Computational Methods and Experiments in Materials Characterisation III

Computational Methods and Experiments in Materials Characterisation III WIT Press publishes leading books in Science ...

Author: A. A. Mammoli | C. A. Brebbia | A. A. Mammoli | C. A. Brebbia

73 downloads 2147 Views 16MB 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

Computational Methods and Experiments in

Materials Characterisation III

WIT Press publishes leading books in Science and Technology. Visit our website for new and current list of titles. www.witpress.com

WITeLibrary Home of the Transactions of the Wessex Institute. Papers presented at Materials Characterisation III are archived in the WIT eLibrary in volume 57 of WIT Transactions on Engineering Sciences (ISSN 1743-3533). The WIT eLibrary provides the international scientific community with immediate and permanent access to individual papers presented at WIT conferences. http://library.witpress.com

THIRD INTERNATIONAL CONFERENCE ON COMPUTATIONAL METHODS AND EXPERIMENTS IN MATERIALS CHARACTERISATION MATERIALS CHARACTERISATION 2007 CONFERENCE CHAIRMEN A.A. Mammoli University of New Mexico, USA C.A. Brebbia Wessex Institute of Technology, UK

INTERNATIONAL SCIENTIFIC ADVISORY COMMITTEE A. Benavent-Climent D. Bernard S. Bordere M. Bush

S. Hernandez J. W. Leggoe P. Prochazka P. Viot

Organised by Wessex Institute of Technology, UK and University of New Mexico, USA

Sponsored by WIT Transactions on Engineering Sciences

WIT Transactions on Engineering Sciences Transactions Editor Carlos Brebbia Wessex Institute of Technology Ashurst Lodge, Ashurst Southampton SO40 7AA, UK Email: [email protected]

Editorial Board B. Abersek University of Maribor Slovenia K S Al Jabri Sultan Qaboos University Oman J A C Ambrosio IDMEC Portugal H Azegami Toyohashi University of Technology Japan G Belingardi Politecnico di Torino Italy S K Bhattacharyya Indian Institute of Technology India A R Bretones University of Granada Spain J Byrne University of Portsmouth UK D J Cartwright Bucknell University USA A Chakrabarti Indian Institute of Science India J J Connor Massachusetts Institute of Technology USA L Debnath University of Texas-Pan American USA S del Giudice University of Udine Italy

B Alzahabi Kettering University USA A G Atkins University of Reading UK A F M Azevedo University of Porto Portugal R Belmans Katholieke Universiteit Leuven Belgium E Blums Latvian Academy of Sciences Latvia F-G Buchholz Universitat Gesanthochschule Paderborn Germany W Cantwell Liverpool University UK S K Chakrabarti Offshore Structure Analysis USA H Choi Kangnung National University Korea L De Biase University of Milan Italy R de Borst Delft University of Technology Netherlands G De Mey Ghent State University Belgium M Domaszewski Universite de Technologie de Belfort-Montbeliard France

I Doltsinis University of Stuttgart Germany J Dominguez University of Seville Spain J P du Plessis University of Stellenbosch South Africa M E M El-Sayed Kettering University USA M Faghri University of Rhode Island USA C J Gantes National Technical University of Athens Greece R Gomez Martin University of Granada Spain R H J Grimshaw Loughborough University UK R Grundmann Technische Universitat Dresden Germany J M Hale University of Newcastle UK L Haydock Newage International Limited UK C Herman John Hopkins University USA M Y Hussaini Florida State University USA D B Ingham The University of Leeds UK Y Jaluria Rutgers University USA D R H Jones University of Cambridge UK S Kim University of Wisconsin-Madison USA A S Kobayashi University of Washington USA S Kotake University of Tokyo Japan

W Dover University College London UK K M Elawadly Alexandria University Egypt F Erdogan Lehigh University USA H J S Fernando Arizona State University USA E E Gdoutos Democritus University of Thrace Greece D Goulias University of Maryland USA D Gross Technische Hochschule Darmstadt Germany R C Gupta National University of Singapore, Singapore K Hameyer Katholieke Universiteit Leuven Belgium P J Heggs UMIST UK D A Hills University of Oxford UK T H Hyde University of Nottingham UK N Ishikawa National Defence Academy Japan N Jones The University of Liverpool UK T Katayama Doshisha University Japan E Kita Nagoya University Japan A Konrad University of Toronto Canada T Krauthammer Penn State University USA F Lattarulo Politecnico di Bari Italy

Y-W Mai M Langseth University of Sydney Norwegian University of Science and Technology Australia Norway B N Mandal S Lomov Indian Statistical Institute Katholieke Universiteit Leuven India Belgium T Matsui G Manara Nagoya University University of Pisa Japan Italy R A W Mines H A Mang The University of Liverpool Technische Universitat Wien UK Austria T Miyoshi A C Mendes Kobe University Univ. de Beira Interior Japan Portugal T B Moodie A Miyamoto University of Alberta Yamaguchi University Canada Japan D Necsulescu G Molinari University of Ottawa University of Genoa Canada Italy H Nisitani D B Murray Kyushu Sangyo University Trinity College Dublin Japan Ireland P O’Donoghue S-I Nishida University College Dublin Saga University Ireland Japan K Onishi B Notaros Ibaraki University University of Massachusetts Japan USA E Outa M Ohkusu Waseda University Kyushu University Japan Japan W Perrie P H Oosthuizen Bedford Institute of Oceanography Queens University Canada Canada D Poljak G Pelosi University of Split University of Florence Croatia Italy H Power H Pina University of Nottingham Instituto Superior Tecnico UK Portugal I S Putra L P Pook Institute of Technology Bandung University College London UK Indonesia D Prandle M Rahman Proudman Oceanographic Laboratory Dalhousie University UK Canada F Rachidi T Rang EMC Group Tallinn Technical University Switzerland Estonia K R Rajagopal B Ribas Texas A & M University Spanish National Centre for Environmental Health USA Spain D N Riahi W Roetzel University of Illinios-Urbana Universitaet der Bundeswehr Hamburg USA Germany

K Richter Graz University of Technology Austria V Roje University of Split Croatia H Ryssel Fraunhofer Institut Integrierte Schaltungen Germany A Savini Universita de Pavia Italy B Scholtes Universitaet of Kassel Germany G C Sih Lehigh University USA P Skerget University of Maribor Slovenia A C M Sousa University of New Brunswick Canada C-L Tan Carleton University Canada A Terranova Politecnico di Milano Italy S Tkachenko Otto-von-Guericke-University Germany E Van den Bulck Katholieke Universiteit Leuven Belgium R Verhoeven Ghent University Belgium B Weiss University of Vienna Austria T X Yu Hong Kong University of Science & Technology Hong Kong M Zamir The University of Western Ontario Canada

S Russenchuck Magnet Group Switzerland B Sarler Nova Gorica Polytechnic Slovenia R Schmidt RWTH Aachen Germany A P S Selvadurai McGill University Canada L C Simoes University of Coimbra Portugal J Sladek Slovak Academy of Sciences Slovakia D B Spalding CHAM UK G E Swaters University of Alberta Canada J Szmyd University of Mining and Metallurgy Poland S Tanimura Aichi University of Technology Japan A G Tijhuis Technische Universiteit Eindhoven Netherlands I Tsukrov University of New Hampshire USA P Vas University of Aberdeen UK S Walker Imperial College UK S Yanniotis Agricultural University of Athens Greece K Zakrzewski Politechnika Lodzka Poland

Computational Methods and Experiments in

Materials Characterisation III Editors A.A. Mammoli University of New Mexico, USA C.A. Brebbia Wessex Institute of Technology, UK

Editors: A.A. Mammoli University of New Mexico, USA C.A. Brebbia Wessex Institute of Technology, UK Published by WIT Press Ashurst Lodge, Ashurst, Southampton, SO40 7AA, UK Tel: 44 (0) 238 029 3223; Fax: 44 (0) 238 029 2853 E-Mail: [email protected] http://www.witpress.com For USA, Canada and Mexico Computational Mechanics Inc 25 Bridge Street, Billerica, MA 01821, USA Tel: 978 667 5841; Fax: 978 667 7582 E-Mail: [email protected] http://www.witpress.com British Library Cataloguing-in-Publication Data A Catalogue record for this book is available from the British Library

ISBN: 978-1-84564-080-4 ISSN: 1746-4471 (print) ISSN: 1743-3533 (on-line) The texts of the papers in this volume were set individually by the authors or under their supervision. Only minor corrections to the text may have been carried out by the publisher. No responsibility is assumed by the Publisher, the Editors and Authors 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 methods, products, instructions or ideas contained in the material herein. © WIT Press 2007 Printed in Great Britain by Athenaeum Press Ltd. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the Publisher.

Preface

In an age of dwindling resources, knowledge of the behavior of materials takes on an even more important role than was traditionally the case. Not only must a material perform its basic function, but it must do so while satisfying constraints given by ecology, economy, safety and durability. Alongside the science of traditional materials, new areas are emerging. At the very small scale, materials are being engineered down to their very microstructure, sometimes even their molecular structure. These microengineered materials promise exceptional performance, however it becomes increasingly difficult to characterize their structure and behavior with traditional methods. In many cases, characterization occurs by indirect means, requiring a computer model to interpret the measurement data to finally recover the material properties sought, for example in the case of nanoindentation of heterogeneous materials. In some cases, it is even difficult to define a property, or at what scale it applies. The second recent trend in materials science is the re-emergence of traditional and natural materials, sometimes in combination with more ‘conventional’ ones, as in the case of natural fibre reinforced composites. These pose particular challenges, as their microstructure and properties can be even more complex than in synthetic materials. The characterization of materials is an extremely broad topic, which could mean different things to different people. We have, nevertheless, endeavoured to structure the book in a logical manner. It comprises three broad areas: papers focusing on the materials and their microstructures, papers focusing on experimental characterization techniques, and papers focusing on computational methods. As in the previous two conferences, we are confident that cross-pollination of ideas and methodologies will occur, leading to new collaboration and new research paths. As always, the editors wish to thank the authors for contributing their work, and the scientific advisory committee in particular, for their help with obtaining and selecting many quality articles. The Editors Bologna, Italy 2007

This page intentionally left blank

Contents Section 1: Microstructures – novel composite materials Study on static and creep properties of CFRP using rubber modified matrix K. Takemura & Y. Yasuda.....................................................................................3 Determination of the fatigue behavior of coatings by means of an improved impact testing evaluation procedure C. David, K. G. Anthymidis, P. Agrianidis & D. N. Tsipas ................................13 Effect of fly ash reinforcement on the corrosion behaviour of cast Al-Mg alloy A535 in 3.5wt% NaCl solution E. R. Obi, I. N. A. Oguocha & R. W. Evitts.........................................................21 Testing of palm fibre as reinforcement material in polyester composites V. V. S. Prasad, D. N. Rao, K. N. S. Suman & N. R. M. R. Bhargava ................31 Section 2: Microstructures – ceramics and advanced materials Experimental study on fracture behaviour of polycrystalline ceramics under shock loading J. T. Zhou & G. W. Yao .......................................................................................43 Blocking and self-locking of superdislocations in intermetallics B. A. Greenberg & M. A. Ivanov.........................................................................51 The properties and performance of polymer fibre reinforced bituminous mixtures I. Kamaruddin & M. Napiah ...............................................................................61 Hardness determination of EBiD-layers containing tungsten and cobalt T. Wich, T. Luttermann & I. Mircea ...................................................................73

Section 3: Microstructures – alloys Thermodynamic modelling of a 6w/o Al P/M processed Ni base superalloy D. A. Akinlade, W. F. Caley, N. L. Richards & M. C. Chaturvedi......................85 An investigation into martensitic transformation in hot stamping process M. Naderi & W. Bleck .........................................................................................95 Quantitative assessment of strain and heat treatment on twin formation in commercially pure nickel Q. Li, J. R. Cahoon & N. L. Richards ...............................................................105 Three-dimensional crystallographic characterization and mechanical modeling of a commercial stainless steel A. C. Lewis, D. J. Rowenhorst, G. Spanos & A. B. Geltmacher .......................115 Section 4: Microstructures – cements and cement based materials Reactive powder concrete: material for the 21st century D. Mestrovic, D. Cizmar & V. Stanilovic..........................................................127 Impedance spectroscopy as a tool to study modifications in the microstructure of concrete in ionic migration experiments G. de Vera, M. A. Climent & I. Sánchez ...........................................................135 Section 5: Experimental methods – imaging and analysis Laser speckle measurements and numerical simulations of the deformation of masonry loaded in compression A. T. Vermeltfoort..............................................................................................147 Quantitative analysis of polyurethane nanocomposites with boehmite structures modified using lactic acid J. Ryszkowska....................................................................................................159 The spatial controlling of Lamb waves excited by a point source on the cylindrical wall V. Sukackas .......................................................................................................169 3D strain mapping inside materials by microstructural tracking in tomographic volumes H. Toda, M. Kobayashi, K. Uesugi, D. S. Wilkinson & T. Kobayashi..............177

Fractal and spectral analysis of fracture surfaces of elastomeric materials D. Ait Aouit & A. Ouahabi ................................................................................187 Multi-scale foam behaviour characterisation P. Viot & D. Bernard ........................................................................................197 Section 6: Experimental methods – thermal analysis Thermo-analytical evaluation of wear debris for thermoplastic and sintered polyimide P. Samyn, I. Van Driessche, G. Schoukens & P. De Baets ...............................209 Analysis of adiabatic heating in high strain rate torsion tests by an iterative method: application to an ultrahigh carbon steel J. Castellanos, I. Rieiro, M. Carsí, J. Muñoz & O. A. Ruano ...........................219 Section 7: Experimental methods – mechanical characterisation and testing Collapse of FRP/syntactic foam sandwich panels M. Perfumo, C. M. Rizzo & M. P. Salio............................................................231 Modelling of viscoelastic properties of a curing adhesive J. de Vreugd, K. M. B. Jansen, L. J. Ernst & J. A. C. M. Pijnenburg...............241 Flexural bond strength of clay brick masonry C. G. Yuen & S. L. Lissel ..................................................................................253 Structural, economic and material comparison of various steel grades under dynamic/fatigue loading I. U. Amobi & H. C. Uzoegbo ...........................................................................263 Mechanical compression tests to model timber structures behaviour V. De Luca & D. Sabia......................................................................................273 Section 8: Experimental methods – new methods Millimeter wave spectroscopy and materials characterization of refractive liquid crystal polymer/titania composites B. R. Dantal, A. Saigal, M. A. Zimmerman, K. A. Korolev, M. N. Afsar & U. A. Khan .................................................................................281

Assessment of surface roughness for the analysis of the water vapour condensation process A. J. Klemm, P. Klemm & I. Ibrahim ................................................................291 Use of impedance spectroscopy to determine the displacement of water in cement paste under small loads I. Sánchez, G. Castro, M. A. Climent & X. R. Nóvoa........................................301 Assimilation of porosity in modern bricks by computational means M. A. Stefanidou................................................................................................313 Dynamic tensile test and specimen design of auto-body steel sheet at the intermediate strain rate S. B. Kim, J. H. Song, H. Huh & J. H. Lim .......................................................319 Utilization of ground coloured glass cullet in construction materials A. Karamberi & A. Moutsatsou ........................................................................329 In situ dynamic characterization of soils by means of measurement uncertainties and random variability G. Vessia & C. Cherubini .................................................................................339 A natural and biodegradable scaffold of electrospun eggshell membrane W. D. Kim, T. Min, S. A. Park, J. H. Park & G. H. Kim ...................................349

Section 9: Computational methods – discrete computational methods Characterization of cementitious materials by advanced concurrent algorithm-based computer simulation systems Z. Q. Guo, M. Stroeven, W. Yang, H. He & P. Stroeven...................................361 A simulation of the behaviour of propane bulks on a grid platform A. Laganà & A. Costantini................................................................................373 Section 10: Computational methods – damage mechanics Failure characterisation of Ti6Al4V gas turbine compressor blades A. Kermanpur, H. Sepehri Amin, S. Ziaei Rad, N. Nourbakhshnia & M. Mosaddeghfar............................................................383 Seismic damage assessment of steel components A. Benavent-Climent .........................................................................................393

A visco-plastic damage model for high temperature creep of single-crystal superalloys A. Staroselsky & B. Cassenti.............................................................................403 Failure mechanics of slope slip with predestinate slip plane J. Vacek & S. Sedláþková..................................................................................413 Section 11: Computational methods – innovative techniques Back analysis of reinforced soil slopes P. Procházka & J. Trckova ...............................................................................423 Towards 3D simulation of sintering processes S. Bordère, D. Bernard, S. Vincent & J.-P. Caltagirone ..................................433 Author Index ...................................................................................................443


Section 1 Microstructures – novel composite materials


Computational Methods and Experiments in Materials Characterisation III

3

Study on static and creep properties of CFRP using rubber modified matrix K. Takemura1 & Y. Yasuda2 1 2

Department of Mechanical Engineering, Kanagawa University, Japan Graduate Student of Kanagawa University, Japan

Abstract In this study, the static and creep properties of Carbon Fiber Reinforced Plastics (CFRP) are examined. Plain woven carbon fabric is used as reinforcement. As the matrix, epoxy resin is modified by using cross-linked rubber particles. Four weight contents (0%, 5%, 10%, and 15%) of rubber modification are used. Three point bending loading is applied to the specimen. Static and creep tests are conducted. The results can be summarized as follows. For epoxy resin bulk and CFRP specimens, the strength and maximum strain decrease by rubber modification at static bending test, but the reduction rates of the strength and maximum strain for CFRP are smaller than those of resin bulk specimens. For example, when the weight content of rubber particles for epoxy resin is 5%, the strength reduces to about 50% and the maximum strain reduces to about 60% in the resin bulk specimen, but the strength and maximum strain reduce to about 35% and 25% respectively in CFRP. For the creep test, the creep strain rate in the secondary state is improved for CFRP with rubber modification. When the weight content of rubber particles is big, the improvement of the creep strain rate in the secondary state is great. For example, for CFRP whose weight content of rubber particles is 15%, the creep strain rate decreases by 25%. When an environmental temperature is beyond 120 degrees centigrade, the creep strain rate at the secondary state increases rapidly for unmodified CFRP, but the creep strain rate for modified CFRP is not so increased. So, an effect of rubber modification is great in the high temperature environment. In the case when the environmental temperature is 190 degrees centigrade, a 5% modification is most effective. Consequently, the rubber modification method for CFRP is effective for creep strain in an elevated temperature environment. Keywords: CFRP, epoxy, rubber modification, creep, bending loading. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070011

4 Computational Methods and Experiments in Materials Characterisation III

1

Introduction

CFRP has high specific strength and modulus compared to conventional metal materials. So, CFRP is widely used in automobile and airplane parts. When CFRP is used as the structural materials of airplane wings, constant load is supplied for several hours. So, the creep property for CFRP is important. Static and creep properties are dependent on environmental temperature because the polymer matrix is used. Therefore, the mechanical properties at elevated temperature are important too. In the case that epoxy matrix of CFRP is modified with cross-linked rubber particles, the static tensile strength and fatigue lives of CFRP have increased [1, 2]. But, as far as the authors know, few papers have been published about the effect of rubber particle to creep properties for CFRP. The objective of this present work is to demonstrate the static and creep properties for CFRP with a rubber modified matrix. The effect of environmental temperature on the properties is also examined.

2

Specimens

Plain woven carbon fabric (Toray Co.) is used as reinforcement. The number of laminates is eight. Epoxy resin is used as the matrix. The matrix is modified by using cross-linked rubber particles. Four weight contents (0%, 5%, 10%, and 15%) of rubber particle are used. Specimens are laminated by the hand lay up method, and they are cured by a hot press facility. The pressure at moulding is about 10MPa. The length, breadth and thickness of the specimens are 100mm, 15mm and 2mm respectively.

3

Experiment

3.1 Static bending test Three point bending tests are conducted with Shimadzu universal testing instruments (AG-IS). Crosshead speeds are 2.0 mm/min for resin bulk and 5.0 mm/min for CFRP specimens. An extensometer (MTS Co.) is used to measure a strain. The number of specimens is five at one test condition based on JIS (Japanese Industrial Standard) 7171 and 7074. 3.2 Creep test Constant temperature oven facility (Advantec Co.) is used for the creep test. Three point bending loading (20N) is applied. The deflection is measured and recorded with a remote scanner (NEC Co.). The creep test continues until failure or near 150 hours. Environmental temperatures are 110, 120, 130 and 190 degrees centigrade.

WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


5

3.3 Observation of flat wise surface In the case of creep test, flat wise surfaces are observed with a scanning electron microscope (SEM-EDX: Hitachi Co.) 㪈㪉㪇

㪪㫋㫉㪼㫊㫊㩿㪤㪧㪸㪀

㪇㩼

㪌㩼

㪐㪇㪍㪇㪊㪇㪇

㪇

㪈

㪉

㪊

㪋

㪪㫋㫉㪸㫀㫅㩿䋦㪀

Figure 1:

Stress–strain curves for epoxy resin bulk.

㪏㪇㪇㪇㩼㪈㪇㩼

㪌㩼㪈㪌㩼

㪪㫋㫉㪼㫊㫊䋨㪤㪧㪸㪀

㪍㪇㪇

㪋㪇㪇

㪉㪇㪇

㪇㪇

Figure 2:

4

㪇㪅㪌

㪈㪪㫋㫉㪸㫀㫅䋨䋦䋩

㪈㪅㪌

㪉

Stress–strain curves for CFRP.

Results and discussion

4.1 Static bending test Figure 1 shows stress-strain curves for a resin bulk specimen which has no reinforcement. Figure 2 shows stress-strain curves for CFRP. Table 1 and table 2 show the mechanical properties for resin bulk and CFRP respectively. From these results, it is understood that the strength (maximum stress) and maximum strain decrease by rubber modification for resin bulk and CFRP. It is because of WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

6 Computational Methods and Experiments in Materials Characterisation III Young’s modulus that the decrease is due to the rubber modification to the matrix. The reduction rates for CFRP are smaller than that of resin bulk specimen. When the weight content of rubber particles for the epoxy resin bulk is 5%, the strength reduces by about 50% and the maximum strain reduces by about 60%. But the strength reduces by about 35% and the maximum strain reduces by about 25% for CFRP. So, the reduction of mechanical properties for CFRP is smaller than that of resin bulk. Table 1:

Mechanical property for epoxy resin bulk at static bending test.

㪤㪸㫏㫀㫄㫌㫄㩷㫊㫋㫉㪼㫊㫊㩷䋨㪤㪧㪸䋩㩷

㪤㪸㫏㫀㫄㫌㫄㩷㪪㫋㫉㪸㫀㫅㩷䋨䋦䋩㩷

㪰㫆㫌㫅㪾㵭㫊㩷㫄㫆㪻㫌㫃㫌㫊㩷䋨㪞㪧㪸䋩㩷

㪇䋦㩷

㪈㪈㪍㩷

㪊㪅㪍㪉㩷

㪊㪅㪏㩷㩷

㪌䋦㩷

㪌㪇㩷

㪈㪅㪌㪌㩷

㪊㪅㪌㩷

㩷㩷

Table 2:

Mechanical property for CFRP at static bending test.

㪤㪸㫏㫀㫄㫌㫄㩷㫊㫋㫉㪼㫊㫊㩷䋨㪤㪧㪸䋩㩷

㪤㪸㫏㫀㫄㫌㫄㩷㪪㫋㫉㪸㫀㫅㩷䋨䋦䋩㩷

㪰㫆㫌㫅㪾㵭㫊㩷㫄㫆㪻㫌㫃㫌㫊㩷䋨㪞㪧㪸䋩㩷

㪇䋦㩷

㪎㪎㪏䋮㪋㩷

㪈䋮㪍㪏㩷㩷

㪋㪎䋮㪇㩷㩷

㪌䋦㩷

㪌㪈㪋䋮㪐㩷

㪈䋮㪉㪈㩷

㪋㪍䋮㪍㩷㩷

㪈㪇䋦㩷

㪋㪋㪇䋮㪍㩷

㪇䋮㪐㪍㩷㩷

㪋㪎䋮㪇㩷㩷

㪈㪌䋦㩷

㪊㪋㪐䋮㪉㩷

㪈䋮㪇㪇㩷㩷

㪊㪋䋮㪊㩷㩷

㩷㩷

4.2 Creep test Figures 2–5 show the relationships between creep strain and time in 110, 120, 130 and 190 degrees centigrade respectively. Creep curve can be divided into three stages. The first stage is the transient creep region which includes the elastic strain region. The second one is constant creep region which is called secondary creep, and the last one is the tertiary creep region. So, creep strain H at first and second stages can be written as follows.

H

H 0 Et n kt

H 0 : elasticity strain t : time Et n : strain of transient creep region kt : strain of secondary creep region WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


7

Figure 7 shows coefficients k of creep curves at each temperature environment. The coefficient k decreases as the rubber contents increase. From fig.3 and 7, when the temperature is 110 degrees centigrade, the strain of the secondary creep region is small. From the viewpoint of static property, Young’s modulus decreases with the increase of rubber content. Therefore, it is thought that the creep strain at 110 degree centigrade is dependent on the static property especially Young’s modulus. 㪇㪅㪋

㪪㫋㫉㪸㫀㫅㩿䋦䋩

㪇㪅㪊

㪇㪅㪉㪇㩼㪈㪇㩼

㪇㪅㪈

㪌㩼㪈㪌㩼

㪇㪇

Figure 3:

㪌㪇

㪫㫀㫄㪼㩿㪿㪀

㪈㪇㪇

㪈㪌㪇

Relationship between strain and time (110°C).

