[Claude Bathias, Hiroshi Fukuda, Kyoshi Kemmoshi, (BookZZ.org)

RepairingStructures using

Composite Wraps

First published in Great Britain and the United States in 2003 by Kogan PageScience, an imprint of Kogan Page LimitedReprinted in 2004 (twice)

Apart from any fair dealing for the purposes of research or private study, or criticismor review, as permitted under the Copyright, Designs and Patents Act 1988, thispublication may only be reproduced, stored or transmitted, in any form or by anymeans, with the prior permission in writing of the publishers, or in the case ofreprographic reproduction in accordance with the terms and licences issued by theCLA. Enquiries concerning reproduction outside these terms should be sent to thepublishers at the undermentioned addresses:

120 Pentonville Road 22883 Quicksilver DriveLondon N1 9JN Sterling VA 20166-2012UK USAwww.koganpagescience.com

© Kogan Page Limited, 2003

The right of Claude Bathias, Hiroshi Fukuda, Kyoshi Kemmoshi, Jacques Renardand Hiroshi Tsuda to be identified as the editors of this work has been asserted bythem in accordance with the Copyright, Designs and Patents Act 1988.

ISBN 1 9039 9649 X

British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library.

Printed and bound by Antony Rowe

RepairingStructures using

Composite Wraps

edited byClaude Bathias, Hiroshi Fukuda,

Kyoshi Kemmoshi, Jacques Renard& Hiroshi Tsuda

KOGANPAGE

SCIENCE

London and Sterling, VA

The 8th Japanese-European Symposiumon Composite Materials

April, 16-17, 2002 - Tokyo University of Science, Tokyo, Japan

Organized by

The Organizing Committee of the Japanese-European Symposium on Composite Materials

Smart Structure Research CenterNational Institute of Advanced Science and Technology

National Institute of Advanced Science and Technology (AIST)

Supported by

Japan Industrial Technology Association (JITA)Embassy of France in Japan

French Association for Composite Materials (AMAC)European Society for Composite Materials (ESCM)

This work was subsidized by the Japan Keirin Associationthrough its Promotion funds from KEIRIN RACE

Organizing Committee

Honorary ChairmenK. KEMMOCHI Shinshu University, JapanC. BATHIAS Conservatoire National des Arts et Metiers, France

ChairmenH. FUKUDAJ. RENARD

Tokyo University of Science, JapanEcole des Mines de Paris, France

Vice-ChairmenK. KEMMOCHI Shinshu University, JapanH. TSUDA National Institute of Advanced Industrial Science &

Technology, Japan

Advisory Board MembersT. KISHI National Institute for Materials Science, JapanI. KIMPARA Kanazawa Institute of Technology, JapanH. MIYAIRI Tokyo Medical & Dental University, Japan

Executive Committee MembersJapanese MembersK. KAGEYAMAM. HOJOQ. NIJ. TAKAHASHIT. ISHIKAWAH. NAGAIK. AMAOKAS. BANDOHK. KIMURAA. HAMAMOTOY. YAMAGUCHIR. HAYASHI

University of TokyoKyoto UniversityKyoto Institute of TechnologyUniversity of TokyoNational Aerospace Laboratory of JapanNational Institute of Advanced Industrial Science & TechnologyFuji Heavy Industries LtdKawasaki Heavy Industries LtdObayashi CorporationIshikawajima-Harima Heavy Industries LtdR&D Institute of Metals & Composites for Future IndustriesJapan Industrial Technology Association

European MembersC. BATHIAS Conservatoire National des Arts et Metiers, FranceC. VISCONTI University of Naples, ItalyC. GALIOTIS University of Patras, GreeceH. LILHOLT Riso National Laboratory, Roskilde, DenmarkMORTON Defense Evaluation and Research Agency, Farnborough, EnglandK. SCHULTE Technical University of Hamburg-Harburg, GermanyA. MARQUES University of Porto, Portugal

This page intentionally left blank

Table of Contents

Introduction 11

Part I. Repairing structures using composite wraps 13

Repairing efficiency of damaged steel structures using composite laminatesK. YAMAGUCHI AND I. KIMPARA 15

RC two-way slabs strengthened with composite materialG. FORET, O. LlMAN AND A. EHRLACHER 25

Structural soundness evaluation of GFRP pedestrian bridgeI. CHOU, K. KAMADA, N. YAMAMOTO, S. SAEKI and K. YAMASHIRO 35

