction kinetic modeling. Chemical Engineering Science, 62, 2007, 2318-2329 9. Panias D., Giannopouloua I.P., Perraki T., Effect of synthesis parameters on the mechanical properties of fly ash-based Geopolymer. Journal of Colloids and Surfaces A: Physicochem. Eng. Aspects,301,2007,246-254 10. Criado, M., Fernandez-Jimenez, A., de la Torre, A., Aranda, M.A.G., Paloma, A., An XRD study of the effect of SiO2/Na2O ratio on the alkali activation of fly ash. Cement and Concrete Research, 37, 2007,671-679 11. van Jaarsveld, J.G.S., van Deventer, J.S.J., Lukey, G., The characterisation of source materials in fly ash-based geopolymers. Materials Letters, 57, 2003, 1272–1280 12. Silva, P. D., Crenstil, K. S., Sirivivatnanon, V., Kinetics of geopolymerization: Role of Al2O3 and SiO2. Cement and Concrete Research, 37, 2007, 512-518 13. Duxson, P., Provis, J.L., Luckey, G.C., Mallicoat, S.W., Kriven, W.M., van Deventer, J.S.J., Understanding the relationship between geopolymer composition, microstructure and mechanical properties. Journal of Colloids and surfaces A: Physiochem.Eng.Aspects,269, 2005, 47-58 14. Temuujin, J., van Riessen, A., Williams, R., Influence of calcium compounds on the mechanical properties of fly ash geopolymer pastes. Journal of Hazardous Materials, doi:10.1016/j.jhazmat.2008.12.121, 2009. 13

18-Structure and.qxd

182

9/19/2009

11:59 AM

Page 182

Deepak RAVIKUMAR, Sulapha PEETHAMPARAN and Narayanan NEITHALATH

15. RILEM CPC11.3. Absorption of water by immersion under vacuum. Materials and Structures, 17, 1984, 391-394. 16. Astutiningsih, S., Liu, Y., Geopolymerisation of Australian slag with effective dissolution by the alkali In: Proceedings of the World Congress Geopolymer Saint Quentin, J. Davidovits eds. France, 28 June–1 July.2005,69-73 17. Fernández-Jiménez. A., Criado, M., Microstructure development of alkali-activated fly ash cement: a descriptive model. Cement and concrete research, 35, 2004, 1204-1209

14

19-Model of.qxd

9/19/2009

11:59 AM

Page 183

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

MODEL OF CONCRETE CARBONATION AS LIMITED PROCESS – EXPERIMENTAL INVESTIGATIONS OF FLUIDAL ASH CONCRETE Lech CZARNECKI, Piotr WOYCIECHOWSKI Department of Building Materials Engineering, Faculty of Civil Engineering Warsaw University of Technology Al. Armii Ludowej 16, 00-637 Warsaw, Poland, e-mail: [email protected]

ABSTRACT The aim of the paper is to develop a theoretical model predicting carbonation depth of concrete made with fluidal ash replacement of cement. Experimentally determined model describes the depth of carbonation depending on time of exposition, w/s ratio and early-age curing conditions. A new mathematical formulation of carbonation depth as a hyperbolic function of time assumes that carbonation process is limitable in depth due to process of fulfilling concrete pores with carbonation products. The depth of carbonation in concretes made with portland cement and various quantities of fluidal ash (30%, 15%, 0% of cement mass replacement) was measured in environment with high CO2 concentration (1%). The test results are shown and used to derive equations for prediction of carbonation depth in such concretes.

Keywords Concrete, carbonation, fluidal fly ash

CONCRETE CARBONATION AND IT`S PROGRESS IN TIME Concrete carbonation can be defined as a complex physicochemical process in concrete under the long-term influence of carbon dioxide. Carbon dioxide is permanently present in atmospheric air, outside and inside the buildings. Volumetric content of CO2 in the air is ca. 0.03%, but in industrial zones and along main roads it could be up to 0.3%, or locally more, so that carbonation can progress continuously in concrete with unprotected surface [1, 2]. Main chemical mechanism of carbonation is reaction of atmospheric CO2 with calcium hydroxide from cement hydration with release of calcium carbonate and water: Ca(OH)2 + CO2 o CaCO3 + H2O Other concrete components, such as lime from hydrated aluminosilicates, can also be partially carbonated, but this process is possible only in high CO2 concentration and its influence on concrete is limited. Basic carbonation product – calcium carbonate – crystallizes in concrete pores. Small quantities of silica, alumina and ferrous oxide are also products of CSH carbonation. Rate of carbonation progress in time depends on external conditions as well as on concrete properties and technological processes of execution of concrete structures. Main negative effect of carbonation is decreasing of pH in the cover, which is one of most frequent causes of corrosion of steel reinforcement. Carbonation can be considered as one of

19-Model of.qxd

9/19/2009

11:59 AM

Page 184

184

Lech CZARNECKI and Piotr WOYCIECHOWSKI

most important factors causing destruction of concrete structures, as stated by Fiertak [4], Da Silva [3] and many other authors. The rate of carbonation is determined by different internal and external factors. Between external factors of great importance are the following: CO2 concentration, and humidity and temperature of atmosphere. Internal factors are related to concrete composition, type and content of binder and water-cement ratio, which determine concrete microstructure. Early curing of concrete plays also an important role. Generally, the rate of carbonation: - increases proportionally to square root of CO2 concentration in atmosphere; - increases with increasing of humidity up to 70-80%, while more important is humidity of concrete than air, but if pores are fully saturated with water the speed of carbonation decreases; - increases with cyclic changes of temperature and pressure; - increases with water-cement ratio; - decreases with compaction of concrete structure; - decreases with increase of cement content in concrete. Mathematical description of carbonation process in time was presented by many authors: Wieczorek [5], Slopkova [6], Papadakis [8], Ishida and Maekawa [10], Gawin [9], Lech [11] and other. The Fick`s low that describes the rate of diffusion is the basis for their models. Final result of carbonation modelling is power function of time with exponent ½. h = A (t)1/2 + B

(1)

where: h – depth of carbonation, t – time of exposition, A, B – coefficients which depend on internal and external factors. The authors of investigations show in their works different forms of such a model, using supplementary variables, such as w/b, CO2 concentration, time of early curing and others [4, 8, 9, 10, 11]. Assumption that the carbonation process is a power function of time of exposition is equivalent to a hypothesis that the process is unlimited in time, and its depth is also unlimited. The Fick`s law refers to gas diffusion through substance immutable in time and for such a case formula (1) would be justified. But concrete is a material in which deep physicochemical changes occur in time, also due to carbonation process. Changes of pores structure with overgrowth of pores by carbonation products, lead to changes ability of CO2 diffusion in time, which can bring about decreasing of carbonation process, down to stabilizing its range on a constant depth. This development of this process was mentioned by Fagerlund [12], Slopkova [6], Fiertak [7]. Researches aimed to prove a limited scope of carbonation process have been conducted for many years in Department of Building Materials Engineering, Warsaw University of Technology. The results were used for publications [14, 15] and a PhD Thesis, WiĊcáawski [13]. This work concludes that concrete carbonation in the urban-industrial conditions can be described with hyperbolic function of time (reciprocal square root of time), which has asymptotic value parallel to time axis. This asymptote is a limit of depth of carbonation. Carbonation model showed by WiĊcáawski [13] had the following form: h = a(w/c) + b(cp) + c(t – 0.5)

(2)

where: h – depth of carbonation, mm, w/c, w/b – water-cement or water-binder ratio, cp – curing period, days, t – time of exposition, years, a, b, c – coefficients which describe relevance of influence of w/c or w/b, early curing and exposition time on depth of carbonation. It was stated that parameters (a, b, c) depend mainly on binder properties and presence of mineral additives. Traditional and hyperbolic models of carbonation are showed on Fig 1.

19-Model of.qxd

9/19/2009

11:59 AM

Page 185

Model of concrete carbonation as limitable process - experimental investigations of fluidal ash concrete

185

depth of carbonation h

A t

asymptote

Asymptotyczna wartosc graniczna

h = a(w/c) + b(cp) + c(t – 0.5) Czas, t

time

Fig. 1 Power and hyperbolic models of carbonation in concrete Similar models were elaborated for different kinds of concrete with portland cement and cement with slag and fly ash. SEM analysis showed different density of concrete structure in carbonated and noncarbonated zone. It is accepted that all results are in accordance with the hyperbolic model: h = f(1/t0.5)

(3)

regardless of binder composition, but various characteristic coefficients are obtained for various cements. Densification of concrete structure with carbonation products, which leads to a blocking barrier for carbon dioxide penetration, was shown by Bahareva et al. [16]. Investigation of microstructure of mortar carbonated in hot water saturated with CO2, conducted by Rimmelé [17] shows large diversity of porosity in sample depth, with sharp limits between well compacted carbonated zone and porous dissolution zone. Porosity distribution in a sample cross-section indicates that near the limit of carbonation depth there is a zone with very small porosity, locally equal to 0, which can be interpreted as possible limit of CO2 diffusion.

INFLUENCE OF FLY ASH ON CARBONATION PROCESS The problem of influence of fly ash on carbonation is widely discussed in publications with reference to „ordinary” fly ash (siliceous ash from stoker furnace), Neville [18]. Very few tests results are available, for calcareous (among them fluidal) fly ash. The results of carbonation tests of fluidal fly ash mortars in CO2 chamber show that there is no evident worsening of mortar resistance to carbonation due to fly ash in comparison with control mortar, Glinicki [19]. Fluidal fly ash can be considered as calcareous because the free lime content in such ash can be higher than in other types of calcareous ashes, cf. Giergiczny [20]. Furthermore, fluidal ashes have less silica content and higher sulfates content. Basic difference between ordinary and fluidal ashes is their form: glassy for ordinary ash and amorphic for fluidal ones. For evaluation of carbonation of fly ash concrete the microstructural aspects of cement paste have to taken into consideration. Fly ash combines a part of Ca(OH)2 due to pozzolanic reaction, so the content of hydroxide which is able to react with carbon dioxide is lower. Thus the depth of carbonation can be bigger, because the same CO2 quantity is sufficient to carbonate a larger volume of concrete, as Neville [18] proposed on the basis of Bier`s research. It can be concluded that presence of fly ash accelerates carbonation and gives deeper

19-Model of.qxd

186

9/19/2009

11:59 AM

Page 186

Lech CZARNECKI and Piotr WOYCIECHOWSKI

carbonated zone. But there is also an opposite effect, related with more compacted structure of hardened paste with fly ash, which decreases diffusivity and reduces rate of carbonation. Generally it can be stated, that influence of fly ash on carbonation consists of two contrary effects: - inhibition – related to densification of structure by fine ash and supplementary products of pozzolanic reaction; - acceleration – related to deficiency of Ca(OH)2 used to pozzolanic reaction, which is supported by CO2, deeper in concrete. Which one of these effects is dominating, depends on interaction of concrete components and external factors. One of more important factors is correct curing, which is necessary for right development of pozzolanic reaction. The speed of carbonation in concrete with unsuitable curing is high; research by Fiertak [7] shows very deep carbonation (20 mm) in concrete B30 with mineral additives after one year exploitation in atmospheric conditions. The same concrete with high content of fly ash and with very good early curing shows almost the same depth of carbonation as concrete without fly ash. Very important is the role of fly ash in concrete: whether it is a replacement of cement mass or it is a supplementary mass of binder. As Papadakis’a research shows [21], in the first case effect the deficit of Ca(OH)2 grows, and in second – the densification is increased. The results obtained by WiĊcáawski [13] proved domination of accelerating effect of ordinary fly ash used as component of cement. Depth of carbonation was by 25% higher for concrete with fly ash cement than with pure portland cement. Higher initial dynamics of rate of carbonation for siliceous fly ash concrete was also observed. Influence of fluidal fly ash on carbonation was not deeply investigated yet. High content of free lime in such ash can suggest that initial accessibility of Ca(OH)2 for carbonation will be high due to high reactivity of free lime in fluidal fly ash, Giergiczny [20], and high initial speed of carbonation can be expected. From the other hand, intensive pozzolanic reaction will use up calcium hydroxide and fulfil pores with reaction products, which ought to decrease CO2 diffusion and inhibit carbonation. Moreover, fluidal fly ash has higher water demand than ordinary ash. This could lead to higher capillary porosity of concrete, to intensification of CO2 diffusion and finally to acceleration of carbonation. Despite the doubts, it is necessary to take into consideration different – amorphic – characteristics of fluidal ash particles. The advantageous influence of this property on fluidal fly ash pozzolanic activity can be important for carbonation. There is high sulfate content in fluidal ash also but there is no simple dependence between carbonation and sulfate content.

RESEARCH ON CARBONATION OF FLUIDAL FLY ASH CONCRETE Aim, subject and scope of investigation The aim of research was experimental determination of the influence of selected fluidal fly ashes from Polish power plants on carbonation process. Concrete with w/b ratio 0.55 (series A) and 0.45 (series B), with two types of fluidal fly ash addition (as replacement of 15 and 30% of cement mass) were tested. Symbol w/b was used to express water-binder ratio, where binder is the sum of cement and ash content. Control concrete was made without ash. Ashes were received from fluidal combustion of hard coal (symbol K) and brown coal (symbol T), so their properties were different (Table 1). Basic difference is total CaO content, but significance of this difference for carbonation process is less important, because the free lime content in both ashes is similar. CEM I 32,5R (adequately 320 kg/m3 for series A and 360 kg/m3 for series B), sand 0/2mm, gravel 2/16 and water reducing admixtures were used for concrete mix with slump

19-Model of.qxd

9/19/2009

11:59 AM

Page 187

Model of concrete carbonation as limitable process - experimental investigations of fluidal ash concrete

187

class S3 (110-150 mm). Compressive strength after 28 days of curing in water was 44-50 MPa for w/ b=0.55 and 52-60 MPa for w/b=0.45. Table 1: Selected properties of fly ashes used in studied concretes

Properties Pozzolanic activity [%] SiO2 CaO Chemical composition CaO free [%] SO3 Loss of ignition SiO2 Phase CaSO4 composition CaO [%] Amorphic phases

Type of fluidal fly ash Turów T Katowice K (from hard coal) (from brown coal) 120 127 36 47 16 6 4.7 3.4 3.8 3.6 2.7 3.4 1.5 15 6.5 6.2 4.7 3.4 82 72

Carbonation process was tested according to standard method PN-EN 13295 [22], using chemical (phenolphthalein) identification of depth of carbonation by PN-EN 14630 [24]. This procedure indicates to test carbonation in accelerated conditions (carbonation chamber – 1% concentration of CO2, temperature 21ÛC, RH = 60%) during 56 days of exposition. In this investigation the time of exposition was extended up to 90 days, in order to increase progression of carbonation in relation to predicted maximal value (asymptote in model (2)). Before carbonation, concrete samples were conditioned 28 days in water and then at least 14 days in laboratory conditions to constant mass. Standard procedure requires comparing carbonation process in tested concrete and in control concrete by PN-EN 1766 [23]. As the control, concrete with w/b = 0.45 and without ash was used (B0). The mentioned standards do not give any criteria or rules of comparison in the control and tested concrete as stated by Czarnecki and Woyciechowski, [25, 26] Analysis of results Results of analysis (Table 2) lead to a conclusion that water-binder ratio has decisive influence on the carbonation process. In concrete with lower w/b carbonation process was slower and after 90 days the depth of carbonation was smaller. Increasing influence of fluidal fly ash content on the depth of carbonation is strongly visible in the case of higher ash content and lower w/b (Fig. 2). No evident differences in influence on carbonation between ashes type K and type T were observed. (Fig. 3).

19-Model of.qxd

9/19/2009

11:59 AM

Page 188

188

Lech CZARNECKI and Piotr WOYCIECHOWSKI

Table 2: Average values of the depth of carbonation, pH = 8.3. Symbols: A- concrete with w/b = 0.55, B – concrete with w/b = 0.45; 0, 15, 30 – content of ash in binder in %; K- ashes from hard coal; T – ashes from brown coal

Symbol

Tested pH

A0 A15K A30K A15T A30T B0 B15K B30K B15T B30T

8.3 8.3 8.3 8.3 8.3 8.3 8.3 8.3 8.3 8.3

Depth of carbonation in mm, after exposition in 1% CO2 during: 0 days 14 days 28 days 56 days 90 days 0 0 2.0 6.0 8.5 0 4.5 5.0 10.5 11.5 0 5.0 8.5 9.0 12.0 0 3.5 5.0 9.0 9.5 0 3.0 5.0 7.0 11.5 0 0 0.5 0.8 0.9 0 0 0.5 1.0 1.0 0 1.4 2.6 5.8 7.0 0 0 0.7 1.0 1.0 0 2.1 4.2 5.6 6.6

Comparison (according to PN EN 13295) of carbonation process in control concrete (concrete B0, Table 3) shows, that only 15% addition of fluidal ash with the same w/b does not significantly increase the depth of carbonation. Increase of ash content or w/b ratio substantially accelerates carbonation and increases its final depth. 12

0.45/0 0.45/15K

10 depth of carbonation,mm

0.45/30K 8

0.45/15T 0.45/30T

6

0.55/0 0.55/15T

4

0.55/30T 2

0.55/15K 0.55/30K

0 14

28

56

90

time of exposition, days

Fig. 2: Carbonation process in time (pH = 8.3), for concrete with w/b = 0.45 (continuous lines) and for concrete with w/b = 0.55 (dashed lines)

19-Model of.qxd

9/19/2009

11:59 AM

Page 189

Model of concrete carbonation as limitable process - experimental investigations of fluidal ash concrete

12

12

0 .45/15 K 10

Depth of carbonation, mm

189

0.45/30K 10

0 .45/15 T

0.45/30T

8

8

6

6

4

4

2

2 0

0 14

28

56

14

90

12

12

10

10

8

8

6

6

4

4

0 .55 /15T 0 .55 /15K

2 0

28

56

90

0.55/30T

2

0.55/30K 0

14

28

56

90

14

28

56

90

Time of exposition, days

Fig 3: Comparison of influence of K and T type fluidal ashes on carbonation process Model of carbonation in fluidal fly ash concrete The observed changes in time of values of depth of carbonation are similar to theoreticalexperimental model of carbonation (2) presented by WiĊcáawski [13, 15]. For obtained results the statistical analysis of conformity to such a model was carried out. Model (2) was simplified to formula: h = a + b(t - 0,5)

(4)

using constant w/s ratio and constant time of early curing cp in model (2). The analysis was carried out separately for each concrete from group A (w/b = 0.55) and B (w/b = 0.45). The models elaborated for all cases (column b in Table 3) exhibit high conformity to the investigation results with regression coefficient 0.93 – 0.99 (column c in Table 3). Asymptote values in col. d can be interpreted as maximal possible value of depth of carbonation according to the proposed model.

19-Model of.qxd

9/19/2009

11:59 AM

Page 190

190

Lech CZARNECKI and Piotr WOYCIECHOWSKI

A15T A30T B0 B15K B30K B15T B30T

Degree of carbonation progression after 90 days [e/d*100%]

A30K

Maximal real value of h after 90 days of exposition [mm]

A15K

h=(12.9951)+(-51.605)/sqrt(t) dh/dt = 25.8/sqrt(t3) h=(16.0901)+(-48.556)/sqrt(t) dh/dt = 24.3/sqrt(t3) h=(15.1312)+(-38.15)/sqrt(t) dh/dt = 19.1/sqrt(t3) h=(13.8364)+(-41.775)/sqrt(t) dh/dt = 20.9/sqrt(t3) h=(13.9993)+(-43.0)/sqrt(t) dh/dt = 21.5/sqrt(t3) h=(1.53144)+(-5.6463)/sqrt(t) dh/dt = 2.8/sqrt(t3) h=(1.77597)+(-6.6216)/sqrt(t) dh/dt = 3.3/sqrt(t3) h=(10.3723)+(-35.51)/sqrt(t) dh/dt = 17.7/sqrt(t3) y=(1.79142)+(-6.4228)/sqrt(t) dh/dt = 3.2/sqrt(t3) y=(9.37236)+(-27.312)/sqrt(t) dh/dt = 13.6/sqrt(t3)

Ordinate of horizontal asymptote

A0

b

Regression coefficient R

a

Model of depth of carbonation h and speed of carbonation dh/dt

Symbol of concrete

Table 3: Models of carbonation for depth of carbonation pH = 8.3 (h – depth of carbonation, mm, t – time of exposition, days)

c

d

e

f

0.96

12.99

8.5

65.4

0.92

16.09

11.5

71.5

0.95

15.13

12.0

79.3

0.96

13.84

9.5

68.6

0.97

14.00

11.5

82.1

0.99

1.53

0.9

58.8

0.99

1.78

1.0

56.2

0.96

10.37

7.0

67.5

0.97

1.79

1.0

55.9

0.999

9.37

6.6

70.4

Progress of carbonation during 90 days of investigation was 55 to 79 % (column f in Table 3), with higher degree of progress for highest fluidal fly ash content. Analysis of rate of carbonation function, defined as first derivative of carbonation depth in respect to time of exposition, confirms this effect (column b in Table 3). This is clearly visible for concrete with low w/b ratio where high initial rate of carbonation can be observed, especially in the case of high fly ash content (Figure 4.B). In the case of higher w/b ratio – initial rate of carbonation is high and practically does not depend on presence of fly ashes (Figure 4.A). For period over 60 days of exposition the rate of carbonation for all tested cases is on the same level.

19-Model of.qxd

9/19/2009

11:59 AM

Page 191

Model of concrete carbonation as limitable process - experimental investigations of fluidal ash concrete

191

Two w/b values used in investigation may not be sufficient for using as variable in models for investigated concretes. Nevertheless, an attempt was made to describe the test results by model (2) with constant time of early curing cp (28 days in water – standard conditions PN EN 12390-2). In all cases the regression coefficient was ca. 0.9 (from 0.88 to 0.93). Series of models in diagram (Figure 5) shows, that if fly ash content in concrete is small, then decreasing of w/b ratio would retard carbonation process. If fly ash content is 30%, carbonation process is initially so intensive, that w/s ratio has minor impact on its final depth. 0,4 0,35

A0

0,3 A15K

0,25 0,2

A30K

Rate of carbonation, mm/day

0,15 A15T

0,1 0,05

A30T

0 20

30

40

50

60

70

80

90

0,4 0,35 B0

0,3 0,25

B15K

0,2

B30K

0,15

B15T

0,1 B30T 0,05 0 20

30

40

50

60

70

80

90

Time of exposition, days Fig.4: Rate of carbonation received from model (3) as dh/dt = -1/2*b/t1,5

19-Model of.qxd

9/19/2009

11:59 AM

Page 192

192

Lech CZARNECKI and Piotr WOYCIECHOWSKI

Model: v5=a+b*v1+c/Sqrt(v2) z=(2,27168)+(32,)*x+(-61,928)/sqrt(y)

Model: v4=a+b*v1+c/Sqrt(v2) z=(-21,758)+(66,6667)*x+(-39,286)/sqrt(y)

15%K

R = 0,9333528

w/b C:1233E3 C:1393E6

R = 0,92528256

Carb. depth time

C:1783E4

C:1783E4

C:1783E4

C:1783E4

C:117

C:1116E3

C:117

Model: v3=a+b*v1+c/Sqrt(v2) z=(-4,9486)+(28,8333)*x+(-38,708)/sqrt(y)

C:542

30%K

C:1233E3 C:1393E6

R =0,8840833

C:1783E4 C:525798 C:787946 C:1783E4

C:542 C:3

C:1116E3 C:1783E4

C:3 C:3

C:3

14 12 10 8 6 4 2

no fly ash 12 10 8 6 4 2

C:1 C:1783E4

C:117

15%T

Model: v6=a+b*v1+c/Sqrt(v2) z=(-18,201)+(60,5)*x+(-43,387)/sqrt(y) R = 0,91837686

C:2 C:3 C:1783E4

C:1783E4 C:1783E4

C:539

Model: v7=a+b*v1+c/Sqrt(v2) z=(7,15888)+(15,)*x+(-49,272)/sqrt(y)

C:542

30%T

R = 0,91956691

C:3

C:1783E4

C:3

C:1783E4 C:1233E3

C:1783E4

C:1

C:1783E4

C:5040E4

C:117 C:1783E4 C:539

C:117 C:542

C:2

C:3

C:1783E4 C:1639E3 C:1771E3 C:1783E4

C:542

C:1783E4 C:3

C:3

C:3 12 10 8 6 4 2

C:3

12 10 8 6 4 2

Fig.5: Carbonation models based on formula (2) for different concretes, with constant time of early curing cp All discussed analysis was prepared with the use of investigation results up to 90 days of exposition in accelerated conditions. This period is longer then standard requirements (56 days in PN-EN 13295). As it could be seen in Table 3 (column f) degree of carbonation progress (by models) after 90 days is ca. 65-75% (more for concretes with fly ash). Consideration of the about asymptote value after 56 days of investigation leads to underestimation of maximal possible depth of carbonation because at this moment the degree of carbonation is not more then 50 %. Furthermore, process of carbonation becomes relatively slow after 90 days, for example after double exposition (180 days) further carbonation progress is only 5-10 %. Exposition for 90 days in accelerated carbonation conditions seems to be a relevant compromise between scientific results accuracy and its engineering usefulness.

19-Model of.qxd

9/19/2009

11:59 AM

Page 193

Model of concrete carbonation as limitable process - experimental investigations of fluidal ash concrete

193

CONCLUSIONS The conclusions from the presented research are as it follows: x fluidal fly ash used as replacement of cement mass accelerates carbonation process and increases final depth of carbonation, but low w/b ratio makes this influence less important; x there was no critical differences between fly ash from hard and brown coal with respect to carbonation process; x equally important for increasing rate and depth of carbonation is growth of w/s ratio from 0.45 to 0.55 and increasing of fluidal fly ash content from 15 to 30%; x carbonation process for concrete with fluidal fly ash can be described with hyperbolic model h = a+b/(t-0,5), and 90 days of exposition in standard conditions make possible to obtain satisfactory engineering accuracy of evaluatiof on the depth of carbonation. The results show importance of future research in order to elaborate dependence between results from accelerated standard tests and real development of carbonation in natural urban conditions for different concretes. Such function is essential for practical use of accelerated tests results. It could also be important to enrich the standard procedure PN-EN 13295 with criterial values of depth of carbonation after fixed time of exposition, which would lead to determination of a degree of concrete carbonation resistance. ACKNOWLEDGMENT The work reported here was supported by R&D Project R04 013 01 “Concrete with ash from coal fluidized-bed combustion”. REFERENCES 1. Czarnecki L. Emmons P.H., Repair and protection of concrete structures (in Polish), Polski Cement, Kraków 2002 2. ĝciĞlewski Z., Protection of RCC structures (in Polish), Arkady, Warszawa 1999 3. Rigo da Silva C., Pedrosa Reis R., Soares Lameiras F.,Vasconcelos W., Carbonation Related Microstructural Changes in Long-term durability concrete, Mat. Res. vol 5 no 3, 2002 4. Fiertak M., Nowak K., Depth of carbonation in structures – theoretical models verification, (in Polish), Ochrona przed Korozją, A, 2008, 51-56 5. Wieczorek G., Steel corrosion induced by chlorides of concrete cover (in Polish), DolnoĞląskie Wydawnictwo Edukacyjne, Wrocáaw 2002 6. Slopkova K., Observing of state of steel reinforcement in concrete and process of its corrosion by effect of carbonization, XIV Conf. KONTRA 2002, Zakopane 2002 7. Fiertak M., Material and structural concrete protection against corrosion (in Polish), XVII Conference Workshop for structures designers, UstroĔ 2002 8. Papadakis V., Vayenas C., Fardis M., Experimental investigation and mathematical modeling of the concret carbonation problem, Chemical Engineering Science 46, 1991, pp. 1333–1339 9. Gawin D., Sanavia L., Mathematical model of heat-moisture phenomena in porous bodies, (in Polish), Building physic in theory and practice, vol. 2, s 53-60, àódĨ 2007 10. Tetsuya Ishida, Koichi Maekawa, Masoud Soltani, Theoretically identified strong coupling of carbonation rate and thermodynamic moisture states in micropores of concrete. Journal of Advanced Concrete Technology vol. 2, 2004, No 2 pp 213-222 11. Lech R., Capillar-porous solid body model for computing carbon dioxide flow through a layer of calcium oxide, Cement Lime Concrete, XIII/LXXV, May-June 2008, No 3, pp 111-123

19-Model of.qxd

194

9/19/2009

11:59 AM

Page 194

Lech CZARNECKI and Piotr WOYCIECHOWSKI

12. Fagerlund G., Durability of concrete structures (in Polish), Arkady, Warszawa 1997 13. WiĊcáawski R., Concrete carbonation process in industrialized urban environment (in Polish) PhD Thesis, Politechnika Warszawska, Warszawa 2002 14. Czarnecki L. WiĊcáawski R., Carbonation of concrete as a limited process (in Polish), VI Conf. MATBUD`2003, Kraków 2003 15. WiĊcáawski R., Garbacz A., Investigation of carbonation profile in concrete exposed in natural environment (in Polish), Conf. Kontra`04, Zakopane 2004 16. Bakhareva T., Sanjayana J. G., Cheng Y., Resistance of alkali-activated slag concrete to carbonation, Cement and Concrete Research, 31, 2001, 1277–1283 17. Rimmelé G., Barlet-Gouédard V. & others.: Heterogeneous porosity distribution in Portland cement exposed to CO2 - rich fluids, Cement and Concrete Research, 38, 2008, 1038–1048 18. Neville A., Properties of concrete (in Polish), Polski Cement, Kraków 2002 19. Glinicki M. A., K. àadyĪyĔski, Activated fly ash from coal fluidized-bed combustion – new additive for concrete, (in Polish), XVIII Conference „Concrete and prefabrication”, Jadwisin 2002 20. Giergiczny Z., Function of siliceous fly ash in monitoring properties of modern binders and cement materials (in Polish), Kraków 2006 21. Papadakis V. G., Effect of supplementary cementing materials on concrete resistance against carbonation and chloride ingress”, Cement and Concrete Research, 30, 2000, 291–299 22. PN-EN 13295:2004 Products and systems for the protection and repair of concrete structures – Test methods – Determination of resistance to carbonation 23. PN-EN 1766:2001 Products and systems for the protection and repair of concrete structures – Test methods – Reference concrete for testing 24. PN-EN 14630:2007 Products and systems for the protection and repair of concrete structures – Test methods – Determination of carbonation depth in hardened concrete by the phenolphthalein method 25. Czarnecki L., Woyciechowski P., Methods of testing concrete carbonation (in Polish), Building Materials, 2, 2008, pp 5-7 26. Czarnecki L., Woyciechowski P., Methods of testing concrete carbonation (in Polish), X Int. Conf. Concrete Days, Aluszta, Ukraine, May 26-31. 2008

20-Modification.qxd

9/19/2009

11:59 AM

Page 195

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

MODIFICATION OF MINERAL BINDING MATRICES CARBON NANOSTRUCTURES 1

Grigory I. YAKOVLEV, 1Grigory N. PERVUSHIN G., 2Jadviga K. KERIENE., 3 Hans-Bertram FISCHER, 3Bernd MÖSER 1 Izhevsk State Technical University, Studencheskaya Str. 7, Izhevsk, 426069, Russia, e-mail: [email protected]; 2 Vilnius Gediminas Technical University, Sauletekio al. 11, LT-2040, Vilnius, Lithuania, e-mail: [email protected]; 3 F.A. Finger-Institut für Baustoffkunde, Bauhaus-Universität Weimar, Coudraystraȕe 11, D-99421 Weimar, Germany, e-mail: [email protected]

ABSTRACT With low concentration of a filler in composite materials and its low dispersion, the boundary layers of distant filler particles do not represent an independent phase in the composite volume capable of influencing its properties. With the increase in the filling stage separate particles in a composite material approach each other and their boundary layers start interacting with the formation of “film” matrix structure in the clearances between particles similar to the filler structure. As a result of the formation of extended film matrix structure the composite starts demonstrating non-additive specific properties expressed in the non-monotonous increase in the strength and density of the structure. The use of carbon nanostructures formations with active surface contributes to the creation of reinforcing structural-directed permolecular shell around nanosystems and formation of structures providing the reinforcement effect in the binding matrix.

Keywords Carbon nanostructures, active surface, synergetics, ultrafine particles, mineral binding

INTRODUCTION Composite materials based on a mineral binding matrix can increase potentially more significantly the mechanical strength due to the structuring of interface layers on the boundary ultra- and nanofine filler – mineral matrix. Taking into account that the bending strength of the mineral matrix is by an order lower than its compressive strength, it is optimal to use as fillers extended structures acting as ultrafine reinforcement. At present polymer fibers [1], mineral fibers [2], fibers from vegetative raw material [3, 4, 5] are usually used for ultrafine reinforcement. Composite materials based on mineral binding matrixes reinforced with fibers of different nature vividly have the features of heterogeneity as there are explicit interfaces between heterogeneous components (Fig. 1ɚ-b). Interface layers in such compositions are basically formed due to adhesion forces and the chemical interactions are weak. The efficiency of such strengthening method is determined by the adhesion of mineral matrix to the filler; at the same time the role of cohesion is insufficient and is not a determining factor of the matrix strengthening.

20-Modification.qxd

196

9/19/2009

11:59 AM

Page 196

Grigory I. YAKOVLEV, Grigory N. PERVUSHIN, Jadviga. KERIENE, Hans-Bertram FISCHER and Bernd MÖSER

ɚ

b

ɫ

Fig. 1. The microstructure (at 200-fold magnification) of the split of cement foam concrete reinforced with polypropylene fiber (ɚ); dense cement concrete reinforced with basalt fiber (b); magnesia concrete reinforced with wood fiber (ɫ) According to main regulations of the synergetics of fine-filled composites [6], the ordering of matrix structure and the formation of orientation-structured shell have been observed in the boundary layer along the ultrafine filler surface (Fig. 2ɚ). At the same time separate particles in composite material approach each other and their boundary layers start interacting thus forming film matrix structure similar to that of the filler in the clearances between the particles (Fig. 2b). As a result of the formation of extended film matrix structure the composite starts demonstrating inherent non-additive specific properties – nonmonotonous increase in strength and durability. Thus with a certain content of the filler the phase transition of the matrix from volumetric state into film one is carried out, boundary layers with oriented matrix structure are formed. The matrix structure becomes uniform and dense, the matrix structuring is observed in the boundary layer with the surface.

a)

b)

c)

Fig. 2. The scheme of forming the structured phase on the surface of ultrafine particle (ɚ): 1 – oriented-structured layer of mineral matrix, 2 – amorphous phase, 3 – ultrafine filler; the scheme of forming the structured film phase in the clearance between particles of ultrafine filler (b), the microstructure of structured mineral matrix on the surface of ultrafine particles of amorphous carbon at 600-fold magnification (ɫ) To provide the structuring of mineral binding matrix it is necessary to use ultrafine fibrous fillers, before the fibers from chrysotile-asbestos were applied. At the same time there is a hidden danger for human health while manufacturing asbestos cement products and exploiting buildings and structures produced with the use of asbestos materials. Due to the above shortcomings the fine reinforcement is necessary able to replace it by properties and at the same time to be ecologically safe in operation. It is assumed that carbon nanoformations [7] are mostly similar to chrysotile-asbestos by the characteristics.

20-Modification.qxd

9/19/2009

11:59 AM

Page 197

Modification of mineral binding matrixes carbon nanostructures

ɚ)

197

b)

Fig. 3. Carbon nanotubulenes obtained by: (ɚ) – method of low-power carbonization in gel-matrixes (transmission electron microscope), (b) – catalytic pyrolysis by gas-phase precipitation of hydrocarbons (ɋɇ4, ɋɯɇɭ) on copper and nickel catalysts (ESEM-FG technique)

MATERIALS AND EXPERIMENTAL To improve the structure of the pores of gas concrete obtained nanofine formations of two types were used: synthesized by the method of low-power carbonization in gel-matrices (Fig. 3ɚ) and by catalytic pyrolysis with gas-phase precipitation of hydrocarbons (ɋɇ4, ɋɯɇɭ) on copper and nickel catalysts (Fig. 3b). To decrease power consumption and wide variations of possibilities of obtaining carbon tubules it has been proposed [8, 9] to synthesize them from hydrocarbon condensed raw material (aromatic hydrocarbons or polymers containing functional groups) by low-power carbonization in gel-matrixes thus allowing to obtain carbon-metal containing tubules with adjustable characteristics [10].

Fig. 4. X-ray photoelectron spectroscopy of carbon nanotubes The nanoproduct was produced from the mixture of metallurgic dust and polyvinyl alcohol in the proportion 1:(1-4). The metallurgical dust represented the black powder containing by phase composition (%): NiO – 81.2; NiS – 8.1; CuO – 6; CuS – 2.5; CoO - 2. The components were mixed by their simultaneous grinding and adding water to “bind” components or by treating fine metallurgical dust with 5-10 % water solution of polyvinyl alcohol. After drying at 50 ºɋ the mixtures obtained were ground and heated step by step in

20-Modification.qxd

198

9/19/2009

11:59 AM

Page 198

Grigory I. YAKOVLEV, Grigory N. PERVUSHIN, Jadviga. KERIENE, Hans-Bertram FISCHER and Bernd MÖSER

closed cups with temperature interval of 50 ºɋ up to 400 ºɋ. The stepped heat-treatment was carried out with holding in each range within 30 minutes, at 200 ºɋ – 60 minutes, at 400 ºɋ – 120 minutes. Tubular, coalescent, globular carbon nanostructures were obtained, the results of investigation of the synthesized carbon nanotubes by X-ray photoelectron spectroscopy showed that they contain 80-90 % of carbon (Fig. 4ɚ, b). The examination of the carbon nanotubes microstructure by electron microscope have shown that the nanotubes have a cylindric form, its diameter ranges up to 50 nm and its length up to 20 ȝm. The microstructure of composites was investigated on raster electron microscope JSM JC 25S produced by JEOL. The split surfaces of the samples sprayed under vacuum with conducting aluminum layer 10 ȝm thick were studied. Nanostructures were investigated on raster electron microscope with field-emission electrode gun (device XL 30 ESEM-FEG produced by Philips). The microstructure was analyzed in optic range on digital microscope Leica DM 4000B-M with 400- and 600-fold magnification.

RESULTS AND DISCUSSION The availability of functional groups capable to direct cluster formations along axes of nanotubes with their further coalescence to crystal-hydrate state allows obtaining, as assumed, anisotropic composite structures with ultrafine reinforcement, for constructional purposes as well [11]. The mechanisms of particle interaction in nanosized range with the formation of macroforms during crystallization are also known [11]. The introduction of mineral binding nanofine structures like tubules during hydration allowed simultaneous formation of disperse reinforcement in composite material and stimulation of structure formation of hardening pastes. At the same time it was found that the formation of linear new-formations with fiber structure was possible both on the surface of nanotubes and inside nanotubules by analogy with the other research results [12]. Thus there is an opportunity of a new approach to the problems of the formation of composite inorganic materials modified with carbon nanotubes as a result of disperse reinforcement.

a)

b)

c)

Fig. 5. Permolecular structures formed on the surface of carbon nanotubes: macroforms of crystals of hydrates of cement minerals on the surface of nanotubes under optical microscope: (ɚ) - with 400-fold magnification, (b) - 600- fold magnification; (c) – fibrillar structures in pore walls of cement foam concrete under raster electron microscope When introducing carbon nanotubes in cement matrix, the content of which did not exceed 0.05% from the matrix weight, we reached structural orientation of the cement matrix around the nanoparticles with the formation of crystal-hydrate new-formations (Fig. 5ɚ, b). Fibrillar permolecular structures, formed in boundary layers of carbon nanostructures,

20-Modification.qxd

9/19/2009

11:59 AM

Page 199

199

Modification of mineral binding matrixes carbon nanostructures

stabilize the foam concrete structure providing the formation of disperse reinforcement in pore walls (Fig. 5c), thus improving its operational characteristics. b)

a)

c)

50 m

50 m

Fig. 6. The structure of pore walls in foam concrete: ɚ – without nanotubes added (perforated), b – stabilized with the addition of 0.05 % of nanotubes, c – aggregations of regularly directed nanotubes coated with products of Portland cement hardening on the pore surface Table 1. Physical and mechanical characteristics of cement foam concrete No

1 2

Content of nanotubes in % from the composition weight 0 0.05

Average Heat Ultimate density, conductivity strength at kg/m3 compression, in coefficient Ȝ, W/mɨK MPa 330 309

0.18 0.306

0.07 0.056

Pore sizes, ȝm

Condition of pore walls

40 - 600 perforated 60 -150 homogeneous

Despite of the low concentration of introduced carbon nanotubes due to their high dispersity caused by nanosizes, the total number of nanotubes for the volume unit of composite material is comparable with the content of basic components of the composition and can influence the structural organization of composite material. Spreading in the volume of the cement matrix, nanosystems play the role of centers of directed crystallization to form ultrafine reinforcement thus resulting in the decrease in the perforation of pore walls of the material (Fig. 6b), providing its continuity and uniformity of the pore walls (Fig. 6c), stabilizing the homogeneity of pore sizes. At the same time the operational characteristics of cement foam concrete are improved due to the increase in its strength and decrease in the heat-conductivity (Tabl. 1) of products manufactures on its basis, [13, 14].

CONCLUSION The application of ultrafine extended formations in the form of carbon nanotubes having an active surface, results in the formation of submolecular structures in the mineral matrix, which form interface layers reinforced with nanoformations. The use of carbon nanotubes is an effective way to control the processes of structure formation in porous cement matrices providing the regulation of interface layer morphology and achievement of increased physicengineering properties of mineral composite materials. The results of the investigation of the reinforced non–autoclave cement foam concrete showed that the use of carbon nanotubes

20-Modification.qxd

200

9/19/2009

11:59 AM

Page 200

Grigory I. YAKOVLEV, Grigory N. PERVUSHIN, Jadviga. KERIENE, Hans-Bertram FISCHER and Bernd MÖSER

(0,05 % by mass) in production of these concretes allows to decrease its heat conductivity up to 12 – 20 % and increase its compressive strength up to 70 %.

REFERENCES 1. Kim, D., Naaman, A.E., El-Tawil, S. Comparative flexural behavior of four fiber reinforced cementitious composites, Cement and Concrete Composites, vol. 30, 10, 2008, pp. 917-928 2. Sim, J., Park, C., Moon, D.Y., Characteristics of basalt fiber as a strengthening material for concrete structures, Composites Part B: Engineering, vol. 36, 6-7, 2005, pp. 504-512 3. Boghossian, E., Wegner, L.D. Use of flax fibres to reduce plastic shrinkage cracking in concrete , Cement and Concrete Composites, vol. 30,. 10, 2008, pp. 929-937 4. Plechanova, T.A., Kerien , Ja., Gailius, A., Yakovlev, G.I. Structural, physical and mechanical properties of modified wood-magnesia composite, Construction and Building Materials, vol. 21, 9, 2007, pp. 1833-1838 5. Ramakrishna, G. & Sundararajan, T., Impact strength of a few natural fibre reinforced cement mortar slabs: a comparative study, Cement & Concrete Composites. vol. 27, 5, 2005, pp. 547-553 6. Bobryshev, Ⱥ.N., Komozov, V.N., Avdeev, R.I., Solomatov V.I., Synergetics of dispersefilled composites. Ɇ.: CKT, 1999, 252 p. 7. Yakovlev, G.I., Kerien , Ja., Plechanova, T.Ⱥ., Krutikov, V.Ⱥ., Nanobewehrung von Schaumbeton, Beton- und Stahlbetonbau, vol. 102, I 2, 2007, pp. 120-124 8. Babushkina, S.N., Kodolov, V.I., Kuznetsov, A.P., Nikolaeva, Ɉ.Ⱥ., Yakovlev, G.I., The way to obtain carbon-metal containing nanostructures. Patent of RF for invention No 2169699. Published: BI, 2001, No 18 9. Kodolov, V.I., Kodolova (Trineeva) V.V., Semakina N.V., Yakovlev G.I., Volkova E.G. The way to obtain carbon nanostructures from organic compound and metal-containing substances, Patent of Russia No 2337062. Bulletin No 30, 28.10.2008 10. Kodolov, V. I., Shabanova, I. N., Makarova, L. G., Khokhryakov, N. V., Kuznetsov, A. P., Nikolaeva, O. A., Kerene, Ja., Yakovlev, G. I., Structure of the products of stimulated carbonization of aromatic hydrocarbons, Journal of Structural Chemistry, vol. 42, 2, 2001, pp. 215-219 11. García-Ruiz, J., Melero-García, E., Hyde, S.T., Morphogenesis of Self-Assembled Nanocrystalline Materials of Barium Carbonate and Silica, Science, vol. 323, 2009, pp. 362–365 12. Kuznetsova, A., Mawhinney, D.B., Naumenko, V. J., Yates T. Jr., Liu, J., Smalley, R.E. Enhancement of adsorption inside of single-walled nanotubes: opening the entry ports, Chem. Phys. Lett. 321, 2000, pp. 292-296 13. Lipanov Ⱥ.Ɇ., Trineeva V.V., Kodolov V.I., Yakovlev G.I., Krutikov V.A., Volkova E.G. Obtaining of carbon metal-containing nanostructures to modify constructional compositions, Alternative Electrical Engineering and Ecology. 8 (64), 2008, pp. 82-85 14. Yakovlev, G., Pervushin, G., Krutikov, V., Makarova, I., Keriene, Ja., Fischer, X.-B. Light-Weight concrete on a basis fluoranhydrite, modified carbon nanostructures, Building Materials. No 3, 2008, pp. 70 -72

21-Mechanical-acoustic.qxd

9/19/2009

11:59 AM

Page 201

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

MECHANICAL-ACOUSTIC AND STRUCTURAL STUDY OF DEGRADATION PROCESSES IN CORUNDUM CERAMICS AND ALUMINOUS PORCELAIN Przemysáaw RANACHOWSKI, Zbigniew RANACHOWSKI, Feliks REJMUND Institute of Fundamental Technological Research Polish Academy of Sciences PawiĔskiego 5B, 02-106 Warsaw, Poland, e-mail: [email protected]

ABSTRACT This paper comprises the results of acoustic emission (AE), microscopic and ultrasonic measurements of samples subjected to slowly increasing compressive stress. On the basis of conducted measurements the successive stages of the materials structural degradation have been recognized. The object of study were samples made of C 799 corundum ceramics and C 130 aluminous porcelain. Both investigated materials have at present wide application in the fabrication of numerous technical elements e.g. overhead power line insulators. In case of such objects not only high mechanical strength but especially elevated durability as well as operational reliability are required. Expected “life time” of net insulators during exploitation is about 40 years. The analysis of obtained mechanoacoustic characteristics pointed out complicated mechanism of degradation of the materials. Microscopic investigation of samples, which were stressed to different levels of load, enabled to specify the various development of gradual growth of microcracks and successive crushing out of elements of the structure. These effects are similar to the ageing processes occurring in the materials during long period of exploitation under working load. Comparison of the results of mechanical, acoustic and microscopic studies revealed that differences recorded for the strength and characteristics are due to inhomogeneities of the materials in the semi-macro as well as in the micro scale. The occurrence of groupings of bigger grains in the structure of the corundum material represents most probably the intermediate state, leading to the known effect of the abnormal grain growth (AGG). The effectiveness of dispersive and fibrous reinforcement of aluminous porcelain C 130 type was described. Strengthening by corundum grains and mullite needle shaped crystals improves mechanical parameters and distinguishes this material from typical aluminosilicate ceramics. Presented results enable to draw up conclusions concerning the resistance of investigated materials to the ageing degradation processes development during long term operation.

Keywords corundum material, aluminous porcelain, structural degradation, acoustic emission.

INTRODUCTION High-alumina ceramics of C 700 group are characterized by high content of Al2O3 (corundum) from 80 % to 99.7 %. This monophase material has wide and increasing application for general technical purposes. Technical objects made of corundum material take advantages of its high mechanical and thermo-mechanical strength as well as abrasive resistance. This ceramics is also resistant to oxidation, chemical corrosion and various types of irradiation. Its thermal conductivity is similar to that of stainless steel. Moreover, the corundum has good electric properties – high dielectric constant and low loss angle. Principal applications are special and low- or ultra-low loss insulators, substrates, metallized parts,

21-Mechanical-acoustic.qxd

202

9/19/2009

11:59 AM

Page 202

Przemys³aw RANACHOWSKI, Zbigniew RANACHOWSKI and Feliks REJMUND

sodium vapour lamp envelopes, [1-3]. In recent times it is used for carrying rods of hybrid insulators, [4]. Aluminous porcelain of C 130 type of high mechanical strength is at present widely applied to produce reliable insulators and small high-strength parts, [1-3]. Particularly line insulators HV and EHV, HV post insulators, medium voltage (MV) line and post insulators of increased mechanical requirements, traction insulators and hollow insulators of high parameters. In case of these products, beside high mechanical strength, a long period of exploitation without breakdown is required. For several years there has been observed increasing tendency for the production of C 130 material, decreasing production of the siliceous porcelain C 110, and to some extent also of aluminous material of C 120. This is the result not only of the present requirements concerning the short-term mechanical strength of the electroinsulating elements. It is more important to guarantee the reliability of power supply, which is determined by the durability, i.e. by long-term mechanical strength of the ceramic material. Evaluation of the operating time of the ceramic materials is based mainly on the analysis of the formation and development of ageing effects in their structure. The essence of ageing degradation is a gradual expansion of the already existing microcracks and the formation of new ones under the influence of the mechanical, structural stresses occurring in the material body. Structural stresses represent the sum of internal stresses and stresses induced by the external factors [5,6]. The internal stresses are created during the technological production processes. Then significant mechanical stresses are formed: on micro scale e.g. on the boundaries of quartz grains and glassy matrix; on semi-macro scale – resulting from textural anisotropy; stresses on the macro scale between the internal and the external regions of the object. An object in operation is subjected to considerable exploitation static stresses, as well as additionally – especially dangerous – dynamic loads. These stresses, when added to the internal ones, accelerate the ageing processes. An additional factor contributing to the propagation of microcracks are the temperature changes in the body, attaining within 24 hours even as much as 45 0C. The most important factor, responsible for gradual degradation of the parameters of ceramic material, are the local stresses occurring at the grains, the interfacial boundaries and the alien inclusions in the ceramic body. Internal stresses in the micro-areas are located in the brittle medium. The only way of their relaxation is an increase of already existing or initiation of a new microcrack. Thus, relaxation of stresses is connected with a decrease of mechanical strength of the material. However, the object in operation is under constant external load. Consequently, the growth of microcracks causes gradual decrease of the cross-section area. So, in the material under load there are internal stresses, which induce increase of microcracks. Development of microcracks causes the degradation of the parameters of the material with ageing. From reports concerning C 120 porcelain it is known that about 35 years long period of exploitation causes over 30% decrease of the mean mechanical strength of insulating material. Besides that, the mechanical strength dispersion of such objects is about 2.5 times greater than in case of the new ones, [7]. In case of ceramic insulators degradation of the mechanical and electric parameters is of great importance, because it decreases reliability of power supply. Appearance of ageing processes in ceramic materials has been also confirmed by several publications, [5,7]. The factor which has important influence on the degradation of the aluminosilicate materials in the process of time is high content of quartz. An additional problem is dispersion of the properties of the material of products, resulting from unsatisfactory repeatability of technological process as well as other effects such as abnormal grain growth (AGG). In case of material of C 130 kind there is not enough experience obtained during a longer period of exploitation of the products. Although production of this

21-Mechanical-acoustic.qxd

9/19/2009

11:59 AM

Page 203

Mechanical-acoustic and structural study of degradation processes in corundum ceramics and aluminous porcelain

203

porcelain in the Polish industry began in the year 1979 (material denoted E-15), it became widely applied not until the 1990’s. The main aim of presented study is description and comparison the effects of structure degradation in two widely applied ceramic materials. Studied materials are generally different. High-alumina, corundum material belongs to the group of monophase ceramics. Electrotechnical porcelain in turn is an aluminosilicate material, containing structural reinforcement by densely dispersed fine corundum grains and needle-shaped mullite crystals in glassy matrix. The obtained results revealed essential differences in mechanical strength and mechanics of degradation process in both materials.

PREPARATION OF THE SAMPLES In this study there are presented the mechanoacoustic and structural investigations of corundum material of C 799 type and Al2O3 content equal to 99.7 %. For the fabrication of the samples the granulated product NM 9922 of Company Nabaltec was used. The size of grains, collected in the aggregates, was lower than 0.5 Pm. The crystallites were smaller by one order of magnitude. Samples for the investigations were formed using the method of single axial pressing (10 MPa) and isostatic condensation (120 MPa). After firing into biscuit at temperature 1250 oC, samples were cut out from a larger element, taking into account grinding and shrinking of the mass. Next, the samples were fired at the maximal temperature 1700 0C and stored for 1.5 hr. at the sintering temperature. The samples were ground to obtain the dimensions 5x5x10 mm, then their density was determined and the absorbability and absence of cracks in the alcohol solution of fuchsine were controlled. The obtained material had density U=3.89 g/cm3 and did not contain any detectable defects. In order to determine the size of grains, some samples were polished, thermally etched at the maximum temperature 1300 0C and stored for 1 hr. at this temperature. Study of C 130 aluminous porcelain was carried out on specimens cut out from typical Polish high voltage line insulator. Traditional and used process of production of long rod insulators consists of many stages [5,8]: selection of components, control and preparation of raw materials (weighing and milling), plasticization of raw material (mixing with water, filtration, seasoning), formation (deairing, pug pulling and profile turning), drying, glazing (marking), firing (sintering), final treatment (cutting and grinding), montage of fitting devices (assembling), tests of the object. The samples had dimensions 7x7x10 mm. The composition of the raw mass applied to fabricate insulator was typical for the materials C 130 type. It comprised approximately 18 % kaolin, 22 % refractory plastics clays, 20 % feldspar fluxes and 35 % ceramic alumina. The cullet constituted 5 % of raw material. Surface of the samples, especially bottom and top, were precisely ground to obtain parallel and smooth planes of the order of 0.05 mm.

ULTRASONIC MEASUREMENTS Ultrasonic control of the homogeneity of the C 799 corundum samples revealed a slight anisotropy as well as some differences of the acoustic parameters and the elasticity modulus between the samples. For example, the velocity of the longitudinal waves cL, measured in the direction perpendicular to the axis for various samples, amounted from 10480 to 10600 m/s (the inaccuracy of measurement was equal to ± 20 m/s). Young’s elasticity modulus E, determined in the same direction was enclosed, depending on the sample, in the range from 364 to 373 GPa. The average E value was 368 GPa, at the measurement inaccuracy equal to

21-Mechanical-acoustic.qxd

204

9/19/2009

11:59 AM

Page 204

Przemys³aw RANACHOWSKI, Zbigniew RANACHOWSKI and Feliks REJMUND

± 2 GPa. The mean value of Young’s modulus in the direction parallel to the sample length was somewhat lower and was equal to 360 GPa. The obtained parameters considerably exceeded the required by standard values of U=3.70 g/cm3 and E=300 GPa for the corundum material of C 799 type, [3]. Ultrasonic control of homogeneity of the porcelain C 130 revealed the anisotropy typical for ceramic objects formed using the screw extrusion method in the vacuum deairing pug mills, [4,8]. Velocities of the longitudinal cL and the transverse cT waves, measured along the lengthwise axis of insulator, were equal to 6980 m/s and 4130 m/s, respectively. The calculated value of Young’s modulus E equaled 111 GPa (density of the material U = 2.64 g/cm3). In the crosswise direction of insulator, cL was equal to 6620 m/s and cT 4100 m/s, respectively. Young elasticity modulus, determined from these data, was equal to 106 GPa. The uncertainty of measurements for cL and cT was r20 m/s, whereas for the calculated values of Young modulus was about r1.0 GPa.

MECHANOACOUSTIC METHOD The method of acoustic emission is a valuable scientific tool when used for monitoring internal structural changes in ceramic materials. This technique allows to obtain numerous data concerning the dynamic processes occurring during change of mechanical, thermal or thermo-mechanical stresses [9]. This is the more essential that AE signals appear already at the threshold stresses, when the generation of microcracks in the material cannot be in practice detected by other methods. The essential factor in AE measurements is to use such descriptors of signals that contain the most relevant information for the evaluation of tested process. The choice of the optimal descriptor for the definite measurement is determined by the following criteria: linear dependence between descriptor and measured physical parameter in time; maximal sensitivity of descriptor concerning changes of measured parameter; good repeatability of measurements – small dispersion of results during testing high number of samples of the same material (the lowest standard deviation); the lowest influence of external conditions such as e.g. discrimination level of AE signals; minimizing the influence caused by signals of the background. Besides the above criteria, there exist additional, very important factor. Acoustic emission descriptor, selected at the research should be optimal for tested material. As it was stated experimentally by the authors, choice of AE descriptor should be different in case of aluminosilicate materials and oxide ceramics. Aluminosilicate materials such as porcelain, steatite or cordierite, are characterized by numerous signals of lower energy. These signals can be effectively recorded using AE events rate. It is specially important during preliminary and subcritical stages of structure degradation. During registration descriptors based on energy of signals, even at low discrimination level, substantial part of them can be lost. In case of corundum material, the authors observed an appropriate correlation between the structure degradation, mainly connected with intergranular cracks development, and the effective value of AE signal (RMS – root mean square). This is connected with high mechanical strength of the corundum and relatively high energy of generated AE signals. Usage the course of RMS rate is more effective than AE events rate in case of this oxide material. The samples were subjected to mechanoacoustic measurements using the technique of acoustic emission (AE) on a special two-channel measuring system. Pieces of small dimensions were submitted to slowly increasing compressive stress (v = 0.02 mm/min), with simultaneous registration of the force in one channel, and AE descriptors in the second one.

21-Mechanical-acoustic.qxd

9/19/2009

11:59 AM

Page 205

Mechanical-acoustic and structural study of degradation processes in corundum ceramics and aluminous porcelain

205

The investigations enabled the recording and description correlation between the increasing external load and the processes of structure degradation, which are reflected in the acoustic activity AE. Acoustic method is suitable for the investigation of the destruction of ceramic materials, because due to the fact that initiation and growth of microcracks belong to the main sources of AE signals. Examination of aluminosilicate and oxide ceramic materials enabled to state that the sum of AE events during the loading period is a good descriptor of the intensity of the processes of cracking, which are the cause of mechanical degradation of the material. There exists a correlation between the rate of the increase of cracks and the rate of AE events (number of AE events per unit of time), [10]. The events rate and the energy of AE signals, however, are not in general a linear function of changes of the mechanical or thermal stresses. The velocity of these changes is an additional factor influencing the acoustic activity, which is difficult to define quantitatively. The measurement of the AE events rate and RMS as descriptors, at very slow increase of mechanical load (of the order of 10-2 mm/min) allows, however, to make the AE investigations almost independent of the influence of other factors on the degradation process of the material. This has been confirmed by the authors during research of porcelain (C 120) and cordierite (C 410) materials, [11,12]. There exist serious analogies between the effects of many years long exploitation under working load applied to the material and material degradation during compressive stresses in a relatively short lasting laboratory test. However, it is necessary to apply a quasi-static, very slow increase of stress and a precise recording of the AE descriptors. This observation has been proved by investigations, carried out by the authors, on electrotechnical porcelain, [11]. Mechanoacoustic tests were carried out using specially constructed two-channel measuring system. The mechanical channel contained testing machine INSTRON with computer control. The steel base on which the sample was placed functioned simultaneously as an acoustic waveguide. In the investigation the velocity of the traverse of the machine equal to 0.02 mm/min was applied. Parallel to the measurement of the load acting on the sample, AE descriptors were recorded. The acoustic recording path contained a broad band transducer WD PAC type (passband 80y1000 kHz), standardized AE analyser and a computer. One second time interval of summing up the signals was applied. The rate of counts, the events rate and the energy of AE signals were recorded. As it has been mentioned above, the most effective AE signal descriptors are presented. Valuable information for the evaluation of the examined processes of material degradation is offered by AE events rate, especially for the porcelain, as well effective value of AE signal – RMS in case of corundum ceramics.

EXAMINATION OF CORUNDUM MATERIAL The compressive strength of the samples has shown an unexpected wide distribution. The mean strength of nine destroyed specimens was 3290 MPa. The least resistant of the samples underwent decohesion already at the stress equal to 2040 MPa. This result was neglected in statistical calculations. The next weakest sample had strength 2660 MPa. It was found impossible to measure the strength of the most resistant sample. For technical reasons it was necessary to stop the process of loading at the stress of 3800 MPa. Relative dispersion of strength equaled 34.6 %. The mechanoacoustic characteristics of the particular samples showed high differentiation – figures 1 and 2. At stresses below 2500 MPa the samples, especially of more inhomogeneous structure, showed AE signals of various intensities. If there appeared the intervals of continuous acoustic activity – the AE signals were most frequently characterized by low energy. Above the stress 2500 MPa the loaded samples

21-Mechanical-acoustic.qxd

206

9/19/2009

11:59 AM

Page 206

Przemys³aw RANACHOWSKI, Zbigniew RANACHOWSKI and Feliks REJMUND

showed AE effects in the form of single signals, sometimes forming intervals of continuous acoustic activity of moderate energy. The interval of subcritical stresses was characterized by diversified width – depending on the sample strength. The wide subcritical stage preceded a short critical interval, containing a group of signals of high energy. The critical interval occurred in the range of stress of some tens of megapascals and it directly preceded the decohesion of the sample. Thus, the stress at which the critical stage took place was closely connected with the sample strength (Fig. 1). 120

AE RMS [s-1]

3500

V[MPa]

3000

100

2500

80

2000 60 1500 40

1000

20

500

0 0

0 500 1000 1500 2000 2500 3000 3500 4000 4500 t [s]

Fig. 1. Typical course of RMS AE rate versus the increase of compressive stress for the sample of the strength 3320 MPa. In order to explain the effects of structural degradation of the material there were carried out accurate microscopic investigations of the samples. Structural analysis concerned size and the uniformity of distribution of the grains and the effects of the compressive stress at various areas of the specimens. Special attention was paid to presence of defects in the structure before and after compression of the samples. 120

V[MPa]

-1

AE RMS [s ]

4000 3500

100

3000

80

2500

60

2000 1500

40 1000 20

500

0 0

1000

2000

3000

t [s]

0 4000

Fig. 2. The course of RMS AE rate versus the increase of compression stress for the most resistant sample. Compression was stopped at 3800 MPa, before the critical stage of degradation. The tests revealed that considerable differences recorded for strength and the mechanoacoustic characteristics of the samples were due to inhomogeneities and faults in the structure in the semi-macro as well as in the micro-scale. The samples contained fine technological defects such as gaseous and solid inclusions, fissures and partly broken grain boundaries – figure 3. Almost all of the samples showed a bimodal size distribution of the grains. The sizes of bigger grains were most often in the range from less than 10 to over

21-Mechanical-acoustic.qxd

9/19/2009

11:59 AM

Page 207

Mechanical-acoustic and structural study of degradation processes in corundum ceramics and aluminous porcelain

207

30 Pm, whereas the diameter of smaller grains was from a fraction to a several micrometers – figure 4. Only few samples demonstrated uniform size of grains and proper internal texture. The grains of the most resistant sample showed one-modal size distribution. Their diameters were enclosed in a narrow range from above 1 to 14 Pm, with the mean value equal to 7 Pm.

Fig. 3. Textural inhomogeneity (clusters of bigger grains surrounded by finer grains), weakened grain boundaries and big pore are visible, magnification 500x. Structure of the corundum material before the beginning of compressive loading. There were observed textural inhomogeneities in the structure of the samples. The bigger grains often formed clusters of the size of the order of tens micrometers (up to 100 Pm), surrounded by smaller areas containing finer grains – figure 3. Another discovered textural inhomogeneity was arrangement of the structure in form of bands. The greater and/or the smaller grains were organized as separate bands, having width of the order of some tens of micrometers. The most regular structure was observed in case of the sample of the highest strength. Percentage of fraction [%]

14 12 10 8 6 4 2 0 1

3

5

7

9 11 13 15 17 19 21 23 25 27 29 31 Diameter of grain [um]

Fig. 4. Typical bimodal distribution of the grains diameters, obtained for the sample of corundum material. After compression tests, several samples of various mechanoacoustic characteristics were selected for microscopic investigations. The loading of speciments was stopped at different stresses. A few was closely before the critical stage of structure destruction. The compressed samples were cut and polished in proper way for microscopic study. In all specimens there were observed the effects of structure loosening. This concerned especially the central part of the samples, where the stresses were cumulated. Propagation of

21-Mechanical-acoustic.qxd

208

9/19/2009

11:59 AM

Page 208

Przemys³aw RANACHOWSKI, Zbigniew RANACHOWSKI and Feliks REJMUND

microcracks occurred almost only along the grain boundaries. Moreover, there have been observed black areas displaying crushed out grains or parts of grains. For samples compressed up to advanced subcritical effects these areas covered 0.2÷0.3 % of the surface of a polished sections at the boundaries and up to 1.0 % in the middle of the samples. The initial porosity of the material in any of the specimens did not exceed 0.1 %. There were present crushed out fragments and even the whole grains. Their size was generally below 20 Pm. The structure of the compressed samples underwent evident but varied loosening. The length of intergranular cracks was contained in the interval equaled 15÷50 Pm, and many of them were closed (around single or group of grains). There occurred also bigger cracks of the character of fissures. The various areas of the samples demonstrated differentiated degree of structural degradation. In the central part of the sample compressed up to beginning of critical stage 3180 MPa, there were observed even long cracks of the length of some hundreds of micrometers – figure 5. The most resistant of the investigated samples (compressing stopped at 3 800 MPa), showed a moderate degree of structure degradation. Even in its central part, the surface area of the crushed out grains was lower than 0.3 %. The mean length of cracks was about 20 Pm and they were shorter than 50 Pm. As it was stated, corundum samples compressed to advanced subcritical stage revealed presence of micro- and macrocracks. Over 90 % of cracks underwent propagation along grain boundaries and only under 10 % within the grains. Points of initiation of the cracks were technological defects (gaseous and solid inclusions, fissures and partly broken grain boundaries), showed in figure 3. The earlier, preliminary stage of AE activity was connected with these defects. Threshold energy of relaxation of the stresses connected with defects is low and such is effective value of AE signals. This phase of degradation is varied for particular samples. It concerns the range of stresses, where signals are registered, and their intensity. The preliminary stage depends closely on number, size and spatial distribution of technological faults. This stage of degradation can take place in very wide range of loading, even up to about 2500 MPa – figures 1 and 2. During the subcritical stage of the material degradation further growth of intergranular cracks takes place. The most easily fracture bounds between bigger grains, especially these elongated, and mainly in the direction perpendicular to the compression. Clusters of joined bigger grains are particularly susceptible to destruction process as well. As a result the strength of samples containing greater number of clusters and bands of bigger grains (10 ÷ 40 Pm) was considerably decreased. Cracking within bigger grains as well as at the boundaries of small grains was considerably less frequent because of much higher threshold energy. AE signals appearing during subcritical stage are strongly diversified. Their energy is moderate but growing and generally higher than in the previous stage. Range of stresses, for which subcritical stage takes place is varied and the most frequently extends to a few hundreds of megapascals. During the critical stage of destruction, propagation of cracks occurred with high velocity and throughout all components of corundum structure. Figure 6 presents strong damages, which take place at the beginning of the critical period of structure degradation. Growth of visible cracks resulted in continuous and high energy AE activity. Sometimes, before destruction of the whole sample, its macroscopic parts (e.g. piece of wall or corner) undergoes separation. The range of the critical stage of the corundum material is generally similar to that of the aluminosilicate ceramics. Textural defects present in the examined ceramics were connected with grains of different size, which were not uniformly distributed in the space of the samples. Sometimes they grouped into clusters or bands. Observed diversification of grains’ size in the structure of the corundum material (Fig. 3 and 4) represents most probably the intermediate state, leading to the known effect of the abnormal grain growth (AGG), [13]. This phenomenon occurs most

21-Mechanical-acoustic.qxd

9/19/2009

11:59 AM

Page 209

Mechanical-acoustic and structural study of degradation processes in corundum ceramics and aluminous porcelain

209

frequently in the oxide ceramic materials. It has a probabilistic character and its origin, despite many years of investigations, has not been sufficiently explained. AGG effect occurs after a longer time of firing than in case of the applied technology of the production of the samples. The quick increase of temperature during thermal treatment favours its occurrence and such temperature raise – of the order of 200 0C by one hour – was realized. In case of obtaining bigger and longer time sintered elements, the AGG effect would cause considerably greater differences in size of the grains.

Fig. 5. Cracks in the central area of the sample loaded up to 3180 MPa, magnification 100x.

Fig. 6. Degradation of the structure visible on surface of the sample, which compression was stopped at 3190 MPa, magnification 200x.

EXAMINATION OF PORCELAIN MATERIAL The compressive strength of eleven samples loaded until a complete destruction showed relatively high dispersion. The mean value of strength was equal to 779 MPa, the lowest – 522 MPa and the highest was 933 MPa. The relative dispersion for these results corresponded to 40.9 %. In calculation the lowest value of strength was neglected and the weakest result was 614 MPa. Besides the broken samples, the group of nine specimens was selected for the microscopic investigation. A compression process of these samples was stopped at different levels of stresses from 50 to 839 MPa. The obtained mechanoacoustic characteristics of the particular samples showed considerable differentiation. Comparison of these results and

21-Mechanical-acoustic.qxd

210

9/19/2009

11:59 AM

Page 210

Przemys³aw RANACHOWSKI, Zbigniew RANACHOWSKI and Feliks REJMUND

microscopic study revealed generally a presence of three stages of the structure degradation. Figure 7 and 8 show the courses of acoustic activity for the samples destroyed at 714 and 614 MPa. Presented range of stresses in Fig. 7 equaled 0 y 707 MPa show only the preliminary and the subcritical stages of the material degradation in more distinct way. 400

AE Events [s-1]

V [MPa]

700 600

300

500 400

200

300 200

100

100 0

0 0

250

500

750

1000

1250

1500

1750

t [s]

Fig. 7. Typical course of the rate of AE events versus increase of compressive stress for the sample of the strength 714 MPa. Displayed are only preliminary and subcritical stages of degradation in the stress range equaled 0 y 707 MPa. The first stage of acoustic activity, described as a preliminary one, takes place approximately from 50 to 300 MPa of compressive stress. This period corresponds to a destruction of particles of cullet, quartz grains and the beginnings of the mullite phase damage. The last effect takes place only in the central part of samples, where the highest concentration of stresses occurs. The first damages of the mullite precipitates are recorded already at about 200 MPa. It is connected with internal cracking, because the mullite precipitates are strongly bound to glassy matrix. In the stress range from 30 to 150 MPa, without an acoustic activity, particles of cullet undergo fracture and separation from the porcelain matrix. Greater part of cullet is separated from matrix even without acting of external stresses. Majority of quartz grains go through a similar process in the approximate stress range 50 ÷ 200 MPa, Though, in case of few bigger quartz grains this process requires much higher loading. Degradation of quartz phase is followed by a weak AE activity. A little stronger acoustic effects follow only the cracks development in quartz grains of greater size (over 20 Pm) and inside mullite precipitates. 4500

-1

V [MPa]

AE Events [s ] 4000 3500

600 500

3000 400 2500 2000

300

1500

200

1000 100

500 0 0

200

400

600

800

0 1000 1200 1400 1600 t [s]

Fig. 8. The course of the rate of AE events versus increase of compressive stress for the weaker sample of the strength 614 MPa. AE signals of the preliminary and the subcritical stages are almost invisible because of the lower amplitude.

21-Mechanical-acoustic.qxd

9/19/2009

11:59 AM

Page 211

Mechanical-acoustic and structural study of degradation processes in corundum ceramics and aluminous porcelain

211

Next stage of structure degradation – the subcritical one is closely connected with the homogeneity of the sample structure in micro and semi-macro scales. The subcritical stage begins at about 450 MPa and lasts up to over 600 MPa, frequently to the beginning of the critical stage. This phase of destruction is considerably varied for particular samples and shows low or moderate intensity of AE activity – figures 7 and 8. During the subcritical period cracks are created and propagated in the mullite precipitates, which are numerously present in the material C 130 type. There are created internal and less frequently peripheral cracks in some mullite precipitates. Their parts undergo crushing out, especially in the central section of the samples. Moreover, at the stresses over about 400 MPa begin a degradation of clusters of agglomerated corundum grains. As the loading grows particular corundum grains undergo a separation from the structure of agglomerate - figure 9. Agglomerates of collected grains of corundum are not numerous in the material structure and the subcritical stage of degradation is connected especially with the internal fracture of mullite precipitates. This process is the most intensive in the central part of the samples. At the end of the subcritical phase, area of damaged, separated and crushed out elements of structure comprises over a dozen percent - figure 9. Majority of them constitutes cullet and quartz damaged already at the preliminary stage.

Fig. 9. Structure of the sample loaded up to the advanced subcritical level, magnification 200x. Black, crushed out parts of grey mullite precipitates, quartz grains, cullet fragments and damaged agglomerate of corundum grains are visible. Black places cover together over a dozen percent of surface. The last – critical interval, showing the highest level of the acoustic activity, begins at loading from several to some tens of megapascals lower than the destructive stress for the particular sample. It lasts up to the destruction of the sample. This interval is characterized by generally good repeatability of the level of AE signals, which have the highest intensity. The extent of its occurrence, however, depends on the strength of the particular sample. Wide range of the critical stresses, observed for the majority of samples – figure 8, results from a fracture and a splitting off greater pieces of sample. This effect is shown on the stress increase curve as a characteristic faults (jumps down), preceding decohesion of the sample. However, some specimens underwent destruction unexpectedly – after very short critical phase. During the critical stage of structure degradation further precipitates of mullite undergo the internal and peripheral fracture. Degradation of the corundum agglomerates is being continued as well. Propagation of cracks in the matrix is the most important and destructive effect, followed by the strong AE signals. These cracks are elongated and not branched – figure 10. They grow initially between mullite precipitates and damaged elements of the structure. The strong

21-Mechanical-acoustic.qxd

212

9/19/2009

11:59 AM

Page 212

Przemys³aw RANACHOWSKI, Zbigniew RANACHOWSKI and Feliks REJMUND

dispersive and fibrous reinforcement of the porcelain structure hampers their increase. At a sufficiently high stress the rapid growth of critical cracks in the glassy-mullite matrix takes place. One of the samples, loaded up to the critical stress – 728 MPa showed also the characteristic branched cascade of cracks.