㪇㪅㪋

㪪㫋㫉㪸㫀㫅㩿䋦䋩

㪇㪅㪊㪇㪅㪉㪇㩼㪈㪇㩼

㪇㪅㪈

㪌㩼㪈㪌㩼

㪇㪇

Figure 4:

㪌㪇


㪈㪇㪇

㪈㪌㪇


In the case that the environment temperature is beyond 120 degrees centigrade, secondary creep strain rate becomes big which are seen in figs 4 and 5. From Fig.7, the effect of rubber modification to the creep strain is great in the WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

8 Computational Methods and Experiments in Materials Characterisation III high temperature environment. Epoxy resin has glassy property under some temperature which is thought 125 degrees centigrade [3]. But, this temperature is affected by a quantity and a kind of hardening agent. Therefore, it is thought that the temperature decrease to 115 degrees centigrade. In addition, it is known that the creep property is improved by crumb rubber modified to bitumen. Therefore, it is thought that the creep property of CFRP is improved with rubber modification. 㪇㪅㪋


㪇㪅㪊㪇㪅㪉㪇㪅㪈

㪇㩼

㪌㩼

㪈㪇㩼

㪈㪌㩼

㪇㪇

Figure 5:

㪌㪇


㪈㪇㪇

㪈㪌㪇


㪈


㪇㪅㪏㪇㪅㪍㪇㪅㪋㪇㪅㪉

㪇㩼㪈㪇㩼

㪌㩼㪈㪌㩼

㪇㪇

㪌㪇

㪈㪇㪇

㪈㪌㪇


Figure 6:


In the case that the environmental temperature is 190 degrees centigrade (Fig.6), unmodified and 15% modified specimen fails rapidly. But, 5% and 10% modified specimens do not fail until 150 hours. So, excess modification is not effective for an extreme high temperature environment. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


9

㪎㪈㪈㪇㷄

㪚㫆㪽㪽㫀㪺㫀㪼㫅㫋㩷㫆㪽㩷㪸㩷㪺㫉㪼㪼㫇㩷㪺㫌㫉㫍㪼㩷㫂㩿㪈㪇㪄㪋㪀

㪍

㪈㪉㪇㷄

㪈㪊㪇㷄

㪌㪋㪊㪉㪈㪇㪇㩼

㪌㩼

㪈㪇㩼

㪈㪌㩼

㪮㪼㫀㪾㪿㫋㩷㪺㫆㫅㫋㪼㫅㫋㩷㫆㪽㩷㫉㫌㪹㪹㪼㫉㩷㫇㪸㫉㫋㫀㪺㫃㪼㫊㩷㩿䋦㪀

Figure 7:

Coefficient of creep curve k .

After the creep test, flat wise surfaces of CFRP are observed with SEM. Figure 8 shows SEM micrograph. The images can be compared with that at 25 degrees centigrade. In the case of 110 degrees centigrade, it looks like the same as that of 25 degrees centigrade. In the case of 120 degrees centigrade, fibers can be seen on the surface. In the case of 190 degrees centigrade, fibers can be seen on the surface clearly. In addition, the weight of specimen in 190 degrees centigrade decreases 1.3 percent. Therefore, it is thought that epoxy resin is removed from the surface when the environmental temperature is 190 degrees centigrade. In the case that the temperature is above 120 degrees centigrade, it is thought the resin may be removed. So, the secondary creep strain rate of CFRP increases rapidly over 120 degrees centigrade.

5

Conclusions

The effect of rubber modification to static and creep properties for CFRP is examined. In the result, following conclusions are obtained. For static test, the mechanical properties decrease due to rubber modification. The reduction rate of CFRP is smaller than that of resin bulk. For creep test at elevated temperature, in the case that the weight content of rubber particles is big, there is an improvement of creep strain rate in the secondary state. Therefore, the rubber modification method has an effect to creep strain rate in secondary region. When the environmental temperature is above 120 degrees centigrade, epoxy resin is removed due to the heat. Therefore, this phenomena affects the acceleration of creep strain rate.



30Ǵm (a)

30Ǵm (b)

30Ǵm (c)

30Ǵm (d)

Figure 8:

(a) 25°C, (b) 110°C, (c) 120°C, (d) 190°C SEM image of CFRP.



11

References [1] [2]

[3] [4]

K. Takemura and T. Fujii, Improvement in Static, Impact and Fatigue Properties of CFRP due to CNBR Modification of Epoxy Matrix, JSME International Journal Series A, Vol.43, No.2, pp.186-195, 2000. M. Higashino, K. Takemura and T. Fujii, Strength and damage accumulation of carbon fabric composites with a cross-linked NBR modified epoxy under static and cyclic loadings, Composite Structures, Vol.32, No.1-4, pp.357-366, 1995 Engineering Materials Handbook, Vol.1 Compoiste, ASM International, pp.66-77, 1987. Sharma, V., Goyal, S., Comparative study of performance of natural fibres crumb rubber modified stone matrix asphalt mixtures. Canadian Journal of Civil Engineering, 33(4), pp134-139, 2006.




13

Determination of the fatigue behavior of coatings by means of an improved impact testing evaluation procedure C. David1, K. G. Anthymidis2, P. Agrianidis1 & D. N. Tsipas3 1

Mechanical Engineering Department, Technical University of Serres, Greece 2 Materials Department in Applied Research Center of Serres, Greece 3 Mechanical Engineering Department, Aristotle University of Thessaloniki, Greece

Abstract Impact testing is an efficient experimental procedure that enables the determination of the fatigue resistance of mono- and multilayer coatings deposited on various substrates, which is not possible with the common testing methods previously available. In this paper an advanced impact tester, capable of assessing the fatigue failure resistance of coatings working under cyclic loading conditions, is presented. The fatigue failure of the tested coating was determined by means of scanning electron and optical microscopy. The test results were recorded in diagrams containing the impact load versus the number of successive impacts that the tested coating can withstand. Keywords: thin films and coatings, materials characterization, fatigue.

1

Introduction

The impact test method has been introduced as a convenient experimental technique to evaluate the fatigue strength of coatings being exposed in alternate impact loads [1–4]. According to this method a coated specimen is cyclically loaded by a hard ball that repetitively impacts on the specimen surface. The superficially developed Hertzian pressure induces a complex stress filled within the coating, as well as, in the interfacial zone. Both these stress states are responsible for distinct failure modes, such as a cohesive or adhesive one. The WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070021

14 Computational Methods and Experiments in Materials Characterisation III exposure of the layered compounds against impulsive stresses creates the real conditions for the appearance of coating fatigue phenomena based upon structural transformation, cracking generation and cracking growth, which are responsible for the gradual microchipping and the degradation of the coating.

2

Experimental procedure:

In this research coatings were characterized using an advanced impact tester system, which is shown in Fig.1. The system consists of three main parts: x The main test device (centre) x The power supply unit (left) x The evaluation and controlling unit (right)

Figure 1:

Impact tester system.

In the present paper characterization of coatings were carried out in such a system. This experimental set up is simple and user friendly and allows the determination of the fatigue behavior of a wide range of single and multielement coatings. The working principle of the impact tester is presented in figure 2 and is based on the alternate Laplace magnetic forces produced by the electromagnetic field, which is induced within the mechanical unit. In order to make the impact tester system more efficient we redesigned the mechanical unit using finite elements to achieve the optimum magnetic flux density, which gives the higher magnetic force of the electromagnetic field and correspondingly an increased impact load (figure 3). Further more the control and the monitoring of the impact tester was improved. All four most important test parameters, the induced impact force, number of successive impacts, the impact frequency and the level of the coil temperature are monitored throughout WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


15

each experiment. The whole mechanical apparatus is flexible and allows the operator to modify the desirable total number of impacts and impact force during the test procedure easily via the front panel of the evaluation and controlling unit (figure 4).

Figure 2:

Figure 3:

Impact tester working principles.

Increased impact load due to design optimisation.



Figure 4:

Front panel of the evaluation and controlling unit.

Figure 5:

Impact crater with the developed coating failure.

The stress strain problem related to the impact test is the Herzian contact, which develops between the spherical indentor (carbide ball) and the examined layered space. Gradual intrinsic coherence release and coating microchipping or abrupt coating fracture and consequent exposure of the substrate material designate the coating failure. In all impact craters resulted from the experiments three different zones inside the impact cavity were identified (figure 5). A central zone in the mid of the impact cavity, where the coating is strained with WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


17

compressive stresses and a gradual cohesive degradation takes place. The intermediate zone inside the piled up rim formed around the impact cavity, where tensile and shear stresses are building up and both cohesive and adhesive delamination arises. Finally, the peripheral zone of the impact cavity, where macrocracks might propagate and coating failure occurs. The coating failure mode and its extent were assessed by SEM observations and EDX analysis. The contact load leading to coating fatigue fracture was recorded in diagrams (endurance strength curves) versus the number of impacts (figure 6). The impact load for which the coating after 106 impacts do not fail is called limit of continues endurance of the coating.

Figure 6:

3

Typical endurance strength curve.


In figure 7 the high cycle fatigue diagram of a Al, Fe pack coating on P92 steel (9% w.t. Cr, 1.8% w.t. W) substrate is shown. This coating consists of an outer Fe14Al84 layer and an inner FeAl13 layer. From impact testing procedure it was concluded that its limit of continues endurance was 100 N (Fig.8, 9). The main failure of the examined coating-substrate compound occurred in the central zone of the impact crater with coating degradation.

4

Conclusions

The work presented here shows a step forward in understanding the failure mechanisms of pack coatings. More specifically the paper reports the results of a novel experimental approach adapted to investigate the endurance performance of coating systems with refer to their mechanical properties and to deliver a semi-empirical design approach. Current impact testing investigations revealed the fatigue strength of Al, Fe pack coating on P92 steel substrate.


18 Computational Methods and Experiments in Materials Characterisation III 700

Impact Force (N)

600 500 400 300 200 100 0 0

500000

1000000

1500000

Number of impacts

Figure 7:

Endurance strength curve of Al, Fe pack coating on P92 steel substrate.

Figure 8:

SEM photo of the Al, Fe coating deposited on P92 steel substrate after 1.000.000 impacts with an impact force of 100 N, failure initiation.



Figure 9:

19

EDX diagram of the Al, Fe coating deposited on P92 steel substrate after 1.000.000 impacts with an impact force of 100 N. Traces of W and Cr indicates failure initiation.

Acknowledgements We express our gratitude to the E.U. for financing this work through the project SUPERCOAT, Contract No: ENK5-CT-2002-00608 and to Technical University of Serres also.

References [1] [2] [3] [4]

Voevodin A.A., Bantle R., Matthews A., Dynamic impact wear of TiCXNY and Ti-DLC composite coatings, Wear, 185 (1995), pp. 151157. Bantle R., Matthews A., Investigation into the impact wear behaviour of ceramic coatings, Surface and Coatings Technology, 74 -75 (1995), pp. 857-868. Heinke W., Leyland A., Matthews A., Berg G., Friedrich C., Broszeit E., Evaluation of PVD nitride coatings, using impact, scratch and Rockwell-C adhesion tests, Thin Solid Films, 270 (1995), pp. 431-438. Ziegele H., Rebholz C., Voevodin A.A., Leyland A., Rohde S. L., Matthews A., Studies of the tribological and mechanical properties of laminated CrC-SiC coatings produced by r.f. and d.c. sputtering, Tribology International, Vol. 30, No. 12 (1997), pp.845-856. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)



21

Effect of fly ash reinforcement on the corrosion behaviour of cast Al-Mg alloy A535 in 3.5 wt% NaCl solution E. R. Obi1, I. N. A. Oguocha1 & R. W. Evitts2 1

Department of Mechanical Engineering, University of Saskatchewan, Canada 2 Department of Chemical Engineering, University of Saskatchewan, Canada

Abstract The effect of fly ash reinforcement on the room temperature corrosion behaviour of cast Al-Mg alloy A535 in 3.5 wt% pH 7 NaCl solution was investigated using an immersion corrosion test, electrochemical tests and optical microscopy. The materials studied were A535 and its metal matrix composites (MMCs) containing 10wt% fly ash, 15wt% fly ash, and a hybrid reinforcement (5wt% fly ash+5wt% SiC). The immersion corrosion test results showed that the corrosion rate of the MMCs increased with increasing fly ash content while the electrochemical test results indicated that their corrosion potential (Ecorr) and critical pitting (breakdown) potential (Ep) decreased with increasing fly ash content. The repassivation potentials of the MMCs were found to be more positive than that of the matrix alloy. The corrosion of the MMCs, which was accompanied by loosening of fly ash particles, was also affected by porosity and the presence of several reaction products. Keywords: Al-Mg alloy, A535, fly ash, MMCs, corrosion rate, corrosion potential, pitting potential, repassivation potential, intermallic compounds, Mg2Si.

1

Introduction

Particle-reinforced aluminum metal matrix composites (MMCs) containing SiC and Al2O3 have received great attention in the past few decades because of their improved wear resistance, reduced coefficient of thermal expansion (CTE), high WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070031

22 Computational Methods and Experiments in Materials Characterisation III elastic modulus, and improved strength compared to unreinforced aluminum alloys [1, 2]. Although they have found potential applications in weight-critical components in automobile, aerospace, and defence systems [2–5], the application base of these particulate MMCs is limited by their high production cost. Recently, inexpensive aluminum alloy MMCs reinforced with fly ash, a waste by-product of coal combustion, has been engineered [6–11] to serve as a substitute for conventional particulate MMCs in several applications in order to widen the application bases of this class of MMCs. The addition of fly ash into aluminum MMCs is a value-added initiative that lowers the disposal cost of fly ash, increases energy savings by reducing the quantity of aluminum produced, and creates a healthier environment. Many potential applications of particulate Al MMCs in naval structures such as ship and boat hulls, offshore structures and desalination plants involve exposure to saline environments with high chloride ion concentrations. Also, particulate Al MMCs used in automobile engine parts usually encounter hostile environments containing chloride, sulphate and nitrate ions as well as exhaust gases like CO2, CO and NOx [12]. Since corrosion resistance is a key design parameter which must be factored in when considering the application potentials of particulate MMCs in structural applications, it is important to understand the corrosion behaviour of these materials in different corrosive environments. The corrosion behaviour of Al-based MMCs reinforced with particles such as Al2O3, garnet, TiC, AlN and SiC particles have been studied by several workers [12–21]. A close look at the results obtained from these studies shows that three types of corrosion can occur in particulate Al MMCs at room temperature. These are galvanic corrosion between the reinforcement and the matrix alloy, crevice corrosion around the reinforcement and in surface pores, and pitting corrosion of the matrix alloy as well as the interface between the matrix and the reinforcement. De Salazar et al [13] investigated the effect of heat treatment and reinforcement volume fraction on the corrosion behaviour of AA6061 and AA7005 reinforced with Al2O3 particles. They found that the pitting corrosion mechanisms of AA6061 MMCs were affected by post-fabrication heat treatment and that the number of corrosion pits increased with increasing Al2O3 volume fraction. Gnecco and Beccaria [14] investigated the corrosion behaviour in sea water of a SiCp/Al-Mg MMC and found that SiC particles acted as cathodic sites with respect to the matrix alloy, which experienced selective aluminium dissolution. They also observed that the MMC suffered localized corrosion of the matrix where Al-Cu intermetallic compounds were present. Also, Gavgali et al [15] studied the effect of reinforcement content on the corrosion behavior of SiCp/Al-Si-Mg MMCs in both aerated and deaerated 3.5wt.% NaCl aqueous solutions. The results showed that the corrosion resistance of the MMCs decreased with increasing SiC particle content. However, Kiourtsidis et al [16] who studied SiCp/AA2024 MMCs reported that the overall performance of the matrix alloy was independent of the volume fraction of SiC particles as they observed no detrimental galvanic attack between the matrix and the particles.



23

Similarly, Aylor and Moran [17] observed that SiC did not alter the corrosion potential of AA6061 in aerated seawater. Although there is a significant amount of research on the corrosion behavior of Al alloys and conventional particulate Al MMCs, there is a dearth of information on the corrosion behaviour of Al alloys reinforced with fly ash particles [9,11]. The present investigation was therefore initiated to contribute to better understanding of the effect of fly ash additions on the corrosion behavior of cast Al-Mg alloy A535 in 3.5wt% NaCl solution. A535 is a non-heat treatable Al-Mg alloy with good combination of strength, machinability, corrosion resistance, weldability and good surface finish. It is used for manufacturing naval vessels, aircraft landing gears, rocket launchers, lightweight armoured vehicles and components of instruments and computing devices.

2 Experimental materials and procedure 2.1 Materials The fly ash reinforced A535 MMCs used in this study were fabricated by the stir casting technique. The MMCs contained 10wt% fly ash (10FA/A535), 15wt% fly ash (15FA/A535) and a hybrid mixture consisting of 5wt% fly ash and 5wt% SiC (5FA5SiC/A535). The composition of the A535 alloy used was 6.17wt% Mg, 0.01wt% Cu, 0.01wt% Si, 0.02wt% Fe, and 0.04wt% Ti, bal. Al while the composition of the raw fly ash used is shown in Table 1. Table 1: Compound Weight %

SiO2 44.8

Weight percent of various oxides in fly ash. Al2O3 22.2

Fe2O3 24.0

MgO 0.9

CaO 1.8

TiO2 0.8

K2O 2.4

Na2O 0.9

SO3 1.4

Balance = oxides of other trace elements. 2.2 Corrosion testing The corrosion behavior of the test materials was evaluated using static immersion test, potentiodynamic and cyclic polarization tests, visual inspection and optical microscopy. The immersion test was conducted at room temperature using conventional weight loss method (ASTM G31) to an accuracy of 0.0001g. Rectangular specimens measuring 10 mm x 10 mm x 5 mm were cut from the A535 and its MMCs, metallurgically polished with emery cloth to high smoothness, cleaned ultrasonically in acetone and methanol, and dried. They were subsequently weighed and immersed in a solution of 3.5wt% NaCl (pH = 7) exposed to the ambient air. The specimens were suspended in the electrolyte using a plastic string and a plastic crocodile clip to avoid crevice and galvanic corrosion. The specimens were removed from the solution at regular intervals and cleaned in accordance with ASTM G1-90 standard, dried and re-weighed. The exposure times used in this study were 1, 3, 5, 7, 10, and 14 days. The surface of each specimen was examined visually and by optical microscopy before and after each exposure test. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

24 Computational Methods and Experiments in Materials Characterisation III Electrochemical polarization measurements were conducted on all specimens of the matrix alloy and composites using a Gamry ECM 8 electrochemical multiplexer with PCI4 potentiostat controlled by a computer. A saturated calomel electrode (SCE) and a graphite electrode were used as reference and auxiliary electrodes, respectively. As in the immersion test, all specimens were metallurgically polished to high smoothness using emery cloth, rinsed in acetone, dried, and immersed in a 3.5wt% pH 7 NaCl electrolyte at room temperature exposed to the atmospheric air. Before starting the measurements, all specimens were allowed to equilibrate for approximately 30 minutes to their corrosion potential (Ecorr). A scan rate of 1 mV/s was used to determine the corrosion potential (Ecorr), pitting potential (Ep) and repassivation potential (Erp). 0.018 A535 5FA5SiC/A535 10FA/A535 15FA/A535

0.016

Corroion Rate (mm/year)

0.014 0.012 0.010 0.008 0.006 0.004 0.002 0.000 0

2

4

6

8

10

12

14

16

Immersion Time (days)

Figure 1:

3

Variation of corrosion rate of A535 and its fly ash reinforced MMCs with time in 3.5wt% NaCl solution (pH = 7).


Fig. 1 shows the variation of corrosion rate with exposure time for specimens immersed in 3.5wt% NaCL (pH = 7) solution at room temperature. It can be seen that the matrix alloy (A535) and its composites showed similar corrosion behaviour. The corrosion rate of all the tested materials decreased rapidly during the first three days of exposure to the electrolyte but, with further exposure time, the decrease was very gradual. Passivation of the matrix alloy is believed to be responsible for the phenomenon of monotonically decreasing corrosion rate with increasing exposure time observed in these materials [12]. It is also seen that 15FA/A535 composite showed the highest rate of corrosion, followed in decreasing order by 10FA/A535, 5FA5SiC/A535, and A535. It was also observed that the corrosion of the composites was accompanied by loosening of fly ash particles, with the amount of loosened fly ash being greatest in 15FA/A535 composite, followed by 10FA/A535 composite. It was believed that the corrosion of the fly ash-matrix interface caused the loosening of fly ash particles which were finally dislodged from the specimens during postimmersion cleaning process. Ramachandra and Radhakrishna [11] have reported that fly ash particles acted as pit initiation sites in Fly ash/Al-Si alloy composites WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


25

and that there was a buildup of corroded fly ash particle debris in corrosion pits. The loss of such particles during contributed to the high weight loss recorded for the MMCs in the present study. The effect of fly ash addition on the corrosion potential of A535 alloy is shown in Fig. 2 where curves of potential versus current density obtained via potentiodynamic polarization measurements are plotted for the tested materials. It can be seen that all the curves are similar indicating that polarization behaviour of unreinforced A535 alloy and its composites is similar. The corrosion potential (Ecorr) of the unreinforced alloy is more positive than that of the composites which tends to increase with increasing fly ash content. The corrosion potential of A535 alloy is -415 mV (SCE) while those of 10FA/A535 and 15FA/A535 composites are -443mV (SCE) and -507mV (SCE), respectively. Hence, the unreinforced A535 alloy is more noble than its MMCs. Similar results have been reported by Bienias et al [9] for fly ash/AL-Si alloy composites. Fig. 3 shows the cyclic potentiodynamic polarization curves obtained for A535 alloy and its fly ash-reinforced MMCs immersed in 3.5wt% NaCl (pH = 7) while Figs 4 and 5 show respectively the variation of critical pitting potential (Ep) and repassivation potential (Erp) with increasing fly ash content. Fig. 4 shows that the Ep of the tested materials became more negative with the addition of fly ash. It decreased from about 166 mV in A535 to -237.8mV in 15FA/A535 composites, indicating that A535 alloy has better pitting corrosion resistance in 3.5wt% NaCl solution than its composites. On the other hand, Fig. 5 shows that Erp increases (in the active direction) with increasing fly ash content. It increased from -822.5mV (SCE) in A535 alloy to -799.3mV (SCE) in 15FA/A535 composite. Since Erp measures the ability of a material to repassivate, the present results show that pit propagation in the composites is retarded more than in the matrix alloy. A measure of the tendency for pits to nucleate in a material is given by the difference between Ep and Ecorr. Thus, the ability of a material to resist pit initiation during localized corrosion increases as the value of Ep – Ecorr becomes larger [20]. Fig. 6 shows a plot of Ep – Ecorr for the materials studied. It can be seen that A535 alloy has superior pitting corrosion resistance to the composites. -0.35

Potential (V)

-0.40

-0.45

-0.50

A535 5FA5SiC/A535 10FA/A535 15FA/A535

-0.55

-0.60 1e-8

1e-7

1e-6

1e-5

1e-4

Current density (Acm-2)

Figure 2:

Potentiodynamic polarization curves for A535 alloy and its fly ash reinforced MMCs in 3.5wt% NaCl solution (pH =7).


26 Computational Methods and Experiments in Materials Characterisation III 0.4

A535 5FA5SiC/A535 10FA/A535 15FA/A535

0.2 0.0

Potential (V)

-0.2 -0.4 -0.6 -0.8 -1.0 -1.2 -1.4 1e-8

1e-7

1e-6

1e-5

1e-4

1e-3

1e-2

1e-1

2

Current density (A/cm )

Figure 3:

Cyclic potentiodynamic polarization curves for A535 and its MMCs in 3.5wt% NaCl solution (pH = 7). 200

Potential (mV)

100

0

-100

-200

-300 A535

5FA5SiC/A535

10FA/A535

15FA/A535

Materials

Figure 4:

Effect of fly ash content on the pitting potential of A535 and its MMCs.

The corrosion behaviour of particulate Al MMCs is influenced by several factors such as porosity, high dislocation densities at the matrix-reinforcement interfaces, the presence of intermetallic compounds (IMCs) and reaction products, and the electrical conductivity of the reinforcing phases [19]. Gikunoo and Oguocha [24] have reported that the amount of dimagnesium silicide, Mg2Si, and spinel, Al2MgO4, in fly ash/A535 composites increased with increasing fly ash content. Mg2Si is produced in the matrix alloy through a solid-state reaction between Si and Mg 2Mg Si l Mg 2Si (1) In the MMCs, the SiO2 phase present in fly ash particles or covering the surface of SiC particles in 5FA5SiC/A535 composite is reduced by molten magnesium through a two-step reaction leading to the formation of the Mg2Si phase: 2Mg + SiO2 l 2MgO + Si (2) (3) 2Mg + Si l Mg2Si



27

The spinel phase is formed in the MMCs via a reaction between elemental magnesium of the matrix alloy and fly ash constituents, particularly the alumina (Al2O3) and quartz phases, following either of the chemical reactions: (4) 3Mg + 4Al2O3 l 2Al + 3MgAl2O4 2SiO2 + 2Al +Mg l MgAl2O4 + 2Si (5) -795 -800

Erp (mV)

-805 -810 -815 -820 -825 A535

5FA5SiC/A535 10FA/A535

15FA/A535

Materials

Figure 5:

Effect of fly ash content on the repassivation potential of A535 and its MMCs. 700 600

Ep - Ecorr (mV)

500 400 300 200 100 0 A535

5FA5SiC/A535

10FA/A535

15FA/A535

Materials

Figure 6:

Plot of Ep–Ecorr for A535 alloy and its MMCs in 3.5wt% NaCl solution (pH = 7).