Analysis of the efficiency of composites in improving serviceability of damagedreinforced concrete structuresS. AVRIL, A. VAUTRIN, P. HAMELIN, Y. SURREL 47

Applications of retrofit and repair using carbon fibersK. KIMURA AND H. KATSUMATA 61

Design and repairing of hydraulic valves using composite materialsN. JUNKER, A. THIONNET, J. RENARD 73

lonomer as toughening and repair material for CFRP laminatesM. HOJO, N. HIROTA, T. ANDO, S. MATSUDA, M. TANAKA, K. AMUNDSEN,S. OCHIAI, A. MURAKAMI 83

Polymer adhesives in civil engineering: Effect of environmental parameters onthermomechanical propertiesK. BENZARTI, M. PASTOR, T. CHAUSSADENT, M.P. THAVEAU 91

Overwrapped structures : a modern approach ?M.J. HINTON, J. COOK, A. GROVES, R. HAYMAND and A. HOWARD 105

8 Repairing Structures using Composite Wraps

Development of scarf joint analysis customized system (SJACS) - a guide forstandard analysis of composite bonded repairsT. ITOH, T. TANIZAWA, S. SAOKA 131

Facing progress of composite materials in the maintenance of aircraftC. BATHIAS 141

Possibility of inverse-manufacturing technology for scrapped wood using wrappingeffect in prepreg sheetK. KEMMOCHI, H. TAKAYANAGI, T. NATSUKI and H. TSUDA 151

High temperature behavior of ceramic matrix composites with a self healing matrixJ. LAMON and PH. FORIO 159

Part II. Development and use of smart techniques for strain measurementor damage monitoring 171

Piezoelectric fiber composites for vibration control applications - development,modelling, characterizationY. VIGIER, C. RICHARD, A. AGBOSSOU, D. GUYOMAR 173

Health monitoring system for CFRP by PZTJ. H. Koo, T. NATSUKI, H. TSUDA, N. TOHYAMA and J. TAKATSUBO 183

Characterization of fibres and composites by Raman microspectrometryPH. COLOMBAN 193

Demonstrator program in Japanese smart material and structures system projectT. SAKURAI, N. TAJIMA, N. TAKEDA and T. KISHI 203

Real-time damage detection in composite laminates with embedded small-diameterfiber Bragg grating sensorsN. TAKEDA, Y. OKABE, S. YASHIRO, S. TAKEDA, T. MIZUTANI andR. TSUJI 215

Measuring the non linear viscoelastic, viscoplastic strain behavior of CFRE usingelectronic speckle pattern interferometry techniqueP.J-P.BOUQUET, A.H. CARDON 225

Mechanical property and application of innovation composites based on shapememory polymerQ. NI, T. OHKI AND M. IWAMOTO 237

Piezoelectric fibers and composites for smart structuresA. SCHONECKER, L. SEFFNER, S. GEBHARDT, W. BECKERT 247

Application of metal core piezoelectric fiber - embedded in CFRPH. SATO, Y. SHIMOJO and T. SEKIYA 257

Table of contents 9

Part III. Process inprovement 265

Cure monitoring of composite using multidetection techniqueM. SALVIA, E. CHAILLEUX, N. JAFFREZIC RENAULT, Y. JAYET 267

Mechanical behavior simulation of glass fiber reinforced polypropylene foamlaminatesT. NISHIWAKI and A. GOTO 281

Short-fibre-reinforced thermoplastic for semi structural parts : process-properties.E. HARAMBURU, F. COLLOMBET, B. FERRET, J.S. VIGNES, P. DEVOS,C. LEVAILLANT, F. SCHMIDT 293

Guidelines for a quality control procedure to ensure sound strengthening andrehabilitation of concrete structures using FRPJ.L. ESTEVES and A.T. MARQUES 305

Numerical simulation of reinforcements forming : the missing link for theimprovement of composite parts virtual prototypingP. DELUCA , Y. BENOIT 315

Monitoring of resin flow and cure using electrical time domain reflectometryK. URABE, T. OKABE and H. TSUDA 323

Effects of manufacturing error on stiffness properties of composite laminatesP. VINCENTI, P. VANNUCCI, G. VERCHERY, F. BELAID 333

Mechanical properties of pultruded CFRPs made of knitted fabricsH. FUKUDA, H. WAKABAYASHI, K. HAYASHI and G. OHSHIMA 343

Introduction

The eight Japanese-european symposium which has been held in Tokyo at theuniversity of science of Tokyo in 2002, continues a serie of symposiums the firstone of which was in 1989. The vocation of these symposiums which take placeevery two years alternatively in Europe and in Japan, is to propose an opportunityfor industries and research centers to analyse fundamental questions dealing with theuse of composite materials and structures and to propose solutions.