Fig. 10. Critical effects of the porcelain structure degradation at loading 839 MPa, magnification 200x. Long crack and damaged precipitates of mullite as well as small agglomerates of corundum are visible. All quartz grains and particles of cullet are crushed out.

CONCLUDING REMARKS There were performed mechanoacoustic, ultrasonic and microscopic investigations of two ceramic materials widely applied in contemporary technology. The corundum samples of lowdimensions were fabricated specially for the study, while the porcelain specimens were cut out from the rod of a high voltage line insulator. Technology of fabrication and internal structure for both ceramic materials is quite dissimilar. The corundum material belongs to monophase oxide ceramics. Aluminous porcelain is aluminosilicate material, characterized by complex multiphase composition, with dispersive and fibrous structural reinforcement. Different was the scale of production and size of the objects. The results enabled to compare mechanical and acoustic properties, microstructure, its texture and homogeneity of investigated ceramics. Additionally, the mechanoacoustic characteristics made it possible to distinguish subsequent stages of structure degradation process. It is supposed that there exists an analogy between the effects of long-term exploitation under working load and material degradation during relatively short compressive test, [11,12]. Very slow, quasistatic growth of loading and sensitive monitoring changes of structure by AE descriptors were applied, as the necessary condition to obtain similar effects as in case of long-term operational degradation. In the both materials there were discovered inhomogeneities. In case of porcelain they were of small importance and restricted to spatial distribution of mullite phase and presence of corundum agglomerates in the material. The best part of corundum samples contained distinct inhomogeneities: inclusions, fissures and partly broken grain boundaries. There was present also bimodal grain size distribution. The bigger grains often formed clusters or bands. This textural defect the most probably resulted from earlier stage of abnormal grain growth (AGG) effect. In spite of these disadvantageous factors, the compressive strength of the corundum samples was high – in average 3290 MPa, relative dispersion of this parameter did

21-Mechanical-acoustic.qxd

9/19/2009

11:59 AM

Page 213

Mechanical-acoustic and structural study of degradation processes in corundum ceramics and aluminous porcelain

213

not exceed 35 %. In case of the porcelain mean strength equaled 779 MPa and relative dispersion – about 41 %. For two tested materials there were distinguished three stages of the structure degradation. The preliminary and subcritical ones show low or moderate intensity of AE signals and considerable variety for the particular samples. The critical stage directly precedes decohesion of the samples. Its width is relatively narrow and contains AE activity of high energy. The primary stage is connected with development of defects, which initiation took place during any technological operations. These defects introduce internal stresses. Their relaxation has low threshold energy and generates AE signals of low intensity. In case of corundum it is related to restricted growth of intergranular cracks. The stress range of this stage is very wide and can approach even 2500 MPa. For the porcelain there occurs primarily degradation of quartz phase as a result of so called quartz stresses. The range of stress comprises up to 300 MPa. The subcritical stage is strongly varied for the particular samples and shows low or moderate intensity of AE activity. Signals occur separately or in groups. The range of stresses and descriptors of AE signals are characteristic for the samples and usually considerably differ between themselves. The interval of stresses for corundum material is much wider because of its high strength, when compared to porcelain. The subcritical stage is connected with the inhomogeneity of the sample structure in micro and semi-macro scales. In the corundum material it is mainly resulted from textural defects – bigger grains grouped into clusters or bands. Clusters of bigger grains are more susceptible to intergranular cracks growth. There is observed propagation of cracks, especially in the central part of samples as well as longwise elongated grains, and in direction perpendicular to compressive load. As a consequence, degree of structure loosening increases as the stress grows. In the structure of porcelain there are propagated cracks in the precipitates of mullite and agglomerates of collected grains of corundum. This results in crushing out pieces of some precipitates and agglomerates. The basic element of porcelain structure, reinforced matrix, during preliminary and subcritical stages remains almost unaffected. The critical stage is the shortest one and closely depends on the sample strength. Propagation of long critical cracks is followed by very strong continuous AE activity. Fracture and splitting off greater pieces (walls and corners) from sample precede destruction of the specimen. This effect is more frequent in porcelain. In case of the corundum material long cracks appear mostly in the areas of bigger grains clusters. At sufficiently high stress very fast propagation of these cracks takes place. This rapid process occurs not only along boundaries but also through the grains of different size. In case of porcelain, the characteristic phenomenon, during the critical stage, is propagation of cracks in matrix reinforced by dispersed mullite and corundum. At the beginning long cracks develop between precipitates of mullite, especially damaged. At suitable high stress the cracks and seldom its cascades propagate trough all phases of porcelain body. The presented sequence of degradation effects, concerning mechanics and components of structure, occurs during long-term exploitation of the material under working load [11]. Performed lastly research of the porcelain C 130 taken from broken insulator of foreign production also appear to confirm described results.

REFERENCES 1. IEC Publication 672-1:1995 Ceramic and glass-insulating materials, Part 1: Definitions and classification.

21-Mechanical-acoustic.qxd

214

9/19/2009

11:59 AM

Page 214

Przemys³aw RANACHOWSKI, Zbigniew RANACHOWSKI and Feliks REJMUND

2. IEC Publication 672-2:1999 Ceramic and glass-insulating materials, Part 2: Methods of test. 3. IEC Publication 672-3:1997 Ceramic and glass-insulating materials, Part 3: Specifications for individual materials. 4. Pohl, Z., Outdoor high voltage insulation in electrical engineering (in Polish), published by Wrocáaw Technical University, 2003. 5. Carty W., Senapati U., Porcelain – Raw Materials, Processing, Phase Evolution and Mechanical Behavior, J. Am. Ceram. Soc., 81 (1), 3-20, 1998 6. Dziadkowiec J., Kupiec E., Aging processes In ceramic insulators (in Polish), Energetyka, 5, 166-170, 1992 7. Liebermann J., Avoiding Quartz in Alumina Porcelain for High-Voltage Insulators, American Ceramic Society Bulletin, 80, No. 6-7, 37-48, 2001 8. Richerson D. W., Modern Ceramic Engineering, Properties, Processing and Use in Design, CRC Taylor & Francis, Boca Raton London New York, 2006 9. Ranachowski J., Rejmund F., Acoustic Emission in Technical Ceramics, in: Acoustic Emission – Sources, Methods, Applications, (in Polish)editors: I. Malecki, J. Ranachowski, edited by Biuro PASCAL, Warsaw, 55-107, 1994 10. Evans A.S., Langdon T. G., Structural Ceramics, in: Progress in Materials Science, editors: Chalmers S., Christian J. W., Massalski T. S. 21, Pergamon Press, 171-441, 1976 11. Ranachowski P., Rejmund F., Paweáek A., Piątkowski A., Structural and acoustic investigation of the quality and degradation processes of electrotechnical insulator porcelain under compressive stress, Proc. of AMAS Workshop on Nondestructive Testing of Materials and Structures II NTM’ 03, IFTR PAS Warsaw, 179-196, 2003 12. Ranachowski P., Rejmund F., Paweáek A., Piątkowski A., Studies of cordierite material under compressive load at different temperatures and after thermal shock (in Polish), Ceramics, 89, 101-115, 2005 13. Rios P.R., Abnormal grain growth development from uniform grain size distributions due to a mobility advantage, Scripta Materialia, 38, 9, 1359-1364, 1998

22-Descriptive.qxd

9/19/2009

11:59 AM

Page 215

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

DESCRIPTIVE MICROSTRUCTURE AND FRACTURE SURFACE OBSERVATIONS OF FIRED VOLCANIC ASH 1

C. LEONELLI1, E. KAMSEU1, 2, U. C. MELO2, A. CORRADI1, G. C. PELLACANI1 Department of Materials and Environmental Engineering, University of Modena and Reggio Emilia, Via Vignolese 905- 41100, Modena, Italy 2 Laboratory of Materials, Local Materials Promotion Authority, 2396 Yaoundé, Cameroon

ABSTRACT Crystals of the pyroxene group (diopside, augite and enstatite, hedenbergite), series of crystals with the general formula: (MgxFe1-x)2SiO4 having various geometry, identified as spinel (and olivine), and plagioclase crystals from anorthite to anorthoclase that grow together in mass having thin parallel groves embedded in a complex matrix together with calcium alumina silicate grains were found to be the descriptive microstructure of fired volcanic ash. Quartz grains were rarely present as confirmed by dilatometry analysis, XRD, SEM and DTA. The presence of dendrites continuously growing to pyroxene crystals indicated the precipitation/crystallization of these crystals from matrix and regions of glass concentration enhance by ions diffusion. Rings of Ti-rich iron micro-crystals observed around spinel (and olivine) suggested the probable nucleating role of these micro-crystals for the precipitation/crystallization phenomenon. The various type of crystals formed, the difference in their geometry and size and their interlocking mechanism results in a contiguous and dense structure with relevant characteristics at relative low temperature (1125-1150°C) confirm volcanic ash as promising alternative raw materials for vitrified ceramic products. It was concluded that controlled precipitation/crystallization of raw volcanic ash results on microstructure similar to that of glass-ceramic materials. The observation of fracture surface allowed comparison of fracture mechanics of volcanic ash ceramic to that of conventional vitrified ceramics. Key words Volcanic ash, microstructure, fracture surface, crystallization, thermal cycle

23-Dehydration.qxd

9/19/2009

12:00 PM

Page 217

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

DEHYDRATION, DEHYDROXYLATION, DENSIFICATION AND DEFORMATION DURING SINTERING OF GEOPOLYMERS BASED ON THE K2O-AL2O3-SIO2 SYSTEM E. KAMSEU1, A. RIZZUTI1, C. LEONELLI1, D. PERERA2 Department of Materials and Environmental Engineering, University of Modena and Reggio Emilia, Via Vignolese 905, 41100 Modena, Italy 2 School of Materials Science and Engineering, University of New South Wales, Sydney, NSW 2052, Australia 1

ABSTRACT Based on the principle of stability of geopolymer gel as refractory binder, a geopolymeric paste in the K2O-Al2O3-SiO2 system was developed and used to produce refractory concretes by adding various amount of D-quartz sand (grain size in the range 0.1Pm-1mm) and fine powder alumina (grain size in the range 0.1-100 Pm). The consolidated samples were characterized before and after sintering using optical dilatometer, DSC, XRD and SEM. The total shrinkage in the range of 25-900°C was less than 3%, reduced with respect to the most diffused potassium or sodium based systems, which generally records a >5% shrinkage. The maximum shrinkage of the base geopolymer was recorded at 1000°C with a 17% shrinkage which is reduced to 12% by alumina addition. The maximum densification temperature was shifted from 1000°C to 1150°C or 1200°C by adding 75 wt% D-quartz sand or fine powder alumina respectively. The sequences of sintering of geopolymer concretes could be resumed as dehydration, dehydroxylation, densification and finally plastic deformation due to liquid phase appearance. The geopolymer formulations developed in this study appeared as promising candidates for high temperature applications. Keywords Geopolymer, non-contact dilatometer, refractory binders, sintering, morphology

24-Multicriterial optimization.qxd

9/19/2009

12:00 PM

Page 219

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

MULTICRITERIAL OPTIMISATION OF AUTOCLAVED AERATED CONCRETE PROPERTIES AND EXPENDITURE OF ENERGY RESOURCES Tatiana V. LYASHENKO, Vitaly A. VOZNESENSKY, Varvara P. GAVRILIUK Odessa State Civil Engineering and Architecture Academy PO Box 76, Main Post Office, 65001 Odessa, Ukraine, e-mail: [email protected]

ABSTRACT The search for compromise between minimal heat conductivity and maximal toughness of autoclaved aerated concrete complying with specified requirements for density and compression strength is presented, the time of isothermal heating in autoclave and operating pressure of steam being minimised as well. To find the compromise iterative random scanning of material property fields in nine composition-process coordinates is carried out. The fields are described by experimental-statistical models built on the data obtained in specially designed experiment at industrial conditions. The models are used in computational experiments with Monte Carlo method. Composition-process solutions found for the materials of three strength classes and equal density are compared.

Keywords Autoclaved aerated concrete, heat conductivity, toughness, compromise, computational materials science, optimisation

INTRODUCTION Multifactor experimental-statistical (ES) modeling came into use in solving the problems of building materials science and technology more than 40 years ago. Among numerous studies carried out at that period the unique ones can be noticed (by hypotheses and approaches, experimental conditions and means, or responses obtained in experiment). However, a great part of useful information (data and knowledge) convoluted in the models could not be extracted from them for the lack of appropriate methods and computer means. It would be unreasonable to lose such information. It makes sense to revert to certain sets of ES-models with new tools, using the methods of computational materials science. These methods enable the complex problems of multicriterial optimisation [1] to be solved. In particular, the multifactor technological solutions, which would provide the compromise optimum for large number of the criteria of material quality and resources, can be found [2-3]. Built on results of physical experiments (fulfilled preferably to optimal design) ESmodels describe the fields of material properties (Y, characteristics of mix and material structure, any quality criteria, etc.) in coordinates of composition and process (CP) parameters (vector x of normalised CP-factors, «xi¸ d1). The models are used (in tandem with Monte Carlo method) to evaluate the levels of property fields for any variant in the CP-region under study, thus "representing" the real material in various computational experiments [4]. To find the optimal compromise between the best levels of optimality criteria (some properties Y and, possibly, some factors x), with requirements for levels of specified criteria

24-Multicriterial optimization.qxd

220

9/19/2009

12:00 PM

Page 220

Tatiana V. LYASHENKO, Vitaly A. VOZNESENSKY and Varvara P. GAVRILIUK

(some other Y) being fulfilled, the iterative random scanning of the fields Y(x) is carried out. In the case presented below the composition-process solutions for autoclaved aerated concrete technology is looked for with new version of the search procedure [5]: generated in computational experiments are only those levels of CP-factors (distributed by discrete uniform law between í1 and +1) which could be fixed within real technology.

CONDITIONS OF EXPERIMENT AND MODELLING The influence of mix proportions and process parameters on the structure and properties of autoclaved aerated concrete (AAC) was studied in cooperation with the institute "NIPISilicatobeton" [6]. The concrete blocks were produced there, at industrial conditions, in keeping with technology regulations for 56 different composition-process variants, according to specially synthesised optimal design of experiment, in which nine composition-process factors were varied. The design has been realised instead of two- or three-level full factor experiments 2k and 3k, with 512 and 19683 trials for number of factors k = 9. The CP-factors are presented in Table 1, with their levels in the experiment. Also indicated for each factor is regulation step hi (i = 1-9) assigned for generating the discreet values of the factors in computational experiment. The magnitude of h is dictated by

Index (i)

Table 1 Nine composition-process factors, their levels in experiment, and regulation conditions for computational experiments Physical experiment

Computational experiment

Factor ɯi = í1 ɯi = 0 ɯi = +1 Step hi

Number of values mi

1

Flowability of mix by Suttard viscosimeter, D (cm)

23

27

31

0.5

17

2

Moisture of sand (when grinded together with lime), ws%

2

5

8

0.5

13

3

Specific surface of the sand Ss (m2/kg)

150

250

350

20

11

4

Time of isothermal heating in autoclave, Ĳ (hour)

4

8

12

0.5

17

5

Operating pressure of steam, p (MPa)

0.8

1

1.2

0.05

9

6

Dosage of aluminum powder A% (by mass of dry components)

0.04

0.07

0.10

0.005

13

7

Mix activity (content of CaO in lime-sand mixture) aCaO%

14

17

20

0.5

13

8

Temperature of mix water tw oC

25

35

45

2

11

9

Quantity of cement C (% of mix mass)

0

10

20

1

21

24-Multicriterial optimization.qxd

9/19/2009

12:00 PM

Page 221

Multicriterial optimization of autoclaved aerated concrete properties and expenditure of energy resources

221

possibilities of specific regulation systems and engineering experience. This defines the number mi of possible values for each factor, with probability of each value being ɪi = 1/mi. Thus the total number of CP-variants of autoclaved aerated concrete that could be produced (in accordance with Table 1) equals 17u13u…u11 = 14.5·109. Physical-mechanical properties and thermo-physical characteristics along with porous space parameters (by the method of laser porosimetry) were determined in experiment for 56 autoclaved aerated concretes. The specimens (of 4u4 cm cross-section) were cut from the blocks made according to 56 variants of technology regulations. Determined, in particular, were the density of concrete Ȗ (kg/m3), coefficient of heat conductivity Ȝ (mW/m/Ʉ), compressive strength Rɫ (MPa), tensile strength Rb (MPa) and ultimate deformability H (mm/m) under bending. These data allowed the quadratic 9-factor ES-models to be built for material properties. The models have been admitted for engineering analysis after eliminating (with sequential regression analysis) the insignificant estimates from initial set of 55 coefficients of each model. Such model for deformability (1), with 22 significant coefficients, describes the field of H in nine CP-coordinates, with maximal level Hmax = 1.6 and minimum Hmin = 0.1 mm/m.

İ =

0.86 +0.04x1 r0x12 r0x1x2 r0x1x3 í0.02x1x4 +0.02x2 í0.13x22 +0.02x2x3 r0x2x4 í0.02x3x4 r0x3 r0x32 +0.03x4 r0x42 +0.08x5 r0x52 +0.04x6 r0x62 í0.03ɯ7 í0.06x72 r0ɯ8 r0x82 í0.04ɯ9 r0x92

r0x1x5 r0x1x6 r0x1x7 r0x2x5 r0x2x6 r0x2x7 r0ɯ3ɯ6 +0.02x3x6 r0x3x7 r0x4x5 r0x4x6 +0.02x4x7 r0x5x6 r0x5x7 r0x6x7

r0x1x8 r0x1x9 r0x2x8 r0x2x9 r0x3x8 +0.04x3x9 r0x4x8 í0.02x4x9 r0x5x8 í0.03x5x9 r0x6x8 í0.02x6x9 í0.02x7x8 r0x7x9 +0.01x8x9 (1)

This and similar models for Ȗ, Rc, Rb, and Ȝ are used to solve the problem of multicriterial optimisation for autoclaved aerated concrete of mark D800 (density Ȗ between 740 and 840) and class 5 (strength Rc between 6.4 and 9.6). The index of toughness H = Rb · H defining crack resistance should be maximised. Minimised are heat conductivity Ȝ and the levels of two energy consuming factors of autoclave technology, heating time Ĳ and steam pressure p (factors ɯ4 and ɯ5).

SEARCH FOR THE COMPROMISE SET Main points of the algorithm At initial stage of the first iteration (stage "1/0") Ng random numbers distributed by discreet uniform law in the full range (í1 d xi d +1) of each factor are generated. The random points thus obtained simulate Ng variants of autoclaved aerated concrete technology within the whole region of CP-parameters under study (k-dimensional cube, k=9). Added to them are 2k vertices of the cube (xi = r1). In this particular case 29=512 points have been added to Ng=10000, i.e., N1/0 = 10512 variants are in competition at stage "1/0" (more than 6 orders less than total number by Table 1). For N1/0 composition-process variants the levels of 6 property fields (of Ȗ, Rc, Ȝ, and ɇ, that should be scanned, Rb and İ under control) are determined by ES-models. Thus the matrix of the results of computational experiment is

24-Multicriterial optimization.qxd

222

9/19/2009

12:00 PM

Page 222

Tatiana V. LYASHENKO, Vitaly A. VOZNESENSKY and Varvara P. GAVRILIUK

F

3/2

3/1

2/2

2/1

1/2

1/1

formed (matrix size being 10512u6). At the next stage "1/1" the rows, which do not satisfy the requirements, firstly, of the mark (740 d Ȗ d 840), then of the class (6.4 d Rc d 9.6), are deleted from the matrix. The CPvariants (696 in this case) complying with requirements of specifications (for Ȗ and Rc) belong to the region of admissible solutions (RAS) remaining for the search. The levels of two optimality criteria (Ȝ, H) for admissible variants turn out to be restricted in their ranges of possible values (RPV). In particular, RPV{ Ȝ } = 211í139 = 72 mW/(m·Ʉ) (Figure 1); succeeding in advancing half the RPV will decrease the upper value of heat conductivity coefficient by near 20%. RAS defines the ranges of admissible levels for each of CP-factors (Figure 2). In this particular case, 8 from 9 factors hold their levels between í1 and +1 at stage "1/1". Only the range of the dosage of gas-former (x6) has contracted by half. At final stage of the 1st iteration ("1/2") the improvement of each optimality criterion within its RPV is carried out, from the worst level (Ȝ = 211, H = 0.16, Figure 1) towards the best one (Ȝ = 139, H = 3.34). Therewith two ranges (of importance for technologist) form: purposefully widened range of the gain in criterion level (RG, vertical shade in Figure 1) and the “pulsing” range of compromise (RC). The steps for each criterion in step-by-step improvement are chosen by technologist in the dialog with a computer. After 1st iteration N1/2 = 5 variants of the technology fall within RC{Ȝ} and RC{H}. The index of effectiveness when minimising Ȝ equals E{Ȝ} = RG{Ȝ} / RPV{Ȝ} = (211í159) / 72 = 0.72, the level of E{ɇ} being practically the same since no priorities have been assigned to the criteria. The factor ranges have shortened.

210

Ȝ

3.2 ɇ 2.7

Stages of iterations

2.89 2.2

185

1.7 1.2

152

160

0.7

F

3/2

3/1

2/2

2/1

135

1/2

1/1

0.2

İ

R 3.5 b 3

1.1

2.5

2.56

2 1.5

Figure 1. Changes of ranges of optimality criteria (Ȝ, H) and the criteria under control (Rb and İ) at stages of the search for compromise

F

3/2

3/1

2/2

2/1

1/2

1/1

F

3/2

3/1

2/2

2/1

1/2

0.1

1/1

1

1.13 0.6

24-Multicriterial optimization.qxd

9/19/2009

12:00 PM

Page 223

Multicriterial optimization of autoclaved aerated concrete properties and expenditure of energy resources

0.5

0.5

0

23.5 cm

-0.5

ɯ6 ( A)

1

ɯ1 ( D)

1

0

-0.5

-1

0.045%

-1

1/1 1/2 2/1 2/2 3/1 3/2

F

1 1

5% 0

-0.5

ɯ7 ( aCaO)

ɯ2 ( WS)

0.5

19.5 %

0.5 0

-0.5

-1

-1

1

ɯ3 ( SS)

0.5 0

1

190 2

m /kg

0.5

ɯ8 ( tW)

-0.5 -1

-0.5

1

-1

ɯ4 ( W)

0.5 0

-0.5

0

35 ɋ 0

8.5 h

1

-1

ɯ9( C)

0.5

1

ɯ5( p)

0.5 0

-0.5 -1

1.1 MPa

0

4%

-0.5 -1 1/1 1/2 2/1 2/2 3/1 3/2

Stages of iterations

Figure 2. Changes in ranges of composition-process factors along the search for compromise optimum

F

223

24-Multicriterial optimization.qxd

224

9/19/2009

12:00 PM

Page 224

Tatiana V. LYASHENKO, Vitaly A. VOZNESENSKY and Varvara P. GAVRILIUK

However, some CP-variants, near the border of multi-factor region, with useful combination of optimality criteria levels and decreased expenditure of resources could be found not generated for random scanning of 9-factor fields at this iteration. So it is reasonable to widen the narrowed ranges at least by one step h{xi} in both directions. In thus widened CP-region new Ng = 10000 random CP-variants are generated at stage "2/0", to which N1/2 = 5 best variants from the 1st iteration are added. At stage "2/1" excluded are not only the variants inadmissible by specifications (by Ȗ and Rc), but also those which have the levels of optimality criteria worse than achieved at previous iteration. Then the procedure is repeated. As compromise ranges of all criteria decrease (to 7-10% of RPV) it is reasonable to seize an opportunity to reduce the expenditure of resources, specifically, by energy consuming factors. In this particular case time Ĳ of isothermal heating in autoclave (x4) and operating pressure of steam p (x5) have been minimised at stage "3/2", practically without sacrifice of material quality. It has been possible to lower the upper limits of compromise ranges RC3/1 (containing 63 variants of technology) by 3.5 h and 1 atm respectively. As a result N3/2 = 4 competing variants have remained in RC3/2. The effectiveness indices of compromise optimisation: by heat conductivity E{Ȝ} = 0.81, by toughness E{H} = 0.83, the properties under control being also improved – ȿ{Rb} = 0.70 and E{İ} = 0.68. The variants remaining at final stage are of equal value from engineering point of view. The choice of final variant (F) could be subjective or based on new additional criteria. The best composition-process solution for ACC of strength class 5 The chosen variant F of composition and process parameters allows competitive autoclaved aerated concrete of mark 800 (Ȗ = 744 kg/m3) and class 5 (Rc = 9.0 MPa) to be produced, with optimised levels (Fig. 1) of heat conductivity (Ȝ = 152 mW/(m·Ʉ)) and crack resistance (characterised by index of toughness H = 2.9 kJ/m3). These values of the quality criteria are provided by CP-variant of reduced energy consumption (Fig. 2): flow of the mixture D = 23.5 ɫm, moisture of sand ws = 5%, specific surface Ss = 190 m2/kg, time of isothermal heating Ĳ = 8.5 h, pressure of steam p = 1.1 MPa, dosage of aluminium powder A = 0.045%, mix activity aCaO = 19.5%, temperature of mixing water tW = 35oC, quantity of cement C = 4%. Comparison of solutions found for equal-in-density autoclave cellular concretes of various classes by strength Technological variants providing the compromise between levels of quality criteria and energy consumption have been found, in particular, for the concretes of the same mark 800 by density but of different class by strength – 3.5, 5, 7.5. The plots in Figures 3 showing the tendencies of changes in optimal compromise solutions from class to class help to compare and analyse these results. Heat conductivity criterion is not shown since values of O (strongly correlated with Ȗ) practically do not vary from class to class. The bending strength Rb of AAC (correlated with Rc) grows linearly with increase of compression strength along abscissa axis. It is quite different with compromise deformability H that does not always follow the increased strength. Furthermore, the best level of H obtained for 3.5-class remains practically the same for class 5. It was noted when analysing the data of natural experiment (on AAC of various density) that with lowered loading capacity of pore walls the deformability could increase. So in case of equally dense AAC too the decrease of strength could be accompanied by retention or increase of deformability. The index of toughness H = Rb · H, which generalises strength and deformability criteria, grows rapidly with strength class, the influence of factors responsible for growth of Rb becoming prior.

24-Multicriterial optimization.qxd

9/19/2009

12:00 PM

Page 225

225

Multicriterial optimization of autoclaved aerated concrete properties and expenditure of energy resources

Y

3.5

Class by Rc 5

xi

7.5

1

4.0

H

x7

0.8

x5

3.5

0.6

x4

3.0

0.4

Rb 0.2

2.5

0 2.0

x2

-0.2 -0.4

1.5

H -0.6

1.0

-0.8

x1

-1

0.5 5

6

7

8

9

10

11

5

6

7

8

9

10

11

Strength Rc (MPa) Figure 3. From class 3.5 to class 7.5 of AAC: the changes in optimal compromise levels of quality criteria and technological factors at the same density and heat conductivity (Ȗ = 744-756, Ȝ = 150-160) Presented in Fig. 3 are also the optimal compromise levels xi.comp of normalised composition-process factors versus strength class of AAC. When moving from class 3.5 to class 7.5 the compromise levels vary by different values 'i = max{xi.comp}– min{xi.comp}. By these values (between 0 and 2) CP-factors have been separated into two groups: with 'i t 0.6 (Fig. 3) forming the increasing series (2) and with 'i d 0.4 written as (3). '7{ɚɋɚɈ} = 2 > '1{D} = '5{p} = 1 > '2{ws} = 0.67 > '4{Ĳ} = 0.62 '6{A} = 0.17 < '9{C} = 0.3 < '3{Ss}= '8{tw} = 0.4

(2) (3)

The factors from series (2) provide the strength of solid phase in AAC. Specifically, the activity of lime-sand mixture (x7) increases from 14% (í1) to 20% (+1), with considerable increase of steam pressure (x5) by 0.1 MPa. Difference in mix spread (x1) is also substantial,

24-Multicriterial optimization.qxd

226

9/19/2009

12:00 PM

Page 226

Tatiana V. LYASHENKO, Vitaly A. VOZNESENSKY and Varvara P. GAVRILIUK

but this parameter needs be decreased by reducing the quantity of water. The factors with less variation of optimal compromise values from class to class (3) influence mainly on the properties of semi-finished product and the porosity of autoclaved aerated concrete. CONCLUSIONS The energy saving solutions have been obtained for nine composition and process parameters of autoclaved aerated concrete technology in the search for compromise between minimal heat conductivity and maximal toughness. To find the solution of the multicriterial problem the algorithm has been used that could be recommended to search for multi-parametric admissible, optimal, and compromise solutions for mix-proportions and parameters of production process, especially when energy consuming factors are to be purposefully minimised. The algorithm of multicriterial optimisation at discrete levels of composition-process variables is well suited and comfortable for adapting, in dialog with a computer, the search parameters to flexible conditions of the problem. REFERENCES 1. Optimization Methods for Material Design of Cement-Based Composites. A.M. Brandt ed. E & FN Spon, London and New York 1998, pp 314 2. Lyashenko, T., Voznesensky, V., Compromise optimization of slag alkaline binders with computational materials science methods. In: Alkali Activated Materials – Research, Production and Utilization. Proc. Int. Conf. Prague 2007, 447-458 3. Lyashenko, T., Gara, A., Podagelis, I., Gailiene, I., Epoxy compositions for protecting road structure units in contact with water-oil mixtures. In: Environmental Engineering. Proc. 7th Int. Conf. Vilnius 2008, V. 3, 1186-1192 4. Voznesensky, V., Lyashenko, T., ES-models in Computational Building Materials Science (in Russian), Astroprint, Odessa 2006, pp 116 5. Voznesensky, V.A., Lyashenko, T.V., Gavriliuk, V.P., Compromise optimisation of gassilicate properties at discrete uniformly distributed levels of nine composition-process factors (in Russian). Bulletin of Donbas National Building and Architecture Academy, 1 (75), 2009, 139-145 6. Voznesensky, V.A., Gavriliuk, V.P., Kersh, V.Y. et al., Autoclaved aerated concrete: Nine-factor quadratic modelling (1981-85) and computational materials science (2007-08) (in Russian). In: Computational Materials Science and Advanced Technologies. Proc. 47th Int. Sem. MOCc47. Odessa 2008, 97-104

25-Application.qxd

9/19/2009

12:00 PM

Page 227

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

APPLICATION OF MACHINE LEARNING FOR PREDICTION OF CONCRETE RESISTANCE TO MIGRATION OF CHLORIDES Maria MARKS1, Daria JÓħWIAK-NIEDħWIEDZKA2, Michaá A.GLINICKI3 Institute of Fundamental Technological Research, Polish Academy of Sciences Pawinskiego 5 B, 02-106 Warszawa, Poland e-mail: 1 [email protected], 2 [email protected], 3 [email protected]

ABSTRACT The objective of this research was to develop rules for automatic categorization of concrete quality using selected artificial intelligence methods based on machine learning. The range of tested materials included concrete containing non-conventional additive of solid residue from coal combustion in fluidized bed boilers (CFBC fly ash). Performed experimental tests on chloride migration provided data for learning and testing of rules discovered by machine learning techniques. The rules generated by computer programs AQ21 and WEKA using J48 algorithm provided means for adequate categorization of plain concrete and concrete modified with CFBC fly ash as materials of good and acceptable resistance to chloride penetration.

Keywords Concrete durability, chloride ion migration, circulated fluidized bed combustion fly ash (CFBC fly ash), machine learning

INTRODUCTION The knowledge of relationships between the composition of concrete, its microstructure and technical properties, including durability in aggressive environments, is a primary objective of research within materials science of concrete. Due to a rapidly increasing number of concrete mix components and its properties, an increased use of complex technologies, a wide range of phase compositions and microstructural features, the simple engineering approach to these relationships might be insufficient. Modern computation methods that belong to the group of artificial intelligence methods could provide practical support to concrete technology. Kasperkiewicz [1, 2] demonstrated a variety of possibilities of using artificial intelligence methods in civil engineering problems. Three basic concepts are artificial neural networks, machine learning and genetic algorithms. In case of all these approaches the user is not obliged to bother about the model of the process or phenomenon, because the system itself gains the knowledge from examples. It can generate thereupon answers in the form of unknown values of the attributes, classification of new examples of the same format or formulation of rules (hypotheses, generalisations) concerning the process under consideration. More details were given in relation to the applied solutions of Fuzzy ARTMAP and ML program AQ19. Further attempts of using machine learning methods to support phase identification in concrete during indentation testing were reported in [3]. The objective of current research was to develop rules for automatic categorization of concrete quality using machine learning techniques.

25-Application.qxd

9/19/2009

12:00 PM

Page 228

Maria MARKS, Daria JÓZ´WIAK-NIEDZ´WIEDZKA and Micha³ A. GLINICKI

228

In the last decade, due to increased use of clean coal technologies in power generation, new types of coal combustion by-products became available. The composition and physical properties of new types of coal combustion by-products are significantly different than properties of well known fly ashes, widely used in concrete technology. The disposal problems of such non-conventional solid residues from coal combustion are growing, therefore some attempts were undertaken to apply such by-products for cement production or concrete mix production. Solid residues from coal combustion in circulated fluidized bed boilers, called circulated fluidized bed combustion (CFBC) fly ash, are characterized by different mineral and phase composition than conventional fly ash, by angular shape of grains and by the lack of glassy phase. In spite of such differences the research on the concrete strength development in time [4, 5] revealed promising perspectives of using fluidized bed fly ash in structural concrete. However, the durability of structural concrete modified with such an additive is still not well known. Therefore the undertaken research was focused on the resistance of concrete with fluidized bed fly ash to chloride ion aggression. Performed experimental tests on chloride migration provided data for learning and testing of rules discovered by machine learning techniques.

COMPOSITION OF CONCRETE MIXES AND TEST RESULTS OF CHLORIDE MIGRATION COEFFICIENT The chloride migration coefficient in concrete specimens with different content of fluidized bed fly ash was measured [6]. Ordinary Portland cement CEM I 32.5 R from Maáogoszcz cement plant, gravel fractions 2÷8 mm and 8÷16 mm, and sand fraction 0÷2 mm, were used. Two kinds of fluidized fly ash were tested: from hard coal combustion in the thermal-electric power station Katowice ‘K’ and from brown coal - lignite in power plant Turów ‘T’. Chemical and physical properties of Portland cement type I and CFBC fly ash are shown in Table 1. Table 1. Chemical composition and physical properties of Portland cement CEM I and fluidized bed fly ash from combustion of hard and brown coal [7] fluidized bed fly ash

Chemical compounds

PC type I

SiO2 Fe2O3 Al2O3 TiO2 CaO MgO SO3 Na2O K2O ClCaOfree

21.4 3.5 5.7 NA 64.1 2.1 2.1 0.5 0.92 0.029 0.9

from hard coal K 47.18 6.8 25.62 1.08 5.84 0.15 3.62 1.18 2.36 0.1 3.4

from brown coal T 36.47 4.4 28.4 3.84 15.95 1.65 3.8 1.64 0.62 0.03 4.75

Specific gravity [g/cm3]

3.15

2.68

2.75

Loss on ignition, 1000oC/1h

1.1

3.4

2.73

Three chemical admixtures were used: a plasticizer (magnesium lignosulfonates), a high range water reducer (polycarboxylane ether) and an air-entraining admixture (synthetic surfactants) were used to achieve approximately the same workability and porosity of fresh

25-Application.qxd

9/19/2009

12:00 PM

Page 229

229

Application of machine learning for prediction of concrete resistance to migration of chlorides

mix. Three concrete mixes were designed: series B with water to binder ratio w/b = 0.45, and air-entrained series C with w/b = 0.45 and series D with w/b = 0.42. In Table 2 the mixture proportions of tested concretes and the compressive strength of hardened concrete are shown. Table 2. Composition of concrete mixes and compressive strength tested after 28 days Concrete mix

Cement

Addition T

K

Aggregate Water

Plasticizer

HRWR

AEA

Content [kg/m3] B0 360 1859 162 3.2 4.3 B15K 306 54 1854 162 3.2 3.2 Series B B30K 252 108 1847 162 3.2 3.2 B15T 306 54 1850 162 3.2 4.7 B30T 252 108 1841 162 3.2 5.6 C0 380 1822 171 3.4 2.7 C15K 323 57 1813 171 3.4 2.5 Series C C30K 266 114 1803 171 3.4 3.4 C15T 323 57 1810 171 3.4 3.8 C30T 266 114 1800 171 3.4 4.8 D0 406 1586 175 0.0 D20T 290 73 1431 151 2.0 Series D D40T 217 145 1423 150 4.0 D20K 323 81 1593 167 2.2 D40K 244 162 1606 157 4.5 HRWR- high range water reducer, AEA- air-entraining admixture, 0-no addition, T - fluidized fly ash from brown coal, K - fluidized fly ash from hard coal

fc28 [MPa]

0.4 0.6 0.6 0.6 0.6 3.2 2.9 5.8 3.2 6.5

55.0 56.2 51.6 60.3 58.7 46.3 47.2 46.8 45.3 46.3 22.7 21.0 26.1 38.3 43.0

The design of concrete mixes was performed according to the experimental method with replacement of cement mass by fluidized fly ash: 15% and 30% in series B and C, 20% and 40% in series D. The specimens were cast in cubical moulds 150 mm and in cylinder moulds ø100 mm x 200 mm. Fresh mixes were consolidated by vibration. After 48 hours the specimens were demoulded and cured in high humidity conditions RH > 90% at temperature 18÷20 0C until the age of 28 days. Table 3. Estimation of the chloride resistance to chloride ions penetration, [9] Diffusion coefficient < 2 x 10-12 m2/s 2 – 8 x 10-12 m2/s 8 – 16 x 10-12 m2/s > 16 x 10-12 m2/s

Resistance to chloride penetration Very good Good Acceptable Unacceptable

The Nordtest Method Build 492 [8] was used to determine the chloride migration coefficient. The principle of the test is that concrete specimen is subjected to external electrical potential applied across it and chloride ions are forced to migrate into concrete. The specimens are then split open and sprayed with silver nitrate solution, which reacts to give white insoluble silver chloride on contact with chloride ions. This provides a possibility to measure the depth to which a sample has been penetrated. The conformity criteria proposed by L. Tang [9] are based on the voltage magnitude, temperature of anolite measured on the beginning and the end of test and the depth of chloride ions penetration are shown in Table 3.

25-Application.qxd

9/19/2009

12:00 PM

Page 230

Maria MARKS, Daria JÓZ´WIAK-NIEDZ´WIEDZKA and Micha³ A. GLINICKI

230

Table 4 presents the results of chloride migration coefficient determined after 28 days of maturity period for concretes series B, C and D. The highest values of Dnssm was obtained for concrete without fluidized fly ash replacement, both for non air-entrained and air-entrained. These values are only acceptable when criteria from Table 3 are used. Table 4. Results of tests of chloride ions penetration, series B, C and D (mean values from 3 specimens) Series B0 B15K B30K B15T B30T C0 C15K C30K C15T C30T D0 D20T D40T D20K D40K

Depth of chloride penetration [mm]

Dnssm [x 10-12 m2/s]

27.2 20.3 15.2 17.9 12.2 26.3 19.0 18.7 23.1 28.2 23.3 22.5 21.7 19.4 14.1

15.25 8.68 4.98 6.40 3.02 13.83 7.53 6.57 9.35 10.08 10.60 7.83 5.69 6.19 1.58

Resistance to chloride penetration Acceptable Acceptable Good Good Good Acceptable Good Good Acceptable Acceptable Acceptable Good Good Good Very good

MACHINE LEARNING METHODS Data mining can be defined as the process of discovering patterns in a dataset. By a dataset we mean a database, i.e. collection of logically related records. Each record can be called an example or instance and each one is characterized by the values of a set of predetermined attributes. A few different styles of learning appear in data mining applications but the most common is a classification [10]. The aim of the classification process is to learn a way of classifying unseen examples based on the knowledge extracted from the provided set of classified examples. In order to extract the knowledge from the provided dataset the attribute set characterizing the examples have to be divided into two groups: the class attribute or attributes and the non-class attributes. It is obvious that for an unseen examples only nonclass attributes are known, therefore the aim of data mining algorithms is to build such a knowledge model that allows predicting the example class membership based only on nonclass attributes. The knowledge model is dependent on the way how the classifier is constructed and it can be represented by decision trees (e.g. algorithm C4.5) or classification rules (the AQ algorithms family). Regardless of the representation both types of algorithms create hypotheses. In order to evaluate the classifier, i.e. to judge the hypotheses generated from the provided training set we have to verify the classifier performance on the independent dataset which is called testing set. Of course, both the training data and the test data should be representative samples of the underlying problem. The classifier predicts the class of each instance from the test set; if it is correct, that is counted as a success; if not - it is an error. In

25-Application.qxd

9/19/2009

12:00 PM

Page 231

Application of machine learning for prediction of concrete resistance to migration of chlorides

231

order to measure the overall performance of the classifier some quantitative analysis should be done. The example of such a quantitative measure can be classification accuracy. This is the number of correct classifications of the instances from the test set divided by the total number of these instances. The measure is expressed as a percentage. In a multiclass prediction, the result on a test set is often displayed as a twodimensional confusion matrix with a row and column for each class. Each matrix element shows the number of test examples for which the actual class is the row and the predicted class is the column. Good results correspond to large numbers down the main diagonal and small, ideally zero, off-diagonal elements. The classification accuracy is the sum of the numbers down the main diagonal divided by the sum of the all numbers in the matrix. Lets consider what can be done when the amount of data for training and testing is limited. The simplest way is to reserve a certain amount for testing and use the remainder for training. Of course the selection should be done randomly. In practical terms, it is common to hold out one-third of the data for testing and use the remaining two-thirds for training. The main disadvantage of this simple method is a problem that this random selection may be not representative. A more general way to mitigate any bias caused by the particular sample chosen for holdout is to repeat the whole process, training and testing, several times with different random samples. This process is called the k-fold cross-validation. In this technique you decide on a fixed number of folds – k. Then the data set U is split into k approximately equal portions ( U E1 ... Ek ) [11]. In each iteration i the set Ei is used for testing and the remainder U \ Ei is used for training. Overall classification accuracy is calculated as an average from the classification accuracy for each iteration K E i , i.e. is defined as: K

1 k ¦K E i . ki 1

(1)

In order to generate rules describing the concrete resistance to chloride penetration many numerical experiments were performed using program AQ21 and algorithm J48 from the WEKA workbench. Algorithm AQ21, invented in the Machine Learning and Inference Laboratory of George Mason University [12], is based on covering approach alike most of the rule-based data mining algorithms. Therefore, the AQ21 algorithm generates subsequent rules until all the examples (sometimes not all) are covered. The idea of adding new rule or new term to existing rule is to include as many instances of the desired class (positive examples) as possible and to exclude as many instances of other classes (negative examples) as possible. The second considered algorithm, J48, is available as a part of WEKA workbench, which was developed at the University of Waikato in New Zealand [13]. Algorithm J48 is an implementation of the last publicly available version of an algorithm C4.5 devised by J. Ross Quinlan. Construction of decision trees is based on a simple divide and conquer approach, which is well known in computer science. The main problem is connected with a selection of tests (splits of attributes) which should be placed in the nodes. The test is good if it allows to shorten the way from the root to the leaves representing classes. Decision trees can be converted to classification rules with ease.

SEEKING FOR THE RULES DESCRIBING CHLORIDE PENETRATION AFTER 28 DAYS Results obtained from AQ21 As the results of the experiments done on specimens of concrete with different content of fluidized fly ash, as shown in tables 2 and 4, the following database consisted of 15 records

25-Application.qxd

9/19/2009

12:00 PM

Page 232

Maria MARKS, Daria JÓZ´WIAK-NIEDZ´WIEDZKA and Micha³ A. GLINICKI

232

was introduced. This database can be used to determine the rules describing the concrete resistance to chloride penetration after 28 days. The database with one nominal and six numerical attributes is presented in Table 5. Table 5. The database pfK W A_fr fc28

C1

pfT

360

0

0

162

2.1

55.0

Resistance Acceptable

306

0

54

162

1.8

56.2

Acceptable

252

0

108

162

1.3

51.6

Good

306

54

0

162

1.6

60.3

Good

252

108

0

162

1.6

58.7

Good

380

0

0

171

6.2

46.3

Acceptable

323

0

57

171

6.8

47.2

Good

266

0

114

171

5.8

46.8

Good

323

57

0

171

6.6

45.3

Acceptable

266

114

0

171

6.2

46.3

Acceptable

406

0

0

175

4.9

22.7

Acceptable

290

73

0

151

6.9

21.0

Good

217

145

0

150

7.8

26.1

Good

323

0

81

167

4.6

38.3

Good

244

0

162

157

4.6

43.0

Good

where: C1 – content of cement, [kg/m3], pfT – content of fluidized fly ash from brown coal (power plant Turów), [kg/m3], pfK – content of fluidized fly ash from hard coal (power station Katowice), [kg/m3], W – content of water, [kg/m3], A_fr – air content in fresh mix, [%], fc28 – compressive strength of concrete tested after 28 days, [MPa], resistance – the resistance of concrete to chloride penetration (Acceptable, Good). The last attribute (resistance) is a nominal one which takes on two possible values (Acceptable, Good). In the considered database to the class [Resistance=Acceptable] belongs 6 examples and to the class [Resistance=Good] belongs 9 examples. The aim of an experiment is to generate the rules, which allow us to determine concrete resistance to chloride ions penetration. As an training set all the instances from the database were considered. The rules generated by an AQ21 algorithm are presented below: [Resistance=Good] # Rule 1 <-- [pfK>=55] : p=5,n=0,q=0.556 # Rule 2 <-- [C1<=258] : p=4,n=0,q=0.444 # Rule 3 <-- [pfT>=27 ] [W<=166] : p=4,n=0,q=0.444

(2)

25-Application.qxd

9/19/2009

12:00 PM

Page 233

Application of machine learning for prediction of concrete resistance to migration of chlorides

233

[Resistance=Acceptable] # Rule 1 <-- [pfK<=55] [A_fr=1.7..6.75 ] : p=6,n=0,q=1 # Rule 2 <-- [pfK<=55] [fc28=44.15..57.45] : p=5,n=0,q=0.833 where p denotes the number of positive examples covered by the rule, n denotes the number of negative examples covered by the rule (i.e. the number of records from the other classes satisfying the rule) and q denotes the quality of the rule. The rules showed in (2) can be interpreted as follows (it should be underlined that the presented rules concern concretes with the overall mass of cement and additions equal 360, 380 or 406 [kg/m3] (Table 2)). [Resistance is Good] IF [pfK >= 55] OR [C1 <= 258] OR [pfT >= 27] i [W <=166] [Resistance is Acceptable] IF [pfK <= 55] i [A_fr = 1.7..6.75] OR [pfK <= 55] i [fc28 = 44.15..57.45] The rules verified on a training set show the 100% classification accuracy, what is illustrated by an confusion matrix:

Acceptable Good

Acceptable 6 0

Good 0 9

To predict the performance of a classifier on new data, we need to assess its classification accuracy on a dataset that played no part in the formation of a classifier – the test set. In order to estimate the performance many numerical experiments where performed both with static set holdout and cross validation. In the first experiment the dataset was divided into training set consisting of two-thirds randomly selected instances and the testing set consisting of remainder instances. The following rules were generated by an AQ21 algorithm: [Resistance=Good] # Rule 1 <-- [pfK>=55] : p=4,n=0,q=0.667 # Rule 2 <-- [C1<=306 ] [pfT<=93 ] [fc28<=53.9] : p=4,n=0,q=0.667 # Rule 3 <-- [C1<=258] : p=3,n=0,q=0.5

25-Application.qxd

234

9/19/2009

12:00 PM

Page 234

Maria MARKS, Daria JÓZ´WIAK-NIEDZ´WIEDZKA and Micha³ A. GLINICKI

[Resistance=Acceptable] # Rule 1 <-- [C1>=298] [pfK<=55] : p=3,n=0,q=0.75 # Rule 2 <-- [fc28=44.15..46.55] : p=2,n=0,q=0.5 The rules generated on a dataset with ten examples taken from all the series (3 from series B, 4 from series C i 3 from series D) and verified on a test set can be described by an confusion matrix:

Acceptable Good so the classification accuracy is 80%.

Acceptable 2 1

Good 0 2

In order to estimate the classification accuracy the k-fold cross validation was also used. Assuming k=3 the classification accuracy obtained for each iteration is equal respectively 60%, 60% i 80%, so the overall classification accuracy is equal 66.7% (1). When the database consists of a very small number of records (less than 100) [11] the suggested value of parameter k is just the number of examples. Assuming k=15 we obtained the classification accuracy equal 53.3%. Results obtained from J48 In order to generate the rules, which allow us to determine concrete resistance to chloride ions penetration the J48 algorithm was also used. As an training set all the instances from the database (Table 5) were considered (the same training set was used in experiment described in section 4.1). The decision tree generated by an J48 algorithm is presented below: C1 <= 323 pfK <= 54 W <= 162: Good (5.0/1.0) W > 162: Acceptable (2.0) pfK > 54: Good (5.0) C1 > 323: Acceptable (3.0), where the first number in brackets denotes the number of examples from the training set covered by a selected leaf, and the second number – just after the sign "/" – indicates the number of incorrectly classified instances (negative examples). When there is only one number in brackets then it indicates the number of examples correctly classified (positive examples). The obtained decision tree can be easily transformed into the following rules: [Resistance=Good] Rule1 [C1 <= 323] i [pfK <= 54] i [W <= 162] Rule2 [C1 <= 323] i [pfK > 54] (3) [Resistance=Acceptable] Rule1 [C1 <= 323] i [pfK <= 54] i [W > 162] Rule2 [C1 > 323]

25-Application.qxd

9/19/2009

12:00 PM

Page 235

Application of machine learning for prediction of concrete resistance to migration of chlorides

235

The classification accuracy for the rules verified on a training set is illustrated by an confusion matrix:

Acceptable Good

Acceptable 5 0

Good 1 9

As one can see one example from Acceptable class is classified incorrectly to Good class, the remaining 14 examples are classified correctly, so the classification accuracy is equal 93.3%. When we held out one-third of the data for testing and used the remaining two-thirds for training (the same division as in section 4.1) we obtained the following results: pfK <= 54 W <= 162: Good (3.0/1.0) W > 162: Acceptable (3.0) pfK > 54: Good (4.0) i.e. [Resistance=Good] Rule1 [pfK <= 54] i [W <= 162] Rule2 [pfK > 54] [Resistance=Acceptable] Rule1 [pfK <= 54] i [W > 162] The rules generated on a training set and verified on a test set can be described by an confusion matrix:

Acceptable Good

Acceptable 1 0

Good 1 3

here also one example from Acceptable class is classified incorrectly to Good class. The classification accuracy is exactly the same as in experiment with static holdout from section 4.1 and is equal 80%. When we used a k-fold cross validation for J48 algorithm the following results were obtained respectively for k equal 3 and 15: for k=3 Acceptable Good Acceptable 2 4 Good 1 8 so the classification accuracy is 66.7%, for k=15 Acceptable Good so the classification accuracy is 60%,

Acceptable 2 2

Good 4 7

25-Application.qxd

236

9/19/2009

12:00 PM

Page 236

Maria MARKS, Daria JÓZ´WIAK-NIEDZ´WIEDZKA and Micha³ A. GLINICKI

CONCLUSIONS The rules generated by computer programs AQ21 and WEKA using J48 algorithm provided means for automatic categorization of plain concrete and concrete modified with CFBC fly ash as materials of good and acceptable resistance to chloride penetration. Due to a small number of tested specimens the rules are applicable only to concrete mix composition of similar binder content and similar values of water to cement ratio. Application of AQ21 and WEKA programs resulted in similar estimation of concrete resistance to chloride ion penetration. Further tests are needed for enlargement of experimental data basis. REFERENCES 1. Kasperkiewicz J., On possibilities of application of artificial inteligence methods in civil engineering (in Polish), Drogi i Mosty, nr 3, 2004, 15-36 2. Kasperkiewicz J., The applications of ANNs in certain materials-analysis problems. Journal of Materials Processing Technology, 106 (2000), 74-79 3. Kasperkiewicz J., Marks M., Woáowicz J., Possibilities of acoustic identification of phases in concrete (in Polish), LIV Konferencja Naukowa KILiW PAN i KN PZITB, Krynica 2008, Tom V, 325-332 4. Roszczynialski W., NocuĔ-Wczelik W., Gawlicki M., Maáolepszy J., Fly ashes from fluidized bed coal combustion as a cement additive, (in Polish) Ceramics 66, 2001, Polish Ceramic Bulletin, 370-377 5. Glinicki M.A., àadyĪyĔski K., The influence of addition of activated fly ash from fluidized bed combustion on properties of structural concrete (in Polish), VIII International Conference „Ashes from power generation”, UPS, Miedzyzdroje 2001, 119133 6. JóĨwiak-NiedĨwiedzka D., Effect of fluidized bed combustion fly ash on the chloride resistance and scaling resistance of concrete, Concrete in Aggressive Aqueous Environments, Performance, Testing and Modeling’, 02-05 June 2009, Toulouse, France, RILEM proceedings PRO 63, vol. 2, 556-563 7. àagosz A., Maáolepszy J, ĝliwiĔski J., Tracz T., Utilization of fly ash from fluidized bed boilers as a mineral additive for concretes, (in Polish), V Konferencja „Dni Betonu – Tradycja i NowoczesnoĞü”, Wisáa, 2008, 719-727 8. Nordtest Method NT Build 492. Concrete. mortar and cement-based repair materials: Chloride migration coefficient from non-steady-state migration experiments. 1999 9. Tang L., Chloride transport in concrete – Measurement and prediction, Publication P-96:6, Chalmers University of Technology, Department of Building Materials, Göteborg 1996 10. Cichosz P., Learning systems (in Polish), WNT, Warszawa 2000 11. Krawiec K., Stefanowski J., Machine learning and neural networks (in Polish), Wydawnictwo Politechniki PoznaĔskiej, 2003 12. Wojtusiak J., AQ21 user’s guide, George Mason University, MLI 04-3, September 2004 13. Witten I.H., Frank E., Data mining. Practical machine learning tools and techniques, Elsevier, 2005

26-Carbon.qxd

9/19/2009

12:00 PM

Page 237

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

CARBON SPHERES AS POSSIBLE MICRO-REINFORCEMENT OF CEMENT-BASED COMPOSITES Jan M. SKOWROēSKI1, Agnieszka ĝLOSARCZYK2 PoznaĔ University of Technology, 1 Institute of Chemistry and Technical Electrochemistry, ul. Piotrowo 3, 60-965 PoznaĔ Poland, e-mail: 2 Institute of Structural Engineering, ul. Piotrowo 5, 60-965 PoznaĔ, Poland e-mail: [email protected], [email protected]

ABSTRACT In this research, the influence of carbon spheres additive on mechanical properties of cement-carbon composites, such as flexural strength, flexural toughness and compression strength, were investigated. Carbon spheres were added to cement mortars in amount of 0.5 wt. %. Chemical treatment in hot nitric acid was applied in order to improve interfacial bonding between carbon fibres and cement mortar. It was found that the addition of carbon spheres in the amount of 0.5 wt.% to cement matrix improved its flexural toughness. The best results were obtained for cement composites subjected to chemical treatment. As proved by electrochemical measurements, during modification of carbons in nitric acid the functional groups, such as carboxyl, hydroxyl or phenol, were build up on the surface of carbon spheres. Consequently, the bonding between carbon fibres and cement matrix was enhanced.

Keywords Carbon spheres; cement-carbon composites, chemical treatment, cyclic voltammetry; mechanical properties

INTRODUCTION While PAN- or pitch-based carbon fibres used as reinforcement in cement-carbon composites have been advanced over the past twenty years [1-5], no attention has been devoted to cement composites modified with carbon spheres (CS) obtained from mesophase pitch. Carbon spheres are commonly used as precursor to obtain high-performance carbon materials or active carbons with high developed specific surface, among others as the components of carbon/carbon composites [6,7]. Carbon spheres are composed of microcrystallites, in which numerous spaces are created during solvent extraction. Due to their specific structural properties, among others spherical shape, carbon spheres enable the formation of isotropic properties of composite. This article focuses on the improvement of mechanical properties, especially flexural toughness, of cement matrix due to the addition of carbon spheres. Carbon spheres were chemically oxidised in nitric acid at the temperature of 120°C in order to enhance the interfacial bonding between carbons and cement paste. Cyclic voltammetry measurements were applied to evaluate the surface properties of unmodified and modified carbon spheres.

26-Carbon.qxd

238

9/19/2009

12:00 PM

Page 238

Jan M. SKOWROÑSKI and Agnieszka ŒLOSARCZYK

EXPERIMENTAL Materials and preparation procedure Portland cement type I (CEM I 42.5 R produced by Gorazdze Corp.), silica sand (Certificate IMMD Krakow, obtained from Kwarcmix Tomaszow Mazowiecki) and distilled water, in relation 1:3:0.5, respectively, were used to make cement mortars. Silica fume with bulk density of 0.65±0.1 kg/dm3 and superplasticizers Addiment FM 95 in amount of 2 wt. % (both components obtained from Sika Poland Corp.) were added to mortar containing carbon spheres. Three types of mortars were used, namely: mortar with 10 wt. % additive of silica fume (w/c = 0.45), mortar with 10 wt. % of silica fume and unmodified CS (w/c = 0.45) and mortar with 10 wt. % of silica fume and modified CS (w/c = 0.45). The mixing process was performed in a Hobart mixer with a flat beater according to procedures described in PN-85/B04500. All components were mixed and poured into oiled moulds. Then the external vibrator was used to remove air bubbles. Specimens were demolded after 24 h and stored in water till tests were performed. CS supplied by Osaka Gas Corp. were added to mortars in amount of 0.5 wt. % in relation to the mass of cement. The chemical treatment involved the oxidation of CS in concentrated nitric acid. The procedure describing the chemical modification of carbon materials was presented earlier [8, 9]. Cyclic voltammetry (CV) measurements Electrochemical measurements were carried out by the cyclic voltammetry (CV) method in a three-electrode system using a scan rate of 10 mV/min. As-received CS and CS exposed to oxidative treatment, used in an amount of 400 mg, played the role of working electrode. The potentials were measured against the Hg/HgSO4/1 M H2SO4 reference electrode while the platinum wire served as a counter electrode. Cyclic voltammetry sweeps were started at the rest potential of the electrode, the value of which was changed in the negative direction down to – 0.7 V. Then, the direction of polarization was reversed and the potential was increased to 0.5 V. All the experiments were performed in 0.25 M aqueous solution of H2SO4 at a temperature of 20 qC. Electrochemical measurements were carried out using a AUTOLAB potentiostat-galvanostat produced by EcoChemie. Scanning electron microscopy (SEM) analysis SEM analysis of cement-carbon composites was performed by the use of scanning electron microscope, Tescan - Vega, model 5135. Mechanical properties The Instron press type 9020 was used in order to determine the flexural strength and flexural toughness of mortars. Six specimens of each type of mortars with dimensions 4x4x16 cm were used to measure the flexural strength. The value of load in relation to displacement at constant rate of 0.01 mm/s was recorded by means of three-point bending tests. The flexural toughness was calculated from the area under the load/deflection curve obtained in the above test. The halves of specimen broken during the flexural strength test underwent compression strength test (on lateral surfaces 40x40 mm).

26-Carbon.qxd

9/19/2009

12:00 PM

Page 239

239

Carbon spheres as possible micro-reinforcement of cement-based composites

RESULTS AND DISCUSSION Cyclic voltammetry Cyclic voltammetry measurements were carried out in order to evaluate chemical modification of carbon spheres. CV is an useful technique allowing the determination of chemical composition of carbon surface, especially the surface carbon oxides. When electroactive substances are present in solution or are reduced/oxidised at the electrode surface, the current peaks are responded from linearly scanned potential. The current peaks are attributed to reversible or irreversible electrochemical reaction. Moreover, unlike conventional chemical treatments used to identify functional groups on carbon surfaces, electrochemical measurements can be made on relatively small samples or samples with low surface area [10, 11]. 0,2 0,1 0,0

I, mA

-0,1 -0,2 -0,3 CS CSox

-0,4 -0,5 -0,6 -0,8

-0,6

-0,4

-0,2

0,0

0,2

0,4

0,6

E, V

Fig. 1 Cyclic voltammograms recorded in 0.25 M aqueous solution of H2SO4 for as-received and oxidised CS Figure 1 presents cyclic voltammograms recorded for unmodified and modified CS in 0.25 M aqueous solution of H2SO4. During the first reduction cycle, which corresponds to the cathodic polarization (potential is changed from the rest potential of electrode towards negative values), for unmodified CS two peaks were observed, the small-one at potential around -0.05 V and the second-one, much larger at about -0.4 V. On the other hand, oxidised CS exhibited the peaks at -0.05 V, -0.35 and -0.55 V. The peak at -0.05 V is associated with the redox reaction of quinone [12-15]. This reaction is reversible giving the anodic peak at the same potential when the polarization direction is changed into the oxidation mode. Two broad cathodic peaks which appear at -0.35 and -0.55 V for chemical treated CS and one peak at 0.4 V recorded for unmodified CS, prove the presence of different electroactive oxides on the carbon spheres surface. The positive shift of peak at -0.35 V noted for oxidised CS as compared with the peak for untreated CS at potential -0.4, is the feature indicating the higher electrochemical activity of oxides built up on the carbon surface during chemical treatment. As stated in ref. [16, 17], surface oxides present on modified carbon spheres can be assigned to organic functional groups, such as carbonyl, phenolic or carboxylic.

26-Carbon.qxd

9/19/2009

12:00 PM

Page 240

240

Jan M. SKOWROÑSKI and Agnieszka ŒLOSARCZYK

Mechanical properties The results of mechanical measurements performed after 28 days of curing cement-carbon composites are shown in Table 1. It is observed that 0.5 wt.% additive of modified and unmodified CS to cement matrix enhances flexural toughness of cement-carbon composites. Flexural toughness obtained for mortars with oxidised carbon spheres was about 10% higher than value corresponding to mortars with untreated carbon spheres and about 55% higher as compared to value for mortars without carbon spheres additive. The values of flexural and compression strength for mortars with additive of chemical treated as well as untreated carbon spheres were insignificantly higher than analogous value obtained for plain binder. Figure 2 presents the typical curves of cement mortars with and without carbon spheres. Unlike conventional short carbon fibres using as reinforcement of cement matrix, the carbon spheres additive influence cement matrix in its elasticity region, namely before the top of crack. Table 1 The comparison of flexural strength, flexural toughness and compression strength for cement mortars in relation to CS additive after 28 days of curing

Flexural strength, MPa Flexural toughness, MPa*cm Compression strength, MPa

B

B+CS

B+CSox

7,5

8,9

7,5

0,263

0,380

0,410

47,3

50,9

48,3

B – mortar with 10 wt.% of silica fume, B+CS - mortar with 10 wt.% addition of silica fume and 0.5 wt.% addition of untreated CS, B+ CSox - mortar with 10 wt.% addition of silica fume and 0.5 wt.% addition of CS exposed to nitric acid 4,0 B B+CS B+CSox

3,5

Load, kN

3,0

2,5

2,0

1,5

1,0

0,5 0,0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

Deflection, mm

Fig. 2 Load/deflection curves recorded for mortars and binders with and without CS

26-Carbon.qxd

9/19/2009

12:00 PM

Page 241

241

Carbon spheres as possible micro-reinforcement of cement-based composites

The analysis of load/deflection curves indicates that the time needed to destroy mortar specimens with oxidised carbon spheres is about 10% longer (displacement was recorded at constant rate of 0.01 mm/s) than for mortar modified with as-received carbon spheres and almost 40% longer than the time required to destroy plain binder. This results prove that the surface treatment of carbon spheres improves the bonding strength between carbons and cement matrix and, as consequence, leads to the increase of composite elasticity. The SEM analysis of cement-carbon composites Figure 3 (a, b) shows the SEM images obtained for plain mortar and mortar with 10 wt.% additive of silica fume. The structure of plain mortar made from Portland cement (CEM I 42,5 R) shows a good homogeneity. The comparison of microstructure recorded for plain mortar with microstructure obtained for mortar with SF indicates, that the use of SF denses the structure of cement paste additionally. The application of SF additive in cement matrix modified with carbon spheres is very important from practical point of view. The consolidation of cement paste improves carbon spheres adhesion to cement paste and enables their better dispersion in mortar.

(a)

(b)

(c)

(d)

Fig. 3 The SEM spectra of cement composites, (a) – plain mortar, (b) – mortar with 10% of SF, (c) – mortar with unmodified CS, (d) – mortar with modified CS

26-Carbon.qxd

242

9/19/2009

12:00 PM

Page 242

Jan M. SKOWROÑSKI and Agnieszka ŒLOSARCZYK

The microstructures of cement mortars with additive of unmodified and modified carbon spheres are presented in Fig. 3 (c, d). In the case of cement mortars reinforced with unmodified CS, as well as CS modified in nitric acid, the microstructures of composites differ significantly from microstructure obtained for plain mortar. In all mortar volume, carbon spheres are present, which correspond to the good dispersion of carbons in cement matrix. The analysis of SEM images shows that the structure of mortar containing chemically treated carbon spheres is much more compacted/condensed than the corresponding structure recorded for mortar with unmodified CS. The presence of carbon spheres in cement matrix contributes to better density of cement paste structure and, as a consequence, better adhesion between cement paste and aggregate, and higher flexural toughness of cement-carbon composite. CONCLUSIONS Mechanical properties of the cement-carbon composites reinforced with unmodified and modified carbon spheres have been investigated. Carbon spheres were surface-treated using nitric acid at the temperature of 120°C. It was found that the presence of 0.5 wt.% of CS in cement matrix improves its flexural toughness. Carbon spheres dense the microstructure of cement paste and cause the increase of mechanical parameters of cement matrix. The best results were obtained for cement composites reinforced with CS exposed to chemical treatment. As proven by the results of electrochemical measurements, during oxidative treatment in nitric acid, functional groups, such as carboxyl, hydroxyl or phenol, were build up on the carbon spheres surface. Consequently, the bonding between oxidised carbon spheres and cement matrix was enhanced. ACKNOWLEDGEMENTS This work was supported by The Committee for Scientific Research (KBN) under the grant No N N506/1566/33/2007. REFERENCES 1. Bentur A., Mindess S., Fibre reinforced cementitious composites, Taylor&Francis, London and New York, 2007 2. Ohama Y., Carbon-cement composites, Carbon, 27, 1989, pp. 729-737 3. Chung D. D. L., Carbon Fiber Composites, Butterworth-Heinemann Publication, 1994 4. Kucharska L., Brandt A. M., Pitch-based carbon fibre reinforced cement composites a review, Archiv. Civil. Eng., 43, 1997, pp. 165-186 5. Zheng Z., Feldman D., Synthetic fibre-reinforced concrete, Prog. Polym. Sci., 20, 1995, pp. 185-210 6. Song Y., Zhai G., Li G., Shi J., Guo Q., Liu L., Carbon/graphite seal materials prepared from mesocarbon microbeads, Carbon, 42, 2004, pp. 1427-1433 7. B. Wu, Z. Wang, Q. M. Gong, H. H. Song, J. Liang, Fabrication and mechanical properties of in situ prepared mesocarbon microbead/carbon nanotube composites, Mater. Sci. Eng. A, 487, 2008, pp. 271-277 8. SkowroĔski J. M., ĝlosarczyk A., The influence of modified and non-modified carbon´s additives on physicomechanical properties of cement composites, Przem. Chem., 85, 2006, pp. 862 9. SkowroĔski J. M., ĝlosarczyk A., Short pitch-based carbon fibres as micro-reinforcement in cement composites, in Press, Przem. Chem., 2009

26-Carbon.qxd

9/19/2009

12:00 PM

Page 243

Carbon spheres as possible micro-reinforcement of cement-based composites

243

10. Sholz F., Electrochemical Methods. Guide to Experiments and Applications, SpringerVerlag Berlin Heidelberg, 2002 11. Bard A. J., Electrochemical Methods. Fundamentals and Applications, J. Wiley&Sons, 1980 12. Kinoshita K., Beet J. A. S., Potentiodynamic analysis of surface oxides on carbon blacks, Carbon, Vol. 11, 1973, pp. 403-411 13. Epstein B. D., Dalle-Molle E., Mattson J. S., Electrochemical investigations of surface functional groups on isotropic pyrolitic carbon, Carbon, 9, 1971, pp. 609-615 14. Zielke U., Hüttinger K. J., Hoffman W. P., Surface-oxidized carbon fibers: I. Surface structure and chemistry, Carbon, 34, 1996, pp. 983-998 15. Moreno-Castilla C., López-Ramón M. V., Carrasco-Marín F., Changes in surface chemistry of activated carbons by wet oxidation, Carbon, 34, 1996, pp. 983-998 16. Pakuáa M., ĝwiątkowski A., Biniak S., Electrochemical behavior of modified activated carbons in aqueous and nonaqueous solutions, J. Appl. Electrochem., 25, 1995, pp. 10381044 17. Yang Y., Lin Z. –G., In situ IR spectroscopic characterization of surface oxide species on glassy carbon electrodes, J. Electroanal. Chem., 364, 1994, pp. 23-30

27-Properties.qxd

9/19/2009

12:00 PM

Page 245

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

PROPERTIES OF FIBER REINFORCED CEMENT COMPOSITES WITH CENOSPHERES FROM COAL ASH Waldemar PICHÓR Department of Building Materials Technology, Faculty of Materials Science and Ceramics AGH University of Science and Technology Mickiewicza 30, 30-059 Kraków, Poland, e-mail:[email protected]

ABSTRACT Cenospheres are lightweight, thin-walled hollow spheres which are by-products of the combustion of pulverized coal at thermal power plants. Due to their properties they are a potentially interesting filler and may be used for lightweight cement-based composites production. Several works show that the cement based composites with addition of cenospheres have good properties. These researches are focused on the properties of such composites with addition of relatively large amount of synthetic fibers. The main effect is manifested by an improvement of flexural behavior after cracking of cement matrix. The results of studies concerning the fiber reinforced cement composites with cenospheres as filler up to 60% of volume are presented in this paper. The influence of cenospheres content on the main mechanical properties (flexural strength, modulus of rupture), water sorption and thermal conductivity of fiber reinforced cement composites with different types of fibers is shown. The SEM observations of interfacial zone between cenosphere-cement matrix and fiber-cement matrix are presented. The results show that an usage of the cenospheres as lightweight filler may be a way to obtain fiber reinforced cement composites of low or moderate density.

Keywords cenospheres, fibers, fiber reinforced cement composite, interfacial transition zone, lightweight cement composite, thermal conductivity, modulus of rupture, toughness index. INTRODUCTION One of the main disadvantage of building materials such as mortars and concrete based on cement is their brittleness. Fracture process of the cement composite starts from a defect of cement matrix – microcracks where the stresses are concentrated. In the early stage the microcracks are generated during shrinkage and premature drying of cement paste. During further exploitation in variable loads, cracks are opening and the process leads to a degradation of the composite. The effect is especially critical for lightweight composites with a rather weak filler, e.g. expanded perlite, because cement matrix is responsible for the strength of composite filling spaces between lightweight aggregates. An effective method of reducing this disadvantageous phenomena in the case of mortars is to introduce randomly oriented synthetic short fibers into cement matrix. After cracking of matrix, a part of load is carried by fibers according to their properties, dimension and volume. Relatively good bond of fibers to cement paste and the absence of interfacial zone enriched in portlandite provides an effective transformation of load from matrix to the fibers. Cement composites with short fibers present multiple improved properties e.g. work of fracture or impact resistance even three orders higher when compared to plain cement paste.

27-Properties.qxd

246

9/19/2009

12:00 PM

Page 246

Waldemar PICHÓR

Mechanical properties are most important for structural materials but are also significant for lightweight cement composites. In this case, the effect of reduction of the bulk density may be obtained by several methods. Most frequently used method is introduction of the lightweight filler into the cement matrix, but in this case the adherence of filler to the paste and the presence of considerable volume of porosity into filler particles are important. The cenospheres from coal ash are the lightweight filler with practically closed porosity. The cenospheres are lightweight hollow spheres composed largely of silica and alumina and filled with air and/or combustion gases, mainly CO2. These are the naturally occurring byproduct of the burning process in coal-fired power plants and have the majority of the properties as manufactured hollow-sphere products have [1-4]. The properties of cenospheres make them possible to be used both in dry state or wet slurry form and to be easily to handled, providing a low surface area to volume ratio. Due to their properties, they are not affected by solvents, water, acids or alkalies. Cenospheres are much lighter than mineral matrix and are used as a filler; they are also lighter than most of polymer resins. The spherical shape of cenospheres improves flowability in most applications and provides better distribution of the filler material. Becouse the lightweight filler-cement paste matrix interaction may influence the composite strength and other properties, the studies of the interfacial zone are important [5-10]. The interfacial zone (ITZ) between hardened cement paste and aggregates has been studied extensively in recent years, after it was discovered that the structure of the paste in the vicinity of solid surfaces may differ distinctly from that of the bulk matrix [11]. A numerous studies concerning the interfacial zone are available in the literature for concrete with normal bulk aggregates but there are only a few referring to concretes with lightweight aggregates. For ordinary concrete with hard aggregate several models are proposed but they have very common elements: sub-layer of CH in direct contact with aggregate surface and CSH sublayer with ettringite crystals backing it, large portlandite crystals and porous layer which are smoothly dense to normal bulk C-S-H paste [12-15]. On the other hand, the interfacial zone was not observed for concretes with water/cement ratio below 0.40 [16]. The cenospheres have different features from typical lightweight filler, e.g. expanded perlite. They are small particles with smooth surface and insignificant open porosity. An interaction between the filler and cement paste similar to the polymer fiber-cement matrix in early stage of cement hydration is suspected [17], but after a long-term period of natural hardening of cement the pozzolanic properties of cenospheres appear. The pozzolanic reaction between cenospheres and Ca(OH)2 may lead to a densification of the interfacial region. Therefore, the mechanism of increasing porosity according to enlargement of water/cement ratio near the surface of cenosphere is unexpected. In this work, the results of studies concerning cement composites with cenospheres from coal ash and synthetic fibers are presented. The influence of cenospheres content on the main mechanical properties (flexural strength, modulus of rupture, toughness indexes) and thermal conductivity of fiber reinforced cement composites with different types of fibers are studied. The SEM observations of interfacial zone between cenosphere-cement matrix and fiber-cement matrix are presented. EXPERIMENTAL DETAILS AND TEST METHODS Materials In this study, different mixes of cement paste having varying cenosphere content and fibers volume were used. Cenospheres were recovered from surface of the sedimentation tank installed in Opole Coal Power Plant. Cenospheres are not commercially recovered in this place, but the method of recovery is under implementation. Before use, cenospheres were

27-Properties.qxd

9/19/2009

12:00 PM

Page 247

247

Properties of fiber reinforced cement composites with cenospheres from coal ash

cleaned by wet method from wastes, crushed parts and fine fly ash particles and dried at 105oC. Basic properties of cenospheres are presented in Table 1. Table 1. Physical properties of cenospheres Relative density, kg/m3 Bulk density, kg/m3 Thermal conductivity, W/(m·K) Particle size distribution, wt. % 0 – 0.063 mm 0.063 – 0,125 mm 0.125 – 0,25 mm 0.25 – 0,50 mm

~800 ~400 0.11 1.2 3.2 54.9 40.7

The cenospheres content in the mixes was maintained at 20%, 40% and 60% by volume. Two kinds of short polypropylene fibers (PP) and polyvinyl alcohol fibers (PVA) were added to cement matrix in constant amounts by volume. Properties of used fibers are shown in Table 2.