The presence of the intermetallic phases and porosities in the MMCs serve as preferential sites for localized corrosion. In the present study, optical microscopy observation of the corroded surfaces of specimens immersed in NaCl solution for several days showed that pits occurred where the Mg2Si phase existed prior to immersion. This was well pronounced in the matrix alloy thus indicating that Mg2Si has a less noble potential than the alloy in 3.5wt% NaCl solution at room temperature. It was reported by Birbilis [22] that Mg2Si does not show any breakdown potential and is capable of corroding freely above its Ecorr, which was WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

28 Computational Methods and Experiments in Materials Characterisation III measured to be -1536 mV (SCE) in 0.6M NaCl [22] and -1530 to -680 mV (SCE) in 3wt% NaCl [23] as compared to -849 mV SCE for pure aluminum in 0.6M NaCl [22] and -760 to -810 mV (SCE) for Al-Mg binary alloy in 53g/l NaCl+3g/l H2O2 solution [23]. Since aluminum is noble to Mg2Si, the microgalvanic couple formed between them in A535 alloy and its MMCs would selectively corrode Mg2Si away. Therefore, the deep pits observed in the A535 alloy are attributed to the dissolution of the Mg2Si

4 1 2 3 4

Conclusions The corrosion rate of fly ash/A535 MMCs immersed in 3.5wt% NaCl solution at room temperature was higher than that of the matrix alloy and increased with increasing fly ash content. Fly ash/A535 MMCs showed increased susceptibility to pitting corrosion compared to the unreinforced A535 alloy in NaCl solution. The pitting potential (Ep) of the composites decreased with increasing fly ash content. The sites for pit initiation in A535 alloy were the intermetallic compounds, especially the Mg2Si phase which dissolved away with increasing immersion time. The predominant pit initiation sites in the MMCs were the interfaces between the matrix alloy and fly ash and intermetallic compounds such as Mg2Si and Al2MgO4.

Acknowledgement This work was supported by the Natural Sciences and Engineering Research Council of Canada via a discovery grant to I. N. A. Oguocha.

References [1] [2] [3] [4] [5] [6]

Chawla, N. & Shen Y., Mechanical Behavior of Particle Reinforced Metal Matrix Composites. Advanced Engineering Materials, 3(6), pp. 357-370, 2001. Lloyd, D.J., Particle-Reinforced Aluminum and Magnesium Matrix Composites. International Materials Reviews, 39(1), pp. 1-23, 1994. Fujine, M, Kaneko, T, & Okijima, J, Adv. Mater. Process, 143(6), pp. 2021, 1993. Akbulut, H, Durman, M & Yilmaz, F, Scripta Materialia, 36, pp. 835-840, 1997. Goni, J, Mitxelena, I, & Coleto, J, Mater. Sci. Technol., 16, pp.743-746, 2000 Rohatgi, P.K, Kim, J.K, Guo, R.Q, Robertson, D.P & GajdardziskaJosifovska, M, Age-hardening characteristics of aluminum alloy-hollow fly ash composites, Metall. Mater. Trans., 33A, pp. 1541-1547, 2002.



[7] [8] [9] [10] [11] [12] [13]

[14] [15] [16] [17] [18] [19]

[20]

[21]

29

Golden, D, Ashalloys: aluminum-fly ash composites. EPRI Journal, 19(1), pp. 46(4), 1994. Guo, R.Q & P K Rohatgi, P.K, Chemical reactions between aluminum and fly ash during synthesis and reheating, Metall. Mater. Trans., 29B, pp. 519-525, 1998. Bienias, J, Walczak, M,. Surowska, B & Sobczak, J., Microstructure and Corrosion Behaviour of Aluminum Fly Ash Composites. Journal of Optoelectronics and Advanced Materials, 5(2), pp. 493-502, 2003. Gikunoo, E., Omotoso, O. & Oguocha, I.N.A., Effects of Fly Ash Particles on the Mechanical Properties of Aluminum Casting Alloy 535. Material Science and Technology, 21(2), pp. 143-152, 2005. Ramachandra, M & Radhakrishna, K., Microstructure, Mechanical Properties, Wear and Corrosion Behaviour of Al-Si/flyash Composite, Materials Science and Technology, 21(11), pp. 1337-43, 2005. Seah, K.H. W, Krishna, M., Vijayalakshmi, V.T. & Uchil, J., Corrosion Behaviour of Garnet Particulate Reinforced LM13 Al Alloy MMCs, Corrosion Science, 44, pp. 917-925, 2002. De Salazar, J.M.G., Urena, U., Manzanedo, S. & Barrena, M.I., Corrosion behaviour of AA6061 and AA7005 reinforced with Al2O3 Particulates in Aerated 3.5% Chloride Solutions: Potentiodynamic Measurements and Microstructure Evaluation, Corrosion Science, 41, pp. 529-545, 1999. Gnecco, F.F, Corrosion Behaviour of Al-Si/SiC Composite in Sea Water. British Corrosion Journal, 34(1), pp. 57-62, 1999. Gavgali, M., Dikici, B. & Tekmen, C., The effect of SiCp Reinforcement on the Corrosion Behaviour of Al Based Metal Matrix Composites, Corrosion Reviews, 24(1-2), pp. 27-37, 2006. Kiourtsidis, G & Skolianos, M., Corrosion Behavior of Squeeze-cast Silicon carbide-2024 composites in aerated 3.5 wt.% sodium chloride. Materials Science and Engineering, A248, pp. 165-172, 1998. Aylor, D.M & Moran, P.J, Effect of Reinforcement on the Pitting Behavior of Aluminum-Base Metal Matrix Composites. Journal of The Electrochemical Society, 321(6), pp. 1277-1281, 1985. Candan, S. & Bilgic, E., Corrosion Behavior of Al-60 Vol.%SiCp Composites in NaCl Solution. Materials Letters, 58, pp. 2787-2790, 2004. Albiter, A., Contreras, A., Salazar, M. & Gonzalez-Rodriguez, J.G, Corrosion Behaviour of Aluminium Metal Matrix Composites Reinforced with TiC Processed by Pressureless Melt Infiltration, Journal of Applied Electrochemistry, 36 (3), pp. 303-308, 2006. Pardo, A., Merino, M.C., Merino, S., Viejo, F., Carboneras, M. & Arrabal, R., Influence of Reinforcement Proportion and Matrix Composition on Pitting Corrosion Behaviour of Cast Aluminium Matrix Composites (A3xxx.x/SiCp). Corrosion Science, Vol. 47, Issue 7, 2005, pp. 17501764. . A. Pardo, M. C. Merino, F. Viejo, S. Feliu, Jr., M. Carboneras and R. Arrabal, Corrosion Behavior of Cast Aluminium Matrix Composites



[22] [23] [24]

(A3xxx.x/SiCp) in Chloride Media. Journal of Electrochemical Society, 152(6), pp. B198-B204, 2005. Birbilis, N.N., Electrochemical Characteristics of Intermetallic Phases in Aluminum Alloys: An Experimental Survey and Discussion. Journal of The Electrochemical Society, 152(4), pp. B140-B151, 2005. Buchheit, R. G, Compilation of Corrosion Potentials Reported For Intermetallic Phases in Aluminium Alloys. Journal of The Electrochemical Society, 142(11), pp. 3994-3996, 1995. Gikunoo, E. & Oguocha, I.N.A. Proc. Of the 6th Joint Canada-Japan Workshop on Composites, ed. J. Lo, T. Nishino, S.V. Hoa, H. Hamada, A. Nakai, C. Poon, DEStech Publications, Inc., Toronto, Canada, pp. 387396, 2006.



31

Testing of palm fibre as reinforcement material in polyester composites V. V. S. Prasad1, D. N. Rao2, K. N. S. Suman2 & N. R. M. R. Bhargava3 1

Department of Marine Engineering, Andhra University, Visakhapatnam-53003, India 2 Department of Mechanical Engineering, Andhra University, Visakhapatnam-53003, India 3 Department of Metallurgical Engineering, Andhra University, Visakhapatnam-53003, India

Abstract In this present work, palm fibre is incorporated in a polyester resin matrix to form unidirectional reinforced composites and bi-directional composites. Samples of different fibre volume fractions are fabricated and specimens with 0°, 45° and 90° fibre orientations are prepared. The specimens are tested on a universal testing machine applying tensile force. The tensile strength is measured as a function of fibre volume fraction. These properties follow “Rule of mixtures” relationship, with the volume fraction of palm. Because of the low density of natural fibers and high electrical resistance, these composites are more suitable for electrical and mechanical applications. Keywords: palm fibre, hand lay up technique, mechanical and electrical properties.

1

Introduction

There is a great interest in the development of new materials which enhance optimal utilization of natural resources, and particularly of renewable resources. Natural fibres such as palm, jute, coir, banana, sisal etc., belong to this category. These fibres are abundantly available in developing countries, particularly in

WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070041

32 Computational Methods and Experiments in Materials Characterisation III India and some places of South Africa [1]. The cotton polymer composite made contributions during the II world war [2,3]. As fibre reinforced plastics, it was first used by the military for radar domes on aircraft. During that period, bearings for ships were made of cotton-phenolic systems; also, brake linings of plywood-phenolic trainer wings and fuselages of British Mosquito bombers, and more than a thousand other items were made. During 1942, the Goodyear Aerospace Corporation for use in aircraft fuel cell produced backing sheet materials made of cotton fabric-polyester. It is recently reported [4] that cotton fabric reinforced phenolic resin composites have been used as bearings in place of phosphor bronze in the roll necks of steel and non-ferrous rolling mills. This resulted in energy savings up to 25%. Reddy, et al [6, 7] studied on fabrication, testing, damage characterization and feasibility of jute-polyester composites. Unprocessed jute yarn and fabric are used as fibers. Twisted jute yarn and fabric, which are semi finished raw materials and commercially produced widely in India, are selected for the work. General-purpose polyester is used as the matrix. Jute reinforced polyester laminates are prepared using 'Hand lay-up' technique to simulate practical production methods. Results indicate that there exists definite correlation between the tensile strength or elastic modulus and fiber volume fraction of the composite and with variation in fiber orientation in the composite. One of the earliest natural fibre-polymer composites are investigated by Paramasivan and Abdulkalam [5] by incorporating sisal fibres and epoxy matrix. The fabrication process attempted by them includes winding and lamination. It is found that the fabrication of these composites is fairly easy and cost of production is quite low. Winding of cylinders with longitudinal or helical and hoop reinforcements is successfully carried out. Tensile strength of the sisalepoxy composites is found to be 250-300 Mpa, which is nearly half the strength of fibre glass–epoxy composites of the same composition. Because of the low density of sisal fibre, however, the specific strength of sisal composites is comparable with that of glass composites. The unidirectional modulus of sisalepoxy composites is found to be about 8.5Gpa. This study indicates the feasibility of developing composites incorporating one of the abundantly available natural fibres, to be used in the field of consumer goods, low-cost housing and civil structures. Lakkad and Patel [8] compared the values of ultimate tensile and compressive strength and young’s modulus of elasticity of bamboo specimens with those of mild steel and glass reinforced plastics. But they have not specified the speciename of the bamboo specimens tested. There are more than 500 species of bamboo available in India and each has different mechanical properties. Extensive literature is available on the production and mechanical behavior of composites obtained by reinforcing epoxy with fibre of glass, boron, carbon silicon carbide etc. Many researchers in the past have developed composites with natural fibres such as sisal henequen, jute, banana, cotton, etc., but the work on the palm reinforced plastic composites and palm reinforced oriented plastic composites are not available in the literature.



2

33

Fabrication of palm reinforced plastic composites

Palm fibre is extracted from the leaf stem of the palm tree, which is not optimally used. Initially the cut stem from the plant is collected in a heap, and the stem is kept wet by spraying water for 48 hours for free release of the husk. Later by tapping lightly with wooden hammer, on the stem the fibre is separated in two forms as coarse fibre (i.e. 150Pm-1500Pm), length up to 500mm and fine fibre (i.e.75Pm-150Pm) length up to 70mm The fibres are flexible compared to the coarser fibres and segregated in the form of bundles. A rectangular thick tapered plastic plate of size (200 X 50) cm2 is used as a mould for making the composite by using “Hand- lay-up technique”. Acetylene is used as a cleaning agent for cleaning the casting surface of the mould, a releasing agent polyvinyl alcohol is used for easy removal of the casting. After thoroughly mixing the resin with hardener, it is applied over the entire sheet using a soft brush and a coat of wax is applied on this resin layer. The finer fibres are inserted in the wax placing them parallel to the longer edges of the mould plate, and brushing is done smoothly so that resin spreads through the yarn. Care is also been taken to see that the yarns are not being displaced from respective positions after brushing. This process is repeated till all the palm fibres are wet properly. The laminates are cured at room temperature for 24 hours. Laminates with approximately 10%, 20%, 30% and 40% of the fibre volume fraction are prepared as shown Fig 2. For the laminates with volume fraction 50% above it is

Figure 1:

Figure 2:

Palm fibre (finer type) bundle.

Palm reinforced plastic laminate.


34 Computational Methods and Experiments in Materials Characterisation III observed that wetting of the fibre is not proper and there is no much improvement in strength. To fabricate the bi-directional composite, the second layer of palm is placed perpendicular to the first layer and the above process is repeated till the resin spreads over the entire surface. Fig.3 shows the bi-directional composites. Oriented fibre composites are prepared by placing second layer of palm at different angles of 15o, 30o and 45o for each composite. Figs.4–6, depict the various orientations respectively.

Figure 3:

Palm bi-directional composites.

Figure 4:

Laminates with 15° fibre orientation with a fibre volume of 20%.

Figure 5:




Figure 6:

3

35


Experimental procedure

3.1 Tensile test As per ASTM standards the specimen are prepared to the required size of 250mmx25.5mmx4mm with 0o, 45o 90o of fibre orientation. The standard specimens are marked with marking scriber and samples are cut to size using power band saw. PFRP samples are tested on tensile testing machine; UNITEK95100 under a load of 25KN and with a cross head speed of 20mm/min. The specimens are held by flat graved grips. To avoid slipping of grippers during load application, the ends of specimens are made rough by filing. The breaking loads and displacements at various loads are measured. The observing results are presented in figs. 7, 9 and 10 for different percentages of volume fraction and orientations of fibre. 3.2 Electrical test The present test is designed to measure the leakage current between two points. Leakage test is conducted on Cascade transformer, 100 kVA, 500 kV, 200 µA and an ammeter is connected to measure the leakage current. Resistance is calculated using the Kirchoff’s law (V=I R) at constant voltage. Specimens in the normal and soaked in sea water for 12hrs, are tested for breakdown voltage, at the High Voltage Laboratory, Department of Electrical Engineering, Jawaharlal Nehru Technical University, Kakinada, Andhra Pradesh, India.

4

Results

From Fig. 7, it is observed that the tensile strength of the composites increases with increase in the fibre volume fraction. Fibres are the main load carrying agents in composites and as the number of load carrying elements increases in a material, its strength increases. The composite tensile strength decreases with increasing the orientation of the fibre from 0° to 90°. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

Breaking load ,kN

36 Computational Methods and Experiments in Materials Characterisation III 0 Degree orientation

5 4.5 4

45 Degree orientation 90 Degree orientation

3.5 3 2.5 2 1.5 1 0.5 0 0

10

20

30

40

50

% Volume of fibre

Figure 7:

Effect of % volume and orientation of palm fibre on breaking load for uni-directional composite.

Figure 8:

Fractured surface of 10 vol % palm fibre reinforced polymer composite 10X.

Figure 8 shows the fractured surface of the 10 vol % palm fibre reinforced composite. Matrix is found to be deformed to a lesser extent while fibres are protruding from the surface. It shows that fibres have been pulled away from matrix indicating poor bonding at the fibre- matrix interface. This effect is much more pronounced at higher percentages of the fibre and has resulted in lower breaking load values compared to the theoretical calculations of rule of mixtures. Figure 9 shows the effect of fibre volumes and the fibre orientation on the breaking load of bi-directional composites. Since the composite is made bidirectional, breaking load values for the fibre orientations of 0o and 900 have shown similar values at all the fibre contents. These values are found to increase with increasing fibre contents. Composites with fibre orientations of 450 have shown a similar trend but have shown lesser strength values. These results are self explanatory as the strength of the fibre in the warp direction is more than that of in the weft direction.



37

Breaking load vs Volume of fibre

Breaking load,kN

0 Dgree orientation

4 3.5 3 2.5 2 1.5 1 0.5 0

45 Degree orientation

0

10

20

30

40

50

% volume of fibre

Effect of palm fibre volume and orientation on breaking load of bidirectional composite.

B reakin g lo ad ,kN

Figure 9:

2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0

10

20

30

40

50

Fibre orientation angle

Figure 10:

Effect of palm fibre volume and orientation on breaking load of palm oriented composites.

Figure 10 shows the effect of fibre orientation angle on the tensile strength of the composites. As the orientation angle increases the tensile strength drops to a minimum at the maximum weft of 450. Figure 11 shows the effect of palm fibres volume on the leakage current. Leakage current found to be increasing with increased fibre volumes initially and almost stabilizes at higher contents. From the literature it is found that presence of voids and air pockets enhance the leakage currents. Since palm is natural one and is also in the thoroughly dried condition, sufficient voids are readily present in it. This might have lead to the increased leakage currents in the reinforced composite. As the fibre volume increases, the presence of these discontinuities also increases which might have lead to the increased leakage currents. During processing increased fiber volumes enhance the chances of void presence due to practical problems. This might have further accentuated the leakage voltage at higher volumes of the fibre. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

38 Computational Methods and Experiments in Materials Characterisation III L.C. vs Applied voltage

Leakage current , M icro am pears

60 50 40 30 20 10 0 22

23

25

28

Applied Voltage,kV

Figure 11:

Effect of palm fibre volume on leakage current. Normal

90

Saline

Break down voltage,kV

100 80 70 60 50 40 30 20 10 0

10

20

30

40

50

% Volume of fibre

Figure 12:

Effect of fibre volume on break down voltage.

Figure 12 shows the effect of fibre volume on breakdown voltage. Specimens soaked in the saline water have shown drastic drop in the breakdown voltages compared to the normal samples. A similar trend of drop in breakdown voltage with increasing fibre volumes has been observed with both the conditions of normal and seawater soaked ones. Since the presence of voids, impurities and the moisture decreases the breakdown voltage, the above discussion holds good for this behaviour as well. Presence of moisture has dropped the values further.

5

Conclusions

1.

Palm fibre can be used as reinforcement and filler in the polymer based composites. It shows a conventional behaviour in mechanical properties depicting higher breakdown strength values with increasing fibre volumes. Composites with fibre orientation in the warp direction exhibit better mechanical properties than the weft direction ones.

2. 3.



4. 5. 6.

39

Leakage current increases with increasing fibre volumes. Breakdown voltage decreases with increasing fibre volumes. Saline water soaked samples exhibit poor breakdown voltage compared to the normal ones.

References [1]

[2] [3] [4] [5] [6]

[7] [8]

Satyanarayana, K.G., Kulkarni, G.Sukumaran, K., Pillai, S.G.K. Cheriyan, K.A. and Rohatgi, P.K., “on the possibility of using natural fiber polymer composites”. Proc. First International Conference on Composite Structures, 16-18(Sept., 1981), ed. 1.H.Marshall. Applied Science publishers, London, pp.618-623. Piggot, M.R., “Load Bearing Fiber Composites”. Pergamon press, Oxford, 1980. Lubin, G. (ed), “Hand Book of Composites”. Van Nostrand Reinhold, New York 1982. “Save energy – Save money – composite news. Composites”, 10 (April 1979) pp.61. Parmasivam, T and Abdulkalam, A.P.J., “On the study of natural fiber composites” Fiber Science and Technology I (1974) pp. 85-88. Govardhan Reddy, B., Rao, D.N., Bhargava, N.R.M.R. Prasad, V.V.S,. “Damage Mechanism under tensile loading of continuous jute reinforced polyester composites” Proc. Third International conference on ‘Advances in composites’ ADCOMP-2000, August 2000, Bangalore, India. pp.24-26. Govardhan Reddy, B., Rao, D.N., and Rao, R.N.S. “Jute-reinforced polyester composites – A study of properties”, Proc. Of 11th AGM, Materials Research Society of India, India, July 2000. Lakkad, S. C. and Patel, J. M., "Mechanical Properties of Bamboo, a Natural Composite," Fibre Science and Technology, Vol. 14, (1980-81) pp. 319-322.



Section 2 Microstructures – ceramics and advanced materials



43

Experimental study on fracture behaviour of polycrystalline ceramics under shock loading J. T. Zhou & G. W. Yao School of Civil Engineering and Architecture, Chongqing Jiaotong University, People’s Republic of China

Abstract Plate impact experiments and impact recovery experiments were performed on 92.93wt% alumina ceramics using a 100-mm-diameter compressed-gas gun. Free surface velocity histories were traced by a VISAR velocity interferometer. There is a recompression signal in free surface velocity, which shows evidence of a failure wave in impacted alumina. The failure wave velocities are 1.27km/s and 1.46km/s at stresses of 7.54GPa and 8.56GPa respectively. It drops to 0.21km/s after the material released. SEM analysis of recovered samples showed the transit of intergranular microcracks to transgranular microcracks with increasing shock loading. The failure wave in impacted ceramics is a continuous fracture zone which may be associated with the damage accumulation process during the propagation of shock waves. Keywords: plate impact experiment, alumina ceramics, failure wave, dynamic fracture, SEM.

1

Introduction

Since failure waves were first observed propagating in glass rods under dynamic compression by Bless et al [1] and in glass plates under high-pressure impulsive loading by Rasorenov et al [2], a series of plate impact experiments, bar impact experiments and impact recovery experiments have been performed on a range of glasses under various impact stresses [3–6]. These experiments show the failure fronts are generated in silicate and filled glasses at a stress near or below their Hugoniot Elastic Limits and propagate from impact surface to interior at velocities in the range of 1.5–2.5km/s. The failed glass has lower acoustic impedance and sound speed than the intact material. The failed layer nearly loses WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070051

44 Computational Methods and Experiments in Materials Characterisation III complete tensile strength, and its shear strength is significantly degraded. The longitudinal stress and transverse strain remain constant cross the failure front, but the transverse stress and longitudinal strain are increasing with time in the region behind the failure front. All these variations of material properties across this front provide experimental evidences for the existence of a failure wave phenomenon for glass under plate normal impact loading. In recent years, there also have been some wide researches made into other brittle materials. Bourne et al [7] and Zhang et al [8] have extended these studies to the polycrystalline ceramics alumina, silicon carbide and titanium diboride, gabbro and 3D-C/SiC composite materials and have postulated similar impact induced fracture front in these brittle materials. There also been recent discussion of the phenomenon of gradual failure behind the elastic wave in mortar by Grote [9]. In the work presented, we have conducted a matrix of plate impact experiments on alumina monitored by VISAR focused on the rear surface of the sample in seeking to pursue the failure wave in brittle materials other than glass.

2

Plate impact experiments

Plate impact experiments on alumina specimens were carried out on the 100mm light gas gun. Impact velocities were measured to an accuracy of 1.5% using three pairs of electric signal pins at different distances away from the impact surface. The copper flyers and targets were circular with different diameters of 94mm and 100mm, with their two cut faces polished in order to ensure smoothness of the impact and measurement area. Free surface velocity histories were traced using VISAR with a fringe constant 101m/s/fringe and a measured response time 1.5ns. The free surface of target was polished and aluminized with a layer 5000 angstroms in thickness to strengthen the reflection of incident laser (see fig. 1, which shows a schematic of the experimental setup). The impact recovery experiments were also performed to study micro-structures of impacted samples. Cushion rubber was filled in target room to absorb the dynamic energy of flyer and target. The flyer and target will be embedded in rubber.

Figure 1:

Plate impact experimental schematic with VISAR.



45

The alumina samples consist of 92.93wt% alumina by weight and a small amount of silicon dioxide, calcium oxide and lanthana analyzed by energy spectrum. The relevant parameters of the specimens are density 3896kg/m3, longitudinal sound speed 9.259km/s, and a shear wave velocity 5.557km/s respectively. The longitudinal sound speed in copper flyer is 4.60km/s and the thickness of flyers and targets range from 4.0mm to 6.1mm. The acoustic impedance ratio of flyer and target is 1.14, then long enough duration pulse generates at the impact surface to avoid the unloading wave propagating into targets from flyers. A summary of experimental conditions and results are presented in table 1. Table 1:

Parameters of plate impact experiments.

Parameters

Impact velocity (m/s)

Impact stress (GPa)

Impactor Thickness (mm)

Target Thickness (mm)

Shot 405

397.8

7.54

4.14

6.08

Shot 425

448.8

8.56

6.10

6.04

Figure 2:

Free surface velocity profiles of shots 405 and 425.

Fig. 2 shows reduced VISAR data by software from the experiments of shots 405 and 425 under shock stresses 7.54GPa and 8.56GPa. These profiles indicate that the alumina specimens did not spall. The distinct feature of note on the traces is the slight recompression signal pointed on top of the stress wave. This velocity jump behaves beyond the elastic behaviours because there is not reflected tensile pulse recorded in the profiles and the time interval between the start of free surface motion and the moment of this reloading signal is less than the elastic wave reverberation time in the sample. And alumina does not behave plasticity in macroscope as typical brittle material, so this inelastic behaviour does not characterized plasticity though the free surface velocity profile has twowave structure. The additional weak compression wave is associated with a WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

46 Computational Methods and Experiments in Materials Characterisation III reflection from a layer of material which has dynamic impedance lower than that of the intact alumina, and this material layer bordering the interface does not pass tensile stresses. So we conclude that the shock-compressed alumina is comminuted behind this interface. This phenomenon is akin to the failure wave which has been observed to occur in glasses under shock compression. On the assumption that the moving speed of failed layer boundary is the failure wave velocity CF, a simple evaluating equation for CF has been derived as the following (see fig. 3, which shows diagram of elastic wave and failure wave propagating). The thickness of the failed layer hf is determined from the measured time interval of ts through the equation

Figure 3:

Propagation of Compression, Rarefaction and Failure waves.

1 (1) h CP ts 2 where h is the sample thickness and CP is the longitudinal wave speed in alumina specimen. Then the failure wave velocity CF can be estimated by hf (2) CF h CP t s 2 It implies that the failure wave has propagated at a velocity of 1.27km/s in shot 405 and 1.46km/s in shot 425 on average before the moment tf. The free surface velocity history from VISAR measurements has shown that the failure front propagates at a speed much lower than longitudinal stress wave velocity, depending on the peak shock stress. The free surface velocity profile from shot 425 is analyzed further in expanded region and there is another smaller recompression signal observed following the first (see fig. 4, which shows second smaller recompression indicated by a narrow). This can be explained if the reflected rarefaction wave from rear surface is reflected again on the failure layer and then reflected on rear surface where a weak jump of velocity is produced at the same time. During the hf



47

time interval ts1 of two recompression signals, the distance of failure layer expanded can be determined through 1 (3) 'h f h f 1 h f h CP ts1 h f 2 Then the average velocity CF1 of failure wave propagating from the moment tf to tf1 can be estimated from the measured time interval of ts1 by 'h f (4) CF 1 (ts ts1 ) 2 This implies that the failure wave has propagated at an average velocity of 210m/s in shot 425 following unloading by the reflected rarefaction wave. This unloading slows down and even eventually arrests the failure procedure in material and results in great lowness in the failure wave propagating.

Figure 4:

3

Expanded region of free surface velocity profile of shot 425.

SEM for samples

To explain the failure process of shock-compressed polycrystalline ceramics in mesoscope, initial and soft-recovered samples were scanned by S530 scanning electron microscope. Each fragment was cut in the centre along a plane parallel to impact surface with 0.2mm distance to impact surface. Fig. 5(A) shows the micro-structures of initial 92.93wt% alumina. Grains and intergranular pores distribute randomly with diameters 1-15µm. Intergranular glassy phase is distinct in compact area. And initial porosity is 5.68% determined by metallurgical analysis software. Pores and glasses weaken mechanical capabilities and these heterogeneous meso-structures result in high singularity in stress distribution. Fig. 5(B) shows intergranular microcracks in recovered sample after 5.76GPa loading and Fig. 5(C) shows transgranular microcracks in recovered sample after 8.65GPa loading. Microcracking transmits from intergranular to transgranular with increasing impact compression. Alumina grains begin to fragment with transgranular microcracks and original pores begin to collapse. And discontinuous microcracks induce dilation after unloading. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


Figure 5:

SEM micrographs of (A) initial and recovered alumina samples under (B) 5.76 GPa and (C) 8.65 GPa shock loading.

The polycrystalline ceramics are heterogeneous in mesoscope. There are many pores, microcracks and other defaults inducing high singularity in stress distribution. Once the local stress exceeds the threshold, the original microcracks will grow up along the pores and crystal boundaries and new microcracks will nucleate in ceramics under shock loading. The original and nucleated microcracks grow up and expand, then excite the neighbour microcracks nucleation and expansion. So the failure wave appears and propagates from impact surface to interior of specimen, and it propagates at higher velocity under stronger shock loading. In essence, the failure wave is characterized by moving damage or fracture zone of material which presented by microcracking system in mesoscope, and it is also called after fracture wave by Resnyansky et al [12].

4

Summary

Ceramics are extensively applied to national defence engineering and military science as effective armour defence with their excellent physical and mechanical capabilities, especially higher dynamic elastic threshold and acoustic velocity than metals. We performed plate impact experiments of 92.93 wt% aluminas with 100-mm-diameter compressed-gas gun and the free surface velocities were traced by VISAR. There is a reloading signal observed in free surface velocity which indicates the failure wave propagation behind the elastic precursor. The failure wave propagates at a speed much lower than longitudinal stress wave velocity, depending on the peak shock stress. And the failed layer has much lower dynamic impedance than that of the intact material. The unloading by the reflected rarefaction wave slows down and even eventually arrests the failure front propagating in alumina. SEM analysis of intact samples shows heterogeneous meso-structures, and SEM analysis of soft-recovered samples shows transit of intergranular microcracks to transgranular microcracks with increasing shock loading. The failure wave is a continuous fracture or damage front which may be associated with nucleation and expansion of microcracks from impact surface to interior during the propagation of shock waves. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


49

References [1] [2] [3] [4] [5]

[6]

[7] [8] [9] [10] [11] [12]

Bless, S.J., Brar, N.S. & Rosenberg Z., Failure of Ceramic and Glass Rods under Dynamic Compression. Shock Compression of Condensed Matter1989, eds. S.C. Schmidt, APS: New Mexico, USA, pp. 939-942, 1990. Rasorenov, S.V., Kanel, G.I., Fortov, V.E., & Abasehov, M.M., The Fracture of Glass under High-pressure Impulsive Loading. High Pressure Research, 6, pp. 225-232, 1991. Rosenberg, Z., Bourne, N.K., & Millett, J., Direct Measurements of Strain in Shock-loaded Glass Specimens. Journal of Applied Physics, 79, pp. 3971-3974, 1996. Bourne, N.K., Millett, J., & Rosenberg, Z., On the Origin of Failure Waves in Glass. Journal of Applied Physics, 81, pp. 6670-6674, 1997. Millett, J., Bourne, N.K., & Rosenberg, Z., Measurements of Strain in a Shock Loaded, High-density Glass. Shock Compression of Condensed Matter-1999, eds. M.D. Furnish, AIP: Utah, USA, 505, pp. 607-610, 2000. Cazamias, J.U., Fiske, P.S., & Bless, S.J., Sound Speeds of Post-failure Wave Glass. Fundamental Issues and Applications of Shock-Wave and High-Strain-Rate Phenomena, EXPLOMET 2000, eds. K.P. Staudhammer, New Mexico, USA, pp. 173-179, 2000. Bourne, N.K., Millett, J., Pickup, I., Delayed failure in shocked silicon carbide. Journal of Applied Physics, 81(9), pp. 6019-6023, 1997. Zhang, Q.M., Huang, F.L., & Han, L.M., Failure Wave Motion of 3DC/SiC Composites Subjected to Shock Compression. Chinese Science Bulletin, 45, pp. 408-411, 2000. Grote, D.L., Park, S.W., & Zhou, M., Experimental Characterization of the Dynamic Failure Behavior of Mortar under Impact Loading. Journal of Applied Physics, 89, pp.2115-2123, 2001. Kanel G.I., Bogatch A.A., Razorenov S.V., & Zhen Chen, Transformation of shock compression pulses in glass due to the failure wave phenomena. Journal of Applied Physics, 92(9), pp. 5045-5052, 2002. Zhao J.H., Sun C.W., Zhao F., Duan Z.P., et al, Velocity overshoot of rear free-surfaces of glass under impact. Explosion and Shock Waves, 22(1), pp. 72-78, 2002. (in Chinese) Resnyansky, A.D., Romensky, E.I., & Bourne, N.K., Constitutive Modeling of Fracture Waves. Journal of Applied Physics, 93, pp. 15371545, 2003.




51

Blocking and self-locking of superdislocations in intermetallics B. A. Greenberg1 & M. A. Ivanov 2 1

Institute of Metal Physics, Ural Division, Russian Academy of Sciences, Ekaterinburg, Russia 2 Institute of Metal Physics, National Academy of Sciences, Kiev, Ukraine

Abstract Superdislocations are carriers of plastic deformation in intermetallics. A large translation vector, different types of stacking faults and antiphase boundaries determine the diversity of dislocation configurations, both glissile and blocked ones. A significant point is that blocked superdislocations, which are formed due to re-splitting of glissile superdislocations or rearrangement of the superpartial dislocation core, have the lowest energy. A new concept about the possibility of thermally activated blocking of superdislocations in the absence of external stresses (self-locking) was proposed. A sufficiently general thermally activated process, which causes the extension of a dislocation in a preferred direction and constitutes a necessary step in dislocation transformations leading to blocking, was revealed. By its nature, this process represents the flip of a dislocation from a shallow valley to a deep valley of the potential relief. Reasons for the multivalley relief and the presence of preferred directions vary for dislocations of different types in different materials. Consecutive stages of the rearrangement of an initial dislocation include the formation of a double kink and its subsequent reorientation along a preferred direction. The driving force of the process was calculated and conditions for its realization in the cases of perfect, superpartial and partial dislocations were formulated. An experimental proof of the proposed concept was obtained: self-locking of dislocations, which were induced by preliminary deformation, was detected in Ni3(Al, Nb) and TiAl during no-load heating. Keywords: dislocations, plastic deformation, potential relief, shallow valley, deep valley, dislocation blocking, self-locking, no-load heating.



1

Introduction

Although dislocation blocking mechanisms, which determine the deformation behavior of materials, are diverse, they can be divided into two groups. In the first group point blocking is due to pinning centers. In the second group linear blocking is explained by different factors, such as dislocation reactions, collisions of dislocations with domain boundaries and, finally, dislocation inherent transformations [1]. Such transformations, which are inherent in dislocations as linear defects, just represent the subject of this study. They are accomplished without participation of other dislocations and result from the rearrangement of the core of a perfect dislocation (BCC metals, TiAl – a single dislocation) or a partial dislocation (semiconductors). In some high-temperature intermetallics such transformations also result from re-splitting of a perfect or a superpartial dislocation [2]. Regardless of transformation details, a common feature is that the dislocation energy is gained at the expense of the dislocation mobility: a glissile dislocation turns to a dislocation barrier, which either remains indestructible or, under certain conditions, can re-transform to a glissile configuration. The barrier axis is the preferential direction along which the transformation to a low-energy configuration is realized. As a result, the potential relief is a multi-valley one for a dislocation: deep valleys extend in the preferential direction and shallow valleys go in other directions (fig. 1). Valleys of different depth along different directions can be distinguished (unlike fig. 1) in a three-dimensional display of the potential relief.

Figure 1:

2

Schematic image of the potential relief; shallow valleys and deep valleys of one type (a) or two types (b).

Flip-process

The flip of a dislocation from a shallow to a deep valley of the potential relief, which causes the extension of the dislocation in the preferential direction, is a sufficiently general thermally activated process and constitutes a necessary step in dislocation transformations leading to blocking. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


Figure 2:

53

Consecutive stages of the rearrangement of the initial dislocation whose direction is close to the preferential direction: a – double kink; b – reorientation in the preferential direction.

The flip process includes the formation of a double kink (fig. 2a) and its subsequent reorientation in the preferential direction resulting in the formation of an asymmetric kink (fig. 2b). The internal structure of the dislocation changes in the preferential direction and, hence, the dislocation energy decreases. Because this process takes place at different points along the dislocation line, the initial dislocation is broken down into long blocked segments. The flip process and the subsequent transfer from a deep to another deep valley ultimately determine the temperature dependence of the yield stress, Vy(T). If the release from deep valleys is possible, Vy(T) will exhibit a normal behavior. If such release is hampered (indestructible barriers), an anomalous trend of Vy(T) will be observed in certain conditions. According to Indenbom et al. [3], a double kink may be viewed as the nucleus of a "new phase" corresponding to the transition to a neighboring valley. We think that a chain [4] of asymmetric kinks rather than a single kink can be formed near the preferential direction (fig. 3a). This chain may be considered as a nucleus capable of transforming to a segment extended along the preferential direction and a multiple kink (fig. 3b). In any case, the extension of a dislocation in the preferential direction is a thermally activated process since it includes formation and propagation of kinks.

3

Nucleation and propagation of kinks

3.1 Perfect dislocations Let us consider a potential relief of the following form: shallow valleys and a deep valley in some preferential direction, with the deep valley separated from the nearest shallow valley by a potential barrier. We shall assume for simplicity that the initial direction and the preferential direction are almost parallel. The double kink consists of initial dislocation segments located in a shallow valley, a segment of the length d flipped to the deep valley, and single kinks connecting WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

54 Computational Methods and Experiments in Materials Characterisation III these segments. According to [5], the energy of the double kink can be written as the sum of the following contributions (with the corresponding signs): the double energy of a single kink; the work of the external stress V consumed for the formation of the double kink; and the interaction energy of the kinks. Furthermore, it is necessary to consider the change of the linear energy as the dislocation is transferred from a shallow valley to the deep valley. This contribution, which is connected with different depth of the valleys, was disregarded earlier and distinguishes the present study from other investigations. The critical configuration of the double kink, which is determined from conditions of the total energy extremum, has the length dc equal to

dc

Ka 2 , (ıba ǻE ) ǻE E0 Ev ,

(1) (2)

Here E0 is the energy of the dislocation in a shallow valley, Ev is the energy of the dislocation in the deep valley (both energies per unit length of the 2

dislocation), K kµb , P is the shear modulus, and k is a coefficient depending on the dislocation orientation. If the external stress is not applied, the unstable configuration, which causes dispersion of the kinks, appears, as can be seen from (1), due to an additional driving force proportional to ǻE . If ı = 0, the condition for the flip process, which causes autoblocking of the dislocation, is the inequality ǻE ! 0.

Figure 3:

Chain of asymmetric kinks (a) and its transformation (b).

Obviously, this driving force simultaneously counteracts the reverse transition from a deep valley to a shallow valley. Therefore, thermally activated formation of indestructible barriers can be expected during no-load heating. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


55

Self-locking is still possible if directions, along which shallow and deep valleys are extended, are not parallel, but the angles between them are not too large. However, the critical configuration is not formed at large angles, because the energy loss during the kink spreading is not compensated by the energy gain during the dislocation flip to a deep trap. Let us estimate, rather roughly, the possibility that the configuration, which appears after the double kink reorientation (fig. 2b), develops or, oppositely, collapses. We shall assume for simplicity that a single kink is perpendicular to the preferential direction. Then the condition of the self-locking is that the energy of the dislocation, which is a broken line and consists of a segment of the length d extended in the preferential direction and a single kink of the length h, is smaller than the energy of the initial rectilinear dislocation of the length l. This condition can be written as Ev d E0 h E0l . (3) Introduce the angle M between the preferential direction and the direction of the initial dislocation (fig. 2b). Using (2) and writing Ev as Ev

E0 'E

where 'E ! 0 is assumed, we obtain from (3):

E0 (1 tgM 1/ cosM) 'E .

(4)

Thus, the condition of self-locking has the form

'E / E0 !

cos M sin M 1 . cos M

(5)

, where It can be easily shown that the condition (5) is fulfilled at angles M M M is determined from the equation 'E / E0

cos M sin M 1 . cos M

(6)

is the limiting angle for auto-locking. If M ! M , this process is Actually, M impossible, naturally in terms of the given model and the adopted approximations. 3.2 Partial and superpartial dislocations It is possible that not perfect, but partial dislocations sink into a deep valley and just partial dislocations have the preferential direction. In this case, the development of the partial dislocation in the deep valley requires an additional energy proportional to the area of the stacking fault. Then, instead of (3), the condition of self-locking takes the form ( Ev E0 tgM 1/ 2JdtgM)d ( E0 / cosM)d , (7) where J is the stacking fault energy (per unit area). The condition (5) changes correspondingly:



· 1 § Jd (8) ) 1¸ . ¨ cos M sin M(1 cos M © 2 E0 ¹ by the relationship (6), then at M ! M and any d If we introduce the angle M 'E / E0 !

the inequality

'E / E0

· 1 § Jd ) 1¸ , M ! M ¨ cos M sin M(1 cos M © 2 E0 ¹

(9)

is fulfilled. This means that the condition (8) does not hold in this case and the flip of the similarly to the partial dislocation to the deep valley is impossible at M ! M perfect dislocation considered above. , the condition (8) holds at small values of the segment length d. If M M However, unlike for perfect dislocations, this condition no longer holds for partial dislocations as the length d increases. The segment length d J , at which the condition becomes invalid, is defined by the relationship

'E / E0

· 1 § 1 J d J ) 1¸ , M M . (10) ¨ cos M sin M(1 cos M © 2 E0 ¹

The existence of the physically reasonable solution of the equation (10) for d J at a preset value of the angle M depends on the relationship between 'E and Ja . To demonstrate this, we shall introduce the critical kink length

hJ

d J tgM . Obviously, the condition hJ / a ! 1 should be fulfilled for the

kink to exist in reality. Using hJ , the equation (10) can be rearranged to the form

hJ a

'E / E0 (cosM sin M 1) / cosM , M M Ja / 2 E0

(11)

Let us consider the case when

'E Ja .

(12)

hJ / a 1, M M .

(13)

Then from (11) we have Hence, if the relationship (12) is fulfilled, a physically reasonable solution of the equation (11) is unavailable, i.e. self-locking does not take place. Self-locking is possible only if the condition 'E ! Ja (14) is fulfilled. Moreover, it can be easily shown that the additional condition M min{'E / E0 , M } (15)



57

should be met. Therefore, if the inequality (14) is fulfilled, the extension of segments, whose orientation satisfies (15), becomes possible. The above expressions also hold when a preferential direction exists for superpartial dislocations connected by the APB band. It is assumed as before that kinks are formed independently (inconsistently) in each of the superpartial dislocations making up the superdislocation. Therefore, the nucleation and the propagation of kinks are determined by the relationship between ǻE and the APB energy in the corresponding plane.

4

Examples

4.1 Blocking of a superpartial dislocation located initially in the cube plane (Ni3Al) The initial superpartial dislocation is not splitted and, therefore, recombination is not required. The superpartial dislocation is blocked due to octahedral splitting. Since octahedral splitting is athermal, we have the only process that requires thermal fluctuations, namely the extension of the superpartial dislocation in the preferential direction. The preferential direction is a direction of the type parallel to the line of intersection between the cube and the octahedral plane. The energy gain from octahedral splitting determines ǻE and in the relationship (2) E0 is the constricted dislocation energy and Ev is the splitted dislocation energy. It is easily shown that ǻE can roughly be written as

ǻE # ȥ(ȕ1 , ȕ 2 ) ln ȥ(ȕ1 , ȕ 2 )

d csf Ȗ csf d csf r0

ȥ(ȕ1 , ȕ 2 ) (ln

µ e1e2 (1 Q )s1s 2 2ʌ(1 Q )

d csf 1) r0

(16)

Here Ȗ csf is the energy of a complex stacking fault, d csf is the equilibrium splitting width of the superpartial dislocation, e and s denote respectively the edge and the screw component of the Burgers vector ȕ of the partial dislocation,

r0 is the dislocation core radius, and v is the Poisson ratio. for self-locking is determined, as before, from the The limiting angle M relationship (6). In this case, the energy J should be replaced by the APB energy Ȣc in the cube plane in the inequality (7) and the subsequent expressions containing J . If the relationship

'E ! Ȣca ,

(17)

which is analogous to (14), is fulfilled, the self-locking process becomes possible.


58 Computational Methods and Experiments in Materials Characterisation III Indeed, experiments on no-load heating of Ni3(Al, Nb) single crystals predeformed at a high temperature demonstrated that superdislocations, which initially glided in the cube plane, turned to dislocation barriers [6, 7].

4.2 Blocking of a superpartial dislocation located initially in the octahedral plane (Ni3Al) A superpartial dislocation is blocked due to a series of consecutive transformations, including the cross slip of the superpartial dislocation from the octahedron to the cube plane, and its octahedral splitting. At each stage the superpartial dislocation extends in the preferential direction of the type, which is parallel to the line of intersection between the octahedron and the cube plane of the cross slip. The effective force

K (effb,nc)

providing the cross slip of the superpartial

dislocation has the form [8]:

K (effb ,nc)

bıf (ĮȢ Ȣc) ,

(18)

where Į 1/ 3 and f is the coefficient dependent on the Schmid factors. It readily follows from (18) that the cross slip of a superpartial dislocation is possible at V 0 too. Therefore, self-locking of a superpartial dislocation, which is initially located in the octahedral plane, is possible if two conditions ĮȢ - Ȣc > 0 , (19)

'E ! Ȣca (20) are fulfilled simultaneously. The condition (19) ensures the transfer of a superpartial dislocation to the cube plane. If (19) is fulfilled, the component ĮȢ of the elastic repulsion force in the cube plane due to another superpartial dislocation is larger than the surface tension Ȣc . The condition (20) is responsible for the subsequent extension of the superpartial dislocation along the line of intersection between the cube and the octahedral plane. Indeed, experiments on no-load heating of Ni3(Al, Nb) single crystals predeformed at a low temperature showed that superdislocations, which initially glided in the octahedron plane, transformed to dislocation barriers [6, 7]. 4.3 Blocking of a single dislocation (TiAl) A single dislocation in TiAl is not connected with either the APB band or the stacking fault band, or another dislocation capable of initiating its blocking. The transfer of a single dislocation from a shallow to a deep valley corresponds to the dislocation core rearrangement [9]. Considering the covalent-like character of TiTi bonds, one may think that a screw dislocation with a narrow core has the lowest energy, because such bonds can be restored thanks to a constricted core. The preferential direction is the 800°C where the driving force for austenite decomposition is low. Another realistic alternative to get martensitic microstructure might be forming at low temperatures, such as < 600°C, i.e. below the ferrite regime. In that case, ferrite formation is not happened, although some enhancement of bainite formation may take place. This may not WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

100 Computational Methods and Experiments in Materials Characterisation III be so detrimental, however, due to the notably smaller strength difference between bainite and martensite. Accordingly, the proper temperature range is quite narrow. Concerning to the before mentioned results and discussions, it can be concluded that the deformation in austenitic region will accelerate thermally activated phase transformations and shifts the CCT diagram to the left. Despite the CCT diagram exhibits that at cooling rates higher than 37°C/s, the final microstructure would be fully martensite, but at faster cooling rates, 50°C/s, due to deformation effects, it will be hard to get full martensite microstructure. Although, it is not so easy to distinguish between bainite and martensite in optical microscopy, but hardness measurements can be suitable method to interpret the microstructures. Ferrite is a much softer phase than martensite. It is given a value of 160 HV for ferrite while the martensite has a hardness of >470 HV10. Hardness measurements in fig. 3.a-d confirmed that the microstructure formed after a high temperature plastic deformation has hardness levels between 330-530 HV10. If the hardness of bainite is about 400 HV10, practically some of the secondary phases in addition to the ferrite might be bainite. The other interesting phenomena to be studied are variation of the dilatation values at different forming conditions. Fig. 4 shows the effect of process parameters on the dilatation curves.

0,1

-1

0,0

TiD = 800°C, dH/dt = 0.1s

-1

950°C-15 min

950°C-5 min -0,2

Dilatation

0.12

0.16

0.13

-0,3

950°C-20 min

Hmax = 0.5

950°C-10 min

(a) 100

200

0.02 0.01

-0,2

Hmax = 0.2

-0,3 0.16

-0,4

Hmax = 0.1

0.24

-0,5

0.13

-0,4

300

(b)

-0,6

400

500

600

700

0

100

200

Temperature (°C)

300

400

500

0,1

TiD = 800°C, Hmax = 0.4

0,0

600

700

Temperature (°C)

0,1 -1

0.1 s

-1

0.2 s

750°C 800°C

-1

-1

0.4 s

dH/dt = 0.1s , Hmax = 0.2

0,0

700°C

-0,1

-0,1

-1

1.0 s

-0,2

0.02

-0,3

0.04 0.09 0.15

-0,2 Dilatation

Dilatation

Hmax = 0.4

-0,1

-0,1 Dilatation

TiD = 800°C, dH/dt = 0.1s

0,0

0.07 0.07 0.16 0.24

-0,3 -0,4

850°C

-0,5

-0,4

-0,6

(c)

-0,5 0

100

200

300

400

500

600

700

800

-0,7

(d) 0

100

200

Temperature (°C)

Figure 4:

300

400

500

600

700

Temperature (°C)

The effect of process parameters on the dilatation values during martensitic transformation, a) austenization soaking time, b) max. strain values, c) strain rate and d) initial deformation temperature TiD.



101

It must be pointed that the dilatation term is mainly the magnitude of the plastic deformation which is occurred by the martensitic transformation during cooling. This dilatation is an invariant plane strain which is the combined effect of a uniaxial dilatation and a simple shear. The higher volume fractions of martensite result in higher dilatation magnitudes. The variations of martensite contents and the Ms and Mf(%) temperatures as well as the magnitude of dilatation as numbers at different conditions can be discussed by using the diagrams in fig. 4(a)-(d). 3.1 The role of applied force The main process parameter in stamping process which can be controlled and monitored is applied force. The applied force in stamping process is determined by load cell on punch and therefore it can be assumed as an independent parameter. 600

400 363

300 200

240 200

Ms 300 Mf(%)

Martensite (%)

100 0

400

2

4

6

8

10

12

14

200 100

(a) 0

0,0

16

18

0 20

0.01

-0,2 Dilatation

400

0,2

500 Hardness (HV10)

Temperature (°C)

500

600 HV10

-0,4

0.24

-0,6

Force (KN)

-0,8

without Force

-1,0 0.29 -1,2

(b)

-1,4 0

100

200

Force (KN)

Figure 5:

300

400

500

600

700

800

900

Temperature (°C)

The effect of applied force in austenite region on martensitic transformation parameters, a) hardness, Ms and Mf(%) and martensite content, b) dilatation values.

Hence, the attempts were focused on finding the relationship between applied force and martensitic transformation during simultaneous forming and quenching process. The achievements are represented in figs. 5 and 6. It is seen that regardless of other process parameters like rate and magnitude of deformation and even TiD, applied force during compression and cooling in austenite region results in: The Ms temperature decreased about 40°C through different applied forces between 0-16 KN while, the Mf(%) temperature were raised about 40°C. By applying higher force levels the dislocation density of austenite matrix increases and as a consequence, more activation energy is needed to make martensitic transformation. Hence, the Ms temperature is lowered. The martensite contents and in the same manner the dilatation magnitudes-as is seen in figs. 5.a and b were lowered by increasing the max. applied forces. The microstructure in absence of applied force was fully martensitic whereas in the case of 16 KN applied force, there was only about 30% martensite in microstructure. There is a minimum force limit which yielded fully martensitic microstructure and higher hardness level. As is seen in fig. 5.a, the minimum WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

102 Computational Methods and Experiments in Materials Characterisation III applied force which gave fully martensitic microstructure and higher hardness in comparison with the forceless samples was 6 KN. The hardness in the sample which tolerated 6 KN was 523 HV10 while in the forceless sample was 500 HV10. It might be due to finer martensite lathes in deformed samples. It can be concluded that lower applied forces during hot stamping process result in more successful martensitic transformation and better material properties.