Type Fiber length, mm Fiber diameter, m Density, g/cm3 Tensile strength, MPa Young’s modulus, GPa

Table 2. Properties of fibers Polyvinyl alcohol (PVA) 6 14 1.30 1500 37

Polypropylene (PP) 6 32 0.91 300-500 3-5

Proportion of fibers by volume in cement matrix and in the composites are presented in Table 3. Portland cement CEM II 32.5R and plasticizer from BV group were used; w/c ratio was equal 0.35 for all mixes. Table 3. Fibers volume in composites, % Total fibers volume in composites due to cenospheres Fibers volume in cement volume, % matrix, % 20% 40% 60% 0.67 0.54 0.40 0.27 1.33 1.06 0.80 0.53 2.67 2.14 1.60 1.07 Preparation of specimens Fibers were added to water with plasticizer and small amount of cement and mixed for 30s. Then, cenospheres and the rest of cement were added and further mixed up for 3 minutes. The mixes were poured into 25×25×100 mm steel prism molds and 200×200×50 mm plate molds for thermal conductivity measurements. After conditioning for 24h in normal laboratory environment all specimens were demolded, then specimens were stored under water for 28 days. For microscopic observation, the specimens after 180 days of hydration in wet condition were also prepared. Test methods The thermal conductivity of composites at mean temperature 23oC was measured by steadystate method using FOX 200 (LaserComp, Inc.). The specimens were dried in 105oC to constant mass and weighted for bulk density before these measurements. The basic

27-Properties.qxd

9/19/2009

12:00 PM

Page 248

248

Waldemar PICHÓR

mechanical properties (limit of proportionality, stress-strain relation, modulus of rupture, toughness indexes) were obtained in third-point bending test method using QC508 B1 (Cometech Testing Machine, Co. Ltd) testing machine. A deflection rate of 1,0 mm/min was used for all tests. Modulus of ruptures was calculated for maximal stress carried by cement matrix after first crack had been initiated. The toughness indexes were calculated basing on the area under stress-strain curve according to proposed in ASTM standard method [18]. Scanning electron microscope JEOL equipped with EDAX LINK system was used to collect the observations of interfacial transition zone (ITZ) between cenosphere and cement matrix and also for such region near fibers surface. RESULTS The bulk density of composites mainly depends on cenospheres volume and observed effect of fibers amount was small in the range of used fibers addition. The results are presented in Fig. 1. Fibers, % vol.: 0.67 PP 1.33 PP 2.67 PP 0.67 PVA 1.33 PVA 2.67 PVA

1800

Bulk density, kg/m

3

1700 1600 1500 1400 1300 1200 1100 0

20

40

60

Volume of cenospheres, %

Fig. 1. Influence of cenospheres volume on the bulk density of composites with different types of fibers Differences between mean value of density obtained for fibers type are small and closed-up in the error range, but the tendency to decrease in the density for composites with polypropylene fibers is visible and probably caused by common effect of smaller density of its fibers. Weak effect of aeration during the mixing process for composites with polypropylene fibers compared to polyvinyl alcohol fibers is also observed, but only relatively large amounts of PVA fibers leaded to the reduction of bulk density of composites [19]. Addition of 20% by volume of cenospheres resulted in decrease of bulk density of about 12% for polyvinyl alcohol fibers and 17% for polypropylene fibers. In the case of 40% addition of cenospheres the reduction of density was 19% and 24%, and for 60% vol. of cenospheres the effect of density decrease was 30% and 32% for different fibers type, respectively. Thermal conductivity also depends on addition of cenospheres and similarly to the results of density the effect of fibers volume was small. Fig. 2 shows the relation between cenospheres content (by volume) and thermal conductivity of the composites. For the composites with smaller amounts of fibers the obtained results were identical. In both cases the small effect of decreasing the thermal conductivity was visible and was probably connected with small aeration of mixes.

27-Properties.qxd

9/19/2009

12:00 PM

Page 249

249

Properties of fiber reinforced cement composites with cenospheres from coal ash

Fibers, % vol.: 0.67 PP 1.33 PP 2.67 PP 0.67 PVA 1.33 PVA 2.67 PVA

Thermal conductivity, W/(mK)

0,60 0,55 0,50 0,45 0,40 0,35 0,30 0

20

40

60

Volume of cenospheres, %

Fig. 2. Influence of cenospheres volume on the thermal conductivity of composites Addition of 20% by volume of cenospheres caused a reduction of thermal conductivity from 0.55 W/(m·K) to 0.49 W/(m·K) for PVA fibers and 0.47 W/(m·K) for PP fibers. Composites with 40% vol. of cenospheres had about 22-23% better insulation properties than plain cement composite. In case of large amount of cenospheres (60% vol.) the thermal conductivity was effectively improved and reached value equal to 0.37 W/(m·K) for composites with PP fibers (about 67% reference value obtained for specimens without cenospheres). Fibers, % vol.: 0.67 PP 1.33 PP 2.67 PP 0.67 PVA 1.33 PVA 2.67 PVA

16 14

LOP, MPa

12 10 8 6 4 2 0 0

20

40

60

Volume of cenospheres, %

Fig. 3. Limit of proportionality (LOP) of stress-strain curve of composites with different volume of cenospheres In the case of composites with large volumes of fibers, the flexural strength of cement matrix may be smaller than the modulus of rupture obtained after brittle matrix cracked, then to compare the properties before brittle cement matrix cracked one may use the limit of proportionality (LOP) on the stress-strain curve [20]. The effect of cenospheres addition was dominant, and for 20% of cenospheres the reduction of the LOP was about 10%, for 40% vol. of cenospheres was about 20% and for maximal used volume (60%) was about 35% with little

27-Properties.qxd

9/19/2009

12:00 PM

Page 250

250

Waldemar PICHÓR

differences in each case depending on fibers content. The changes of the LOP value caused by addition of fibers were small with decrease of LOP obtained for composites with 2.67% vol. of fibers and were probably as a result of locally arrested air voids between bundled fibers. Flexural strength values of composites without fibers due to the cenospheres content were 10.9±0.3 MPa, 9.0±0.2 MPa, 7.9±0.3 MPa and 6.8±0.4 MPa, respectively. a) Modulus of rupture, MPa

6 5 4

Volume of cenospheres: without 20% 40% 60%

3 2 1 0 0,5

1,0

1,5

2,0

2,5

3,0

2,5

3,0

Volume of fibers, %

b)

16 Modulus of rupture, MPa

14 12 10

Volume of cenospheres: without 20% 40% 60%

8 6 4 2 0 0,5

1,0

1,5

2,0

Volume of fibers, %

Fig. 4. Modulus of rupture of composites with cenospheres according to volume of fibers: (a) polypropylene, (b) polyvinyl alcohol The modulus of rupture strongly depends on two factors: cenospheres content and fibers volume introduced to the cement matrix. Generally, as one may expect, in both cases the value of modulus of rupture decreased due to the volume of cenospheres but major effect was related to the load carried by fibers after brittle matrix had been cracked. Addition of polypropylene fibers to the plain cement paste resulted in the modulus of rupture value equal to 1.4 MPa for 0.67% vol. of fibers up to 4.4 MPa for 2.67% vol. (about 40% of flexural strength of plain cement matrix). Values of modulus of ruptures obtained for samples with maximal used volume of polypropylene fibers and cenospheres were 3.1 MPa for 20% vol. of cenospheres, 3.0 MPa for 40% and 2.3 MPa for composite with 60% vol. of cenospheres (about 35% of flexural strength of composites without fibers). Polyvinyl alcohol fibers have

27-Properties.qxd

9/19/2009

12:00 PM

Page 251

251

Properties of fiber reinforced cement composites with cenospheres from coal ash

about five times greater tensile strength than polypropylene fibers and also have relatively good adherence to the cement matrix, which caused much higher reinforcement effect after hardened cement matrix cracked. For samples without cenospheres and with 20% and 40% addition, the values of modulus of rupture were higher than flexural strength of matrix without fibers or larger amount of fibers. For these composites the polynomial fits of curves of the relation between modulus of rupture and fibers volume are proposed, with R-square at least 0.99. From the equations (1), (2) and (3) the point where the modulus of rupture was equal to flexural strength were calculated, respectively 1.8624V f2 10.398V f 0.1415

MOR0

(1)

2 f

(2)

2 f

(3)

1.4517V 7.9V f 0.2028

MOR20

1.1393V 5.9054V f 0.1936

MOR40

where MOR0 is a value of modulus of rupture obtained for composites without cenospheres, MOR20 with 20% vol., MOR40 for composites with 40% of cenospheres and Vf is the volume of fibers. a)

Fibers, % vol.: 0.67 PP 1.33 PP 2.67 PP 0.67 PVA 1.33 PVA 2.67 PVA

16

Toughness index I5

14 12 10 8 6 4 2 0 0

20

40

60

Volume of cenospheres, %

b)

Fibers, % vol.: 0.67 PP 1.33 PP 2.67 PP 0.67 PVA 1.33 PVA 2.67 PVA

16

Toughness index I10

14 12 10 8 6 4 2 0 0

20

40

60

Volume of cenospheres, %

Fig. 5. Toughness indexes I5 (a) and I10 (b) of composites due to cenospheres content

27-Properties.qxd

252

9/19/2009

12:00 PM

Page 252

Waldemar PICHÓR

For samples without cenospheres, the calculated effect of reinforcement of the flexural behavior of composite was from 1.37% of fibers volume, 1.56% for 20% of cenospheres and for samples with 40% addition of cenospheres was equal to 2.39%. For composites with 60% of volume of cenospheres flexural strength of composites without fibers were not reached. In Fig. 5 the toughness indexes calculated for all composites are presented. The indexes strongly depend on fibers volume and effect of cenospheres content were insignificant. For PVA fibers both the indexes (I5 and I10) were about twice greater than obtained for PP fibers. In the case of used fibers (with good adherence to the cement matrix), the dominant mechanism of increasing the work of fracture measured here by toughness indexes was rather connected to breaking fibers than the pull out process.

Fig. 6. SEM image of Interfacial Transition Zone (ITZ) of cenosphere and cement matrix Fig. 6 shows the region of Interfacial Transition Zone after 28 days of hydration of cement paste. The bulk paste on the cenosphere surface is formed. Because of the smooth and nonporous surface of cenosphere, the amount of surface-adsorbed water is very small. Thus, the effect of formation of porous layers around the cenospheres is not observed. As the consequence of the absence of the porous layer during next period of cement hydration, the ITZ enriched in Portlandite crystals were only local or not formed at all. After a long period of hydration the microstructure of cement paste in the vicinity of cenosphere was compact and similar to distant region from its surface.

Fig. 7. SEM observation of the polished sample after 180 days of hydration and dot maps of Si, Al, Ca and Fe distribution. The arrows indicate Ca-rich area

27-Properties.qxd

9/19/2009

12:00 PM

Page 253

Properties of fiber reinforced cement composites with cenospheres from coal ash

253

In Fig. 7 the SEM observation of the polished sample of naturally hardened cement paste with cenospheres for 180 days is shown. In this case the effect of calcium concentration on the interfacial region was observed only locally for cenosphere with diameter about 400-500 Pm. Generally, if the ITZ exists, especially enriched in large Portlandite crystals, it has a rather discontinuous form. During the hydration process the pozzolanic reaction between cenospheres surfaces and cement may also occurred. The presented X-ray analysis as dot maps of Ca and Si shows that on the surface of cenosphere the Interfacial Transition Zone (where the Ca/Si ratio is higher) may locally exist. The higher concentration of calcium in this case is probably caused by wall effect. Despite the pozzolanic properties of cenospheres, after 180 days of hydration the Ca-enriched zone was still present, however it can be noted again, that the phenomena has a rather occasional character. CONCLUSIONS Basing on the obtained results, following conclusions can be drawn: Bulk density of composites was proportionally decreased due to the volume of added cenospheres up to 1170 kg/m3 for composites with 60% by volume (about 70% of density of composite without cenospheres). Small decrease of bulk density was observed for composites with relatively high volume of added fibers (2.67% vol.) and the reduction of the bulk density was greater for polypropylene fibers. Thermal conductivity of composites with cenospheres strongly depends on the volume of cenospheres added, for maximal used addition (60% vol.) the value of thermal coefficient is about 0.37 W/(m·K). In this case, a small influence of fibers type was also observed. Mechanical properties of composites with cenospheres decrease with the volume of cenospheres, but the effect of fibers is dominant and in the case of polyvinyl alcohol fibers leads to the improvement of ultimate flexural behavior for composites with relatively large amount of fibers. Interfacial Transition Zone, especially rich in portlandite, was observed only for relatively large diameter cenosphere and had rather occasional character than continuous form. Generally, good adherence between cenospheres and cement matrix was observed. The results show that an usage of the cenospheres as lightweight filler may be a right way to obtain low or moderate density fiber reinforced cement composites with good thermal conductivity and sufficient mechanical properties for many applications. ACKNOWLEDGEMENTS This work was supported by the Ministry of Scientific Research and Higher Education Grant no. N506 1597 33. REFERENCES 1. Kolay P.K., Singh D.N., Physical, chemical, mineralogical and thermal properties of cenospheres from an ash lagoon. Cement and Concrete Research, 31, 2001, 539-542 2. Fisher G.L., Chang D.P.Y., Brummer M., Fly ash collected from electrostatic precipitators: Microcrystalline structures and they mystery of the spheres. Science, 192, 1976, 553-555

27-Properties.qxd

254

9/19/2009

12:00 PM

Page 254

Waldemar PICHÓR

3. Matsunaga T., Kim J.K., Hardcastle S., Rohatgi P.K., Crystallinity and selected properties of fly ash particles. Materials Science and Engineering: A, 325, 2002, 333-343 4. Vassilev S.V., Menendez R., Diaz-Somoano M., Martinez-Tarazona M.R., Phase-mineral and chemical composition of coal fly ashes as basis for their multicomponent utilization. 2. Characterization of ceramic cenosphere and salt concentrates. Fuel, 83, 2004 585-603 5. Wasserman R., Bentur A., Interfacial interactions in lightweight aggregate concretes and their influence on the concrete strength. Cement and Concrete Composites, 18, 1996, 67-76 6. Suryavanshi A.K., Swamy R.N., Development of lightweight mixes using ceramic microspheres as fillers. Cement and Concrete Research, 32, 2002, 1783-1789 7. Lilkov V., Djabarov N., Bechev G., Kolev K., Properties and hydration products of lightweight and expansive cements. Part I: Physical and mechanical properties. Cement and Concrete Research, 29, 1999,1635-1640 8. Lilkov V., Djabarov N., Bechev G., Petrov O., Properties and hydration products of lightweight and expansive cements. Part II: Hydration products. Cement and Concrete Research, 29, 1999, 1641-1646 9. Mc Bride S.P., Shukla A., Processing and characterization of lightweight concrete using cenospheres. Journal of Materials Science, 37, 2002, 4217-4255 10. Pichór W., The interfacial transition zone between filler and matrix in cement based composites with cenospheres. Kompozyty/Composites, 3, 2006, 71-77 11. Farrat J., Contribution minéralogique à l’étude de l’adhérence entre constituants hydratés des ciments et les matériaux enrobés. Matériaux et Constructions, 490-491,1956, 155-172 12. Maso J.C., The bond between aggregates and hydrates cement pastes. Proc. 7th Int. Congress Chemistry of Cement, Paris, 1980, vol.3, 7, 3-15 13. Breton D., Carles-Gibergues A., Ballivy G., Grandet J., Contribution to the formation mechanism of the transition zone between rock-cement paste. Cement and Concrete Research, 23,1993, 335-346 14. Zimbelman R., A Contribution to the problem of cement-aggregate bond. Cement and Concrete Research, 15, 1985, 801-808 15. Monteiro P.J.M., Improvement of the aggregate-cement paste transition zone by grain refinement of hydration products. Proc. 8th Int. Congress Chemistry of Cement, Rio, 1986, vol.3, 433-437 16. Diamond S., Huang J., The interfacial transition zone: reality or myth? The Interfacial Transition Zone in Cementitious Composites, London, 1998, 3-39 17. Pichór W., Dyczek J., Early formation of the interfacial zone in FRC with PAN fibers. Proc. Int. Symp. “Brittle Matrix Composites 5”, A.M. Brandt, V.C. Li and L.H. Marshall eds., Woodhead Publishing Ltd., Cambridge and Warsaw 1997, 74-78 18. ASTM C1018-97 Test Method for Flexural Toughness and First-Crack Strength of FiberReinforced Concrete (using beam with third-point loading). American Society of Testing and Materials,1988, 544-551 19. Bezerra E.M., Joaquim A.P., Savastano Jr. H., John V.M., Agopyan V., The effect of different mineral additions and synthetic fiber contents on properties of cement based composites. Cement and Concrete Composites, 28, 2006, 555-563 20. Kim D., Najman A.E., El-Tawil S., Comparative flexural behavior of four fiber reinforced cementitious composites. Cement and Concrete Composites, 30, 2008, 917-928

28-How to get.qxd

9/19/2009

12:00 PM

Page 255

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

HOW TO GET RELIABLE 3D INFORMATION ON CONCRETE POROSITY? Piet STROEVEN Faculty of Civil Engineering and Geosciences, Delft University of Technology Stevinweg 1, 2628 CN Delft, the Netherlands; e-mail: [email protected]

ABSTRACT Pores are an inherent constituent of concretes. Particularly the part of porosity that connects external surfaces of concrete elements is of research interest because harmful substances will be allowed traveling from one side to the other, or to the main reinforcement. Size distribution of pores (SDP) and the related critical pore size are generally considered important parameters. Further, of course, the process of pore de-percolation during hardening, and the de-percolation threshold could be considered relevant phenomena. Conventional methods for assessment of SDP like MIP, Wood’s metal intrusion porosimetry, traditional 2D image analysis and 3D reconstruction of serially sectioned model concretes are evaluated against the method of subjecting single sections to the mathematical morphology “opening” operator. An alternative is application of star volume measurements in random points on pores displayed in single sections. These techniques are superior in economy (restricting labor intensity dramatically). Moreover, they yield reliable 3D information on the size distribution of pores.

Keywords Concrete, pore size distribution, opening operator, intrusion porosimetry, image analysis

INTRODUCTION Porosity and pore structure are of paramount importance with respect to the mechanical and durability properties of cementitious materials. The micro-structural development of cementitious materials and the relationship between structure and material properties have been extensively studied by experimental techniques and computer modelling approaches. However, accurate quantitative characterization of pore structure remains a challenge due to the complex and interconnected nature of the pore network in cement pastes and concretes. The reliability of most experimental techniques is limited since the interpretation of experimental data is based on assumptions of pore geometry that are largely deviating from reality. Numerical modelling of cement paste and concrete, when based on non-realistic simulation of particle packing, cannot yield correct simulation results either. This situation is pertinent to all commonly used systems that are making use of random generators. This inevitably leads to considerable biases in the spatial dispersion of particles at densities relevant for cement and concrete. Hence, the relevance of this study is derived from the impact a fundamental, methodologically sound approach to porosity analysis of concrete will have on durability studies with obvious economic interest. This aspect has received major interest in an earlier BMC paper, to which the interested reader is referred [1]. This contribution will therefore briefly concentrate on a modern, readily available and more reliable alternative for experimental approaches provided by the so-called opening operation method, and applies the technique to section images of concrete specimens for the

28-How to get.qxd

256

9/19/2009

12:00 PM

Page 256

Piet STROEVEN

purpose of illustrating proper determination of SDP. Mathematical morphology techniques are based on measuring the changes produced when a binary image (reflecting the phase of interest) is transformed. The simplest mathematical morphology transformations are erosion, dilation, opening and closing. For a detailed description, see Serra [2]. In this paper, the opening distribution technique is utilised to characterise porosity and assess the SDP in two concrete specimens. In one, water is employed containing 3% NaCl, denoted as N sample, whereas demineralised water is used for the second, D, sample. The reliability of this method is discussed with the help of mutually comparing outcomes obtained further by conventional MIP, Wood’s Metal intrusion porosimetry (WMIP), area histogram method, and by serial section simulation approach. Finally, characterisation of porosity connectivity in concrete is briefly outlined on the basis of a simple example. This comparison study aims at indicating the proper approach to characterising pore size distribution, and to providing reliable information on material structure for models to predict permeability of cementitious materials. Insight into durability aspects of cementitious materials will be promoted by following the indicated procedure.

CONVENTIONAL APPROACHES TO PORE SIZE DISTRIBUTION Mercury intrusion porosimetry (MIP) It has been customary for quite some years to evaluate pore size distribution in cementitious materials by using MIP. Based on the mercury intrusion data, the well-known Washburn equation [3,4] is thereupon applied to estimate the diameter of supposedly cylindrical pores intruded at each pressuring step. The Washburn model invokes two distinct assumptions as to the pores, (1) they are cylindrical, and (2) they are accessible from the outer surface of the specimen. However, it has become increasingly apparent that the intrinsic pores in hydrated cement systems fail to conform to the requirements of the model [5]; pores are clearly not cylindrical, instead, most of their boundaries are obviously convoluted. Wang and Diamond [6] have been evaluating pore shapes in cement pastes by quantitative image analysis (QIA). Very irregular and elongated shapes were determined by means of standard form factors. Furthermore, they found pore profiles to exhibit appreciably fractal nature by applying a standard progressive dilation technique. These findings demonstrate convincingly that the Washburn equation based on MIP data departs enormously from reality (as revealed by QIA). Most pores in this approach are two orders of magnitude smaller than they actually are. However, these effects appear even less important than due to violations of the accessibility condition (the so-called ink-bottle effect). The Washburn equation renders possible calculating the pressure required to fill a cylindrical pore of a given cross-sectional diameter with mercury, provided the pore is directly accessible to the mercury. This issue of accessibility to mercury appears to be at the root of the failure of MIP to provide realistic pore size distributions. The accessibility issue does not arise because individual pores may have ink-bottle shapes, as occasionally stated in the literature [7], but is far more fundamental than that. It arises from but a small proportion of the pores in hydrated cement specimens being directly connected to the exterior surface of the specimen that is subjected to the pressurized mercury. Nearly all of the pores are in the interior of the specimen, and most of them can be reached by mercury only through a long percolation chain of intermediate pores of varying sizes and shapes. This problem is explicitly explained and illustrated by Diamond [3]. Diamond [3] evaluated the deficiencies of MIP technique by comparing the SDP results obtained by MIP and by image analysis on 28 days old paste with w/c ratio of 0.4. The author reported that almost all of the pores obtained by image analysis were smaller than the MIP

28-How to get.qxd

9/19/2009

12:00 PM

Page 257

257

How to get reliable 3D information on concrete porosity?

threshold diameter of the cement paste. Fig. 1a [3] presents a comparison between the two techniques for a cement paste containing air voids (deliberately designed for the experiment). It is clear that MIP data includes the air void volume as part of the measured porosity, but disguises the air void space by not intruding the air voids until the threshold diameter pressure is reached. Contrary, image analysis technique, as a direct observation of the pore space, shows the air void space at appropriate diameters and therefore, reveals a more realistic picture of the actual pore size distribution in cementitious materials. The aforementioned deficiencies of MIP technique have been extensively discussed and at least accepted by part of the cement researchers; among them, QIA of specimen sections is drawing rising interest as a more reliable alternative for pore structure analysis in cementitious materials.

(a)

(b)

Fig. 1. (a) Comparison of MIP and QIA pore size distribution plots for 28 days old cement paste (w/c = 0.4) containing air voids, after Diamond [1]; (b) Comparison of pore size distributions results obtained by WMIP and MIP, respectively, for the aforementioned mortar but at the age of 14 days (w/c = 0.4), after Willis, et al. [7]

Image-based characterization by Wood’s metal intrusion (WMIP) Due to the aforementioned limitations of the MIP technique, some researchers replaced Wood’s metal for mercury as the intruding liquid. The molten Wood’s metal solidifies within the pore structure of the samples. This additionally allows for application of scanning electronic microscopy (SEM) and QIA to specimen sections. Scrivener and Nemati [8] employed Wood’s metal for visualising percolation of pore space in the interfacial transition zone (ITZ) in concretes, and found direct evidence of disproportionately higher pore connectivity, which should have direct impact on transport properties. Recent computer simulation studies by SPACE reveal a similar phenomenon [9,10]. Willis et al. [7] investigated pore structure in an experimental approach also employing WMIP. It was found that molten Wood's metal has a contact angle on hydrated cement that is similar to that of mercury, and that its pressure vs. volume intrusion curve into a mortar was essentially identical to that of mercury under the same conditions. However, with Wood's metal, the pressure applied can be released at any given level, so that after cooling down the location of the solidified Wood's metal within the specimen can be determined. The pore size distribution

28-How to get.qxd

258

9/19/2009

12:00 PM

Page 258

Piet STROEVEN

in the cement mortar was assessed thereupon by QIA. Fig. 1 provides results obtained by application of conventional MIP and of QIA to sections of the Wood’s metal intruded mortar specimens. A practical observation, to which we will refer later, is that the intrinsic non-air void pores, revealed by the solidified Wood’s metal, are mostly in the size range between 1 and 10 Pm. Obviously, MIP data present incorrect information on pore size distribution, since pore size is derived on the basis of the diameter of the access throat through which the mercury has to penetrate the material to reach internal pores. As a result of this so-called inkbottle effect, the MIP pore size distribution curve is shifted towards smaller size range by two or three orders of magnitude. The SDP information obtained by QIA and presented in Fig. 1 is based on the aforementioned area histogram approach. Although far more realistic SDP information is obtained by WMIP technique than by the conventional MIP, the area histogram method is still restricted by the 2D character of pore features in sections. This straightforward approach by area classification belongs to individual granulometric analysis [2]. The quality of SDP characterization can be improved, nevertheless, by the opening distribution technique. Three-dimensional (3D) pore size distribution in model cement paste Ye [11] developed a software package to directly derive three-dimensional (3D) information on SDP in model material generated by the random generator-based HYMOSTRUC computer simulation system. This system is based on a so-called sequential random addition (SRA) algorithm that seriously limits the scope of application possibilities as will be discussed later. The pore structure in simulated cement paste is interpreted as the free space between expanding cement particles during the hydration process. The volume fraction of porosity accessible to imaginary testing spheres of radius r, denoted as U(r), can be computed by gradually filling testing spheres into pore space and densely packing of these spheres. As a fractality issue, it can be expected that U(r) is a monotonously decreasing function of r. It should be noted that the assessed pore volume at a certain value of r constitutes a lower bound estimate. However, when the minimum radius and the interval between radii of testing spheres are small enough at the specified resolution, this method can provide reasonable results [11]. Unfortunately, the calculation results also depend on the starting point of the testing process, an effect of which the impact is larger the larger the radius of the test spheres. Only a randomly selected starting point and averaging over different trials could solve this problem. This would make the approach even more time-consuming, however. The cumulative pore size distribution curve is obtained by directly plotting U (r ) versus r. The derivative of the cumulative curve yields the pore size distribution curve. For further details, see Ye et al. [11]. For model cement pastes with w/c ratios of 0.6 and 0.3, Ye’s data revealed critical pore sizes of about 3 Pm and 1.5 Pm, respectively, at 4 days of hydration (corresponding to a degree of hydration of 0.64). This very time-consuming approach provides a rather straightforward 3D measurement of volume-based pore size distribution. It should be emphasized that for continuous phases consisting of a complex and convoluted system of tunnels and cavities, as in the case of pore space in cementitious materials, conventional measurements of volume and surface area are of little practical use in describing ‘size’ [2]. An available and feasible definition of size for this type of phase including pores in cementitious materials - is the so-called star volume, i.e., the volume of a (3-D!) sphere with its center at a randomly generated point x, situated on a 2-D pore feature, and a radius corresponding to the average value of a series of unobstructed distances in radial directions (creating a “star”) to the pore perimeter [2]. Fig. 2(a) presents part of a schematized field image. The light grey area, Yx, is the pore area that can be observed in an unobstructed way from point x. In a properly designed experiment, a series of such fields is taken from vertical uniformly random (VUR) sections for the practical case in which the 3-D pore structure can be assumed displaying an axis of symmetry. The most general (unpractical) case

28-How to get.qxd

9/19/2009

12:00 PM

Page 259

259

How to get reliable 3D information on concrete porosity?

would require sampling isotropic uniform random (IUR) sections [12]. Such sampling designs allow deriving 3-D estimates from the 2-D field images. When x moves over the connected pore area in Fig. 2(a), the zone of direct vision (Yx) will fluctuate. The star v* at point x can yield a local unbiased estimate for 3-D pore size on the basis of the linear intercept lengths li for a set of random directions (as shown in Fig. 2b). Porosity is obtained by averaging star volume for a random set of points. In addition, the histogram of measured star values represents the actual SDP in 3D. Obviously, this method allows deriving pore size distribution in a far more efficient and economic way than in the former approach.

li

Yx .x

x v*

(a)

S 3

li3

(b)

Fig. 2. (a) Part of schematised field image in which the connected pore area (Yx) that can be observed from random point x is indicated in light gray (b) the star in x measures in random directions the intercept lengths, li. Their average value is used to define the star volume v*.

Opening distribution technique Opening distribution is a type of granulometric analysis by measuring changes in the amount of phase of interest after an opening operation. Clustering of objects in space can be studied by volume fraction measurements (of the phase of interest) after applying a dilation operation with structural elements of increasing size [13]. In a similar approach, pore size distribution in cement paste can be determined by means of area fraction measurements after an opening operation is accomplished on pore space with structural elements of increasing size [14,15]. This is defined as granulometric distribution functions in mathematical morphology. Two main categories of the so-called granulometric distribution methods exist: measurement by individual analysis or by morphological opening. Individual granulometric analysis is well known: the size of each pore feature in the binary image is measured. Then they are classified according to their size to yield the size distribution function. In contrast to individual granulometric analysis, the morphological opening distribution is a more suitable approach to analyzing a complex, interconnected structural phase like pores, since it can be used independently on the nature of the phase of interest (i.e., pore space). On the basis of mathematical morphology transformations, various distributions reflecting pore phase can be obtained by using a sequence of similarly shaped structuring elements (a square in this study) of increasing size. The opening distribution curve of pore space is obtained by plotting the pore area fraction after an opening operation (as illustrated in Fig. 3) versus the linear dimension of structuring element. This gives a type of size classification in the case of an interconnected structure, in contrast to single features as in

28-How to get.qxd

260

9/19/2009

12:00 PM

Page 260

Piet STROEVEN

the case of individual granulometric analysis (e.g., area histogram). This is a very economic way to get consistent and reliable information on SDP.

Fig. 3. Illustration of an opening operation, i.e., an erosion operation followed by a dilation operation with the same structuring element. Please note the structural changes in the circled regions after application of the opening operation.

Determination of critical pore size The definition of critical pore size (denoted as lc) can be found in the literature [4]. It is conceived as the diameter of the pore that completes the first interconnected pore pathway in a network developed by a procedure of sequentially adding pores of diminishing size to this network. It is generally accepted that the smaller the critical pore size, the finer the pore structure. The critical pore size is a unique transport length scale of major significance for permeability properties. Katz and Thompson [4] supported the conjecture that mercury injection is of percolation geometry and considered the characteristic length lc as corresponding to the percolation threshold. So, it can be experimentally assessed by MIP. A typical mercury injection curve (as shown in [4]), reflecting the volume percentage of mercury intruded as a function of the applied pressure, is S-shaped; after a gradual increase under rising pressure, a rapid rise in volume percentage of mercury intrusion occurs at a certain pressure, whereupon gradually a plateau value is reached. This stage of rapid rise is interpreted as the moment when the intruded mercury has formed the first connected cluster spanning the sample. The inflection point of the rapidly rising portion of the curve can be seen as the percolation threshold. Similarly, the critical pore size lc can be associated with the inflection point of the opening distribution curve, and with the peak of the derivative curve of the opening distribution curve. The optical magnification required for visualization of the material microstructure at the level of the researcher’s interest defines both the micro-structural level displayed by the field,

28-How to get.qxd

9/19/2009

12:00 PM

Page 261

261

How to get reliable 3D information on concrete porosity?

and the geometric structure of the representative volume element (RVE). Hence, optical resolution and observed level of the microstructure are intimately connected: higher densities of geometric parameters will be observed under higher optical resolution. It is necessary to strike a balance between a representative area element revealing sufficiently large pore section, and a satisfactory resolution for detection of small capillary pores. The influence of resolution on the accuracy of critical pore size determined by the opening distribution was studied in [15]. It was found that it is reasonable to study cement permeability with image analysis techniques, as long as a satisfactory resolution of about 0.20 Pm per pixel is ensured.

RESULTS AND DISCUSSION

Derivative of opening distribution

SDP obtained by opening distribution technique Porosity p is fairly straightforwardly obtained as the average pore area fraction measured on 2D sections [16]. Fig. 4a presents the opening distribution curve for the tested bulk concretes (two specimens in each group) and the averaged cumulative pore size distribution curved for each group of concretes, D and N, respectively. Total porosity equals the value of the opening distribution curve at zero size of structuring element, leading to 17.5% (D specimen) and 8.4% (N specimen), respectively. Since chloride ions accelerate the hydration process, a lower porosity in the N specimen can be expected. The critical pore size lc is determined as shown in Fig. 4b. The value of lc is 0.79 Pm in the N specimen, slightly lower than the 0.95 Pm in the D specimen. The critical pore size will decrease with hydration time. Further, the higher w/c ratio, the larger will be critical pore size. It is not possible to convert the opening distribution directly into a 3D representation of pore size distribution. Yet, comparison with 3D calculations (see [15]) revealed good correspondence between the curve patterns and size ranges. This demonstrates that the SDP from opening distribution offers realistic 3D structural information of pore space in cementitious materials.

30

D

25

N

20 15 10 5 0 0.1

(a)

1 Size of structuring element ( Pm)

10

(b)

Fig. 4. (a) Pore size distribution (opening distribution) curves for concrete with w/c ratio of 0.6 at 14 days of hydration; (b) determination of critical pore size from derivatives of pore size distribution curves; peak point corresponds to the inflection point on the pore size distribution curve.

28-How to get.qxd

9/19/2009

12:00 PM

Page 262

262

Piet STROEVEN

Figure 5 provides a comparison between pore size measurements by opening distribution and by the conventional area histogram. Since a square is used as structuring element in the opening distribution technique, pore size (x-axis) in Fig. 5 is converted by l A. Obviously, shape and size range covered by the SDP curve due to opening distribution resembles WMIP results to a better extent than defined by the area histogram. This confirms that the conventional area histogram is not suitable for characterizing size distribution of a highly interconnected structure like pore space. Instead, a mathematical morphology technique like opening distribution provides far more realistic structural information.

40 opening distribution area histogram

Area frac tion of poros ity (%

30

20

10

0 0,1

1

Pore size (m )

10

100

Fig. 5. Pore size distribution (SDP) curves obtained by opening distribution technique and by area histogram, for a cement paste with w/c ratio of 0.5 at 3 days hydration.

Porosity connectivity This subject was extensively covered in an earlier contribution to BMC [1]. Section images of cement paste and concrete specimens cannot provide direct information on pore connectivity. However, structural evolution in the cementitious materials can be considered a geometricalstatistical (=stereological) process. The solid phase (cement particles plus hydration products) gradually develops into a connected network, which contributes significantly to the material strength. In contrast, the pore space changes to a de-percolated structure in the matured material as a result of the continuous decline in porosity and in pore connectivity during the hydration process. Spacing parameters defined in stereological theory [16], characterize spatial dispersion of the solid phase and the pores, providing insight into the connectivity of

28-How to get.qxd

9/19/2009

12:00 PM

Page 263

How to get reliable 3D information on concrete porosity?

263

pore space in matured cementitious materials. From this point of view, the de-percolation threshold of porosity for model cement pastes with different water cement ratios and cement finesses levels was discussed on the basis of mean free spacing between elements of the solid phase [17]. The involved model cements were generated with the SPACE system [18], which has been demonstrated capable of realistically simulating particle packing (mutual spacing, configuration of grains) and cement hydration in concrete composites, since based on concurrent algorithms. The pore structure in properly (by SPACE) simulated matured cement paste was shown easier to de-percolate (corresponding to lower pore connectivity) although total porosity was slightly higher than in RSA-based model cements at the same hydration age. This confirms that correct information on pore topology demands realistic modelling of particle configuration in the fresh state that is underlying the hydration simulation.

SUMMARY AND CONCLUSIONS

This contribution presents a critical evaluation of different methods applicable for characterising pore size distribution of cementitious materials, among which are the conventional experimental techniques as mercury intrusion porosimetry (MIP), and a modified experimental approach with Wood’s Metal (WMIP), the conventional area histogram of pore features displayed in section images, and the opening distribution technique. It is argued that quantitative image analysis, when combined with mathematical morphology operations (specifically, opening), will provide relatively realistic structural information of the pore space in cementitious materials. Critical pore size can be derived from the opening distribution curve. Values of critical pore size (also referred to as threshold diameter in literature) and of porosity are useful indexes of pore structure for mutually comparing between cement pastes or mortars. Total porosity varies with w/c ratio and hydration time in a similar way as the critical pore size. The critical pore size seems to be a valid comparative parameter, related to permeability and ion diffusion in cement systems. The proper characterisation of pore structure is expected to provide important information for modelling permeability of cementitious materials; hence, it is highly relevant to studies of durability aspects of the materials. The stereological spacing parameter (mean free spacing) can be used to characterise the structural evolution and to estimate pore connectivity in cementitious materials during the hydration process. REFERENCES

1.

2. 3.

4. 5.

Stroeven, P., Chen, H., Stroeven, M. On connectivity of porosity in model cement paste. In: Proc. Int. Symp. “Brittle Matrix Composites 8”, A.M. Brandt, V.C. Li and I.H.Marshall eds., Cambridge and Warsaw, Woodhead Publ. Co. Ltd., 2006, 25-34 Serra, J., Image Analysis and Mathematical Morphology. London, Academic Press, 1882 Diamond, S. Mercury porosimetry; an inappropriate method for the measurement of pore size distributions in cement-based materials. Cem. Concr. Res., 30 (10) 2000,1517-1525 Katz, A.J., Thompson A.H. Quantitative prediction of permeability in porous rock. Phys. Rev. B, 34 (11) 1986, 8179-8181 Lange, D.A., Jennings H.M., Shah S.P. Image analysis techniques for charac-terization of pore structure of cement-based materials. Cem. Concr. Res., 24 (5) 1994, 841-853

28-How to get.qxd

264

6.

7. 8. 9. 10. 11.

12.

13. 14.

15. 16. 17. 18.

9/19/2009

12:00 PM

Page 264

Piet STROEVEN

Wang, Y., Diamond S. An approach to quantitative image analysis for cement pastes. In: Proc. Int. Symp. “Microstructure of Cement Based Systems/Bonding and Interfaces in Cementitious Materials”, S. Diamond, S. Mindess, F.P. Glasser, L.W. Roberts, J.P. Skalny, L.D. Wakeley eds., Pittsburgh, Mat. Res. Soc., 1995, 23-32 Willis, K.L., Abell A.B., Lange D.A. Image-based characterization of cement pore structure using Wood’s metal intrusion. Cem. Concr. Res., 28 (12), 1998, 1695-1705 Scrivener, K.L., Nemati K.M. The percolation of pore space in the cement paste/ aggregate interfacial zone of concrete. Cem. Concr. Res., 26 (1), 1996, 35-40 Hu, J., Stroeven, P., Properties of the interfacial transition zone in model concrete. Interface Sci., 12, 2004, 389-397 Chen, H., Stroeven, P., Ye, G., Stroeven, M., Influence of boundary conditions on pore percolation in model cement paste. Key Eng. Mater. 302-303, 2006, 486-492 Ye, G., Hu, J., van Breugel, K., Stroeven, P., Characterization of the development of microstructure and porosity of cement-based materials by numerical simulation and ESEM image analysis. Mat. Struct., 35 (4) 2002, 603-613 Gundersen, H.J.G., Jensen, E.B., Stereological estimation of the volume-weighted mean volume of arbitrary particles observed on random sections. J. Microsc., 138, 1985, 127142 Jeulin, D., Random texture models for material structures. Statist. and Comp., 10, 2000, 121-132 Scrivener, K.L., The use of backscattered electron microscopy and image analysis to study the porosity of cement paste. In: Proc. Int. Symp. “Pore Structure and Permeability of Cementitious materials”, L.R. Roberts, J.P. Skalny eds., Pittsburgh, Mat. Res. Soc. Symposium Proc., 137, 1989, 129-140 Hu, J., Stroeven, P., Application of image analysis to assessing critical pore size for permeability prediction on cement paste. Im. Anal. & Stereol., 22 (2) 2003, 97-103 Underwood, E.E., Quantitative Stereology. New Jersey, Addison-Wesley Publ. Co., 1968 Hu, J., Stroeven, P., De-percolation threshold of porosity in model cement; approach by morphological evolution during hydration. Cem. Concr. Comp., 27 (1) 2005, 19-25 Stroeven, M., Discrete numerical modelling of composite materials. PhD Thesis, Delft University of Technology, Delft, Meinema, 1999.

29-Analysis.qxd

9/19/2009

12:00 PM

Page 265

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

ANALYSIS OF THE INFLUENCE OF TYPE, AMOUNT AND WAY OF INTRODUCTION OF ANTI-FOAMING ADMIXTURE (AFA) ON THE PROPERTIES OF SELF-COMPACTING CONCRETE MIX Beata àAħNIEWSKA-PIEKARCZYK Department of Building Processes, Faculty of Civil Engineering, Silesian University of Technology, Akademicka 5, 44-100 Gliwice, Poland, e-mail: [email protected] ABSTRACT The properties of the self-compacting concrete mix depend on an automatic introduction of air bubbles caught during the process of mixing. What is interesting, the criterion for self-compactibility is not taken into consideration in commonly used self-compacting tests. On the basis of different tests concerning selfcompacting concrete mixes, it has been found out that too high air content in their volume was the result of superplasticizer, in spite of meeting the self-compactibility criteria. For the decrease of too high air volume in SCC, the use of anti-foaming admixture (AFA) is proposed. As a result, the effect of AFA mix flow diameter is increased and the flow time is decreased. Moreover, the workability loss is lower. In case of mix incorporating AFA, their high flowability does not cause segregation of the mix, what is possible in case of SCC incorporating only superplasticizer. However, the time of the introduction of AFA and its type is essential to get higher flowability degree, but is not important to achieve low air volume in SCC. Keywords

superplasticizer, anti-foaming admixture (AFA), air-volume, self-compacting concrete, rheological properties INTRODUCTION The characteristic of self-compacting concrete mix is an effective elimination of air bubbles caught during the process of mixing. The condition of the self-compacting of mix depends on the size of the rheological parameters: yield stress and plastic viscosity of cement paste [1, 2]. Because the availability of the direct measurement of rheological properties is limited, technological tests are used in building practice, assessing the self-compactibility of the concrete mix (SCC), such as: flow test (Table 1 and 2). The value of SCC flow diameter depends on the mix yield stress W0m, whereas SCC time flow depends on its plastic viscosity Kpl. The diameter and time flow of SCC should correspond with the classes presented in Tables 1 and 2. In the European guidelines for self-compacting concrete [3], detailed outlines in respect of SCC classes and other technical tests of the self-compacting concrete mix depending on its purpose are given.

29-Analysis.qxd

9/19/2009

12:00 PM

Page 266

Beata £AZ´ NIEWSKA-PIEKARCZYK

266

Table 2. Viscosity classes [3]

Table 1. Slump-Flow classes [3] Class SF1 SF2 SF3

[mm] from 550 to 650 from 660 to 750 from 760 to 850

[s]

Class

T500 2 >2

VS1/VF1 VS2/VF2

V-funnel 8 from 9 to 25

On the basis of the results of various tests it was stated that in numerous cases a problem of excessive air-entrainment of concrete mix appears, [4, 5], despite the fact that fresh mix has achieved recommended flow in suitable time according to [3] (Table 1 and 2). Tests of the porosity characteristics of the concrete proved that the excessive air-entrainment of the mix influences the air-entrainment of concrete (during the process of concrete hardening, formed pores are not filled with the hydration products, because C-S-H gel may form only in water), [4]. Also, the tests results [6] (presented in Table 3), [4] and [7], prove that new generations of superplasticizers (SP) show air-entrainment effect. It should be emphasized that according to standard requirements concerning chemical additives of the concrete, superplasticizers should not cause air formation in the mix higher than 2%. Table 3. The influence of the superplasticizer on the mix air-entrainment [6]

SP type

Aircontent

Sulfonated LignoNaphtalene sulfian, Formaldehyde LS Condensate, SNF ++

+

Sulfonated Melanine Formaldehyde Condensate, SMF 0

New Generation Superplasticizers Amino PolyCarboxylate Phosphonate Polyoxyethylene, Polyoxyethylene, PCP AAP ++

++

The reason for the PCP air-entrained superplasticizer effect is its influence on the decrease of the surface tension of the liquid phase in the paste, as it was proved by other tests, [8]. The air content developed with the superplasticizer effect increases with the increase of w/s ratio (Fig. 1 and 2). With the increase of the liquid phase part in mix, the airentrainment effect is higher, similarly as in case of the air-entraining admixture. On the basis of tests results shown in [9], we can conclude the influence of etheric and poly-carboxyl superplasticizer the SCC porosity structure. The test results of A÷D fresh mix properties are presented in Table 4, whereas in table 5 test results of the porosity characteristics of the hardened concrete are shown. Table 4. Test results of the rheological properties of concrete mix with the use of Abrams cone, [9] Symbol A B C D

Binder type CEM I 42,5, fly ash

SP based on: polycarboxylic ether (PCE) polycarboxylate (PCP)

w/b 0.34 0.45 0.34 0.45

T500, [s] 5.0 4.6 4.9 4.1

Slump-Flow, [mm] 680 660 690 710

Tests results presented in Table 5 prove that superplasticizers on the basis of polycarboxyl ether cause considerable SCC air-entrainment. The air-entrainment of concrete is

29-Analysis.qxd

9/19/2009

12:00 PM

Page 267

Analysis of the influence of the type, amount and way of introduction of anti-foaming admixture (AFA)...

267

higher in case of higher w/c value and amounts to even 8.30%. It should be emphasized that mix achieved relatively high flow, that is 660 mm (which corresponds with class SF2, Table 1). Despite this fact, no suitable self-compactibility took place (self-venting). The airentrainment of fresh mix was probably higher than 8.3%, because the air content marked in hardened concrete is approximately by 1%÷2% lower than that found in the mix. The reason for the excessive air-entrainment of the fresh mix is probably the influence of the mechanism of working and structure of superplasticizer on the formation and behavior of air bubbles in its volume. In the case of poly-carboxyl superplasticizer the air content was lower and amounted to 4.45% with w/c = 0.45. Test results concerning the effects of poly-carboxyl superplasticizer comply with other test results published in [4]. The conclusions from the analysis of the test results presented in Table 5 also confirm test results of mercurial porosimeter. And in this case, the superplasticizer on the basis of poly-carboxyl ether causes higher SCC air-entrainment. An excessive air content appeared although the fresh mix achieved even 710 mm flow. A well-founded question is formed concerning the effectiveness of commonly used tests aiming to qualify the mix as self-compacting, [10, 11], in case when a superplasticizer shows the air-entraining effect. Table 5. The set of test results of air pores structure in concretes, [9] Porosity structure parameter air-content, A [%] content of micropores below 0.3 mm, A300 [%] spacing factor, CL [mm] specific surface, D [mm-1]

A 6.7 1.50 0.26 17

Series B C 8.30 2.90 2.96 0.70 0.11 0.33 36 21

D 4.45 1.74 0.13 45

NEGATIVE EFFECTS OF THE EXCESSIVE AIR-ENTRAINMENT IN CASE OF SELF-COMPACTING MIX AND CONCRETE The air-entrainment of the fresh mix may decrease its flow depending on the degree of the initial fluidity, as the result of internal compression of the air bubbles and lower density of the fresh mix. The air-entrainment may also initially increase the flow when the fresh mix characterizes with originally low fluidity, [12, 13]. However, other amounts of the airentrained admixture cause the decrease of the diameter of flow of fresh mix. The sizes of pores formed during the effect of superplasticizer in hardened concrete characterize with too big sizes (Fig. 1 and 2). They are the reason of the decrease of concrete mechanical parameters and are not beneficial from its frost resistance and absorptivity point of view [4]. In order to protect concrete against the effects of cyclic freezing and thawing it is beneficial when the bubbles are characterized with diameters of 0.05÷0.10 mm and are in the paste volume in the range of 0.15÷0.20 mm from each other, [14]. Although the problem of the critical value of the pores range in frost resistant concrete, depending on its type, is still a considered issue [2]. Analyzing the results of other tests shown in [4], it is proved that 4% of the airentrainment (being the result of a superplasticizer) cause the resistance decreases down to 24% when SCC concrete with ratio w/c = 0.4 with zero air volume is considered.

29-Analysis.qxd

268

9/19/2009

12:00 PM

Page 268

Beata £AZ´ NIEWSKA-PIEKARCZYK

Figure 1. The porosity characteristics of non air-entrained SCC (CEM I 32.5 R + 10% silica fume; w/b = 0.41), [4]

Figure 2. The porosity characteristics of non air-entrained SCC (CEM II B-S 32.5 R B-S; w/b = 0.29), [4] Considering these above mentioned tests results it may stated that certain superplasticizers of new generation cause excessive air-entrainment which remains in the selfcompacting volume of the fresh mix and concrete, causing deterioration of their properties, although the mix meets commonly accepted criteria of technical tests (Table 1 and 2). Desired suitable fluidity of the fresh mix, essential for its efficient self-compacting, is not included in any commonly used technical tests. Commonly accepted criteria for such tests are insufficient in this scope and do not guarantee effective self-compacting. It can be obtained by increasing the fluidity of the fresh mix with the superplasticizer, however it may cause its segregation. Due to this fact, in order to prevent the presence of the excessive air-entrainment, the superplasticizers should not only be compatible with cement, but also do not create airentraining effect in the paste. In order to counteract the excessive air-entrainment anti-foaming admixtures (AFA) may be used against the formation of air bubbles. Such admixtures are not commonly used in building practice. Hence, the mechanism of their functioning is not well-known, as well as their effectiveness in decreasing the air content in fresh mix and its influence on fresh mix properties and on properties of hardened concrete. So, it is advisable to carry on proper tests aiming at verification of the influence of anti-foaming admixtures on air-entrainment, rheological properties and stability of self-compacting concrete mix, also depending on the time of their introduction.

29-Analysis.qxd

9/19/2009

12:00 PM

Page 269

Analysis of the influence of the type, amount and way of introduction of anti-foaming admixture (AFA)...

269

ANTI-FOAMING ADMIXTURES Components and their proportions used in the anti-foaming admixtures, as in superplasticizer, are known only to their producer. They may be: mineral oils, silicone oils, organic modified silicones, hydrophobic constant molecules (silica, waxes, higher fatty acids soaps, alcohols and fatty acids), emulsifiers, polyalcohol, alcohol derivatives of organic compounds. They may be mixes of above active components acting in a synergetic way. Unfortunately, high price and not widely tested influence on properties of the fresh mix and concrete do not favour wider use of the anti-foaming admixtures. To analyze the influence of properties of those admixtures on properties of the fresh mix and concrete, first their mechanism of functioning should be known. Research results (Fig. 3) show that the effectiveness in decreasing air content in cement paste of the antifoaming admixtures does not consists in increasing the value of surface tension of its liquid phase. Surface tension of water solution of anti-foaming admixture, superplasticizer and antifoaming admixture characterize with even smaller value than the surface tension of the water solution of air-entraining superplasticizer (Fig. 3).

Figure 3. Influence of type admixture on surface tension of water; SP based on polycarboxylate (PCP); and AFA based on polyalcohol The mechanism of functioning of anti-foaming admixture may be explained in the following way. The active components are distributed around gas bubbles, displacing surfactant molecules. In result, the thickness of the lamella wall built from surfactant causes its destabilization and results in the fracture or coalescence of the bubble (Fig. 4).

Figure 4. Stages of the mechanism of action of an anti-foaming admixture (AFA) To identify the wider unknown influence of the anti-foaming admixtures on airentrainment of the self-compacting mix and its properties, suitable tests were carried out.

29-Analysis.qxd

9/19/2009

12:00 PM

Page 270

Beata £AZ´ NIEWSKA-PIEKARCZYK

270

METHODOLOGY OF THE RESEARCH The mix proportions of the tested self-compacting concretes is showed in Table 6. The process of mixing started with dry components (about 0.5 min). Then water was added where superplasticizer was earlier distributed At the end of the process of mixing, the anti-foaming admixture was added (in case of one series such admixture was introduced after 20 minutes, Table 7), and all ingredients were mixed for another 6 minutes in case of ordinary concrete and 12 minutes in case of high-performance concrete. After 15 minutes, the fresh mix was subjected to another short mixing and then rheological measurements were carried out and the air content in the fresh mix was checked. After filling up the container, the fresh mix was kept for 10 minutes. The air content was defined after proceedings described after EN 12350-7, mix density after EN 12350-6, whereas the flow and its time after ASTMC 143. Table 6. The mix composition of self-compacting mixes

Cement: Series CEM II B-S 3

[kg/m ]

w/b

W/b W= water + liquid chase of admixture

Sand 0-2

Gravelly aggregates [kg/m3] 2-4 [mm]

2-8

SP*

AFA **

8-16 [% m.C.]

0.41 0.00 M1 0.42 2.02 M1-f 541 0.40 890 200 228 256 0.77 M1-f0.42 2.02 t*** * based on polycarboxylate (PCP), ** based on polyalcohol, *** AFA added after 20 min. The effectiveness of anti-foaming admixture depending on its type and the most airentrained superplasticizer (identified on the basis of earlier tests of fresh mix, Table 6 and 7), was checked by the tests of flow and air-entrainment of mortar after EN 10153:2000/A2:2007 and EN 1015-7:2000 respectively. The volume of mortars corresponded to that used in the fresh mix (Table 6). The process of mixing of the mortars started with dry ingredients, and then superplasticizer was added and next anti-foaming admixture of particular type (Table 7). The mixing of components of mortars was carried out after the proceedings accepted for standard mortar after EN 197-1/2002. Mortar Z , because of containing SP, was subjected to densification through shaking. The mortar was put in three layers in the container being the part of the apparatus that tests the air content in its volume. Each layer was shaken before laying another one. After 20 minutes, in case of all tested mortars, the air-entrainment measurements and mortar flow were carried out, because after such time the effectiveness of superplasticizer is the highest. To assess AFA influence on the workability loss of the mortar in relation to time, the assessment of paste flow was checked after 20 and 60 minutes, counting from the time of mixing the remaining components.

29-Analysis.qxd

9/19/2009

12:00 PM

Page 271

Analysis of the influence of the type, amount and way of introduction of anti-foaming admixture (AFA)...

271

Table 7. Mix composition of mortars; cement CEM II B-S: 541 [kg/m3], w/b = 0.40; sand: 890 [kg/m3] SP AFA Series w1/b AFA based on: [%m.C] [%m.C] Z2* 0.77 0.00 0.00 Z 2.21 froth breaker on the PDMS basis/ silicone oil/ Z2-a hydrophobic silica 4.42 Z2-a1 froth breaker on the basis of mineral oil or amidol wax Z2-b froth breaker on the basis of alcohol derivative of 2.21 Z2-c saturated fatty alcohol, mineral oil and PE wax 0.40 Z2-d 0.77 fiakyl derivative of saturated fatty alcohol/mineral 4.42 Z2-d1 oil/PE and amidol wax 1.11 Z2-d2 alkoxyl derivative of fatty alcohol, 100% Z2-e polyalcohol Z2-f 2.21 oxiakyl derivative of saturated fatty alcohol/mineral Z1** oil/PE and amidol wax Z3*** 0.00 0.41 0.56 okiakyl derivative of saturated fatty alcohol/ mineral Z32.21 oil/PE and amidol wax a*** * SP based on based on polycarboxylic ether (PCE), ** SP based on polycarboxylate (PCP), *** SP based on polycarboxylic ether (PCE) (another type than *), 1water + admixtures liquid part RESEARCH RESULTS AND THEIR DISCUSSION Research results presented in Table 8 show the effectiveness of anti-foaming admixture in decreasing the air content in the fresh mix. The air content in case of tested mixes may be decreased even by about 3% and more in case of mixes of low w/b. Additional advantage that was achieved due to the use of anti-foaming admixtures, was high increase of fresh mix flow, and what was essential, in considerably shorter time (Table 8). Moreover, in case of the fresh mix containing AFA, the loss of initial consistency was considerably slower. Despite shorter time and low viscosity of the mix the segregation did not occur. On the contrary, the fresh mix containing anti-foaming admixture was more stable than the mix with similar diameter of flow, but containing higher amount of superplasticizer. The advantage achieved as a result of the usage of anti-foaming admixture due to its fluxing effect was the decrease of the volume of the necessary superplasticizer. In result, similar consistency was obtained but, what should be emphasized, of non air-entrained excessively self-compacting mix. In Table 8 test results on technology of preparation of the fresh mix containing antifoaming admixture are shown. In one case, the admixture was immediately introduced after the superplasticizer, in another one, 20 minutes after starting the process of mixing and introduction of superplasticizer to the mix volume. The test results prove that the time of introduction the anti-foaming admixture is not essential for the effectiveness of this admixture in decreasing the air-entrainment of the mix. This conclusion is very important because the anti-foaming admixtures may be used in order to: - prevent the occurrence of the excessive air-entrainment of the mix (introducing antifoaming admixture with air-entrained superplasticizer),

29-Analysis.qxd

9/19/2009

12:00 PM

Page 272

Beata £AZ´ NIEWSKA-PIEKARCZYK

272

-

decrease already existing air content (caused by air-entrained superplasticizer) in fresh mix. Other tests [15] proved that it was not possible to air-entrain the fresh mix for the second time, if the anti-foaming admixture was introduced. So, anti-foaming admixtures should not be used in a case of mixes that were air-entrained on purpose. The test results presented in Table 8 show that the time of introduction of anti-foaming admixture influences significantly the diameter and flow time of the fresh mix. In order to achieve the highest fluxing of mix, the anti-foaming admixture should be introduced as fast as possible. Table 8. Test results on the properties of mixes Series M1 M1-f M1-f-t

Air-volume Ac, [%] 4.0 2.8 2.8

Flow diameter after 20 min, [mm] 63 70 65

Flow time T after 20 min, [s] 3 1 1

In Table 9 and 10 the test results on mortars are presented. It was noticed, that different type of superplasticizers causes radically different volume of air in mortars (Table 9 and 10). In case of mortars subjected to concentration (Z), 4% of air volume was observed. The mortar, in which air-entraining superplasticizer was used, was characterized with even about 12% air volume. Both mortars were considered as a reference in the assessment of AFA effectiveness. Tests results of Z2-a÷Z2-f mortars prove that due to the reduction of the air-entrainment in mortar and its rheological properties, the most effective ones were characterized with AFA on the basis of polyalcohol. Comparing test results of Z2-d and Z2 mortars it can be concluded that the superplasticizer type is essential because of the AFA effectiveness. Table 9. Test results on the properties of mortars Flow time T Flow Flow Flow time T diameter after after diameter after after 60 min, [mm] 60 min, [s] 20 min, [mm] 20 min, [s] Z3*** 3.0 29.5 5 24.5 7 Z3-a*** 2.2 31.0 4 27.0 6 *** SP based on polycarboxylic ether (PCE) (another type than *). Series

Air-volume Ac, [%]

Test results of Z2-d, Z2-d1 and Z2-d2 mortars prove that the effectiveness of AFA depends on its volume, and when certain point of saturation is exceeded, no further reduction of the airentrainment is possible, but only the improvement of the fluidity of the mortar. The reduction of the air-entrainment with the use of AFA may be higher than in case of the use of mechanic concentration of mortar (Z). The mortar despite considerable fluidity and containing AFA, does not undergo segregation, as it is in case of the mortar with no AFA of similar fluidity level. Moreover, test results of Z3 and Z3-a mortars prove that mortars with AFA maintain initial consistency longer in comparison to mortars with only superplasticizer (Table 10). Other tests [15] prove that conclusions on AFA based the tests on mortars may be successfully accepted in fresh mixes.

29-Analysis.qxd

9/19/2009

12:00 PM

Page 273

Analysis of the influence of the type, amount and way of introduction of anti-foaming admixture (AFA)...

273

Table 10. Test results on the properties of mortars Air-volume Ac, Flow diameter, after 20 min, Flow time T after 20 min, [%] [mm] [s] Z2* 12.0 26.0 2 4.0 11.0 4 Z 3.0 29.9 6 Z2-a 3.0 27.0 6 Z2-a1 5.2 36.3 5 Z2-b 3.8 39.1 4 Z2-c 3.4 37.0 6 Z2-d 3.4 34.0 6 Z2-d1 4.0 32.0 6 Z2-d2 5.4 30.6 6 Z2-e 2.8 36.0 5 Z2-f Z1** 3.0 31.0 6 * SP based on polycarboxylic ether (PCE), ** SP based on polycarboxylate (PCP). Series

CONCLUSIONS In the scope of carried out tests on anti-foaming admixtures, the following conclusions may be proposed: 1) Anti-foaming admixtures may be used in order to prevent excessive air-entrainment of the mix (introducing anti-foaming admixture with air-entrained superplasticizer), or in order to decrease already existing air volume, caused by air-entrained superplasticizer in the fresh mix. 2) As to the results of use of AFA, the increase of the flow of mixture achieved in shorter time (the bigger the smaller w/s is) follows, and without its segregation. Moreover, in the case of fresh mix with AFA, the loss of its initial consistency is slower, as it was proved in the latest tests [15]. So, due to the use of anti-foaming admixtures, it is possible to decrease the need of a superplasticizer in order to achieve suitable flow of the mix. 3) The essential for mentioned above rheological properties modifications developed by the effect of anti-foaming admixtures, is the moment of its introduction to the fresh mix. The best effectiveness is obtained by the introduction of that admixture immediately after superplasticizer. However, the moment of introduction of this admixture is not essential to achieve the low air volume in the fresh mix. 4) Test results of mortars show that because of the reduction of the air-entrainment in mortars and their rheological properties, AFA on the basis of polyalcohol is characterized by the highest effectiveness. The effectiveness of AFA depends on the used amount of superplasticizer, and after exceeding some point of saturation, there is no further reduction, only improvement of the fluidity of the mortar. It should be emphasized, that the mortar, despite considerable fluidity and containing AFA, does not undergo segregation, as it is in the case of mortars with no AFA of similar degree of fluidity (caused by higher amount of superplasticizer). Moreover, test results prove that mortars with AFA maintain initial consistency for longer time in comparison to mortar with only superplasticizer.

29-Analysis.qxd

9/19/2009

12:00 PM

Page 274

Beata £AZ´ NIEWSKA-PIEKARCZYK

274

Because certain producers already use anti-foaming admixtures, the problem of compatibility of AFA depending on the SP type and other admixtures types in their content, and also introduced to fresh mix on purpose, will be the subject of future tests, where the influence of AFA and other chemical admixtures on the properties of hardened selfcompacting concrete will be analyzed. REFERENCES 1. 2.

3.

4.

5.

6. 7.

8.

9. 10.

11.

12.

13. 14. 15.

Szwabowski, J., Rheology of cement based mixes (in Polish), Wydawnictwo Politechniki ĝląskiej. 1999 Gliwice, pp. 239 Szwabowski J., àaĨniewska-Piekarczyk B., The suggested values of parameters of porosity structure of self-compacting concrete (SCC), Cement-Wapno-Beton, 3, 2008, 155-165 European Project Group. 2005, The European guidelines for self-compacting concrete: specification, production and use, Available from Internet: http://www. efnarc. org/ pdf/SCCGuidelinesMay2005.pdf àaĨniewska-Piekarczyk B., The influence of superplasticizer on porosity structure characteristic and durability of SCC (in Polish), Konferencja “Dni Betonu, Tradycja i NowoczesnoĞü”, 13-15 Oct. 2008, Wisáa, 567-576 Szwabowski, J., àaĨniewska, B., Influence of the properties of self-compacting concrete on the effect of air entrainment, The 9th International Conference Modern Building Materials, Structures and Techniques. May 16-18 2007, Vilnius, Lithuania. Selected papers, 182-189 Mosquet, M., The new generation admixtures (in Polish), Budownictwo Technologie Architektura, special edition 2003, 21-23 Ramachandran V.S, Concrete Admixtures Handbook. Properties, Science and Technology, Ed. V.S., Ramachandran, Noyes Publications, Park Ridge, New Jersey, USA 1995, pp. 1154 àaĨniewska-Piekarczyk B., The surface tension of cement paste and its affects to formation air bubbles, 6th International Conference AMCM’2008 Analytical Models and Concepts in Concrete and Masonry Structures, àódĨ 2008, 229 -230 (abstract with paper in CD) GorzelaĔczyk T.: The assessment of the failure of self-compacting concretes by acoustic techniques (in Polish), PhD dissertation, Wrocáaw 2007, pp. 134 Szwabowski J., àaĨniewska-Piekarczyk B., The increase of air content in SCC mixes under the influence of carboxylate superplasticizer, Cement-Wapno-Beton, 4, 2008. 205-215 Sakai E., Kasuga T., Sugiyama T, Asaga K., Daimon M., Influence of superplasticizers on the hydration of cement and the pore structure of hardened cement, Cement and Concrete Research, 36, 2006, 2049–2053 àaĨniewska-Piekarczyk B., The influence of air-entrainment on rheological properties of SCC (in Polish), X Sympozjum Naukowo-techniczne Reologia w Technologii Betonu, Gliwice 2008, 113-124 Neville A.M., Properties of concrete (in Polish), Polski Cement, Kraków 2000, pp. 874 Fagerlund, G., Durability of concrete structures (in Polish), Arkady, Warsaw 1999, pp. 93 àaĨniewska-Piekarczyk B., Recent test results (non published), 2009

30-Predicting.qxd

9/19/2009

12:00 PM

Page 275

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

PREDICTING THE ELASTIC MODULI OF ENHANCED POROSITY (PERVIOUS) CONCRETES USING RECONSTRUCTED 3D MATERIAL STRUCTURES Milani S. SUMANASOORIYA, Omkar DEO, and Narayanan NEITHALATH* Department of Civil and Environmental Engineering Clarkson University, Potsdam, NY 13699, USA *

e-mail: [email protected]

ABSTRACT Enhanced Porosity Concrete (EPC), also known as pervious concrete is a macroporous material that is finding applications in parking areas and low volume pavements because of its ability to transport a large amount of storm water through its porous material structure. Majority of the current research focuses on the functional performance of this material, with scant research on its mechanical behavior. In this study, a computational procedure is implemented on two-dimensional planar images of EPC to reconstruct three-dimensional material structures. The 3D reconstructed digital image data is used as an input to a finite element program to calculate the effective linear elastic properties of the material when subjected to applied macroscopic strains. EPC consists of three phases – the aggregates, the paste surrounding the aggregates, and the pores. In order to reduce the complications associated with assigning each of these phases different elastic moduli, each aggregate (assumed to be spherical in shape) surrounded by a paste shell, is mapped into an effective particle having a uniform elastic modulus, which is then input into the program to calculate the effective elastic modulus of the composite. The paste thicknesses for different EPC mixtures are obtained from an image analysis procedure. Ultrasonic pulse velocity method is used to experimentally determine the elastic modulus of several EPC mixtures proportioned using different aggregate sizes and blends. The results of the predictions using the computational method, and the experimental values are in good agreement.

Keywords Pervious concrete, Enhanced Porosity Concrete (EPC), planar images, three-dimensional structure, digital image data

INTRODUCTION Enhanced Porosity Concrete (EPC) or pervious concrete is a relatively new type of concrete in which gap-graded coarse aggregates and low water-to-cement ratio (w/c) is used to create a network of pores in the material that can sustain water flow through it (ACI 522R-06 [1], Neithalath 2004 [2], Neithalath et al. 2005 [3]). This characteristic has made EPC a sustainable infrastructural material in areas of high rainfall intensity where rain water can seep in through the interconnected pore structure into the underlying gravel base or drainage/storage structures. EPC has several other advantages, including its better acoustic absorption characteristics, ability to enhance the skid resistance, and better thermal absorption characteristics (ACI 522R-06 [1], Neithalath et al. 2006 [4], Yang and Jiang 2003 [5]).

30-Predicting.qxd

276

9/19/2009

12:00 PM

Page 276

Milani S. SUMANASOORIYA, Omkar DEO and Narayanan NEITHALATH

Porosity is the most definitive pore structure feature for most porous materials and EPC is no exception. Porosity, along with the other pore structure features such as pore sizes, pore connectivity, and pore phase distribution dictates the material behavior. The porosity of EPC is typically in the range of 15 to 30 % (ACI 522R-06 [1], Tennis et al. 2004 [6], Neithalath et al. 2006 [4], Marolf et al. 2004 [7]), and the pores are of the order of 2-8 mm in size (Neithalath 2004 [2], Sumanasooriya and Neithalath 2009 [8]). While the primary functional performance of EPC is related to its water transport ability, in order to be successfully used in pavement overlays and low volume pavements, the structural performance of the material needs to be carefully considered. Though the current specifications rarely mention strength (or any structural property) as a performance feature, it is anticipated that, with large scale use of this material and more demanding requirements, these properties will be considered. In a random two-phase composite material such as EPC, with a dominant “zero-property” phase (pore phase), the properties are expected to be highly variable. The influence of pore structure features on the material response is better studied using computational material science based approaches in such cases (Bentz and Martys 1994 [9], Bentz 2005 [10], Garboczi et al. 1999 [11], Bentz and Martys 2007 [12]). The linear elastic properties of cement-based materials have been predicted based on the modeled three-dimensional microstructures (Meille and Garboczi 2001 [13], Garboczi and Berryman 2001 [14]). Recent studies (Bentz et al. 2009 [15], Sumanasooriya et al. 2009 [16]) have used computational permeability prediction of EPC based on two-dimensional images of the material structure. This study uses a similar approach for the elastic modulus predictions of EPC. Pore structure features are extracted from 2D images of EPC sections in order to obtain effective material properties that are used as inputs to the computational procedure that was developed at the National Institute of Standards and Technology (NIST).

EXPERIMENTAL PROGRAM: MIXTURE PROPORTIONS, PORE STRUCTURE FEATURE EXTRACTION AND ELASTIC MODULUS MEASUREMENT Materials and mixture proportions Six different EPC mixtures were proportioned in this study using three different single sized aggregates and a few of their blends. Type I ordinary Portland cement conforming to ASTM C 150 was used as the binder and the water-to-cement ratio (w/c) was kept constant at 0.33 for all the mixtures. The aggregate-cement ratio for all the mixtures was maintained as 5. The single (mono) sized aggregates used were: #8 (passing 4.75 mm sieve and retained on 2.36 mm sieve), #4 (passing 9.5 mm sieve and retained on 4.75 mm sieve), and 3/8” (passing 12.5 mm sieve and retained on 9.5 mm). The remaining three mixtures were made using blends of #4 and 3/8” aggregates, with 75%, 50%, or 25% 3/8” aggregates replacing #4 aggregates by mass. The mixtures were prepared in a laboratory mixer and cast in 75 mm diameter x 150 mm long (3 in x 6 in) cylindrical molds. Porosity and pore size determination using image analysis technique Image analysis techniques have been successfully employed to obtain the porosity and other pore structure features such as effective pore sizes and pore spacing of EPC specimens (Neithalath 2004 [2], Low et al. 2008 [17], Sumanasooriya and Neithalath 2009 [8], Marolf et al. 2004 [7]).

30-Predicting.qxd

9/19/2009

12:00 PM

Page 277

Predicting the elastic moduli of enhanced porosity (pervious) concretes using reconstructed 3D material structures

277

To prepare the specimens for image analysis, 12.5 mm thick slices were removed from the top and bottom of the 75 mm diameter x 150 mm long cylindrical specimens, and the remaining portion of the specimens were divided into three slices of equal thickness. By using the inside surface of the top and bottom 12.5 mm thick slices along with two faces for each of the three sections provided a total of eight faces for the pore structure analysis. The surfaces were ground to remove surface deficiencies and scanned in a flatbed scanner in the gray scale mode. The images were processed and analyzed using Image J ™ software package (freely downloadable from www.rsb.info.nih.gov). The grayscale images were converted into binary images (pore phase in black and solid phase in white) with a threshoding operation, and the noise in the images removed. The pore area fractions of these images were obtained as a fraction of total pore area to the total area of the circular images. Thus, eight different values of the pore area fraction were obtained for each EPC cylindrical specimen, and the average value recorded. According to stereological theory, if random samples are used, the area fraction of pores should be equal to the volume fraction of pores (Hu and Stroeven 2005 [18]), and thus the pore area fraction of images can be considered to represent the volumetric porosity of a particular EPC mixture. Further details of comparison between the measured volumetric porosities and pore area fractions of several EPC mixtures could be found in (Sumanasooriya and Neithalath 2009 [8]). Since square images are required to generate the three dimensional material structure from a planar images, it was decided to use the square images to determine the pore size distribution in EPC specimens. The square images were obtained by inscribing the shape in the original circular image. Figure 1 shows the steps in the image analysis procedure.