Figure 6:

Microstructure evolution by applying different force levels in austenite region; a) 0 KN, b) 6.1 KN, c) 12 KN and d) 16 KN.

As can be seen in fig. 6.a-d, the higher applied forces yielded on more bainite networks and some ferrite islands as the secondary phase. It can be concluded that by increasing the applied stress level in the austenite region, the nose of bainite in the CCT diagram not only shifts to the left but also to the lower temperatures (due to decreasing of the Ms temperature). It can be interpreted that by applying higher force levels on constant shape or volume of samples, more dislocations will be formed. Accordingly, these forests of dislocations will accelerate the thermally activated phase transformations and therefore, the nose of ferrite and bainite phases shift more and more to the left. It was also seen that the applied force in the austenite region altered the bainite nose more than ferrite zone. That is, to get fully martensitic microstructure in the final sample in hot stamping, cooling rate must be sufficiently higher than the minimum cooling rate which is mentioned in the CCT diagram, i.e. higher than 37°C/s. It can be simply concluded that to get full martensite in the end product, WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


103

cooling rates are strongly dependent on the maximum applied force level. Instead, the higher force levels need higher cooling rates to get fully martensitic microstructure. The above mentioned facts gives the best key points to control martensitic transformation during hot stamping process, because the applied force is the best parameter which can be in hand and in control during the process.

4

Conclusions

The martensitic transformation which occurs during hot stamping process was investigated by means of simultaneous hot compression and cooling in dilatometry machine, resulting in the following conclusions. 1- The Ms temperature is decreased by increasing austenization soaking time. The variation of martensite content is negligible. Due to coarser martensite lathes at longer soaking time, hardness is decreased. Hence, to have a successful hot stamping process, the optimum soaking time is required to get fully fine lath martensitic microstructure. 2- At constant initial deformation temperatures TiD and constant rates, higher deformation levels are ended at lower temperatures and therefore, the possibilities to cross ferrite and bainite noses are raised. Accordingly, the martensite fractions are lowered and as a result lower hardness values are achieved. The decrease in Ms is about 15°C but owing to martensite content reduction, the Mf(%) temperature is increased about 50°C. 3- The higher rates of stamping, the higher temperatures to finish the deformation. At constant deformation magnitudes and constant initial deformation temperatures, higher strain rates results in more martensite fractions and higher hardness. The possibility of the presence of ferrite islands is increased by increasing the rates of deformation. 4- The higher initial deformation temperatures results in higher volume fraction of martensite which is favorite in hot stamping process. It means the time to transfer material from furnace to press must be as short as possible. The 850°C800°C ranges is recommended as the best start temperature to deform the studied steel through hot stamping process. 5- Regardless of rate, magnitude and start temperature of deformation, maximum applied force as an outstanding parameter in hot stamping process can give the best interpretations. The higher forces are applied in austenite region, the martensite volume fractions and the hardness levels are decreased. The Ms temperature is decreased about 40°C by applying 16 KN force to compression samples. In hot stamping process, the lowest possible level of force must be applied to achieve the favorite microstructure. 6- It can be concluded that by applying deformation in austenite region the CCT diagram shifts not only to the left but also to the lower temperatures. Accordingly, the criteria to get full martensitic microstructure in final product through hot stamping process do not follow the CCT boundary conditions and must be cared. Instead, the higher applied force, the faster cooling rate to get fully fine lath martensitic microstructure. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


References [1] [2] [3]

[4] [5] [6] [7] [8]

C. Capdevila, F.G. Caballero and C. Garcia de Andres, Analysis of the effect of alloying elements on the martensite-start temperature of the steels, Materials science and Technology, 2003, Vol.15, no.5, 581-586. P.J. Brofman, G.S. Ansell, The Effect of Fine Grain Size on the Ms Temperature in Fe-27Ni-0.025C Alloys, Metal. Trans. A, 1983, 14A, 1929-1931. M.C. Somani, L.P. Karjalainen and M. Eriksson, Dimensional changes and microstructural evolution in a B-bearing steel in the simulated forming and quenching process, ISIJ International, Vol. 41, No.4, 2001, 361-367. M. Unemoto, W.S. Owen, Effects of austenitizing temperature and austenite grain size on the formation of athermal martensite in an ironnickel and an iron-nickel-carbon alloy, Metal. Trans., 1975, 5, 2041-2053. O.A. Ankara, A.S. Sastri, D.R.F. West, Some effects of austenitizing conditions on martensite formation in an iron-20% nickel alloy, J. Iron and steel institute, 1966, 509-511. S. Denis, E. Gautier, A. Simon, and G. Beck, Stress-phase-transformation interactions-basic principles, modelling and calculation of internal stresses, Mater.Sci.Technol. 1, 1985, 805-814. I. Tamura, C. Ouchi, and T. Tanaka, Thermo-mechanical processing of high strength low alloy steels, Butterworth, London, 1988, 99. C.S. Lee, and W.Y. Choo, Effects of austenite conditioning and hardenability on mechanical properties of B-containing high strength steel, ISIJ Int. 40, 2000, S189-S193.



105

Quantitative assessment of strain and heat treatment on twin formation in commercially pure nickel Q. Li, J. R. Cahoon & N. L. Richards Department of Mechanical and Manufacturing Engineering, University of Manitoba, Winnipeg, Manitoba, Canada

Abstract Thermomechanical treatment comprising cold working to 6% strain followed by annealing at temperatures in the range 700 oC to 1000 oC greatly increases the fraction of special boundaries, primarily 63 type. The accompanying generation of annealing twins is analysed using the accident growth model due to Gleiter and the phenomenological model of Pande et al. It is shown that Gleiter’s model, which contains no scaling factors, when used correctly, predicts the twin density in Cu alloys and pure nickel. The model due to Pande et al also predicts the twin density in Ni but this model incorporates two scaling factors that detract from its generality. Keywords: special grain boundaries, annealing twins, copper alloys, nickel.

1

Introduction

It has been long accepted that thermomechanical treatments can alter the size, shape and crystallographic orientation of grains. In 1984, Watanabe [1] proposed that the type of grain boundaries present could also be controlled sand thus the concept of “grain boundary engineering” was initiated. Grain boundary engineering has been used to increase the fraction of “special grain boundaries”, usually defined in terms of coincidence lattice site (CSL) boundaries. A high fraction of low 6 CSL boundaries {6 ( 29)} has been shown to improve resistance to intergranular crack propagation in Inconel 600 [2, 3] and to substantially improve the elevated temperature crack growth resistance of a nickel-based superalloy [4]. For type 304 stainless steel and alloy 600, high WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070111

106 Computational Methods and Experiments in Materials Characterisation III 6 CSL boundary fractions resulted in reduced crack growth rates. High special grain boundary fractions also resulted in a significant reduction of intergranular stress corrosion cracking in irradiated type 304 stainless steel [5]. It has been clearly demonstrated that high fractions for 6 CSL special boundaries can substantially improve the resistance to creep, fatigue, and environmental degradation of engineering alloys. The purpose of the present investigation was to increase the fraction of 6CSL special boundaries in commercially pure nickel by employing an appropriate thermomechanical treatment. Following treatment, the fraction of special boundaries was determined using orientation imaging microscopy and the annealing twin density was also determined. The experimental values for annealing twin density are compared to the theoretical models of Gleiter [6] and Pande et al [7].

2

Experimental

The material used for this investigation was commercial purity nickel sheet, 3mm thick, mill annealed at 750qC. Work completed by Guyot et al [8] indicated that cold rolling 6% reduction in area would be a good starting point for the investigation. Therefore, strips of the as-received sheet were cold rolled 6% reduction in area and annealed for various times at temperatures from 500qC to 1000qC to obtain a range of grain sizes. Following the anneals, the fraction of special grain boundaries was determined using orientation imaging software EDAX/TSL, version 3.5, on a JEOL 5900 SEM equipped with an ultra thin, Oxford EDS window. Grain sizes and annealing twin densities were determined using the linear intercept method on a Zeiss microscope equipped with a Clemex Vision Image Analysis System. The results from at least 600 intercepts were averaged to obtain the values.

3


The fraction of special boundaries after ten minutes at the annealing temperature is shown in fig 1. Over 80% of the special boundaries were of the 63 type with small fractions of 69, 627 and other types. The effect of annealing time at 700qC on the fraction of special boundaries is given in fig 2 which shows that the optimum annealing time at 700qC is 2.2 X 105 s or about 61 hours (220,000 s). The maximum 75% of special boundaries obtained by annealing at 700qC for 61 hours is the same as the maximum percentage of 75% obtained after annealing at 900qC for 10 minutes (600s). If the formation of special boundaries is diffusion controlled, then the product of Dt, where D is the diffusion coefficient and t is the annealing time, should be similar for the optimum times at the various annealing temperatures. However, the formation of special boundaries could be controlled by either grain boundary diffusion or lattice diffusion so both cases must be considered. For the grain boundary diffusion coefficients, we use the results of ýermák [9] who obtained Dgb = 5.1 X 10-15 exp (-120,000/RT) m2/s where R = 8.3145 J·K-1· mole-1 and T is the absolute WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


107

Percentage of Special Boundaries

temperature. (Actually, ýermák’s results were primarily for the grain boundary diffusion of Co in Ni, but he obtained a few results for the grain boundary diffusion of Ni in Ni that were essentially identical to the values for the diffusion of Co in Ni. Since Co and Ni atoms are very similar, ýermák concluded that the grain boundary diffusion of Ni in Ni was similar to that for the grain boundary diffusion of Co in Ni. Further, ýermák’s results agree well with the earlier results of Wazzan [10].)

80 70 60 50 40 30 20 10 0 500

700

800

900

1000

Annealing Temperature (oC) Figure 1:

Percentage of special boundaries after cold rolling 6% reduction in area and annealing for 10 minutes.

Percentage of Special Boundaries

100 80 60 40 20 0 0.E+00

2.E+05

4.E+05

6.E+05

8.E+05

Annealing Time (sec) Figure 2:

Percentage of special boundaries after cold rolling 6% reduction in area and annealing at 700qC.


108 Computational Methods and Experiments in Materials Characterisation III For 700qC, the grain boundary diffusion coefficient is 1.84 X 10-21 m2/s and for an annealing time of 220,000 s, the product Dt = 4.1 X 10-16 m2. For 900qC, the grain boundary diffusion coefficient is 2.31 X 10-20 m2/s and for an annealing time of 600s, the product Dt = 1.4 X 10-17 m2. For grain boundary diffusion, the Dt values for the two temperatures differ by a factor of about 40. The lattice diffusion coefficient for the diffusion of Ni in Ni is given by Dlatt = 1.82 X 10-4 exp(-285,000/RT) m2/s [11]. For 700qC, the lattice diffusion coefficient is 9.13 X 10-20 m2/s and for an annealing time of 220,000 s, the product Dt = 2.0 X 10-14 m2. For 900qC, the lattice diffusion coefficient is 3.71 X 10-17 m2/s and for an annealing time of 600s, the product Dt = 2.2 X 10-14 m2. For lattice diffusion, the Dt values for the two temperatures are almost identical. Therefore, it appears that the formation of special grain boundaries in commercially pure Ni is controlled by lattice diffusion. The annealing twin densities for annealing temperatures of 700qC, 800qC, and 1000qC are plotted versus grain size in figs 3-5 respectively. Two models were used to calculate the theoretical values for the twin density. The first is due to Gleiter [6] who proposed an atomistic model for the formation of annealing twins. Gleiter proposed that the probability, p, of a {111} plane being a coherent twin plane is given by:

p

° § 'G 0 exp®V z ¨¨ Q kT ln kT °¯ ©

· ª SkTH 2 h 2 ¸¸ / «kTV z 0 Q kT ln 'kTG ¹ ¬«

º ½° »¾ . ¼» °¿

(1)

In eqn (1), Vz is the surface energy of a coherent twin boundary, Q is the activation energy for grain boundary migration, 'G0 is the difference in the Gibbs’ free energy between the growing and the shrinking grain, h is the step height formed by the twin nucleus (taken as the distance between {111} planes), H is the surface energy of a step of height h, and k and T have their usual meanings. The twin density is simply calculated from the product of the number of {111} planes per unit length and the probability that any {111} is a coherent twin plane. Gleiter, noting that kT 180°C. Two subsequent heating cycles from 23 to WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


213

450°C and 23 to 590°C at 20°C/min are applied in nitrogen atmosphere, showing less variation compared to original polyimide through homogenisation under heating. 0

unworn -1

Heat flow (µV)

-2 -3

-4 -5

60°C 80°C 100°C

-6

-7

180°C 260°C

-8 0

100

200

300

400

500

600

Temperature (°C)

Figure 4:

Differential thermal analysis (DTA) of SP wear debris after sliding at 60 to 260°C.

No significant thermal degradation of the SP-1 wear debris is noted: the debris weight loss for each sliding temperature is restricted to 2% during the first heating cycle. The weight loss during the second heating cycle is higher and solely concentrated at 450 to 590°C. The endotherm peak temperature or dehydration temperature Thydrat (maximum dehydration intensity) is determined from fitting curves in Figure 4 and it shifts towards higher temperatures for wear debris relatively to the unworn SP. The dehydration reaction of wear products depends on sliding temperature and normal loads: there is a general trend that the dehydration temperature increases for high sliding temperatures while this trend is stronger for high normal loads; at low loads, however, a critical bulk temperature of 180°C must be exceeded to increase the dehydration temperature. The upward shift in dehydration temperature of wear debris indicates chemical changes after sliding, such as imidisation, that cause a delay in dehydration. After formation of polyimide networks modified by sliding, higher temperatures or activation energy is needed for dehydration. Nevertheless, this process remains reversible during subsequent heating-cooling-heating cycles. According to Nagai [4] it is known that polyimides are sensitive to water absorption, but its relation to polyimide structures that are modified by wear was not yet illustrated.


214 Computational Methods and Experiments in Materials Characterisation III Table 1:

TGA weight loss and DTA endotherm peak position of SP wear debris after sliding at 50 and 200 N, 0.3 m/s at high temperature. 50 N normal load sliding test

Bulk temperature (°C)

Weight loss (%) st

1 heating*

nd

2 heating*

Thydrat (°C)

200 N normal load sliding test Weight loss (%) 1st heating*

2nd heating*

Thydrat (°C)

original

0.72

13

182

0.72

13

182

60°C

0.64

16

204

1.45

32

200

100°C

1.51

18

199

1.82

32

220

180°C

1.59

18

192

0.94

31

222

220°C

1.15

18

197

0.23

30

228

260°C

1.20

18

200

0.20

30

230

* First heating between 23 to 430°C, second heating between 23 to 590°C

4

Thermo-analytical analysis of thermoplastic wear debris

The variations in the structure of thermoplastic polyimide wear debris are more important than for sintered polyimides and can be well-correlated to different transitions in friction and wear. It mainly indicates thermal changes in the amorphous phase through crystallisation and/or cross-linking. The shift in dehydration temperature at 180°C as noted for sintered polyimides is less clear for thermoplastics through interference with complex phase changes. Thermo-analytical DTA measurements of TP debris under similar conditions to SP are presented in Figure 5, during a first heating step (23 to 450°C) and a second heating step (23 to 590°C) at 20°C/min. The corresponding TGA curves are given in Figure 6. The original TP has clear transition temperatures, but they smoothen for wear debris. Chemical reactions progressively change the thermoplastic polyimide structure into the properties of sintered polyimides that lack transitions and/or melting. Both sintering and/or sliding are thus considered as a ‘thermal treatment’ altering the polymer structure. The first heating step in DTA (Figure 5(a)) shows that the glass transition temperature Tg increases or finally disappears when it becomes smoothened over a broad interval from 230 to 280°C. The glass transition temperature Tg is representative for linear molecular structures in an amorphous ordering and shifting or disappearance indicates that the amorphous zone is affected by crosslinking into the formation of a more stable and ordered structure after sliding at high temperature. This evolution is also reflected in the disappearance of a crystallisation peak Tc. The melting exotherm Tm decreases in intensity and its maximum value increases from 387°C (unworn) to 391°C (after sliding at 140 to 180°C) in parallel to the formation of crystalline and strongly cross-linked structures. The recrystallisation peak during cooling disappears because crystal WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


215

0

(a)

Tc

unworn

Heat flow (µV)

-2 -4

Trex

-6

Tg Tm

-8 -10

< 100°C 120°C

140°C 180°C 220°C 260°C

-12 -14 -16 0

100

200

300

400

500

Temperature (°C) 0

(b)

unworn Tc

-2

Heat flow (µV)

-4

Tg

-6

Tm

-8 -10

< 100°C 180°C 120°C 140°C 220°C 260°C

-12 -14 -16 0

100

200

300

400

500

600

Temperature (°C)

Figure 5:

Differential thermal analysis of TP wear debris after sliding at 60 to 260°C, (a) first DTA heating cycle, (b) second DTA heating cycle.


216 Computational Methods and Experiments in Materials Characterisation III nucleation lacks in the modified amorphous phase. The conclusion that the amorphous phase modifies under sliding agrees to visual observations that debris particles change from transparent to opaque yellow colour. Debris becomes brittle (and consequently acting more abrasive) after sliding in parallel to the behaviour of sintered polyimides. The second heating step in DTA (Figure 5(b)) shows that the structural modifications of the amorphous phase are partially reversible after a first heating step. The transition zones in tribological properties of TP (Figure 1(b)) are compared to wear debris evaluation and parallel transitions in thermo-analytical analysis are found: x

After sliding at 100 to 120°C (increasing friction), dark coloured particles indicate chemical degradation by hydrolysis. This is confirmed by TGA analysis showing lowest thermal stability for 120°C wear debris.

x

After sliding at 120 to 180°C (decreasing friction), flake-like particles indicate an increase in polyimide strength by imidisation. This is confirmed by TGA analysis showing high thermal stability of those particles. The DTA analysis also indicates crystallisation or cross-linking of debris particles after 120 to 180°C sliding tests, although at lower temperature relatively to unworn TP.

x

After sliding at 220 to 260°C (increasing friction), rolled debris particles indicate melting. As those particles are rapidly removed out of the sliding interface, TGA analysis shows high thermal stability.

5

Thermo-analytical mass spectroscopy (MS) of sintered wear debris

The emission of gaseous species during thermal decomposition of SP wear debris is analysed with a mass spectrometer coupled to DTA/TGA measurements. One single heating step from 23 to 600°C at 20°C/min is applied, using argon carrier stream. It will demonstrate that the instrumentation is sensitive to detect water volatilisation (only 1% weight loss) and confirms that the endothermic reaction at 180°C is surely related to dehydration. Other degradation products are characterised and have lower intensity, which further decreases when the sliding temperature was higher. Some spectra for SP wear debris after sliding at 50 N, 0.3 m/s and 100, 180, 260°C, are given in Figure 7. Spectra represent the volatilisation intensity for a specific emission product as a function of the heating temperature. Each decomposition product is characterised by its atomic mass unit (a.m.u.) and intensities are related to the ion current (A) in the mass detector (scaled to sample weight), which is proportional to the concentration of decomposition product in the carrier stream.



217

105

220 to 260°C

1 Residual weight weight (%) Residual (%)

100

2

3

95

140 to < 100°C 180°C 120°C

90 85 80 75 70 0

100

200

300

400

500

600

Temperature (°C) Temperature (°C)

Thermogravimetric (TGA) analysis of TP wear debris after sliding at 60 to 260°C over a heating cycle 1, cooling cycle 2, heating cycle 3.

(a)

-9

6.0 10

-10

-9

5.0 10

-9

4.0 10

-9

3.0 10

-9

2.0 10

-9

1.0 10

Intensity a.m.u. = 18 (water)

5.8 10

-10

5.7 10

-10

5.6 10

-10

5.5 10

-10

5.4 10

-10

5.3 10

Wear debris 100°C

-9

0

5.2 10

0

0.0 10 100

200

300

400

500

600

Temperature Temperature(°C) (°C)

(b)

5.0 10

-11

4.0 10

-11

3.0 10

-11

2.0 10

-11

1.0 10

-11

0.0 10

0

Intensity a.m.u. = 30 (nitrogen

Wear debris 180°C

Wear debris 260°C Wear debris 100°C

0

100

200

300

400

500

600

Tem perature (°C) (°C) Temperature

Figure 7:

Mass spectroscopy for sintered polyimide wear debris.


Intensity a.m.u. = 16 (oxygen)

Figure 6:

218 Computational Methods and Experiments in Materials Characterisation III Most important decomposition fraction for tribological performance is the production of water (a.m.u. = 18, Figure 7(a)). Water volatilisation occurs at 180 to 200°C heating temperatures as a peak intensity for the debris samples after 100°C sliding or a maximum intensity for debris samples after 180 and 260°C sliding (not shown). Water volatilisation at 200 to 600°C heating temperatures depends on the sliding temperatures, showing: (i) a constant or increasing tendency of water volatilisation for 100°C debris indicating progressive imidisation during heating, and (ii) a decreasing trend for 180 to 260°C wear debris indicating that debris is more inert by imidisation during sliding. The peak in water emission for high temperature wear debris broadens through variations in structure (and molecular weight) after wear. Water volatilisation agrees to the previously noted shift of dehydration temperature Thydrat in DTA thermographs, rising for debris after 180 to 260°C sliding. A second small peak in a.m.u. = 18 intensities for debris after 100°C sliding occurs at 380°C and corresponds to small weight loss. The oxygen intensity (a.m.u. = 16) is plotted over the intensities of water volatilisation in Figure 7(a) and confirms that dehydration or water condensation is responsible for the noted endothermic reaction at 180°C. The nitrogen-monoxide NO fraction (a.m.u. = 30, Figure 7(b)) has similar features to nitrogen N fraction (a.m.u. = 14). It is firstly stressed that the concentration of released nitrogen is a factor 10-2 to 10-3 smaller than water concentrations. The decomposition of polyimide wear debris into nitrogen monoxide is postponed and finally disappears with increasing sliding temperatures through formation of a strong imide structure after sliding.

6

Conclusions

For sintered polyimides, hydrolysis and imidisation reveals from thermoanalytical analysis of wear debris and explains transitions in friction and wear. For thermoplastic polyimides, crystallisation and melting effects are most important.

References [1] Jacko M.G., Tsang P.H.S., Rhee S.K., Wear debris compaction and friction film formation of polymer composites, Wear, 133, pp. 23-38, 1989. [2] Sharf T.W., Singer I.L., Monitoring transfer films and friction instabilities with in situ Raman tribometry, Tribology Letters, 14, pp. 3-8, 2003. [3] Tewari U.S., Bijwe J., Tribological behaviour of polyimides. Polyimides, ed. M. Dekker, Marcel Dekker: New York, pp. 533-583, 1996. [4] Nagai N., Hironaka T., Study of interaction between polyimide and Cu under high humidity condition, Applied Surface Science, 171, pp. 101-105, 2001.



219

Analysis of adiabatic heating in high strain rate torsion tests by an iterative method: application to an ultrahigh carbon steel J. Castellanos1, I. Rieiro1, M. Carsí2, J. Muñoz1 & O. A. Ruano2 1

Department of Mathematics, University of Castilla - La Mancha, Spain Department of Physical Metallurgy, National Center for Metallurgical Research (C.E.N.I.M.), Spain

2

Abstract An iterative algorithm has been developed to establish the adiabatic heating correction of flow curves for torsion tests of an ultrahigh carbon steel containing 1.3% C. High temperatures (1223 to 1473 K) and high strain rates (2, 5, 10 and 26 s-1) were used. The curves are corrected in a finite and discrete set of strain data by means of parametric derivatives and integration on the initial curve without correction. The process is repeated until the termination tolerance for the stress is less than 10-2 MPa. Usually, four iterations are needed to reach this tolerance. The corrections are bounded by the maximum of mechanical energy available to be converted into heat. The corrections are carried out until a true strain H 4 in order to avoid the effects of flow localization in the material. Keywords: adiabatic heating, torsion test, modelling, simulation, Garofalo equation, hot working.

1

Introduction

Torsion tests at high temperatures and strain rates of materials usually show a strong increment of temperature during the test above the programmed temperature that is attributed to adiabatic heating [1]. The temperature correction due to adiabatic heating has been discussed in various works [1–6]. The following expression is usually considered: 'Ti (H )

K. A U C

H

V H dH ³ H max


(1)

220 Computational Methods and Experiments in Materials Characterisation III where T is the test temperature, C is the specific heat capacity, U is the density, K and A are efficiency coefficients of the energetic conversions and V (H ) is the stress-strain relation. Some authors assume a variable energy performance in eqn (1) [2] or a constant one [6]. Other authors use the relation V (H ) without considering the intrinsic error due to the adiabatic heating itself [3]. In general, it is not considered that determination of the term V (H ) implies derivatives at constant temperature, which is not true under the effect of adiabatic heating. These derivatives appear in the calculation of the strain rate sensitivity and the strain hardening coefficients. In addition, constant values for U and C are used in the entire working range. In this work, we consider the following expression for determining the true value of the corrected relation V c , in contrast to the experimental value V wc :

V ic (H ) V iwc H

. wV iwc (H , H , T ) 'Ti (H ) '[ wT H ,H

(2)

where '[ is the error associated to the experimental value of the stress that should be bounded to avoid wrong answers. The goal of this work is to design a modular and iterative algorithmic method that guarantees the convergence of the experimental function V iwc (H ) to the nominal function V ic (H ) . This method is based in eqns (1) and (2). The validity ranges of the algorithms are adjusted taken into account physical fundaments on flow localization [3,5] and bounds of the performance for the conversion on mechanical energy into heat.

2

Material and experimental procedure

The UHC-1.3%C steel studied in this investigation has the following composition: 1.3% C, 0.5% Mn, 0.6% Si, 0.18% Cr and balance Fe [1]. The manganese was added to neutralize the deleterious effects of sulphur and phosphorus. The steel was obtained at Sidenor Industry as a cast of 8 litres by means of an induction furnace. The as-cast ingot was initially soaked at 1050ºC and forged into a bar of 60 mm x 55 mm cross section. Simulation of the forming process of forged parts was carried out by means of torsion tests. An induction furnace heats the test sample and the temperature is continuously measured by means of a two-color pyrometer. A silica tube with argon atmosphere ensures protection against oxidation. A helium atmosphere is used to obtain, after testing, a cooling rate of 325 K/s. The torsion samples have an effective gage length of 17 mm and a radius of 3 mm. The density and specific heat are 7800 kg·m-3 and 670 J·kg-1·K-1, respectively. Strain rates in the range 2 to 26 s–1 were used. The temperature WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


221

range was 900 to 1200 ºC. The samples were deformed in a SETARAM torsion machine at CENIM (National Center for Metallurgical Research) in Madrid, Spain.

3 Theoretical approach Two main processes limit the conversion of mechanical energy into heat in an adiabatic framework: changes in the internal energy of the material and flow localization. Both processes are related to the start of catastrophic failure [2–4]. A differential expression for the first law of thermodynamics V dH U c dT du mst where du mst is the variation of the microstructural internal energy can be considered. The plastic work carried out by the material is transformed into heat that is used to increase the internal energy of the material. Some authors assume du mst 0 [3]. Other authors consider du mst z 0 leading to the general expression [2]: dT

V dH § 1 du mst · ¨1 ¸ U c © V dH ¹

V dH K H , H, T U c

(3)

where K H , H, T 1 (1 V ) >du mst dH @ is the performance of the conversion and it is variable. A constant value for K of 0,90 or 0,95 may be taken but an iterative procedure would be necessary to eliminate the effect of this approximation. The approach of Prasad et al. [4] is convenient to estimate the upper limit bound of the increase of temperatures due to adiabatic heating. A simple constitutive equation for the energy dissipation is the following: H

V H

³

V

V dH H dV

0

³

GJ

(4)

0

where G is the dissipator content and J is the dissipator co-content. Part of the power dissipated by the plastic flow, G, can be converted into heat. The quantity J is related to the processes of form change. The limit for G is Gmax (V max Hmax ) 2 . The following expression can be deduced from eqn (4): K G max (5) 'Tmax U C V that represents the limit for the adiabatic 'T in a volume V. The increment in temperature can be expressed as [3]: Hp'

'T

K V dH U c

³

H0


(6)

222 Computational Methods and Experiments in Materials Characterisation III where K is constant and H p ' is the deformation limit where the plastic instability starts [3]. Therefore, a critical deformation can be considered above which it is not possible to apply this kind of corrections [5]. Values of dT dH 165 K from eqn (6) are obtained for the UHC-1.3%C steel. Assuming the analysis of Armstrong et al. [3] and considering the stressstrain relation V K1 .H T , together with the definition of the stress exponent, n, in the Garofalo equation and the definition of constant strain rate tests, H p H p t , the condition for plastic instability gives the following equation: ª dT º « » «¬ d H p »¼H p '

1· § ¨ T ¸ (H p ' E ) n¹ ©

(7)

Under stability conditions, dV ! 0 . The instability starts at dV be expressed as [7, 8]: dV

0 , that can

§ wV · § wV · § wV · dH ¨ dH ¨ dT ¨ ¸ ¸ ¸ w w H H © ¹ H , H © ¹ H , H © wT ¹ H , H

0

(8)

Under adiabatic heating conditions, and assuming H constant, by means of Q § · ¨ ¸ R the Garofalo equation ¨ H A.e T sinh(D V ) n ¸ it is obtained [7] that ¨ ¸ © ¹

wV

wH H,T

T V H

and

wV

wT H ,H

V Q R T 2 .

Taken

these

expressions into the plastic instability condition, an expression for the flow stress at which the instability starts, V in , can be obtained:

V in

T R T 2 U c K Q H

(9)

This expression will be applied later to the UHC-1.3%C steel.

4

Basic methodology

The following assumptions are used in the algorithm developed in this work: 1) adiabatic conditions in the deformation process, 2) U and C do not vary with T, 3) K and A are constant with strain, and 4) adiabatic heating has an important effect from the peak stress of the curve V (H ) , to a value H f . For a given test at constant H j and Ts , the initial temperature, and for a given value of H , the uncorrected stress, V 0 wc (experimental stress), can be expressed as a function of the corrected stress, V 0c , as: WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


V 0 wc (T ) V 0c (T 'T )

223 (10)

Applying Taylor expansion of the function V 0c in eqn (10) about the point T:

V 0 c (T ) V 0 wc (T )

wV 0 wc (T ) 'T wT

(11)

where it is assumed that wV 0wc wT | wV 0c wT . On the other hand, the temperature increment for each measured strain H i , for a given test at constant H j , can be expressed as: 'T0 H i

KA U C H

Hi

0

³H

V 0wc H , H j , T d H

(12)

max

Substituting eqn (12) in eqn (11):

V H i , H j , T V 0wc H i , H j , T c 0

wV 0wc H i , H j , T K A H i ³ V 0wc H i , H j , T d H (13) wT U c H 0 H max

A single time application of these equations would result in a value of V H , H, T that is not accurate. This is due, as mentioned in the introduction, to c 0

the associated error in the determination of V 0 wc and consequently in the integral part of eqn (13). Furthermore, calculation of wV 0wc H , H, T wT in eqn (13) is also not accurate since the function V 0 wc H is a warped curve in the space

^H , V , T ` . To minimize the inaccuracies of the calculation carried out in eqn (13), an iterative algorithm based in eqns (11) and (12) has been developed using as initial value the stress V 0c H , H, T . It is a modular algorithm in three steps. In the first step, the value of 'Tk H i in the iteration k (from k=1 to the number of iterations) is calculated according to the following expression: Hi

'Tk (H i )

KA V wc V kc1 d H U C H ³ k 1

(14)

max

where V kwc1 , is the uncorrected stress in the iteration k-1, and V kc1 is the corrected stress in k-1. The integration interval is divided in sufficiently enough small parts. In this work, we have worked step by step with all the output data WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

224 Computational Methods and Experiments in Materials Characterisation III given by the machine. Therefore, a trapezoidal rule is used in order to compute the numerical integration in eqn (14). In the second step, the partial derivative with respect to T of the uncorrected stress, wV kwc H i , H j , T wT in the iteration k, is calculated by means of the H ,H

expression: wV kwc Ts , H i wT

#

V kwc Ts 1 , H i h V kwc Ts , H i h Ts 1 Ts

(15)

where s 1 refers to the test conducted at the same strain rate but at a temperature Tk+1, belonging to the temperature set ^Ts `s 1, N , that is next in the ascendent sequence. This approximation is good enough since the discretization intervals are small. Finally, the value of the corrected stress in the iteration k is given as:

V kc H i , H j , T V kwc H i , H j , T

wV kwc H i , H j , T K A 'Tk (H i ) d H wT Uc

(16)

The criterion adopted for stopping the algorithm, i.e. the termination tolerance, is 'V d 10 2 MPa for a given control strain. By means of this procedure, the final measured temperature is reached at a given iteration for a value H but the correction is used only up to H f , a value at which the flow localization is not considered important to distort our correction.

5

Results and conclusions

5.1 Analytical basis The results obtained in section 3 for the bound limits are applied in our model to establish the adiabatic correction of the UHC-1.3%C steel. For this steel, at H 10 s 1 and T=1323 K, the peak stress is V max 100 MPa and, according to eqn (5), 'Tmax 172 K . For H 2 s 1 and T=1323 K, 'Tmax 24 K . This gives an idea of the upper bounds of the uncorrected values. Using the data of Castellanos et al. for this steel [9], the relation V (T ) K 2 e 0,0037T is obtained. The values of the Garofalo equation are: Q=274,3 kJ/mol and n=4,66. A value T | 0.2 is obtained for an integration on all deformation paths. For comparison, it was obtained for H 1,5 , T | 0.11

and dT (H 'p ) d H p dT (H 'p ) dH p

28 H p ' K

and for H

next to the peak, T | 0 and

58 H p ' K .



225

Table 1 shows the limit strains, H 'p , for the start of plastic instability or flow localization for the UHC-1.3%C steel. The values were calculated by means of eqn (9) using T

0,90 . The values H 'p are determined in the

0.2 and K

point where the experimental curve V (H ) cross the curve V in H . The table shows that the plastic instability is delayed at high temperatures and low strain rates. It is worth noting that the local instability develops progressively with plastic strain. At H =5 a clear change of behavior of the flow curves is observed characterized by oscillations of the derivates of the stress with respect to the strain. A value of H f 4 was chosen because up to this value the corrections were meaningful. Table 1:

Limit strains, H 'p , for the start of plastic instability for the UHC1.3%C steel as a function of strain rate and temperature.

T(K) 1223 1273 1323 1373 1423 1473

from eqn (9)

V in V in V in V in V in V in

H

2 s 1

H

5 s 1

H 10 s 1

H

26 s 1

52.7 / H

0.45

---

0.3

0.3

57.1 / H

0.6

0.5

0.45

0.4

61.7 / H

0.8

0.65

0.55

0.45

66.4 / H

1.05

0.8

0.7

0.5

71.3 / H

1.5

1.2

0.9

0.7

76.4 / H

1.9

1.5

1.1

0.9

5.2 Correction of flow curves for the UHC-1.3%C steel The flow curves of the UHC-1.3%C steel have been modified to consider the adiabatic heating. The curves were conducted at H 2, 5, 10, and 26 s-1 and T from 1223 to 1473 K, with a variation of 50 K. A maximum of four iterations were conducted for the attainment of the final measured temperatures for H >6,8@ although the corrections were carried out up to H f 4. Figure 1 shows true stress vs. true strain curves at various strain rates and temperatures for the UHC-1.3%C steel. The solid lines represent the correction for adiabatic heating according to eqn (12). The corrections agree with those carried out by other authors [10, 11]. However, somewhat different results were obtained when compared with other investigations where unreliable approximations were conducted [12, 13]. Figure 2 shows the evolution of 'T , according to eqn (11), with strain for H 26 s 1 at various temperatures. All the temperature increments are inside the bounds established for the maximum increments. The convergence of the iterative algorithm was reached at a maximum of four iterations. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


Figure 1:

Figure 2:

Flow curves for the UHC-1.3%C steel. Solid lines are corrected curves for adiabatic heating and dotted lines are uncorrected curves.

Evolution of 'T with strain for several T0 ' s at a H =26 s-1.



Table 2:

H

Values of 'V at H (a) 2 s 1

H 5 s 1

1 (a) and H

3 (b) for the corrected tests.

1223 K 5.38

1273 K 2.06

1323 K 1.91

1373 K 1.85

1423 K 0.9

1473 K 0.9

---

3.68

2.69

1.39

1.15

1.02

1

5.44

2.17

2.62

2.34

0.86

0.99

1

(b) H 2 s 1

6.54 1223 K 12.16

2.76 1273 K 6.21

2.49 1323 K 5.90

2.37 1373 K 5.70

1.41 1423 K 3.32

1.28 1473 K 2.79

H 5 s 1

H 10 s H

26 s

H 10 s H

---

8.92

9.25

6.08

3.08

2.67

1

14.11

8.46

6.32

7.57

3.48

3.29

1

16.81

9.4

7.73

7.78

7.32

6.23

26 s

227

Table 2 shows a summary of all the results obtained in this work. The accumulated values of 'V are given for each pair ^H, T ` at H 1 and 3. Values at H 5,6 are higher but were not considered due to flow localization. It can be concluded that the method, and the implemented algorithm, that we have developed in this work is reliable and convergent. The corrected stressstrain curves are efficient and reliable and take all the experimental data set without the need of average approximations. In addition, the method provides the detailed corrections at the discretization level given by the machine. The main conclusions of this work are: 1. A new iterative approach for the adiabatic heating correction for torsion tests has been established. It is a natural generalization of a previous approach where the correction was carried out in a single run. 2. The new approach brings an improvement in the precision of the corrected flow curves. The relative errors associated to determination of the experimental stresses are minimized. 3. The temperature increments obtained for the UHC-1.3%C steel are inside the bounds established for the maximum increments due to adiabatic heating.

Acknowledgement The work was carried out through the Project PBC-05-010-1 from JCCM (Castilla-La Mancha, Spain).

References [1]

Fernández-Vicente, A., Carsí, M., Peñalba, F., Carreño, F. & Ruano, O.A., Deformation behavior during hot torsion of and ultrahigh carbon steel containing 1.3 wt.% C. Zeitschrift für Metallkunde, 94(8), pp. 922929, 2003. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

228 Computational Methods and Experiments in Materials Characterisation III [2] [3] [4]

[5] [6] [7] [8] [9]

[10] [11] [12] [13]

Pantleon, W., Francke, D. & Klimanek, P., Modelling adiabatic heating during high-speed deformation. Computational Materials Science, 7, pp. 75-81, 1996. Armstrong, R.W., Coffey, C.S. & Elban, W.L., Adiabatic heating at a dislocation pile-up avalanche. Acta Metallurgica, 30, pp. 2111-2116, 1982. Prasad, Y.V.R.K., Gegel, H.L., Doraivelu, S.M., Malas, J.C., Morgan, J.T., Lark, K.A. & Baker, D. R., Modelling of dynamic materials behavior in hot deformation: Forging of Ti-6242. Metallurgical Transactions A, 15A, pp. 1883-1892, 1984. Lindholm, U.S., Mechanical Properties at High Rates of Strain. Conference Series nº 21, ed. J. Harding, Institute of Physics: London and Bristol, pp. 3-21, 1974. Bhattacharyya, A., Rittel, D. & Ravichandran, G., Strain rate effect on the evolution of deformation texture for D -Fe, Metallurgical and Materials Transactions A, 37(A), pp. 1137-1145, 2006 Semiatin, S.L., Staker, M.R., & Jonas, J.J., Plastic instability and flow localization in shear at high rates of deformation. Acta Metallurgica, 32 (9), pp. 1347-1354, 1984. Staker, M.R., The relation between adiabatic shear instability strain and material properties. Acta Metallurgica, 29, pp. 683-689, 1981 Castellanos, J., Rieiro, I., Carsí, M, Muñoz, J., Ruano, O.A., Analysis of several methods for the data conversion and fitting of the Garofalo equation applied to an ultrahigh carbon steel. Journal of Achievements in Materials and Manufacturing Engineering, 18(1-2), pp. 447-454, 2006. Wei-Guo, G., Nemat-Nasser, S., Flow stress of Nitronic-50 stainless steel over a wide range of strain rates and temperatures. Mechanics of Materials, 38, pp. 1090-1103, 2006. Zhou, M., Clode, M.P., Thermal analysis of the torsion test under hotworking conditions. Computational Materials Science, 9, pp. 411-419, 1998. Holzer, A.J. & Wright, P.K., Dynamic plasticity: a comparison between results from mechanical testing and machining. Materials Science and Engineering, 51, pp. 81- 92, 1981. Venugopal, P., Venugopal, S. & Seetharaman, V., Some aspects of the dependence of the flow curve of commercially pure titanium on the forming temperature and the strain-rate. Journal of Materials Processing Technology, 21, pp. 201-217, 1990.


Section 7 Experimental methods – mechanical characterisation and testing



231

Collapse of FRP/syntactic foam sandwich panels M. Perfumo1, C. M. Rizzo2 & M. P. Salio2 1

Cantieri SANLORENZO S.p.a., La Spezia, Italy Department of Naval Architecture and Marine Technologies (DINAV), Genoa University, Italy

2

Abstract In the framework of a wider research project, large scale testing of composite sandwich panels has been carried out at the DINAV shipbuilding laboratory. The skins of the sandwich are made of fibre glass epoxy prepreg and the core consists of a syntactic epoxy foam. Strain gages have been bonded on the outer skins and also located in between the core and the skins. The captioned material is currently used for small components of naval ships (e.g. shields, stanchions, etc.) either in single skin laminates or sandwiches: the final goal of the project is to study its applicability in building pleasure craft hulls, taking advantage of its high strength. The large scale tests have been completed by usual testing on small scale specimens, according to well-known international standards and analytical and finite elements (FE) numerical models have been calibrated with the experimental data. Different options of FE codes have been investigated in order to catch their capabilities and approximations in modelling the composite material and their damage up to collapse. Some advice on the behaviour of quite large sandwich panels is reported, highlighting the effects of the size of the structure on the material mechanical properties. Keywords: FRP, prepreg, syntactic epoxy foam, composite sandwich, laminates, mechanical tests, large scale tests, numerical simulation (FEM).

1

Introduction

Composite sandwiches are commonly adopted in marine and aeronautical engineering for structures or structural elements requiring high stiffness and strength, mainly to flexural loads, together with low specific weight. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070231

232 Computational Methods and Experiments in Materials Characterisation III This paper presents the main results of an experimental and numerical study on the mechanical behaviour of a type of sandwich currently used for small components of naval ships (e.g. shields, stanchions, etc.). The external facings of the sandwich (skins) are prepreg glass-fibre/epoxymatrix composites whereas the central part of the sandwich (core) is a syntactic foam consisting of hollow glass microspheres embedded in an epoxy resin matrix. The final goal of the project is to study the applicability of such material in building entire hulls of pleasure craft, taking advantage of its high strength. It is remarked that prepregs have very high mechanical properties, also against fatigue and shock and syntactic foam is a core fabric with superior physical properties, (Greene [1]). Another significant advantage concerns prepreg low environmental impact, with no styrene emission. In fact, more and more reducing VOC (Volatile Organic Content) requirements force builders to look for alternative construction methods; it is therefore expected that demand will drive more prepreg manufacturers towards the development of products specifically suited for the marine industry. Other distinct advantages are ease of handling and excellent resistance against water, seawater, oil and hydrocarbons, (Greene [1]). Main advantages of the syntactic foam adopted are lightweight, high resistance against stability loss due to compression, quite high strength against impact loads. An attractive option for structural optimization seemed to limit the stiffening of the shell plates using sandwich panels and gradually varying the lamination sequences of the skins and of the core thickness in the different hull areas, according to loads demands. Design of such structures needs a reliable and quite precise numerical model of the whole hull shell. Therefore, analytical and numerical finite elements (FE) models have been studied as well. The mechanical characterization of this highly heterogeneous material (or rather, structural element) has been carried out at the Department of Naval Architecture and Marine Technologies (DINAV), Genoa University, with the collaboration of Centro Tecnologico Sperimentale S.r.l., La Spezia for small scale testing and Nuova Connavi S.r.l. for experimental data about the syntactic foam, through the following sequence of steps: (a) experimental testing on small specimens of the material adopted for the skins; (b) collecting data about the syntactic foam material adopted for the core; (c) experimental testing of the sandwich panels, both on large and small scale; (d) development of analytical and numerical FE models calibrated with the experimental data, firstly simulating the small scale tests, then the large scale ones. The paper is organised as follows. In Section 2, the sandwich under study is fully described. Section 3 is devoted to the construction of the numerical model with reference to the theoretical formulations used and the judgement of their applicability. The numerical simulations of tests carried out on small scale specimens and the description of large scale tests together with relevant results are presented respectively in Section 4 and Section 5. Lessons learned are briefly resumed in Section 6. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


2

233

The sandwich under study

The FRP/syntactic-foam sandwich under study was manufactured by Nuova Connavi s.r.l. (Italy). The sandwich structure is represented in Figure 1. The materials adopted for the skins are called EPREG UD 52TM and EPREG DIAG 43TM and are prepregs obtained by impregnation with an epoxy resin system of an E-glass tissue. EPREG UD 52TM is a unidirectional composite with 97% of fibres oriented longitudinally and 3% transversally whereas EPREG DIAG 43TM is bidirectional and has ±45° fibres.

Skins: prepreg composite

Figure 1:

Core: syntactic foam

The sandwich under study.

The syntactic foam core, whose trademark is EFOAMTM, is assembled with the same epoxy matrix as EPREGTM which embeds hollow air-filled glass microspheres, mixing resin and hardener under vacuum and by adding microspheres repeatedly until full homogenization. Bubbles have an average diameter of 70 mm and an average wall thickness of 0.58 mm. The density of the resulting syntactic foam averages 0.53 g/cm3 (see [2] for all details).

3

Material modelling

To analyze a sandwich structure, many challenging issues need to be addressed such as the complexity of the mechanical interactions between material constituents, particularly when applied loads produce local damage and sequential failure. The mechanisms of failure in FRP sandwich structures are entirely different from that of conventional steel structures. Static/dynamic failure involves matrix cracking, fibre buckling and rupture, and layer delamination in an interrelated manner. The complexity of the mechanical response of FRP sandwich structures presents great difficulties in predicting reliably composite’s performance, nevertheless, finite element method (FEM) is becoming a very popular and powerful tool for simulating an engineering system. After a preliminary study of a few commercial finite element codes, the software ANSYS® has been adopted for all the numerical simulations performed. This code allows to model composite materials with specialized elements called layered elements. Several formulations are available: linear and nonlinear, shell and solid, with different capabilities. SHELL91 and SHELL99 in particular have been used because fitting better the material under study. SHELL 91 is an WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

234 Computational Methods and Experiments in Materials Characterisation III 8-node, nonlinear, layered element with 6 degrees of freedom at each node that supports plasticity and large-strain whereas SHELL 99 is an 8-node, linear, layered element, without the nonlinear capabilities of SHELL91. Each of these shell elements is shear deformable and allows failure criterion calculations, [3]. The first input required within the software is the definition of the layered configuration, obtained by specifying, layer-by-layer, ply thickness, ply orientation and material properties. To this aim, being the sandwich skins assumed made of an orthotropic material, the widely known micromechanics formulations have been applied, by superimposition of elementary layers. These equivalent layers have unidirectional fibres and are characterized by the same content of reinforcement as a given layer, whatever the type of reinforcement used. In order to determine the elastic characteristics of that equivalent layer, classical rule-of-mixtures equations for longitudinal moduli and modified equations for transverse and shear moduli have been then used, (Tsai and Hahn [4]). It is remarked that similar formulations are adopted within the HSC Code, [5], whereas semiempiric formulations are adopted by Class Society, estimating average properties but not accounting for fiber orientation, lay-up method (e.g. manual, prepreg or infusion), stacking sequence, etc. The material used for the sandwich core has been considered as homogeneous and isotropic. Failure analysis has been carried out as well, using the capabilities of the software adopted. Within ANSYS®, possible failure of the material can be evaluated by up to six different criteria, of which three are predefined (max strain, max stress and Tsai-Wu). In this study, two failure criteria were examined, max stress and Tsai-Wu, but, since a complete analysis of the sequential collapse is quite difficult to be implemented in the ANSYS® environment, this tool has been used to determine only the first ply failure, leaving to further developments of the research the automatization of the procedure for the progressive failure. Concerning the sandwich core, Drucker Prager criterion has been considered, supported by the code as well. The elastic properties for the materials under study are presented in Table 1: as regards the sandwich skins they are calculated as previously mentioned whereas the core characteristics have been provided by the manufacturer. Table 1:

EPREG UD 52TM EPREG DIAG 43TM EFOAMTM

Elastic properties for the materials under study. Ex (MPa) 29966

Ey (MPa) 12584

Ez (MPa) 10833

Gxy=Gyz=Gxz (MPa) 4282

Qxy

Qyz

Qxz

0.207

0.208

0.127

27125

10613

8783

3413

0.212

0.204

0.121

1512

582


0.300


4

235

Small scale testing

The mechanical behaviour of the sandwich and its components (skins and core) has been investigated through the following series of tests on specimens directly prepared by the manufacturer: tension, compression, three point bending tests and short beam tests as regards the skins, [6], three point and four point bending tests, uniaxial compression, uniaxial tension, constrained compressive tests on the core, (Cecchinelli [2]), and, concerning the specimens taken from the sandwich panels, three and four point bending tests, [6]. For each group of tests, specimen shapes and sizes have been chosen according to the relevant standards. FE models of all tests have been developed as mentioned before and nominal dimensions have been considered. A few significant results are presented as an example in Table 2, Figure 2 and Figure 3, comparing the averaged experimental data for the three point bending tests and short beam tests on EPREG UD 52TM. Satisfactory agreement between tests and calculations was found for skins laminates while larger difference exists for the sandwich specimens. Such discrepancies may be explained taking into account that small single skin specimens were specifically made for tests while large sandwich panels, from which small specimens were taken, were built according to the usual shipyard practice. Table 2:

Weft Warp

Weft Warp

Figure 2:

Comparison between averaged experimental data and FEM results for the three point bending tests (TPB) and short beam tests (SBT). fmax exp (mm) 8.54 2.12

Vmax exp (MPa) 638 54

Wmax exp (MPa) 47.77 8.15

TPB - EPREG UD 52TM fmax FEM Vmax FEM (mm) (MPa) 8.85 656 2.82 54 SBT - EPREG UD 52TM Wmax FEM (MPa) 59.00 11.00

Error fmax exp/FEM 4% 25%

Error Vmax exp/FEM 3% 0%

Error Wmax exp/FEM 19% 26%

Example of a FE model with the corresponding experimental test.


236 Computational Methods and Experiments in Materials Characterisation III Tau XZ

Tau Il XZ 15

15

14

14

13

13

12

12

11

11

10

10 9

9 strati

8

Tau IL XZ

7

strati

8

Tau XZ

7

6

6

5

5

4

4

3

3

2

2 1

1

0

0 0,00

20,00

40,00

60,00

0,00

80,00

20,00

40,00

30

30

25

25

20

20

15

strati 15

Sigma X

10

10

5

5

0 -1500

-1000

-500

Tau IL XZ

0

0

500

1000

1500

0

5

[MPa]

Figure 3:

5

80,00

Tau IL XZ

Sigma X

strati

60,00

10

15

20

25

[MPa]

Examples of distributions of stresses in the layers from FE analyses (interlaminar shear and shear of short beam test, tension and interlaminar shear of three points bending).