Thresholding Removing noise Scanned and cropped image (570 pixels diameter)

Processed image (570 pixels diameter)

Square image for 3D reconstruction (400 x 400 pixels)

80

Frequency, %

60

40

20

0 0

1

2

3

4

5

6

7

8

Pore size, mm

Pore size distribution

Figure 1: Schematic of the image analyis procedure for porosity and por size destribution

30-Predicting.qxd

9/19/2009

12:00 PM

278

Page 278

Milani S. SUMANASOORIYA, Omkar DEO and Narayanan NEITHALATH

The frequency distributions of pore sizes were obtained for the square images. The pores in the two-dimensional images were assumed to be circular in shape and the equivalent diameter of each of the individual pores were obtained. A typical frequency plot is shown in Figure 2. The pore size corresponding to 50 % cumulative frequency is defined as the effective pore size (d50) for a particular EPC mixture. d50 for all images were obtained and the average value reported as the corresponding d50 for a particular EPC mixture. The results will be discussed in a later section.

80

Frequency, %

60 60 40 40 20 20

Cumulative frequency, %

100

80

0

0 0

1

2

3

4

5

6

7

8

Pore size, mm

Figure 2: Frequency distribution of pore sizes

Determination of paste thickness Along with the aggregate phase, the cement paste phase constitutes the solid phase in EPC, the effective property and volume fraction of which is required for elastic modulus prediction. The thickness of cement paste between any two aggregates of an EPC specimen was measured using a digital stereo microscope. The same EPC sections used for porosity determination were used to determine the paste thickness also. The ground and polished EPC sections were placed on the stage plate of the microscope, which was connected to a personal computer. Images of the sections were captured at various locations using a 2.0 Mega Pixel camera attached to the microscope. The magnification used was 10x. Figure 3 shows a typical image used to determine the paste thickness. The cement paste between the aggregates was carefully identified in the images and their thicknesses were measured using a pre-calibrated and user-defined scale bar. From number of sliced sections corresponding to an EPC mixture, images were captured from which several paste thickness values were obtained at different locations. Overall, about 100 paste thickness measurement values were obtained from every specimen.

30-Predicting.qxd

9/19/2009

12:00 PM

Page 279

Predicting the elastic moduli of enhanced porosity (pervious) concretes using reconstructed 3D material structures

279

Aggregate Void

Void

Paste layer Aggregate

Figure 3: A typical EPC image used for paste thickness measurement Determination of elastic modulus The dynamic elastic moduli of the EPC specimens were determined in 75 mm diameter x 125 mm long cylindrical specimens using the ultrasonic pulse velocity method after 28 days of moist curing. The ends of the specimen were smoothened so as to provide proper contact with the transducers of the pulse velocity meter. The ultrasonic pulse velocity (V) was determined based on the time of travel of the ultrasonic pulse between the two transducers kept at opposite ends of the specimen. From the measured specimen density (ǒ) and the Poisson’s ratio assumed (Ȟ = 0.2), the elastic modulus (E) was calculated as: (1) Measurements were carried out on three EPC cylindrical specimens, and the average value is reported as the elastic modulus of the particular specimen.

RESULTS AND DISCUSSION Porosity, effective pore sizes, and paste thickness Porosity is one of the most significant pore structure features of any porous material that dictates its mechanical and transport properties. For a macroporous material such as EPC, the overall porosity of the specimen is easy to measure, making it a useful tool for in-situ material evaluation and quality control. In fact, currently porosity is the only pore structure feature that is routinely measured for EPC in order to ascertain its permeability, which is its most important functional performance characteristic. The average porosity values obtained from the image analysis procedure for different EPC mixtures are shown in Figure 4. As explained earlier, the reported porosity value is the average of eight images from a single EPC specimen. The average porosities of the EPC mixtures ranged from 17 % to 28 %. The error bars indicate one standard deviation of the observed porosities for different specimens. A general trend of increasing porosity with increasing amount of smaller aggregates in the mixture is observed. This could be an attribute of the compaction method adopted.

30-Predicting.qxd

9/19/2009

280

12:00 PM

Page 280

Milani S. SUMANASOORIYA, Omkar DEO and Narayanan NEITHALATH

The pore sizes and its distribution are also significant in the performance of porous materials. EPC consists of relatively large pores (2-8 mm), and a highly connected pore structure (Neithalath 2004 [2], Neithalath et al. 2006 [4], Marolf et al. 2004 [7]). A recent study (Sumanasooriya and Neithalath 2009 [8]) has dealt with stereology and mathematical morphology based evaluation of pore structure characteristics of EPC. The effective pore sizes (d50) obtained from frequency distributions of pores are shown in Figure 5. It can be seen that the effective pore sizes are lower for EPC specimen made with increasing amounts of smaller aggregates, as expected. Even though the effective pore size is lower for these mixtures, it is believed that the higher porosity values observed in Figure 4 could be due to the presence of a larger number of pores in EPC mixtures made with smaller sized aggregates. EPC specimens made with 100 % 3/8” aggregates showed the maximum effective pore size (2.80 mm) and the specimen made with 100 % #8 aggregates showed the lowest effective pore size (1.78 mm). The pore sizes for the blended mixes are found to lie in between the pore sizes corresponding to the single-sized aggregate mixtures. It can also be noticed from Figure 5 that the relative variation in the pore sizes becomes larger as the aggregate size increases. This is because, for the larger aggregate EPC mixtures, there are only fewer pores present, and as a result the material becomes increasingly heterogeneous. 30

Porosity, %

25

20

15

10 100 - #8

100 - #4

25 - 3/8" - 75 - #4

50 - 3/8" 75 - 3/8" 100 - 3/8" - 50 - #4 - 25 - #4

Aggregate composition

Figure 4: Porosity of different EPC mixtures

30-Predicting.qxd

9/19/2009

12:00 PM

Page 281

Predicting the elastic moduli of enhanced porosity (pervious) concretes using reconstructed 3D material structures

281

Effective pore size (d50), mm

4.0 3.5 3.0 2.5 2.0 1.5 1.0 100 - #8

100 - #4

25 - 3/8" - 75 - #4

50 - 3/8" 75 - 3/8" 100 - 3/8" - 50 - #4 - 25 - #4

Aggregate composition

Figure 5: Effective pore size (d50) for different EPC mixtures The median paste thickness values for the EPC mixtures used in this study as obtained from image analysis ranged from 0.31mm to 0.85 mm. The paste thickness values were found to be normally distributed for almost every specimen. A typical distribution of the paste thickness is shown in Figure 6. 30

Frequency, %

25 20 15 10 5 0 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Paste thickness, mm

Figure 6: Paste thickness distribution In order to computationally predict the elastic modulus of the composite material, it is necessary to have an understanding of the elastic moduli of the constituent phases and their volume fractions. The volume fraction of the aggregate in the mixture can be easily obtained. To obtain the effective modulus of a “solid” particle (aggregate + paste that surrounds it) that can be used in computational prediction, the thickness of the paste layer that surrounds the aggregate is used along with the effective aggregate size (deff-agg). To represent the aggregate size in a

30-Predicting.qxd

9/19/2009

12:00 PM

282

Page 282

Milani S. SUMANASOORIYA, Omkar DEO and Narayanan NEITHALATH

particular EPC mixture, “effective aggregate size” (deff-agg) is used assuming the aggregate particles to be spherical. For single-sized aggregate mixtures, the deff-agg value is the nominal aggregate size itself (2.36 mm for 100 % #8 aggregate mixture, 4.75 mm for 100 % #4 aggregate mixture and 9.5 mm for 100 % 3/8” aggregate mixture). For blended aggregate mixtures, deff-agg was calculated using the nominal sizes of the aggregates in the mixture and their respective volume fractions. For example, the effective aggregate size of 50 % - 3/8” - 50 % - #4 blended mixture was calculated as (0.5 × 9.5 + 0.5 × 4.75 = 7.13 mm). Effective aggregate sizes for different EPC mixtures are shown in Table 1. Figure 7 depicts the variation of median paste thickness with the effective aggregate size. It can be seen that the paste thickness increases with increase in effective aggregate size. For a constant paste volume (aggregate content, aggregate-to-cement ratio and water-to-cement ratio are same for all the mixtures), since the smaller aggregates have more surface area, the paste thickness for mixtures made using smaller aggregates is lower.

Median paste thickness (t), mm

1.4 1.2 t = 0.22e0.12deff-agg R2 = 0.89

1.0 0.8 0.6 0.4 0.2 0.0 0

2

4

6

8

10

12

Effective aggregate size (deff-agg), mm

Figure 7: Variation of paste thickness (t) with effective aggregate size (deff-agg) Dynamic elastic modulus from pulse velocities The experimentally determined elastic moduli for different EPC mixtures are shown in Figure 8. It is well known that elastic modulus is directly related to the porosity of concrete (Mehta and Monteiro 2006 [19]). As expected, it can be clearly seen from Figure 9 that the mixtures made with smaller aggregates (having a higher porosity) have lower elastic moduli. The experimentally determined elastic modulus was found to lie between 9 GPa and 19 GPa with a maximum standard deviation of 1.7 GPa. It can be seen that the elastic modulus of EPC are lower when compared to normal concretes (20 GPa – 30 GPa) (Mehta and Monteiro 2006 [19]) due to its highly porous material structure.

30-Predicting.qxd

9/19/2009

12:00 PM

Page 283

Predicting the elastic moduli of enhanced porosity (pervious) concretes using reconstructed 3D material structures

283

Elastic modulus, GPa

20

16

12

8

4 100 - #8

100 - #4

25 - 3/8" - 75 - #4

50 - 3/8" 75 - 3/8" 100 - 3/8" - 50 - #4 - 25 - #4

Aggregate composition

Figure 8: Variation of elastic modulus for different EPC mixtures

Computational elastic modulus prediction from planar images of EPC Effective modulus of the solid phase: The hardened EPC effectively consists of three phases aggregate phase, cement paste phase surrounding the aggregates, and the pore phase. The aggregate phase and the cement paste phase have different elastic moduli while the elastic modulus of the pore phase is zero. When predicting elastic modulus based on the reconstructed 3D material structures (which will be discussed in the next section), it is necessary to assign elastic moduli values for paste phases would necessitate calculation different phases. Assigning separate elastic moduli values for the aggregate and the cement for a three-phase composite. To simplify this process, the aggregate and the cement paste phases are mapped into an effective spherical “solid” particle. A spherical particle and a surrounding shell, both having different elastic moduli can be accurately mapped into an effective particle having a uniform elastic modulus (Garboczi and Berryman 2001 [14]). It is assumed that there is no overlap between the paste shells belonging to the adjacent aggregates. Figure 9 shows the schematic of effective particle mapping for an aggregate particle of diameter deff-agg (effective aggregate size) and paste thickness t. The elastic modulus for aggregates (Eagg) was taken as 45 GPa (Mehta and Monteiro 2006 [19]) and for the cement paste with 0.33 w/c, the elastic modulus (Epaste) was taken as 22 GPa (de Larrard 1999 [20]). Based on the paste thickness and effective aggregate size, the corresponding volume fractions of aggregate (Vf-agg) and cement paste (Vf-paste) were calculated for the effective particle and are shown in Table 1. The effective elastic modulus (Eeff) of the solid phase was calculated using a simple Voigt model as shown in Equation 2, and the values are given in Table 1. (2)

30-Predicting.qxd

284

9/19/2009

12:00 PM

Page 284

Milani S. SUMANASOORIYA, Omkar DEO and Narayanan NEITHALATH

deff-agg

deff-agg+ 2t

Figure 9: Mapping of an aggregate and the paste surrounding it into an effective particle Table 1: Properties of aggregates and cement pastes for different EPC mixtures Effective Median paste thickness, mm Aggregate Vf-paste aggregate Vf-agg Standard composition Average value size, mm deviation 100 - #8 2.36 0.31 0.11 0.50 0.50 100 - #4 4.75 0.42 0.20 0.61 0.39 25-3/8 - 75-#4 5.94 0.43 0.23 0.67 0.33 50-3/8 - 50-#4 7.13 0.48 0.19 0.68 0.32 75-3/8 - 25-#4 8.31 0.55 0.27 0.69 0.31 100 -3/8 9.50 0.85 0.39 0.61 0.39

Eeff, GPa 33.5 36.1 37.3 37.7 37.9 36.1

3D Material structure reconstruction from planar images: Digitized 3D reconstructed material structures of the EPC specimens were developed using the square images from real EPC specimens (Figure 1). A set of C and FORTRAN computer programs, developed at NIST, and freely downloadable at NIST ftp sites (ftp://ftp.nist.gov/pub/bfrl/bentz/permsolver and ftp://ftp.nist.gov/pub/bfrl/garbocz/FDFEMANUAL) were used for 3D reconstruction and elastic modulus predictions using a parallel computing cluster. For a particular EPC mixture, two point correlation functions (TPC) for the square images were initially obtained. The TPC function is a measure of the probability of finding two randomly selected points, both lying in the same phase, in a two-phase random composite media. Details on TPC functions and their applications in computational materials science based material property predictions can be found elsewhere (Bentz and Martys 1994 [9], Garboczi et al. 1999 [11], Berryman 1985 [21], Berryman and Blair 1986 [22]). The TPC function was then utilized to reconstruct 100 x 100 x 100 pixels 3D material structures having the same porosity (average image porosity – Figure 4) using a three dimensional reconstruction algorithm. The system size of 100 x 100 x 100 pixels was chosen in order to expedite the execution of computer programs used for elastic modulus predictions. This algorithm generates a 3D material structure with similar properties as that of the real material (Bentz and Martys 1994 [9], Bentz et al. 2009 [15], Bentz 2008 [23]). However the hydraulic radius of the reconstructed microstructure was found to be consistently lower than that of the original image, and hence a sintering algorithm (Bentz and Martys 1994 [9], Bentz and Forney 2000 [24]) was employed in order to modify the reconstructed material structure to match the hydraulic radius of the original square image. The planar slices extracted from the modified

30-Predicting.qxd

9/19/2009

12:00 PM

Page 285

Predicting the elastic moduli of enhanced porosity (pervious) concretes using reconstructed 3D material structures

285

material structures were found to have similar pore structure features as that of the starting square images. Elastic modulus prediction: A finite element program (computer program elas3d.f, available at ftp://ftp.nist.gov/pub/bfrl/garbocz/FDFEMANUAL) was employed on the modified 3D reconstructed material structures to obtain the effective elastic modulus of the various EPC mixtures. In this process the digitized 3D material structure was subjected to applied microscopic strains and the linear elastic equations were solved to minimize the total energy stored in the microstructure through a variational principle (Garboczi 1998 [25]). Each cubic pixel (voxel) in the reconstructed 3D material structure was treated as a cubic tri-linear finite element, and the elastic displacements calculated for each pixel. The average stresses and strains were calculated for each pixel and the six different components (normal and shear components) of effective stresses and strains obtained by volume averaging those values. The details of this process can be found in (Garboczi 1998 [25], Garboczi and Day 1995 [26]). When computing the effective elastic modulus of the composite, the elastic modulus and Poisson’s ratios need to be defined for each phase in the material structure. As described earlier, in this study the aggregate surrounded by the cement paste was treated as one effective spherical particle and hence EPC consists of two phases - the solid phase, and the pore phase. The elastic properties for pore phase can be taken as zero, and the effective elastic moduli for the solid phase are given in Table 1 for the different EPC mixtures used in this study. The Poisson’s ratio (Ȟ) for the solid phase was taken as 0.2 and normal strains (İxx, İyy, İzz) in x (i), y (j) and z (k) directions of 0.1 and shear strains of (İxz = 0.1), (İyz = 0.2), and (İxy = 0.3) were applied to each cubic trilinear element. The effective stresses and strains were obtained and the effective elastic modulus was calculated using Equations 3 and 4. Figure 10 shows the steps involved in 3D reconstruction and elastic modulus predictions. (3) (4) 0.25

0.20

0.15

3D reconstruction

s2

TPC function

0.10

0.05

0.00 0

1

2

3

4

5

6

7

8

r, mm

400 x 400 pixel square image

TPC function Input Eeff and Ȟ for each phases Predicted Elastic modulus

Reconstructed 3D image (100 x 100 x 100 pixels)

Computed İij and ıij

Figure 10: Steps in 3D reconstruction and elastic modulus predictions

30-Predicting.qxd

9/19/2009

12:00 PM

286

Page 286

Milani S. SUMANASOORIYA, Omkar DEO and Narayanan NEITHALATH

Comparison of experimental and predicted elastic moduli: At least two 3D material structures were generated for a given EPC mixture with different starting 2D images. The elastic moduli were predicted for each of these material structures and the average value is reported as the predicted elastic modulus for the corresponding EPC mixture. Figure 11 shows the comparison between the experimentally measured elastic modulus and the elastic modulus predicted from reconstructed 3D material structures. It can be seen that the predicted elastic moduli are in good agreement with the experimentally determined values. The assumption of spherical “solid” phases, and the elastic modulus values considered for the aggregate and cement paste phases are found not to adversely affect the composite elastic moduli predictions. The computational modeling procedure also has assumed that the material is homogeneous and isotropic. It is likely that, in EPC specimens made with large sized aggregates (and hence larger pores), this assumption might not be completely satisfied, especially in the 100 x 100 x 100 pixel reconstructed material structure. However, the predicted values are not found to be significantly affected. Elastic modulus - predicted, GPa

20 18 16 14 12 10 8 6 6

8

10

12

14

16

18

20

Elastic modulus- experimental, GPa

Figure 11: Variation of predicted elastic modulus with experimentally measured elastic modulus Figure 12 depicts the variation of experimental and predicted elastic modulus with the average porosity for different EPC mixtures. The experimental elastic modulus – porosity relationship is found to be better than the predicted elastic modulus – porosity relationship. However, the prediction methodology has shown that it is possible to obtain reasonable elastic modulus values from 2D images of EPC, even when the material structure is very heterogeneous even at a macroscale.

30-Predicting.qxd

9/19/2009

12:00 PM

Page 287

Predicting the elastic moduli of enhanced porosity (pervious) concretes using reconstructed 3D material structures

287

Elastic modulus (E), GPa

20 Experimental Predicted

18 16 14

Eexp = 51.8e-0.07) R2 = 0.90

12 10

Epredict = 38.2e-0.05) R2 = 0.83

8 6 12

14

16

18

20

22

24

26

28

30

Porosity ()), %

Figure 12: Variation of elastic moduli with porosity

CONCLUSIONS This paper has detailed a computational methodology for elastic modulus prediction of six different EPC mixtures made using different aggregate sizes or combinations. The conclusions of the study are itemized below: i. Determination of pore structure features such as porosity, pore sizes, and paste thickness were carried out so as to be used in the computational procedure as well as to determine the effective properties of the solid phase in the composite. EPC specimens made with higher proportion of smaller aggregates were found to have higher porosities. The reported porosities of the studied mixtures ranged from 17% to 28%. The effective pore size (d50) was found to be higher for EPC specimens made with larger aggregates. For EPC specimens made with blended aggregates, d50 was found to lie in between those of the single sized aggregates used in the blend. Larger sized aggregate EPC mixtures were also found to have higher paste thickness values. ii. An effective aggregate size (deff-agg) was defined to represent the aggregate sizes in different EPC mixtures, especially those made using blended aggregates. deff-agg was obtained using the nominal sizes of the aggregates that constituted the blend, and their volume fractions. The deffagg values were used along with the median paste thickness to represent an “effective particle”, for which the elastic modulus was determined using a simple Voigt model. Combining the aggregate and the paste into an effective solid particle reduced the composite into a two-phase system, which is computationally more convenient. iii. Three dimensional material structures (100 x 100 x 100 pixels) for different EPC mixtures were successfully developed using two dimensional square images from real EPC cross sections. Finite element elastic programs (developed at NIST) were utilized to obtain the effective stresses and strains in the 3D material structure, and thus to predict the elastic modulus of different EPC mixtures. The predicted elastic moduli of the 3D material structures of EPC

30-Predicting.qxd

288

9/19/2009

12:00 PM

Page 288

Milani S. SUMANASOORIYA, Omkar DEO and Narayanan NEITHALATH

were found to be in good agreement with the dynamic elastic modulus values for different EPC mixtures experimentally determined using ultrasonic pulse velocities.

ACKNOWLEDGEMENTS The authors gratefully acknowledge the financial support for this work from the National Science Foundation (NSF) through a CAREER award (CMMI 0747897) to the third author. REFERENCES 1. ACI 522R – 06, Pervious concrete. American Concrete Institute Committee, 2006 2. Neithalath, N., Development and characterization of acoustically efficient cementitious materials. PhD thesis, Purdue University, West Lafayette, Indiana 2004, pp 269 3. Neithalath, N., Marolf, A., Weiss, J., Olek, J., Modeling the influence of pore structure on the acoustic absorption of enhanced porosity concrete. J. of Advanced Concrete Technology, 3, 2005, 29-40 4. Neithalath, N., Weiss, J., Olek, J., Characterizing enhanced porosity concrete using electrical impedance to predict acoustic and hydraulic performance. Cement and Concrete Research, 36, 2006, 2074-2085 5. Yang, J., Jiang, G., Experimental study on properties of pervious concrete pavement materials. Cement and Concrete Research, 33, 2003, 381-386 6. Tennis, P.D., Leming, M.L., Akers, D.J., Pervious concrete pavements. Portland Cement Association, Skokie IL 2004, pp 28 7. Marolf, A., Neithalath, N., Sell, E., Wegner, K., Weiss, J., Olek, J., Influence of aggregate size and gradation on acoustic absorption of enhanced porosity concrete. ACI Materials Journal, 101, 2004, 82-91 8. Sumanasooriya, M.S., Neithalath, N., Stereology and morphology based pore structure descriptors of Enhanced Porosity (pervious) Concretes. ACI Materials Journal, accepted for publication in 2009 9. Bentz, D.P., Martys, N.S., Hydraulic radius and transport in reconstructed model porous media, Transport in Porous Media, 17, 1994, 221-238 10. Bentz, D.P., CEMHYD3D: A three-dimensional cement hydration and microstructure development modeling package. Version 3.0, National Institute of Standards and Technology Interagency Report 7232, Technology Administration, U.S. Department of Commerce, June 2005 11. Garboczi, E.J., Bentz, D.P., Martys, N.S., Digital images and computer modeling. Experimental Methods in the Physical Science, Methods in the Physics of Porous Media, 35, 1999, 1-41 12. Bentz, D.P., Martys, N.S., A stokes permeability solver for three dimensional porous media. National Institute of Standards and Technology Interagency Report 7416, Technology Administration, U.S. Department of Commerce, April 2007 13. Meille, S., Garboczi, E.J., Linear elastic properties of 2D and 3D models of porous materials made from elongated objects. Modeling and Simulation of Materials Science and Engineering, 9, 2001, 371-390

30-Predicting.qxd

9/19/2009

12:00 PM

Page 289

Predicting the elastic moduli of enhanced porosity (pervious) concretes using reconstructed 3D material structures

289

14. Garboczi, E.J., Berryman, J.G., Elastic moduli of a material containing composite inclusions: effective medium theory and finite element computations. Mechanics of Materials, 33, 2001, 455-470 15. Bentz, D.P., Garboczi, E.J., Martys, N., Snyder, K.A., Guthrie, W.S., Kyritsis, K., Neithalath, N., Virtual testing of concrete transport properties. Submitted to ACI Fall 2009 session. “Material Science Modeling as a Solution to Concrete Problems”, 2009 16. Sumanasooriya, M.S., Bentz, D.P., Neithalath, N., Predicting the permeability of pervious concretes from planar images. Submitted to 2009 Concrete Technology Forum, National Ready Mix Concrete Association, Cincinnati 2009 17. Low, K., Harz, D., Neithalath, N., Statistical characterization of the pore structure of enhanced porosity concrete. Proceedings in CD of the 2008 Concrete Technology Forum, National Ready Mix Concrete Association, Denver 2008 18. Hu, J., Stroeven, P., Local porosity analysis of pore structure in cement paste. Cement and Concrete Research, 35, 2005, 233-242 19. Mehta, P. K., Monteiro, P.J.M., Concrete Microstructure, Properties and Materials. 3rd edition McGraw-Hill, New York 2006, pp 659 20. de Larrard, F., Concrete Mixture Proportioning - A Scientific Approach. Modern Concrete Technology, E & FN Spon, An imprint of Routledge London-New York 1999, pp 421 21. Berryman, J.G., Measurement of spatial correlation functions using image processing techniques. J. of Applied Physics, 57, 1985, 2374-2384 22. Berryman, J.G., Blair, S.C., Use of digital image analysis to estimate fluid permeability of porous materials: application of two-point correlation functions. J. of Applied Physics, 60, 1986, 1930-1938 23. Bentz, D.P., Virtual pervious concrete: Microstructure, percolation, and permeability.ACI Materials Journal, 105, 2008, 297-301 24. Bentz, D.P., Forney, G.P., User's guide to the NIST virtual cement and concrete testing laboratory. Version 1.0, National Institute of Standards and Technology Interagency Report 6583, Technology Administration, U.S. Department of Commerce, November 2000 25. Garboczi, E.J., Finite element and finite difference programs for computing the linear electric and elastic properties of digital images of random materials. National Institute of Standards and Technology Interagency Report 6269, Technology Administration, U.S. Department of Commerce, December 1998 26. Garboczi, E.J., Day, A.R., An algorithm for computing the effective linear elastic properties of heterogeneous materials: Three-dimensional results for composites with equal phase Poisson ratios. J. of the Mechanics and Physics of Solids, 43, 1995, 1349-1362

31-Enhanced.qxd

9/19/2009

12:00 PM

Page 291

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

KEYNOTE PAPER

KEYNOTE PAPER ENHANCED DURABILITY OF CONCRETE BY SUPERABSORBENT POLYMERS Hans W. REINHARDT, Alexander ASSMANN Department of Construction Materials University of Stuttgart Pfaffenwaldring 4, 70569 Stuttgart, Germany, e-mail: [email protected]

ABSTRACT Superabsorbent polymers (SAP) are a new type of admixture. They can store many times more water than their own weight. They can be used for internal curing of concrete. When they dry out they stay as air-filled pores of about 100 Pm diameter and act similarly as pores generated by air-entraining agent. They interrupt the capillary pores and influence the transport properties. The paper presents and discusses test results with respect to water permeability, oxygen permeability, and capillary suction which are relevant for concrete durability.

Keywords Concrete, superabsorbent polymers, water permeability, oxygen permeability, capillary suction

INTRODUCTION The durability of concrete structures depends strongly on the physical properties of concrete. The water permeability governs the ingress of deleterious substances such as chlorides, the gas diffusion and permeability is responsible for carbonation of concrete and corrosion of reinforcing steel, water supply triggers alkali-silica reaction and leaching, freeze-thaw problems occur with high water saturation. Superabsorbent polymers (SAP) can store many times more water than their own weight. When mixing in fresh concrete they absorb water which can serve as interior water reservoir for the hydration of cement. Thus, it is available for internal curing which may cause a denser structure of the cement matrix. On the other hand, when the SAP dries out it remains as empty pore which can act similarly as air voids due to air-entraining agent. One can assume that the freeze-thaw resistance can be improved. In order to know more about the transport properties of SAP modified concrete test series have been started at the University of Stuttgart. In the first series, the oxygen and water permeability, and the capillary suction were determined wheras, in the second series, the freeze-thaw resistance and the chloride migration have been investigated. Over the first series shall be reported in the following. The results of the second series can be found elsewhere [1].

SUPERABSORBENT POLYMERS The used superabsorbent polymer (SAP) was an acrylic type which can be divided into three classes of particle diameter in dry condition shown in Fig. 1 as particle-radius distribution

31-Enhanced.qxd

9/19/2009

12:00 PM

Page 292

292

Hans W. REINHARDT and Alexander ASSMANN

curves. The water absorption of the particles was determined to be 24 g water/g polymer which makes the particle diameters to swell by a factor of 3.26. The three classes get the ranges < 63, 63 to 125 and 125 to 250 ҏPm in the saturated condition. 100

Passing [% by mass]

90 80 70 60 50 40 30 20 10 0 0

10 20 30 40 50 60 70 80 90 100 110 120 130 140 150

Radius of dry particle [m]

Figure 1: Particle size distribution of dry SAP From the radius distribution curve and the added amount of SAP, the total SAP pore volume and the total number of pores in 1cm³ concrete can be estimated. The values are listed in Table 1. Table 1: Concrete mixtures produced for the experimental series on permeability Mixture Water

NWA

M1

M2

M3

M4

M5

M6

M7

M8

w

l/m³

225

189

225

225

225

162

225

189

w stored

l/m³

-

-

36

36

36

-

63

27

(w/c)tot

-

0.50

0.42

0.50

0.50

0.50

0.36

0.50

0.42

(w/c)eff

-

0.50

0.42

0.42

0.42

0.42

0.36

0.36

0.36

1603 1697

1600

1600

1600

1768

1598

1695

Total

kg/m3

fdry

kg/m3

-

-

1.50

1.50

1.50

-

2.63

1.13

m-% b.c

-

-

0.333

0.333

0.333

-

0.58

0.25

Size

Pm

-

-

< 63 63
Median pore radius rsm

m

-

-

65

147

244

-

147

147

Pore volume % b. V.

-

-

3.75

3.75

3.75

-

6.6

2.8

Total number of SAP pores

p. cm³

-

-

120170

4970

785

-

8745

3710

kg/m3

0.68

1.80

1.80

2.03

2.03

5.63

4.05

3.60

PCE PCE

PCE

PCE

PCE

PCE

PCE

PCE

0.40

0.45

0.45

1.25

0.90

0.80

Admixtures SAP

Fraction

SuperAmount plasticizer Type Fraction

m-% b.c 0.15

0.40

-

63
31-Enhanced.qxd

9/19/2009

12:00 PM

Page 293

293

Enhanced durability of concrete by superabsorbent polymers

CONCRETE COMPOSITION Three reference mixtures with w/c-ratios of 0.50 (M1), 0.42 (M2) and 0.36 (M6) and five mixtures with SAP addition (M3-M5, M7, M8), varying content and particle size distribution, were produced. To evaluate the influence of SAP on concrete properties, for each mixture with SAP addition two reference mixtures can be taken for comparison: one reference mixture with same w/c-ratio and a second with w/c according to w/ceff of the SAP-mixture. w/ceff is defined as w/c-ratio by disregarding the water stored in SAP pores. Table 1 shows the mixture composition in detail. The used cement type was a CEM I 42.5 R with a constant amount of 450 kg/m³. The chosen grading curve of silicious Rhine gravel aggregates was an approximated ‘A/B’ curve according to DIN 1045-2 with a maximum grain size of 8 mm in diameter. To achieve the same consistency for all mixtures, superplastizicer Glenium ACE 30 was used.

PROPERTIES OF FRESH CONCRETE The total mixing time was 5 min 30 s by normal speed in a pan-type mixer with a capacity of 150 dm³. After mixing aggregates, cement and SAP in dry condition for the first 60 s - to achieve a better dispersal of SAP - half of the water was added. Within the following 1 min 30 s, the water absorption of SAP started. Superplasticizer and remaining water were added and the mixer was mixing for another 3 min. Former experience confirm the chosen mixing time to be effective. To achieve higher workability of M8, 0.3% superplasticizer by mass of cement was added supplementary. Therefore the total mixing time was extended to 7 min 30 s. Observations have shown, that later added superplasticizer is less effective due to lack of free water. Spread, temperature, air content and bulk density of the fresh concretes were measured directly after mixing. All measured values are listed in Table 2. Table 2: Fresh concrete properties Mixture M1

M2

M3

M4

M5

M6

M7

M8

Spread

38

36

40

39

50

53

52

2.29

2.27

2.27

2.28

-

2.29

-

Bulk density

[cm]

44 3

[kg/dm ] 2.28 o

Concrete temperature [ C]

26.9

26.8

25.2

24.5

25.6

22.4

22.9

23.5

Air content

1.95

3.40

2.95

2.35

2.50

0.70

1.30

0.80

[%]

The target value of concrete spread was 45cm. Actual values range between 36 and 53cm. The measured air content is correlated inversely with the spread: a more flowable concrete means better compactibility, and consequently a lower air content. A change of air content over time has not been observed.

STORAGE CONDITIONS All specimens were left in their moulds for 24 hours at 20°C and 100% RH. To simulate different curing periods, half of the specimens were stored for three more days at 20°C and 100% RH. The other half was put at 23°C and 50% RH. Specimens used for capillary suction, water and oxygen permeability were sealed at the perimeter with epoxy one week after

31-Enhanced.qxd

9/19/2009

12:00 PM

Page 294

294

Hans W. REINHARDT and Alexander ASSMANN

production. The specimens for water permeability were stored in water from the 7th to the 28th day. The tests were performed at an age of 28 days. Specimens used for oxygen permeability were used later for mercury intrusion porosimetry.

COMPRESSIVE STRENGTH Three cubes with 150 mm edge length were tested at an age of 28 days. The average values of measured compressive strength fc in N/mm² are given in Fig. 2. The standard deviation ranges between 0.5 N/mm² (M7) and 2.7 N/mm² (M1). 90

Compressive strength fc [N/mm²]

80 70 60 50 40 30 20 10 0 M1 58.29

M1(2.) 46.60

M2 67.13

M3 50.67

M4 53.06

M5 51.24

M6 84.11

M7 53.60

M8 73.89

Figure 2: 28d compressive strength fc [N/mm²] The strength of SAP modified concrete with low SAP content tends towards strength of reference concretes with the same w/c. The results of M3, M4, M5 and M7 verify this hypothesis: all SAP mixtures with w/c = 0.50 and SAP pore volumes of 3.7% to 6.50% by total volume, show almost same values for fc. The difference in pore size as well as in number of pores (M3-M5) doesn’t seem to have an effect, because of little polymer pore volume compared to volume of air pores. The reference mixture M1 shows a higher value for fc than expected. Due to this result which could not be explained, 3 more cubes were cast in a second batch with 2.9% air content, 44 cm spread and a fresh concrete temperature of 18.2°C. Their 28 days fc reached only 46.6 N/mm², which is an acceptable value.

TESTING PROCEDURE A cut through the cell for water permeability experiments is shown in Fig. 3. Its leak-tightness was secured by the rubber tube (5) with a pressure of 11 bar. The cell for oxygen permeability is analogous [2]. Gas flow through porous materials like concrete can be described by Hagen-Poiseuille’s law. The specific gas permeability coefficient is given by:

K

Q K L 2p 2 A pi po 2

(1)

31-Enhanced.qxd

9/19/2009

12:00 PM

Page 295

Enhanced durability of concrete by superabsorbent polymers

295

For oxygen applies Q [cm³/s] = flow rate, K [m²] = specific permeability coefficient, A [mm²] = test surface, L [mm] = depth of specimen, po= p [10-5N/mm²] = outlet and atmospheric pressure, pi [10-5N/mm²] = inlet pressure, Ș [10-5Ns/m²] = viscosity of test gas. Oxygen was chosen as testing gas: first of all, it causes no change of microstructure like CO2 for example. Second, it is the oxidizing agent in corrosion processes of reinforcement and therefore relevant for durability of concrete structures and, third, it can be handled safely in the laboratory.

Figure 3: Test cell for water permeability 1. aluminium housing; 2. seal pressure sleeve; 3. fixing screw; 4. concrete sample; 5. pressure tube; 6. deaeriation valve; 7. measuring capillary (water pressure); 8. water inlet; 9. backup ring; 10. support; 11. measuring capillary (flow rate) Flow rates Q of oxygen through cylindrical specimen with diameters of 100 mm and 50 mm in depth were determined by a bubble counter. Before cutting the specimen at an age of 7days, the perimeter was sealed with epoxy to eliminate gas flow due to shrinkage cracks. At the end of preparation, the testing surfaces were polished. At 20°C room temperature, the viscosity of oxygen is approximately Ș = 2.02 · 10-5 Ns/m² and the atmospheric pressure is nearly 1 bar. Flow rates were measured by bubble counter for three inlet pressures: 2.0, 2.5 and 3.0 bar. To reach constant values for Q, oxygen flows were not measured until 5 min up to 30 min after the specimens had been put under the adjusted inlet pressure. For laminar flow, Q has to be proportional to (pi²-1). The specific permeability coefficient K can be regarded as gradient of the graphs in Figs. 4a and 4b. Each measuring point is the average of two tested specimens. Cylinders stored 1 day at 20°C/100% RH were named Z3, cylinders with 4 days curing time were named Z4. The dotted lines represent the best fitting regression lines. As the results show, the assumption of a laminar flow is acceptable. The water permeability was determined on specimens with 30 mm in depth and 150 mm in diameter. They were put in the cell and fixed with a metal ring towards the top cover. The whole cell was closed up airtight by pressure tube and filled with water. To apply a water pressure of 7 bar, nitrogen gas was used as a buffer. The water flow through the specimen was read on measuring capillaries.

31-Enhanced.qxd

9/19/2009

12:00 PM

Page 296

296

Hans W. REINHARDT and Alexander ASSMANN

0.7

0.7

0.6

0.6

M1 Z3

0.4

M4 Z3 M5 Z3

0.3

M6 Z3 M7 Z3

0.2

M1 Z4

0.5

M2 Z3 M3 Z3

Q [ml/s]

Q [ml/s]

0.5

M2 Z4 M3 Z4

0.4

M4 Z4 M5 Z4

0.3

M6 Z4 M7 Z4

0.2

M8 Z3

0.1

M8 Z4

0.1

0

0

8

9

10

a)

11

12

(pi ²-1) [bar²]

13

14

15

8

9

10

b)

11

12

13

14

15

(pi ²-1) [bar²]

Figure 4: a) Oxygen flow rates of specimens with 1 day curing time in dependence on inlet pressure (pi²-1); b) Oxygen flow rates of specimens with 4 days curing time in dependence on inlet pressure (pi²-1) Before the measurements started, the specimen in the test cell was under water pressure of 7 bar for 24 hours to make sure that it is completely water saturated. After these 24 hours the preparations had been finished and the measurements were started. Flow rates were determined after 1 hour and 24 hours, according to [2]. In case of laminar flow, Eq. (1) is valid for water permeability, too. Since water is assumed to be incompressible, it follows:

kW

QL A 'h

(2)

with kw [m/s] = water permeability coefficient, Q [m³/s] = flow rate, A [m²] = test surface, L [m] = depth of specimen, ǻh [m] = water pressure.

OXYGEN PERMEABILITY

K [10-15m²]

The measured oxygen permeability coefficients are presented in Fig. 5. They have coefficients of variation between 6 and 13%. In all cases, an extension of curing time leads to a reduction of permeability: 4 d show smaller values of K than 1 d. Average values of 4 days cured specimens range between 60 and 70% of 1 day cured specimens. 0.12 0.11 0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 M1 M1 M2 M2 M3 M3 M4 M4 M5 M5 M6 M6 M7 M7 M8 M8 1d 4d 1d 4d 1d 4d 1d 4d 1d 4d 1d 4d 1d 4d 1d 4d

Figure 5: Oxygen permeability coefficients [10-15m²]

31-Enhanced.qxd

9/19/2009

12:00 PM

Page 297

297

Enhanced durability of concrete by superabsorbent polymers

As SAP is concerned a look at mixtures M3 toM5 is conspicuous, that an increasing particle size distribution of SAP causes less permeability. Indeed the total number of pores is less by same total pore volume, but large SAP pores seem to generate a greater density of microstructure than small SAP pores do. The most notable effect could be that smaller numbers of SAP pores do not create a continous structure. Furthermore, the volume of a spherical pore depends on the third power of the diameter whereas the surface of a sphere grows only with the second power. Therefore, the larger the pores with the same total porosity then less accessible towards oxygen and fluids. Another speculation is that larger pores become denser because of their better water supply to the surrounding cement paste, accompanied by a stronger growth of hydration products into the pores. In general, the determined permeability coefficients agree with values given by literature [3]. Beside the strength class, K highly depends on moisture content of tested specimens [4]. In [5] specimens of same mixtures were exposed to different climates. Higher moisture contents caused a decrease of K up to one or two orders of magnitude. To compare different concretes with each other, same storage conditions have to be kept. In our case, all specimens were dried in the oven at 60°C till they reached constant weight. If standard climate of 20°C and 65% would have kept it could be anticipated that SAP mixtures would have shown lower values of K because of delayed desiccation in comparison to reference mixtures.

WATER PERMEABILITY

10

1

9

0.9

8

0.8

7

0.7

kw [10-12m/s]

kw [10-12m/s]

The outcome of the measured water flow rates after 1 hour respectively 24 hours, leads to the values for kw presented in Fig. 6.

6 5 4

0.5 0.4

3

0.3

2

0.2

1

a)

0.6

0.1 M1

M2

M3

M4

M5

b)

M1

M2

M3

M4

M5

Figure 6: a) Water permeability coefficients after 1 hour; b) Water permeability coefficients after 24 hours It can be clearly seen that kw decreases with increasing time. Some reasons are the reactivity of water with unhydrated cement particles and the swelling of the concrete itself. This causes further density of the matrix. The results of the water permeability coefficients after 24 hours confirm the tendency deduced from oxygen permeability measurements: increasing particle size of SAP causes less permeability. However values of kw determined after 1 hour show high variations. It seems likely, that the specimen were not saturated completely at the beginning of the tests. This assumption gets affirmed by another observation: air bubbles rose up at the non-pressurized surface during the measurements. On the one hand they might be coming from the dissolved pressure gas, on the other hand they might have emerged from the

31-Enhanced.qxd

9/19/2009

12:00 PM

Page 298

298

Hans W. REINHARDT and Alexander ASSMANN

air in the larger pores which is squeezed out. Both reasons weaken the value of the results. To get reliable data, the investigation about water permeability will be extended.

CAPILLARY SUCTION

In case of water absorption of dry concrete, capillary suction is the most prominent transport mechanism. Cylindrical specimens with 150 mm in diameter were sealed with epoxy on the perimeter and cut to length. The cut surface was polished till all specimens had the same weight of 5000 g. Figs. 7a and 7b show the absolute mass of absorbed water in g as well as the relative water absorption in % by mass over a period of 3 days. M2

M3

M4

M5

M6

M7

M8

M1

M3

M4

M5

M6

M7

M8 2

90

1.8

90

1.8

80

1.6

80

1.6

70

1.4

70

1.4

60

1.2

60

1.2

50

1

50

1

40

0.8

40

0.8

30

0.6

30

0.6

20

0.4

20

0.4

10

0.2

10

0.2

0

water absorption [g]

0

0 0

a)

M2

100

water absorption [% by mass]

water absorption [g]

2

10

20

30

40 time [h]

50

60

70

0 0

80

b)

water absorption [% by mass]

M1 100

10

20

30

40

50

60

70

80

time [h]

Figure 7: a) Water absorption by capillary suction vs. time for cylinders with 1 d curing time; b) Water absorption by capillary suction vs. time for cylinders with 4 d curing time An extension of curing time causes only a small decrease of water absorption by capillary suction. In general, the results of the SAP mixtures show the same tendency known from permeability testing: added SAP cause densification of the matrix and lead to less capillary porosity. High numbers of small particles create a coherent pore structure which absorbs water easier than a structure caused by a small number of larger particles. However, the difference between M4 and M5 is less pronounced than expected. Both mixtures show nearly the same performance. This finding refers to the fact that capillary pores are scaled below SAP pores. For capillary pores, which have diameters of 10-5 to 0.1 mm, capillary rise is inversely proportional to the pore radius. With growing pore sizes (> 0.1 mm), capillary action decreases till the influence of gravity prevails. Consequently high sized fractions of SAP pores are not relevant for capillary suction. Anyway, SAP pores get filled with water and have a share in the increase of mass. Different moisture contents in the specimens could have influenced the results, too. Due to epoxy sealing and large size of the samples, a period of 21 days at 60°C was not enough to dry out completely. The graphs of water absorption get flattened over the time. The reasons are the same as mentioned with water permeability: water reacts with unhydrated cement particles and causes further densification of the matrix. Secondly the concrete swells when getting in contact with water [6]. Finally the gravity interferes with capillary action. According to theory, the capillary uptake follows a square root of time law [7] except for water at longer duration as stated above. So the water-absorption coefficient w has the unit kg/m2h1/2. Picking values of water absorption in kg/m² in dependence on square root of time over a period of the first 4 hours, a linear relationship appears. The gradient of the dotted

31-Enhanced.qxd

9/19/2009

12:00 PM

Page 299

299

Enhanced durability of concrete by superabsorbent polymers

linear regression lines in Figs. 8a and 8b stand for the water absorption coefficients w. Values for w [kg/m2 h1/2] are given in Table 3. 3

3

M1

M1

M2 2.5

M2 2.5

M3

M3 M4

water absorption [kg/m²]

water absorption [kg/m²]

M4 M5

2

M6 M7 M8

1.5

1

0.5

M5

2

M6 M7 M8

1.5

1

0.5

0

0

0

20

40

60

80

100

1/2

time [s ]

a)

120

0

20

40

60

1/2

80

100

120

time [s ]

b)

Figure 8: a) Water absorption of 1 d cured specimen per m² vs. square root of time; b) Water absorption of 4 d cured specimen per m² vs. square root of time

Mixture Curing time 1 day 4 days

M1

Table 3: Capillary suction w [kg/m2 h½] M2 M3 M4 M5 M6

M7

M8

1.42 1.35

0.97 0.87

0.92 0.91

0.74 0.71

1.23 1.17

1.04 1.04

1.18 1.14

0.61 0.58

It seems to be evident, that the presented results are subject to variance because the experiments comprise only one sample of each mixture. However, the rate of water absorption during the first 4 hours is almost independent of curing time. This phenomenon may be explained: measured median pore radius, which acts as an indicator for capillary pore size, does not change with an extended curing time of 3 days. The reason why mixture M4 has a smaller water absorption rate compared with M5 at the beginning, whereas both mixtures adopt same level at the end of the test, is yet unknown.

CONCLUSIONS

Superabsorbent polymers (SAP) have an effect on strength and transport properties of concrete. (i) Related to the total water-cement ratio the strength is increased due to internal curing. Related to the effective water-cement ratio the strength decreases due to the extra SAP pores. (ii) Oxygen permeability decreases strongly with longer curing time. (iii) SAP reduces oxygen permeability considerably. The larger the SAP pores the more is the oxygen permeability reduced. (iv) Water permeability depends strongly on the flow time. After 24 h the permeability is one order of magnitude less compared to 1 h flow time. (v) SAP reduces water permeability. As with oxygen permeability, larger SAP pores have a more pronounced effect. (vi) Capillary suction is reduced by SAP. The medium size SAP pores have the greatest effect.

31-Enhanced.qxd

300

9/19/2009

12:00 PM

Page 300

Hans W. REINHARDT and Alexander ASSMANN

(vii) Since the SAP pore size has a sgnificant effect on transport properties research has to be focussed on this aspect. ACKNOWLEDGEMENT

The assistance of Dr. Kazuo Ichimiya during the tests is gratefully acknowledged. Financial support is due to Deutsche Forschungsgemeinschaft (German Research Community). REFERENCES

1. Reinhardt, H.W., Assmann, A., Mönnig, S. Superabsorbent polymers (SAPs) – An admixture to increase the durability of concrete. Microstructure related durability of cementitious composites. Ed. by Wei Sun, K. van Breugel, Changwen Miao, Guang Ye, Huisu Chen, RILEM Publ. Proc. 61, Bagneux, 2008, pp. 313-322 2. Jooss, M., Reinhardt, H.W.: Permeability and diffusivity of concrete as function of temperature. Cement and Concrete Research 32 (2002), No. 9, pp. 1497-1504 3. Grube H.: Einfluss der Nachbehandlung auf die Porosität des Betons, Mitteilungen aus dem Forschungsinstitut des Vereins der österreichischen Zementfabrikanten Heft 36, pp. 54-59 4. Jacobs F. P.: Permeabilität und Porengefüge Zementgebundener Werkstoffe, Dissertation Eidgenössische Technische Hochschule Zürich 1994 5. Graef H., Grube H.: Verfahren zur Prüfung der Durchlässigkeit von Mörtel und Beton gegenüber Gasen und Wasser, Beton 36 (1986) Heft 5 pp.184-187 and Heft 6 pp. 222-226 6. Hall, C., Hoff, W.S ., Taylor, S. C., Wilson, M. A., Beom-Gi Yoon, Reinhardt, H.-W., Sosoro, M., Meredith, P., Donald, A. M.: Water anomaly in capillary absorption by cement-based materials. In: Journal of Materials Science Letters 14 (1995), S. 1178-1181 7. Reinhardt, H.W. (Ed.) Penetration and permeability of concrete. Barriers to organic contaminating liquids. RILEM Report No. 16, E&FN SPON, London 1997

32-Volumetric.qxd

9/19/2009

12:00 PM

Page 301

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

VOLUMETRIC STABILITY OF CONCRETE USING RECYCLED CONCRETE AGGREGATES Yogini DESHPANDE1, Jacob E. HILLER2 and Cory J. SHORKEY3 Department of Civil and Environmental Engineering, Michigan Technological University, Houghton, MI, 49921, USA e-mail: [email protected], [email protected],[email protected] ABSTRACT Recent studies show that many pavements built in the United States of America using recycled concrete aggregate (RCA)-based concretes have experienced high amount of shrinkage cracking. Fresh and hardened concrete properties of RCA-based concretes are linked to the porosity and absorption capacity of the RCA aggregate. In this study, detailed evaluation of the porosity and absorption capacity using petrographic techniques of the selected aggregates was performed. This paper presents a study performed to evaluate the volumetric stability of RCA-based concretes using two types of locally manufactured RCA and one virgin aggregate type to assess the effect of recycling operations on the concrete performance. Concrete mixtures with water to cement ratio (w/c of 0.42 were prepared using the selected virgin aggregates and RCA. Fresh concrete properties such as slump and air content were assessed in addition to hardened concrete properties at 28 days such as compressive strength, flexural strength, and elastic modulus. One-dimensional shrinkage of the virgin aggregate and RCA-based concretes was also measured using sealed and unsealed specimens stored at 50% RH to assess drying and total shrinkage respectively. The difference between the sealed and unsealed prisms gave an indication of the drying shrinkage potential of both the RCA-based and virgin aggregate concretes. Restrained shrinkage behavior and its associated cracking susceptibility was evaluated using the ring test as per AASHTO PP 34. Petrographic studies showed that RCA-based aggregates exhibited high amount of porosity. In the volumetric stability evaluation, it was observed that the shrinkage behavior of RCA-based concretes was linked to the porosity and the amount of existing unhydrated cement material on the RCA through the use of scanning electron microscopy. For the same w/cm, it was observed that the RCA concretes containing slag components exhibited higher values of shrinkage strains in comparison with RCA concretes that were originally limestone aggregates.

Keywords: volumetric stability, shrinkage cracking, recycled aggregate, ring test

INTRODUCTION The use of recycled concrete aggregates (RCA) in concrete has been adopted by the construction industry since the 1950’s to overcome the shortage of natural aggregates. Due to the increasing cost for landfills, the scarcity of natural resources coupled with a substantial increase in aggregate requirements for construction, the use of recycled aggregates to partially replace the virgin aggregate has, therefore, become more common. The recent thrust on building sustainable infrastructure has also increased the usage of RCA in new concrete construction [1]. Volumetric stability of concrete is an important parameter and is linked closely to durability and long term performance of concrete. Studies of jointed concrete pavements in the USA have shown excessive shrinkage cracking and premature deterioration of sections

32-Volumetric.qxd

302

9/19/2009

12:00 PM

Page 302

Yogini DESHPANDE, Jacob E. HILLER and Cory J. SHORKEY

containing RCA as the sole aggregates in the concrete slabs [2]. The effect of RCA on concrete properties is related to various factors such as type of virgin aggregate used in the original concrete, age of crushed concrete, amount of mortar adhering the aggregates and the type of mortar adhering the virgin aggregates [3,4]. All these factors mainly affect the porosity and the absorption capacity of the RCA. Higher porosity and absorption capacity in turn affects the workability, compressive strength and shrinkage of the RCA-based concrete. Various studies have shown that use of RCA as partial or full substitution in concrete can potentially lead to higher shrinkage and creep deformations than those compared with normal concrete with same mixture proportions [3,5]. The main reason for higher shrinkage has been related to the presence of paste matrix on the aggregates. Studies have shown that the paste matrix may or may not enhance bond between the new paste and the existing paste on the aggregate [6]. One of the reasons for high shrinkage and low strength could be potentially related to the porosity inherent in the recycled concrete aggregate [5]. Excess water absorbed by the aggregate during the mixing procedure as well during hydration process can lead to deterioration of concrete properties. It is necessary that absorption capacity is correctly measured so that accurate mix design and batching can be developed for the concrete. Studies have shown that current standardized methods adopted for measuring porosity of RCA aggregates may not be accurate due to the presence of mortar on the aggregate [4]. Standardized methods require water submersion of aggregates for 24 hours, which can lead to removal of loose mortar from the aggregates, hydration of mortar leading to inaccurate values of absorption. In this study, two types of recycled concrete aggregates were evaluated in detail. The porosity and absorption capacity was evaluated using standard as well as non standard methods. The non standard test methods adopted were based on usage of helium pycnometer and envelope density analyzer as testing equipments [7]. These test methods do not require water submersion thereby reducing the uncertainty caused by hydration or loss of mortar due to water submersion. The porous nature and porosity was studied using petrographic methods to understand the effect of porosity on the volumetric stability of concretes made using such aggregates.

EXPERIMENTAL METHODS Material In this study, three mixtures with water to cement ratio (w/c) of 0.42 were cast using three different types of aggregates. One mixture contained virgin crushed gravel rock whereas two mixtures were made using recycled concrete aggregates. The RCA used in this study were of two types- i) recycled concrete aggregate manufactured using blast furnace slag aggregate as the original virgin aggregate (referred as slag aggregate RCA in this study) and ii) recycled concrete aggregate manufactured using limestone aggregate as the original virgin aggregate (referred to as limestone aggregate RCA in this study). The gradation of the aggregates was developed based on the gradation specified by Michigan Department of Transportation (MDOT) [8] with a nominal maximum aggregate size of 25 mm (refer Figure 1). The specific gravity and absorption capacity of the coarse aggregates were measured using the helium pycnometer and envelope density analyzer as well as per ASTM C 127 [9]. A detailed petrographic study of the RCA adopted in this study was conducted which included preparation of thin sections of individual aggregate and observed under the optical microscope. Local natural sand conforming to MDOT Standard Specifications #2NS was used as fine aggregate [8]. The gradation of the sand used is provided in Figure 1. The specific

32-Volumetric.qxd

9/19/2009

12:00 PM

Page 303

303

Volumetric stability of concrete using recycled concrete aggregates

gravity and absorption capacity of fine aggregate used was measured as per ASTM C 128 and is provided in the results section [10]. The cement used in this study was of Type I cement. Table 1 provides the mixture proportions for the different types of aggregates used in this study. The water to cement ratio and the bulk volume of coarse aggregates used in the mixture was kept constant at 72 % to adhere to MDOT paving specifications for all mixtures. Air entraining agent (AEA) conforming to ASTM C 260 [11] at the rate of 0.8 % by weight of cement was used to ensure presence of adequate air voids in the mixture. For developing the mixture proportions specific gravity and absorption capacity of coarse aggregates measured using the helium pycnometer and the envelope density analyzer was utilized. Fine aggregate

Coarse Aggregate

Cumulative Passing (%)

100 80 60 40 20 0 100

10

1

0.1

0.01

Sieve Size (mm)

Figure 1. Gradation of aggregates used in the study Table 1. Mixture proportions of concretes Material (kg/m3) Type I Cement Water Coarse Aggregate Fine Aggregate

Crushed Gravel 339 142 1166 688

Slag Aggregate RCA 367 154 1072 495

Limestone Aggregate RCA 367 154 1089 568

Mixing Methodology A 115 L-capacity rotating drum mixer was used for mixing concrete. In field conditions, the RCA is typically dry and exposed to ambient atmosphere for long periods of time. In order to ensure dryness of the coarse aggregate, the RCA was conditioned in an oven at 50oC for 24 hours prior to mixing. The mixing methodology adopted was as follows: RCA+ water required to bring the RCA to surface saturated condition Æmix for 1 minute Æsand+air entraining agent+Type I cement+remaining water Æmix for 3 minutes Æmixer stopped for 2 minutes Æmix for 2 minutes Testing Procedures In this section, test procedures adopted for the study are discussed. Details about the petrographic methodology and non-standardized methods of calculating specific gravity and absorption capacity are provided in detail.

32-Volumetric.qxd

304

9/19/2009

12:00 PM

Page 304

Yogini DESHPANDE, Jacob E. HILLER and Cory J. SHORKEY

Petrographic Methods For creating thin sections of RCA, aggregate samples were chosen randomly with the exception of size. It was important in the process to choose samples of an appropriate size to match the equipment used (approximately 25 x 45 mm). Three samples for each aggregate type were chosen. Samples were then oven dried and placed in a vacuum to remove water and air respectively. The samples were covered with a coarse silica sand to reduce the necessary epoxy resin used in the embedding process (this sand accounts for the noticeably rounded material surrounding each sample, it was chosen for ease of visibility and high contrast to the sample itself.). Initially, the samples were embedded in a blue epoxy resin, followed by curing for 24 hours. The blue epoxy resin can easily penetrate the surface pores and expose their presence when the sample is cut in cross-section and mounted to glass slides using a quick-set epoxy. Next, the samples were ground to a flat polished surface and the exposed face was again embedded with a green epoxy resin following similar drying and vacuuming procedures as above. The re-emersion of the sample in green epoxy assists in exposing the secondary porosity or internal porosity of the aggregate. All cutting and grinding of samples was performed using a kerosene cooled saw and grinder to eliminate hydration of cement or slag particles. Following the green epoxy resin, samples were cured, the exposed face polished to a bare surface and the sample mounted to slides using a clear epoxy for viewing. Finally, samples were ground and polished to the final viewing thickness and observed under an optical microscope. Specific Gravity Assessment The specific gravity of the coarse aggregates was measured using a helium pycnometer to determine the apparent specific gravity and an envelope density analyzer to assess the bulk specific gravity. For detailed information on the two equipments and procedure to conduct the tests, the reader is requested to refer study conducted by Vitton et.al [7]. For this study, a fully automated helium pycnometer with a sample cell volume of 100 cm3 (manufactured by Micromeritics, Inc. of Norcross, Georgia, USA). Each type of aggregate was randomly sampled for each sieve size for use in the helium pycnometer. The average value of five aggregate specimens for each sieve size was then assessed. To conduct a test, the mass of the sample was accurately measured and the sample entered into the helium pycnometer where it calculates the volume of the sample. Dividing the sample weight by its apparent volume gives its apparent specific gravity, GS. An envelope density analyzer was used for evaluating the bulk specific gravity of each aggregate sample. The envelope density analyzer determines the bulk volume of a sample, and given the sample mass, calculates the material’s bulk specific gravity, GB. This is accomplished by initially compacting a micro-grained material called DryFlow¤ (Micromeritics, Norcross, Georgia, USA) in the sample chamber of the envelope density analyzer with a plunger. After compaction the plunger is pulled out of the chamber and aggregate sample is placed in the cylinder. Small amounts of the fine material are removed to accommodate the sample. The plunger is then re-inserted into the chamber and the finegrained material and the sample are again compacted to the previous compaction pressure. The difference in volume between the two tests then provides an estimate of the bulk volume of the sample. Measurement of Fresh and Hardened Concrete Properties The fresh properties measured for all concretes were slump, air content and unit weight. The hardened concrete properties measured were compressive strength, shrinkage and elastic modulus. Three samples were tested for evaluation of each hardened concrete property. Compressive strength and elastic modulus was measured at 28 days as per ASTM C 39 [12] and ASTM C 469 [13] respectively. Free shrinkage of concrete samples in various curing

32-Volumetric.qxd

9/19/2009

12:00 PM

Page 305

305

Volumetric stability of concrete using recycled concrete aggregates

conditions (unsealed and 100% RH, 50% RH and unsealed as well as sealed and 50% RH at 23oC) was measured as per ASTM C 157 [14]. The first measurement was taken at 18 h after addition of mix water to the concrete mixture. Cracking susceptibility of concrete was measured as per AASHTO PP 34 [15] by casting concrete around a ring as shown in Figure 2. The specimens were maintained in 50 % RH and 23oC chamber during the entire duration of the test. Initial readings were taken within 20 minutes of addition of mix water to the concrete mixture. The rings were sealed at the top and bottom as shown in Figure 2. The resistance to chloride ion penetrability was measured as per ASTM C 1202 [16] at the age of 56-days.

Figure 2. Unsealed, sealed, and restrained shrinkage specimens RESULTS AND ANALYSIS Aggregate Properties The correct measurement of absorption capacity of the aggregates is an important factor in developing the mixture proportions of concrete. Amount of water added for absorption affects workability as well as long term properties such as shrinkage and strength. The specific gravity results of coarse aggregates using helium pycnometer and envelope density analyzer as well as standard ASTM procedures are shown in Table 2. Both types of RCA exhibit high values of absorption capacity using either testing methods. While blast furnace slag is inherently a porous material, using this aggregate type from a recycled concrete material only adds to this porosity. The absorption capacity of limestone aggregate RCA is also on the higher side though inherently limestone aggregate does not exhibit a porous nature. Table 2. Characteristics of coarse aggregate Aggregate Type Crushed Gravel Slag Aggregate RCA Limestone Aggregate RCA

Using helium and envelope density analyzer Absorption Porosity Capacity GS GB (%) (%)

ASTM C 127 GS

GB

Absorption Capacity (%)

2.8

2.78

0.27

0.77

2.92

2.82

0.67

2.26

2.21

7.78

17.08

2.5

2.34

4.76

2.60

2.34

4.23

9.80

2.6

2.44

4.25

32-Volumetric.qxd

306

9/19/2009

12:00 PM

Page 306

Yogini DESHPANDE, Jacob E. HILLER and Cory J. SHORKEY

The main difference in values amongst the different methods of testing was absorption capacity. For the slag aggregate RCA, the absorption capacity was twice as much using the helium pycnometer and envelope density analyzer in comparison to the ASTM standards. Helium gas used in the helium pycnometer as well as the micro grain material used in the envelope density analyzer is easily absorbed into the aggregate’s effective pore space than water. The high porosity of the slag aggregate RCA is evident through petrographical observations. Pictorial observation of thin sections of aggregates under the optical microscope is presented in Figure 3. Figure 3a and 3b represents thin sections of slag aggregate RCA. The blue epoxy resin scattered throughout the thin section specimen indicates a large amount of surface porosity for the slag aggregate RCA. Specifically, the slag particles in Figure 2a and 2b (marked by arrows) show presence of porosity in the form of surface (blue-gray color) as well as internal porosity (light yellowish green color). The internal porosity is mostly in the form of unconnected pores. Figure 3c represents thin slide section of limestone aggregate RCA. In the limestone aggregate RCA, the dominant presence of the green epoxy resin indicates a larger percentage of internal porosity but lower amount of surface porosity. Thin sections of the crushed gravel were also made and analyzed under the optical microscope. The natural aggregate had very low porosity and hence significant intrusion of the epoxy resin was not observed. Though in this study a quantitative analysis of the petrographical studies using tools such as image analysis are not presented, the petrographic image observations can be made to conclude that the slag aggregate RCA has an inherent high porosity and in turn higher absorptivity which may or may not be easily quantified by the ASTM methods of measuring absorption capacity of aggregates. The absorption capacity values as per ASTM C 127 for both RCA are comparable but the petrographical analysis show a different picture. Based on these observations the selection of absorption capacity of aggregates measured using helium pycnometer for calculation of mixture proportions of concretes is justifiable.

Surface porosity Figure 3a. Thin section of slag aggregate RCA

Figure 3b Thin section of slag aggregate RCA

32-Volumetric.qxd

9/19/2009

12:00 PM

Page 307

307

Volumetric stability of concrete using recycled concrete aggregates

Internal porosity

Figure 3c. Thin section of limestone aggregate RCA Fresh Concrete Properties The fresh property details of the concrete mixtures are provided in Table 3. The slump was lowest for the concrete mixture containing crushed gravel aggregates and was highest for mixture made with slag aggregate RCA. The absorption capacity of the slag aggregate RCA was highest at 7.78 % amongst the three aggregates used in this study. The high slump values of the slag aggregate RCA mixture could be due to the presence of large amount of water in the mixture which was not easily absorbed by the aggregates during the short mixing time. Table 3. Fresh properties of concrete mixtures Fresh Property

w/c Slump (mm) Air content (%) Unit Weight (kg/m3)

Crushed Gravel 0.42 127 3.75 2469

Concrete Type Slag Aggregate Limestone RCA Aggregate RCA 0.42 0.42 228 178 3.5 5.75 2124 2200

Hardened Concrete Properties The 28-day compressive strength for all the mixtures is shown in Figure 4a. The least compressive strength was achieved by the mixture containing slag aggregate RCA (17 MPa). The highly porous nature of the slag aggregate RCA resulted in low compressive strengths for the mixture. The limestone aggregate RCA mixture also exhibited lower compressive strengths in comparison to the mixture containing virgin crushed gravel aggregates. Some studies have shown that higher compressive strengths are achieved by concretes containing RCA, however, it is dependent on the maximum size of RCA used in the mixture as well as the amount of mortar adhering to the RCA particles. Cabo et. al. postulate that when completely dry RCA are used, due to the high absorption capacity, the effective water to cement ratio is reduced and the compressive strength is increased [4]. In the study conducted by them, it was observed that increase in the amount of RCA substitution resulted in higher compressive strength values. However, in this study the higher absorption capacities of the RCA meant higher amount of water present in aggregate itself resulting in reduced concrete strength. If the water content in aggregate (¨w) is expressed by the following function: [5] w

CA ACCA 100 ACCA

where, CA= coarse aggregate content and ACCA is absorption capacity

32-Volumetric.qxd

9/19/2009

12:00 PM

308

Page 308

Yogini DESHPANDE, Jacob E. HILLER and Cory J. SHORKEY

then, corresponding values for the three mixtures are 13.02 % for crushed gravel, 139.73% for slag aggregate RCA and 76.78% for limestone aggregate RCA. Highest percentage of water present in aggregate was calculated for slag aggregate RCA which also exhibits poorest compressive strength. Elastic modulus at 28-days of the concretes is shown in Figure 4b. The elastic modulus of RCA based mixtures is comparable with each other but is about 20% lower than the mixture containing virgin crushed gravel aggregate. Conflicting experimental results of elastic modulus values of RCA based concretes can be found in the literature (Cabo) [17]. Some studies have shown that concretes made with 100% substitution of RCA exhibit lower elastic modulus but this value is also dependent on the absorption capacity and porosity of the RCA. However, some studies have suggested an improved matrix of RCA based concretes can result in higher elastic modulus values [18]. The flexural strength at 28-days is lower for mixtures with RCA than the mixture containing virgin aggregates. Failure through the RCA was observed in the flexural strength specimens of concrete made with slag aggregate RCA. Similar observations were not made for the other two types of concrete. Due to the weak nature of slag aggregates and slag’s high surface porosity, the plane of weakness generally is through and not around coarse aggregates, tending to lead towards lower flexural strength and fracture energy [2]. Crushed Gravel Slag Aggregate RCA Limestone Aggregate RCA

40 35 30 25 20 15 10

25000 20000 15000 10000 5000

5 0

0

Aggregate type

Figure 4a. Compressive Strength

Crushed Gravel Slag Aggregate RCA Limestone Aggregate RCA Modulus of Rupture (MPa)

45

Elastic Modulus (MPa)

Compressive Strength (MPa)

Crushed Gravel Slag Aggregate RCA Limestone Aggregate RCA

7 6 5 4 3 2 1 0

Aggregate type

Figure 4b. Elastic Modulus

Aggregate type

Figure 4c. Modulus of Rupture

Resistance to chloride ion penetration was measured as per ASTM C 1202 at the age of 56-days (Figure 5). Concretes made with RCA exhibited very high permeability to chloride ions in comparison to crushed gravel concrete. Limestone aggregate RCA based concrete exhibited the highest permeability values and as per ASTM C 1202 falls in the range of high chloride ion penetrability whereas virgin aggregate concrete exhibited moderate resistance to chloride ion penetrability. The chloride permeability is dependent upon the quality of the mortar adhering the RCA. Residual mortar in RCA has previously been suggested as a conduit for water transport [2,4,5]. But in this study, such behavior is not observed and both concretes exhibit comparable quality of concrete for chloride penetrability.

32-Volumetric.qxd

9/19/2009

12:00 PM

Page 309

309

Volumetric stability of concrete using recycled concrete aggregates

Crushed Gravel Slag Aggregate RCA Limestone Aggregate RCA Average Coulombs

7000 6000 5000 4000 3000 2000 1000 0

Aggregate type

Figure 5. Chloride permeability using the Rapid Chloride Penetration Test at 56 days In Figure 6a, the length change of concrete prism samples placed in a 100% relative humidity (RH) moist curing room is shown. The first shrinkage measurement was taken 18 hours after addition of mix water. An expansion was observed in the concretes made with RCA in comparison to concrete made with crushed gravel. Higher strains and weight loss was observed in concrete made with slag aggregate RCA (Refer Figure 6a and 6b) in comparison with the other concretes. Crushed Gravel Slag Aggregate RCA Limestone Aggregate RCA

Crushed Gravel Slag Aggregate RCA Limestone Aggregate RCA 0

250 200

-0.01

Weight Gain (kg)

150

Strain (İ)

100 50 0 -50

-100 -150

-0.02 -0.03 -0.04 -0.05

-200 -250

-0.06 0

20 40 Age (days)

6a

60

0

20 40 Age (days)

60

6b

Figure 6a. Shrinkage strains in concretes cured at 100% RH and Figure 6b. Weight loss in concretes cured at 100% RH Figure 7a and 7b show the drying shrinkage and weight loss of concrete samples obtained based on shrinkage strains developed in sealed and unsealed condition placed at 50% RH and 23oC. Highest drying shrinkage strains and weight loss is observed in slag aggregate based RCA. Both RCA-based mixtures exhibit substantial weight loss until 20 days after which a slightly higher weight loss is observed for slag aggregate RCA mixture. At the end of 45 days both concretes exhibit comparable weight loss. Initially, the slag aggregate RCA concrete exhibits some amount of expansion followed by shrinkage. Limestone aggregate RCA-based concrete exhibit higher amounts of drying shrinkage at 28 days but thereafter,

32-Volumetric.qxd

9/19/2009

12:00 PM

310

Page 310

Yogini DESHPANDE, Jacob E. HILLER and Cory J. SHORKEY

slag aggregate RCA based concrete exhibits higher shrinkage. Lower shrinkage strain in slag aggregate RCA concrete at initial ages could be due to higher amount of water content in the aggregate. Thus, as the weight loss in the slag aggregate RCA concrete after 28 days increased the drying shrinkage strains also increased. Crushed Gravel Slag Aggregate RCA Limestone Aggregate RCA

Crushed Gravel Slag Aggregate RCA Limestone Aggregate RCA 0.15

100

0.125

0 -50

0.1 Weight Loss (kg)

Drying Shrinkage (ȝm/m)

50

-100 -150 -200 -250 -300 -350

0.075 0.05 0.025 0

-400 0

10

20

30

40

0

50

20

40

60

Age (days)

Age (days)

Figure 7a. Drying shrinkage at 50 % RH

Figure 7b. Weight loss at 50% RH

Figure 8 shows the strains developed in ring specimens cast to measure the cracking susceptibility of concrete. At the end of 50 days, none of the specimens had cracked. The highest shrinkage strains were developed in slag aggregate RCA concrete at the end of 50 days, whereas, lowest strains were developed in crushed gravel concrete. The RCA-based concretes exhibit similar phenomenon under restrained condition as those observed in free shrinkage specimens. Both RCA-based concretes exhibit comparable increase in strains from the time of addition of mixing water, but beyond 20 days the slag aggregate RCA, thereby demonstrating the volumetric stability issues of RCA-based concretes that should be addressed in pavement or structural design.

Crsuhed gravel

Slag Aggregate RCA

Limestone Aggregate RCA 20

Strain (ȝİ)

0 -20 -40 -60 -80 -100 0

20

40

60

Age (days)

Figure 8. Strain development in ring specimens

32-Volumetric.qxd

9/19/2009

12:00 PM

Page 311

Volumetric stability of concrete using recycled concrete aggregates

311

Studies have shown that if effective porosity is not accounted properly, shrinkage of concretes with slag aggregate RCA can be high [3,4] and can result in significant premature cracking of pavements structures [19]. In this paper, the effective porosity was correctly accounted by the measurement techniques of the helium pycnometer and the equivalent density analyzer. Therefore, even though water content in the slag aggregate RCA-based mixture was highest to account for high porosity of the slag aggregate, the overall shrinkage was still higher.

CONCLUSIONS AND FUTURE WORK 1. Concretes made with 100 % substitution of RCA are poor in quality in comparison with concrete made with the same w/c and virgin aggregates. RCA based concretes in general exhibited lower compressive and flexural strength, lower elastic modulus and high shrinkage strains. 2. Petrographical observations of slag aggregate RCA show a large amount of surface and internal porosity which may not be accounted by standard ASTM methods of calculating absorption capacity. 3. Recycled limestone and slag aggregate-based concrete mixtures exhibited comparable shrinkage strains in the first 25 days after casting. At the end of 45 days, the slag aggregate RCA concrete exhibited overall higher shrinkage strains. 4. Proper measurement of absorption capacity is an important aspect in the use of slag aggregate RCA as the issues of external and internal porosity become more complex than for natural virgin aggregates. REFERENCES 1. Mehta, PK. Greening of the concrete industry for sustainable development. Concrete International, 2002, vol. 24 (7), pp 23-28 2. Buch, N. Frabizzio, MA. and Hiller, JE. Impact of coarse aggregates on transverse crack performance in jointed concrete pavements, ACI Materials Journal, 2000, vol. (97) 3, MayJune, pp 325-332 3. Berndt, ML. Properties of sustainable concrete containing fly ash, slag and recycled concrete aggregate. Construction and Building Materials, 2009, vol. 23, pp 2602-2613 4. Tam, VWY. Gao, XF. Tam, CM and Chan, CH. New approaches in measuring water absorption of recycled aggregates. Construction and Building Materials, 2008, vol. 22, pp 364-369 5. Ayano, T. Fujii, A. and Sakata, K. Drying shrinkage of recycled concrete. Proceedings of Creep, Shrinkage and Durability Mechanics of Concrete and Concrete Structures. Edited by Tanabe. 2009, Japan, pp 805-810 6. Tam, VWY. Gao, XF. Tam, CM and Ng, KM. Physio-chemical reactions in recycle aggregate concrete. Journal of Hazardous Materials, 2009, vol.163, pp 823-828 7. Vitton, SJ. Lehman, MA. and Van Dam, TJ. Automated soil particle specific gravity analysis using bulk flow and helium pycnometery, Nondestructive and Automated Testing for Soil and Rock Properties, 1998, ASTM STP 1350, W.A. Marr and C.E. Fairhurst, Eds., American Society for Testing and Materials 8. Michigan Department of Transportation, Standard Specification for Construction, 2008 9. ASTM C 127, Standard Test Method for Density, Relative Density (Specific Gravity), and Absorption of Coarse Aggregate, Annual Book of ASTM Standards, American Society for Testing and Materials, vol. 04.02, 2008

32-Volumetric.qxd

312

9/19/2009

12:00 PM

Page 312

Yogini DESHPANDE, Jacob E. HILLER and Cory J. SHORKEY

10. ASTM C 128, Standard Test Method for Density, Relative Density (Specific Gravity), and Absorption of Fine Aggregate, Annual Book of ASTM Standards, American Society for Testing and Materials, vol. 04.02, 2008 11.ASTM C 260, Standard Specification for Air-Entraining Admixtures for Concrete, Annual Book of ASTM Standards, American Society for Testing and Materials, vol. 04.02, 2008 12. ASTM C 39, Standard Test Method for Compressive Strength of Cylindrical Concrete Specimens, Annual Book of ASTM Standards, American Society for Testing and Materials, vol. 04.02, 2008 13. ASTM C 469, Standard Test Method for Static Modulus of Elasticity and Poisson's Ratio of Concrete in Compression, Annual Book of ASTM Standards, American Society for Testing and Materials, vol. 04.02, 2008 14. AASHTO PP 34, Estimating the Cracking Tendency of Concrete, American Association of State Highway and Transportation Officials, 2005 15.ASTM C 157, Standard Test Method for Length Change of Hardened Hydraulic-Cement Mortar and Concrete, Annual Book of ASTM Standards, American Society for Testing and Materials, vol. 04.02, 2008 16.ASTM C 1202, Standard Test Method for Electrical Indication of Concrete's Ability to Resist Chloride Ion Penetration, Annual Book of ASTM Standards, American Society for Testing and Materials, vol. 04.02, 2008 17. Domingo-Cabo, A. Lazaro, C. Lopez-Gayarre, F. Serrano-Lopez, MA. Serna, P. and Castano-Tabares, JO. Creep and shrinkage of recycled aggregate concrete. Construction and Building Materials, 2009, vol. 23, pp 2545-2553 18. Kakizaki, M. Harada, M. Soshiroda, T. Kubota, S. and Ikeda, YK. Strength and elastic modulus of recycled concrete aggregate. Demolition and Reuse of Concrete and Masonry, 1988, Chapman and Hall, London, pp 565-574 19.Vancura, M. Tompkins, D. and Khazanovich, L. Reappraisal of Recycled Concrete Aggregate as Coarse Aggregate in Concretes for Rigid Pavements. Pre-print CD-ROM Transportation Research Board Annual Meeting, 2009, National Research Council, Washington, D.C., USA

33-Damage.qxd

9/19/2009

12:00 PM

Page 313

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

DAMAGE ASSESSMENT IN SECTIONS FOR DURABILITY PURPOSES: TWO ARGUMENTS NOT TO OPT FOR AUTOMATION Piet STROEVEN, Huan HE and Martijn STROEVEN Faculty of Civil Engineering and Geosciences, Delft University of Technology Stevinweg 1, 2628 CN Delft, the Netherlands; e-mail: [email protected]

ABSTRACT Prediction of remaining service life and form maintenance schemes of an infrastructural facility can be pursued on the basis of quantitative image analysis of section images prepared from core samples. It is discussed in this paper that economic and reliability arguments plead for manually executing simple intersection counting in analogue images. Digitization of images leads generally to overestimation of total crack extent and to biases in degree and direction of orientation. This paper advises to model damage as a partially linear-planar structure. This allows elegantly visualizing differences in roses of the number of intersections (or of intersection density) between analogue and the digitized images of the same crack pattern. Moreover, the paper demonstrates that this approach points towards a possible methodological way out of the problems faced when opting for automation. Adjustment of traditional four-connexity digitization with prevailing crack orientation direction should be considered crucial in this methodology.

Keywords Concrete, digitization biases, quantitative image analysis, sweeping test line

INTRODUCTION Image analysis can provide quantitative information on the damage structure in sections of cores of a concrete structure that has undergone some degradation during an elapsed period of service. In a multi-stage sampling scheme researchers can pursue correlating extension and degree and prevailing orientation of damage in field images to characteristics of the concrete structure and to environmental conditions. This could serve giving predictions for the remaining service life period and for maintenance schemes to guarantee durability of the infrastructural facility. For such a multi-stage sampling scheme holds the general saying: “Do more less well!” [1]; this implies increasing numbers in the first sampling stage (i.e., of core samples) and reducing numbers in the last stage (i.e., of intersections in a directed secants approach to field images). Hence, this constitutes an economic argument not to opt for automation of the quantitative image analysis (QIA) approach in the multi-stage sampling scheme. The simple point counting operation of limited extent is readily executable in a manual way. Moreover, by taking this decision it is avoided obtaining biased results because of digitization of the field images. This phenomenon has been recognized earlier [2,3], but is somewhat neglected in concrete technology [4]. Focus in this paper will not be on giving an outline of all aspects of QIA, but merely on the geometrical statistical (stereological) background of QIA by means of the sweeping test line approach. This will allow demonstrating that biases can be expected in total extent of damage as well as in the

33-Damage.qxd

9/19/2009

12:00 PM

Page 314

314

Piet STROEVEN, Huan HE and Martijn STROEVEN

prevailing direction and degree (strength) of crack orientation [3,6]. Apart from a mathematical treatment, the sphere model introduced by Underwood for roses of intersections [7] will be employed to visualize the differences between manual approaches to analogue images and automated approaches to digitized but otherwise similar images. Together with the aforementioned multi-stage sampling argument, this points toward opting for the manual approach whereby economy and reliability go hand in hand. Yet, a way out of the problems will be indicated; the methodological and modeling conditions will be outlined that should be fulfilled when still opting for unbiased information in an automation approach.

STEREOLOGICAL FRAMEWORK FOR QIA Conventional approach Sections of concrete specimens in laboratory studies or of cores drawn from the engineering structure can be used to get information on damage. When contrast of cracks and cementitious background is improved, images can be submitted to the traditional sweeping test line approach [7-10]. This involves counting the number of intersections P between cracks and test lines. The result is normalized when divided by the total test line length, L, yielding intersection density PL. Classical stereological theory offers the very simple relationships [7]

S

PL and SV 2 PL (1) 2 which allow estimating two-dimensional (2D) crack density in the image plane (L/A=LA) and three-dimensional (3D) crack density in space (S/V=SV) on the basis of intersection density measurements. However, the tricky character is due to the averaging operations expressed by the horizontal bar on the measurements. When confronted with non-isotropic structures, as is generally the case with damage in concrete, this averaging in 3D would require “random sampling”. This is a highly unpractical and expensive operation when dealing with cores or prismatic concrete specimens. So, this requirement should be avoided, though not neglected! The averaging in 2D is simple: the sweeping test line system is applied in a number of different directions (systematically or randomly selected), observations normalized and averaged. However, the real world (the realcrete) is three-dimensional; any interpretation of LA into 3D would be sheer speculation, and can therefore not be considered the best option. A step forward would offer an ortrip (=orthogonal tripod) sampling scheme, which involves making sections in three orthogonal directions. Still quite some work. Moreover, only reasonable information can be expected when the prevailing orientation of damage will coincide with one of the Cartesian coordinate directions. This would be possible in laboratory research of specimens subjected to simple loading schemes. However, fulfillment of this condition for the assessment of damage by cores drawn from the engineering structure will be a more complicated. When some reasonable modeling assumption can be made as to the damage structure, the huge problems could be released somewhat. When confronted with locations in engineering structures where high uniaxial compressive (or tensile) stresses are expected – a situation likely to be encountered in laboratory research – an axis of symmetry in the damage structure could be assumed. This would dramatically simplify the sampling scheme and reduce the efforts required. Under the given conditions, QIA can be restricted to so-called vertical sections; i.e., sections parallel to the axis of symmetry. Nevertheless, in the vertical section the application of the sweeping test line should be sine-weighed when pursuing 3D information. This is to account for the diminishing length of circles (as sinș) upon approach of zenith in the unit sphere model [11], whereby ș defines the angle enclosed by the vertical axis of the unit sphere and the radius to the circle. This is illustrated in Fig. 1. This sineLA

33-Damage.qxd

9/19/2009

12:00 PM

Page 315

Damage assessment in sections for durability purposes: two arguments not to opt for automation

315

weighed line scanning is mostly achieved by using cycloids instead of straight grid lines [7,9], which renders possible directly estimating SV on the basis of eq. (1) at the right, so can be considered the optimum methodological approach.

Fig. 1. Unit sphere: infinitely small isotropic uniformly random (IUR) distributed crack elements, 'S - hypothetically transferred from the real material maintaining spatial orientation - will uniformly random cover the surface of this sphere. Relative frequency of crack elements enclosing the same angle with the vertical axis declines from “equator” to “zenith” as sin ș The discussion on biases introduced by digitization in an automated set up will therefore be based on the expressions in eq. (1). However, methodology can be further improved when modeling the damage structure as a partially linear-planar structure (of partly interconnected) surfaces scattered in 3D space. This is the so-called Stroeven-concept (derived from a suggestion by Saltikov) [6]. Partially linear-planar damage model The crack surfaces are hypothetically subdivided in infinitely small flat elements (as pursued in unit sphere model of Fig. 1). The basic assumption is that these flat elements can be allotted to three portions: a 3D isotropic uniformly random (IUR), a 2D planar random and a 1D lineal random one. The crack elements in the 3D portion are randomly distributed as to location and uniformly random as to orientation. Crack elements are also randomly located in 2D and 1D portions. In the 2D portion the only restriction is that all elements are parallel to an orienttation plane, whereas in the 1D portion they are parallel to an orientation axis. The orientation axis can develop in a situation whereby relatively high compressive stresses prevail in the concrete; the orientation plane does so when relatively high tensile stresses occur. This set up requires assessment of the three portions in SV, i.e., SV3, SV2 and SV1, which are the specific surface areas in the 3D, the 2D and the 1D portions, respectively. SV is the sum of these three portions. Once determined by QIA approach, the orientation distribution is also known.

DIGITIZATION-INDUCED BIASES

Conventional four-connexity digitization transforms a smooth crack contour in the section image into a discrete line composed of orthogonal “sticks” that are oriented in orthogonal directions, as shown in Fig. 2. Higher order digitization techniques lead to digitized contours

33-Damage.qxd

316

9/19/2009

12:00 PM

Page 316

Piet STROEVEN, Huan HE and Martijn STROEVEN

that can approximate the analogue crack more closely. Obviously, digitization will influence intersection counts except those coinciding with the orthogonal directions of the sticks.

Fig. 2. Four-connexity digitized crack This is as an example illustrated in Fig. 3 for a crack pattern in a vertical section of a partially linear damage structure. So, the rose of intersections (or of intersection densities) is biased. This was earlier demonstrated for the case of shrinkage cracking at a free surface, which was also modeled as a partially linear structure [12,13]. Moreover, the averaging operations required for assessment of 2D or 3D damage characteristics lead to biases, so that misleading information on total crack length or total crack surface area will always be resulting from conventional QIA approach to digitized image. Fig. 3 presents the roses of intersection densities of the assumed damage components in the vertical section. The 3D and 1D portions in space lead to a 2D random (indicated by index “r”) and an oriented portion (indicated by index “o”) in the image plane. Hence, PL (T ) PLr PLomax sin T governs the orientation distribution in the analogue images, with index “max” for the largest value of the intersection density of the oriented portion. Total crack length is readily obtained for this concept by application of LA S PL 2 . This yield for:

Fig. 3. Roses of intersection densities for analogue (left) and digitized image of the same section crack pattern; damage is composed of 1D and 3D portions in space. In the section plane, the oriented crack portion leads to two circles through the origin, the random one to a dashed circle around the origin in the analogue image and four dashed circles through the origin in the digitized image (yielding upon addition the dashed line of the flower-like curve). The outer curves represent the addition of oriented and “random” portions.

33-Damage.qxd

9/19/2009

12:00 PM

Page 317

Damage assessment in sections for durability purposes: two arguments not to opt for automation

S 2

LA

Analogue image:

³ P

S

0

Lr

PLo max sin T dT

2

³

S 2

0

S 2

Digitized image:

LA

S

³ P 0

Lr

S 2

dT

PLr PLo max

sin T PLr cos T PLo max sin T dT

2

³

S 2

0

2 PLr PLo max

dT

317

(2)

(3)

Both are based on the assumption of the Stroeven-concept (here of the partially linear system), which has been proven an accurate way of describing orientation distribution on the basis of two orthogonal observations [14]. Similarly, for the crack surface area we have

Analogue: SV (T , E )

2 PL

S 2

S 2

0

0

³ ³ P P 2 S S ³ ³ Lr

0

Digitized: SV (T , E )

2 PL

Lo max

2

S 2

S 2

0

0

³ ³ P 2

Lr

0

2

sin T cos T dT dE

cos T dT dE

2 PLr PLo max

(4)

sin T PLr cos T PLo max sin T cos T dT dE S 2

S 2

0

0

³ ³

cos T dT dE

S

( 1) PLr PLo max 2

(5)

Over-estimation of total crack length and total crack surface area when based on digitized images is obvious (in both cases by slightly less than 30% in this specific case). It should be understood, however, that this is still a simplified concept. In [4,14] the orientation distribution is considered of elongated particles of a tillite. Since the surfaces of the particles were the subject of interest, the step towards crack surface areas is not a fundamental one, hence direct interpretation is possible. The prevailing direction of structural orientation (of the surfaces) was the prime research parameter. Digitization direction could not be adjusted to that of structural orientation because of the weakness of the signal; a situation also pertinent to crack analysis even at a modest level of optical magnification [15-17]. This introduces complications, and leads to additional biases. The presented example in Fig. 3 is also not dealing with the more general case of a partially linear-planar system, which is composed of three instead of two portions. The orientation line of the 1D portion could coincide with the 2D orientation plane where compressive and tensile stresses are orthogonally oriented. They could also be perpendicular to each other in cyclic loading cases. Mostly, two properly selected orthogonal sections will suffice in such cases, because providing three independent observations on the three damage portions. We will come back to this conclusion later.

WAY OUT OF THE DIGITIZATION PROBLEMS

The best solution is not to opt for automation. However, when an automatic approach to QIA is pursued, this last sampling stage should be properly designed. Estimation of the preferred direction of orientation in the damage structure is thereby crucial. This would allow fourconnexity digitization adjusted to the damage orientation direction. Next, the assumption of

33-Damage.qxd

9/19/2009

12:00 PM

Page 318

318

Piet STROEVEN, Huan HE and Martijn STROEVEN

the actual damage structure conforming to the partially linear-planar system should be adopted. This brings the researcher in the situation where ortrip sampling can be executed; whether two or three orthogonal sections should be prepared depends on whether further assumptions could be made as to the damage structure. In [6], the simplest but highly practical cases of prevailing uniaxial compression and tension are elaborated. The efforts could be reduced under such conditions to a sampling design consisting of a single vertical section. Of course, more sections could be possible to reduce scatter and improve the reliability of the average. However, the aforementioned adjustment of the digitization direction to the prevailing orientation direction of the damage structure is the most crucial one for the automatic QIA of the section image(s). In a more complicated case of low-cyclic compression loading-induced damage structure, one could be confronted with a lineal crack portion due to the compressive loading and with a planar portion due to stress release under relatively large amplitudes, as we have witnessed in research [15]. The planar (2D) portion’s orientation plane coincides with, say, {x,y}-plane, is thus perpendicular to the 1D orientation axis that runs in the z-direction. Obviously, this system is axially symmetric around the orientation axis, so that a vertical section and a “horizontal” one should provide the independent and unbiased information when sampled by a line grid in the successive Cartesian coordinate directions. The crack pattern in the {x,y}-plane is 2D random, hence providing for just a single observation. Unfortunately, the expression in crack portions in the vertical section with the grid in the direction of the orientation plane is similar, so we end up with a system of two equations with three unknown parameters. This can only be solved by adding another independent observation in the vertical section in a direction not corresponding to one of the coordinate directions. Unfortunately, this will be biased due to the digitization effect, as we have seen. So, this is a problem with a dead end when opting for automation. The situation is different under biaxial compression-tension (compression in z-direction, tension in y-direction, see Fig. 4). Here the axial symmetry is lacking. The vertical {x,z}plane reveals the 3D and 1D crack portions in the z-direction, whereas the orthogonally oriented {y,z}-plane will show the 3D, 2D and 1D crack portions in z-direction. These two orthogonal sections suffice for obtaining an unbiased solution. This situation is displayed in Fig. 4. The observations with a grid perpendicular to the orientation axis yield PL ( y, z )

PL ( x, z )

1 SV 3 2

(6)

The other two observations in the sections are PL A ( x, z )

PL A ( y, z )

1 2 SV 3 SV 1 SV 2 2 S 1 2 SV 3 SV 1 2 S

(7)

(8)

This can provide a proper solution for the three crack portions as to total crack surface area in space, as well as for the crack orientation distribution, when based on the Stroeven-concept of the partially linear-planar system (as demonstrated earlier). Alternatively, a two-dimensional solution in crack length portions could be formulated in a similar way.

33-Damage.qxd

9/19/2009

12:00 PM

Page 319

Damage assessment in sections for durability purposes: two arguments not to opt for automation

{x,z}-plane

319

{y,z}-plane

Fig. 4. Ortrip sampling design adjusted to the Cartesian coordinate directions. An orthogonal set of vertical sections are (four-connexity) digitized in the in-plane coordinate directions. Observations are made with grids successively in the two relevant coordinate directions; this yields four observations that are not affected by digitization. Since two observations are identically related to crack portions, a system of three equations with three unknown parameters is resulting.

DISCUSSION

The strength of the signal is expressed by the degree of orientation, Ȧ. This is defined as the ratio of “parasitic portions” over total value. Hence, SV 1 SV 2 SV

Z

(9)

Alternatively, also a 2D approach can be followed, employing the LA-portions. As practical examples we can take the partially linear and the partially planar concepts. These are suitable approximations in prevailing direct tension or direct compression, respectively. From [5] the following simple expressions are taken, which also result from eqs. (6) to (8) upon substitution of SV 1 0 (direct tension) and SV 2 0 (direct compression). tension:

compression:

PL PL A

SV 2 SV 1

S 2

P

LA

PL

and

and

SV SV

PL PL A

S 2

S· § PL A ¨ 2 ¸ PL 2¹ ©

(10) (11)

Hence, the strength of the parasitic signal is in the two cases defined by tension:

Z

PL PL A PL PL A

and

Z

PL A PL §4 · PL A ¨ 1¸ PL ©S ¹

(11)

33-Damage.qxd

320

9/19/2009

12:00 PM

Page 320

Piet STROEVEN, Huan HE and Martijn STROEVEN

This is the same for a digitized and an analogue image provided the aforementioned “way out”-conditions are realized. Otherwise, as derived in [3], biases of complicated nature arise due to digitization. The strength of the signal is optical magnification-dependent (due to fractal nature of cracking), i.e., strength declines at higher magnification. Generally, a positive value is found for Ȧ. When a negative value arises due to changes in loading conditions, the assumed crack mechanism is false. Hence, Ȧ can function as a detection parameter. In lowcycle compression fatigue research, such a modification occurred at increased load amplitude, witnessing the occurrence of tensile cracking due to load release in compression [15]. This caused a drop by one order of magnitude in the number of load cycles to fracture. The partially linear damage system at smaller load amplitudes had to be modified into a partially linear-planar one at higher load amplitudes!

CONCLUSIONS

Concrete core samples drawn from the engineering structure or laboratory specimens contain large numbers of cracks due to various types of loading. They can be visualized in sections (by improving contrast). Insight into the local stress state helps to design an economic sampling scheme. Moreover, when opting for automation, the only way out of obtaining biased information due to (conventional four-connexity) digitization is by adjusting digitization direction to prevailing orientation of damage in the section plane. By modeling damage as a partially linear planar structure the economy of the experiment can be improved, because random sampling can be replaced by ortrip sampling and in many cases even vertical sections may suffice. However, major profit is reliability. All interesting damage parameters, like total crack length or surface area, orientation distribution and degree of orientation, can be expressed in orthogonal intersection counts by sweeping test line system. These values are in the direction of digitization and as a consequence unbiased.

REFERENCES

1. Gundersen, H.G., Osterby, R., Optimizing sampling efficiency of stereological studies in biology, or: “Do more less well”. J. Microsc, 121, 1981, 65-73 2. Chaix, J.M., Grillon, F., On the rose of direction measurements on the discrete grid of an automatic image analyzer. J. Microsc., 184, 1996, 208-213 3. Stroeven, P., Stroeven, A.P., Dalhuisen, D.H., Image analysis of ‘natural’ concrete samples by automated and manual procedures. Cem. Concr. Comp., 23, 2001, 227-236 4. Bisschop J., Drying shrinkage microcracking in cement-based materials. PhD Thesis. Delft, Delft University Press, Delft 2002 5. Stroeven, P., He, H., Computers and concrete: not always a good marriage. Proc. Int. Conference on Computational Technologies in Concrete Structures. Jeju 24-27 May, 2009 (to be published) 6. Underwood E.E., Quantitative Stereology, Addison-Wesley, Reading (MA), 1970 7. Stroeven P., Geometric probability approach to the examination of microcracking in plain concrete. J. Mat. Sc. 14, 1979, 1141-1151 8. Stroeven P. Some aspects of the micromechanics of concrete. PhD Thesis, Delft University of Technology, Delft, 1973 9. Saltikov, S.A., Stereometric metallography (Second edition), Metallurgizdat, Moscow, 1945

33-Damage.qxd

9/19/2009

12:00 PM

Page 321

Damage assessment in sections for durability purposes: two arguments not to opt for automation

321

10. Stroeven, P., Hu, J., Stereology: Historical perspective and applicability to concrete technology. Mat. Struct., 39, 2005, 127-135 11. Stroeven, P., Hu, J., Gradient structures in cementitious materials, Cem. Concr. Comp., 29, 2007, 313-323 12. Bisschop J, van Mier JGM., Effect of aggregates on drying shrinkage microcracking in cement-based composites. Mat. Struct., 35, 2002, 453-461 13. Stroeven, P., Use of radiography-image analysis for steel fiber reinforced concrete. Closure of Discussion. In: Testing and Test Methods of Fiber Cement Composites”, R.N. Swamy, ed., Constr. Press, Lancaster, 1978, pp. 308-311. 14. Stroeven, A.P., Stroeven, P., van der Meer, J.J.M., Microfabric analysis by manual and automated stereological procedures: a methodological approach to Antarctic tillite. Sedimentology, 52, 2005, 219-233 15. Reinhardt, H.W., Stroeven, P., den Uijl, J.A., Kooistra, T.R., Vrencken, J.H.A.M., Einfluss von Schwingbreite, Belastungshöhe und Frequenz auf die Schwingfestigkeit von Beton bei niedrigen Bruchlastwechselzahlen. Betonw. & Fertigteil-Techn., 44, 1978, 498503 16. Stroeven, P., Some observations on microcracking in concrete subjected to various loading regimes. Engr. Fract. Mech., 35, 1990, 775-782 17. Stroeven P., Damage evolution in concrete; application of stereology to quantitative image analysis and modelling. In: Proc. Int. Symp. “Advanced Materials for Future Industries: Needs and Seeds”, I. Kimpara, K. Kageyama and Y. Kagawa eds. Chiba 11-14 Dec. 1991, SAMPE, Tokyo, 1991, 1436-1443

34-The effects.qxd

9/19/2009

12:22 PM

Page 323

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

THE EFFECTS OF LASER CLEANING PROCESS ON GEOMETRICAL MICROSTRUCTURE OF CEMENTITIOUS COMPOSITES Agnieszka J. KLEMM1, Poologanathan SANJEEVAN1, Piotr KLEMM2 School of Built and Natural Environment, Glasgow Caledonian University, 70 Cowcaddens Rd, Glasgow, G4 0BA, UK; e-mail: [email protected] 2 Technical University of Lodz, Poland

1

ABSTRACT The paper presents part of the larger study on laser cleaning process. A special emphasis is placed on the modification of geometrical microstructure of cementitious composites resulting from cleaning process. A wide range of samples characterised with different internal microstructures, surface roughness, and moisture content were subjected to laser cleaning process, and a subsequent assessment. Removal of mortar, cracks and glassy patches were the characteristic features of all laser-cleaned areas. Systematic analysis of surface modifications resulting from laser cleaning confirmed a strong relationship between initial roughness of surfaces and their end conditions. An increase in initial surface roughness leads to more pronounced alterations in roughness and reduced tendency towards crack formation.

Keywords Cementitious materials, laser cleaning, surface modification

INTRODUCTION Traditional cleaning methods for cementitious materials, such as air-abrasive and steam cleaning can lead to severe damage of surfaces. This unavoidable damage is caused mainly by the direct contact of cleaning agents with surface and the difficulty in identifying borderline between contaminating material and substrate. The difficulties during the cleaning process increase with the increase of thickness of the contaminating layer and the lack of cohesion of the surface deposit. Base materials lying below may be altered so much that the removal of the unwanted deposit may be extremely difficult even with careful mechanical or chemical intervention. Laser cleaning allows overcoming most of such problems, Cooper [1]. The relationship between laser cleaning processes and substrate parameters is a twoway relationship. While the influence of microstructural features of surfaces on laser cleaning process has been previously researched, Matsui, et al [2], Sanjeevan, et al [3],[4], McStay, et al [5], the effect of laser radiation on characteristics of modified cementitious surfaces has been slightly overlooked. Laser cleaning is often described as a self-limiting process. The mechanism of this process is based on the difference between the monochromatic reflection (absorption) of photons by the contaminator and the background, Liu et al [6]. The great variation in absorptivity of highly developed surfaces of cementitious materials results in substantial differences in their responses to laser irradiation. Even though

34-The effects.qxd

9/19/2009

324

12:22 PM

Page 324

Agnieszka J. KLEMM, Poologanathan SANJEEVAN and Piotr KLEMM

lasers can be successfully used to remove paint from mortar surfaces, there are always some residual surface alterations, associated with the removal of mortar itself, formation of cracks and glazing (melted mortar). Despite the fact that these effects are generally microscopic and only visible at high magnification, their importance should not be underestimated, especially on national heritage buildings as they may expose surfaces to further accelerated environmental deterioration, RoĪniakowski, et al [7].

EXPERIMENTAL DETAILS All tested samples differ in their microstructure (low porosity - LP and high porosity - HP), surface roughness (A, B and C) and moisture content (WET and DRY). WET samples were immersed in water for 24 hours to become fully saturated before cleaning while DRY samples were in equilibrium with surrounding air in a laboratory (temp. 23 ± 0.5 ºC and RH 50 ± 5%). The mortar specimens had the following composition: cement to sand ratio 1:1 and water/cement (w/c) ratio 0.4 with a different air-entraining admixture (AEA). The admixture content, selected mechanical properties, surface roughness, and moisture content of the mortar samples are shown in Table 1, below. Surface roughness was measured by a stylus device and represented by average surface roughness (Ra). Table 1. Compositions, mechanical properties, surface roughness and moisture content of the mortar samples.

Mix

Conplast AE380 (l/100 kg OPC)

Surface roughness (μm)

Compressive strength (N/mm2)

Flexural Strength (N/mm2)

Bulk density (g/ml)

Porosity (İHg) %

DRY

WET

Moisture content

LP-A

0

2.28-2.49

79.4

8.3

2.16

11.9

2.3

8.2

LP-B

0

7.70-8.49

79.4

8.3

2.16

11.9

2.3

8.2

LP-C

0

15.58-17.89

79.4

8.3

2.16

11.9

2.3

8.2

HP-A HP-B HP-C

1.3 1.3 1.3

2.28-2.49 7.70-8.49 15.58-17.89

57.5 57.5 57.5

7.4 7.4 7.4

1.8 1.8 1.8

26.4 26.4 26.4

3.2 3.2 3.2

10.7 10.7 10.7

The cleaning of spray paint from mortar surfaces was done with application of Nd:YAG laser with the following characteristics: wavelength: 1.06μm, energy: 500 mJ, pulse duration: 10ns and pulse repetition rate: 1Hz.

34-The effects.qxd

9/19/2009

12:22 PM

Page 325

The effects of laser cleaning process on geometrical microstructure of cementitious composites

Fig.1.

325

Mortar sample after cleaning process.

Each sample has been subjected to laser radiation of the same laser fluence F = 3.06 J/cm2. Increasing number of pulses has been applied to 11 areas as shown in Fig. 1. RESULTS AND DISCUSSION Crack formation Comprehensive SEM analysis of surfaces revealed a high number of micro cracks and glazing as shown in Fig. 2. Cracks were quantified by their density, defined as the length of crack per area. BSE image analyses have been used to calculate crack densities. The actual lengths of cracks were approximated by straight lines. The white lines on Fig. 2, consisting of short straight lines represent actual cracks. Glazing Crack

Fig. 2.

BSE image of HP-C(WET) after the application of 26 laser pulses

Any interaction of the laser with mortar leads to the production of a significant amount of heat. The heat produced dissipates through the formations of cracks, glazing, removal of paint or mortar and other losses. Mortar surfaces can develop cracks in two different ways. Cracks may form, when thermal stresses on the cementitious material due to application of laser, exceed its tensile strength or when the tensile stresses due to expansion of water or air inside the pores, exceed the tensile strength of material (Fig. 3), Hertz [8].

34-The effects.qxd

9/19/2009

326

12:22 PM

Page 326

Agnieszka J. KLEMM, Poologanathan SANJEEVAN and Piotr KLEMM

Laser Beam

Laser Beam

T

T

Crack

P

T - Thermal stress

P - Outward pressure due to expansion of water

Mechanism 1

Mechanism 2

Fig. 3. The mechanisms of formation of cracks due to laser cleaning. Since, thermal conductivity of a wet mortar is higher than a dry one; the temperature difference between two points is greater for dry samples, Khan [9]. Thus, the development of thermal stress, resulting from laser cleaning of dry samples, is more pronounced. The wetness of the samples impedes the crack formation in the first mechanism. On the other hand, since development of pressure in pores depends on the amount of water inside them, wetness of mortar facilitates crack formation in the second mechanism. Figs. 4 to 6 show the relationship between number of pulses applied to the surface and crack density in the laser affected area.

Fig. 4. Relationship between number of pulses applied and crack density; Ra = 2.28-2.49 μm.

34-The effects.qxd

9/19/2009

12:22 PM

Page 327

The effects of laser cleaning process on geometrical microstructure of cementitious composites

327

Fig. 5. Relationship between number of pulses applied and crack density; Ra =7.70-8.49 μm.

Fig. 6. Relationship between number of pulses applied and crack density; Ra=15.58-17.89μm As the number of laser pulses increased, the crack density increased up to a certain level, followed by their disappearance, which was caused by the removal of mortar. Further application of pulses lead to the formation of new cracks. Regardless of the surface roughness and moisture content of mortar, cracks always form in more porous samples, mainly due to the presence of pit holes. On the other hand, no crack formation was observed when the laser was applied to low porosity samples (LP-C(WET), LP-C(DRY) and LP-B(DRY)). Pit holes have better absorption characteristics than the other parts of mortar. A high temperature

34-The effects.qxd

328

9/19/2009

12:22 PM

Page 328

Agnieszka J. KLEMM, Poologanathan SANJEEVAN and Piotr KLEMM

gradient due to the geometrical shape of holes leads to a higher concentration of cracks around them. An example is shown in Fig. 7 below.

Fig. 7.

BSE image of HP-A(WET) (1000x) and LP-A(WET) (150x) after the application of 31 laser pulses.

Fig. 8 shows the effect of moisture content and porosity of mortar on the crack formation on surfaces with low surface roughness, between 2.28-2.49 μm (A).

Fig. 8. The effect of surface moisture content and porosity of mortar on the crack formation; Ra = 2.28-2.49 μm; F = 3.06 J/cm2. The formation of cracks due to laser cleaning was found to be more pronounced in denser samples. For example, crack density of LP-A(WET) after the application of 31 laser pulses was 2.7 mm/mm2 and the crack density of HP-A(WET) was only 1.17 mm/mm2. Moreover, presence of water in mortar facilitated the crack formation in low porosity samples. Highly porous samples experienced the opposite. This leads to the conclusion that low porosity is responsible for the crack formation, mainly due to build up of internal pressures caused by expansion of water/air inside the pores. However, in more porous samples, crack formation is mainly caused by thermal stress on the mortar surface. Fig. 7 above shows the spalling caused by laser cleaning of a less porous wet mortar sample. Since spalling of mortar results from built up of internal pressure due to expansion of water inside the pores, the above surmise is correct.

34-The effects.qxd

9/19/2009

12:22 PM

Page 329

The effects of laser cleaning process on geometrical microstructure of cementitious composites

329

Fig. 9 shows the effect of moisture content and porosity on the crack formation of mortars with surface roughness 7.70-8.49 μm (B) and 15.58-17.89 μm (C). In the first case (top), similarly to case (A), sample wetness facilitates crack formation in low porosity mortars and impedes formation in high porosity ones. The crack formation in HP-B(DRY) is more pronounced than in LP-B(DRY) mainly because of the presence of pit holes in the first one.

Fig. 9. The effect of surface moisture content and porosity of mortar on the crack formation; Ra = 7.70-8.49 μm and Ra = 15.58-17.89 μm; F = 3.06 J/cm2. No crack formation was observed in denser/rougher surfaces (bottom). The effect of surface roughness of mortar on the crack formation due to laser cleaning is shown in Fig. 10. Crack density on smooth surfaces was higher than on rough ones, probably due to the higher energy dissipation through mortar removal rather than crack formation. Since crack density depends on the amount of energy used to produce cracks, the number of cracks developed on the smooth surface was higher than the rough one. The width of surface cracks varied from 0 to 7μm. The width of cracks formed by thermal expansion (mechanism 1) is smaller than for those formed by internal pressures (mechanism 2). For example, the width of a crack in sample LP-A(WET) was higher than HP-A(WET).

34-The effects.qxd

330

9/19/2009

12:22 PM

Page 330

Agnieszka J. KLEMM, Poologanathan SANJEEVAN and Piotr KLEMM

Fig. 10. The effect of surface roughness on the crack formation during laser cleaning.

Surface roughness Surface roughness varies greatly with an increasing number of laser pulses applied. Fig. 11 presents results of analysis of smooth samples (A) (initial roughness, before cleaning of 2.282.49μm). The relationship can be divided into two zones - the higher rate of increase of surface roughness in Zone 1 was probably caused by the removal of paint, while the lower increment in Zone 2 could be attributed to the removal of mortar. Mortars of initial roughness 7.70-8.49μm (samples B) showed a very similar relationship.

Fig. 11. Variation of surface roughness due to application of laser pulses; samples A.

34-The effects.qxd

9/19/2009

12:22 PM

Page 331

The effects of laser cleaning process on geometrical microstructure of cementitious composites

331

However, all samples C, with Ra = 15.58-17.89 μm, behaved differently (Fig. 12). Here the relationship could be divided into three/four zones, depending on the samples’ porosity: Zone 1: increment in surface roughness at lower rate; Zone 2: increment in surface roughness at higher rate than in Zone 1; Zone 3: drop in surface roughness; Zone 4: drop in surface roughness at smaller rate than in Zone 3. Four zones were observed only in highly porous samples. In low porosity samples only three zones could be distinguished. The observed drop in the surface roughness indicates high possibility of laser damage on rough surfaces.

Fig. 12. Variation of surface roughness due to the application of laser pulses; samples C. Variations of surface roughness with increasing number of pulses applied can also be influenced by near-surface porosity and moisture content of mortars (Figs. 13-15). Data presented in Fig. 13 relates to samples with initial surface roughness between 2.28-2.49 μm (A) and 7.70-8.49 μm (B). The effect of moisture content and porosity was noticeable in both cases although more prominent after different number of pulses (4 to 6 laser pulses for sample A and 17 – 21 pulses for B). Generally the increment of the surface roughness due to laser cleaning was higher in samples of higher porosity. Due to the lower strength of highly porous samples, the removal of mortar resulting from laser cleaning was more pronounced. Moreover, the absence of water in mortar appeared to promote the increment of surface roughness.

34-The effects.qxd

332

9/19/2009

12:22 PM

Page 332

Agnieszka J. KLEMM, Poologanathan SANJEEVAN and Piotr KLEMM

Fig. 13. Effects of porosity and moisture content on the surface roughness as a function of number of pulsed applied; Ra = 2.28-2.49μm

Fig. 14. Effects of porosity and moisture content on the surface roughness as a function of number of pulsed applied; Ra = 7.70-8.49 μm. The effects of moisture content and porosity on the variation of surface roughness in mortar samples with the initial surface roughness between 15.58-17.89μm (C) were not clearly defined (Fig. 15).

34-The effects.qxd

9/19/2009

12:22 PM

Page 333

333

The effects of laser cleaning process on geometrical microstructure of cementitious composites

. Fig. 15. Effects of porosity and moisture content on the surface roughness as a function of number of applied pulses; Ra = 15.58-17.89 μm; F = 3.06 J/cm2. Nevertheless, any changes in surface roughness were more pronounced in highly porous samples. This may lead to the conclusion that the removal of mortar caused by laser cleaning in highly porous samples is much more substantial and therefore may cause more severe damage to the substrate. Surface roughness and crack appearance In the case of rougher surfaces with low porosity, no cracks were observed. Cracks were not observed also on dry/low porous mortars with a moderate roughness. However, all low porous samples with smoother surfaces developed cracks (both wet and dry). This demonstrated that the energy dissipation through crack formation becomes less predominant with the increase of initial surface roughness. Moreover, due to the presence of pit-holes, cracks always developed on highly porous surfaces (HP). As initial surface roughness increased, alterations in roughness resulting from laser cleaning became more pronounced and the tendency of crack formation reduced. The summary of effects of laser cleaning on surface roughness and crack formation in mortar surfaces are shown in Fig. 16.

A Surface roughness change

Development of crack

Ra

Surface roughness B Ra

N Low

High

C

Porosity LP HP

Moisture content DRY WET

Ra N Moderate

Moderate

N High

Low

Low

High

High

Low

High

Low

High(HP) Low(LP)

Low(HP) High(LP)

Fig. 16. Effect of surface roughness, porosity, and moisture content of mortars on roughness increment and crack development.

34-The effects.qxd

334

9/19/2009

12:22 PM

Page 334

Agnieszka J. KLEMM, Poologanathan SANJEEVAN and Piotr KLEMM

Furthermore, in less porous samples, formation of cracks was more visible, whilst increased surface roughness was low. As a whole, the results indicated that whenever change in surface roughness was high, changes in crack density would be low.

CONCLUSIONS Based on experimental investigations, the following conclusions can be formulated: Thermal stresses on the cementitious materials, resulting from the application of the laser may lead, in extreme cases, to severe crack formation, particularly around the pit-holes on highly porous surfaces. Cracks of wider openings may also form as a consequence of the expansion of water/air inside pores. The second mechanism is more prominent on smoother/denser mortars. The changes of surface roughness, with the number of pulses applied, depend mainly on the initial surface roughness of the mortar. The changes in surface roughness are mainly due to removal of mortar. In the case of smoother surfaces (Ra = 2.28-2.49 μm and 7.70-8.49 μm), surface roughness of the mortar increases with the number of laser pulses applied. The characteristic feature in rougher surfaces (Ra = 15.58-17.89 μm) is an increase of roughness, followed by a sudden decrease. The average change of surface roughness per pulse is high for rough, highly porous, and/or dry surfaces. As initial surface roughness of mortars increases, the alterations in roughness resulting from laser cleaning become more pronounced and the tendency towards crack formation reduces. Whenever change in surface roughness is great, changes in crack density are low. Decrease in surface tensile strength results in the removal of mortar, which reduces crack formation. Laser-cleaned areas proved to be generally denser and more consolidated than the reference surface, due to vitrification of some parts of mortar following the laser cleaning process. REFERENCES 1. Cooper, M., Laser cleaning in conservation, an introduction, Bath Press, UK, 1998. 2. Matsui, I., Nagai, K., Yuasa, N., Ishigami, Y., Removing graffiti on concrete surface by laser, Proc. International Conference, the University of Dundee, Scotland, UK, 2002. 3. Sanjeevan, P., Klemm, A.J., Klemm, P., Removal of graffiti from the mortar by using Qswitched Nd:YAG laser, Applied Surface Science, Vol 253, Issue 20, pp 8543-8553, 2007 4. Sanjeevan, P., Klemm, A.J., Klemm, P., The effects of microstructural features of mortars on the laser cleaning process, Proc. of the 8th International Symposium On Brittle Matrix Composites, 2006, Warsaw, Woodhead Publishing Ltd, pp 45-54 5. McStay, D., Wakefield, R., Murray, M., Houston, J., Laser stone cleaning in Scotland, Historic Scotland, Edinburgh, 2005. 6. Liu, I., Garmire, E., Paint removal using lasers, applied optics, Optical Society of America, Vol. 34, No. 21, 1995. 7. RoĪniakowski K., Klemm, P., Klemm, A.J., Some experimental result of laser beam interaction with surface layer of brick, Building and Environment, Vol. 36, 2001, pp. 485-491. 8. Hertz, K.D., Limits of spalling of fire-exposed concrete, Fire Safety Journal, Volume 38, Issue 2, March 2003, pp 103-116 9. Khan, M.I., Factors affecting the thermal properties of concrete and applicability of its prediction models, Building and Environment 37, 2002, pp 607 – 614

35-Assessing.qxd

9/19/2009

12:01 PM

Page 335

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

ASSESSING THE REPEATABILITY OF THE RESTRAINED RING TEST Aleksandra RADLINSKA1, Jason WEISS2 Villanova University, 800 Lancaster Ave, Villanova, PA 19085-1603, USA e-mail: [email protected] 2 Purdue University, 550 Stadium Mall Drive, West Lafayette, IN 47907-1284, USA e-mail: [email protected] 1

ABSTRACT The restrained ring test is frequently used to assess the susceptibility of a concrete mixture to restrained * shrinkage cracking. The test method has been recently standardized as ASTM C 1581 (2004). Despite many useful applications of this test, concerns have been raised with respect to the repeatability and interpretation of the results when some of the rings made from a concrete mixture crack, while other specimens from the same mixture do not exhibit cracking at all. This paper provides explanation on why even in a properly performed experiment, not all the rings are always expected to crack. The work presented in this paper combines experimental and stochastic approach to analyze the repeatability of the test method and quantify variability in shrinkage measurements. In each of the experiments performed in this work, six rings were cast simultaneously from each batch of mortar prepared. Two water-tocement ratios were considered: w/c=0.30 and w/c=0.40. Additionally, the effect of shrinkage reducing admixtures on variability in the time of cracking was evaluated. The results of the experiment and the simulations are presented to explain why in a given experiment not all the rings may necessarily crack. Additionally, a probabilistic approach is used to describe the probability of cracking in a restrained concrete element. The effect of variability in material properties is included and its effect on concrete cracking prediction explained. This work can be further used by engineers as a tool to evaluate different materials performance and deliver cracking prediction. Once information about cracking probability is obtained, material properties or mixing procedures can be modified and potential for cracking minimized. As such, this approach meets the need for accurate assessment tools that can be implemented in performance-related specifications.

Keywords early-age cracking, restrained ring test, probability, variability, performance-related specifications

INTRODUCTION Background Concrete is material of choice for the worldwide construction of the civil infrastructure, including pavements and bridges. Current goals of achieving sustainability in construction applications underline the need to focus on the most efficient and environmentally friendly application of this versatile material [1]. One way to foster a sustainable approach is by using concrete mixtures that are less prone to crack through application of mixtures with low *

ASTM – American Society for Testing and Materials

35-Assessing.qxd

336

9/19/2009

12:01 PM

Page 336

Aleksandra RADLINSKA and Jason WEISS

shrinkage and minimized cracking susceptibility. It has been shown in the past that shrinkage cracking is one of the main causes of bridge decks distress and premature deterioration. It has been estimated in 2000 by the Federal Highway Administration (FHWA) that approximately 90 billion dollars would be needed for the repairs and rehabilitation of existing bridges [2]. While these data may seem almost a decade old, 2009 ASCE† Report Card for America’s Infrastructure assigned grades C - for bridges and D - for roads [ 3 ] which shows no improvement from 2005 report for bridges (also grade C) and even lower grade for roads (grade D in 2005). The costs of maintenance and repairs costs could be significantly reduced if the life-cycle of the structure was prolonged, i.e., concrete was less prone to cracking. In order to adequately assess susceptibility of cement and concrete mixtures to cracking, adequate tools and techniques must be employed. One of these techniques is standardized in 2004 ‘Standard Test Method for Determining Age at Cracking and Induced Tensile Stress Characteristics of Mortar and Concrete under Restrained Shrinkage’ (ASTM C 1581-04). Restrained Shrinkage Test The ring test was introduced to concrete community [ 4 ] as a test method allowing measurements of concrete susceptibility to cracking under restrained shrinkage. In the test procedure, a cement, mortar or concrete annulus is cast around a steel ring. As time progresses, compressive strain develops in the steel ring as a result of the restrained shrinkage of the cementitious specimen. Whenever the developing residual stress exceed material’s resistance, cracking occurs. The strain development is measured by four strain gages mounted on the inner surface of the steel ring and continuously recorded by data acquisition system. The time of cracking can be simply denoted as a sudden decrease in the strain measurements. Despite the simplicity and potential for versatile applications of the test, some concerns has been raised with respect to the analysis of the result in cases where two or more specimens were cast and some rings cracked, while other rings remain uncracked throughout the experiment. This works presents results obtained in the experimental work together with Monte Carlo simulations in order to: - quantify repeatability of the restrained ring test - explain why in some properly conducted experiments it is not necessary expected to observe cracking in all the specimens under investigation.

EXPERIMENTAL PROCEDURE TO DETERMINE TIME OF CRACKING Ring Geometry The experiments presented in this work were performed in accordance with ASTM C1581-04 [5]. The typical view of the sample used in the experiments together with dimensions of the ring is shown in Figure 1. It should be noted here that in addition to the ASTM specification, there exist an AASHTO‡ Standard Practice for Estimating the Cracking Tendency of Concrete (AASHTO PP-99, 2004) which uses a similar to ASTM testing setup, however different ring dimensions. As such, the AASHTO ring provides a smaller degree of restraint comparing to the ASTM ring and consequently, the results obtained from these two different methods cannot be directly compared [6,7].

† ‡

ASCE – American Society of Civil Engineers AASHTO – American Association of State Highway and Transportation Officials

35-Assessing.qxd

9/19/2009

12:01 PM

Page 337

337

Assessing the repeatability of the restrained ring test

Sealed surface

Steel ring (thickness: 12.5 ± 0.13 mm)

330± 3 mm 406± 3 mm

150± 6 mm Nonabsorptive base

a) b) Figure 1: a) Typical view of the sample used in the testing procedure and b) geometry of the ring sample Each ring used in this work had four strain gages placed at the mid height of the inner surface of the steel ring. These strain gages were mounted on a previously prepared (ground and cleaned) surface, as shown in 2. After the gages were placed on the steel surface, they were secured with aluminum tape and adhesive/sealant to avoid potential damage during specimen casting and demolding procedures.

a) b) Figure 2: Strain gages being placed on the prepared surface of the steel ring: a) polished surface, b) gage being mounted For each mixture six rings were cast simultaneously. To ensure that replicable data are obtained between the tests and between the six rings in each test, rings were always cast in the same order and were placed in the environmental chamber always in the same designated location. The labeling system was chosen where the rings were color-coded to ensure that the rings can be easily differentiated. Mixture Proportions A total of six mortar mixtures were tested in this work, including plain mortar with water-tocement ratio of 0.30 (w/c=0.30) and plain mortar with water-to-cement ratio of 0.40 (w/c=0.40). Additionally, mortar mixtures were examined where 5% of water was replaced by shrinkage reducing admixture (w/c=0.30+5%SRA and w/c=0.40+5%SRA). Ordinary portland cement was used (Type I) that met requirements of ASTM C 150-05. The cement had fineness of 367 m2/kg and an estimated Bogue composition of a 61% C3S, 13% C2S, 10% C3A, loss on ignition of 1.75%, insoluble residue 0.5%, and the total alkali content of 0.61%. The mortar used in this work had 55% of aggregate by volume. The fine aggregate used in all the mixtures was local natural river sand with a specific gravity of 2.62, an absorption capacity of 1.87%, and a fineness modulus of 3.23.

35-Assessing.qxd

338

9/19/2009

12:01 PM

Page 338

Aleksandra RADLINSKA and Jason WEISS

Testing Program Once a mortar mixture was prepared, all six rings were cast, each time in the same order. Each ring was cast in two layers and each layer was externally vibrated. Once all the rings were cast, they were placed in an environmental chamber that allowed ambient control with an accuracy of ± 0.1qC and ±1%RH. Each ring had then top surface covered with plastic to prevent evaporation. As mentioned before, rings were instrumented with four strain gages per each ring. These strain gages were connected to the data acquisition system (Vishay® System 5000) which recorded strains in 5-minute intervals. The outer steel ring was removed and the top surface of the concrete ring was sealed with aluminum tape to ensure moisture loss only from the concrete’s outer circumference 24 hours from the time when cement came into contact with water. Two drying conditions were evaluated for the mixture with water to cement ratio of 0.30 (plain and SRA): in the first series of testing, rings were kept at 23.0 ± 0.1qC and 50%RH (w/c=0.30 and w/c=0.3+5%SRA). In the second series of tests, ambient conditions remained the same, but rings were completely sealed with aluminum tape to prevent moisture loss (w/c=0.30_sealed and w/c=0.3+5%SRA_selaed).

EXPERIMENTAL RESULTS Repeatability and Accuracy of Strain Readings Presentation of the experimental results starts with the analysis of the repeatability of the readings within one ring specimen. The sample strain readings obtained from four gages mounted on a single ring (plain mortar: w/c=0.30, and mortar with 5% SRA: w/c=0.30+5%SRA) are presented in Figure 3. It can be seen that for randomly reported ring results (results for ring #1 ‘black’ and ring #2 ‘yellow’ presented here), consistent and repeatable readings are obtained from all four gages mounted on the ring. More information on repeatability of the results obtained within single mixture and between different experiments conducted on the same mortar can be found in [8].

a) b) Figure 3: Sample strain readings for a single ring specimen (four gages mounted at mid-height of the inner steel ring) a) results for mortar w/c=0.30 and b) results for mortar w/c=0.30+5%SRA

35-Assessing.qxd

9/19/2009

12:01 PM

Page 339

339

Assessing the repeatability of the restrained ring test

Additionally, the analysis of the average standard deviation (SD) of the strain readings within a single ring (readings from the four gages) was conducted for mixtures w/c=0.30, w/c=0.30+5%SRA, w/c=0.40, and w/c=0.40+5%SRA, as shown in Figure 4a. It can be noticed that random pattern is being observed for average values of standard deviation. The four different mixtures analyzed have similar values of SD ranging from a minimum of 0.9 PH for yellow ring in mixture w/c=0.40+5%SRA to a maximum of 5.2 PH for white ring in mixture w/c=0.30+5%SRA. The value of coefficient of variation (COV) was calculated for each of the four gage readings and the maximum value obtained within 95% of data was reported in Figure 4b. The remaining 5% of data was excluded to eliminate the influence of outliers. It can be seen that the value of COV between 0.2 and 4.4 was observed and similar data spread was observed for the four analyzed mixtures.

6

5

Ring 1 Ring 2 Ring 3 Ring 4 Ring 5 Ring 6

5 4

COV for Strain Readings

Avg. Standard Deviation (PH)

Strain at the time of cracking In the next step of analysis, strains recorded at the time of cracking (the last recorded value before the sudden drop in the strain readings) were analyzed. Figure 5 presents the strains at the time of cracking for each of the rings in the tests conducted on mortars w/c=0.30, w/c=0.30+5%SRA, w/c=0.40, and w/c=0.40+5%SRA. It should be noticed that none of the reported data presents any noticeable pattern, which means that data are randomly distributed. As such the ring used or ring’s placement in the environmental chamber does not have any effect on strain readings.

3 2 1 0

Ring 1 Ring 2 Ring 3 Ring 4 Ring 5 Ring 6

4

3

2

1

0

0

1

2

3

4

Mixture Number

5

0

1

2

3

4

5

Mixture Number

a) b) Figure 4: The summary of the results obtained for each of the six rings in four tests performed (Mixture notations: #1: w/c=0.30,#2: w/c=0.30+5%SRA, #3: w/c=0.40, #4: w/c=0.40+5%SRA) a) average standard deviation between the readings obtained in each test, b) maximum coefficient of variation for the results (95% of the data included)

35-Assessing.qxd

9/19/2009

12:01 PM

Page 340

340

Strain at Time of Cracking (PH)

Strain at Time of Cracking (PH)

Aleksandra RADLINSKA and Jason WEISS

0 w/c=0.30, 0% SRA w/c=0.30, 5% SRA

-10 -20 -30 -40 -50 -60 -70 0

1

2

3

4

5

Ring Number

6

7

0 w/c=0.40, 0% SRA w/c=0.40, 5% SRA

-10 -20 -30 -40 -50 -60 -70 0

1

2

3

4

5

6

7

Ring Number

a) b) Figure 5: Strains recorded at the time of cracking a) results for mortar mixtures w/c=0.30 and w/c=0.30=5%SRA and b) results for mortar mixtures w/c=0.40 and w/c=0.40=5%SRA The Effect of SRA and Change in the External Relative Humidity Condition For each mortar mixture described in this work, the effect of 5% of SRA was evaluated. In addition to RH conditions required by ASTM C1581-04 (50%RH ± 4%) a series of 0.30 mortars (w/c=0.30 and w/c=0.30+5%SRA) were covered with aluminum tape to ensure sealed conditions. Figure 6 presents the results of strain measurements obtained for mortars: w/c=0.30 and w/c=0.30+5%SRA (stored at 50% RH), w/c=0.30 and w/c=0.30+5%SRA (sealed), and w/c=0.40 and w/c=0.40+5%SRA (stored at 50% RH). It can be seen that consistent data are obtained between the different rings within each of the tests. Analyzing the results presented in Figure 6a it can be in seen that for the plain mortar mixture cracking occurred between 1.6 and 2.2 days. The rings with mortar with addition of 5% SRA cracked between 5.1 and 6.7 days which indicates that the addition of SRA delays the formation of a crack. Figure 6b presents the results of restrained ring test conducted in sealed condition on mortar with w/c=0.30, with and without SRA. It can be noticed that for plain mortar mixture cracking occurred between 5.8 and 6.8 days. The addition of SRA, however, significantly delayed the time of cracking (first ring cracked at 15.5 days) and when the test was stopped at 25.7 days, two of the six rings remained uncracked. The discussion on why in the properly performed experiment some rings crack while other remain uncracked will be presented in the Results Interpretation section.

35-Assessing.qxd

9/19/2009

12:01 PM

Page 341

341

Assessing the repeatability of the restrained ring test

20

20 Demolding

Average Strain (PH)

Average Strain (PH)

10 0 -10 -20 -30 -40 -50 -60 -70 0

24

48

72

0 -10 -20 -30 -40 -50 -60 w/c=0.30 Sealed w/c=0.30 Sealed, 5% SRA

-70

w/c=0.30 w/c=0.30, 5% SRA

-80

Demolding

10

-80 0

96 120 144 168

96

192 288 384 480 576

Age of Specimen (Hours)

Age of Specimen (Hours)

a)

b) 20

Demolding

10

Average Strain (PH)

Average Strain (PH)

20 0 -10 -20 -30 -40 -50 -60 -70

w/c=0.40 w/c=0.40, 5% SRA

-80 0

48

96

144

192

240

Age of Specimen (Hours)

10 0 -10 0.30 0.30+5%SRA 0.30 Sealed 0.30+5%SRASeal. 0.40 0.40+5%SRA

-20 -30 -40 -50 -60 -70 -80 0

100

200

300

400

500

Age of Specimen (Hours)

c) d) Figure 6: Strain development in restrained ring test a) mortar mixture w/c=0.30 and w/c=0.30+5%SRA stored at 50% RH, b) mortar mixture w/c=0.30 and w/c=0.30+5%SRA sealed condition, c) mortar mixture w/c=0.40 and w/c=0.40+5%SRA stored at 50% RH, d) average strain from six rings for all mortar mixtures It is also interesting to notice in Figure 6c that for w/c=0.40 mixture the addition of SRA delayed the time of cracking: from the range between 3.6 and 4.6 days for plain mixture, to the time frame of 8.2 and 9.3 days for mixture containing SRA. To enable direct comparison of the test results obtained in the six aforementioned mixtures, the average readings from six rings in a single experimental procedure have been summarized in Figure 6d.

RESULTS INTERPRETATION The Effect of SRA Significant work has been already performed to document and explain the mechanism by which SRA causes shrinkage reduction and cracking mitigation [9,10,11,12]. Consistently with previous research, the results of this study present that SRA can effectively reduce the

35-Assessing.qxd

9/19/2009

12:01 PM

342

Page 342

Aleksandra RADLINSKA and Jason WEISS

magnitude of shrinkage in a cementitious system. This work looks more closely at the effect of SRA on the variability in strain measurements. Based on the multiple tests conducted it can be noticed that similar repeatability of measurements is being observed for plain and SRA mixture. It is also interesting to see, that for the case where w/c=0.30 mortar was stored in a sealed condition, not all the rings cracked within the time designated for the experimental work. As such, it can be concluded that in case where water is not allowed to leave the system, different mechanism governs the shrinkage process. The following section discusses the difference between sealed and unsealed drying. Sealed vs. Unsealed Behavior Shrinkage occurs in cementitious systems in response to drying, chemical reaction, or temperature reduction. A distinction should be made here between autogenous shrinkage that occurs in a sealed system is due to internal drying (i.e., self-desiccation) [13,14] and drying shrinkage that is caused by the loss of water to the surrounding environment (i.e., external drying). During the hydration reaction between cement and water, a reduction of volume takes place by approximately 8% to 9% which is related to volume of hydration products being smaller than the volume of reactants. This volume reduction results in the generation of vapor-filled voids as shown in Figure 7a. This is referred to as self-dessication and occurs in the sealed sample. However if the sample is allowed to lose moisture externally (i.e., unsealed system), as shown in Figure 7b, liquid-vapor menisci form at the surface as well and the sample simultaneously undergoes both internal (self-dessiccation) and external drying (water loss). More details on the mechanism of moisture loss and void formation can be found in [16]. It is also worth mentioning here that that for mixtures where the water is lost externally, a moisture gradient exist across the ring sample while for the self dessication case, the stress is more uniform through the cross section with a variation just due to geometry [15]. Sealed

a) Sealed solid

Drying

liquid

b) Unsealed vapor

Figure 7: Drying mechanisms in sealed and unsealed systems: a) Sealed – only internal drying, b) Unsealed – internal plus external drying (adapted from [16]) Data presented in Figure 6b and Figure 6d show that in case of plain mortar samples stored in sealed conditions (w/c=0.30_sealed) cracking was delayed comparing to samples stored at 50%RH (w/c=0.30). This was due to the higher internal humidity maintained by plain sealed system comparing to the unsealed system. Additionally, because presence of SRA in the system lowers capillary pressure [16], lower shrinkage is being observed in mixtures with SRA than in plain mixtures. A comparison of restrained shrinkage test results for plain system

35-Assessing.qxd

9/19/2009

12:01 PM

Page 343

Assessing the repeatability of the restrained ring test

343

with SRA (w/c=0.30+5%SRA) and sealed system with SRA (w/c=0.30+5%SRA_sealed) shows that cracking is further delayed in case of the sealed system. This is because the mixture w/c=0.30+5%SRA_sealed is characterized by the lowest shrinkage of the four cases analyzed here. Probability of Cracking in Restrained Ring Test Despite the adoption of ASTM C1581-04 as a standard testing procedure, users of the test frequently happen to raise a question: what it means when only some percent of the rings crack, while other specimens (performed using the same mixture and stored in the same environmental conditions) do not experience cracking at all. In this work, the modeling approach developed earlier [17,18] was utilized to compare time-dependent stress and strength development and predict cracking occurrence in restrained concrete element. While details of the modeling approach are explained elsewhere [8,19], this work will focus mainly on the importance of including probability information in the modeling calculations and design considerations. The results presented in Figure 8 were obtained using Monte Carlo simulations (10,000 iterations per simulations) assuming that the shrinkage value measured in the specimens has inherent material variability, based on the previous studies [8] assumed to be 12%. As such, shrinkage parameter in this work is being treated as random variable with assigned normal distribution. During Monte Carlo simulation process, the value of shrinkage is being randomly drawn from predefined distribution and probability of cracking is calculated. It should be noted here that the model assumes that cracking took place whenever the calculated tensile strength exceeds the strength of the material. As can be noticed in Figure 8a probability of cracking for mixture with water-to-cement ratio of 0.30 increases over time, however different slope is being observed for plain mixture and mixture with 5% SRA. This means that lower probability of cracking is observed for mixtures containing SRA. The probability of cracking for both mixtures ultimately reaches the value of 100% (cumulative distribution functions, CDF, reaches the value of 1) and this is consistent with results obtained in the experimental work where all the rings cracked by the age of 10 days. Results for mortar with water-to-cement ratio of 0.40 have been presented in Figure 8b. As in the previous case the effect of SRA is clearly visible, as lower probability of cracking is being observed for the w/c=0.40+5%SRA mixture. The analysis of Figure 8 additionally allows explaining why in a given experiment where multiple rings were cast, not all of the specimens are expected to crack. Since concrete possess some inherent material variability, not only shrinkage can be treated as random variable, but the stress and strength development as well. As such, instead of single curve describing material properties development, a range including information about material variability should be used. For mixtures with low magnitude of shrinkage this would result in CDF having lower slope and reaching ultimate value of cracking probability less than 100%. A case like that was seen for mixture w/c=0.30+5%SRA stored in sealed condition, where only 4 out of six rings cracked.

35-Assessing.qxd

9/19/2009

12:01 PM

Page 344

344

Aleksandra RADLINSKA and Jason WEISS

1.0

0.8 0.6 0.4 0.2 w/c=0.30, 0% SRA w/c=0.30, 5% SRA

0.0 0

4

8

12

16

20

24

28

Probability of Cracking

Probability of Cracking

1.0

0.8 0.6 0.4 0.2 w/c=0.40, 0% SRA w/c=0.40, 5% SRA

0.0 0

Age of Specimen (Days)

4

8

12

16

20

24

28

Age of Specimen (Days)

a) b) Figure 8: Results of simulations for mortar mixtures: a) probability of cracking for 0.30 mortar (plain and with SRA), b) probability of cracking for 0.40 mortar (plain and with SRA) CONCLUSIONS The work presented in this paper leads to the following concluding remarks: x The restrained ring test is an accurate and repeatable testing method to assess cracking

susceptibility of concrete mixture. While some of the rings prepared from the same batch might crack while other remain uncracked, this occurs due to inherent variability present in concrete and the random nature of concrete material properties; x The probability of cracking in cement, mortar or concrete mixture can by minimized if

the magnitude of shrinkage can be lowered, for example by addition of shrinkage reducing admixture. With higher magnitude of shrinkage, probability of cracking increases; x A model has been developed that allows cracking prediction using Monte Carlo

simulations. This model allows time-dependent stress and strain calculations and potential for cracking estimation. This model can be further used to develop a tool that would allow quantifying future mixture performance. This would find very useful application as an assessment tool in performance-related specifications. ACKNOWLEDGEMENTS The authors gratefully acknowledge support for this research which has come from the Center for Advanced Cement-Based Materials. The experimental work presented in this study was conducted in the Materials Characterization and Sensing Laboratory at Purdue University. The authors gratefully acknowledge the support that has made this laboratory and its operation possible. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors.

35-Assessing.qxd

9/19/2009

12:01 PM

Page 345

Assessing the repeatability of the restrained ring test

345

REFERENCES 1. Sakai, K., Sordyl, D., ACI St. Louis Workshop in Sustainability. Planning the effects of green building and international standars, Concrete International, 31(2), 2009, 34-37 2. Kirkpatrick, T. J., Weyers, R. E., Sprinkel, M. M., and Anderson-Cook, C. M. Impact of specification changes on chloride-induced corrosion service life of bridge decks. Cement and Concrete Research, 32(8), 2002, pp. 1189-1197 3. http://www.infrastructurereportcard.org (retrieved on April 2nd, 2009) 4. Carlson, R.W. Cracking of concrete, Boston Society of Civil Engineers 29(2), 1942, pp. 98-109 5. ASTM, Standard C1581-04, Standard Test Method for Determining Age at Cracking and Induced Tensile Stress Characteristics of Mortar and Concrete under Restrained Shrinkage, Annual Book of ASTM Standards, Vol. 4(2), ASTM International, West Conshohocken, PA, 2006, 787-792 6. Hossain, A.B., and Weiss, J. Assessing residual stress development and stress relaxation in restrained concrete ring specimens, Cement & Concrete Composites 26 (2004), pp. 531–540 7. Hossain, A.B, Pease, B., and Weiss, W. J., Quantifying Early-Age Stress Development and Cracking in Low w/c Concrete Using the Restrained Ring Test with Acoustic Emission, Transportation Research Record, Concrete Materials and Construction 1834, 2003, pp. 24-33 8. Radlinska, A., Bucher, B., and Weiss, J. Comments on the Interpretation of Results from the Restrained Ring Test, Journal of ASTM International, 5(10), 2008, pp. 11 9. Weiss, J., and N.S. Berke. Shrinkage Reducing Admixtures. A. Bentur (Ed.), Early Age Cracking in Cementitious Systems, RILEM State of the Art Report, 2002 10 . Rajabipour, F., Sant, G., and Weiss, J. Interactions between Shrinkage Reducing Admixtures and Cement Paste’s Pore Solution. Cement and Concrete Research 38(5), 2008, pp. 606-615 11. Bentz, D., Geiker, M., and K. Hansen. Shrinkage-Reducing Admixtures and Early Age Desiccation in Cement Pastes and Mortars, Cement and Concrete Research, 31(7), 2001, pp. 1075-1085 12. Sato T., Goto T., Sakai K. Mechanism for reducing drying shrinkage of hardened cement by organic additives, CAJ Review, Cement Association of Japan, 1983, pp. 52-55 13. Jensen, O.M., and P.F. Hansen, Autogenous Deformation and RH-change in Perspective, Cement and Concrete Research, 31, 2001, pp. 1859-1865 14 . Geiker, M.R., Bentz, D.P. and O.M. Jensen. Mitigating Autogenous Shrinkage by Internal Curing. High-Performance Structural Lightweight Concrete, American Concrete Institute Special Publication 218, J.P. Ries and T.A. Holm, eds., 2004, pp. 143154 15. Moon J.H., and Weiss, J. Estimating residual stress in the restrained ring test under circumferential drying, Cement and Concrete Composites, 28, 2006, pp. 486–496 16. Radlinska, A., Rajabipour, F., Bucher, B., Henkensiefken, R., Sant, G., and Weiss, J. Shrinkage Mitigation Strategies in Cementitious Systems: a Closer Look at Differences in Sealed and Unsealed Behavior, Transportation Research Record Vol. 2070, (2008), pp. 59-67 17 . Weiss, J. ‘Shrinkage cracking in restrained concrete slabs: Test Methods, Material Compositions, Shrinkage Reducing Admixtures and Theoretical Modeling’, M.S. Thesis, Northwestern University, Evanston, IL, 1997

35-Assessing.qxd

346

9/19/2009

12:01 PM

Page 346

Aleksandra RADLINSKA and Jason WEISS

18. Radlinska, A., Pease, B., and Weiss, J. ‘A Preliminary Numerical Investigation on the Influence of Material Variability in the Early-Age Cracking Behavior of Restrained Concrete’, RILEM Materials and Structures Vol.40(4), 2007, 375-386 19. Radlinska, A. (2008)‘Reliability-based analysis of early-age cracking in concrete, Ph.D. Dissertation, Purdue University, West Lafayette, IN

36-Performance.qxd

9/19/2009

12:01 PM

Page 347

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

PERFORMANCE EVALUATION OF RAPID-SETTING MATERIALS FOR CONCRETE PAVEMENT/BRIDGE DECK REPAIRS: LABORATORY AND FIELD PERSPECTIVES Prashant V. RAM1 and Jan OLEK2 Applied Pavement Technology, Inc. 115 W. Main St., Champaign, Illinois-61821, USA, email: [email protected] 2 Purdue University 550 Stadium Mall Drive, West Lafayette, Indiana-47907, USA, e-mail: [email protected] 1

ABSTRACT Maintenance and repair of concrete structures has been the focus of the construction industry activities for nearly half a century. The ability to rapidly repair and rehabilitate deteriorated bridge decks and highway pavements minimizes interference with traffic in heavily traveled areas and reduces travel delays. In many cases, the needed repairs have been accomplished using rapid-setting materials. Extensive review of literature has indicated that repair materials which meet the required performance criteria in the laboratory do not necessarily perform well in the field. This paper presents results and analyses of the laboratory investigations of rapid-setting repair materials as well as issues related to their field installation. It was seen that while, in most cases, the controlled laboratory conditions yielded consistent mixes and acceptable performance, the properties of mixes produced on site were more variable. This variability was the result of changes in the amount of aggregate extension used, moisture content of the aggregates, amount water added and ambient temperature conditions. Suggestions regarding the possible improvements in quality control procedures of field mixes to ensure adequate performance of the repair patches are also presented.

Keywords Concrete repair, rapid-setting materials, durability

INTRODUCTION Background Concrete repair is a complex process, presenting unique challenges that are different from those associated with new concrete construction. The repair system must successfully integrate new materials with the existing concrete to form a durable composite system which is capable of enduring exposure to service loads, environment and time (Vaysburd et al. [1], Vaysburd [2] and McDonald et al. [3]). Rapid-setting materials have been in great demand for repair applications as they considerably reduce the construction downtime. Repairs performed under high traffic volumes and aggressive environmental conditions require materials that will cure rapidly while developing adequate strength and durability. The main factors which cause premature failure of repairs include exposure to freezing and thawing cycles, aggressive chemical exposure, mechanical abrasion, loss of bond between existing concrete and repair material, and lack of dimensional stability of the repair

36-Performance.qxd

9/19/2009

12:01 PM

Page 348

348

Prashant V. RAM and Jan OLEK

material. While some of the problems associated with premature deterioration of repairs are due to structural failures, most of the problems are durability related. Research Objectives The primary objectives of the present study were to identify the critical properties based on the laboratory tests that could be correlated to the field performance of the repair materials. The laboratory test results, together with the field performance data, could be used in the future to develop performance-based specifications for the selection and use of rapid setting materials for the Indiana Department of Transportation (INDOT).

LABORATORY TESTING Materials Six commercial rapid-setting repair materials (RMs) were chosen for this study based upon the extensive laboratory investigations carried out earlier at Purdue University and on the review of other literature in the topic (Barde et al. [4], Deshpande and Olek [5]). The first four repair materials (RM1-RM4) were pre-packaged mortars. Repair Material 5 was supplied in two formulations – regular (RM5-R) and air-entrained (RM5-AE). Each of these formulations was supplied in the form of powder, without any pre-mixed aggregate. RM1 through RM3 were portland cement based materials and RM4 was an alumina-cement based repair material. RM5-R and RM5-AE were gypsum cement based binders. Table 1 provides information on the mixture composition of all repair materials. Table 1. Mixture Proportions of the Repair Materials

Material RM1 RM2 RM3 RM4 RM5-R RM-AE

Aggregate Extension (%) PG* Sand 60 NA 60 NA 60 NA 60 NA 100 100 100 100

Mix Water (l per bag** of RM) Actual*** 3.22 2.70 3.00 3.07 6.62 6.62

Recommended 3.22 2.60 3.00 2.60 5.58 5.58

% Extra 0 3.8 0 18.1 16.5 16.5

*PG: Pea Gravel ** RMs were available in 22.7 kg bags *** Extra water was added to achieve adequate workability

The fine aggregate used was natural sand with absorption of 1.85% and a specific gravity of 2.63 while the coarse aggregate used was locally available pea gravel with max diameter of 9.5 mm with absorption of 2.36% and a specific gravity of 2.64. Mixing Sequence A small capacity (28 L) portable mortar mixer was used to carry out the mixing operations in the laboratory. The following mixing sequence was adopted: Pea Gravel (+ Sand for RM5 R/AE) Æ Mix for 20 seconds Æ RM + ¾ Water Æ Mix for 120-150 seconds. The remaining water was added during mixing. Test Procedures All laboratory concretes were prepared, cured and tested at a constant temperature of 23±2°C. The fresh and hardened properties evaluated are listed in Table 2, which also provides information on the performance requirements adopted for this project.