Large scale tests

Large scale tests have been carried out at DINAV ship structures laboratory on two 2000x1000 mm sandwich panels supplied by Nuova Connavi S.r.l. The three point bending test has been deemed the most significant for the mechanical characterization and for comparisons with small scale tests. 5.1 Panel 1 Panel 1 has a lower skin (in tension) with a 5-ply [0/90/±452/0] staking sequence and a 4-ply [0/90/±452] staking sequence upper skin (in compression); each layer has a nominal thickness of 0.4 mm, whereas the core is 50 mm thick. Strain gages have been bonded on outer skins following the map of Figure 4: the three mid-span channels are rosettes, placed to evaluate the on-plane shear stress as well as the longitudinal stress induced by bending moment. This layout has been repeated also in between the lower skin and the core to evaluate interlaminar shear stresses. Signals of gages have been recorded using a routine developed on purpose in Labview® and analysed by means of some Matlab® routines: some examples are shown in the following Figure 6 to Figure 8. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


Figure 4:

237

Strain gages layout bonded on Panel 1.

FPF

Figure 5:

Comparison of Load-Displacement experimental data of panel 1 with FEM calculation of First Ply Failure (FPF).

Panel collapsed at 52 kN with 130 mm displacement and FPF (First Ply Failure) has been reached at 25 kN with 45 mm displacement. Figure 6 shows the behaviour of some significant gages and FPF may be noted. Such curves highlight that some areas of panel collapsed at 25 kN and others maintained residual strength up to the final collapse. Shear stresses have been evaluated using the rosettes signals (Figure 7). Moreover, three constantan wires (Ch.0, Ch.1, Ch.2) have been inserted between the lower skin and the core to obtain the bending average deformation. It is worth to point out that all wires, other than Ch.2 whose signals went lost due to wiring connection problems, behave in the same way: they all failed to provide electrical signals only when the panel collapsed, reaching a strain of nearly 5000 PHFPF may be noted when the slope of plots in Figure 8 suddenly changes.



Figure 6:

Figure 7:

Examples of plots representing gages signals vs. load.

Load vs. shear stress calculated by internal east and west strain gages (IntW & IntE) and by external center strain gage (ExtC).

Figure 8:

Constantan wires signals of panel 1.

Figure 9:

Large scale test and panel collapse.



239

5.2 Panel 2 Panel 2 has both skins with 8-ply [0/90/±452]2 staking sequence; each layer has a nominal thickness of 0.4 mm, whereas the core is 30 mm thick. Strain gages have been bonded according to the map of Figure 4. Constantan wires have been also placed and recorded data are shown in Figure 11. The final collapse occurred just after the FPF, probably because of the symmetry of the skins and of the lower thickness of core with respect to panel 1. Strain gages provided signals similar to the ones of panel 1, not reported here for sake of shortness. FEM calculation estimates exactly the collapse load of panel 2 (60 kN) but overvalued the displacement (191 mm instead of 180 mm).

FPF

Figure 10:

Comparison of Load-Displacement experimental data of panel 2 with FEM calculation of First Ply Failure (FPF).

Figure 11:

Constantan wires signals of panel 2.

6 Lessons learned The study presented in this paper highlights that material characterization needs to be carried out looking towards the overall size and behaviour of the structure. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

240 Computational Methods and Experiments in Materials Characterisation III While failure modes of small scale tests are clearly identifiable, the failure of the large scale specimens follows a progressive collapse where different areas of the sandwich are affected by different failure modes. The interaction of different failure modes might not be simply superimposed. FE models are in substantial agreement with the small scale tests while larger differences have been found with the large scale ones. Further to the manufacturing defects, whose density may be higher in larger structures, interaction of failure modes may lead to lower material strength. A larger number and size of defects have been noted in the 50 mm thick core with respect to the 30 mm one because of the different manufacturing procedures. This impacted onto the strength of the panels. The gages and the constantan wires inserted between the lower skin and the core allows estimating the interlaminar shear stresses and shows that bonding of skins and core is better than in traditional sandwich used for pleasure craft, probably because of the same origin of the constituent materials. Finally, it is believed that constantan wires can be used to realize a cheap and very light system for structural monitoring of very large areas of FRP hulls. Constantan wires would be weaved in glass reinforcement fabric as weft or warp. Of course, prototypes cited in this paper need to be further developed and tested.

Acknowledgements The present paper originated from the research project no. 23 founded by the European Union, in the framework of PRAI-Liguria (Programma Regionale di Azioni Innovative). At that time, the author M.P. was an employee of DINAV. The authors wish to acknowledge the invaluable support of Professor Giovanni Carrera.

References [1] [2] [3] [4] [5] [6]

Greene, E., Marine Composites, Eric Greene Associates, Inc: Annapolis, p. 73, pp. 272-273, 1999. Cecchinelli, A., Mechanical characterization of an epoxy syntactic foam, MSc thesis, Pisa University, 2005. ANSYS® Release 8.0 Documentation, ANSYS Inc: Canonsburg, 2003. Tsai, S.W., Hahn H.T., Introduction to composite materials, Technomic Publishing Company: Lancaster, pp. 392-399, 1980. Rules for the Construction and Classification of High Speed Craft, HSC Code, EEIG UNITAS, 2002. Della Biancia, C., Reports of small scale tests CTS job no. 416/06, Centro Tecnologico Sperimentale S.r.l.: La Spezia, 2006.



241

Modelling of viscoelastic properties of a curing adhesive J. de Vreugd1, K. M. B. Jansen1, L. J. Ernst1 & J. A. C. M. Pijnenburg2 1 2

Delft University of Technology, Delft, The Netherlands TNO Science and Industry, Delft, The Netherlands

Abstract Thermoset adhesives are widely used in high tech applications to join two bodies together. The main advantages of using adhesives are the low weight of the construction and the easy way to apply the adhesive to the surfaces which have to be fixed together. The disadvantage of thermoset adhesives however is that cure shrinkage occurs. Shrinkage and evolution of mechanical properties during cure leads to development of internal stresses. In this paper, the mechanical behaviour of a curing adhesive is investigated. In the case of using a thermoset adhesive in high precision applications like optical instruments, care should be taken. Small displacements and distortions of important components caused by cure shrinkage may already lead to malfunctioning. For this reason a material model suitable for implementation in a finite element program is developed to predict stresses and strains in glued objects. The temperature and cure dependent viscoelastic shear modulus of the adhesive are obtained by using Dynamic Mechanical Analyzing methods. The bulk modulus is obtained at fully cured state with a high pressure dilatometer. Curing-time–time superposition is applied to model the shear modulus at any state of cure. It is assumed that the bulk modulus remains constant during cure. The kinetics of the adhesive is investigated by using Dynamic Scanning Colorimetric techniques. The relation between time and degree of cure is modelled by making use of the KamalSourour equation. Also diffusion limitation is added to this equation. The cure shrinkage of the adhesive is experimentally determined by making use of the principle of Archimedes. Finally a validation experiment is performed. The validation experiment is simulated in the finite element program ABAQUS and compared with the experiment. It turned out that the developed material model is accurate enough to predict reaction forces, stresses and strains caused by cure shrinkage. Keywords: adhesive, cure shrinkage, DMA, DSC, viscoelastic properties. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070241


1

Introduction

Thermoset adhesives are used in many high-tech applications to fix two bodies together instead of other bonding techniques. Thermoset adhesives are often used when a low construction weight is required. Another advantage is the easy way of applying the glue to the surfaces. Next to above mentioned advantages of using an adhesive, also some negative properties are present. An important disadvantage of using an adhesive for bonding is the shrinkage of the adhesive during the transformation from a fluid to a solid material. The shrinkage results in distortions and internal stresses. In instruments where a high precision is required like optical instruments, cure shrinkage might cause problems. Displacements and rotations of important parts in high precision instruments are undesirable because of the necessary accurate position. It is even possible that cure shrinkage leads to cracks in a glued object. An example is shown in figure 1 where a glass plate is glued to a metal surrounding. In this example shrinkage forces were that high that the glass plate is cracked.

Figure 1:

Cracks caused by cure shrinkage.

In order to avoid the mentioned problems, cure shrinkage should be taken into account at the design state of instruments where a high precision is required. To be able to produce a fail-proof design, a reliable model of both cure shrinkage and viscoelastic material properties is needed to predict stresses and strains during cure. The Araldite AV 138 M adhesive is a frequently used adhesive in aerospace. The cure shrinkage of this adhesive caused many problems in the past. For this reason is chosen to investigate the viscoelastic behaviour of this adhesive during the transformation from a fluid to a solid material. The mechanical behaviour of the adhesive is completely characterized and modelled such that all viscoelastic properties are available at a large range in time, temperature and at a certain degree of conversion. The obtained model is used to solve the linear viscoelastic stress-strain relation: t

V ij t 2 G t [ Hij d[

³

f

t

2

³ «¬K t [ 3 Gt [ »¼H ª

º

eff 11

eff eff H 22 H33 G ij

(1)

f

In equation (1) the variables G and K refer to shear and bulk modulus respectively, which are a function of time (t), temperature (T) and degree of WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


243

conversion (Į). The variable H iieff is the effective volumetric strain which includes cure shrinkage and thermal contributions. A model of both shear G(t,T,Į) and bulk K(T) modulus as well as cure shrinkage İcure is proposed in this paper. The mechanical properties are experimentally found by using DMA (Dynamic Mechanical Analyzing) techniques and by using a high pressure dilatometer. The chemical reaction model is found by using DSC (Dynamic Scanning Calorimetric) measurements. Finally the decrease in density of the adhesive is measured during cure by using the method of Archimedes. All experimentally found properties are modelled so that they can be implemented in a finite element program. In order to validate the material model of the adhesive during cure, a validation experiment was done. This experiment showed that the determined material model is accurate enough to do reliable predictions of stresses and strains.

2

Cure kinetics

Chemical reaction is started by applying a thermal loading to an uncured or not fully cured material. During this reaction, the individual epoxy monomers transform to a three dimensional network. This network prevents the molecules to slide past each other; this is the reason that a fluid like material transforms into a solid. The rate of reaction is dependent on the applied temperature. At the instant that the curing reaction is finished, one speaks about a fully cured material. The states between un- and fully cured situation are expressed by the expression: degree of cure or degree of conversion which is represented by the symbol Į. The value of Į varies between 0 and 1. To be able to model the cure reaction it is necessary to describe the progress of the reaction, such that it is possible to calculate the degree of conversion at any moment of time and temperature. The reaction progress is measured in this research project by a DSC 2920 of TA instruments. 2.1 Degree of cure determination During cure, heat comes free because of the chemical process (cross-linking). The degree of cure Į is related to the maximum heat which comes free after a complete reaction (Hmax) and the heat which comes free after a certain state of reaction (H). The degree of cure (Į) is defined as 1-H/Hmax. In order to measure the degree of conversion of a partly cured sample, it is necessary to calculate the total heat generated by a complete cure reaction. A temperature ramp is applied to an uncured sample and the rate of heat generation dH/dt is measured. Several heating rates (ȕ) are applied: 1, 2, 5, 10ºC/min. (dynamic scanning). The heat of reaction H is the amount of heat generated during dynamic scanning. The total generated heat, caused by the reaction is calculated for every measurement by the following equation:


244 Computational Methods and Experiments in Materials Characterisation III t

Hu

§ dH ·

³ ¨© dt ¸¹dt

(2)

0

The thus obtained total heat of reaction varied between 130.27 and 135.42 J/g. A value of 135.42 J/g is used in further calculations. 2.2 Cure dependent Tg determination Viscoelastic materials have the property that at low temperatures the material behaves glassy and at high temperatures more rubbery. The temperature at which this behaviour changes from a glassy to a rubbery behaviour is the glass transition temperature (Tg). The glass transition temperature is a cure dependent property and is therefore measured as a function of degree of conversion. This property is measured as a sudden change in heat capacity Cp (Seifi et al [1]). The heat capacity is measured by applying DSC scans to samples of different conversion levels. The results of these tests are shown in figure 2.

Figure 2:

Glass transition temperature as a function of degree of conversion.

The above measured data points are fitted to the Di-Benedetto equation: T gf T g 0 O D (3) T g D T g 0 1 1 O D Tgf and Tg0 represent Tg at fully and uncured state respectively. The measurements showed that, Tgf = 77.5°C and Tg0 = -32.1°C. The value Ȝ is a material dependent parameter. For the adhesive studied in this paper, Ȝ = 0.474.

2.3 Kinetic model The chemical reaction is described by the model of Kamal and Sourour. This model is given in equation (4).



245

§ Ea ·

¨¨ ¸¸ dD k 0 e © RT ¹ D m 1 D (4) dt In this equation Ea denotes an activation energy (Starink [2]), R denotes the universal gas constant § 8.314 J/(mol·K) and T represents the absolute temperature as a function of time. The parameters k0, m and n are fit variables. The following values are found: k0 = 3.2982·105, m = 0.185, n = 1.5154 During isothermal curing, a thermosetting resin vitrifies if the reaction temperature is lower than the maximum glass transition temperature of the fully cured material. Due to the vitrification process, the kinetics becomes diffusion controlled. This phenomenon is also observed in the studied adhesive. It turned out that a sample cured at a room temperature could not reach maximum conversion. The maximum conversion level turned out to be 81%. Therefore, the kinetics model is modified to:

dD ª dD º fd (5) dt «¬ dt »¼ chem [da/dt] describes the chemically controlled kinetics. Kamal-Sourour’s equation is used here. fd denotes the diffusion control function (Schawe [3]). If the reaction is chemical controlled fd is equal to unity. In case of diffusion controlled reaction fd will have a value between 0 and 1. The diffusion control function has to show an inflection point if the glass transition temperature is equal to the reaction temperature. A model for this function has to describe the inflection point properly. The following equation is fulfilling the mentioned requirements: 1

3· § ¨ 1 §¨ Treact 'T T g D ·¸ ¸ (6) f d D 1 ¨1 ¨ ¸ ¸¸ 'T ¨ 2© ¹ © ¹ Treact is the temperature where the reaction place. ǻT is a fitting parameter; for this adhesive is found by trial and error that ǻT is 21.5 °C.

3

Mechanical properties

To be able to predict the stresses in a glued object, it is necessary to know the mechanical properties of the adhesive. The properties have to be known at a fully cured state, as well as during the curing trajectory. For the viscoelastic elongation- and shear-modulus a Dynamic Mechanical Analyzer (DMA) is used. The used test device for these measurements is a DMA Q800 of TA-instruments. This instrument has a displacement resolution of 1nm and a force resolution of 1mN. The bulk modulus is measured by a high pressure dilatometer. A GNOMIX dilatometer with a pressure range of 200MPa is used. 3.1 Tensile modulus of fully cured adhesive In order to measure the viscoelastic properties of the fully cured material, a test bar was required. The used dimensions are [22.89 x 3.1 x 0.82 mm]. The test bar WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

246 Computational Methods and Experiments in Materials Characterisation III is produced by curing the adhesive in a suitable mold. Before curing the material submitted to vacuum to subtract the gas bubbles in the uncured resin. A cure temperature of 75 ºC is applied for 5 hours. The test bar is exposed to a sinusoidal strain with different frequencies: 0.3, 0.65, 1.4, 3, 6.5, 13.8, 30, 64.6, 130 Hz. During the frequency sweeps a temperature ramp is applied from -50ºC to 220ºC with a heating rate of 1ºC/min. The result of this experiment is shown in figure 3.

Figure 3:

Result of DMA experiment to a fully cured bar of adhesive.

From this figure it is concluded that the glass-plateau of this material is: E glass 5517 10 6 30.452 10 6 T [MPa]

(7)

By applying the time-temperature superposition principle (Nielsen and Landel [4]) to the rough data, a mastercurve is obtained. The shiftfactors (at) are fitted to the Williams-Landel-Ferry equation: C1 T T R loga t (8) C 2 T TR The constants are treated as fit variables; Tg is taken as the reference temperature TR. By applying a non-linear fit, it is found that C1 = 1.51, C2 = 28.6. Tg is the value where tan(į) at 1 Hz reaches a maximum. Tg turned out to be 84ºC. 3.2 Bulk modulus measurement A GNOMIX high pressure dilatometer is used to measure the bulk modulus. A sample with a typical mass of 1.5 gram is contained in a rigid cell, closed by flexible bellows. This cell contains mercury to fill the cell completely. The cell is placed into a vessel which can be heated. A hydrostatic pressure applied to the cell, will result in a deflection of the bellows. This deflection can be related to deformations inside the cell. The pressure range that can be applied to the sample varies between 10 and 200MPa, the highest applicable temperature is 400ºC. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


247

In order to measure the bulk modulus of the material, a stepwise temperature scan is applied to the material. At every temperature step, steps of 10MPa are applied. Equation (9) (Fung [5]) is used to calculate the bulk modulus: v K 'p (9) 'v The result of this measurement is presented in figure 4.

Figure 4:

Bulk modulus as a function of temperature.

In figure 4 is shown that the maximum modulus value is about 2900 MPa, the lowest about 1800 MPa. For temperatures above 50ºC the material is time dependent so for these temperatures the measured values cannot be used for finite element simulations but it gives an estimate of the modulus at those particular temperatures. 3.3 Shear modulus during cure To be able to predict stresses and strains in a glued object during cure, it is necessary to know the viscoelastic properties during cure. The cure dependent shear modulus is determined by measuring the change in stiffness of a droplet of adhesive which is clamped between 2 plates. To one of the plates a sinusoidal strain of 5 µm is applied. By recording the forces and amplitudes of the plate during the experiment, the stiffness K of the sample is calculated. With the known dimensions of the droplet of adhesive the shear modulus is calculated: § 2h · (10) G Z K Z ¨ ¸ © A¹ In equation (10), h and A represents the gap between the plates and the crosssectional surface of the adhesive sample respectively. Different isothermal loadings are applied to sample such that the adhesive cures during the experiment. Three experiments are done with isothermal loadings of 40ºC, 45ºC and 50ºC. The results of these experiments are plotted in figure 5. In figure 5 the viscoelastic shear modulus is plotted as a function of degree of conversion. At conversion levels lower than 0.55, the material is still a fluid. The shear modulus in this region is 0. In order to obtain a mastercurve of the shearWIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

248 Computational Methods and Experiments in Materials Characterisation III modulus, the cure-time–time superposition principle (Yongsung [6]) is used. The shiftfactors which are used for determining the mastercurves are fitted to the following equation: Shift D , T 10 C1 C2 T C3 e

C 4 D

(11)

The following values for the fit factors are found: C1 = -2.361, C2 = -0.150, C3 = 16.17, C4 = -3.66.

Figure 5:

4

Result of shear tests during cure.

Cure shrinkage

The cure shrinkage is experimentally found by making use of Archimedes' principle. An apparatus is designed which makes use of buoyancy forces caused by immersing a body in a fluid. By knowing the mass of the sample, the weight of the mass immersed in the fluid, and the density of the fluid, the density of the sample can be calculated. The mass of the sample and the density of the fluid should be known before doing the measurement. As an immersing fluid, silicone oil is used with a density of 0.9670 g/cm2 and a CTE of 8.20·10-4/K. Different isothermal loadings are applied to samples: 20ºC for 70 hrs, 40ºC for 18 hrs and 50ºC for 16 hrs. The result of the measurement at 20ºC is shown in figure 6. From figure 6 is concluded that there is linear relation between degree of cure and density. The total volumetric cure shrinkage Ȗv is calculated with the following equation: 'v 'U 0.0704 Jv 0.0133 (12) v fully 3 U ful 3 1.7674



249

ȡ = 1.697 + 0.0704·Į

Figure 6:

Results of density measurement.

5 Validation experiment In order to check the accuracy of the obtained material model, a validation experiment is done, see Figure 7. In this experiment a droplet of adhesive is applied in the middle of a glass-plate which is fixed at both ends. Dimensions of the glass-plate were: [40 x 10 x 2 mm]. Glass-plate

Adhesive bump Figure 7:

Schematic drawing of validation experiment.

Due to the shrinkage of the adhesive, the glass-plate will deflect. A temperature load is applied to the adhesive, firstly a temperature of 40ºC is applied for 35 hours, after that the temperature is changed to 80ºC for 20 hours. In the validation experiment, the reaction force at the bottom of the adhesive bump is measured. Simultaneously the force is calculated by using the finite element program ABAQUS. User-subroutines were used to implement the obtained material model. A picture of the used mesh is given in figure 8. Shell-elements are used for the glass-plate, and solid elements are used for the adhesive bump. Measured and calculated forces are presented in figure 15. In figure 9, the calculated forces are compared to the measured forces. The calculated forces are about 30% too high. This is probably due to the bulk modulus which was implemented in the simulation model as a non-time and noncure dependent value. Another reason is most probably due to friction at the boundary conditions. This is not modelled due to a lack of time. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


Mesh of the validation experiment. 100

10 8 6 4 2 0 -2 -4 -6

80 60 40 20

0

1000

2000

3000

Temperature °C

Force [N]

Figure 8:

Measured force Calculated force

Temperature

0 4000

Time [min.]

Figure 9:

Result of validation experiment.

6 Conclusions and recommendations In this research, a first start is made in characterising the mechanical properties of the adhesive Araldite AV 138M. The mechanical properties which are a function of time, temperature and degree of conversion are studied and fitted in a material model. The kinetics of this material is well described in a relation in which also diffusion limitation is implemented. The cure shrinkage is found and modelled. The material model is implemented in ABAQUS. The following properties of the investigated adhesive were established during this work: Tg varies between -32°C and 77.5°C during cure. The relation between Tg and degree of conversion is well described by Di-Benedetto’s equation. The kinetics is modelled by making use of Kamal-Sourours’ equation. Diffusion limitation is added to this model. The fully cured elongation modulus varies between 6500 MPa at -50°C and 65 MPa at 200°C. The bulk modulus varies between 2700 MPa at 30°C to 1800 MPa at 90°C. A simple cure and temperature dependent shiftfactor is obtained, with which mastercurves at other conversion levels can be found. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


251

The decrease in density is about 4%. It turned out that there is a linear relationship between degree of conversion and density. From the validation experiment is concluded that the obtained material model is accurate enough for predicting stresses and strains in glued objects. For future work some recommendations are listed below: More validation experiments should be done. Some parameters of the validation experiment can be changed. For instance the thickness of the adhesive layer. It might be that the reaction forces caused by cure shrinkage are very sensitive to the applied layer thickness. It turned out that there is an error of about 30% between the measured and simulated reaction forces. This is probably caused by an inaccurate bulk modulus. So, bulk modulus has to be found as a function of time, temperature and degree of conversion.

References [1] Seifi, R., Hojjati, M., Heat of reaction, cure kinetics and viscosity of araldite LY-556 resin, Journal of composite materials, 39(11), pp. 10271039, 2005. [2] Starink, M.J., The determination of activation energy from linear heating rate experiments: a comparison of the accuracy of isoconversion methods, Thermochimica acta, 404(1-2), pp. 163-176, 2003. [3] Schawe, J.E.K., A description of chemical and diffusion control in isothermal kinetics of cure kinetics, Thermochimca Acta 388(1-2) pp. 299312, 2002. [4] Nielsen, L.E., Landel, R.F., Mechanical properties of polymers and composites, pp.110, 1988. [5] Fung, Y.C., Foundations of solid mechanics, pp. 113, 1984. [6] Yongsung E.O.M., Louis Boogh et al., Time-cure-temperature superposition for the prediction of instantaneous viscoelastic properties during cure, Polymer engineering and science, 60(6), pp. 521-528, 2002.




253

Flexural bond strength of clay brick masonry C. G. Yuen & S. L. Lissel Civil Engineering Department. University of Calgary, Canada

Abstract The intent of a new parametric study at the University of Calgary is to investigate the influence of several factors on the flexural bond strength of clay brick masonry. These factors include the absorption characteristics of the brick units, and varying construction and curing methods. A preliminary study was performed with a series of clay brick prisms built from different types of brick with various absorption characteristics, and cured at different conditions. The bond wrench test was used to determine the flexural bond strength between the mortar and brick. The results showed high variation, but did provide some indication of which factors may be contributing to the highly variable findings. The objectives of this continuing study are to eliminate the possible parameters that were causing the highly variable results, and to determine correlations between brick properties and bond strength. The results are presented in this paper. Further research will be ongoing to establish a more definite relationship between the various parameters and bond strength and to investigate effects of mortar, and curing conditions. Keywords: masonry, clay brick, flexural bond strength, bond wrench, initial rate of absorption, sorptivity, absorption, construction method.

1

Introduction

Masonry is a composite material of clay or concrete units held together by mortar. An ineffective bond between the unit and mortar will cause cracking when subjected to lateral loading. Cracks increase susceptibility to moisture ingress, which leads to freeze-thaw damage, and corrosion of metal connectors. Therefore, the bond strength between the brick and mortar acts as an indicator of the overall quality of the masonry structure [1]. For many years, researchers have been interested in determining the factors that affect the bond at the interface of the two materials. These parameters WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070251

254 Computational Methods and Experiments in Materials Characterisation III include the absorption characteristics of the units (initial rate of absorption or IRA, sorptivity, and cold and hot water absorption), mortar properties (type, flow, and retentivity), curing conditions, and workmanship. It is difficult to determine the significance of one factor and its effect on the bond strength because no factor alone is responsible for good bond, making it difficult to devise an experiment to produce consistent results [2]. Numerous studies [3–7] have shown large differences in strengths and highly variable results, and no conclusive findings were obtained from replicating experimental procedures [1]. A preliminary study was performed at the University of Calgary in the summer of 2005 investigating the effects of various absorption properties of clay bricks (particularly the IRA property) and curing conditions on the flexural bond strength of clay brick masonry. Although the variability of the results was high, it did provide some indication of which factors may be contributing to these highly variable findings. The continuing study involves the investigation of two possible causes of variability with the objective to eliminate these factors, and also aims to determine correlations between brick properties and bond strength. This paper identifies the possible causes of variability, and presents the results when new methods were applied to eliminate these factors. Apparent correlations between different brick properties are also presented.