36-Performance.qxd

9/19/2009

12:01 PM

Page 349

Performance evaluation of rapid-setting materials for concrete pavements/bridge decks repair: laboratory and field perspectives

349

The temperature signature data (for laboratory mixes only) were collected using thermocouples placed in 76x152 mm cylinders. The compressive strength was tested using 76x152 mm cylinders at the ages of 2, 4, 24 hours and 28 days. The specimens tested at the age of 2 hours were de-molded right before testing. The other specimens were de-molded 2 hours after addition of water and were moistcured at 23ºC and 100% relative humidity (RH) until tested. The compressive strength specimens cast in the field were de-molded and tested after 12 hours (this was the only strength data collected). Table 2. Testing Procedures and Performance Requirements Property Test Method Requirement Workability1 (mm) ASTM C 143 (no rodding)*** Slump 200-250 Spread/Flow 400-700 Set Time (Min.) ASTM C 266 Initial 10-20 Final 20-40 Temperature Signature Compressive Strength (MPa) 2h 14 ASTM C 39* 4h 21 24 h 28 35 28 days Slant Shear Bond Strength2 (MPa) ASTM C 882 modified by ASTM C 1 day 7 928 [6] 7 days 10 Iowa Shear Bond Strength (MPa) IOWA 406 C-2000*** [7] 1 day 7 days 28 Day Free Shrinkage2,3 ASTM C 157 <750 ȝİ Restrained Shrinkage AASHTO PP 34 [8] No Cracking upto 28 Days* Freeze-Thaw Resistance ASTM C 666 Procedure A Min 60% RDM after 300 cycles* Scaling Resistance ASTM C 672 (25 Cycles) Rating - 0 28-Day Rapid Chloride Permeability AASHTO T 277 [9] < 3000 Coloumbs 1 Slump or flow was measured, depending on material consistency 2Modified by ASTM C 928 3Initial measurement performed after 1.75 hours of addition of water 4Procedure A was adopted; specimens de-molded after 2 hours. *Minimum requirements ** Maximum requirements *** Non-standard test procedures

The bond strength was evaluated at the age of 1 and 7-days. The effect of the environmental exposure on the bond-strength was also evaluated. The bond strength was tested after the specimens were moist cured for 7 days and then exposed to the natural environment for 60 days (Dec 6 2007 – Feb 5 2008). In addition to the ASTM C 882 method, the bond strength was also evaluated using a non-standard Iowa shear test (Iowa DOT Test Method 406-C-2000) which is typically used to test the shear bond strength between asphalt and Portland cement concrete. The freeze-thaw specimens cast in the field were de-molded 12 hours after addition of water and were exposed to the natural environment (stored at the repair site) for 12-days. The specimens were immersed in lime water 48 hours before the test. The rapid chloride

36-Performance.qxd

9/19/2009

12:01 PM

Page 350

350

Prashant V. RAM and Jan OLEK

permeability specimens were moist cured at 23±2°C and 100% RH for 28 days after which the test was performed.

FIELD INSTALLATION Introduction In order to verify the performance of the RMs in the field, they were used to repair deteriorated sections on the bridge deck over Wabash River. The bridge deck is located on Interstate 65 near the town of Lafayette, Indiana, USA. The installation took place in October 2007. Due to traffic control and scheduling problems, the field installation of RM5-R could not be carried out. Representatives from the material manufacturers were present on site to ensure proper mixing and placing of the materials. Preparation of Repair Areas In preparation for the repairs, the surface of the bridge deck was sounded with a hammer to identify the boundaries of deteriorated areas. The removal of concrete from these areas was accomplished using concrete saws and pneumatic jack hammers. If rebars were present, the concrete was removed to the depth of 25mm below the rebars. The repair area was then cleared using a blast of high pressure air. The cleared surface was then wetted to avoid removal of mixture water from the repair material. An example of repair area prepared for patching is shown in Figure 1.

Figure 1. Repair Area Prepared for Patching

Figure 2. Finished Patch

Repair Material Installation After the preparation of the patch area had been completed, the repair material was mixed onsite, typically in a drum mixer and placed in one or more batches, depending on the size of the patch. Consolidation was carried out by tamping the repair concrete using the shovels. The surface of the newly placed patches was finished with concrete trowels and textured using a broom. Except for RM4, no specific curing regime was recommended by the representatives of the material manufacturers present on site. As a result, those patches were not cured after placement. For RM4, the placed material was covered with a plastic sheet. Figure 2 shows a sample patch after finishing.

36-Performance.qxd

9/19/2009

12:01 PM

Page 351

Performance evaluation of rapid-setting materials for concrete pavements/bridge decks repair: laboratory and field perspectives

351

General Comments x Due to time constraints (the bridge-deck had to be open to traffic in 8 hours), none of the materials used to prepare repair mixtures were actually weighed. Instead, a volume-based approximation of the mass was used to proportion the mixtures. x The water was added using 3.78L (1 Gallon) containers. Depending on the results of visual observation of the consistency of the mix inside the mixer, extra water was added upon the recommendation of the material manufacturer representative. The amount of mix water varied from batch to batch. x The consistency of mixes sometimes varied significantly from batch to batch due to the variations in the amount of aggregate extension and mix water used. x The moisture content of the aggregates used was not taken into account when batching the materials. x The ambient temperature at the time of placement was around 10°C and, as a result, the setting time of the materials was extended compared to the laboratory mixtures

RESULTS AND ANALYSIS Setting Time, Workability and Temperature Signature Table 3 shows a comparison of the initial and final laboratory setting times for all the materials. The initial setting time for all the materials was between 10 and 20 minutes, except RM4 which had an initial setting time of 34 minutes. RM4 and RM5-AE are the only materials with the final setting site of over 45 minutes. The rest of the materials had a final setting time of at most 35 minutes. Table 3. Setting Time and Workability Material RM1 RM2 RM3 RM4 RM5-R RM-AE

Initial Setting Final Setting Slump (S, mm), Time (min) Time (min) Flow (F, mm) 15 22 F 15 27 S 13 31 S 34 47 F 16 35 F 18 48 F

Depending upon the “wetness” of the mixtures, their workability was measured either in terms of slump or slump flow. Table 3 shows that RM1, RM4, RM5-R and RM5-AE displayed good flow characteristics and may be considered self-leveling products. However, (although not shown in Table 3) RM2 and RM3 experienced rapid slump loss. These materials should be placed within 4-8 min. after mixing to avoid consolidation problems. The temperature signature curves for the RMs are shown in Figure 3. There is a steep surge in the temperature for RM1, RM2 and RM3, which occurs within 20 minutes after addition of mix water. RM4 shows a slightly delayed response. The rate of temperature evolution in RM5-R/AE was considerably slower than that in other materials. When compared to RM5-R, the curve for RM5-AE is slightly shifted to the right which may be explained by the fact that RM5-AE has admixed air-entraining agent which may retard the rate of hydration. High dosages of air entraining agent may reduce rate of cement hydration

36-Performance.qxd

9/19/2009

12:01 PM

Page 352

352

Prashant V. RAM and Jan OLEK

(Ramachandran et al. [10]). After 6 hours, all materials returned to the level of ambient laboratory temperature. 70

RM1

RM2

RM3

RM5-R

RM5-AE

Requirement

RM4

10

100

70

60 RM2 Compressive Strength (MPa)

Temperature (°C)

RM3 50

RM4

40

RM1

30

RM5-AE RM5-R

60 50 40 30 20 10

20

0 1

10 0

1

2

3 Time (Hours)

4

5

Figure 3. Temperature Signature Curves (Lab Mixes)

1000

Specimen Age (Hours)

6

Figure 4. Rate of Compressive Strength Gain (Lab Mixes)

Compressive Strength The rate of compressive strength gain for laboratory mixes used in this study is shown in Figure 4. At the ages of 2, 4 and 24 hours, RM5-R and RM5-AE were the only two materials with strengths well below the project specified requirements. At the age of 28 days, only RM5-AE had strength lower than the required value of 35 MPa. The other repair materials showed excellent rates of strength gain under laboratory conditions of 23±2°C. The lower rate of strength gains (at least at early ages) observed for RM5-R and RM5-AE materials can be attributed to their longer set times. This fact is also corroborated by the temperature signature curves (Figure 3) which show a significantly delayed response. Table 4 shows a comparison between the 12-hour compressive strength of the field specimens and the predicted 12-hour compressive strength of the laboratory mixes. Table 4. 12 Hr. Compressive Strengh: Lab vs. Field Mixes

Material RM1 RM2 RM3 RM4 RM5-R RM-AE

Field (MPa) 15.1 10.5 36.1 41.2 No Data 10.6

Lab (MPa) 27.4 36.1 37.5 42.9 20.1 15.4

% Difference 81.6 244.1 3.9 4.1 44.7

The laboratory RM1, RM2 and RM5-AE mixtures have the 12-Hr. compressive strengths which were, respectively, around 80%, 240% and 45% higher than those of the field mixes. For RM3 and RM4, the strengths from the laboratory and the field mixes are very comparable. This indicates that some of the RMs are more sensitive to temperature variations during early hydration than the others.

36-Performance.qxd

9/19/2009

12:01 PM

Page 353

Performance evaluation of rapid-setting materials for concrete pavements/bridge decks repair: laboratory and field perspectives

353

Bond Strength The slant-shear bond strength results are presented in Figure 5. All materials tested achieved the project-required bond-strength. The increase in bond strength from 1-day to 7-day is negligible for RM1 and RM2. In order to study the effect of the natural freeze-thaw exposure, another set of specimens were moist-cured for 7 days and then exposed to the natural environment from Dec 6, 2007 – Feb 5, 2008. The temperature history for this environmental exposure is shown in Figure 7. Although the temperature during the exposure period varied from -20°C to over +20°C, these variations did not seem to have any negative effect on the slant-shear bond strength values. In fact, there was an increase in bond strength for all the repair materials.

1-Day MC

7-Day MC

7 Day MC & 60 Day Environmental Exposure

1-Day MC

25 ASTM C928 1-Day Requirement

7-Day MC & 60 Day Environmental Exposure

RM2

RM3

ASTM C928 7-Day Requirement

20

4 Bond Strength (MPa)

Bond Strength (MPa)

7-Day MC

5

15

10

5

3

2

1

0

0

FX

SQ

HD

TR

Dur-R

Dur-AE

RM1

Figure 5. Slant Shear Bond Strength (Lab Mixes)

RM4

RM5-R

RM5-AE

Figure 6. Iowa Shear Bond Strength (Lab Mixes)

Maximum Temperature

Minimum Temperature

25 20 15

Temperature (°C)

10 5 0 -5 -10 -15 -20 -25 24-Nov

4-Dec 14-Dec 24-Dec

3-Jan

13-Jan 23-Jan

2-Feb

12-Feb

Figure 7. Ambient Temperature History for the for the Bond Durability Specimens (Source:http://www.wunderground.com) The Iowa-shear bond strength results are presented in Figure 6. The same curing/exposure regime was adopted as in the case of the slant-shear test. It must be noted that, in the slant-shear test, the repair concrete is installed on substrate mortar, where there is no coarse aggregate exposure along the bonding plane. However in the Iowa shear test, a significant number of coarse aggregate grains are present along the bonding plane (smooth saw-cut surface). Also, in the slant-shear test, the stress configuration at the interface is a state

36-Performance.qxd

9/19/2009

12:01 PM

Page 354

354

Prashant V. RAM and Jan OLEK

of combined shear and compression, whereas, in the Iowa-shear test, the interface is in a state of pure shear. Hence, one would expect significantly lower bond-strengths in the case of the Iowa-shear test. For RM1 though RM4, the increase in the Iowa shear bond strength from 1-Day to 7day is not very significant. However, for RM5-R and RM5-AE there is an evident increase in the 1-day and 7-day bond strengths (around 75%). Another interesting observation was that, the 60-day environmental exposure resulted in a very conspicuous decrease in the Iowa-shear bond strength for RM5-R and RM5-AE. However, the same was not observed in the case of the slant-shear test. Extensive field investigations must be carried out to verify these results. Shrinkage The free-shrinkage of the repair materials is presented in Figure 8. From Figure 8, it is observed that the majority of the shrinkage occurred within the first two days. After the initial reading, subsequent measurements were performed at the ages of 1, 2, 4, 7, 14 and 28 days. RM2 and RM3 displayed the maximum shrinkage after 28 days, with the values being around 600 and 700 micro-strains respectively. RM1 had the minimum shrinkage of around 300 micro-strains after 28 days. It must be noted that the initial measurement of the shrinkage beams was performed 1 hr. 45 mins. after the addition of mix-water. Hence, the dimensional changes of the material from time zero until 1 hr. 45 mins. has not been captured. Since these materials undergo a significant increase in temperature at very early ages, they will undergo considerable expansion which, in-turn may result in lower values of net shrinkage. Alternate test methods for dimensional changes at very early ages like the corrugated tube method (Sant et al. [11]) must be considered when evaluating these materials. RM1

RM2

RM3

RM4

RM5-R

RM5-AE

0

Shrinkage (Microstrains)

-100 -200 -300 -400 -500 -600 -700 -800 0

5

10

15

20

25

30

Specimen Age (Days)

Figure 8. Free Shrinkage (Lab Mixes) To study the cracking tendency of the repair materials, the restrained shrinkage test (AASHTO PP 32 [8]) was performed. The strains were monitored for a period of 35 days in a controlled environment of 23°C and 50% RH. None of the repair materials cracked within the time frame of testing. The overall strains for all the materials was generally low and hence, the restrained shrinkage cracking tendency of these materials may be considered to be low.

36-Performance.qxd

9/19/2009

12:01 PM

Page 355

Performance evaluation of rapid-setting materials for concrete pavements/bridge decks repair: laboratory and field perspectives

355

Freeze-Thaw Resistance Table 5 provides results on the change in relative dynamic modulus (RDM) of the specimens subjected to accelerated freeze-thaw (F-T) conditions in the laboratory. Table 5. Freeze-Thaw Resistance Material RM1 RM2 RM3 RM4 RM5-R RM-AE

RDM after 100 Cycles (%) Lab Field 93 98 99 90 95 98 * 97 * NDA 75 93

RDM after 200 Cycles (%) Lab Field 91 99 99 * 91 96 * 98 * NDA 60 85

RDM after 300 Cycles (%) Lab Field 90 99 98 * 80 92 * 99 * NDA 35 74

The laboratory mixes of RM1, RM2 and RM3 showed excellent freeze-thaw resistance (RDM>80% after 300 F-T cycles). The laboratory specimens of RM4 and RM5-R mixes failed after 30 cycles of freezing and thawing. The laboratory specimens of RM5-AE reached a RDM of 60% after 220 cycles of freezing and thawing. The specimens also underwent severe scaling. This could be attributed to the excess water added (~17%) above the manufacturer’s recommendation. Also, the laboratory specimens were consolidated by vibration, which might have altered the air-void system resulting in poor F-T resistance for RM4 and RM5-R and reduced F-T resistance for RM5-AE. RM2 specimens cast in the field failed after 160 cycles of freezing and thawing. RM1, RM3 and RM4 specimens cast in the field displayed excellent freeze-thaw resistance. RM5-AE specimens cast in the field had a RDM of 74% after 300 cycles. Resistance to Chloride-Ion Penetration Figure 9 provides information on the relative resistance of tested mixtures to chloride-ion penetration. 9000

Total Charge Passed (Coloumb)

8000 7000 6000 5000 4000 3000 2000 1000 0 RM1

RM2

RM3

RM4

RM5-R

RM5-AE

Figure 9. Resistance to Chloride-Ion Penetration (Lab Mixes) The material RM3 had the lowest value of the total charge passed after 6 hours (336 coulombs) which corresponds to very low chloride-ion permeability according to AASHTO T 277, RM2 had low chloride-ion permeability and RM1, RM4 and RM5-R had moderate chloride-ion permeability. RM5-AE had the highest value (8159 coulombs) for the total

36-Performance.qxd

9/19/2009

12:01 PM

Page 356

356

Prashant V. RAM and Jan OLEK

charge passed after 6 hours indicating that it could potentially be unsafe to be used in direct contact with rebars under condition of salt exposure.

CONCLUDING REMARKS Laboratory Study x The repair materials studied in this project had varying chemical compositions which influenced performance in terms of mechanical and durability parameters. x RM1, RM4, RM5-R and RM5-AE materials displayed good flow characteristics and can be considered as having self-leveling characteristics. RM2 and RM3 experienced rapid slump loss. x All the materials (except RM5-AE) showed acceptable rate of strength gain. x In general, all the materials displayed good bond to substrate concrete. x Relatively poor freeze-thaw durability was observed for RM4, RM5-R and RM5-AE; possibly the result of additional water that needed to be added to maintain adequate workability of extended mixtures. x RM5-AE displayed poor resistance to chloride-ion penetration x The cracking potential of all materials was low. Field Installations x Ambient temperature conditions played a vital role in controlling rate of strength gain. In hot weather conditions, cold water should be used to prevent flash-set of materials studied. In cold weather conditions, warm water should be used to accelerate hydration. x The uniformity of the field installations can be improved by accounting for moisture content of the aggregate during the batching process. x Calibrated buckets should be used to control the amount of water and aggregates added into each mix. ACKNOWLEDGEMENTS This work was supported by the Joint Transportation Research Program administered by the Indiana Department of Transportation and Purdue University. The contents of this paper reflect the views of the authors, who are responsible for the facts and the accuracy of the data presented herein, and do not necessarily reflect the official views or policies of the Federal Highway Administration and the Indiana Department of Transportation, nor do the contents constitute a standard, specification, or regulation. The authors would like to thank all the material manufacturers for supplying their materials for testing and also assisting with the field installations. REFERENCES 1. Vaysburd, A.M., Emmons, P.H., Mailvaganam, N.P., McDonald, J.E. and Bissonnette, B., Concrete Repair Technology – A revised Approach is Needed, Concrete International, Jan. 2004, 59 – 64 2. Vaysburd, A.M. Holistic system approach to design and implementation of concrete repair, Cement and Concrete Composites Vol. 28, 2006, 671-678 3. McDonald, J.E., Vaysburd., A.M., Emmons, P.H., Poston, R.W., and Kesner, K., 2001. Selecting durable repair materials: Performance Criteria – Summary, Concrete International, Jan 2001, 37-44

36-Performance.qxd

9/19/2009

12:01 PM

Page 357

Performance evaluation of rapid-setting materials for concrete pavements/bridge decks repair: laboratory and field perspectives

357

4. Barde, A., Parameswaran, S., Chariton, T., Weiss, J., Cohen, M., and Newbolds, S. Evaluation of Rapid Setting Cement-Based Materials for Patching and Repair (Phase-I), Joint Transportation Research Program Project SPR-2648, Purdue University, West Lafayette, IN, 2006 5. Deshpande, Y., Nantung, T. and Olek, J. Dowel Bar Retrofit Mix Design and Specification, Joint Transportation Research Program Project SPR-2789. Purdue University, West Lafayette, IN., 2006 6. ASTM C 928, Standard Specification for Packaged, Dry, Rapid-Hardening Cementitious Materials for Concrete Repairs, ASTM International, 2006 7. Iowa Department of Transportation, Test Method No. Iowa 406-C 2000. Method of Test for determining the shearing strength of bonded concrete, 2000 8. AASHTO PP 34 Standard Practice for Estimating the Cracking Tendency of Concrete., 2005 9. AASHTO T 277. Standard Practice for Electrical Indication of Concrete’s Ability to Resist Chloride Ion Penetration, 2005 10. Ramachandran, V.S., Paroli, R.M., Beaudoin, J.J., and Delgado, A.H. Handbook of Thermal Analysis of Construction Materials. Noyes Publication/William Andrew Publishing, Norwich, New York USA, 2003 11. Sant, G., Lura, P., and Weiss, W.J., Measurement of Volume Change in Cementitious Materials at Early Ages: Review of Testing Protocols and Interpretation of Results: Transportation Research Record, No. 1979, 2007

37-The effect of.qxd

9/19/2009

12:01 PM

Page 359

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

THE EFFECT OF TEMPERATURE ON THE RHEOLOGICAL PROPERTIES OF SELF-COMPACTING CONCRETE Jacek GOàASZEWSKI, Grzegorz CYGAN Silesian University of Technology, Faculty of Civil Engineering Akademicka 5, 44 - 100 Gliwice, Poland, e-mail:[email protected]

ABSTRACT In the paper the methodology and results of investigation of the influence of temperature and time on the rheological properties of cement paste and different in cement paste volume fresh self compacting concrete are presented and discussed. The influence of temperature on the rheology of mortar analogous to mortar that fills voids of coarse aggregate in the self compacting concretes was additionally the subject of investigation. It was found that the influence of temperature on rheological parameters of fresh cement paste and concrete may be significantly different, whereas on rheological parameters of fresh mortar and concrete is generally similar for both these materials. Discussion on the test results concern possible mechanisms of temperature influence on rheological properties of fresh cement paste, mortar and concrete. It is concluded that testing cement paste is not sufficient for designing and development of self-compacting concrete in respect to changing temperature. For predicting trends of changes in rheological behaviour of fresh self compacting concrete in changing temperature tests made on mortars are necessary.

Keywords Self-Compacting Concrete, rheology, temperature

INTRODUCTION Self-Compacting Concrete (SCC) is defined as concrete that has an ability to flow under its own weight, to fill the required space or formwork completely and to produce a dense and adequately homogenous material without a need for vibrating compaction, [1]. Rheological properties are the dominant feature of SCC with regard to its designing and processing. It is accepted that cement based mixtures, such as fresh mortars and concretes, as well as SCC, behave as Bingham material, [1]. The characterization of rheological properties of cement paste generally demands more complex models, [2]. Nevertheless, from the practical point of view Bingham model is in most cases accurate enough. Bingham material properties are expressed according to the formula:

W W K J o

pl

(1)

. where W (Pa) is the shear stress at shear rate J (1/s) and W (Pa) and Kpl (Pa.s) are the rheological parameters of yield stress and plastic viscosity respectively. Rheological parameters of fresh SCC should be properly matched to enable mixture to free flow and self release of trapped air without segregation and bleeding. Comparing with concrete compacted by vibration, the range of suitable rheological properties for a SCC in given conditions is considerably narrower.

37-The effect of.qxd

360

9/19/2009

12:01 PM

Page 360

Jacek GO£ASZEWSKI and Grzegorz CYGAN

It is common knowledge that rheological behaviour of fresh concrete is closely affected by the material temperature. That problem becomes especially important when it concerns the self compacting concrete. Then even small changes in temperature may lead to significant loss of flowing properties and may cause proper casting of concrete difficult. However, it should be noticed that temperature is a factor, which up to now has not received the attention it deserves, [1]. The influence of the material temperature on rheology of fresh SCC is investigated rarely, and only limited data are available in literature. The sensitivity of SCC rheological properties to material temperature changes is usually attributed to the properties of cement and superplasticizer. It was demonstrated in [3-6] that the character of changes in yield stress and plastic viscosity of mixtures with temperature showing ambiguous trends depending on superplasticizer and cement type, specific surface of cement, C3A and Na2Oe content in cement and interaction of these factors. Due to the variety of factors and existing interactions, it is difficult to predict the influence of temperature on rheological properties of SCC made with specific components. Therefore, the examination of that influence is recommended as an essential stage of SCC designing and development. Because the material temperature has a significant impact on the assessment of the cement and superplasticizer compatibility, the examination of possible variations in temperature of the mixture should be taken into account. Most of published methods for designing of SCC try to optimise grading envelope and later to optimise the flow and stability of the paste, mortar and fresh concrete successively, [1]. Often after designing the cement paste, a mixture composition is selected. The compatibility of cement and superplasticizer is traditionally tested using cement paste. Therefore, one may say that designing and development of SCC is based on testing the rheology of cement paste. However, it was stated in [7] that the effects of the influence of different technological factors on rheological properties of cement paste, mortar and fresh concrete not always are unequivocal. It was also demonstrated in [8], that it is difficult to predict the effects of the superplasticizer content and its variations with time basing only on tests made on cement paste, and not making the allowance on the factor of aggregate filling by cement paste Mk/z in this mortar or concrete. In the light of necessity for temperature consideration during the selection of superplasticizer and designing of SCC, basic question can be asked: does temperature influence rheological properties of cement paste and SCC mixture similarly? In the paper the investigation of the influence of temperature and time on rheological properties of cement paste and for self compacting concretes are presented and discussed. The rheology of mortars is additionally studied.

EXPERIMENTAL PROGRAM In the first part of the research, the influence of temperature on rheological properties of cement paste with the addition of different superplasticizers was defined in order to select the superplasticizer for next parts of the research. Then, coupled effect of temperature and time on rheological properties of cement pastes different in w/c was determined. These cement pastes were used in the following parts of research. In the second part of research, the influence of temperature and cement paste volume on rheological properties of concrete was investigated. The type of an aggregate and grading were kept constant. Cement paste volume was expressed in terms of factor of uncompacted aggregate filling by cement paste Mk/z. Methods of Mk/z computing are presented in existing literature [2]. In the third part of the research, the influence of temperature on rheological properties of mortars analogical to

37-The effect of.qxd

9/19/2009

12:01 PM

Page 361

361

The effect of temperature on the rheological properties of self - compacting concrete

mortars filling the voids of aggregate of concretes was defined additionally. Experimental program is presented in Table 1. Table 1. Experimental program Material

Temperature o

Cement paste

10, 20, 30 C

Fresh concrete

10, 20, 30 oC

w/b ratio* 0.30 0.30, 0.40

SP type** SP1, SP2, SP3, SP4, SP5 SP1

0.30, 0.40

SP1

Factor Mk/z 1.35, 1.45, 1.55, 1.65, 1.75 -

Fresh mortar 10, 20, 30 oC 0.30 SP1 * b = cement + ground limestone; ** SP according to Table 4.

MATERIALS AND METHODS Measurements of rheological parameters of fresh mortars The measurement of rheological parameters of cement paste has been performed using flow test (cylinder height of 100 mm and diameter of 50 mm). The spread diameter and flow time to diameter of 250 mm were measured. It was proved in existing literature [1,2], that spread diameter and flow time corresponds with yield stress and plastic viscosity respectively. Rheological parameters of mortars were measured using Viskomat PC rheometer detailed described in [9]. The rheological parameters are determined by the regression analysis according to the relation: T=g+Nh

(2)

where T is the shear resistance of sample measured at rotation speed N and g and h are constants corresponding to Wo and Kpl respectively. Measuring procedure was as follow: rotation speed N = 120 rev/min held constant for 3 min and next measurements of shear resistance T at decreasing N from 120 to20 rev/min. Total time of measurement was 260 sec. The measurement of rheological parameters of fresh concrete has been performed using BT2 rheometer. It is described in detail in [9]. To execute the measurement, a sample of testing material is placed in a sample container, the BT2 rheometer is placed in the middle of the measurement container, and subsequently one full turn is performed. During this turn the moment is measured on two probes, as well as angular velocity. On this basis, the values of rheological parameters of the mixture are calculated according to the modified relation (2). Materials and mixes Cement CEM II 32,5R B-S, ground limestone and polyether based superplasticizers were used for the investigations. Their main properties are given in Tables 2, 3 and 4 respectively. Natural sand 0 - 2 mm and crushed syenite aggregate 2 – 8 mm were used. Aggregate grading is presented in Table 5. Mixtures proportioning are presented in Table 6. Table 2. Properties of cement CEM II 32,5R B-S Cement ingredients [%] SiO2 CaO Al2O3 Fe2O3 MgO Na2Oe 24.7 56.7 6.3 2.3 2.9 0.70

SO3 3.2

Specific surface [m2/kg] 325

37-The effect of.qxd

9/19/2009

12:01 PM

Page 362

362

Jacek GO£ASZEWSKI and Grzegorz CYGAN

Table 3. Properties of limestone CaCO3 SiO2 MgCO3 Fe2O3 Al2O3 Na2O 96.0

1.5

1.4

0.1

0.1

Specific surface [m2/kg] 0.03 43.7 226

K2O

0.02

S

0.03

LOI

Table 4. Properties of superplasticizer Superplasticizer SP1 SP2 SP3 SP4 SP5

Density [g/cm3] 1,04 1,04 1,04 1,06 1,08

Major constituent polycarboxylate acid polyether polycarboxylate acid polyether polyether

Concentration [%] 40 18 20 32 18

Table 5. Aggregate grading Sieve size, mm Passing by sieve, %

0.125 0.7

0.25 6.2

0.5 26.8

1 40.6

2 50.1

4 75.9

8 98.6

16 100

Table 6. Mixtures proportioning Component Cement Ground limestone Water SP Aggregate

w/b = 0.30

w/b = 0.40

1.35 486

1.40 497

1.45 507

1.55 526

Factor ĳkz 1.65 1.75 1.35 543 560 367

122

124

127

131

136

1.40 374

1.45 382

1.55 396

1.65 409

1.75 422

140

157

160

164

170

175

181

182 186 190 197 204 210 4.3 4.4 4.5 4.7 4.8 5.0 1549 1525 1501 1457 1415 1377

209 3.0 880

214 3.1 866

218 3.2 853

226 3.3 828

234 3.4 804

241 3.5 782

Mortar mixing and testing procedures All tested mixtures were prepared in order to obtain initial temperature 10, 20 or 30°C. Temperature was kept constant during tests. Cement pastes were prepared using standard mixer after PN EN 196-1:1996. Mixing time was 3 min at rotation speed 140 rev/min. Mortars were prepared after PN EN 196-1:1996, superplasticizer was added with water. Concretes were mixed in pan mixer of volume 50 dm3. Mixing time was 5 min; dry constituents were first mixed during 1 min and next water and superplasticizer were added. For all tested mixtures at least three measurements of rheological parameters have been carried out after 20 and 60 min after the end of mixing.

TEST RESULTS AND DISCUSSION The influence of temperature on rheological properties of cement paste is presented in Fig. 1 and 2. It may be stated that irrespectively of the superplasticizer used, the temperature increase does not influence considerably the slump flow (yield stress) of cement paste. Appearing tendencies are in the range of measurement accuracy. In the same time increasing temperature clearly decreases the flow time (plastic viscosity) of cement paste. Such effect is stronger in the range from 10 to 20oC. For further test SP1 superplasticizer has been selected.

37-The effect of.qxd

9/19/2009

12:01 PM

Page 363

363

The effect of temperature on the rheological properties of self - compacting concrete

As can be seen in Fig. 2 the changes in the slump flow (yield stress) in time of cement pastes with SP1 may be considered as negligible, and independent on temperature. At the same time, the plastic viscosity of these cement pastes decreases with time. The range of such changes is the lowest in 30°C. 430 390

8

SP1

7

SP2 SP3 SP4 SP5

6

370 350

SP1 SP2

330 310

SP3

290

Flow time, s

Slump flow, mm

410

4 3 2

SP4 SP5

270

5

1

250

0 10

20

30

10

o

Temperature, C

20

30 o

Temperature, C

430

8

410

7

390

6

370 350 330 310

20 min

60 min

w/c = 0.30

290

Flow time, s

Slump flow, mm

Fig. 1. Influence of temperature on slump flow (yield stress) and flow time (plastic viscosity) of w/b = 0.30 cement pastes 20 min

60 min

w/c = 0.30 w/c = 0.40

5 4 3 2

w/c = 0.40

1

270 250

0 10

20

30 o

Temperature, C

10

20 Temperature,

30 o

C

Fig. 2. Influence of temperature and time on slump flow (yield stress) and flow time (plastic viscosity) of cement pastes with SP1 superplasticizer The influence of temperature on yield stress g of concrete is presented in Fig. 3 and 4. In the concretes made of w/b = 0.30 cement paste the increase of yield stress g with the temperature increase is observed. Such increase is the higher the less cement paste is in the mixture and more intense in the range from 20 to 30°C. It is noticed that in such case only the mixture containing most of the cement paste maintains appropriate for SCC fluidity at 30°C. The yield stress g shows the higher increase with time the lower content of paste is in the mixture and the higher temperature of mixture is. Such increase is relatively small at 10°C, whereas at 30°C none of tested mixtures kept fluidity required for self compacting concrete.

37-The effect of.qxd

9/19/2009

12:01 PM

Page 364

364

Jacek GO£ASZEWSKI and Grzegorz CYGAN

The character of the temperature influence on yield stress g of w/b = 0.40 concrete is analogous. However, in that case the range of changes of yield stress g as the result of temperature changes is considerably lower. 800

800 ĳkz = 1,35

ĳkz = 1,35

20 min

ĳkz = 1,55

Yield stess g, Nmm

Yield stress g, Nmm

ĳkz = 1,45 600

60 min

ĳkz = 1,45

ĳkz = 1,65 ĳkz = 1,75 400

SCC 200

600

ĳkz = 1,55 ĳkz = 1,65 ĳkz = 1,75

400

SCC 200

0

0 10

20

Temperature,

10

30 O

20

Temperature,

C

30 O

C

Fig. 3. Influence of temperature on yield stress g of w/b = 0.30 self compacting concretes. 20 min

800

800

ĳkz = 1,35

ĳkz = 1,55

ĳkz = 1,55

Yield stress g, Nmm

Yield stress g, Nmm

60 min

ĳkz = 1,45

ĳkz = 1,45

600

ĳkz = 1,35

ĳkz = 1,65 ĳkz = 1,75

400

SCC 200

0

600

ĳkz = 1,65 ĳkz = 1,75

400

SCC 200

0

10

20

Temperature,

30 O

C

10

20

Temperature,

30 O

C

Fig. 4. The influence of temperature on yield stress g of w/b = 0.40 self compacting concretes The influence of temperature on plastic viscosity h of fresh concrete is presented in Fig. 5 and 6. Plastic viscosity h of w/b = 0,30 mixtures increases with the temperature increase. The effect of temperature on plastic viscosity h is the lower the higher volume of cement paste is. It can be noticed that the temperature influences plastic viscosity h in lesser degree than yield stress g. Plastic viscosity h increases with the temperature increase from 10 to 30°C two times, yield stress g at least ten times at the same time. Plastic viscosity h of w/b=0.30 mixtures increases with time, and such increase is the higher the higher temperature is.

37-The effect of.qxd

9/19/2009

12:01 PM

Page 365

365

The effect of temperature on the rheological properties of self - compacting concrete

8000

8000

6000

4000

ĳkz = 1,35 ĳkz = 1,45 ĳkz = 1,55 ĳkz = 1,65 ĳkz = 1,75

2000

Plastic viscosity h, Nmmmin

Plastic viscosity h, Nmmmin

20 min

0

60 min 6000

4000

ĳkz = 1,35 ĳkz = 1,45 ĳkz = 1,55 ĳkz = 1,65 ĳkz = 1,75

2000

0

10

20

30

Temperature,

O

10

C

20

30 O

Temperature,

C

Fig. 5. Influence of temperature on plastic viscosity h of w/b = 0.30 self compacting concretes 8000

8000

20 min

ĳkz = 1,45 ĳkz = 1,55

6000

60 min

ĳkz = 1,35

Plastic viscosity h, Nmmmin

Plastic viscosity h, Nmmmin

ĳkz = 1,35

ĳkz = 1,65 ĳkz = 1,75 4000

2000

0

ĳkz = 1,45 ĳkz = 1,55

6000

ĳkz = 1,65 ĳkz = 1,75 4000

2000

0

10

20

Temperature,

30 O

C

10

20

Temperature,

30 O

C

Fig. 6. Influence of temperature on plastic viscosity h of w/b = 0.40 SCC Plastic viscosity h of w/b=0.40 concretes is considerably lower than that of w/b = 0.30. Generally, the tendency of the plastic viscosity h increase with the temperature increase is retained, however the range of these changes is considerably smaller. Plastic viscosity of w/b = 0.40 mixtures increases with time, but the range of this increase depends on the temperature only to a slight degree. Obtained results show that the character of temperature influence on rheological properties of cement paste and SCC prepared from the analogous cement paste can be significantly different. It suggests, that in the selection of superplasticizer compatible with cement and in designing and developing of SCC mixtures with the consideration of the

37-The effect of.qxd

9/19/2009

12:01 PM

Page 366

366

Jacek GO£ASZEWSKI and Grzegorz CYGAN

temperature, tests performed on the cement paste may be inappropriate. Due to that reason, the influence of temperature on rheological properties of mortars analogous to mortars in SCC was examined. Presented in Fig. 7 and 8 the influence of temperature on rheological parameters of mortars show significant qualitative similarity to those obtained for concrete. It suggests, that the results obtained with mortars are useful for predicting trends in temperature influence on rheology of fresh SCC. However, further tests are necessary, comprising mortars and concretes of different type and proportions of constituents. 50

50

ĳkz = 1,35

ĳkz = 1,45

ĳkz = 1,45 40

ĳkz = 1,55

Yield stress g, Nmm

Yield stress g, Nmm

40

ĳkz = 1,65

30

ĳkz = 1,35

ĳkz = 1,75

20

10

ĳkz = 1,55 ĳkz = 1,65

30

ĳkz = 1,75

20

10

0

0

10

20

Temperature,

30 O

10

C

20

Temperature,

30 O

C

50

50

40

40

Plastic viscosity h, Nmms

Plastic viscosity h, Nmms

Fig. 7. Influence of temperature on yield stress g of mortars analogical to mortars filling the voids of aggregate in w/b = 0.30 self compacting concretes

30

20

ĳkz = 1,35 ĳkz = 1,45

10

ĳkz = 1,55

30

20

ĳkz = 1,35 ĳkz = 1,45

10

ĳkz = 1,65

ĳkz = 1,55 ĳkz = 1,65

ĳkz = 1,75

ĳkz = 1,75

0

0 10

20

Temperature,

30 O

C

10

20

Temperature,

30 O

C

Fig. 8. Influence of temperature on plastic viscosity h of mortars analogical to mortars filling the voids of aggregate in w/b = 0.30 self compacting concretes

37-The effect of.qxd

9/19/2009

12:01 PM

Page 367

The effect of temperature on the rheological properties of self - compacting concrete

367

The analysis of published papers [2,5,6,10] indicates, that dissimilar influence of temperature on rheological properties of cement paste and fresh mortar and concrete may be explained as follows. The influence of temperature on cement paste rheology derived from the competition between adsorption of superplasticizer on cement particles surface and the rate of cement hydration process. Adsorption of superplasticizer on the cement surface increases with increasing temperature, enhancing the steric hindrance effect and slowing down cement hydration due to efficient inhibition of nucleation and growth of hydrate products. Such effect promotes better dispersion of cement particles and enhances fluidity and fluidity retention. On the other side, the increase in temperature accelerates cement hydration process considerably, which contributes to the reduction of the amount of the free water in the mortar and faster burying of superplasticizer in the layer of hydration products. This effect results in decrease of cement paste fluidity. Additionally, it is also suggested in [10] that higher concentration of sulphate ions in the solution at higher temperatures may contribute to shrinking of the steric size of superplasticizer polymers, resulting in decrease of cement paste fluidity. Summarizing, in case of cement pastes made from cements of low initial chemical activity the influence of temperature on rheological properties of the mixture is less intensive, because the increase in temperature does not cause any increase in the hydration process rate of such importance that would outbalance increased adsorption of the superplasticizer. In same cases even enhancing of such cement paste fluidity with increasing temperature can be observed (like in the presented research, see Fig. 2). Simultaneously, increasing temperature causes stiffening of cement paste made of cements of higher chemical activity. In case of fresh concrete besides characterized above effects, the temperature significantly influences the capillary cohesion of mixture. It is shown in [2] that the mixture cohesion depends on water content in mixture, aggregate and binder grading and temperature. Increasing temperature causes the increase of the air pressure in the bubbles trapped in concrete mixture, which leads to an increase in capillary cohesion of the mixture. This effect can result in stiffening of fresh concrete and may considerably decrease its fluidity. The influence of that effect is stronger for mixtures with lower w/c, and/or with lower degree of filling aggregate with the cement paste, and/or lower superplasticizer content. Additionally, the increase of the pressure of cohesion may be significantly intensified by water evaporation from the mixture. The increase of capillary cohesion of mixture with increasing temperature is most likely responsible for different rheological behaviour of cement pastes and fresh mortars and concretes in changeable temperatures.

CONCLUSIONS The temperature increase does not influence considerably the yield stress and decrease plastic viscosity of the tested cement pastes. The self-compacting concrete made with analogous cement pastes show the increase of the yield stress and plastic viscosity with the increasing temperature; the higher the less volume of cement paste is in the mixture. The changes of the yield stress of cement pastes with time may be considered as negligible in the whole range of temperatures. Simultaneously, the plastic viscosity of these cement pastes decreases with time and the range of such decrease is the lowest in 30°C. In case of self-compacting concretes made with analogous cement pastes yield stress and plastic viscosity increase with time considerably, and the range of such increase is the higher the higher mixture temperature is (and the less cement paste is in the mixture). Because character of temperature influence on rheological parameters of cement paste and self compacting concrete may be considerably different, tests made on cement pastes are inappropriate for designing and developing workability of self-compacting concrete in

37-The effect of.qxd

368

9/19/2009

12:01 PM

Page 368

Jacek GO£ASZEWSKI and Grzegorz CYGAN

changing temperatures. Simultaneously, the influence of temperature on rheological parameters of mortars and self-compacting is generally similar for both these materials. Thus, tests made on mortars are necessary for designing self-compacting concrete when the influence of temperature is considered. REFERENCES 1 2 3

4

5 6

7 8

9 10

De Schutter, G, Bartos, P.J.M, Domone, P, Gibbs, J. Self compacting concrete. Whittles Publishing, Dunbeath, 2008 Szwabowski, J. Rheology of mixes on cement binders (in Polish), Wydawnictwo Politechniki ĝląskiej, Gliwice, 1999 Brameshuber, W, Uebachs, S. The influence of the Temperature on the Rheological properties of Self Compacting Concrete. 3rd International RILEM Symposium on SelfCompacting Concrete, Reykjavik, Iceland, 2003, 174-183 Goáaszewski, J. Effect of temperature on rheological properties of superplasticized cement mortars. 8ht CANMET/ACI Conference Superplasticizers and other Admixtures for Concrete, Ed. Malhotra V.M., ACI SP 239, Italy, 2006, 423 - 440 Petit, J-Y, Khayat, K.H, Wirquin E. Yield stress and viscosity equations for mortars and self-consolidating concrete. Cement and Concrete Research, Vol.37, No5, 2007, 655-670 Petit, J-Y, Khayat, K.H, Wirquin E. Coupled effect of time and temperature on variations of plastic viscosity of highly flowable mortar, Cement and Concrete Research, Vol. 39, No. 3, 2009, 165-170 Aïtcin, P-C. High Performance Concrete, EF&N SPON, London, 1998 Goáaszewski, J. The influence of cement paste volume in mortar on the rheological effects of the addition of superplasticizer. 8th International Conference “Brittle Matrix Composites”, Ed. A.M. Brandt, Poland, 2006, 441 - 449 Greim, M. Rheological measurement on building materials, a comprehensive research program. Annual Transactions of the Nordic Rheology Society, Vol. 5, 1997 Nawa, T., Ichiboji, H., Kinoshita, M., Influence of temperature on fluidity of cement Paste containing superplasticizer with polyethylene oxide graft chains, 6th CANMET/ACI International Conference “Superplasticizers and Other Chemical Admixtures in Concrete”, ACI SP 195, Ed. Malhotra V.M., France, 2000, 195 - 210

38-Assessment of the.qxd

9/19/2009

12:02 PM

Page 369

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

ASSESSMENT OF THE RHEOLOGICAL PROPERTIES OF CEMENT MIXES USING ELECTRICAL RESISTANCE Dominik LOGOē Institute of Building Engineering Technical University of Wrocáaw Plac Grunwaldzki 11, 50-372 Wrocáaw, Poland, e-mail:[email protected]

ABSTRACT This paper shows relations between the rheological properties and electrical conductivity of cement mixes. The traditional measurements of the cement mixes consistence give insufficient information about the rheological properties to optimize cement composites. The tests confirmed that the electrical resistance provides information to control and optimize the mixture proportion of cement matrixes with admixtures. The tests gave varying results depending on the manner of measuring the resistance. In order to compare the obtained results in different research centres it is necessary to develop a standardised method. The paper shows the relation between traditional methods of consistence measurement (Novikov and Abrams cone test) as well as Bingham’s rheological parameters (g, h) and electrical resistance. There are no studies proving a correlation between rheological properties and electrical resistance. The existing studies on electrical resistance measurement have mainly referred to the control of hydrated or hardened matrix. It was determined in this paper that electrical resistance of cement mixes can be effectively used to control rheological properties of cement mixes.

Keywords Rheology, electrical resistance, cement mixes, fibres

INTRODUCTION The major problem in cement composites mixes are rheological properties. The traditional measurements of the consistence (slump, Ve-Be, flow table) give insufficient information about them, [1,2,3]. A development of cement composites causes changes in the methods of measurement of rheological properties. The influence of admixtures on the rheological properties of mixes may be determined using Bingham model by two parameters: g-Bingham yield value (Ĳo) and h–plastic viscosity (Șpl), fig.1. Execution of composites with dispersed reinforcement FRC and HPFRC (High Performance Fibre Reinforced Cement Composites) requires proper dispersing of microfibres [4,5,6]. The maximum fibre volume that can be dispersed is related to the initial rheological properties of mixes and can be determined by the rheological tests [7,8,9]. It has been shown that the influence of the fibre content on the mix flowability decreases linearly with increasing fibre content [7,8]. According to Kucharska and LogoĔ [10,11], the linear relation between (Vf - g) ends when properties of the mix are no longer controlled by the matrix but by the fibre interaction. It was found that the fibres influence mainly the g parameter and their influence on plastic viscosity (the h parameter) is not significant.

38-Assessment of the.qxd

9/19/2009

12:02 PM

Page 370

370

Dominik LOGOÑ

Measurement of electrical resistance is a new method of assessing rheological properties of the mix. It can be used to assess mix flowability, adequate volume admixtures or sedimentation of components. The majority of studies focus on testing electrical conductivity at early ages of hydration [12,13] or in hardened cement composites. Electrical conductivity of hardened cement pastes was measured and the composition of the used mineral additives to the cement during hydration was taken into account [14]. It was determined that the relative dielectric constant of cement paste is decreased by silica fume addition and by steel fibre addition [15]. It is increased by latex and by carbon fibre addition due to the interface between cement and these admixtures [15]. Electrical resistance is used to determine the quantity of fibres [16,17,18], and their proper dispersion in hardened cement matrixes as well as to control damage in static and dynamic loading [19]. It is shown also that the addition of a small quantity of carbon fibres to cement mortar specimens can produce a significant increase in the electrical conductivity of the composite material [18]. A review of relevant literature shows that there are no established methods to measure the resistivity of fresh mixes. Quality control of dispersing microfibres in cement pastes using rheological parameters and electrical resistance was presented in [20] and the results confirmed the previous conclusions. The influence of admixtures on the resistance of cement mixes was also observed. The relation was determined between traditional rheological measurements of the consistence (Novikov and Abrams cone test) as well as Bingham parameter (the yield value) and resistivity of the cement pastes, mortars and concretes. There are no studies proving a correlation between the consistence of cement mixes and their electrical resistance, in particular with regard to non-conductive and conductive fibres. In various studies different values of electrical resistance of mixes were observed depending on the size of measuring vessel, distance between electrodes, their immersion etc. The observed differences in measurements indicate a need to develop a standardised method for reliable test results.

EXPERIMENTS Materials Table 1 shows the composition of paste and mortars, table 2 – composition of concretes. The materials for preparation of the cement composites and the symbols used were: - portland cement CEM I 42.5 (c), - silica fume (SF), - superplasticizer (Sp): ViscoCrete 5-600, - coarse aggregate (A) 2-16mm, - sand (s) 0-2mm, - water (w), - carbon fibres, Pitch-based (Kureha): L=3mm, d=0.18mm, - steel fibres (Chircu): L=25mm, d=0.4mm (hooked), - polypropylene fibres (Schomburg-Fibrin23): L=6mm, d=0.18Pm, - fibre volume Vf = 0-3 %. Testing procedures The traditional measurements of the consistence were conducted using Novikov cone test for mortars and Abrams cone test for concrete. The components of pastes and mortars were mixed in a two-speed Hobart mixer. In the preliminary tests quantities of SP were determined and introduced by halves to mixing water

38-Assessment of the.qxd

9/19/2009

12:02 PM

Page 371

371

Assessment of the rheological properties of cement mixes using electrical resistance

and to the paste with a three minute delay. Immediately after mixing the flow curves and values from Novikov cone test were determined. To reduce the influence of uncontrolled factors on the properties of fresh mixes, all steps in the procedure of mix preparation, rheological measurements and sampling were kept the same. Table1: Composition of paste and mortars w/(c+SF)

Paste (P)

sand/cement

SF

Sp

Vf

s/c

[%]c

[%]c

[%]

Mortar (M) P0.3

0.3

-

10

1

0-2.5

M1

0.4;0.5;0.6;0.7

0.33

0

0

-

M0.5

0.5

0.33

10

1

0-3

Table 2: Composition of the concretes Vp - volume of paste increase [%]

Sp

SF

[kg/m3]

[%]c

[%]c

w/c

Concrete

c

C1

0.45; 0.5

350

-

0-5

-

C2

0.45; 0.5

280

34.17-136.7

0

-

Workability of the mixes for pastes and mortars was determined using a viscometer “Viskomat PC” with controlled speed of rotation and measurement of moment due to the resistance of the mix. All test parameters were controlled and recorded. Obtained values of the parameters g and h, corresponding to yield value and plastic viscosity (g - Bingham yield value (Ĳo) and h – plastic viscosity (Șpl)), Fig. 1. The components of concrete were mixed in a concrete mixer. Immediately after mixing the components, slump by Abrams cone test and resistivity were measured.

.

T [Nmm]

Ĳ = Ĳo + Kpl . Ȗ

Ĳo - yield stress [Pa], Șpl - plastic viscosity [Pa . s-1]. T = g + h. N

g tgD h N [min-1]

Fig. 1: Graphic interpretation of the rheological parameters g and h

38-Assessment of the.qxd

9/19/2009

12:02 PM

Page 372

372

Dominik LOGOÑ

For electrical resistance (R) measurements an ELC-3131D meter was used, Fig. 2. This 10,000-count Bench type L/C/R meter is a special microprocessor-controlled meter for measuring functions of inductance/capacitance/resistance. The meter provides direct and accurate measurements of inductors, capacitors and resistors with dual testing frequencies of 120Hz and 1KHz. To measure resistance two copper electrodes were used (I 5 mm). The distance between the electrodes was 40mm. The diameter of the glass cylindrical container where the paste was poured was 92mm, and its depth was 50mm. The container was filled with paste, mortar or concrete up to the upper rim and covered with a plate 5mm thick, and the electrodes were immersed 32mm deep in the mix.

Fig. 2: ELC-3131D meter for measuring functions of inductance (L) / capacitance (C) / electrical resistance (R)

TEST RESULTS Influence of w/c in mortar on electrical resistance (R) and Novikov cone test As Fig. 3 shows, for mortars the influence on the Novikov cone test and electrical resistance is (w/c=0.3-0.7, without superplasticizer) nonlinear. There is approximately linear relation for w/c=0.3-0.5 which corresponds to 0-30mm cone test and, respectively, 1000-180: value of resistivity. For w/c = 0.6 and 0.7 there is an inflection of curve, a decrease of resistivity is slower, respectively, up to 120 and 100: while the cone’s immersion increases, respectively, to 46 and 106 mm. As Fig. 3 indicates also small changes in resistance (of 10-20: with resistivity measured below 150:, correspond to significant differences in the Novikov cone test. Observations during the studies show that inadequate rheological properties (w/c below 0.4 – mortars without Sp) correspond to high resistance values (300-1000: . Values of R over 500: indicate a lack of proper flowability. Resistivity measurements with bad rheological properties of mixes are characterised by considerable scatter of obtained values R.

38-Assessment of the.qxd

9/19/2009

12:02 PM

Page 373

373

Assessment of the rheological properties of cement mixes using electrical resistance

R [:] 450

Mortar M1

400

c - 540kg/m3 c/s =0.333

350 300

w/c = 0.4

250 200

w/c = 0.5

150

w/c = 0.6

w/c = 0.7

100 50 0

20

40

60

80

100

120

Novikov cone test [mm]

Fig. 3: Influence of w/c on electrical resistance (R) and Novikov cone test Influence of Sp and volume paste amount on electrical resistance (R) and slump in (Abrams cone test) The influence of superplasticizer on slump and resistance of concrete mixes w/c = 0.45 and w/c = 0.5 is presented in Fig. 4. Concrete mix with a higher content of water is characterised by a more significant slump and resistance decrease with the same increase of the quantity of superplasticizer. R [:] 450

Concrete C1

400

c -350kg/m3 +Sp (%c )

350

0%Sp

300

w/c =0.5 w/c =0.45

1%Sp

1% 2% 3% 200 4%Sp 250

5%Sp

150

2%Sp 3%Sp

100 50 0

25

50

75

100

125

150

175 200

225

250

Abrams cone test [mm]

Fig. 4: Influence of superplasticizer on electrical resistance (R) and slump (Abrams cone test)

38-Assessment of the.qxd

9/19/2009

12:02 PM

Page 374

374

Dominik LOGOÑ

Fig. 5 presents an influence of the increase of the paste content (in concrete without superplasticizer) for w/c = 0.45 and w/c = 0.5 on slump and resistance. The influence of paste volume increase on the Abrams cone test is nonlinear. There is approximately linear relation for a small quantity of paste volume (not exceeding 34%, which corresponds to 0-20 mm slump and, respectively, 1000-350: value of resistivity). A further increase of the paste volume results in an inflection of curve with a corresponding decrease of resistivity and an increase of slump. As Fig. 5 shows in the case of concrete mixes (similarly to the results obtained for mortars, Fig. 3), small changes in resistivity (10-20: with resistivity measured below 180: correspond to significant differences in the flowability of mixes. R [: ] 450

Concrete C2

400

+34.17% Vp

c -280 kg/m3 +%Vp

350

w/c = 0.5 w/c =0.45

300 250

+68.34% Vp

200

+102.51% Vp

150 100

+136.7% Vp

50 0

25

50

75

100 125 150 175 200 225 250

Abrams cone test [mm]

Fig. 5: Influence of Vp (volume of paste) increase in concrete mix on electrical resistance (R) and slump (Abrams cone test) As Figs. 3, 4 and 5 indicate, inadequate rheological properties of mortars and concretes correspond to R=300-500: Range R = 150-250: can be considered sufficiently good. Acceptable rheological properties correspond to R = 100 -150: Observations showed that resistance R over 500: indicates insufficient rheological properties of cement mixes for appropriate casting and the measurement results varied significantly. The influence of fibres volume on electrical resistance (R) The relation between electrical resistance (R) as a function of the fibre volume V in paste and mortar mixes is shown in Fig. 6., and the relation between the yield value (parameter g) and fibres volume - in Fig. 7. The approximately linear relation R-Vf for pastes is confirmed as it is shown in Fig. 6a [20]. A similar relation was determined for mortars (fig. 6b). Irregularities in that relation R-Vf are been attributed mainly to the entrapped of air into the mix with fibres. As figure 6 indicates, for conductive fibres (carbon, steel) in cement pastes and mortars the relation R-Vf is a linear decreasing function, and for non-conductive (polypropylene) fibres it is a linear increasing function. The end of the linear relation R-Vf indicates a maximum volume of fibres that can be correctly dispersed (Figs. 6-9) and corresponds to the rheological parameter gmax (constant). It has been confirmed that the end of linear relation g-Vf (Fig. 7), indicates that the

38-Assessment of the.qxd

9/19/2009

12:02 PM

Page 375

375

Assessment of the rheological properties of cement mixes using electrical resistance

rheological parameters of the mix are controlled by the fibres and that relation is more significant for pastes than for mortars. R [: ]

R [:]

100

230

Paste P0.3 w/c =0.33

a) 90

Mortar M0.5 w/c = 0.5

b) 200

polypropylene

polypropylene

80

170

70

140

carbon

60

carbon

110

steel

steel 50

80

40

50 0

0,5

1

1,5

2

2,5 V f 3[%]

0

0,5

1

1,5

2

2,5 V [%] 3 f

Fig. 6: Electrical resistance (R) as a function of the fibre volume Vf [%]: a) paste, b) mortar g [Nmm]

R [: ]

400

a)

300

230

Mortar M0.5 w/c = 0.5

b)

Paste P0.3 w/c =0.33

200

polypropylene

steel 170

polypropylene 140

200

carbon

carbon

110

steel

100 80

50

0 0

0,5

1

1,5

2

V f 2,5 [%]

0

0,5

1

1,5

2

2,5 V [%] 3 f

Fig. 7: Relations between the yield value (parameter g) and fibres volume Vf : a) paste, b) mortar The volume of fibres that can be dispersed is determined by rheological parameters of mixes which can be controlled by resistivity measurement Ri (Fig. 9). The conducted measurements

38-Assessment of the.qxd

9/19/2009

12:02 PM

Page 376

376

Dominik LOGOÑ

show that the value Ri should be as low as possible but not causing a sedimentation of components of the mix and should not exceed 150 :.

g [Nmm]

Fibre reinforced cement paste/mortar

gmax

g-constant

gcr

Vcr

Vfmax

Vf [%]

Fig. 8: Rheological parameter g as a function of the fibre volume Vf [%]

R

Fibre reinforced cement paste/mortar

Electrical resistance

non-conductive fibres

Fibres cannot be dispersed Ri conductive fibres

Vfmax

Vf [%]

Fig. 9: Electrical resistance (R) as a function of the fibre volume Vf [%]

38-Assessment of the.qxd

9/19/2009

12:02 PM

Page 377

Assessment of the rheological properties of cement mixes using electrical resistance

377

CONCLUSIONS It was determined that workability of cement mixes can be controlled by the measurement of electrical resistance. Increase of water to cement ratio (w/c) or the volume of superplasticizers result in an increase of flowability and in a non-linear decrease of resistivity of cement mortars and concrete. The relation between resistance (R) and fibre volume Vf in cement pastes and mortars is approximately linear. The resistance decreases for conductive fibres and increases for nonconductive fibres. A rapid increase of electrical resistance for cement pastes and mortars with conductive fibres indicates that the maximum volume of fibres that can be correctly dispersed has been exceeded, which corresponds to gmax parameter (yield value in the Bingham model). The electrical resistance gives possibilities to determine and optimize the proportions of water, superplasticizers, fibres and other admixtures in cement mixes. REFERENCES 1. Banfill, P.F.G., The rheology of fresh mortar. Magazine Concrete Research. 43, No.154, 1991, 13-21 2. Banfill, P.F.G., Rheological methods for assessing the flow properties of mortar and related materials. Construction and Building Materials, 8 (1),1994, 43-50 3. Goáaszewski, J., Szwabowski, J., Influence of superplasticizers on rheological behaviour of fresh cement mortars. Cement and Concrete Research, 34, 2004, 235-248 4. Kucharska, L., Brandt, A.M., Pitch-based carbon fibre reinforced cement composites, in: Proc. Materials Engineering Conference ASCE, Materials for the New Millenium, Ed. K.P. Chong, Washington 1996, 1, 1271-280 5. Nishioka, K., Yamakawa, S. and Shirakawa, K., Properties and application of carbon reinforced cement composites, in: FRC-86, RILEM Symp. Developments in Fibre Reinforced Cement and Concrete, 13-17, Sheffield 1986, 95-104 6. Naaman, A.E., Reinhardt, H.W., Characterization of high performance fibre reinforced cement composites-HPFRCC: 2nd International Workshop, v.2 (HPFRCC-95), Ed. A.E. Naaman and H.W. Reinhardt, RILEM, 1995, 1-21 7. Ando, T., Sakai, H., Takahashi, K., Hoshijima, T. Awata, M. Oka S., Fabrication and properties for a new carbon fibre reinforced cement product. ACI SP 124.1990, 40-60 8. Banfill, P.F.G., Starrs, G., Derruau, D., McCarter, W.J., Chrisp, T.M., Rheology of low carbon fibre content reinforced cement mortar. Cement and Concrete Composites 28, 2006,773-780 9. Park, C.K., Noh, M.H., Park, T.H., Rheological properties of cementitious materials containing mineral admixtures. Cement and Concrete Research, 35, 2005, 842-849 10. Kucharska, L., LogoĔ, D., The influence of fly ash on rheological and mechanical properties of cement mortars reinforced with pitch-based carbon fibres. Brittle Matrix Composites 5, Ed. A.M. Brandt, V.C.Li and I.H. Marshall, Woodhead and Bigraf, Cambridge and Warsaw, 1997, 113-122 11. Kucharska, L., LogoĔ, D., Relation of the mechanical properties of HPFRC on the rheological parameter of cement pastes and mortars. Proc. Int. Conf. on durability of highperformance concrete, Essen 23-24 sept. AEDIFICATIO, Ed. M.J. Setzer and S. Palecki, 2004, 107-117

38-Assessment of the.qxd

378

9/19/2009

12:02 PM

Page 378

Dominik LOGOÑ

12. Salem, Th.M., Electrical conductivity and rheological properties ot ordinary Portland cement-silica fume and calcium hydroxide-silica fume pastes. Cement and Concrete Research 32, 2002, 1473-1481 13. Xiao, S., Li, Z., Wei, X., Selection of superplasticizer in concrete mix design by measuring the early electrical resistivities of pastes. Cement and Concrete Composites 29, 2007, 350-356 14. Boychuk, W., Giergiczny, Z., Electrical conductivity of fresh cement pastes (in polish). VII Sympozjum Naukowo-Techniczne Reologia w Technologii Betonu, GóraĪdĪe, June 2005, 107-112 15. Wen, S, Chung D.D.L., Effect of admixtures on the dielectric constant of cement paste. Cement and Concrete Research 31, 2002, 673-677 16. Cao, J., Chung, D.D.L., Improving the dispersion of steel fibers in cement mortar by the addition of silane. Cement and Concrete Research 32, 2001, 1473-1481 17. Chen, B., Liu, J., Wu, K., Electrical responses of carbon fiber reinforced cementitious composites to monotonic and cyclic loading. Cement and Concrete Research 35, 2005, 2183-2191 18. Chiarello, M., Zinno, R., Electrical conductivity of self monitoring CFRC. Cement and Concrete Composites 27, 2004, 463-469 19. ASTM D., Test method for fiber content of unidirectional fiber-resin composites by electrical resistivity. 1980,790-84a 20. LogoĔ, D., Quality control of dispersing microfibres in cement pastes using rheological parameters and electrical resistance. Non-Traditional Cement & Concrete III, Brno 2008, 470-476

39-Experimenta.qxd

9/19/2009

12:02 PM

Page 379

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

EXPERIMENTAL TECHNIQUES FOR MULTI-SCALE CHARACTERIZATION OF MECHANICAL RESPONSE IN CEMENT-BASED MATERIALS Joseph J. BIERNACKI Department of Chemical Engineering Tennessee Technological University Box 5013, Cookeville, Tennessee 38505, e-mail: [email protected]

ABSTRACT Multi-scale phase-resolved computational modeling approaches are far in advance of supporting experimental methodologies for predicting the mechanical response of cement-based materials. Recently, however, a variety of techniques including low-energy x-ray diffraction, neutron diffraction and photostimulated luminescence spectroscopy (PLS) have been applied to cement-based materials demonstrating that these techniques are viable multi-scale probes for the study of micro- and meso-scale mechanics of portland cement. Stress-strain responses for applied loading were demonstrated using both x-ray and neutron diffraction by utilizing the native portlandite in neat portland cement paste as in situ strain sensors. Likewise, micrometer-scale particles of aluminum oxide (corundum) were used as strain gauges in neat cement paste to illustrate PSL response to mechanically applied loads. This technique utilizes the strain sensitive response of the luminescence spectrum of Cr+3 doped corundum. Individual micrometer-sized particles were shown to respond to applied loads making it possible to directly study the transference of mechanical stresses at the interface of individual grains. This suite of new tools can now be used in conjunction with modeling to further the understanding of multi-scale mechanical response and prediction of macroscopic properties and performance of cement-based systems.

Keywords X-ray diffraction, strain measurements, micro-mechanics

BACKGROUND AND MOTIVATION Portland cement concrete is a complex, multi-scale, multi-phase material. Unfortunately, a complete theoretical description of its mechanical behavior does not exist as compared to the body of knowledge, describing the behavior of metals or polymers. None-the-less it is well known that macroscopic failure modes in portland cement-based materials have micro-scale origins [1]. Furthermore, phase-resolved and multi-scale simulations are beginning to populate the literature [2, 3]. A general lack of experimental data, however, on micro- and even mesoscale mechanics is conspicuously absent as techniques have either not been developed or have not been ported for use with portland cements-based materials. Since portland cement phases are inherently distributed on the micro-scale, it is impractical to utilize techniques such as attached strain sensors, though breakthroughs in sensor technology have miniaturized such devices considerably in the past decade. It was the ambition of this research to abandon the traditional methods since they are largely limited to macro and

39-Experimenta.qxd

380

9/19/2009

12:02 PM

Page 380

Joseph J. BIERNACKI

mesoscopic measurements and rather to focus on techniques that would permit direct measurement of in situ strains by utilizing signals from the phases native to the cement paste or mortar or from a suitable surrogate that could harmlessly be added. Among an array of techniques that could be used, several presented unique opportunities, these included diffraction based methods, i.e. x-ray and neutron diffraction, photo-stimulated luminescence spectroscopy (PSLS) and direct image-based methods. A number of prior studies suggested that diffraction might be used for measuring in situ strains in cement-based systems. Rybakova, et al and Shchukin, et al reported the development of hydration strains inferred from diffraction of barium sulfate (BaSO4) crystals introduced as strain sensors to neat portland cement paste [4, 5]. Their use of laboratory x-rays (Cu KD), their choice of low angle reflections and absence of statistical analysis and details of their method makes their findings somewhat questionable, though they do report compressive stains on the order of 1/1000 which seem plausible. Schulson, et al [6, 7,] demonstrated that stresses due to thermal loads could be observed as strains using neutron diffraction, however, they were unable to quantify their results due to lack of instrumental resolution. Finally, Hyurutnan, et al observed a compressive strain on the order of 1/1000 and interpreted them as drying related strains in the CH phase while attempting measurement of crystal orientation at the fracture interface between paste and aggregate [8]. While none of these studies provided a definitive demonstration of a diffraction-based method, they all suggested that such a method may be resolved enough for making strain measurements on the order of 1/1000 a value that is near the tensile strain to failure for CH and other portland cement phases. Asmus et al offers the only known publication wherein photo-stimulated luminescence spectroscopy has been applied to cements. Here the authors used Cr+3 doped aluminum oxide particles as in situ strain sensors [9]. This work illustrates that strain sensitivity and signal intensity of nano-dispersed alumina particles can be used as strain sensors. Unfortunately, their mortars contained as much as one part nano-alumina to four parts cement by weight, an amount that surely perturbs the behavior of the composite. None-the-less, this first demonstration of the technique provided a baseline for the present research. Image-based strain measurements have been the focus of considerable research including metals, polymers and cement-based materials. The techniques for analyzing images are similarly numerous, however, digital image correlation (DIC) has been applied by researchers to the study of cementitiuos systems [10]. Strain measurements on portland cement-based systems, however, have been limited to high strain environments, i.e. the study of crack formation and propagation [11, 12] and the study of drying shrinkage [13]. Such have demonstrated the use of both optical and electron microscope-based imaging.

SUMMARY OF EXPERIMENTAL TECHNIQUES The experimental effort was categorized according to technique and so was divided into: (1) diffraction methods (X-ray and neutron techniques); (2) photo-stimulated luminescent spectroscopy (PSLS); and (3) image-based methods. The array of samples used was likewise varied. All samples were prepared by hand mixing water and cement in a ratio of 0.45. Quartz (Q) mortars used a 2:1 aggregate to cement ratio and alumina (A) mortars were prepared with 1 weight percent aggregate. Samples were typically cured at between 25 and 35 oC. Curing procedures utilized for specific experiments will be described below. The cement was an ordinary Type-I portland cement, the Quartz was a high

39-Experimenta.qxd

9/19/2009

12:02 PM

Page 381

Experimental techniques for multi-scale characterization of mechanical response in cement-based materials

381

purity natural sand, ball milled to an unspecified fineness. Two alumina sources were used, a Sumitomo Chemical AK 502 powder used as received and a Coors 99.8% densified aluminum oxide. All samples were prepared using distilled water except for samples to be used for neutron diffraction which were prepared using D2O (heavy water, deuterium oxide). Prisms were cut from a nominal 2.5×5 cm billet. Prisms prepared with D2O were cut under saturated Ca(DO)2 (calcium deuteriate, CD) and then returned to a saturated CD solution for storage until use. The orientation of the prisms cut from the billet was not recorded. Details of all the experimental procedures and results can be found elsewhere [14, 15, 16, 17]. The following is intended to summarize the methods and outcomes only. Some of the present results have yet to be published [18, 19, 20, 21, 22]. Diffraction Coherent diffraction from crystalline materials, i.e. x-ray and neutron diffraction from single crystals or ensembles of crystals (powders), offers a direct measure of the distance between atoms in the lattice via Bragg’s Law: nO 2dSin(T ) , where n is an integer, O=the wavelength of the diffracting energy, d=the distance between the atomic planes responsible for diffraction at angle T. When a macroscopic object, such as a piece of hydrated portland cement (cement paste) or mortar is subjected to some form of mechanical stress, the observed deformation, strain, must ultimately propagate to atomic displacements. In practice, diffraction measurements are capable of resolving strains as small as 100 ppm (standard deviation~10 ppm). While the details of such methods can be somewhat cumbersome, the concept is simple, measure the location of a diffraction peak under condition 1, i.e. no load, measure the location of a diffraction peak under condition 2, i.e. while applying load, determine the peak position for a suitable no strain condition and reference all stains to this state. Strains (H) for conditions 1 and 2 and other stress d0 d states are thus computed by: H , where d=the d-spacing under some experimental d0 condition(s) and do=the d-spacing for some reference state, ideally a suitable no-stress state. Low energy synchrotron diffraction was carried out at the NSLS beam-line X14A using a reflected beam geometry and O=0.15467 nm. Neutron diffraction (transmission geometry, O=0.19201 nm) at the NIST Center for Neutron Research (CNR) beam-line BT8. Both X14A and BT8 were equipped with a point detector. Since X14A was configured in a reflection geometry, samples were scanned at various \ angles (tilt angles) and the Sin2(\) method of analysis was applied. In this case Equation (2) above is a simplification since the tensoral nature of strain must be considered. Detailed descriptions of this method have been reported in numerous texts and papers [23, 24]. Photo-luminescent spectroscopy When substituted for an aluminum atom in the aluminum oxide crystal lattice, trivalent chromium ion (Cr+3) is know to luminesce intensely in the visible spectrum, e.g. Cr+3 substituted in the Al2O3 lattice has two luminescence peaks, 14400 cm-1 (R1) and 14430 cm-1 (R2). This phenomenon is strain sensitive, as luminescence spectral energies are associated with the electronic configurational geometry and can be resolved with sufficient precision to permit strain determination as small as 100 ppm in some cases. The luminescence spectra of Cr+3 doped alumina has been well characterized and thus alumina particles were used as strain sensors in low volume alumina-based mortars. Strains were computed by measuring the shift in the R2 peak of the Cr+3 ion since it has been better characterized in the literature. The incident radiation was a

39-Experimenta.qxd

9/19/2009

12:02 PM

Page 382

382

Joseph J. BIERNACKI

5145.32 Å, 10 mW, argon laser with a spot size of about 5 Pm. Peak shifts were directly 1 converted into stress using the following relationship: 'Q A 3V A , where 'Q A is the mean 3 observed strain-induced shift in wave number, V A is the mean stress experienced by the randomly oriented crystallites and 3 is the overall piezospectroscopic coefficient. Ma and Clark [25] report that for uniaxial loading, 3 =0.00762 MPa-1cm-1. Image-Based The naked eye cannot discern strains as small as those expected in cement paste, mortar or concrete when subjected to loads below the yield stress of the materials. Digital image correlation, however, may have suitable resolution to discriminate minute strains. This hypothesis was tested using cement paste and various mortars. Images were collected with a Nikon Eclips LV150 compound microscope at a nominal magnification of 20×. Digital Image Correlation [26] was used to post processes images of samples under various load conditions. DIC software is well developed and readily available commercially and so neither the theory of image correlation nor software developments were objectives of the present study. Load Sequences In cases wherein the response to applied loading was studied, loads were applied with a 4470 N (1000 lbf) gear driven frame. Loads were typically applied in sequence of increasing load flowed by steps of decreasing load in the reverse sequence, returning to the zero load state, i.e. 44.7 N. The specific load sequences will be mentioned in the discussion where the sequence is relevant to the interpretations.

FINDINGS

The findings are summarized below first by technique and then discussed as a collective body of knowledge. Diffraction A series of experiments were conducted in an effort to test the hypothesis that in situ phase resolved strains could be measured directly without the introduction of an instrument-based sensor, e.g. by using the response of the material itself and in particular being able to distinguish strains in individual phases. Both x-ray and neutron diffraction were considered. Synchrotron X-rays with a wavelength of 1.5467 Å (an energy of 8.0161 keV, roughly that of Cu KD radiation) were used. Prisms measuring nominally 1 cm × 1.5 cm were used. Such were cut from a larger billet as described above. The parent materials were cured at 35 oC for at least 28 days prior to cutting, however, the cure time was not strictly controlled in these experiments and could be as long as several months. Details of the experimental set-up are provided elsewhere, in summary however, the samples were loaded in a cycle of 0.43, 8.6, 0.43, 17.2, 0.43, 25.9, 0.43, 34.5 and 0.43 MPa. For each load cycle the \ (tilt) angle was held contend. X-ray scans from 129.5o to 145.5o 2T were made at tilt angles of 0o, 20.7o, 30o, 27.7o, 45o, 52.5o and 60o. The peak position for the (214) (nominal d-spacing of 0.85 Å) and (312) (nominal d-spacing of 0.81 Å) reflections of calcium hydroxide were used as strain sensors.

39-Experimenta.qxd

9/19/2009

12:02 PM

Page 383

Experimental techniques for multi-scale characterization of mechanical response in cement-based materials

383

Figure 1 typifies the results, illustrating that strains as small as 1/10,000 can be confidently discerned. Here, Poisson’s effect (tensile behavior due to expansion in the direction orthogonal to the applied load) is seen for strains below tilt angles with a Sin2\ less than about 0.25. Compressive behavior is noted above this point, in response to the uniaxial loading. Pre load and post load strains are also plotted suggesting subtle variations in no-load stain states likely due to plastic matrix and associated stress relaxation. Similar datasets at higher load (shown and discussed in detail elsewhere) also suggests increased evidence of plastic behavior indicated by an increasingly non-linear response in both loaded and un-loaded states. 3.0E-04 2.0E-04 1.0E-04

Strain

0 -1.0E-04 -2.0E-04 -3.0E-04 -4.0E-04

62.5 psi

-5.0E-04

3750 psi

-6.0E-04

62.5 psi

-7.0E-04 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

2

Sin +

Figure 1. Strains in the (214) CH plane as measured using the Sin2\ technique and low energy synchrotron diffraction [14]. Similar results were also found for mortars made with finely dispersed quartz sand, further demonstrating that the technique can be extended to look at other phases native to cement-based materials. Low energy X-rays proved to be a reliable and sensitive near surface probe. This technique, however, is limited to near surface behavior due to the shallow depth of penetration of the low energy X-rays and requires tilting due to the reflection geometry of the experiment. Furthermore, highly resolved strain measurements can only be made when using very high angle reflections, necessitating the interpretation of a limited number of acceptably strong, yet relatively weak, high angle peaks, thus limiting the choices to a few CH reflections. Finally, strains in the axial direction, cannot be directly observed since tilting to \=90o is impossible, thus, axial strains must be determined by extrapolation of Sin2ҏ\ to 1.0. Some of these limitation were overcome by using neutrons. Figure 2 illustrates typical data for neutron-based strains measured in transmission mode. Due to the high energy of the neutron beam, sufficient transmitted energy could be detected to reliably define the Bragg peaks. This technique, however, also required high angle reflections for strain resolution. When using transmission geometry, tilting is not necessary since the axial strain (strain in the direction of the applied load) can be directly observed. It is notable that coherent neutron scatting requires the use of deuterium oxide (D2O, heavy water) and that samples were prepared long in advance of the experiment date, some year or more in some cases. Such were cut from larger billets as were the X-ray samples however,

39-Experimenta.qxd

9/19/2009

12:02 PM

Page 384

384

Joseph J. BIERNACKI

these were kept under heavy water until exposure to the neutron beam to prevent exchange of D2O with H2O. Though the strain response observed with neutrons is similar to that observed with Xrays, there is a significant and largely unexplainable compressive strain off-set on the order of 1/1000. Since it would be expected that large strains would eventually be relaxed by creep effects, the presence of enduring compressive or tensile residual strains of a significant magnitude remains to be further investigated. 4000

0

Strain - Unloaded

3000

Load

2500

-0.001

2000 -0.0015

1500

Load (N)

Axial Lattice Strain

3500

Strain - Loaded

-0.0005

1000 -0.002 500 -0.0025

0 0

2

4

6

8

10

Load Sequence Step

Figure 2. Neutron-based strain measurements of the (211) CH reflection [17]. Similar, but less extensive studies were done with fine quartz mortars. Both low energy synchrotron X-rays and neutrons were again used. The results, shown elsewhere [15, 1717], show that strains in a suitably fine distributed aggregate phase can confidently be observed. Surface strains (low energy X-rays) and bulk strains (neutrons) suggest the influence of both plastic deformations and strain relaxation. For example, highly non-linear Sin2\ plots from surface X-rays indicate extensive plastic interactions between the aggregate and matrix. Intermediate no load states also indicate the build-up of incomplete strain relaxation in the form of increasingly tensile no load states, though relatively small. These results suggest a new methodology to study micromechanical, phase resolved interaction in portland cement based materials. Photo-stimulate luminescence spectroscopy The illuminated area in this case was on the order of a few square microns, much smaller that an embedded individual alumina particle. This small beam footprint makes PSLS a true micronscale probe. Unfortunately, this probe, like low energy X-rays, is limited to near surface measurements and other problems such as localized laser heating. A study was performed to characterize the applicability of this technique for determination of within particle strains for very low volume fraction mortars seeded with aluminum oxide particles. Figure 3 illustrates a typical response to a loading sequence. Here, a 1 cm × 1.5 cm prism was loaded at a rate of 1341 N/min to 3576 N, then unloaded at –1341 N/min. The response is clear and systematic as expected

39-Experimenta.qxd

9/19/2009

12:02 PM

Page 385

Experimental techniques for multi-scale characterization of mechanical response in cement-based materials

385

illustrating a compressive stress. Here, peak shifts are converted directly to stress using Equation (3). 20

Uniaxial Stress (MPa)

0 -20 -40 -60 -80 -100 0

200

400

600

800

1000

Time (s)

Figure 3. Typical expression of within particle stress experienced as a result of loading as determined from PSLS Cr+3 doped alumina R2 peak shift. 5002.2 5002

Wave Number

5001.8 5001.6 5001.4 5001.2 5001 5000.8 5000.6 0

20

40

60

80

100

Position (Pm)

Bulk Monolithic Sintered Dense Alumina Sintered Dense Particles in Cement Agglomerated Particles in Cement

Figure 4. Comparison of residual stresses within various alumina samples illustrating that nanoparticulate agglomerates have a uniform stress distribution compared to dense sintered alumina particles. The as calculated stress represents the mean stress experienced on randomly oriented alumina crystals within the target particle. If the particle were a monolithic, theoretically dense polycrystalline mass, and if the modulus were known, then the strain could be determined within that particle. It will be shown here, that some or all of this information is not available, yet the calculated stress must be the average stress experienced by the luminscing alumina grains since the observed shift is directly correlated to the experienced crystallite stress, much as the X-ray peak shift is directly related to d-spacing strain.

39-Experimenta.qxd

386

9/19/2009

12:02 PM

Page 386

Joseph J. BIERNACKI

Two types of alumina particles were considered as aggregate strain sensors: (1) lightly sintered nano-particle agglomerates (Sumitomo Chemical AK 502 powder) and (2) high density alumina particles made by micronizing a dense sintered body (pulverized Coors 99.8% alumina). Figure 4 illustrates the residual stress states, here shown as variation in wave number as a function of location within a particle of nominally 100 I m in diameter. The results illustrated in Figure 4 clearly show that densified monolithic aluminum oxide particles contain a significantly higher level of residual stress than do the lightly sintered nanoparticle agglomerates. Furthermore, the variation in wave number within the densified particles is on the order of that expected for applied-stress induced peak shift, thus, densified alumina particles were rejected as in situ strain sensors for the present research. Figures 5a and 5b shows that on the micro-scale, the nano-particle agglomerates are less dense as seen by a darker gray-scale in backscatter image. Higher magnification, Figure 5c, reveals that the agglomerate is actually a low density mass of nano-scale particles connecting larger islands of more dense particles on the order of 1 Pm in size. This microstructural analysis offers one possible reason for the difference in variation in residual stress levels within these two particle forms. Individual grains that make up the densified alumina particles are as much as an order of magnitude larger than those that make up the low density agglomerate and approach the size of the laser footprint. Thus, variations observed may reflect grain orientation as well as grain-to-grain variation in residual stress. Since the densified alumina is sintered from powder, grain interactions likely contribute to the observed residual stresses.

(a) (b) (c) Figure 5. Various SEM images illustrating the microstructure of: (a) an alumina nano-particle agglomerate, (b) a dense alumina particle and (c) the low density nano-structure of the alumina nano-particle agglomerates. On the contrary, the individual particles that make up the lightly sintered nano-particle agglomerates appear randomly oriented with respect to the much larger incident laser beam and since grain-to-grain interactions are limited, there is less apparent point-to-point variation in apparent residual stress. This feature of the agglomerate made it the more attractive material for this research and thus was used in all additional studies. Finally, Figure 6 illustrates the type of response found when various particles were surveyed at constant load. Particles of various size, orientation (when shape is other than nominally circular) and association with other matrix features, i.e. cracks as illustrated in Figure 5a, manifest with a different apparent stress (peak shift). The original objective of this work was to demonstrate and assess PSLS as a potential method for doing true micro-scale particle-based micromechanics. It appears that PSLS is a

39-Experimenta.qxd

9/19/2009

12:02 PM

Page 387

Experimental techniques for multi-scale characterization of mechanical response in cement-based materials

387

suitable tool, however, there are limitations including surface accessibility only, the concern over laser heating and interpretation of the observed stresses. Image-Based Image-based methods were found to be the least tractable means for investigating the mechanical behavior of cement-based materials since the objective was to study the behavior prior to exceeding the yield stress of the material. Notwithstanding, image-based methods are well developed and have successfully been used to study post crack propagation and drying shrinkage, wherein large deformations, comparatively, are observed.

0.00018

Normalized Peak Shift

0.00016 0.00014

Particle B

0.00012

Particle F

Particle D

0.0001 0.00008 0.00006 0.00004 0.00002 0 0

200

400

600

800

1000

1200

-0.00002 -0.00004

Time (s)

Figure 6. Stress in various particles reported here as uncalibrated, normalized peak shift [17]. It should be noted that whereas diffraction-based methods are a direct measure of strain in a given phase and PSLS provides the stress in an embedded surrogate (the Cr+3 doped alumina aggregate), digital image correlation provides displacements. With sufficient image resolution to define individual particles and where the modulus of those particles are known, in theory, it is possible to extract strains. Due to mechanical anisotropy in some of the phases in hydrated cement, i.e. CH, and the lack of knowledge about crystal orientation, it is not presently possible to extract phase resolved strains. Furthermore, it was concluded by the present study that the current state-of-the-art in DIC is not sufficiently resolved to reliably extract even displacement data for loads below the yield stress for portland cements wherein significant creep deformations are not present. Despite the remarks above which indicate the DIC is unreliable for small (elastic) deformations in cement-based systems, limited success was, none-the-less, experienced. Figure 7 illustrates a typical load/no-load sequence suggesting that displacements are easily measured. The conclusion that such is reproducible, as suggested by the small error bars, is misleading. While on some occasions, such observations could be made, reproducibility problems were encountered as well as unphysically low Poisson’s ratios.

39-Experimenta.qxd

9/19/2009

12:02 PM

Page 388

388

Joseph J. BIERNACKI

890

44.5

1179

44.5

2669

44.5

3558

44.5

0 -0.002

Pixel Displacement

Load Sequence No. 1 -0.004

Load Sequence No. 2 Intermediate No-Load States

-0.006

Intermediate No-Load States

-0.008 -0.01 -0.012 Applied Load (N)

Figure 7. Displacements in pixel fractions as a function of load, here showing the loaded deformations and unloaded intermediates in a load/no-load sequence. Here, error bars represent displacement variations observed for replicate images of the same region wherein the sample was displaced under no load so that the DIC software would be forced to map identical regions subject to translation (and possibly minor rotation) only. The results are, however, encouraging, in that they suggest that such a method can be resolved enough to observe displacements that are of the order of those seen in cementitious system prior to the yield point. Details of this study can be found in Batiste [16].

DISCUSSION AND SUMMARY OF OUTCOMES

Three classes of techniques were investigated for the determination of strains in portland cement based materials: (1) diffraction, (2) photostimulated luminescence spectroscopy and (3) direct imaging. By far, the most reliable and sensitive technique is diffraction wherein the method was demonstrated using low and high energy X-rays and neutrons. Strains as small as 1/10,000 were shown to be discernable from diffraction data with reasonable statistical confidence (errors on the order of 1/100,000). This technique was suitably demonstrated and can now be applied to a wide range of applications including applied loading, drying shrinkage and autogeneous hydration effects as well as for neat cement and fine aggregate quart-containing mortars. A number of immediate applications for the diffraction-based techniques come to mind. Dynamic mechanical effects such as creep and associated relaxation phenomena may be studied using CH or a similar native phase as an in situ strain sensor in bulk materials (using high energy X-ray or neutrons) or at the free surface (using low energy X-rays). Drying strains and concomitant creep effects may be studied in the same manner. CH crystals appear to have some form of inherent apparent strain at early ages, possibly less than 10 d, whereas thereafter, a component of mechanical strain can be isolated. Both the mechanical and inherent strains appear to be tensile. While the early age inherent strains appear to vary anisotropically and possibly with the hydration environment (availability of water or similar), the mechanical strain component appears to be small and tensile. Such may lead to experiments based on diffraction

39-Experimenta.qxd

9/19/2009

12:02 PM

Page 389

Experimental techniques for multi-scale characterization of mechanical response in cement-based materials

389

that help to unify and elucidate hydration kinetics and early age autogeneous effects such as plastic shrinkage cracking and self desiccation. PSLS is a promising method, and while the R2 shift in Cr+3 doped aluminum is pronounced and extremely strain sensitive, as an in situ cement-grain sized strain sensor, the range of applications may be limited. It appears that rigid, low bulk density alumina nanoparticle agglomerates are effective stress sensors that may be applied to study matrix behavior by inference however, the technique requires considerable development at this stage prior to finding applications. While stresses can be seen, the interaction between the matrix and the aggregate particle is extremely variable. It may be that this variation is, however, the factor to be studied in that it may reveal new insights into how aggregate, i.e. residual unreacted cement particles, interact with the matrix. Finally, direct imaging is the least reliable method of the three explored. While the present results suggest a promising method may exist, difficulty reproducing results day-to-day leads to the general conclusion that this method is presently limited to the study of events having large strains, i.e. drying shrinkage, post yield fracture, crack growth and opening, etc. ACKNOWLEDGEMENTS

The author would like to acknowledge the National Science Foundation (NSF) for support through Grant No. CMS-02346161. Supplemental support from the Tennessee Technological University (TTU) Center for Manufacturing Research (CMR) is also gratefully acknowledged. The author would also like to acknowledge the input of numerous colleagues at the National Laboratories including those at Oak Ridge National Laboratory (ORNL) C. Hubbard, T. Watkins, M. Lance and J. Bai (also at the University of Tennessee at Knoxville); and the National Institute of Standards and Technology (NIST) T. Gnaeupal-Harold. Finally, the contribution of the many students that contributed to the research is acknowledged: S. Mikel (MS 2007), J. Batisti (MS 2006), C. Parnham (MS 2005), Jennifer Sinopoli (REU 2006, Vanderbilt University, Chemical Engineering), Ruoya Wang (REU 2005, Georgia Institute of Technology, Physics/Mecahnical Engineering) and Matthew Turner (REU 2004, Clemson University, Civil Engineering). REFERENCES

1. S. P. Shah, S. Swartz and C. Ouyang, “Fracture Mechanics of Concrete: Applications of Fracture Mechanics to Concrete, Rock and Other Quasi-Brittle Materials,” Wiley, New York, p. 592, 1995. 2. E. Schlangen and E. Garboczi, “Fracture Simulations of Concrete Using Lattice Models: Computer Aspects,” Eng. Fract. Mech. vol 57, no. 2, pp. 319-332, 1997. 3. E. Shlangen and E. Garboczi, “New Method for Simulating Fracture Using an Elastically Uniform Random Geometry Lattice,” Int. J. Eng. Soc. vol. 34, no. 10, pp. 1131-1144, 1996. 4. L. M. Rybakova, E. A. Amelina, I. Kuksenova, and E. D. Shekukin, “Investigation of Residual Internal Stresses of the I and II Modes in Cement-Hardening Structures,” Colloids Surfaces A, vol. 160, pp. 163-170, 1999. 5. E. D. Shchukin, L. M. Rybakova, L. I. Kukenova, and E. A. Amelina, “X-Ray Diffraction Method for the Determination of Residual Stresses in Cement, “ Kolloidn, Zh., vol. 59, pp. 96-101, 1997. 6. E. M. Schulson, I. Swainson, T. Holden and C. J. Korhonen, “Hexagonal Ice in Hardened Cement,” Cem. Concr. Res., vol. 30, pp.191-196, 2000. 1 Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors(s) and do not necessarily reflect the views of the National Science Foundation.

39-Experimenta.qxd

390

9/19/2009

12:02 PM

Page 390

Joseph J. BIERNACKI

7. I. Swainson and E. M. Schulson, “A Neutron Diffraction Study of Ice and Water within a Hardened Cement Paste During Freeze-Thaw,” Cem. Concr. Res., vol. 31, pp. 1821-1830, 2001. 8. V. Harutyunyan, E. Abovyan, P. Monteiro, V. Mkrtchyan, M. Balyan and A. Aivazyan, “XRay Diffraction Investigations of Microstructure of Caclium Hydroxie Crystallites in the Interfacial Transition Zone of Concrete,” J. Am. Ceram. Soc., vol. 86, no. 12, pp. 2162-2166, 2003. 9. S. M. F. Asmus, G. Pezzotti, “Analysis of microstresses in Cement Paste by Flourescence Piezospectroscopy,” Phys Rev E, vol. 66, pp. 1-4, 2002. 10. Y. Xi, T. B. Bergstrom and H. M. Jennings, “Image Intensity matching Technique: Application to the Environmental Scanning Electron Microscope,” Comp. Mat. Sci., vol. 2, pp. 249-260, 1994. 11. S, Choi and S. P. Shah, “Measurement of Deformations on Concrete Subjected to Compression Using Image Correlation,” Experimental Mechanics, vol. 37, no. 3, pp. 307313, 1997. 12. J. S. Lawler, D. T. Keane, and S. P. Shah, “Measuring Three-Dimensional Damage in Concrete Under Compression,” ACI Mat. J., vol. 98, no. 6, pp. 458-464, 2001. 13. C. M. Neubauer, T. B. Bergstrom, K. Sujata, Y. Xi, E. J. Garboczi, and H. M. Jennings, “Drying Shrinkage of Cement Paste as Measured in an Environmental Scanning Electron Microscope and Comparison with Microstructural Models,” J. Mat. Sci., vol. 32, pp. 64156427, 1997. 14. J. J. Biernacki, C. J. Parnham, J. Bai, T. Watkins and C. Hubbard, “Phase resolved Strain Measurements in Hydrated Ordinary Portland Cement Using Synchrotron X-Rays,” J. Am Cer. Soc., vol. 89, no. 9, pp. 2853-2859, 2006. 15. C. J. Parnham, “Strain Measurements in Cement-Based Materials Using Synchrotron XRays,” MS Thesis, Tennessee Technological University, p. 133, 2005. 16. J. W. Batiste, “Image-Based Strain Analysis on Cement-Based Materials,” MS Thesis, Tennessee Technological University, p. 83, 2006. 17. S. E. Mikel, “Measurement of Induced and Residual Strain States in Portland Cement Using Various Novel Techniques,” MS Thesis, Tennessee Technological University, in press, 2007. 18. J. J. Biernacki, C. J. Parnham, S. E. Mikel, T. Gnaeupal-Harold, and J. Bai, “A Diffractionbased Study of the Strain Response of Mortar,” in preparation. 19. S. E. Mikel, J. J. Biernacki, T. Gnaeupal-Harold and J. Almer, “Using Neutron Diffraction to Study the Strain Response of Portland Cement,” in preparation. 20. S. E. Mikel, J. J. Biernacki, T. Gnaeupal-Harold and J. Almer, “A Diffraction-based Study of Early Autogenous Strain Development in Portland Cement,” in preparation. 21. S. E. Mikel, J. J. Biernacki and M. Lance, “The Strain Response of Single Particles in Portland Cement Using Photostimulate Luminescence Spectroscoy,” in preparation. 22. J. Batiste, J. J. Biernacki and J. Sinopoli, “Using Images to Determine Meso-scale Strains in Portland Cement,” in preparation. 23. I. C. Noyan, and J. B. Cohen, “Residual Stress Measurement by Diffraction and Interpretation,” Springer-Verlag, New York, p. 267, 1997. 24. V. Hauk, “Structural and Residual Stress Analysis by Nondestructive Methods,” Elsevier, Amsterdam, The Netherlands, p. 640, 1997. 25. Q. Ma and D. R. Clark, “Piezospectroscopic Determination of Residual Stresses in Polycrystalline Alumina,” J. Am. Ceram. Soc., vol. 76, no. 6, pp. 1433-1440, 1993. 26. Matrox Inspector vs. 4.

40-Influence o.qxd

9/19/2009

12:02 PM

Page 391

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

INFLUENCE OF HYDROTHERMAL CURING ON MICRO STRUCTURE AND MECHANICAL PROPERTIES OF ULTRA-HIGH PERFORMANCE CONCRETE Patrick FONTANA, Christian LEHMANN and Urs MÜLLER Federal Institute for Materials Research and Testing (BAM) Unter den Eichen 87, 12205 Berlin, Germany, e-mail: [email protected]

ABSTRACT Thermal curing of Ultra-High Performance Concrete (UHPC) has a significant influence on its mechanical properties. When additionally a water vapour saturation pressure is applied (autoclaving), the curing conditions are strongly enhanced, leading to an improved hydration of cement and secondary cementitious materials. Autoclaved UHPC exhibits a cement paste matrix of close networked C-S-H crystal fibres. Flaws are filled with crystalline reaction products generating a more homogenous micro structure. As a consequence, the mechanical properties of UHPC are improved significantly. The partial reaction of quartz filler grains may contribute by improving the cohesion with the cement paste.

Keywords UHPC, hydrothermal curing, mechanical properties, micro structure, phase composition

INTRODUCTION One of the most recent developments in concrete technology is Ultra High Performance Concrete (UHPC). Due to its high density it exhibits exceptional high strength as well as excellent chemical durability and the interest in this material is steadily increasing. So far UHPC is still object of intensive research and its use is not covered by standards or guidelines. Nevertheless it was already applied successfully as construction material, such as for pedestrian and motorway bridges [1-3], lightweight roof constructions, façade elements, pre-cast protection panels [4-6] as well as retrofitting and protection of concrete structures [7]. The high strength of UHPC of 150 MPa and more is the result of a high packing density based on an optimised particle size distribution and significant reduction of water in the cement paste compared to ordinary concrete [8]. The good workability of the UHPC, in many cases even self compacting properties, is adjusted by adding highly efficient plasticizers during mixing. UHPC is also characterised by a high elastic modulus and an almost linear-elastic stress-strain-behaviour until failure. Therefore steel fibre reinforcement is commonly used to improve the ductility of UHPC [7], [9]. Due to the better controlled conditions pre-cast plants are suited very well for the production of UHPC [10], [11]. In pre-cast industry heat curing of concrete with temperatures up to 90 °C is already common practice to shorten the setting time of the concrete and to improve the early strength. Previous studies on heat treated UHPC have shown that due to the higher temperature the reaction kinetics are accelerated and the pozzolanic activity is intensified. Between 200 °C and 250 °C the formation of crystalline calcium silicate hydrates was ob-

40-Influence o.qxd

392

9/19/2009

12:02 PM

Page 392

Patrick FONTANA, Christian LEHMANN and Urs MÜLLER

served [12], but the crystallisation process is strongly dependent on water vapour pressure [13]. So, heat curing above 100 °C under atmospheric pressure seems be a rather uncontrollable process since large differences in local water vapour pressure may be present in the cross section of a concrete element. Therefore our approach was to maintain a homogenous water vapour pressure in a concrete during heat curing by applying an external water vapour saturation pressure (autoclaving). This technique is used for the industrial production of autoclaved aerated concrete (AAC) and sand-lime bricks. Besides the higher degree of cement hydration due to the controlled heat curing process we aimed for the utilisation of secondary cementitious materials to reduce the typically high cement content in UHPC which would benefit the reduction of costs and the increase of sustainability of UHPC. The presented paper deals with the mechanical properties of autoclaved UHPC in comparison to standard curing and simple heat treatment under atmospheric pressure (1 bar). The mechanical properties are related to changes of micro structure and phase composition.

EXPERIMENTAL Materials and curing conditions In order to obtain results with practical relevance the mix design of the UHPC was based on commercial raw materials. As reactive components a white cement CEM I 42.5-R, a micro fly ash (median particle diameter 0.2 μm) and silica fume were used. The maximum size of the quartz aggregate was 2 mm. In order to optimize the particle size distribution a quartz filler with a median size of 50 μm was added. The water cement ratio was 0.26 and the total water binder ratio resulted in 0.22. The mix design of the UHPC is given in Table 1.

Table 1: Composition of the UHPC. material content (kg/m³) cement 745 silica fume 72.4 fly ash 74.5 quartz filler 243 quartz aggregate 0-0.5 mm 238 quartz aggregate 0.5-1.0 mm 357 quartz aggregate 1.0-2.0 mm 357 water 168 superplasticiser * 42.1 * including 70 % water by weight (29.5 kg/m3) The self-compacting properties of the UHPC were adjusted using a polycarboxylate based superplasticiser. The slump-flow was 260 mm (small cone according to EN 1015-3). First, the solid components were dry mixed in a high shear mixer to homogenise the material. Then, water and superplasticiser were added and the material was mixed thoroughly. The fresh concrete was cast in prismatic steel moulds (160 x 40 x 40 mm³). The specimens were demoulded after 1 day and cured under six different conditions (Table 2). Standard curing at 23 °C under water was used as reference (series 1). After heat curing the specimens were stored at 20 °C / 65 % r.h. until testing and sample preparation.

40-Influence o.qxd

9/19/2009

12:02 PM

Page 393

Influence of hydrothermal curing on microstructure and mechanical properties of ultra-high performance concrete

393

Table 2: Curing regimes for the UHPC series. Curing condition Curing time Series 1 – reference 23 °C / 1 bar * 6 days Series 2 – heat treated 90 °C / 1 bar 2 days Series 3 – heat treated 150 °C / 1 bar 2 days Series 4 – heat treated 200 °C / 1 bar 2 days Series 5 – autoclaved 150 °C / 5 bar 8 hours Series 6 – autoclaved 200 °C / 15 bar 8 hours

Methods The mechanical tests were performed after 7 days on standard specimens 160 x 40 x 40 mm³. Flexural strength was determined according to EN 196-1 with a three-point bending test and compressive strength was tested according to DIN 1048-5 in order to determine the ultimate load for elastic modulus tests. The elastic modulus as defined by DIN 1048-5 was the secant modulus after three loading cycles with a maximum load of one third of the ultimate load. For phase analysis a combination of several techniques was used to achieve reliable results. Samples were analyzed using a Philips PW 1710 X-ray diffractometer with Cu KĮ radiation. To optimize the identification of the pozzolanic reaction of portlandite with the mineral additions, differential thermal gravimetry (DTG) was used (Netzsch – STA 449 Jupiter). For micro chemical and structural analysis a scanning electron microscope (Leo Gemini 1530 VP) was employed. Analysis was performed on polished cross sections. Mercury intrusion porosimetry (ThermoFischer Pascal 140-440) was performed to examine the evolution of pore sizes under the different curing conditions.

RESULTS AND DISCUSSION Mechanical properties Fig. 1 shows on the left hand side that the heat curing increased the compressive strength of the UHPC from 85 MPa (reference series 1) to more than 130 MPa in the case of heat treatment at atmospheric pressure (series 2 to 4) and autoclaving at 150 °C / 5 bar (series 5). With autoclaving at 200 °C / 15 bar (series 6) the compressive strength was increased to 157 MPa which corresponds to a strength gain of 85 %. With curing at 200 °C under atmospheric pressure (1 bar) the specimens of series 4 showed a relatively large scatter in test results. The maximum value was 163 MPa, the minimum was 102 MPa. Interesting was the evolution of flexural strength. The heat treated series 2 to 4 showed lower values (around 13 MPa) than the reference series 1 (14.6 MPa). The autoclaved series 6 reached the highest values with 16.9 MPa. Due to the very brittle behaviour of non-reinforced UHPC the scatter of the test results (coefficient of variation 5-12 %) was acceptable.

40-Influence o.qxd

9/19/2009

12:02 PM

394

Page 394

Patrick FONTANA, Christian LEHMANN and Urs MÜLLER

Compressive strength (MPa)

20

200

Flexural strength (MPa)

16.9 157

150 132

15.1

14.6

15

135 138

13.5 12.7

138 12.5

10

100

reference reference

heattreated treated heat

0

autoclaved autoclaved

reference reference

heat treated treated heat

S6 (200 °C / 15 bar)

S5 (150 °C / 5 bar)

S4 (200 °C / 1 bar)

S2 (90 °C / 1 bar)

5

S3 (150 °C / 1 bar)

S1 (23°C / 1 bar)

S6 (200 °C / 15 bar)

S5 (150 °C / 5 bar)

S3 (150 °C / 1 bar)

S4 (200 °C / 1 bar)

S1 (23°C / 1 bar)

50

0

S2 (90 °C / 1 bar)

85

autoclaved autoclaved

Figure 1: Influence of heat curing on compressive (left) and flexural (right) strength of UHPC (average of 6 single values, deviation bars indicate minimum and maximum value). While a clear influence of heat curing on the strength of the UHPC was observed, the elastic modulus was not affected significantly (Fig. 2). Similar results are reported in [14] for UHPC cured for 24 hours at 250 °C under atmospheric pressure. Therefore it can be assumed that heat cured UHPC achieves higher strains at failure. The observed pattern for strength and elastic modulus follows the one of a previous study [15], where the macroscopic mechanical properties were confirmed by nano indentation tests. 50

40.8 38.4

40.3

39.6

39.3 37.7

0

reference reference

heattreated treated heat

Series 6 (200 °C / 15 bar)

Series 5 (150 °C / 5 bar)

Series 4 (200 °C / 1 bar)

10

Series 3 (150 °C / 1 bar)

20

Series 2 (90 °C / 1 bar)

30

Series 1 (23°C / 1 bar)

Elastic modulus (GPa)

40

autoclaved autoclaved

Figure 2: Influence of heat curing on the elastic modulus of UHPC (average of 3 single values, standard deviation is indicated).

40-Influence o.qxd

9/19/2009

12:02 PM

Page 395

Influence of hydrothermal curing on microstructure and mechanical properties of ultra-high performance concrete

395

Microstructure and phase analysis The various curing conditions resulted in significantly different micro structures of the UHPC. The series 6 autoclaved at 200 °C / 15 bar showed clearly the lowest amount of remnant clinker grains and the overall texture of its cement paste was denser with only few larger pores. The higher degree of hydration of the two autoclaved series was also evident from the pozzolanic reaction of the fly ash. The fly ash particles remained almost unreacted in series 1 to 4. White arrows in Fig. 3a indicate still intact fly ash particles. Black arrows indicate portlandite and black circles AFm phases. In the autoclaved series 5 and 6 the fly ash particles have reacted almost completely, sometimes filled up with reaction products (white arrows in Fig. 3b).

b

a

20 μm

20 μm

Figure 3: a. UHPC cured at 23 °C/1 bar (series 1). b. UHPC cured at 200 °C/15 bar (series 6). The higher pozzolanic activity of the fly ash in the autoclaved series is also evident from the amount of portlandite (Ca(OH)2) left after curing. Results of differential thermal gravimetry (DTG) are shown in Fig. 4. In series 1 and 4 portlandite was still observed, but in series 6 it was almost completely consumed by the pozzolanic reaction (no dehydration peak of portlandite at 460 °C).

0.00

Series 6 (200 °C / 15 bar)

DTG (M.-%/min)

-0.05

-0.10 Series 4 (200 °C / 1 bar) Series 1 (23 °C / 1 bar) -0.15

portlandite -0.20 0

100

200

300

400

500

600

Temperature (°C)

Figure 4: DTG curves of UHPC series 1, 4 and 6.

700

40-Influence o.qxd

9/19/2009

396

12:02 PM

Page 396

Patrick FONTANA, Christian LEHMANN and Urs MÜLLER

Simple heat treatment produced a slightly denser micro structure than standard curing, whereas the autoclaved UHPC showed a homogeneous and dense cement paste of close networked crystal fibres with a length up to one micrometer. Cracks and small pores were filled with crystalline calcium silicate hydrates (C-S-H) during autoclaving (black arrows in Fig. 5a). Furthermore reaction of quartz grains was observed in the autoclaved series and in the series 4 heat treated at 200 °C. Small quartz particles were surrounded by a rim of crystalline C-S-H (Fig. 5b). This reaction of quartz in the cement paste system at higher temperatures was already observed in a previous study [15]. Micro chemical analysis confirmed that the reaction product was xonotlite Ca6Si6O17(OH)2.

a

b

20 μm

4 μm

Figure 5: Micro structure of UHPC autoclaved at 200 °C/15 bar (series 6).

Cum. pore volume (mm3/g)

The pore structure was severely influenced by the heat curing. Standard curing at 23 °C/1 bar (reference series 1) resulted in the lowest porosity. Simple heat treatment (series 2 to 4) produced a remarkable increase of pore volume and average pore size what may explain the reduction in flexural strength compared to the standard cured reference series 1. Autoclaving (series 5 and 6) increased the pore volume too, but the pore sizes were significantly reduced. The largest fraction of the pores had a diameter below 10 nm (Fig. 6).

Series 4 (200 °C/1 bar) Series 6 (200 °C/ 15 bar) Series 1 (23 °C/1 bar)

Pore diameter (nm)

Figure 6: Pore size distribution of UHPC series 1, 4 and 6. Hydration products in the cement paste were analyzed by XRD and SEM-EDX. The results indicated the presence of phases though exact analysis of the phase composition in the

40-Influence o.qxd

9/19/2009

12:02 PM

Page 397

Influence of hydrothermal curing on microstructure and mechanical properties of ultra-high performance concrete

397

cement paste was difficult due to the presence of quartz filler. From the results it was evident that heat curing resulted generally in a transformation from amorphous to crystalline C-S-H phases. In conjunction with the slightly denser microstructure this might be the main reason for the increase in compressive strength of the simply heat treated series 2 to 4. It was observed that the level of dehydration of the phases increased with increasing curing temperature. Simple heat treatment under atmospheric pressure increased the Si/Ca atom ratio slightly from 0.65 to 0.75. At 200 °C portlandite was still present, but the amount was reduced to a large extent. The XRD results confirmed the total absence of portlandite in the autoclaved UHPC series, indicating again the high pozzolanic activity of the fly ash under these curing conditions. Results from micro chemistry analysis by SEM-EDX revealed a significant increase of the Si/Ca atom ratio to 1.1 in the autoclaved UHPC series.

CONCLUSIONS Hydrothermal curing (autoclaving) improves significantly the mechanical properties of UHPC. Due to autoclaving the degree of hydration of the cementitious materials is substantially enhanced. Portlandite is consumed completely and Si-rich phases are formed. The main reasons for the improved mechanical properties are the formation of a homogenous cement paste matrix, consisting of close networked C-S-H crystal fibres, which develop a more stable structure than amorphous C-S-H phases, and the “healing” of flaws, such as cracks and small pores, by filling them with C-S-H crystals. In addition autoclaving results in increased cohesion between cement paste and fillers by the partial reaction of quartz grains and a distinctive reduction of pore sizes. While autoclaving increases the strength of UHPC, the elastic modulus is not affected. Therefore autoclaving may improve the strain capacity of UHPC. Thermal treatment of C-S-H phases results generally in the development of crystalline phases at which the water content of the phases decreases with increasing temperature. This might be the main reason for the increased compressive strength when simple heat curing is applied. The observed reduction in flexural strength may be explained by the increase of large sized porosity. Since the pozzolanic reaction of fly ash is significantly accelerated, autoclaving may help to reduce the high cement and silica fume content in UHPC. Future work will focus on different types of secondary cementitious materials and on higher replacement levels in order to improve the sustainability of UHPC.

ACKNOWLEDGEMENTS The authors would like to thank Werner Österle, Sascha Dieter and Romeo Saliwan-Neumann for their help with the electron microscopy and X-ray diffraction. The support of Xella International GmbH by autoclaving the samples is gratefully appreciated.

REFERENCES 1. Aïtcin, P.C., Lachemi, M., Adeline, R., Richard, P., The Sherbrooke reactive powder concrete footbridge, Struct. Eng. Int. (IABSE) Zürich, 8 (2), 1998, 140–144 2. Fehling, E., Bunje, K., Schmidt, M. and Schreiber, W., The "Gärtnerplatzbrücke" Design of first hybrid UHPC-steel bridge across the river Fulda in Kassel, Germany. In: Proc. 2nd Int. Symp. on Ultra High Performance Concrete, Fehling, E., Schmidt, M. and Stürwald,

40-Influence o.qxd

398

9/19/2009

12:02 PM

Page 398

Patrick FONTANA, Christian LEHMANN and Urs MÜLLER

S. eds. Kassel, Germany, 5-7 Mar 2008, Schriftenreihe Baustoffe und Massivbau (10), Kassel University Press, 581-588 3. Ulm, F.-J. and Acker, P., UHPC in the U.S. highway transportation system. In: Proc. 2nd Int. Symp. on Ultra High Performance Concrete, Fehling, E., Schmidt, M. and Stürwald, S. eds. Kassel, Germany, 5-7 March 2008, Schriftenreihe Baustoffe und Massivbau (10), Kassel University Press, 3-10 4. Acker, P. and Behloul, M., Ductal® technology: a large spectrum of properties, a wide range of applications. In: Proc. Int. Symp. on Ultra High Performance Concrete. Schmidt, M., Fehling, E. and Geisenhanslüke, C. eds. Kassel, Germany, 13-15 Sep 2004, Schriftenreihe Baustoffe und Massivbau (3), Kassel University Press, 11-23 5. Behloul, M. and Batoz, J.-F., Ductal® applications over the last Olympiad. In: Proc. 2nd Int. Symp. on Ultra High Performance Concrete, Fehling, E., Schmidt, M. and Stürwald, S. eds. Kassel, Germany, 5-7 Mar 2008, Schriftenreihe Baustoffe und Massivbau (10), Kassel University Press, 855-862 6. Rebentrost, M. & Wight, G. (2008): Behaviour and Resistance of Ultra High Performance Concrete to Blast Effects. In: Proc. 2nd Int. Symp. on Ultra High Performance Concrete, Fehling, E., Schmidt, M. and Stürwald, S. eds. Kassel, Germany, 5-7 Mar 2008, Schriftenreihe Baustoffe und Massivbau (10), Kassel University Press, 735-742 7. Brühwiler, E. and Denarié, E., Rehabilitation of concrete structures using Ultra-High Performance Fibre Reinforced Concrete In: Proc. 2nd Int. Symp. on Ultra High Performance Concrete, Fehling, E., Schmidt, M. and Stürwald, S. eds. Kassel, Germany, 5-7 Mar 2008, Schriftenreihe Baustoffe und Massivbau (10), Kassel University Press, 895-902 8. de Larrard, F. and Sedran, T., Optimization of ultra-high-performance concrete by the use of a packing model. Cement and Concrete Research, 24, 1994, 997-1009 9. Bornemann, R. and Schmidt, M., Ultrahochfester Beton (UHPC) - Herstellung und Anwendung. Betonwerk + Fertigteil-Technik, 68, 2002, 10-12 10. Schmidt, M., Fehling, E., Teichmann, T., Bunje, K. and Bornemann, R., Ultra-Hochfester Beton: Perspektive für die Betonfertigteilindustrie. Beton, 53 (3), 2003, 16-19 11. Rebentrost, M. and Wight, G., Experiences and applications on Ultra-high Performance Concrete in Asia. In: Proc. 2nd Int. Symp. on Ultra High Performance Concrete, Fehling, E., Schmidt, M. and Stürwald, S. eds. Kassel, Germany, 5-7 Mar 2008, Schriftenreihe Baustoffe und Massivbau (10), Kassel University Press, 19-30 12. Cheyrezy, M., Maret, V., Frouin, L., Microstructual analysis of RPC (Reactive Powder Concrete). Cement and Concrete Research, 25 (7), 1995, 1491-1500 13. Feylessoufi, A., Crespin, M., Dion, P., Bergaya, F., Van Damme, H. and Richard, P., Controlled rate thermal treatment of reactive powder concretes. Advanced Cement Based Materials, 6 (1), 1997, 21-27 14. Hegger, J., Tuchlinski, D., Kommer, B., Bond anchorage behavior and shear capacity of Ultra High Performance Concrete beams. In: Proc. Int. Symp. on Ultra High Performance Concrete. Schmidt, M., Fehling, E. and Geisenhanslüke, C. eds. Kassel, Germany, 13-15 Sep 2004, Schriftenreihe Baustoffe und Massivbau (3), Kassel University Press, 351-360 15. Müller, U., Kühne, H.-C., Meng, B., Nemecek, J. and Fontana, P., Micro texture and mechanical properties of heat treated and autoclaved Ultra High Performance Concrete (UHPC). In: Proc. 2nd Int. Symp. on Ultra High Performance Concrete, Fehling, E., Schmidt, M. and Stürwald, S. eds. Kassel, Germany, 5-7 Mar 2008, Schriftenreihe Baustoffe und Massivbau (10), Kassel University Press, 213-220

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 399

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

PATCHES IN CONCRETE: RECENT EXPERIMENTAL DISCOVERY OF A NATURAL PHENOMENON – SUPPORTING EVIDEDENCE BY DEM Piet STROEVEN and Huan HE Faculty of Civil Engineering and Geosciences, Delft University of Technology Stevinweg 1, 2628 CN Delft, the Netherlands; e-mail: [email protected]

ABSTRACT Microscopy investigations have revealed a so-called patch microstructure. This paper generalizes the patchy character of concrete by demonstrating the occurrence of patches on different levels of the microstructure. This is achieved on the basis of virtual concrete produced by concurrent algorithm-based DEM. The irregularities in appearance in sections of aggregate, fibers, pores or unhydrated cement are in conformity with the stochastic nature of the material.

Keywords Concrete, cement, patches, fibers, pores, unhydrated cement

INTRODUCTION Microscopy observations by Diamond & Thaulow have proven the physical existence of a socalled patch microstructure in concrete [1]. Further experimental studies on meso-level form supporting evidence for the occurrence of patches on different levels of the micro-structure [2,3]. So, the conclusion can be drawn from such experiments that patches seem a natural phenomenon in particulate materials such as concrete. Patchy should be understood according to MerriamWebster dictionary as “irregular in appearance, make up or quality”. This paper aims offering additional material obtained in virtual reality, whereby computer concrete (=compucrete) is investigated for this structural feature. This requires the DEM approach to be able realistically producing particle dispersion. Therefore, use was made of concurrent algorithm-based SPACE and HADES systems, since so-called sequential random (particle) addition (SRA) systems would offer biased information [4]. Concrete is a particulate material on different levels of the micro-structure whereby relative high particle densities occur. On meso-level, the aggregate grains are dispersed in the fresh cementitious matrix. Aggregate density falls commonly in the 70 to 80% range, whereby the state of dense random packing is obtained. Cement particle density in the fresh state can be up to 60% (for low water to cement ratio concretes). At low particle density (and small particles), particle distribution is governed by a Poisson process reflecting the state of pure “chaos”. This can be described mathematically (based on probability theory) [4-6]. Pure “order” is found in crystalline structures, whereby particle

1

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 400

400

Piet STROEVEN and Huan HE

dispersion is governed by the theory of geometry. The aforementioned concurrent systems can cover the full density range from chaos to order [7,8]. Actual particle dispersions dealing with the aforementioned components on different levels of the microstructure should therefore reveal intermediate characteristics. So, the typical density fluctuations of the Poisson field should still be an inherent feature of particulate structure of concrete on the different levels of the microstructure. This paper is aiming to provide in virtual reality by physical discrete element (computer) modeling (DEM) supporting evidence as well as a more general perspective on the patchy nature of concrete on the different levels of the microstructure, as witnessed in some experimental microscopy efforts. Figs. 1 and 2 may form an introduction to this topic.

20 mm

20 mm

Fig. 1. Concrete containing about 10% by volume of 16 mm coarse aggregate (ceramic spheres (Stroeven, 1973)); boundary disturbance-free 200 mm section image of test specimen (left), and of comparable SPACE-generated compucrete are displayed (right) [3]. Patch microstructure is obviously revealed. Particularly in cement paste, changing w/c ratio, or introducing flocculation or particle repulsion, will change the size and spacing of the patches, however this will not fundamentally affect the patchy nature due to particle packing, as we have been showing [3]. This is in conformity with observations of Diamond on concrete and superplasticized concrete: “The porous patches appeared to be relatively fewer in number and generally smaller in this heavily superplasticized concrete, but they are definitely present.”

VIRTUAL REALITY CONCRETE: COMPUCRETE Particle packing Available physical computer simulation methods for forming granular packing of hard particles can be placed in two distinctive groups. The first group, based on so-called concurrent algorithms, involves the densification of a fixed number of particles. The SPACE and HADES systems, on which our studies rely, are representative examples of this category in present-day concrete technology [3-4, 7-11]. The second group is based on sequential algorithms, which has been proven an inappropriate approach to the targeted problems [4]. Some upgraded SRA systems are described in the literatures, such as the one in which particles that tend to overlap are subjected to random shifts [12]. This is an example of a static approach to inclusion of particle interference. The SPACE and HADES systems, developed at 2

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 401

Patches in concrete: recent experimental discovery of a natural phenomenon – supporting evidence by DEM

10 Pm

401

10 Pm

10 Pm

10 Pm

(a) 0.25L from surface of aggregate

(b) 0.5L from surface of aggregate

Fig. 2 (top) Sections of fresh SPACE-generated cement paste (w/c=0.5; 40000 particles; 508 m2/kg) in container with rigid boundaries; (bottom) after one-year hydration [13]. Patches in gel (grey), water (blue), air (pink) and unhydrated cement (black) are obvious; L is the linear length of the specimen Delft University of Technology realizes compaction by a dynamic algorithm, which is also supposed to imitate the production stage of the material. The forces added to the particles can be manipulated, so that “sticky” particle contacts (or particle repulsion) during the production of the model material can be simulated. Also gravity effects can simply be included. This dynamic (Newtonian) simulation mechanism has no significance after completion of the simulation, hence, is not connected with the rheological properties of the cementitious model material. SPACE is based on spherical particles only. However, HADES is allowing the use of arbitrarily shaped particles [8,14]. This is to account for particle shape effects that seem to have more serious impact on packing than so far assumed in concrete technology. SPACE is impulse-based, whereas HADES is a force-based system. The compucrete is produced in limited quantities in cubic moulds. Basically, containers with rigid walls are used, or with so-called periodic boundaries. The first situation conforms to a molded aggregate, or to cement paste pocketed between surfaces of aggregate grains, so 3

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 402

402

Piet STROEVEN and Huan HE

encompassing also gradient structures like in the ITZ. The second type of boundary conditions is used to simulate bulk material. Also mixed situations, whereby two rigid surfaces are combined with four periodic ones, can be simulated [14]. Hydration The kinetic hydration model used to describe the hydration behaviour of single cement particles consist of two subsequent stages, following van Breugel [15] and Navi and Pignat [16]. The first stage, in which a phase boundary mechanism controls the hydration rate, is followed by a stage in which the reaction rate is controlled by a diffusion mechanism. A uniform deposition of the gel products onto the available gel surface, a uniform decrease of the entire cement surface, and a uniform consumption of the water from the available water-air surface are assumed. Subsequently, the cement-gel interfaces as well as the gel-air and air-water interfaces remain located on concentric spheres throughout the hydration process. The simultaneous expansion of multiple neighbouring particles and the limited amount of available water will significantly affect the hydration behaviour of a single particle. First, growing contact areas between the gel products of neighbouring particles diminishes the surface area that is available for product deposition. Finally, mutual contacts and local water shortage will influence the inward growth of the cement. A detailed description of the hydration algorithms is available in Stroeven and Stroeven [17].

EXAMPLES OF PATCHES Patch formation occurs naturally in cementitious particulate materials on different levels of the microstructure. Examples will be presented of our research on aggregate and fibers on meso-level, and on pores and unhydrated cement on micro-level. Meso-level: as to the aggregate distribution, we can refer here to Fig. 1, revealing denser and less dense areas. Further examples can be found in [5]. Additionally, Fig. 3 offers a HADESproduced compacted angular-shaped coarse aggregate packing and a section image revealing its patchy meso-structure. Density differences can be quantified as indicated for pores and unhydrated cement in the forthcoming section on “size of patches”.

(a)

(b)

Fig. 3 (a) Visual model of HADES-compacted package of arbitrary octahedrons and (b) image of section trough package, revealing patchy structure. 4

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 403

Patches in concrete: recent experimental discovery of a natural phenomenon – supporting evidence by DEM

403

At the same meso-level, the natural density variations in steel fiber distribution in mortar and concrete can be mentioned. A logic possibility to demonstrate this patchy fiber structure is to superimpose an orthogonal grid on the section image of SFRC and determine the number of fibers per box [18]. This methodology will be illustrated in the next section. An example is shown in Fig. 4 (left). An alternative is using X-ray projection images (Fig. 4 (right)) of which the patchy nature can be quantified by counting intersections with a superimposed line grid [19]. A more sophisticated approach is by second order stereology. This involves determination of the average number of fibers in identical circles around all fibers in the section plane at steadily increasing circle radius, r. The plot of NA versus r give insight into “inner order”, and is thus a way of quantifying the “irregularities in the appearances” (here displayed in an X-ray radiograph) [20,21].

Fig. 4 (left) Section image (100x100mm) of 8 mm maximum grain size mortar containing 1.5% by volume of fibers, and (right) X-ray radiograph of vertical slice (70x200 mm) of SFRC. Patchy nature is in both appearances obvious. As an example on micro-level, a Portland cement with Blaine density of 508 cm2/kg was chosen for illustrative purposes. The sieve curve in Fig. 5 (left) complies with the so-called Rosin-Rammler size distribution function G (d ) 1 exp( bd a ) [7], in which a and b are constants. Three fresh cement specimens with different water to cement ratios (in 0.2 to 0.6 ranges) were produced in molds with rigid boundaries by the concurrent algorithm-based SPACE system, employing 40000 particles. The smallest linear length of the model cube is 80.64 Pm, which is more than 4 times exceeding maximum size of the particles (18.6 Pm). Accordingly, models are assumed to be representative (have RVE size) for composition. Dynamic mixing accompanied by compaction was terminated when density corresponding to the required w/c ratio was reached. Thereupon, 28 days of hydration was simulated. Three equidistant sections were selected, as shown in Fig. 5 (right). Two of the images of the specimen with w/c=0.5 were earlier shown in Fig. 2 for the fresh and hydrated states, respectively. Additional pictures of other cases can be found in [13]. However, the presented material is sufficient to demontrate the occurrence of patches in fresh and hardened states. A comparison could learn that the patch phenomenon in pore structure is more pronounced in a mixture with high w/c ratio. The reversed conclusion could be drawn as to the unhydrated cement. This is a logic consequence of density differences in the two components left in hydrated model paste, and could have been predicted from the theory of stochastic heterogeneity that will be touched upon in the next section [22]. Although the way quantitative

5

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 404

404

Piet STROEVEN and Huan HE

information can be obtained as to the patch structure has been indicated for meso-structures, in a next section we will come more explicitly back to this problem for the intimately related case of micro-structures.

Cum.volume fraction (-)

1.00 0.80 0.60 0.40 0.20

C508

0.00 0

5

10 Size ( P m )

15

20

Fig. 5 (left) Cumulative size distribution function of particles of model cement C508, and (right) equidistant sections prepared from model cement paste.

SIZE OF PATCHES By varying technological parameters in research, the size of the appropriate representative volume element (RVE) and of the representative area element (RAE) is affected. This governs the degree of stochastic heterogeneity in descriptors of patch microstructure obtained in samples (e.g., fields) of given size. Of course, the same holds for the visual impression the patch structure undergoes after changing a technological parameter. When the ratio of linear dimensions of RAE and of the field would be kept constant and image magnification would also be adjusted accordingly, the patch structure would seem similar (in stochastic perspective) to the observer. Hence, sizing in this paper is a means to visualize more explicitly concrete’s patch structure. For the appropriate underlying theory on stochastic inhomogeneity in the case of configurationsensitive features of material structure, see [22]; this paper does not allow going into more detail. To obtain quantitative information on the scale of the patches in the case of the hydrate structure, one could simply determine areal fraction of porosity pertaining to a (by superimposed orthogonal grid produced) field system covering the section image (Fig. 6 (left)). This could be employed for different grid spacing, to produce a so-called local porosity distribution, shown for a real cement paste in Fig. 6 (right). This approach is further elaborated in [23,24], where it is demonstrated that application of a grid spacing that is adapted to the involved linear changes in respective RAE sizes during hydration would have resulted in similar local porosity curves. So, patch sizes would have been similar too! The ITZ is an example of gradient structures in concrete that can be studied by producing cement paste in containers with rigid boundaries. By plotting various descriptors of simulated ITZ structure, the associated extent of the ITZ can be assessed [25,26]. A composition parameter (porosity) analyzed on SPACE produced cement paste with various water to cement ratios is plotted as an example in Fig. 7. Data are average values of patchy structure in container sections

6

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 405

Patches in concrete: recent experimental discovery of a natural phenomenon – supporting evidence by DEM

405

parallel to aggregate grain surface. Porosity data still fluctuate, reflecting patchy nature, particularly at high w/c-ratio.

Local porosity distribution

25

3d 7d

20

14d 15 10 5 0 0

20

40

60

80

100

Porosity (%)

Fig. 6 (left) Grid superimposed on section pattern of Fig. 2 (left), allowing determination of local porosity distribution P (right) for paste with w/c=0.6 at 3, 7 and 14 days of hydration, respectively, using the same grid. Patchy nature of the cement is obvious. 1.0

VV of pore (-)

0.8

w/c=0.2

w/c=0.3

w/c=0.4

w/c=0.5

w/c=0.6

0.6 0.4

0.2 0.0 0

10 20 30 Distance to aggregate surface (P m )

40

Fig. 7. Gradient in porosity of cement paste perpendicular to aggregate grain surface, 1 year’s matured cement pastes with Blaine number 497 at different w/c-ratio.

To fulfill the requirements of stochastic heterogeneity, the observations on single specimens - as plotted in Fig. 7 - should be averaged over a significant number of independent experiments (20 in [7]). The number of experiments required depends on the configurationsensitivity of the descriptor [22]. Effectively, the patchy nature is contracted away in doing so. Of course, for the unhydrated cement structure, a similar approach can be followed [27]. The pattern of densities obtained by way of the grid method shown in Fig. 6 (left) can be further elaborated by constructing density contour lines, as shown for a single case in Fig. 8,

7

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 406

406

Piet STROEVEN and Huan HE

however dealing with the amount of unhydrated cement shown in sections of Figs. 2 and 6. Again, the result is magnification-dependent according to the theory of stochastic heterogeneity [22]. Hence, direct application is only justified for similar technological parameters (like w/c ratio and hydration time). The given section patterns deal with the same specimen, so direct evaluation of the successive patchy patterns in sections is justified. 5.200 16.28 27.35 38.43 49.50 60.58 71.65 82.73 93.80 1.000 12.05 23.10 34.15 45.20 56.25 67.30 78.35 89.40

70

70

60 50

50

Y Axis (Pm)

Y Axis (Pm)

60

40 30

40 30 20

20

10

10 10

20

30

40

50

60

10

70

20

30

40

50

60

70

X Axis (Pm)

X Axis (Pm) 4.800 14.55 24.30 34.05 43.80 53.55 63.30 73.05 82.80

70 60

Y Axis (Pm)

50 40

30 20 10 10

20

30

40

50

60

70

X Axis (Pm)

Fig. 8. Quantification by way of contour lines of areal density of unhydrated cement paste in successive cement paste images of a single specimen with w/c=0.2.

CONCLUSIONS The concurrent algorithm-based DEM approaches used herein present supporting evidence for the patchy nature of concrete on meso-level (aggregate, fibres), as well as on micro-level (unhydrated cement, pores). This way of tackling the problem renders possible investigating systematically how technological parameters affect number and size of patches in samples of given size. This was not pursued herein. To do this correctly, the grid spacing should be adapted

8

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 407

Patches in concrete: recent experimental discovery of a natural phenomenon – supporting evidence by DEM

407

to changes in RAE sizes resulting from modifications in investigated technological parameters. This will inevitably be different for descriptors with different configuration-sensitivity applied to the same image. REFERENCES 1. Diamond, S., Thaulow, N., The patch microstructure in concrete: Evidence that it exists and is not a backscattered SEM artefact. Cem. Concr. Compos., 28, 2006, 606-612 2. Diamond, S., The patch microstructure in concrete: the effect of superplasticizer. Cem. Concr. Res., 36 (4), 2006, 776-779 3. Stroeven, P., He, H., Guo, Z., Stroeven, M., Particle packing in a model concrete at different levels of the microstructure: Evidence of an intrinsic patchy nature. Mater. Charact., 2009 (in press) 4. Stroeven, P., J. Hu, J., Stroeven, M., On the usefulness of discrete element computer modeling of particle packing for material characterization in concrete technology. Comput. Concr., 2009 (in press) 5. Stroeven, P., Some Aspects of the Micro-mechanics of Concrete. PhD Thesis, Delft University Press, Delft 1973 6. Kendall, M.G., Moran, P.A.P., Geometrical probability. Charles Griffin & Co., London 1963 7. Stroeven, M., Discrete numerical modeling of composite materials. PhD Thesis, Meinema BV, Delft 1999 8. He, H., Guo, Z., Stroeven, M., Stroeven, P., Sluys, L.J., Particle packing characteristics in concrete assessed by a discrete element method. In: Proc. Int. Conf. “Particle 2009”, 25-27 Nov. 2009 (to be published) 9. Stroeven, P., Stroeven, M., Assessment of packing characteristics by computer simulation. Cem. Concr. Res., 29 (8), 1999, 1201-1206 10. Stroeven, M., Stroeven, P., Computer-simulation of internal structure of materials. Acta Stereol., 15 (3), 1996, 247-252 11. Stroeven, P., Stroeven, M., SPACE approach to concrete’s space structure and its mechanical properties. Heron, 46, 2001, 265-289 12. Williams, S.R., Philipse, A.P., Random packings of spheres and spherocylinders simulated by mechanical contraction. Phys. Rev. E, 67, 2003, 1-9 13. He, H., Stroeven, P., Guo, Z., Stroeven, M., Virtual reality approach to concrete’s patchy structure. In: Proc. Int. Conf. “Microstructure Related Durability of Cementitious Composites”, W. Sun, K. van Breugel, C. Miao, G. Ye and H. Chen eds., RILEM Publ. S.A.R.L., Bagneux, 2008, 1189-1198 14. Stroeven, P., Sluys, L.J., Guo, Z., Stroeven, M., Virtual reality studies of concrete. Forma, 21 (3), 2006, 227-242 15. Breugel, K. van, Simulation of hydration and formation of structure in hardening cementbased materials. PhD Thesis, Delft University Press, Delft 1991 16. Navi, P., Pignat, C., Simulation of cement hydration and the connectivity of the capillary pore space. Adv. Cem. Based Mater., 4, 1996, 58-67 17. Stroeven, M., Stroeven, P., SPACE system for simulation of aggregated matter application to cement hydration. Cem. Concr. Res., 29, 1999, 1299-1304 18. Stroeven, P., Babut, R., Fracture mechanics and structural aspects of concrete. Heron, 31 (2), 1986,15-44

9

41-Patches in concrete.qxd

9/19/2009

12:02 PM

Page 408

408

Piet STROEVEN and Huan HE

19. Stroeven, P., Shah, S.P., Use of radiography-image analysis for steel fibre reinforced concrete. In: “Testing and Test Methods of Fibre Reinforced Composites”, R.N. Swamy ed. Construction Press, Lancaster, 1978, 308-311 20. Stroeven, P. and Stroeven, M., Application of second order stereological method to steel wire reinforced concrete. Acta Stereol., 11 (1), 1991, 605-610 21. Hanisch, K.H., König, D., Stoyan, D., The pair correlation function for point and fibre systems and its determination in planar sections. J. Microsc. 140, 1985, 361-370 22. Stroeven, P., Hu, J., Stochastic heterogeneity as fundamental basis for the design and evaluation of experiments. Cem. Concr. Comp., 30, 2008, 506-514 23. Hu, J., Stroeven, P., Local porosity analysis of pore structure in cement paste. Cem Concr Res. 35 (2) 2005, 233-242 24. Hu, J., Porosity of Concrete, Morphological Study of Model Concrete. PhD Thesis, OPTIMA, Delft, 2004 25. Chen, H, Stroeven, P, Ye, G., Stroeven, M., Influence of boundary conditions on pore percolation in model cement paste. Key Eng. Mater. 302-303, 2006, 486-492 26. Hu, J., Stroeven, P., Properties of the Interfacial Transition Zone in Model Concrete, Interface Sci., 12, 2004, 389-397 27. He, H., Guo, Z., Stroeven, P., Stroeven M., Sluys L.J., Self-healing capacity of concretecomputer simulation study of unhydrated cement structure. Image Anal. Stereol., 26 (3), 2007, 137-143

10

42-Relationships.qxd

9/19/2009

12:02 PM

Page 409

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

RELATIONSHIPS BETWEEN FRACTAL DIMENSION AND THE MECHANICAL AND STRUCTURAL PARAMETERS OF CONCRETES WITH BASALT AGGREGATE Janusz KONKOL1, Grzegorz PROKOPSKI2 Department of Materials Engineering and Technology of Building Rzeszów University of Technology PowstaĔców Warszawy 6, 35-959 Rzeszów, Poland, e-mail: [email protected], [email protected]

ABSTRACT The paper presents the results of strength, stereological and fractal tests and examinations carried out on concretes based on coarse basalt aggregates. The testing programme was set up based on the two-factor design with two repetitions of the test at the central point. The variables in the composition of concretes were watercement (w/c) ratio in the range from 0.41 to 0.61 and amount of coarse aggregate in relation to fine aggregate in the range from 1.6 to 3.0. A relationship was demonstrated to exist between the strength characteristics of concretes and their fractal dimension, as determined by the chord method. The concrete of poorer strength properties had higher fractal dimension values of its fractures, whereas the concrete of the best strength parameters was characterized by the lowest fractal dimension value. A statistical significant correlation was found to exist between the fractal dimension DC, as determined by the chord method, and Young's modulus E, the unit failure work JIc, and the compressive strength fc. With the increase in fractal dimension, all the above-mentioned basalt concrete characteristics declined. No significant correlations are, however, found between the fractal dimension and the S

critical values of stress intensity factor K Ic .

Keywords Concrete, fractal dimension, fracture surface, profile line development factor, fracture surface development factor, mechanical parameters

INTRODUCTION The contemporary scientific achievements enable analyzing the structure of concrete using quantitative methods, among which the analysis of surface morphology of fractures originated from fracture toughness tests can be ranked. The degree of complexity of a concrete fracture surface can be described with the profile line development factor RL, or the fracture surface development factor RS, as well as with the fractal dimension D, defined by Mandelbrot [1] in 1980. Early studies on concretes and mortars, aimed at using fractal geometry, confirmed the usefulness of fractal geometry for the description of the fracture surface of concretes and mortars [2, 3]. Further studies demonstrated a relationship to exist between the fractal dimension D, and concrete properties, such as compressive strength fc [4, 5]; fracture toughness, as expressed with the critical stress intensity factor KIc [4, 6, 7]; fracture energy GF [6, 8]; or the modulus of elasticity E [4].

42-Relationships.qxd

9/19/2009

12:02 PM

410

Page 410

Janusz KONKOL and Grzegorz PROKOPSKI

Analysis of results of the compressive strength and fracture toughness tests of concretes made from three different aggregates (dolomite, gravel, and basalt) and the relationship between fracture mechanics parameters and compressive strength and the fractal dimension, as determined by the chord method, are reported in [4]. The tests were carried out after 7, 14, 28 and 90 days of concrete curing. For the 28 days' concrete, the following correlation equations were obtained: K IcS 61.69-57.59 D MN/m 3/2

fc

1947.6-1842.2 D MPa

(1)

E 970.44-912.08 D GPa A linear relationship was found to exist between the concrete characteristics from the tests (i.e. the critical stress intensity factor K IcS , compressive strength fc and Young's modulus E) and the fractal dimension, as determined by the chord method. With increasing fractal dimension, all concrete characteristics impaired. Hence, the conclusion was prepared that for the concretes examined, the degree of fracture surface complexity was the greater, the poorer the aggregate/cement paste contact was. The influence of the cross-grain fracture may be here considered. A similar trend was also found by Saouma and Barton [6]. In their tests of gravel and basalt aggregate-based concretes, they showed that with growing fractal dimension of fracture surfaces, the values of the parameters GF and KIc decreased following the relationships below: G F 1380.7 1043.9 D N/m (2) K Ic 6.837 5.239 D MNm -3/2 In the case of two concretes made with the same aggregate and with two different water/binder (w/b) ratios, Yan et al obtained liner relationships between fracture energy and the fractal dimension, in the form as follows: GF 12.48 1802 ( D 2), w / b 0.44 (3) GF 21.00 2730 ( D 2), w / b 0.26 In both cases, the coefficient of linear correlation was R = 0.986. The increase in fracture energy with increasing fractal dimension was also confirmed in the study by Issa et al [8]. Fractal tests were also used for the analysis of cracks on concrete surfaces. In their investigation of concrete reinforced with steel filaments, Yan et al [5] obtained a good correlation between compressive strength and the fractal dimension of those cracks, as determined by the coating method. They found a greater number of cracks on the concrete specimen surfaces of specimens with the highest content of steel filaments, which caused an increase in fractal dimension. They proposed a formula relating compressive strength with the fractal dimension [5]: f c 79.9 81.4 ( D 1 ) , (4) The correlation coefficient for this relationship was R = 0.995.

TEST PROGRAMME

In the present work, an attempt was made to determine the relationship of fracture surface morphology with fracture mechanics parameters and the compressive strength of concretes. Specimens were made with concretes based on broken basalt aggregate with grains up to 16 mm. Portland cement CEM I 32.4 R was used for the tests. The variables in concrete composition were: water-cement (w/c) ratio, ranging from 0.41 to 0.61, and the mass relation

42-Relationships.qxd

9/19/2009

12:02 PM

Page 411

Relationships between fractal dimension and the mechanical and structural parameters of basalt...

411

of coarse aggregate to fine aggregate (Cagg./Fagg.), varying from 1.6 to 3.0. A constant cement/fine aggregate ratio of 1:1.76 was applied. Ten series of concretes differing in their mix composition (Table 1) were prepared, and within each item of the testing plan four repetitions were made for fracture toughness tests according to fracture Mode I, and six repetitions for compressive strength tests. The test programme was drawn up based on a two-factor design with two repetitions of the experiment in the central point (Fig. 1). The plan is optimal in respect of three selection criteria, namely: realizability, informativeness, and effectiveness.

Fig. 1. The arrangement of points in the testing plan. Tests were carried out for five intermediate values of two input values (variables in the testing plan), determined for normalized values equal to: 0, r1, r1.414. Table 1. Mix proportions of tested concretes. Real variables Series no.

w/c

Cagg./Fagg.

[-] 1 2 3 4 5 6 7 8 9 10

[-] 0.44 0.58 0.41 0.61 0.51 0.51 0.51 0.44 0.58 0.51

[-] 1.81 2.80 2.30 2.30 1.60 3.00 2.30 2.80 1.81 2.30

Proportions of components per 1 m3 of concrete mix Fine Coarse Cement Water aggregate aggregate c W Fagg. Cagg. 3 [kg/m ] [kg/m3] [kg/m3] [kg/m3] 396 174 698 1261 306 178 540 1508 358 147 631 1450 334 204 588 1353 405 206 714 1142 301 154 531 1592 345 176 609 1400 320 141 564 1576 375 218 661 1194 345 176 609 1400

42-Relationships.qxd

9/19/2009

12:02 PM

Page 412

412

Janusz KONKOL and Grzegorz PROKOPSKI

STRENGTH TESTING OF CONCRETES

The fracture toughness test according to Mode I was carried out on specimens of dimensions of 8×15×70 cm (Fig. 2), while 10 cm-edge cubic specimens were used for the compressive strength test. The specimens were stored in air with RH=95 %. The strength tests were performed at age of 28 days of concrete curing.

a0=5cm

d = 15 cm

P

CMOD

b=8cm

S = 60 cm L = 70 cm

Fig. 2. Notched-bar fracture toughness test specimen The following were determined in the fracture toughness tests, according to [9]: 9critical stress intensity factor K IcS , 9unit failure work JIc, 9modulus of elasticity at bending E. Table 2 summarizes the obtained strength test results together with the mean value standard deviation s. Table 2. Mean values of the mechanical characteristics of concretes. Series no. [-] 1 2 3 4 5 6 7 8 9 10

K Ics ±s [MN/m3/2] 1.54±0.04 1.36±0.02 1.51±0.12 1.28±0.04 1.53±0.06 1.65±0.06 1.48±0.05 1.55±0.08 1.21±0.04 1.53±0.07

Mean values ± mean standard deviation J Ic ±s E ±s [N/m] [GPa] 31.8±1.5 34.4±1.2 25.0±0.4 32.0±1.7 40.9±1.1 35.3±1.3 25.3±0.9 30.4±1.3 26.8±1.9 31.3±1.6 35.3±3.2 34.8±0.6 29.5±4.1 35.0±0.8 36.9±1.4 34.5±1.2 24.6±1.6 27.8±0.9 38.3±2.4 34.3±1.8

f c ±s [MPa] 56.4±1.1 52.2±1.1 72.6±1.0 44.4±0.4 44.3±1.3 55.2±2.0 54.4±1.1 63.5±1.1 43.8±0.8 50.3±1.0

It can be noticed from the obtained results shown in Table 2 that the increased water-cement ratio w/c, and the lower coarse aggregate fraction of concrete mix result in a drop in the mechanical concrete characteristics tested, i.e.: the critical stress intensity factor K IcS ; the unit failure work JIc; the coefficient of elasticity, E; and the compressive strength fc.

42-Relationships.qxd

9/19/2009

12:02 PM

Page 413

Relationships between fractal dimension and the mechanical and structural parameters of basalt...

413

EXAMINATION OF FRACTURE SURFACE MORPHOLOGY

The fractal and fractographical examinations of fracture surfaces were performed on specially prepared gypsum replicas made on the fractures of concrete bars previously used in fracture toughness testing. Twenty four fracture surfaces were randomly chosen for examination. Stained gypsum was poured on the concrete fracture replicas made of white gypsum. So prepared specimens were slit along the longer side into 5 mm-thick 10 strips to obtain 20 profile lines for each fracture (Fig. 3). This adoption of the slitting direction was intended to obtain profile lines that would be roughly consistent with the direction of crack propagation. The slit specimens were scanned at a resolution of 600 dpi, and the profile line coordinates were obtained using the FRAKTAL_Digit1) software application (Fig. 4). The computation of the fractal dimension D was carried out by the chord method using the FRAKTAL_Wymiar2D2). The profile line coordinates were also used for the determination of the profile line development factor RL, being the ratio of the profile line L to the length of its projection onto the reference line L' (Fig. 3), and the fracture surface development factor RS, or the ratio of the surface area to the area of surface projection onto the reference plane.

PROFILE LINES

L

L’ Fig. 3. Direction of scanning the fracture surfaces. Table 3. Results of fractal and fractographical examinations. Series no. [-] 1 2 3 4 5 6 7 8 9 10 1) 2)

Variable

w/c [-] 0.44 0.58 0.41 0.61 0.51 0.51 0.51 0.44 0.58 0.51

Cagg./Fagg. [-] 1.81 2.80 2.30 2.30 1.60 3.00 2.30 2.80 1.81 2.30

Fractal dimension

DC ± s [-] 1.0241±0.0008 1.0240±0.0007 1.0216±0.0008 1.0235±0.0009 1.0238±0.0010 1.0240±0.0010 1.0227±0.0008 1.0227±0.0006 1.0253±0.0010 1.0227±0.0007

Konkol, J.: FRAKTAL_Digit, a software application, 2001. Konkol, J.: FRAKTAL_Wymiar2D, a software application, 2000.

Degree of profile line/surface development RL ± s RS ± s [-] [-] 1.246±0.009 1.442±0.016 1.248±0.007 1.441±0.012 1.229±0.006 1.406±0.010 1.244±0.007 1.426±0.011 1.251±0.010 1.437±0.016 1.267±0.012 1.474±0.019 1.250±0.009 1.439±0.015 1.248±0.007 1.439±0.011 1.290±0.010 1.505±0.015 1.240±0.007 1.425±0.011

42-Relationships.qxd

414

9/19/2009

12:02 PM

Page 414

Janusz KONKOL and Grzegorz PROKOPSKI

The computation of the fracture surface development factor RS was made by the cycloid method provided by Wojnar [10]. Table 3 summarizes results obtained from both the fractal and the stereological examinations. The extreme values of computed parameters are marked in bold in the table.

Fig. 4. A fracture profile scanned at a resolution of 600 dpi and the result of profile line digitalization with the FRAKTAL_Digit programme. The least values the fractal dimension, DC, as well as those of the profile line development factor RL, and the fracture surface development factor RS, were obtained for Series 3 of concrete with the lowest water-cement ratio (w/c = 0.41) and of the best mechanical properties (the highest values of E, JIc and fc). An agreement between the extreme values of the fractal dimension DC and the profile line development factor RL were also observed for the highest values of these parameters. The greatest values of the parameters D, RL and RS were obtained for Series 9 of concrete that exhibited the poorest mechanical properties. The obtained results demonstrate that the parameters D, RL and RS are the higher, the poorer mechanical properties are possessed by a concrete, whereas the minimal values of these parameters are obtained for concretes of the best mechanical properties. Obtaining the least values of all fractographical parameters for that concrete was due to the low cohesion forces at the coarse aggregate – cement paste interface resulting in the occurrence of a grain-wise fracture with high roughness. The fracture surfaces of a concrete exhibiting a higher strength (with the aggregate/cement paste contact layer with an increased strength) are more flat as a result of the fracture occurring across the coarse aggregate grains (the inter-grain fracture, Fig. 5).

Fig. 5. Microstructure of concrete (w/c = 0.51, Cagg./Fagg. = 3.0, Series 6). A cross-fractured basalt grain is visible.

42-Relationships.qxd

9/19/2009

12:02 PM

Page 415

Relationships between fractal dimension and the mechanical and structural parameters of basalt...

415

Thus, for example, the fractures of the basalt aggregate-based concretes of Series 9 with the lowest values of all of the mechanical parameters ( K IcS , JIc, E and fc), had the greatest value of the fractal dimension DC = 1.0253. The dependence of the fractal dimension DC on the water-cement ratio w/c and on the ratio of coarse aggregate to fine aggregate Cagg./Fagg. is shown in Fig. 6.