2 Stage 1: eliminate construction factors 2.1 Identifying the factors The first factor identified from the preliminary study is that the mortar was mixed by an experienced mason. It was observed that the color and texture of the mortar joints varied. An explanation for this is that the amount of water added into the mortar was based on the experience of the mason, and each new batch of mortar may have differed slightly. Another possible factor that contributed to the highly variable results is the height of the prism. For the preliminary study, 5-brick high prisms were used. It was hypothesized that the varying weight on each mortar joint along the height of the prism may have caused stress variation at the joints. Lastly, CSA A371-04 [8] requires mortar joints to be 10 mm thick with a tolerance of ±3 mm. Although the joints were fairly consistent with all the prisms, this is also considered as a contributing factor to the high variations. With these identified factors, the first stage of the current study was to attempt to reduce the variability by eliminating these factors using nonconventional construction methods. 2.2 Materials 2.2.1 Bricks Three types of brick were used with various IRA values: tan (IRA = 10 g/min/200cm2); red (IRA = 23 g/min/200cm2), and cream (IRA = 42 g/min/200cm2). These same brick types were also used in the preliminary study. All units are metric modular with dimensions 90 x 190 x 57 mm (W x L x H). WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


255

2.2.2 Mortar A general purpose Type S 1:0.5:4.5 (Portland cement:lime:sand by volume) mortar was used. The contents were proportioned and mixed in accordance with CSA 179-04 [9]. The amount of water added was measured (by weight) and recorded for future mixes. This entire procedure was performed by the researcher. 2.3 Specimen preparation To eliminate the varying weight on the mortar joints along the height of the prism, it was decided to build 2-brick high prisms. To have all the prisms built the same, a simple jig was designed for proper alignment of the units and to ensure a 10 mm mortar joint in between. The jig consists of four wooden right angled pieces, with an M6 hex screw embedded in the middle. The screw head has a diameter of 9.8 mm and sits on the bed face of the brick, and all the corner pieces are held together by a heavy-duty elastic. A full bed of mortar is then placed, and the top brick is added (Figure 1). Afterwards, the elastic is removed and the corner pieces are pulled out. A total of 110 prisms were made, and all were air-cured at ambient laboratory conditions (temperature of 20ºC and relative humidity 21%).

Figure 1:

Prism construction with jig.

2.4 Test method The bond wrench method described in CSA S304.1 Annex E [10] was used to determine the bond strength of the masonry prisms. The test apparatus and method have been used in other studies as well [11]. 2.5 Results and discussion Specimens were tested at 7 and 28 days. The average bond strengths and standard deviations are plotted, and shown in Figure 2. It can be seen that the variability for the 7-day cure is much greater than the 28-day cure. It is also surprising to see that bond strengths tend to be greater at 7 days then at 28 days. But due to the highly variable results, it is difficult to make any conclusions. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

256 Computational Methods and Experiments in Materials Characterisation III Bond Strength vs IRA at 7 and 28 Days 2.50

Bond Strength (MPa)

2.00 1.50

7 Days 28 Days

1.00

2

R = 0.0114

0.50 2

R = 0.832 0.00 0

10

20

30

40

50

IRA (g/min/200cm2)

Figure 2:

3

Bond strengths of three brick types at 7 and 28 days.

Stage 2: eliminate brick-to-brick variability

3.1 A new factor to consider It is obvious from Stage 1 that despite controlling most workmanship variables, large variability still exists. Therefore, in this stage the variability in brick absorption properties for each individual brick were taken into consideration. This variability has been reported in the literature as well. With a sample of 20 of the same brick type, Lauersdorf and Robinson [12] observed that individual brick IRA ranged from 14.9 to 39.4 g/min/194 cm2. Bailey et al. [13] reported that an individual brick unit may even exhibit significantly different IRA values from one bed face to another. Therefore, for the second stage, brick couplets were paired up according to their individual IRA and sorptivity values. Care was taken to ensure that the tested bed face was the one on which mortar was placed. 3.2 Materials 3.2.1 Bricks Two types of bricks were used: tan and red (both these brick types were used in the previous tests). Prior to construction, each individual brick was tested for its IRA and sorptivity properties in accordance with CSA A82-06 [14], and ASTM C1585 [15]. More than 600 brick units were tested for IRA. For the red brick, individual IRA values ranged from 25.6 to 59.2 g/min/200cm2 (difference of 33.6 g/min/200cm2), whereas for the tan brick IRA ranged from 6.6 to 20.1 g/min/200cm2 (difference of 13.5 g/min/200cm2). Therefore, it was decided to pair up the red bricks according to IRA and the tan bricks according to sorptivity. Approximately 200 of the tan bricks were further tested for sorptivity which WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)


257

ranged from 0.0226 to 0.099 mm/s0.5. Bricks with matching values were paired up (tolerance of ±0.2 g/min/200cm2 for IRA, and 0.0001 mm/s0.5 for sorptivity) for prism construction. 3.2.2 Mortar The same mortar preparation used in Stage 1 was used in Stage 2. 3.2.3 Specimen preparation The same method from Stage 1 was used in Stage 2 to construct the prisms. A total of 50 matched IRA prisms were constructed, and were cured for 14 days: 7 days air-cured at ambient laboratory conditions, and 7 days covered with a sheet of plastic. Only 25 matched sorptivity prisms were built, and all were cured for 14 days, and covered with a plastic sheet for the whole curing duration. 3.3 Test method As in all the previous tests, the same bond wrench method was used to determine the bond strength of the masonry prisms. 3.4 Results and discussion The results from the matched IRA prisms are shown in Figure 3. Bond Strength vs Matched IRA 0.45 Minimum 0.2 MPa bond strength requirement per CSA S304.1

0.40 Bond Strength (MPa)

0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 25

30

35

40

45

50

55

IRA (g/min/200cm2)

Figure 3:

Bond strengths of the matched IRA prisms.

CSA A371-04 [8] uses IRA as a guideline to ensure proper bond strength in masonry construction. It suggests that a brick unit with IRA of 30 g/min/194 cm2 is considered a high IRA brick. Without prewetting, the brick will absorb excessive amount of water and improper curing of the mortar will occur leading to poor bond strength. Despite controlling a number of identified factors, no correlation can be seen in Figure 3 between bond strength and IRA. Although more than half the prisms failed at bond strength below 0.2 MPa (the minimum bond strength required by WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

258 Computational Methods and Experiments in Materials Characterisation III CSA S304.1 [10]), it is possible to have good bond strength without prewetting high IRA bricks. It is also noted that the curing length and conditions did not conform exactly to the procedure outlined in CSA S304.1 [10]. Although the tan prisms were matched according to sorptivity, it was of interest to examine the relationship between bond strength and IRA of these prisms. Interestingly, the IRA values were also quite closely matched. The largest IRA difference was 5.4 g/min/200cm2. The IRAs for each prism were averaged, and then plotted with the corresponding bond strength (Figure 4). Similar to the matched IRA prisms, no correlation is observed between IRA and bond strength. Figure 5 presents the results of the matched sorptivity prisms. It can be seen that no correlation is apparent between sorptivity and bond strength either. Bond Strength vs Average IRA 1.20

Bond Strength (MPa)

1.00 0.80 0.60

Minimum 0.2 MPa bond strength requirement per CSA S304.1

0.40 0.20 0.00 8

9

10

11

12

13

14

15

16

17

18

19

2

IRA (g/min/200cm )

Figure 4:

Bond Strength vs. average IRA from the matched sorptivity prisms. Bond Strength vs Matched Sorptivity 1.20

Bond Strength (MPa)

1.00 0.80 0.60

Minimum 0.2 MPa bond strength requirement per CSA S304.1

0.40 0.20 0.00 0.02

0.04

0.06

0.08

0.1

Sorptivity (mm/s 0.5)

Figure 5:

Bond strengths of matched sorptivity prisms.



4

259

Relationship between various brick properties

As a supplement to this current study, it was of interest to determine whether typical brick unit properties (IRA, sorptivity, 24 hour cold water absorption, and compressive strength) correlate to each other. 4.1 Bricks Two types of bricks were chosen: tan, a relatively low absorption brick (same type that was previously used) and light tweed, a relatively high absorption brick that had not been used before. Ten bricks of each type were randomly chosen from the pallet for various property comparisons. 4.2 Properties Four brick properties were determined: IRA, 24 hour cold water absorption, and compressive strength were determined in accordance with CSA A82-06 [14]; and sorptivity was determined in accordance with ASTM C1585 [15]. 4.3 Results and discussions Sorptivity and 24 hour cold water absorption were plotted against IRA for each individual brick, and is shown in Figure 6. It can be seen that the sorptivity correlates well with IRA, but 24 hour cold water absorption does not appear to correlate to IRA. Compressive strength of the unit was plotted against each absorption property for each individual brick (Figures 7–9). It can be seen that compressive strength correlates well with the IRA and sorptivity properties, but not the 24 hour cold water absorption property.

Sorptivity (mm/s0.5)

0.3000 R2 = 0.1576

0.2500 0.2000

R2 = 0.9313

0.1500 0.1000 0.0500 0.0000 0.0

10.0

20.0

30.0

40.0

50.0

60.0

8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 70.0

Absorption (%)

Sorptivity and 24 h Absorption vs IRA

IRA (g/min/200cm2) Sorptivity

Figure 6:

24h Absorption

Relationship between various absorption properties with IRA.


260 Computational Methods and Experiments in Materials Characterisation III Compressive Strength vs IRA

Comp. Strength (MPa)

100.0 80.0 60.0 40.0 R2 = 0.8342

20.0 0.0 0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

IRA (g/min/200cm2)

Figure 7:

Relationship between compressive strength and IRA. Compressive Strength vs Sorptivity


100.0 80.0 60.0 40.0 R2 = 0.7732

20.0 0.0 0.0000

0.0500

0.1000

0.1500

0.2000

Sorptivity (mm/s

Figure 8:

0.5

0.2500

0.3000

)

Relationship between compressive strength and sorptivity. Compressive Strength vs 24h Absorption


100.0 80.0 60.0

R2 = 0.0601

40.0 20.0 0.0 4.0

4.5

5.0

5.5

6.0

6.5

7.0

7.5

Absorption (%)

Figure 9:

Relationship between compressive strength and absorption.



5

261

Conclusion

This study has shown that despite controlling various factors that affect the bond strength of masonry, such as construction methods, and brick-to-brick variability, there is still a lack of correlation between the flexural bond strength and IRA. Therefore, the question is raised whether Canadian standards should use IRA as a guideline to ensure good bond strength. In addition, no correlation was found between the bond strength and the sorptivity property of brick units, however, relationships between typical brick unit properties were found. More research is needed to determine how these relationships can be applied to the flexural bond strength, and how other factors such as mortar, and curing may also affect bond strength.

References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

Lawrence, S.J. and Page, A.W., Bond Studies in Masonry. Proc. of the 10th IB2MaC, eds. N.G. Shrive and A. Huizer, University of Calgary: Calgary, pp. 909-917, 1994. Goodwin, J.F. and West, H.W.H., A Review of the Literature on Brick/Mortar Bond. Proc. Of British Ceramic Society, 30(23), pp. 23-37, 1982. Baker, L.R., Some Factors Affecting the Bond Strength of Brickwork. Proc. Of 5th International Brick Masonry Conference, Washington, DC, pp. 62-72, 1979. Sise, A., Flexural Bond Strength of Masonry, MSc Thesis, University of Calgary, Canada, 139 pages, 1984. Venu Madhava Rao, K., Venkatarama Reddy, B.V., & Jagadish, K.S., Flexural Bond Strength of Masonry Using Various Blocks and Mortars. J. of Materials and Structures, 29(2), pp. 119-124, 1996 McGinley, W.M., IRA and the Flexural Bond Strength of Clay Brick Masonry. Masonry. Components to Assemblages: ASTM STP 1063, ed. J.H. Matthys, ASTM: Philadelphia, pp 217-229, 1990. Meslin, M. & Brzev, S., Effect of Mortar Type on Flexural Bond Strength of Brick Masonry. Civil Engineering Research Project from BCIT, Report No. CERP – 2006/01, 40 pages, 2006. Canadian Standards Association, CSA A371-04, Masonry Construction for Buildings. Mississauga, Canada: Canadian Standards Association, 2004 Canadian Standards Association, CSA A179-04, Mortar and Grout for Unit Masonry. Mississauga, Canada: Canadian Standards Association, 2004 Canadian Standards Association, CSA304.1-04 Design of Masonry Structures, 2004. Shrive, N.G. & Tilleman, D., A Simple Apparatus and Method for Measuring On-Site Flexural Bond Strength. Proc. of the 6th Canadian Masonry Symposium, University of Saskatchewan: Saskatoon, pp 283294, 1992. WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

262 Computational Methods and Experiments in Materials Characterisation III [12]

[13]

[14] [15]

Lauersdorf, L.R. & Robinson, G.C., Discussion of Paper, “Initial Rate of Absorption of Clay Brick Considering Both Bed Surfaces in the As Received Condition and After Outside Exposure. Masonry: Components to Assemblages: ASTM STP 1063 eds. J.H. Matthys, ASTM: Philadelphia, pp. 22-26, 1990. Bailey, W.G., Matthys, J.H., & Edwards, J.E., Initial Rate of Absorption of Clay Brick Considering Both Bed Surfaces in the As Received Condition and After Outside Exposure. Masonry: Components to Assemblages: ASTM STP 1063, ed. J.H. Matthys, Philadelphia, ASTM, pp. 5-21. 1990. Canadian Standards Association, CSA A82-06 Fired Masonry Brick Made from Clay or Shale, Public Review Draft 2005. American Society for Testing and Materials. ASTM C 1585, Standard Test Method for Measurement of Rate of Absorption of Water by HydraulicCement Concretes, ASTM: Pennsylvania, USA, 2004.



263

Structural, economic and material comparison of various steel grades under dynamic/fatigue loading I. U. Amobi & H. C. Uzoegbo School of Civil and Environmental Engineering, University of the Witwatersrand, Johannesburg, South Africa

Abstract As industries are upgrading rapidly from a lower steel grade to higher ones it has become necessary to study the effect of changing from lower steel grades to higher grades. This paper reports on fatigue life and behaviour, economic implications and material composition of these higher strength steels (HSS) as compared to the conventional grades. Three grades are commercially available in South Africa: 300W, 350W and 460W. These different steel grades (conventional and HSS) with the same moment capacities were subjected to constant dynamic stresses and the fatigue crack growth of the overloading and unloading were monitored and compared with each other. The influences of the overloading and unloading made standard grades perform better under repeated loading than the HSS, since HSS have been proved to have poor ductility, resulting in a lower number of cycles to failure. An 85% increase in material cost was generated as HSS replaces the conventional lower steel grades. A reduction in the number of cycles to failure in HSS was over 500%. Keywords: steel grade, HSS, fatigue, low-cycle fatigue, high-cycle fatigue, load capacity, cycles to failure.

1

Introduction

There is a trend towards increasing the strength grade of the general purpose steel for construction in most countries. This trend was prompted by increased loading on structures, larger spans and architectural designs that require smaller sections. Australia and other countries around the world have in recent years changed from lower steel grades to higher ones. In 2005, South Africa changed WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line) doi:10.2495/MC070261

264 Computational Methods and Experiments in Materials Characterisation III from grade 300W steel to grade 350W steel. The current investigation is mainly concerned with the study of the dynamic behaviour of the three main grades in South Africa.

2

Specimen section determination and Loading

Since various steel grades (300W, 350W and 460W) were tested under the same conditions, the load capacities of these steel grades were designed to be equal. To achieve this, an initial I-section of the 300W grade was assumed and its load capacity determined. The load capacity was then imposed on the other sections, 350W and 460W, and their different sections determined. 100 mm

16 mm

250 mm

218 mm

16 mm

16 mm

Figure 1:

Initial 300W section.

In order to maintain consistency, only the overall depth d, of the initial section grade was changed to suit the equivalent load capacity of grades 350W and 460W. As a result, the equation for the section modulus, Zpl became:

Z pl

ª100 xd 2 º ª 42 xd 32 2 º » « » 2« 4 ¬ 4 ¼ ¬« ¼»

(1)

As d = 250 mm for the initial section grade, the moment capacity was easily determined from the following equation:

Mp

0.9 xZ pl xf y


(2)


265

The moment capacity was then imposed on 350W and 460W accordingly and their respective overall depths were determined. A summary of the section dimensions is shown in the following table below: The samples are assumed to be laterally supported. Plates with 16 mm thickness were ordered in all three grades. The samples were factory fabricated to specification and delivered to the laboratory for testing. Table 1:

Specimen Breadth of top flange, btf Breadth of bottom flange, bbf Depth of top flange, ttf Depth of bottom flange, tbf Length of top flange, ltf Length of bottom flange, lbf Breadth of web, bw Depth of web, tw Length of web, lw Overall Depth, d

Section dimensions of the various grades.

300W

350W

460W

100 mm

100 mm

100 mm

100 mm

100 mm

100 mm

16 mm

16 mm

16 mm

16 mm

16 mm

16 mm

2500 mm

2500 mm

2500 mm

2500 mm

2500 mm

2500 mm

16 mm

16 mm

16 mm

218 mm

193 mm

155 mm

2500 mm

2500 mm

2500 mm

250 mm

225 mm

187 mm

2.1 Instrumentation The setup was done in such a way that a constant force was maintained throughout each experiment and the strain measurements were periodically taken. A 100 mm LVDT which was firmly fixed to the specimen was connected serially to both the DC voltage power supply and the memory card of an ‘Agilent’ Data Logger in order that voltage output can be measured during testing. The data logger was then connected to a computer which has the Agilent programme installed in order that the measurements can be adjusted and stored appropriately. Although the setup was not based on maximum deflection method but rather on force method, the LVDT was firmly fixed at mid-span directly under the WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

266 Computational Methods and Experiments in Materials Characterisation III point of load application. This enables us to ascertain the behaviour of the various specimens under specific loadings. Strain gauges were precisely glued at various sensitive places on the specimen and connected to a strain gauge reader. Since the MTS actuator measures its capacity in percentages, a load calibrator was used to convert the load percentage to actual readable loading quantities. 2.2 Loading There were two specimens for each grade of steel and two loading capacities for these various grades of steel. The specimens were tested under the same constant load as tabulated below. The load was applied at mid-span for all cases and supported at the supports. Proper bracings were provided in order to avoid lateral displacement. The applied loads were expressed as a percentage of the static load capacity of the sections. Table 2:

Steel Grade 300W 350W 460W

Load capacities.

Cyclic Load 1

Cyclic Load 2

0.50P = 122 kN 0.50P = 122 kN 0.50P = 122 kN

0.75P = 184 kN 0.75P = 184 kN 0.75P = 184 kN

Figure 2:

Test setup.



3

267

Results

Due to varying section compositions, the stresses generated at particular points were peculiar to each specimen. The strain values at support and zone of loading were measured after every 100,000 cycles at a frequency of 1 Hz using strain gauges. The stresses generated at the mid-span until failure is shown in the table below: Table 3:

Mid-span stresses during failure.

Stress generated (N/mm2) 165 192 251 249 290 378

Specimen loading 300W @ 122 kN 350W @ 122 kN 460W @ 122 kN 300W @ 184 kN 350W @ 184 kN 460W @ 184 kN

Top Flange of 300W @ 122KN 17500

17400

0 17380

17300

micro strain

17200

17100

17000

200000 16980

16900

100000 16865

300000 16875

400000 16855

500000 16860

600000 16860

700000 16840

800000 16865

900000 16870

1000000 1100000 16860 16850

16800 1200000 16750 16700 0

200000

400000

600000

800000

1000000

1200000

cycles to failure

Figure 3:

Microstrain result for top flange.

The stresses were constant until failure occurred. Since there were no changes in stresses, i.e. a constant stress was applied to the structure throughout the testing until failure; the stress-strain curve yielded a straight line. The changes in strain were monitored periodically during the life of the experiment. Microstrains were read off at every 100,000 cycles of loading. The behaviour of strain with respect to the number of cycles occurred in the same pattern for all cases. For the microstrain at the top flange, there was a sharp WIT Transactions on Engineering Sciences, Vol 57, © 2007 WIT Press www.witpress.com, ISSN 1743-3533 (on-line)

268 Computational Methods and Experiments in Materials Characterisation III contraction after the first 100,000 cycles and afterwards stabilization, showing that the top flange was under compression. For the microstrain in the web, there was a sharp increment of the specimen at the web as measured using the strain gauges. This shows that the web was under tension. Its behaviour was constant for all specimens, although the values were varying due to different loading for each specimen. Figures 1 and 2 show a consistent pattern of the microstrain results with respect to cycles to failure. Figures 3 and 4 are results for 300W at 122 kN. For the other experiments, the graph curve remains constant but with varying results at all points. Web - D of 300W @ 122KN cyclic loading 18550 1200000 18500

18500 18450 18400 100000 18365

micro strain

18350

200000 18390

300000 18370

400000 18390

500000 18395

600000 18380

700000 18410

800000 18410

900000 18390

1000000 1100000 18395 18395

18300 18250 18200 18150 0 18105

18100 18050 0

200000

400000

600000

800000

1000000

cycles to failure

Figure 4:

Microstrain result for web.

Figure 5:

Failed beam.


1200000


269

There were several initiations of micro-cracks which eventually grew and formed one macro-crack that caused instant fracture of the beams. Failure occurred instantaneously without warning because the beams where tested within their elastic region. When the beams were unloaded, they returned to their original form without any visual deformation. 3.1 Discussions It is observed that as the steel grades increase in yield stress with lower web depth, their cycle to failure reduces. As a wrap up, the tables below show in summary the points of failure for the various steel grades tested under the same load factor. Table 4:

Under 0.5P = 122 kN.

Steel Grades

Failure Cycle

Testing Time

300W

1,200,000

333 hours

350W

786,000

218 hours

460W

182,400

51 hours

Table 5:

Under 0.75P = 184 kN.

Steel Grades

Failure Cycle

Testing Time

300W

322,023

89 hours

350W

60

1 minute

460W

50

50 seconds

Figure 6:

Failed beam.


270 Computational Methods and Experiments in Materials Characterisation III Grades 350W and 460W failed by deformation under 75% of their capacity indicating poor ductility in material composition. All the specimens failed, whether fracture or deformation, within the zone of loading. The stresses induced at the point of failure where close to the maximum stresses induced at the midspan of the specimens. Even for specimen 1 (300W @ 122 kN) which failed by global buckling, the maximum deformation occurred at the zone of loading.

y, Load capacity (KN)

3.1.1 Structural consideration From tables 4 and 5 above, we can conclude that an increase in the steel grade reduces the capacity for the steel to withstand fatigue loading. The experiment shows clearly a reduction in cycles to failure as the steel grades increase. This statement can be shown graphically in the following graph. Failure behaviour of the specimens under cyclic Load 200

50 184KN

180

322023 184KN

60 184KN

160

y = -7E-05x + 206.74

y = -0.0003x + 184.02

140

y = -8E-05x + 184 182400 122KN

120

786000 122KN

1200000 122KN

100 80 60 40 20 0 0

200000

400000

600000

800000

1000000

1200000

1400000

x, Cycles to failure 300W

Figure 7:

350W

460W

Linear (300W)

Linear (350W)

Linear (460W)

Failure behaviour of specimens under cyclic loading.

Graphic equations were derived from the results in order that predictions for various load capacities of the steel grades can be fairly determined. The graph showed a linear relationship because only two load capacities for each steel grade were tested. As a result, estimations for cycles to failure of various load capacities can only be done within the limits of the experiment. For 300W, y = -7x10-05x + 206.64 For 350W, y = -8x10-05x + 184 For 460W, y = -0.0003x + 184.02 Where x = Cycles to failure and y = Load capacity in kN Range: 122 kN

Computational Methods and Experiments in Materials Characterisation III

Computational Methods And Experiments In Materials Characterisation II

Physical methods for materials characterisation

Physical methods for materials characterisation

Materials Characterisation V: Computational Methods and Experiments (Wit Transactions on Engineering Sciences)

Materials Characterisation IV: Computational Methods and Experiments (Wit Transactions on Engineering Sciences)

Computational Spectroscopy: Methods, Experiments and Applications

Computational Materials Chemistry: Methods and Applications

Advanced Computational Methods and Experiments in Heat Transfer X

Microwave Electronics: Measurement and Materials Characterisation

Surface Effects and Contact Mechanics IX: Computational Methods and Experiments

Composites with Micro- and Nano-Structure: Computational Modeling and Experiments (Computational Methods in Applied Sciences)

Composites with Micro- and Nano-Structure: Computational Modeling and Experiments (Computational Methods in Applied Sciences)

Computational Materials

Computational Methods in Photochemistry

Computational Methods in Earthquake Engineering (Computational Methods in Applied Sciences)

Computational Plasticity (Computational Methods in Applied Sciences)

Nanoporous Materials III

Computational Materials Chemistry: Methods and Applications (Bioelectric Engineering)

Iii-nitride Semiconductor Materials

Computational Ballistics III

Computational materials science

Magnetism: Molecules to Materials: Models and Experiments

Computational methods in structural dynamics

Computational Methods in Biomedical Research

Computational Methods in Systems Biology

Computational Methods in Plasma Physics

Computational Methods in Process Simulation

Computational Materials Science

Computational Materials Science

Computational Methods in Economic Dynamics

Computational Methods and Experiments in Materials Characterisation III