Fig. 6. Dependence of the fractal dimension DC on the water-cement ratio w/c and the coarse aggregate to fine aggregate ratio Cagg./Fagg.. The value of the coefficient of correlation for the adopted regression model amounted to 0.558 for all results being taken into account, and 0.865 when being based on mean values. From the obtained results, an increase in the fractal dimension with decreasing w/c is found, which is caused by the formation of a cement paste structure of increased strength. An important factor determining the strength of concrete is also a proportion of coarse aggregate to sand such as to obtain an aggregate stack with the highest density. The obtained diagram shape (Fig. 6) for the relationship of the fractal dimension DC versus coarse aggregate, and the observed optimum for the coarse aggregate to fine aggregate fraction Cagg./Fagg., equalling to approx. 2.4, indicate that the most dense aggregate stack has been obtained for the combination of sand with coarse aggregate at a weight proportion of 1:2.4.

RELATION BETWEEN MECHANICAL CONCRETE CHARACTERISTICS AND THE FRACTAL DIMENSION DC

Based on the obtained results of fractal examination and strength tests (Tables 2 and 3), an analysis of correlation between the fractal dimension and the mechanical properties of concretes was carried out. A statistically significant correlation (at a significance level of p = 0.05) was found to

42-Relationships.qxd

9/19/2009

12:02 PM

Page 416

416

Janusz KONKOL and Grzegorz PROKOPSKI

exist between the fractal dimension DC, as determined by the chord method, the modulus of elasticity E, the unit failure work JIc, and the compressive strength fc. With increasing fractal dimension, all of the above-mentioned characteristics of basalt concrete declined (Fig. 7). Relationships of the following form were obtained: E 1850.0 1775.0 DC at R -0.740,

J Ic

4426.4 4294.0 DC

at R

-0.733,

fc

6542.5 6340.0 DC

at R

-0.720.

(5)

For the critical stress intensity factor K IcS , no significant correlations between this parameter and the fractal dimension DC, were obtained.

Fig. 7. Dependence of a) the modulus E, b) unit failure work JIc, and c) compressive strength fc, on the fractal dimension DC.

42-Relationships.qxd

9/19/2009

12:02 PM

Page 417

Relationships between fractal dimension and the mechanical and structural parameters of basalt...

417

CONCLUSIONS

The tests and analyses carried out have enabled the following conclusions to be drawn. 1. The fractal analysis has confirmed that the profile lines of concrete fractures are fractals and therefore they can be subjected to any operations appropriate for fractal geometry. The employed method of determining the fractal dimension of concrete fracture profile lines has proved to be sensitive to change both in the water-cement ratio and in the coarse aggregate fraction of concrete. Greater fractal dimension values were obtained for a concrete with a higher w/c ratio. The relationship between the fractal dimension, as determined by the chord method, and the water-cement ratio is linear. Increasing the coarse aggregate fraction of concrete from the values of Cagg./Fagg. = 1.6 to the value of 2.3 caused a decrease in the fractal dimension DC. Further increasing the Cagg./Fagg. resulted in an increase in the fractal dimension. The least fractal dimension value obtained for the concrete of Series 3 with the lowest w/c and, at the same time, the best strength parameters, is likely to have a relationship with the aggregate stack being optimal in terms of tightness. 2. The fractographical examinations have confirmed the relationship between the nature of the fracture and the properties of concretes. The higher level of fracture surface complexity (greater values of RL and RS), the poorer mechanical properties were exhibited by concretes. This is caused by smaller cohesion forces existing at the aggregate/cement paste interface and the occurrence of a grain-wise fracture. In concretes of higher strength, cracking occurred across the aggregate grains (the inter-grain fracture), and the fracture surfaces were less rough (more flat) in that case. 3. The investigation carried out has demonstrated a relationship existing between the fractal dimension of the fracture surface of concretes examined and their mechanical properties, including fracture toughness. Significant correlations between the fractal dimension DC as determined by the chord method, and the coefficient of elasticity E, unit failure work JIc, and compressive strength fc, were only obtained for a broken aggregate-based concrete. However, no significant correlations between the fractal dimension and the critical stress intensity factor K IcS were found in that case. 4. The obtained investigation results indicate an opportunity of application of this method for concrete structure modeling and concrete behaviour simulation in cracking process. 5. Analytic relationships make possible identification of each parameter (E, JIc and fc) on the basis of fractal dimension of fracture profile line. REFERENCES

1. Mandelbrot B. B.: Fractals. Form, chance and dimension. Freeman, San Francisco 1977 2. Winslow D. N.: The fractal nature of the surface of cement paste. Cem. Concr. Res., Vol. 15, pp. 817-824, 1985 3. Brandt A. M., Prokopski G.: On the fractal dimension of fracture surfaces of concrete elements. Journal of Materials Science, 28, pp. 4762-4766, 1993 4. Prokopski G., Konkol J.: The fractal analysis of the fracture surface of concretes made from different coarse aggregates. Computers and Concrete, Vol. 2, No. 3, pp. 239-248, 2005 5. Yan A., Wu K., Zhang X.: A quantitative study on the surface crack pattern of concrete with high content of steel fiber. Cement and Concrete Research, Vol. 32, pp. 1371-1375, 2002 6. Saouma V. E, Barton C. C.: Fractals, fractures, and size effects in concrete. J.Enging Mech., Vol. 120, No. 4, pp. 835-854, 1994

42-Relationships.qxd

418

9/19/2009

12:02 PM

Page 418

Janusz KONKOL and Grzegorz PROKOPSKI

7. Carpinteri A., Spagnoli A., Vantadori S., Viappiani D.: Influence of the crack morphology on fatigue crack growth rate: A continuously-kinked crack model based on fractals. Engineering Fracture Mechanics, 75, pp. 579-589, 2008 8. Issa M. A., Hammad A. M., Chudnovsky A.: Correlation between crack tortuosity and fracture toughness in cementitious material. Int.J.Fract., 60, pp. 97-105, 1993 9. Determination of fracture parameters (KSIc and CTODc) of plain concrete using three-point bend test. RILEM Draft Recommendations, TC 89 - FMT Fracture Mechanics of Concrete Test Methods, Materials and Structures, 23, 1990 10. Wojnar, L.: Quantitative fractography. Basic principles and computer aided research. (in Polish) Scientific Booklets of the Cracow. Univ. of Techn., Mechanical Series, Booklet no. 2, Cracow 1990

43-T-stress values.qxd

9/19/2009

12:02 PM

Page 419

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

T-STRESS VALUES DURING FRACTURE IN WEDGE SPLITTING TEST GEOMETRIES: A NUMERICAL STUDY Stanislav SEITL1, Pavel HUTAě2, Václav VESELÝ3, ZbynČk KERŠNER4 Institute of Physics of Materials, Academy of Sciences of the Czech Republic Žižkova 22, 616 62 Brno, Czech Republic, e-mail: [email protected]; [email protected] 3,4 Institute of Structural Mechanics, Faculty of Civil Engineering Brno University of Technology, VeveĜí 331/95, 602 00 Brno e-mail: [email protected]; [email protected] 1,2

ABSTRACT The paper is focused on a detailed numerical analysis of the stress field in specimens used for the wedge splitting test (WST), which is a convenient alternative to the classical bending and tensile tests within the area of determination of the fracture-mechanical parameters of quasi-brittle building materials, particularly cementitious composites. The near-crack-tip stress field in the WST specimen is described by means of constraint-based twoparameter fracture mechanics in the paper. Particular attention is paid to the influence of usual variants of boundary conditions used for this kind of testing procedure on the stress field in the cracked body. The next part of the paper aims at investigation of how much the detailed description of the near-crack-tip stress field obtained by applying the two-parameter fracture mechanics approach is then utilizable for an estimation of the size and shape of the non-linear failure zone in quasi-brittle materials, i.e. the fracture process zone (FPZ). The results obtained with regard to the near-crack-tip stress field approximation are compared with data taken from the literature. An attempt is made to exploit the estimation of the FPZ extent within the determination of fracturemechanical characteristics of cementitious composites.

Keywords Concrete, wedge splitting test, T-stress, fracture process zone

INTRODUCTION Cement-based composites have been used in civil engineering construction for more than a century as the main construction material due to their significant resistance in compression. Much attention has been paid to the mechanical and fracture properties of cement-based composites recently, including particularly, tensile strength, toughness/brittleness and fracture energy. For determination of the values of the fracture-mechanical parameters of the materials, special experimental tests, e.g. three- or four-point bending of notched beams or tensile tests of notched or dog-bone specimens, are usually used, see e.g. Karihaloo [1], Shah et al. [2], Bažant and Planas [3], van Mier [4]. However, the influence of the bent specimen’s own weight at the performed test’s record is significant and cannot be ignored, especially in the case of materials with low tensile strength, e.g. early-aged concretes and mortars. Moreover, a substantial portion of the testing specimen’s volume remains elastic and does not directly participate in the failure test, which causes a useless increase of the fresh concrete amount necessary for the testing specimens casting or unnecessarily increases the demands on the size of specimens taken from existing structures for the test (e.g. by core drilling).

43-T-stress values.qxd

420

9/19/2009

12:02 PM

Page 420

Stanislav SEITL, Pavel HUTAØ, Václav VESELÝ and Zbynìk KERŠNER

The so called wedge splitting test (WST, Linsbauer and Tschegg [5], Brühwiler and Wittmann [6]) presents a convenient alternative to the above-mentioned bending or tensile tests. The WST is an interesting experimental procedure aimed at the determination of values of concrete fracture parameters by applying a compressive load to a compact notched specimen similar to that used for compact tension tests (CT) standard within fracture testing of metals. The compressive load applied to the specimen from an ordinary testing machine is transformed to a tensile loading opening the initial notch via a special testing arrangement based on the wedge mechanism (Fig. 1).

Figure 1: Wedge splitting test geometry: testing arrangement (left), variants of specimen shapes (right) (inspired by [5]) As is obvious from Fig. 1, the test specimens can be prepared either from standard cube or cylindrical specimens cast into standard moulds, or as prismatic or cylindrical specimens taken from existing structures by sawing or core-drilling, respectively. Nowadays the WST is extensively used for various experimental studies, e.g. [7, 8, 9, 10, 11].

MOTIVATION For determination of the values of the fracture-mechanical properties of cementitious composites by means of WST procedure, relevant information regarding the stress field in the tested specimen is necessary. A finite element analysis of the near-crack-tip stress field of WST geometry can be found for classical fracture mechanics e.g. in Guinea et al. [12]. However, as the fracture of quasi-brittle materials is non-linear, its description via classical fracture mechanics approach is insufficient and incorrect. Approximate expressions of the Williams series, i.e. the power expansion approximating the stress and displacement fields in the cracked specimen, up to the terms of order 5 were introduced in Karihaloo et al. [13]. The latter paper covers a rather wide range of exploitable dimensions of cubic-shaped WST specimens with two types of boundary conditions in the area of application of the load to the specimen by the wedge mechanism (see Fig. 2b and 2c). This paper introduces results of the numerical study of the stress field in the wedge splitting geometry conducted in the framework of two-parameter fracture mechanics and supplements the work [13] in some aspects, particularly in the variants of boundary conditions on the opposite part of the WST in relation to the load applied (see Fig. 2f). On the other hand, only the values of the stress intensity factor K and the T-stress along the crack propagation through the specimen ligament are computed, as is noted above.

43-T-stress values.qxd

9/19/2009

12:02 PM

Page 421

T-stress values during fracture in wedge splitting test geometries: A numerical study

421

Another improvement covered by the proposed paper in comparison to the results published in Karihaloo et al. [13] is the investigation of the influence of the vertical component of the loading force on the values of the near-crack-tip stress state parameters (see Fig. 2e, 2f, 2g). This issue is partly elaborated in previous works by the authors (e.g. [14, 15]).

Figure 2: Schema of WST specimen: Dimensions and possible variants of boundary conditions (a), dimensions and boundary conditions considered in Karihaloo et al [13] (b and c), dimensions and boundary condition considered in the study (d, e, f, and g) Knowledge of the stress field in the cracked specimen, or at least its accurate enough estimation, provides the possibility of constructing the size and shape of the fracture process zone (FPZ) characteristic for the fracture of quasi-brittle cementitious composites, as is proposed by Veselý et al. [16]. A technique of this kind enables the specification of the energy dissipated in the FPZ to its volume, which might contribute to a more precise determination of the fracture-mechanical parameters of quasi-brittle materials. The numerical analysis proposed in the paper has been motivated also by this issue.

THEORETICAL BACKGROUND Stress field description in a cracked body It is well known that the stress field ahead of the crack tip is not only influenced by the loading conditions, but also depends considerably upon the crack (specimen) geometry. This effect is described by constraint. Since the crack initiation of a structure depends upon the stress field in the crack tip region, the fracture parameters values depend on the geometry as

43-T-stress values.qxd

9/19/2009

422

12:02 PM

Page 422

Stanislav SEITL, Pavel HUTAØ, Václav VESELÝ and Zbynìk KERŠNER

well as the load. Due to this fact, the fracture toughness KIc in the case of brittle fracture (or other parameters, e.g. Jc or CTOD in the case of elasto-plastic fracture) is not a material parameter, but also a geometry dependent parameter. The T-stress and the Q-parameter, in the case of brittle and ductile fracture, respectively, are the two most widely used approaches to measure the constraint. The T-stress is used as a measure of constraint in the analyses presented in this paper. The T-stress is the first non-singular term in William’s solution of the stress field [17, 18], as shown in eq. (1). The T-stress helps to characterise the local crack tip stress field for materials with the restriction of small-scale yielding (SSY) conditions:

V xx

KI § T · § 3T ·º § T ·ª cos¨ ¸ «1 sin ¨ ¸ sin ¨ ¸» T 2 2Sr © 2 ¹ © 2 ¹¼ © ¹¬

V yy

KI § T · § 3T ·º § T ·ª cos¨ ¸ «1 sin ¨ ¸ sin ¨ ¸» 2Sr © 2 ¹ © 2 ¹¼ © 2 ¹¬

W xy

KI § T · § T · § 3T · cos¨ ¸ sin ¨ ¸ cos¨ ¸ 2Sr ©2¹ ©2¹ © 2 ¹

,

(1)

where r is the distance to the crack tip, T is the orientation angle, KI is the stress intensity factor (SIF) for normal mode I and T is the second constant term corresponding to a uniform parallel stress Vxx = T. Thus, in two-parameter-based fracture mechanics, the stress field in the cracked body is expressed by means of the two parameters, the stress intensity factor KI and the T-stress (see e.g. [19]). The SIF and the T-stress according to Leevers and Radon [20] may be normalized to be dimensionless as follows:

B1

KI , where K 0 K0

Psp t W

(2)

and

B2

T Sa , KI

(3)

where Psp is the splitting loading force, t is the thickness, W is the fundamental specimen’s dimension and a represents the crack length. In the present numerical study the crack length a is defined as a distance from the point of the force application to the crack tip, i.e. a = c + dn – h (see Fig. 2a). Both expected components of the loading force of WST specimen are shown in Fig. 2a. The horizontal component of the applied load – the splitting force Psp – is related to the vertical component – the compressive load Pv – by the following formula, see e.g. [21, 22]:

1 Psp k , 2

(4)

2 tan D w P c . 1 P c tan D w

(5)

Pv where k

43-T-stress values.qxd

9/19/2009

12:02 PM

Page 423

423

T-stress values during fracture in wedge splitting test geometries: A numerical study

Symbol Dw represents the wedge angle and Pc refers to the friction in the roller bearings (the values of Pc can be found e.g. in IEC [23]). To compare the values of normalized parameters B1 and B2, for all considered specimens, the variable parameter D was defined as a ratio of the crack length, i.e. a, and the distance from the point of the load application to the back face of the specimen, i.e. the effective width Wef = W – h:

D

c ( d n h) W h

a . Wef

(6)

The values of normalized parameters B1 and B2 are then calculated and plotted as a function of parameterD. Fracture process zone estimation The size and shape of the FPZ can be constructed based on an amalgamation of several approaches according to the concept proposed in Veselý et al. [16]. For the stress field approximation the multi-parameter linear elastic fracture mechanics (Williams [17]) is used. For the description of the stress field in the more distant surroundings of the crack tip, which is essential for quasi-brittle materials with a large FPZ, higher order terms of the Williams series are taken into account. The estimation of the effective crack tip during the fracture process is performed via equivalent elastic crack models [1]. The effective crack model enables calculation of the effective crack length in the cracked body from the change in the compliance of the body between the beginning and the current stage of the fracture process. Finally the cohesive crack models [24, 2, 3] are employed within the procedure of the FPZ range estimation. This approach helps with taking into account the non-linear (cohesive) material behaviour in the FPZ. These approaches are used within the processing of fracture test records; typically load– displacement diagrams (P–d diagrams). For individual stages of the fracture process the length of the equivalent elastic crack will be estimated by means of the effective crack model. The stress state in a body with an effective crack is approximated through Williams power series, whereas the number of terms of the series should correspond with respect to the mutual relation between the assumed FPZ size/shape and the size/shape of the body (with respect to the distance of the FPZ to the free boundaries of the body). The extent of the zone where the until-now elastic material starts to fail, is determined by comparing the tensile strength ft of the material to the proper characteristics of the stress state around the crack tip (e.g. principal stress or equivalent stress from any proper failure criterion). The crack opening displacement values at the propagating crack faces are calculated from appropriate LEFM formulas. In agreement with the cohesive crack approach, the FPZ is supposed to extend from the zone of the current failure around the crack tip, where the selected stress state characteristic exceeds the tensile strength ft up to a point at the crack faces where the value of crack opening displacement reaches its critical value (i.e. the value of cohesive stress drops to zero).

NUMERICAL STUDY

For numerical analysis of the stress field in the WST geometries, the following values of the dimensions were used: W = 100 mm, e = 35 mm, f = 30 mm, h = 10 mm, dn = 20 mm (see Fig. 2). The initiation notch of length c was modelled as a crack where its length varied so that parameter D lies in the interval from approx. 0.2 to 0.9. The material input data values for

43-T-stress values.qxd

424

9/19/2009

12:02 PM

Page 424

Stanislav SEITL, Pavel HUTAØ, Václav VESELÝ and Zbynìk KERŠNER

the concrete used in the numerical simulation were: Young’s modulus E = 44 GPa and Poisson’s ratio Q = 0.2. The coefficient of friction Pc usually varies between 0.001 and 0.005 for bearings used within the experimental setup; see e.g. IEC [23]. A detailed discussion concerning the influence of friction can be found in [21, 14]. In this study the value of coefficient Pc was chosen to be equal to 0. The stress intensity factor K and the T-stress values were computed either using direct methods or determined by means of quarter-point crack-tip elements (e.g. Tan and Wang [25]). For direct method the estimation of the fracture parameters is derived directly from the singular stress description, see eq. (1). As a first step, the 2D finite element method solution is employed to verify the accuracy of the numerical model used and to compare the data with those published by the authors in [14]. Note that the geometries are symmetric (including boundary conditions), therefore only one half of the body was modelled. A typical finite element mesh used in the computations [26] is shown in Fig. 3 (left), together with boundary conditions. A detailed view of the small region near the crack tip is shown in Fig. 3 (right). The size of the smallest element in the crack tip is 5 × 10-5 mm.

Figure 3: The finite element mesh used in the computations: One half of the WST specimen with detailed view of the small region near the crack tip (Quarter-point crack-tip element was used) Results The comparison of numerically determined values of the normalized stress intensity factor KI (i.e. B1 see eq. (2)) and T-stress (B2, see eq. (3)) with the values published by Karihaloo et al. [13] are given in graphs in Figs. 4 and 5. There are five curves plotted: i) the curve from [13] corresponding to the configuration depicted in Fig. 2b, ii) the calculated curve for WST loaded only by the splitting force Psp, see Fig. 2d, iii) the calculated curve for WST loaded by both components of the loading force, i.e. the splitting force Psp and the vertical compressive force Pv, see Fig. 2e, iv) and v) the calculated curves for WST loaded by both the splitting and the vertical compressive force with two supports with different distances, see Figs. 2f and 2g, respectively. The dependences of the normalized stress intensity factor B1 as a function of D for the considered WST specimens with various boundary conditions are presented in Fig. 4, where

43-T-stress values.qxd

9/19/2009

12:02 PM

Page 425

425

T-stress values during fracture in wedge splitting test geometries: A numerical study

90

Fig. 2b [Karihaloo et al. 2003] Fig. 2d [Seitl et al. 2009] Fig. 2e [Seitl et al. 2009] Fig. 2f Fig. 2g

80 70

B 1 [-]

60 50 40 30 20 10 0 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

D [-]

Figure 4: Normalized values B1of the stress intensity factor KI for considered specimen dimensions and boundary conditions

1

Fig. 2b [Karihaloo et al. 2003] Fig. 2d [Seitl et al. 2009] Fig. 2e [Seitl et al. 2009] Fig. 2f Fig. 2g

0.8

B 2 [-]

0.6 0.4 0.2 0 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

-0.2 -0.4

D [-]

Figure 5: Normalized values B2 of the T-stress for considered specimen dimensions and boundary conditions stress intensity factor KI is expressed by B1 defined in eq. (2). The graph shows that the normalized stress intensity factor values are similar for all cases studied. The normalized Tstress values from the conducted FEM analysis, expressed as B2 defined as a function of Tstress in eq. (3), are plotted in Fig. 5. According to the results obtained, it is obvious that the B2(D) functions for the WST geometry vary from negative values for short cracks to positive values for longer than approximately 0.25 in all cases studied. The influence of boundary

43-T-stress values.qxd

9/19/2009

426

12:02 PM

Page 426

Stanislav SEITL, Pavel HUTAØ, Václav VESELÝ and Zbynìk KERŠNER

conditions is significant especially in the interval D (0.3; 0.8), which is the interval commonly used within the measurement of fracture parameters in the case of mentioned WST. It should be noted, before the comparison of numerically obtained results with the data from literature [13] is made, that the authors of the mentioned paper [13] did not specify whether the vertical component of the load was considered in their computations or not. Fig. 5 shows good agreement of the WST variants depicted in Fig. 2b and 2d. This indicates that the vertical component of the loading force was not considered in [13]. Utilization of results, evaluation of FPZ size and shape The construction of an FPZ evolving during fracture in a quasi-brittle specimen is illustratively depicted in Fig. 6. It corresponds to an example of the geometry of three-point bending (TPB) of a notched beam of width W = 0.32 m with an initial relative crack/notch length a0/W = 0.1 and actual crack length a/W = 0.84. The specimen was supposed to be made of concrete with the following values fracture-mechanical parameters: tensile strength ft = 3.7 MPa, fracture energy GF = 93 Jm-2 and with exponential (Hordijk’s, see e.g. [27]) cohesive law. The fracture process was simulated numerically using ATENA software [27]. For the construction of the FPZ 1 and 4 terms of the Williams power series, respectively, and the Drucker-Prager failure criterion (in a plane stress state) are taken into account. α [-]

y [mm]

0.00 0.10

0.05

0.00 0.00

0.25

0.50

0.75

1.00

1.25

specimen boundaries FPZ - 1 term FPZ - 4 terms

0.05

0.10

0.15

0.20 0.25 x [mm]

0.30

0.35

0.40

Figure 6: A FPZ extent (a symmetric one-half) in a quasi-brittle specimen The authors of this paper are currently working on the determination of values of higher order terms of the stress field approximation via Williams’ series (similarly to [13]) for the WST geometry with different variants of boundary conditions. As is evident from Fig. 5, the boundary conditions (load and support) influence the values of T-stress significantly and it can be expected that higher order terms will be influenced even more, which follows from the trend indicated by Figs. 4 and 5. Therefore, the effect of the accuracy of the near-crack tip approximation on the size and shape of the FPZ in the quasi-brittle specimen is illustrated on the case of TPB geometry (instead of the WST geometry analysed here) for which the higher order terms are known [28, 29] but the these results did not take into account various boundary conditions. Determination of the higher order terms of the crack tip asymptotic field and the subsequent estimation of the FPZ size and shape are currently at the work-in-process stage.

CONCLUSIONS

Finite element analyses have been conducted to calculate the SIF and T-stress (B1, B2) for cracks in wedge splitting tests geometries with various boundary conditions. The analysis procedures and results were compared with existing solutions from the literature. From the

43-T-stress values.qxd

9/19/2009

12:02 PM

Page 427

T-stress values during fracture in wedge splitting test geometries: A numerical study

427

part of the work concerning the numerical analysis of the stress field in the WST specimen the following conclusions may be drawn: Normalized stress intensity factor values K(B1) are almost the same in the whole range the relative crack length for all cases of the WST geometry studied and are in good agreement with the results published in [13]. Normalized values of the T-stress (B2) depend on both consider boundary conditions in the part of application of the load (based on considering the compressive force or not) and on the opposite side of the specimen (one or two supports, different distances of the supports). Neglecting of the vertical component of the loading force increases stress constraint at the crack tip, this tends to underestimation of the size of the zone of failure. It consequently causes an overestimation of the values of the determined fracturemechanical parameters. An increase in the distance of the two supports on the bottom side of the specimen leads to an increase in the value of the T-stress. The results of the utilization of the procedure for the FPZ size and shape estimation shows that knowledge of the higher order terms of the crack tip stress field is necessary.

ACKNOWLEDGEMENTS

This outcome has been achieved with the financial support of the Ministry of Education, Youth and Sports, project No. 1M0579 (CIDEAS). In this undertaking, theoretical results gained in the project Grant Agency of Academy of Sciences, No. KJB200410901, and project Grant Agency of Czech Republic, No. 101/08/1623, were partially exploited. REFERENCES

1. 2.

3. 4. 5. 6. 7.

8.

9.

Karihaloo, B. L., Fracture mechanics of concrete. Longman Scientific & Technical, New York 1995 Shah, S. P., Swartz, S. E., Ouyang, C., Fracture mechanics of structural concrete: aplications of fracture mechanics to concrete, rock, and other quasi-brittle materials. John Wiley & Sons, Inc., New York 1995 Bažant, Z. P., Planas, J., Fracture and size effect in concrete and other quasi-brittle materials. CRC Press, Boca Raton 1998 van Mier, J. G. M., Fracture processes of concrete: Assessment of material parameters for fracture models. CRC Press, Boca Raton 1997 Linsbauer, H. N., Tschegg, E. K., Fracture energy determination of concrete with cubeshaped specimens. Zement und Beton, 31, 1986, 38–40 Brühwiler, E., Wittmann, F. H., The wedge splitting test, a new method of performing stable fracture mechanics test. Engineering Fracture Mechanics, 35, 1990, 117–125 Elser, M., Tschegg, E. K., Finger, N., Stanzl-Tschegg, S. E., Fracture behaviour of polypropylene-fibre reinforced concrete: an experimental investigation. Composite Science and Technology, 56, 1996, 933–945 Löfgren, I., Stang, H., Olesen, J. F., Fracture properties of FRC determined through inverse analysis of wedge splitting and three-point bending tests. Journal of Advanced Concrete Technology, 3, 2005, 423–434 Xu, S., Bu, D., Gao, H., Yin, S., Liu, Y., Direct measurement of double-K fracture parameters and fracture energy using wedge-splitting test on compact tension specimens

43-T-stress values.qxd

428

10.

11. 12. 13. 14.

15. 16.

17. 18.

19. 20. 21. 22. 23. 24.

25.

26. 27. 28.

29.

9/19/2009

12:02 PM

Page 428

Stanislav SEITL, Pavel HUTAØ, Václav VESELÝ and Zbynìk KERŠNER

with different size. In: proc. of Fracture Mechanics of Concrete and Concrete Structures – New Trends in Fracture, Al. Carpinteri et al. (eds), Catania, Italy, 271–278, 2007 Xiao, J., Schneider, H., Donnecke, C., Konig, G., Wedge splitting test on fracture behaviour of ultra high strength concrete. Construction and Building Materials, 18, 2004, 359–365 Kim, J. K., Kim,Y. Y., Fatigue crack growth of high-strength concrete in wedge-splitting test. Cement and Concrete Research, 29, 1999, 705–712 Guinea, G. V., Elices, M., Planas, J., Stress intensity factors for wedge-splitting geometry. International Journal of Fracture, 81, 1996, 113–124 Karihaloo, B. L., Abdalla, H., Xiao, Q. Z., Coefficients of the crack tip asymptotic field for wedge splitting specimens. Engineering Fracture Mechanics, 70, 2003, 2407–2420 Seitl, S., Dymáþek, P., Klusák, J., ěoutil, L., Veselý, V., Two-parameter fracture analysis of wedge splitting test specimen. In: Proc 12th Int. Conf. on Civil, Structural and Environmental Engineering Computing, B. H. V. Topping, L. F. Costa Neves and R. C. Barros (eds), Civil-Comp Press, Stirlingshire, 2009 Seitl, S. ěoutil, L., Veselý, V., Numerical analysis of stress field for wedge splitting geometry. In: Proc. Applied Mechanics 2009, Smolenice, Slovakia, 270–278, 2009 Veselý, V., Frantík, P., Keršner, Z., Cracked volume specified work of fracture. In: Proc 12th Int. Conf. on Civil, Structural and Environmental Engineering Computing, B. H.V. Topping, L. F. Costa Neves and R. C. Barros (eds), Civil-Comp Press, Stirlingshire, 2009 Williams, M. L., On the stress distribution at the base of stationary crack. ASME Journal of Applied Mechanics, 24, 1957, 109–114 Larsson, S. G., and Carlsson, A. J., Influence of non-singular stress terms and specimen geometry on small scale yielding at crack tips in elastic-plastic material. Journal of Mechanics and Physics of Solids, 21, 1973, 263–277 Knésl, Z., BednáĜ, K., Radon, J. C., Influence of T-stress on the rate of propagation of fatigue crack. Physical Mesomechanics, 2000, 5–9 Leevers, P. S. and Radon, J. C. Inherent stress biaxiality in various fracture specimen geometries. International Journal of Fracture, 19, 1983, 311–325 RILEM Report 5: Fracture Mechanics Test Methods for Concrete. S. P. Shah and A. Carpinteri (eds.), Chapman and Hall, London 1991 Skoþek, J., Stang, H., Inverse analysis of the wedge-splitting test. Engineering Fracture Mechanics, 75, 2008, 3173–3188 Interactive Engineering Catalogue: http://iec.skf.com. Hillerborg, A., Modéer, M., Petersson, P.-E., Analysis of crack formation and crack growth in concrete by means of fracture mechanics and finite elements. Cement and Concrete Research, 6, 1976, 773–782 Tan, C. L., Wang, X. The use of quarter-point crack-tip elements for T–stress determination in boundary element method analysis. Engineering Fracture Mechanics, 70, 2003, 2247–2252 ANSYS Users manual version 10.0, Swanson Analysis System, Inc., Houston 2005 ýervenka, V. et al., ATENA Program Documentation, Theory and User manual. Cervenka Consulting, Prague 2005 Knésl, Z., BednáĜ, K., Two parameter fracture mechanics: calculation of parameters and their values. Institute of Physics of Materials of Academy of Sciences of the Czech Republic, 1997 Karihaloo, B. L., Xiao, Q. Z., Higher order terms of the crack tip asymptotic field for a notched three-point bend beam. International Journal of Fracture, 112, 2001, 111–128

44-The influence.qxd

9/19/2009

12:02 PM

Page 429

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

THE INFLUENCE OF AGGREGATE SIZE ON THE WIDTH OF FRACTURE PROCESS ZONE IN CONCRETE MEMBERS Marta SàOWIK, Ewa BàAZIK-BOROWA Lublin University of Technology Nadbystrzycka 40, 20-618 Lublin, Poland, e-mail: [email protected], [email protected]

ABSTRACT In this paper the problem of the influence of aggregate size on concrete parameters is considered. The discussed parameters include fracture energy and width of the fracture process zone. The results of the numerical simulations concerning bent concrete members with different width of the fracture process zone are analyzed.

Keywords Fracture mechanics, concrete structures, aggregate size, width of fracture process zone

INTRODUCTION As the crack formation and crack growth play an important role in the rational design of concrete structures they should be based on realistic theoretical models. Recent advances in nonlinear fracture mechanics give a possibility to analyse crack propagation in concrete structures. It is possible to apply fracture mechanics rules to practical cases using finite element methods FEM. There are two ways of modelling cracking in concrete structures using a finite element analysis. In the first concept crack is considered as densely distributed throughout the finite area of element. The alternative approach assumes an isolated sharp interelement crack. The first concept of smeared crack is mostly used in numerical computations. As concrete is not a perfectly brittle material, not only tensile strength, but also tensile toughness is of particular importance to make safe concrete structures. The way of quantifying the toughness under tension is proposed by means of the fracture energy, as it is defined for example by Hillerborg [1]. However, in classical theories of designing concrete structures, tensile toughness is not taken into account. In general, material fracture properties should be characterized by three main parameters - axial tensile strength (fct ), fracture energy (GF ) and shape of the stress-deformation diagram (VG ) – see Fig. 1. The stress-deformation properties of concrete are given by two curves: stress-strain (VH) and stress-crack opening curve (Vw becauseafterreaching the tensile strength, stress starts to decrease whereas deformation increases within the damage zone and decreases in the remaining part of a member.

44-The influence.qxd

9/19/2009

12:02 PM

Page 430

430

Marta S£OWIK and Ewa B£AZIK-BOROWA

V

V

V

f ct

f ct

fct

w

=

+ GF G

H

w

Fig. 1. Geometrical interpretation of concrete fracture parameters

wc

The decrease in stress under increasing deformation is called strain softening and it takes place in the narrow zone where the progressive microcracking appears – see Fig. 2. The width of the microcrack band, which is called the width of fracture process zone (wc), is the additional parameter taken into account when fracture in concrete is modelled as a smeared crack band.

Fig. 2. The width of fracture zone wc Main fracture parameters of tensile concrete should be determined in a deformationcontrolled tensile test. Unfortunately such a test is difficult to perform, so in practice tensile strength of concrete is determined in splitting tensile test and fracture energy in three-point bend test. These tests were chosen for the proposed RILEM recommendations because it is much easier to perform them than a stable tensile test. The determination of the width and the length of the fracture process zone is a difficult experimental problem and there are no standard methods of its measurement. The extent of this zone can be revealed by measuring the resulting strain localization while strain gauges can only measure average values of the deformation over their base length. Other methods, like conventional interferometry techniques, do not have the necessary sensitivity. Some experiments on standardization of fracture process zone measurement have been performed, for example by Cedolin, Poli, Iori [2] and Tang, Yang, Zollinger [3], yet they have not led to finalization.

44-The influence.qxd

9/19/2009

12:02 PM

Page 431

The influence of aggregate size on the width of fracture process zone in concrete members

431

THE INFLUENCE OF AGGREGATE SIZE ON CONCRETE PROPERTIES Fracture energy GF, which characterizes softening branch of stress-strain curve, is the basic parameter of tensile concrete. In the model of crack band, the width of fracture process zone wc is an additional parameter of concrete fracture. In the fracture process zone, microcracking takes place in mortar and in the contact zone where mortar is connected with aggregate grains and so a real crack may develop in the loose material. According to previous experimental and analytical studies, the fracture properties of concrete depend on the aggregate sizes. The influence of the maximum aggregate size on fracture energy GF was considered in several papers. Zhao, Kwon, Shah [4, 5] performed three-point bend tests for a notched beam and wedge splitting tests of different size specimens for concretes with different maximum size of aggregate. Comparing fracture energies, they discovered that the fracture energy increased with an increase of the maximum aggregate size and with an increase of the specimen size. In Kleinschrodt and Winkler experiment [6] the doubling of the maximum aggregate from Dmax= 8 mm to Dmax= 16 mm caused an increase of GF value by 25%. On the basis on the test results, Rossello, Elices, Guinea [7] observed that the effectiveness of the bond between aggregate and a matrix had a significant influence on fracture energy. Higher GF values were obtained in concrete with aggregate well bonded to the matrix. It may be concluded that, because of the fact that the intensity of crystal joints depends on grain-size distribution of aggregate, also the strength of matrix-aggregate interface correlates with aggregate sizes. The experimental investigation performed in Lublin University of Technology by Golewski [8] confirmed the influence of maximum aggregate size on fracture parameters of concrete. Two types of limestone aggregate compositions, with Dmax= 8 mm and Dmax= 16 mm, were used in the tests. It was found out that higher fracture energy was obtained in concrete with aggregate size up to 16 mm. The difference of obtained values of GF was nearly 60%. The influence of maximum aggregate size on fracture energy is reflected in analytic proposition of estimating the GF, for example the formula proposed by Bažant, Oh [9]:

GF

3.1u10

6

f ct 2.57 f ct2

Dmax , Ec

(1)

where GF is given in [Nm/m2], fct and Ec in [Pa], and Dmax in [m]. Also in CEB-FIP Model Code [10] there is the formula for estimating GF:

GF

D F f c0.7 [Nm/m2],

(2)

where DF depends on Dmax and is equal DF=4; 6; 10 when Dmax=8; 16; 32 mm, fc is compressive strength of concrete in [MPa]. For measurements of energy absorption, a three-point bend test on a beam with a central notch has been proposed by RILEM Technical Committee 50 [11]. The sizes of the recommended standard specimens are shown in table 1 and they depend on a maximum aggregate size Dmax.

44-The influence.qxd

9/19/2009

12:02 PM

Page 432

432

Marta S£OWIK and Ewa B£AZIK-BOROWA

Table 1. Size of specimens for measuring GF [11] Dmax [mm] 1 y 16 16.1 y 32 32.1 y 48 48.1 y 64

Depth d [mm] 100 r 5 200 r 5 300 r 5 400 r 5

Width b [mm] 100 r 5 100 r 5 150 r 5 200 r 5

Length L [mm] 840 r 10 1190 r 10 1450 r 10 1640 r 10

Span l [mm] 800 r 5 1130 r 5 1385 r 5 1600 r 5

As already pointed out, there are no standard methods to determine the width of fracture process zone experimentally. It may be approximately calculated from the equation:

wc

2GF f ct2

§1 1 · ¨¨ ¸¸ © E Et ¹

1

(3)

where E is elastic modulus of concrete, and Et is tangent softening modulus of a declining segment of the stress-strain diagram. This is just an effective width corresponding to linear stress-strain diagram and to assumed uniform strain distribution within the fracture process zone. Some authors point out that wc is an independent material parameter, which can differ from concrete to concrete, and it depends on the maximum size of aggregate Dmax. The ratio of wc to Dmax, presented in professional technical literature, ranges from 1.0 to 5.0 for various kinds of concrete. Bažant and Oh [9] came to a conclusion that the boundary of the localized cracking region should not be limited only as the boundary of visible microcracks but as the boundary of the whole strain-softening region. In their opinion the strain softening is caused not only by microcracking but also by ruptures of the bond between aggregate and concrete matrix and so the fracture process zone could be wider than the region of visible microcracks. Analyzing several test data from the literature, they have concluded that in practical cases it is generally possible to assume that the optimum width of the crack band is about three-times the maximum aggregate size. The value of wc/Dmax= 3 is only an approximation and, as it may be expected, is a function of concrete strength. It depends on the difference between the elastic modulus of the large aggregate pieces and the surrounding mortar. When the difference becomes smaller, the material becomes more homogenous and wc/Dmax should decrease. Another possible reason for the influence of aggregate graining on the value of fracture energy and the width of fracture process zone, given by Hu, Duan [12], is the non-uniform distribution of local fracture energy. This is because the cohesive or crack bridging stresses in concrete have to come from a frictional pull of aggregates and a tearing of various unbroken connections over the uneven main crack surface created by multiple cracking, which occurs in fracture process zone of certain width and length. The presence of large size aggregates prevents the crack from opening and results in wider fracture process zone. Hu and Duan have noticed that the length of the fracture process zone or the cohesive stress zone is also closely related to wc. Interesting experiments were performed by Otsuka and Date [13]. To investigate the behavior of the fracture process zone in concrete, they used X-rays with contrast medium and three-dimensional acoustic emission techniques. They carried out experiments on four differently sized specimens. All specimens were made with concrete with specified cylinder strength of 20 MPa but different maximum aggregate size equaling 5, 10, 15 and 20 mm.

44-The influence.qxd

9/19/2009

12:02 PM

Page 433

The influence of aggregate size on the width of fracture process zone in concrete members

433

When comparing fracture process zone traced from X-ray films of concrete with different Dmax, they observed a significant influence of aggregate size on the width of the microcrack zone. The results obtained by acoustic emission technique showed the relationship between the width and the length of fracture process zone. Based on the results of experiments performed on specimens of identical size, they concluded that, with the increase of maximum aggregate size, the width of fracture process zone increased whereas the length of fracture process zone decreased. The comparison of fracture process zone dimensions, obtained by acoustic emission of concrete with two different Dmax: 10 mm and 20 mm, is shown in Fig. 3.

Fig.3 Comparison of fracture process zone of concrete with two different Dmax ,[13] Quite different experimental results were obtained in tests performed by WoliĔski [14]. He determined fracture mechanics parameters of concrete on the basis of deformationcontrolled uniaxial tensile tests for five different concretes of the same class but with different maximum aggregate size (Dmax= 2, 4, 8, 16 and 32 mm). He did not find marked relationships between the fracture parameters and maximum aggregate size. The mean value of the width of fracture process zone obtained by WoliĔski was 26.6 mm and it did not depend on Dmax. There are no consistent conclusions as to whether the width of fracture process zone depends on the maximum aggregate size. The task of standardizing the testing procedure and the method of estimating the width of fracture process zone has not been undertaken yet. Therefore, there are difficulties with performing numerical simulations of concrete structures based on crack band model of nonlinear fracture mechanics in which it is necessary to model the width of fracture process zone. The question arises as to how the choice of the width of fracture process zone influences the results of numerical calculations. This problem is discussed in this paper.

NUMERICAL SIMULATIONS

To analyze how the choice of the width of the fracture process zone influences the results of numerical calculations in the case of concrete beams, our own calculations have been performed using the commercial program ALGOR, which is based on finite element method. When performing the numerical simulations, six concrete beams were computed with the rectangular cross section and the following dimensions: width b = 0.15 m, height h = 0.30 m, total length L = 3.00 m, span l = 2.70 m. The four-point bend scheme was chosen for the simulations. Beams were loaded symmetrically by two concentrated forces, which were applied from bottom towards the top. Beam geometry and the static scheme of the analyzed specimen are shown in Fig. 4.

44-The influence.qxd

434

9/19/2009

12:02 PM

Page 434

Marta S£OWIK and Ewa B£AZIK-BOROWA

Fig. 4. The analyzed specimen The FEM-analysis was performed on one half of the concrete beam since the four-point test is symmetrical. The fracture process zone was modeled in the region of the biggest bending moment. To analyze the influence of the width of the fracture process zone on the results of numerical calculations different widths were taken at modeling this zone: wc = 5; 10; 20; 26.5; 50 and 100 mm. The finite element mesh for the analyzed beam in the case of wc = 10 mm is shown in Fig. 5.

Fig. 5. The FEM-mesh for a beam with wc=10 mm While performing FEM calculations, the following material properties were taken: - the tensile strength fctm = 1.5 MPa; - the compressive strength fcm = 20.5 MPa; - the modulus of elasticity Ecm = 22 GPa; - the fracture energy GF = 83 Nm/m2; - the maximum size of aggregate Dmax = 32 mm. In the region of the fracture zone the concrete was modeled as nonlinear material whereas outside this zone it was modeled as linear elastic one. To describe the fracture region of concrete the relations VH and V-w proposed in CEB-FIP Model Code [10] were used (Fig. 6). It has been shown in [15], that such diagrams are suitable for FEM analysis of the cracking in flexural members.

44-The influence.qxd

9/19/2009

12:02 PM

Page 435

435

The influence of aggregate size on the width of fracture process zone in concrete members

a)

b)

fct 0.9 fct

fct

w0, w1- from the CEB-FIP Model Code

E 1 0.15 fct

GF w1

0.00015

w0

w

Fig. 6. Diagrams recommended by CEB [10] for the tensile zone of concrete: (a) stress versus strain, (b) stress versus crack opening As a result of the numerical simulations, the dislocations of nodes and stress components along three axes of the global coordinate system were obtained. On that basis, the elongation on the base 250 mm long, situated in the tensile zone where the crack appeared, was calculated with different width of fracture process zone. The results of the calculations for all beams were compared in succeeding load stages and presented in Fig. 7. 7.0

- the beam is modelled by an elastic material - the beam with wc=5.0 mm - the beam with wc=10.0 mm - the beam with wc=20.0 mm

6.0

- the beam with wc=26.5 mm - the beam with wc=50.0 mm - the beam with wc=100.0 mm

F [kN]

5.0

4.0

3.0

2.0

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

Elongation [mm]

Fig. 7. Comparison of the elongation for beams with different wc Furthermore, diagrams of normal stress distribution along the fracture process zone for analyzed beams have been made. To compare all obtained diagrams of normal stress for concrete beams with different wc, they have been juxtaposed (Fig. 8).

44-The influence.qxd

9/19/2009

12:02 PM

Page 436

436

Marta S£OWIK and Ewa B£AZIK-BOROWA

0.30 - the beam with wc=5.0 mm

0.27

- the beam with wc=10.0 mm - the beam with wc=20.0 mm

0.24

- the beam with wc=26.5 mm - the beam with wc=50.0 mm

z [m]

0.21

- the beam with wc=100.0 mm

0.18 0.15 0.12 0.09 0.06 0.03 -2000

-1500

-1000

-500

0

500

1000

1500

Vxx [kPa]

Fig. 8. Comparison of normal stress distribution along the fracture zone in beams with different width wc at the same load stage F = 6 kN When analyzing the diagrams presented in Fig. 7 and 8, we can observe the differences in calculation results when compared with the width of the fracture process zone taken in FEMcalculations. Greater concrete elongation has been obtained in the cases of the modeled beams where wc was more than 20 mm. Comparing the stress distributions presented in Fig. 8, we can noticed that the greater the width of fracture zone, taken in FEM-calculation, the less intensive strain softening of tension concrete. In order to analyze the influence of the fracture zone width on numerical results more precisely, concrete strains within the modeled fracture process zone have been calculated (Fig. 9). 2.0 - F= 3.00 kN - F= 4.50 kN - F= 5.10 kN - F= 6.00 kN - F= 6.30 kN - F=6.72 kN - F=6.96 kN

1000•strain [/]

1.6

1.2

0.8

0.4

0.0 0

10

20

30

40

50

60

70

80

90

100

wc [mm]

Fig. 9. Concrete strain within the fracture zone at different wc Taking into account the minimum potential energy in a member, it may be said that the most rational thing to do is to take the smallest elongation within the localized microcracking where the crack appears. In analyzed beams this condition takes place when wc is between 5

44-The influence.qxd

9/19/2009

12:02 PM

Page 437

The influence of aggregate size on the width of fracture process zone in concrete members

437

and 10 mm. If we take wc = 3Dmax as proposed in the literature [9] (in analyzed beam it would be wc = 100 mm, because Dmax = 32 mm) such an assumption does not fit this criterion. Also the width from the experiment performed by WoliĔski [14], wc = 26.5 mm, is not in good relation to obtained FEM-calculation results.

CONCLUSIONS

On the basis of the state of knowledge presented after several papers, it may be concluded that the influence of maximum aggregate size on the basic fracture parameters of concrete does exist. In the case of tensile strength and fracture energy of concrete this problem has been described in depth. There are no definite conclusions as far as the influence of aggregate size on the width of the fracture process zone is concerned. The proper choice of this parameter during the numerical calculation is a condition of obtaining correct results performed by finite element method. The numerical analysis shows that the width of the fracture process zone has an influence on the FEM results but it does not confirm that wc depends on the maximum aggregate size. This problem requires further systematic experiments. REFERENCES

1. Hillerborg A., Modeer M., Petersson P. E.: Analysis of crack formation and crack growth in concrete by means of fracture mechanics and finite elements. Cement and Concrete Research, Vol. 6, 1976, 773-782 2. Cedolin L., Poli S. D. and Iori I.: Experimental determination of the fracture process zone in concrete. Cement and Concrete Research, Vol. 13, 1983, 557-567 3. Tang T., Yang S., and Zollinger D. G.: Determination of fracture energy and process zone length using variable-notch one-size specimens. ACI Materials Journal, Vol. 96, No 1, 1999, 3-10 4. Zhao Z., Kwon S. H. and Shah S. P.: Effect of specimen size on fracture energy and softening curve of concrete: Part I. Experiments and fracture energy. Cement and Concrete Research, Vol. 38, Issue 8-9, 2008, 1049-1060 5. Kwon H., Zhao Z. and Shah S. P.: Effect of specimen size on fracture energy and softening curve of concrete: Part II. Inverse analysis and softening curve. Cement and Concrete Research, Vol. 38, Issue 8-9, 2008, 1061-1069 6. Kleinschrodt H. D. and Winkler H.: The influence of the maximum aggregate size and the size of specimen on fracture mechanics parameters, fracture toughness and fracture energy of concrete. Edited by F. H. Wittmann, Elsevier Science Publishers B. V., Amsterdam 1986, 391-402 7. Rossello C., Elices M. and Guinea G. V.: Fracture of model concrete: 2. Fracture energy and characteristic length. Cement and Concrete Research, Vol. 36, Issue 7, 2006, 13451353 8. Golewski G. J.: Influence of Dmax on fracture mechanics parameters of concrete made of limestone aggregate at three point bending (in Polish). Civil Engineering and Architecture, Vol.1/2007, Lublin 2007, 5-16 9. Bažant Z. P. and Oh B. H.: Crack band theory for fracture of concrete. Matériaux et Constructions, Vol. 16, 193, 1983, 155-177 10. CEB-FIP Model Code 1990 Bulletin d’information No. 196 11. RILEM Draft Recommendation: Determination of the fracture energy of mortar and concrete by means of three-point bent tests on notched beams. Matériaux et Constructions, Vol. 18, No 106, 1985, 258-290

44-The influence.qxd

438

9/19/2009

12:02 PM

Page 438

Marta S£OWIK and Ewa B£AZIK-BOROWA

12. Hu X. and Duan K.: Influence of fracture process zone height on fracture energy of concrete. Cement and Concrete Research, Vol. 34, Issue 8, 2004, 1321-1330 13. Otsuka K. and Date H.: Fracture process zone in concrete tension specimen. Engineering Fracture Mechanics, Vol. 65, 2000, 111-131 14. WoliĔski Sz.: Tensile behaviour of concrete and their applications in nonlinear fracture mechanics of concrete (in Polish). Scientific Works of Rzeszów University of Technology, No 15/91, Rzeszów1991, pp 210 15. Sáowik M., Báazik-Borowa E.: Models of concrete in tension – experimental verification (in Polish). XLV Scientific Conference KILiW PAN i KN PZITB, Krynica 1999, Vol. 4, 59-66

45-Effect of steel.qxd

9/19/2009

12:02 PM

Page 439

Proc. Int. Symp. "Brittle Matrix Composites 9" A.M. Brandt, J. Olek and I. H. Marshall, eds. Warsaw, October 25-28, 2009 IFTR and Woodhead Publ., Warsaw 2009

EFFECT OF STEEL STRIP GEOMETRY ON PULL-OUT STRENGTH OF AERATED CONCRETE Dariusz ALTERMAN1, Juan VILCHES1, Thomas NEITZERT1, Hiroshi AKITA2 1 AUT University, 24 St Paul Street, Auckland, New Zealand e-mail: [email protected], [email protected], [email protected] 2 Tohoku Institute of Technology, 35-1 Yagiyama Kasumicho, Sendai 982-8577, Japan e-mail: [email protected]

ABSTRACT An investigation of shearing strength between steel strips and concrete specimens is presented in this paper. Pullout tests were carried out to check the influence of various geometrical parameters of steel strips on the shearing strength in aerated concrete. The size-effect of the various strip widths with and without holes, area of holes and circumference area of holes were analysed. All these parameters were compared to a total area of the strip among different sets and ratio/coefficients were proposed. The tests were performed on aerated concrete cubes with galvanized steel strips of 0.75mm thickness.

Keywords Aerated concrete, pull-out test, steel strips, shearing force

INTRODUCTION Composite structures made of various materials can fail in several ways depending on material properties of components, design methods and loading cases. Thus experiments, which can assist the understanding of the fundamental failure phenomena between components of composites, are very important. In this paper, an effort was undertaken to determine forces between steel elements and concrete samples through pull-out tests. Most of the previous investigations on pull-out tests have been performed to analyse only the shear stress between steel bars and concrete specimens. The lack of compatibility between steel bars and concrete specimens leads to bond failure, sliding of reinforcement bars or strips, local deformations and finally cracking. Generally, experimental and theoretical studies focused on the shearing behaviour of reinforcing bars and analytical bond-slip models [1-8]. Destructive measurement of the shear strength by pull-out, push-in, and related testing methods is commonly used to assess the quality of a connection between steel elements and concrete [3]. The shear strength depends upon the mechanical properties of steel , concrete and the surface properties of steel bars and concrete [1]. The bond behaviour can be characterized by a mode of failure, bond strength and bond– slip relationships [9]. A new study was undertaken in this paper to find the shearing strength between the steel strips and concrete specimens. Pull-out tests have been carried out to evaluate the influence of various geometrical parameters of steel strips embedded in foam concrete.

45-Effect of steel.qxd

440

9/19/2009

12:02 PM

Page 440

Dariusz ALTERMAN, Juan VILCHES, Thomas NEITZERT and Hiroshi AKITA

EXPERIMENTAL METHOD This paper investigates the methodologies used to evaluate the behaviour between steel strips and concrete samples [10, 11]. The direct shearing strength was obtained through pull-out tests. For this purpose, the concrete specimens were prepared with the dimension of 100x100x100 mm for steel strips of 25 and 50 mm in width and with the dimension of 200x200x100 mm for steel strips of 75 and 100 mm in width. Steel strips were placed in the middle of specimens and embedded 100 mm into the foam concrete as in Fig. 1. Lightweight concrete consisting of foam agent, cement, fly ash, PVA fibers and water was used as well as galvanized steel strips (Bluescope NZ Steel product G250) of 0.75 mm in thickness with various diameters, numbers and distribution of holes. Three specimens for each single set of strips were prepared and finally 75 specimens were tested to determine the shearing strength through pull-out tests after 28 days from casting.

Fig. 1 Placement of steel strips in concrete specimens The pull-out tests were carried out using a universal testing machine equipped with a frame, which holds cubic samples as shown in Fig. 2. The load was applied to a strip through a mechanical joint and evenly increased while controlling the displacement at a rate of 0.1 mm/s. The pull-out forces had to be centric to the strips to avoid any eccentricity through loading.

45-Effect of steel.qxd

9/19/2009

12:02 PM

Page 441

Effect of steel strip geometry on pull-out strength of aerated concrete

441

Fig. 2 View of the testing rig for pull-out tests Both, the load applied and displacement of a steel strip were measured and recorded on a computer until a strip was removed from a cubic sample, which allows evaluating the energy needed to pull a strip out. Thus, a load-deformation curve was obtained from each experiment. In this paper maximum loads are only analysed. Pull-out strength was related to the whole area of a strip. The first tests failed for all cubic specimens with the dimension of 100x100x100mm and steel strips of 75 and 100 mm in width and had to be repeated. Every experiment was performed for three similar sets of steel strips and the average result is reported this in this paper. Although, the composition of concrete mixtures and density were the same but compressive strength was different and finally the analysis of results was limited to the set of strips of 50 mm in width in this paper. Another advantage for choosing steel strips of 50mm width was the lowest standard deviation factor for all tests of the same set of holes.

ANALYSIS OF RESULTS Size effect of steel strips The tests were proposed to find out whether there is any correlation between widths of steel strips without holes and pull-out forces. A difference of 25% at pull-out forces between strips of 75mm and 100mm in width was observed as in Fig. 3 and this is exactly related to the surface area between these sets. Thus, pull-out forces increase proportional with the width of strips. There was not noticed the same difference between strips of 25mm and 50mm in width because variation of results was about 50% for 25 mm wide strips.

45-Effect of steel.qxd

442

9/19/2009

12:02 PM

Page 442

Dariusz ALTERMAN, Juan VILCHES, Thomas NEITZERT and Hiroshi AKITA

Fig. 3 Size effect between pull-out forces and width of strips Influence of the number of holes with the same area on pull-out forces The influence of the number of holes with the same hole area on pull-out forces was studied. The pull-out force increased to 13% for strips with increasing the number of holes from 2 to 4 holes as in Fig. 4 and to 16% for strips from 3 and 6 holes as in Fig. 5.

Fig. 4 Influence of 2 and 4 holes with the same area on pull-out forces

45-Effect of steel.qxd

9/19/2009

12:02 PM

Page 443

Effect of steel strip geometry on pull-out strength of aerated concrete

443

Fig. 5 Influence of 3 and 6 holes with the same area on pull-out forces

PROPOSITION OF HOLE RATIOS Experiments confirmed the advantage of holes cut out from strips for increasing pull-out forces. An attempt of finding appropriate parameters/ratios, which could describe relations between the contact strip area and the geometrical parameters of holes, was undertaken. Also an analysis of geometrical parameters of holes and shearing strength of concrete filling the holes was made. Three most suitable factors are proposed as presented in the followings paragraphs. Contact-strip-area ratio The influence of total contact strip area on pull-out behaviour was studied and the Contact Strip Area Ratio (CSAR), (1) describing such relation is proposed. CSAR = (As-Ah)/As where:

(1)

As - steel area Ah - hole area.

Pull-out forces increased when a contact-strip-area between concrete and steel strips was diminished. It just shows the favourable influence of concrete shearing strength for final bonding strength, which is presented in Fig. 6.

45-Effect of steel.qxd

444

9/19/2009

12:02 PM

Page 444

Dariusz ALTERMAN, Juan VILCHES, Thomas NEITZERT and Hiroshi AKITA

Fig. 6 Influence of contact strip area ratio on relative pull-out forces Hole circumference area ratio The relationship between hole circumference area and pull-out forces was analysed and an appropriate Hole Circumference Area Ratio (HCAR), (2) was considered. HCAR = CAh/(As-Ah) where:

CAh – circumference area of holes

The strips with the greatest circumferences area of all holes occurred to have higher pullout strength, which is presented in Fig. 7

Fig. 7 Influence of hole circumference area on relative pull-out forces

(2)

45-Effect of steel.qxd

9/19/2009

12:02 PM

Page 445

Effect of steel strip geometry on pull-out strength of aerated concrete

445

Hole-area ratio The total hole area of strips was analysed and a third ratio to describe relations between shearing strength of concrete and area of holes was proposed as the Hole Area Ratio (HAR), (3). HAR = Ah/(As-Ah)

(3)

Strips with higher area of holes occurred to be stronger then strips without holes. The relations are shown in Fig. 8.

Fig. 8 Influence of the hole area ratio on relative pull-out forces

EFFECT OF OTHER PARAMETERS ON PULL-OUT STRENGTH Loading velocity The influence of loading velocity on pull-out tests has been examined after 7 days from casting of concrete samples. For this purpose concrete specimens with embedded steel strips of 25 mm in width and without holes were used. Four values of velocity were selected as shown in Table 2. Table 2 The effect of velocity on pull out force Velocity [mm/s] 0.5 0.1 0.05 0.01

Average Max. Load [kN] 1.62 1.59 1.39 2.19

Standard Deviation [kN] 0.66 0.53 0.08 0.12

45-Effect of steel.qxd

446

9/19/2009

12:02 PM

Page 446

Dariusz ALTERMAN, Juan VILCHES, Thomas NEITZERT and Hiroshi AKITA

The velocities for values of 0.05 and 0.01 mm/s are acceptable based on the lower standard deviation factor between 3 tested samples for each case. These experiments will be revisited through further tests. Failure of steel strips The steel strips were made from G250 steel with a minimum guaranteed ultimate tensile strength of 320 [MPa] and the minimum ultimate strength for each set of strips was calculated. For some cases the pull-out forces occurred to be higher than the strength of the strips because of the positive influence of the contact bonding strength between strips and concrete. The failure cases for the steel strips with two rows of holes of 75 mm in width are presented in Fig. 10.

Failure of steel strips

Figure 10 View of steel strips failure FEM analysis has been also undertaken to confirm the maximum stress for a failure of the steel strips and the results were nearly similar to the real cases of pull-out tests. Splitting of concrete sample First tests were performed with the same size of concrete samples of 100x100x100 mm for all set of strips. Unfortunately, samples for steel strips of 75 and 100 mm in width were splitting at the beginning of loading as shown in Fig. 11. Other concrete samples of the dimension of 200x200x100 mm were cast and tests were repeated. Thus, there is some critical distance between a strip edge and cubic concrete sample edges which has to be ensured to get reliable results and properly performed tests.

45-Effect of steel.qxd

9/19/2009

12:02 PM

Page 447

Effect of steel strip geometry on pull-out strength of aerated concrete

447

Figure 11 View of a split sample for a steel strip of 100 mm in width

CONCLUSIONS AND FURTHER RECOMMENDATIONS The influence of several geometrical parameters as number of holes, size, total area, and pattern of steel strips were investigated and the following conclusions are drawn: x No influence of the size effect of steel strips for pull-out forces was observed. Thus, pullout strength rises in proportion to the area of a strip for cases without holes. x Pull-out strength can be increased by about 16% with doubling the number of holes while keeping the same area of holes. This was observed in two cases by increasing the number of holes from 2 to 4 and from 3 to 6. x The pull-out strength was increased by 70% for strips with holes in comparison to strips without holes. This means adhesion of a contact layer between steel strips and concrete is less important than the strength of concrete, which filled up the holes. x An analysis of three ratios as the contact-strip-area ratio, the hole-circumference- area ratio and the hole area ratio, describing geometrical parameters of steel strips was undertaken. x Strength of steel strips has to be considered for each set of strips to avoid failure. Next steps: x An optimisation between area of holes and thickness, trength of steel strips should be investigated and a factor based on such parameters should be developed. x The influence of mechanical properties of concrete samples as compressive strength, modulus of elasticity, Poisson’s ratio and splitting strength should be analysed to find out the shear strength of concrete. x Results from such tests should be used for further numerical simulations of pull-out behaviour between concrete samples and steel strips.

45-Effect of steel.qxd

448

9/19/2009

12:02 PM

Page 448

Dariusz ALTERMAN, Juan VILCHES, Thomas NEITZERT and Hiroshi AKITA

x Different types of supports e.g. an adjustable ball-joint, a plate with bolts and nuts placed at the bottom of a sample and holding arms should be studied to avoid a lateral influence of loading (clamping forces) as well as the eccentricity of steel strips acting from a testing machine. ACKNOWLEDGEMENTS The work presented in these paper origins from a collaborative research program Composite Structure Assemblies. The authors are grateful to Dr Wolfgang Scholz from the Heavy Engineering Research Association of New Zealand for his advice and The Foundation for Research, Science and Technology (FRST) in New Zealand for the financial support. REFERENCES 1. Bouazaoui L., Li A., Analysis of steel/concrete interfacial shear stress by means of pull out test, International Journal of Adhesion & Adhesives, Vol. 28, Issue: 3, April, 2008, pp. 101-108 2. Banholzer B., Brameshuber W., and Jung W., Analytical evaluation of pull-out tests-the inverse problem, Cement & Concrete Composites, 2006. Vol. 28: pp. 564–571 3. Cao J. and Chung D., Degradation of the bond between concrete and steel under cyclic shear loading, monitored by contact electrical resistance measurement, Cement and Concrete Research, 2001. Vol. 31: pp. 669-671 4. Won J., et al., Effect of fibers on the bonds between FRP reinforcing bars and highstrength concrete, Composites: Part B, 2008. Vol. 39: pp. 747–755 5. Al-Mahmoud F., et al., Effect of surface pre-conditioning on bond of carbon fibre reinforced polymer rods to concrete, Cement & Concrete Composites, 2007. Vol. 29: pp. 677–689 6. Schilde K. and Seim W., Experimental and numerical investigations of bond between CFRP and concrete, Construction and Building Materials, 2007. Vol. 21: pp. 709–726 7. Tanyildizi H. and Coskun A., Performance of lightweight concrete with silica fume after high temperature, Construction and Building Materials, 2007 8. Sena J. and Barros J., Modeling of bond between near-surface mounted CFRP laminate strips and concrete, Computers and Structures, 2004. Vol. 82: pp. 1513–1521 9. Tang W., Lo T., and Balendran R., Bond performance of polystyrene aggregate concrete (PAC) reinforced with glass-fibre-reinforced polymer (GFRP) bars, Building and Environment, 2008. Vol. 43: pp. 98-107 10. De Lorenzis L., Rizzo A., and La Tegola A., A modified pull-out test for bond of nearsurface mounted FRP rods in concrete, Composites: Part B, 2002. Vol. 33: p. 589–603. 11. Chu X. and Neitzert T., Experimental and numerical modelling of interfacial behaviour between galvanised steel and aerated concrete, International Journal of Modelling, Identification and Control (IJMIC), 2007: pp. 208-218

46-index of contributions.qxd

2009-09-22

21:52

Page 449

INDEX OF CONTRIBUTORS Hiroshi AKITA, Japan 439 Dariusz ALTERMAN, Poland 439 Salah ALTOUBAT, UAE 53 Yazdanbakhsh ARDAVAN, USA 53 Alexander ASSMANN, Germany 291 Nemkumar BANTHIA, Canada 1 Samer BARAKAT, UAE 53 Joseph J. BIERNACKI, USA 379 J. BLOM, Belgium 111 Ewa BàAZIK-BOROWA, Poland 429 Andrew J. BOYD, Canada 1 Till BÜTTNER, Germany 101 Bernardino CHIAIA, Italy 121 A. CORRADI, Italy 215 Grzegorz CYGAN, Poland 359 Lech CZARNECKI, Poland 183 J. DEGRIECK, Belgium 111 Omkar DEO, USA 277 Yogini DESHPANDE, USA 301 Frédéric DUPRAT, France 11 Alessandro P. FANTILLI, Italy 121 Hans-Bertram FISCHER, Germany 195 Patrick FONTANA, Germany 391 Varvara P. GAVRILIUK, Ukraine 219 Khosrow GHAVAMI, Brazil 71 MICHAà A. GLINICKI, Poland 227 Jacek GOàASZEWSKI, Poland 359 Izabela HAGER, Poland 63 Rashid HAMEED, France 11 Huan HE, The Netherlands 315, 399 Jacob E. HILLER, USA 301 Pavel HUTAě, Czech Republic 419 Chang-Il JANG, Republic of Korea 161 Daria JÓħWIAK-NIEDħWIEDZKA, Poland 227

D. KAKOGIANNIS, Belgium 111 E. KAMSEU, Italy, Cameroon 215, 217 Jacek KATZER, Poland 139 Jadviga K. KERIENE, Lithuania 195 ZbynČk KERŠNER, Czech Republic 419 Wan-Young KIM, Republic of Korea 161 Agnieszka KLEMM, UK 323 Piotr KLEMM, Poland 323 Janusz KONKOL, Poland 409 Sang-Woo LEE, Republic of Korea 161 Christian LEHMANN, Germany 391 C. LEONELLI, Italy 215, 217 Dominik LOGOē, Poland 369 Tatiana V. LYASHENKO, Ukraine 219 Beata àAħNIEWSKA-PIEKARCZYK, Poland 265 Maria MARKS, Poland 227 U.C. MELO, Cameroon 215 Hirozo MIHASHI, Japan 121 Sidney MINDESS, Canada 1 Fausto MINELLI, Italy 23 Barzin MOBASHER, USA 81 Bernd MÖSER, Germany 195 Urs MÜLLER, Germany 391 Tommy NANTUNG, USA 149 Narayanan NEITHALATH, USA 169, 275 Thomas NEITZERT, New Zealand 439 Richard M. NEWELL, USA 149 Jan OLEK, USA 149, 347 Jeanette ORLOWSKY, Germany 101 S. PALANIVELU, Belgium 111 Sulapha PEETHAMPARAN, USA 169 G.C. PELLACANI, Italy 215 D. PERERA, Australia 217 Grigory N. PERVUSHIN, Russia 195 Waldemar PICHÓR, Poland 245 Giovanni A. PLIZZARI, Italy 23 Tomasz PONIKIEWSKI, Poland 131 Grzegorz PROKOPSKI, Poland 409

46-index of contributions.qxd

2009-09-22

21:52

Prashant V. RAM, USA 347 Przemysáaw RANACHOWSKI, Poland 201 Zbigniew RANACHOWSKI, Poland 201 Michael RAUPACH, Germany 101 Deepak RAVIKUMAR, USA 169 Feliks REJMUND, Poland 201 Klaus-Alexander RIEDER, Germany 53 A. RIZZUTI, Italy 217 Aleksandra RADLINSKA, USA 335 Hans W. REINHARDT, Germany 291 Conrado S. RODRIGUES, Brazil 71 Adam RUDY, USA 149

Page 450

Piet STROEVEN, The Netherlands 71, 255, 313, 399 Milani S. SUMANASOORIYA, USA 277 Jacek ĝLIWIēSKI, Poland 33 Agnieszka ĝLOSARCZYK, Poland 237 Romildo D. TOLEDO FILHO, Brazil 81 Tomasz TRACZ, Poland 63 Anaclet TURATSINZE, France11 Paolo VALLINI, Italy 121 J. Van ACKEREN, Belgium 111 D. Van HEMELRIJCK, Belgium 111 W. Van PAEPEGEM, Belgium 111 J. VANTOMME, Belgium 111 Václav VESELÝ, Czech Republic 419 Juan VILCHES, New Zealand 439 Vitaly A. VOZNESENSKY, Ukraine 219

Poologanathan SANJEEVAN, UK 323 Erik SCHLANGEN, The Netherlands 43 Stanislav SEITL, Czech Republic 419 Alain SELLIER, France 11 Cory J. SHORKEY, USA 301 Mercedes G. SIERRA BELTRAN, The Netherlands 43 Flavio de Andrade SILVA, Brazil 81 Surinder P. SINGH, India 91 Jan M. SKOWROēSKI, Poland 237 Marta SàOWIK, Poland 429 Martijn STROEVEN, The Netherlands 313

J. WASTIELS, Belgium 111 Jason WEISS, USA 335 Jong-Pil WON, Republic of Korea 161 Piotr WOYCIECHOWSKI, Poland 183 Ggrigory I. YAKOVLEV, Russia 195 Tomasz ZDEB, Poland 33

*

*

*

47_subject of index.qxd

9/21/2009

8:12 PM

Page 451

Subject Index The numbers refer to the opening pages of the relevant papers acoustic emission 201 aerated concrete 439 air-volume 265 aggregate size 429 aluminous porcelain 201 anti-foaming admixture (AFA) 265 AR glass fibres 101 autoclaved aerated concrete 219 beams 53 bond 101 bonding 43 capillary suction 291 carbon nanostructures 195 carbon spheres 237 carbonation 183 cellulose-cement composites 71 cement 397 cement composite 139; 245 cement matrix 43 cement mixes 369 cement-carbon composites 237 cementitious composites 81 cementitious materials 323 cenospheres 245 chemical treatment 237 chloride ion migration 227 circulated fluidized bed 227 combustion fly ash (CFBC fly ash) CMOD 11 compromise 219 computational materials science 219 concrete 3; 183; 255; 291; 313 399; 409; 419 concrete durability 227 concrete repair 347 concrete structures 429 confining pressure 121 corundum material 201 cracking 23 crystallization 215 curing conditions 33 cyclic voltammetry 237 damage mechanisms 111 deflection 11 digital image data 275 digitization biases 313 durability 161; 347 early-age cracking 335 electrical resistance 369 Enhanced Porosity Concrete 275 (EPC) energy profile 111 fatigue 81 fibres 245; 369; 399 fiber reinforced cement 245 composite Fibre-Reinforced Concrete 23; 91 fibre volume ratio 33 fine aggregate 139 flexural toughness 91 fluidal fly ash 183 fly ash 149; 161; 169 fractal dimension 409 fracture energy 33 fracture mechanics 429 fracture process zone 419 fracture surface development 409 factor fracture surface 215; 409 fracture toughness index 33

fracture toughness 121 gas permeability 63 geopolymer 217 ground granulated blast 149; 169 furnace slag (GGBFS) Glass Fibre Reinforced Plastics 1 (GFRP) heat conductivity 219 high performance concretes 63; 121 high temperature 63 hybridisation 11 hydrothermal curing 391 image analysis 255 interfacial adhesion 71 interfacial transition zone 245 intrusion porosimetry 255 laser cleaning 323 lightweight 245 low-velocity impact 111 machine learning 227 macro synthetic fibers 53 mechanical behaviour 71 mechanical parameters 409 mechanical properties 237; 391 metallic fibers 11 micro-mechanics 377 micro structure 215; 391 mineral binding 195 modification 195 modulus of rupture 245 moment capacity 11 morphology 217 non-contact dilatometer 217 opening operator 255 optimisation 149; 219 oxygen permeability 291 patches 399 performance-related 335 specifications permeability 139 pervious concrete 275 phase composition 391 planar images 275 pores 399 pore size distribution 255 porosity 169 PP fibres 63 probability 335 profile line development factor 409 pull-out tests 43; 439 quantitative image analysis 313 rapid-setting materials 347 reactive powder concrete 33 recycled aggregate 301 refractory binders 217 reinforced fibrous concrete beam 11 restrained ring test 335 retrofitting RC beams 1 rheology 131; 359; 369 rheological properties 265 ring test 301 roller compacted concrete 161 self-compacting concrete 121;131 265; 359 semi empirical formulation 71 shear behavior 53 shear strength 53 shearing force 439 shrinkage 161 shrinkage cracking 301 silica fume 139

sintering 217 sisal fiber 81 sodium hydroxide (NaOH) 169 spalling 63 sprayed FRP 1 standard test methods 23 statistical methods 149 steel fibres 33; 131; 139 steel strips 439 strain measurements 379 strength 33; 169 structural degradation 201 superabsorbent polymers 291 superplasticizer 265 surface modification 323 sweeping test line 313 synergetics 195 temperature 359 ternary mixtures 149 textile reinforced cements 101; 111 thermal conductivity 245 thermal cycle 215 toughness 23; 219 toughness index 245 three-dimensional structure 275 triaxial tests 121 T-stress 419 ultrafine particles 195 ultra high performance 33; 389 cementitious composites variability 335 volcanic ash 215 volumetric stability 301 unhydrated cement 399 water permeability 291 wedge splitting test 419 width of fracture process zone 429 wood fibre 43 workability 131 X-ray diffraction 379