The main theme of the eight Japanese-european symposium «Repairing structuresusing composites wraps» is a major question for a variety of structural applications,where it is desired to increase service life of their components. If damaged area islocalized and in small compared with the whole size of the structure, it is aneconomical way to arrest the damage extension by a local repair while assuringsafety and reliability. For several years many investigations have been conducted forreinforcement and rehabilitation of damaged infrastructures by their repair andpreservation with fiber reinforced plastics wraps or sheets.

During this symposium differents themes has been discussed concerning :

- Application fields:

- Compensation of civil infrastructures for stabilization or quake-resistance.

- Repair of composite structures.

- Repair of steel structures.

- Different types of reinforcements and techniques of wrapping


- Theoretical and experimental investigations :

- Characterization of the reinforcing effect

- Strength of structural members reinforced with bonding sheets- Design and optimisation strategy

- Use of health monitoring techniques :

- To secure structures and to find optimal processing conditions

- To detect damage state and damage evolution according to different typesof loading.

The participation of different european countries as the Japanese participation duringall sessions has been the opportunity for fruitfull exchanges sometimes leading thissymposium to looks like a workshop during discussion.

To end, the editors would like to thank all institutions, associations, ministry andembassy which supported this symposium and contributed to this successfulmeeting.

Part I:

Repairing structuresusing composite wraps

Repairing efficiency of damaged steelstructures using composite laminate

Koji Yamaguchi — Isao Kimpara

AMS R&D Center, Kanazawa institute of technology3-1, Yatsukaho, Matto 924-0838, Ishikawa Japan

yamagu@neptune. kanazawa-it. ac.jp

kimpara@neptune. kanazawa-it. ac.jp

ABSTRACT; Upgrading was required due to changes in usage of buildings, due to factors suchas deterioration and aging and change in occupancy. Composite (laminate) patch repairingtechnique has gained widespread acceptance as an excellent method for repairing andupgrading of existing structures because of the high strength to weight ratio, ease ofinstallation on site and the improved durability and corrosion resistance of the compositematerial. In this study, composite patch repairing system was applied to crack arrester ofsingle notch steel beam, using two types of carbon fibers: first is a high strength carbon (HS),and second is a high modulus carbon (HM). Effect of externally bonded composite patch onresistance of crack propagation was experimentally and theoretically showed based on linearelastic fracture mechanics. Stress intensity factor and energy release rate in single notchedsteel beam repaired with composite patch are obtained in the closed-form equations. Underfatigue loading, resistance of crack propagation of test specimen repaired with HM washigher than that of test specimen repaired with HS. However, delamination growth of HMwas more rapid than that of HM. Simulation of crack propagation and delamination growthbased on proposed theoretical analysis was in good accordance with experimental result ofthose. It was shown that repairing efficiency and repairing life depend on material propertiesof composite patch and characteristic bonding strength between base material and compositepatch.

KEY WORDS: composite laminate, repairing, fracture mechanics, bonding strength,delamination growth


1. Introduction

Composite patch repairing system has been widely used in several fields. Inaeronautic engineering, composite patch repairing system has been applied to crackarrester in a damaged aluminium plate.

Crack growth behaviour in a plate repaired with reinforcing patch was predictedbased on the finite element analysis and the integral equation approach. The effectsof adhesive thickness and patch thickness on crack growth behaviour were discussed(Ratwani 1977). Under consideration of residual thermal stress induced by thebonding process and effect of bending load, crack growth behaviour in the repairedplate with composite patch was analysed theoretically (Rose 1982). Fromexperimental aspects of composite patch system, effects of adhesive curetemperature, surface treatments before bonding on adhesive fatigue wereinvestigated based on studies on overlap joints, which were simulating repairs andcrack propagation behaviour in patched panels (Baker et al, 1984, Baker 1984).Crack growth behaviour was undertaken to assess the effect on patching efficiencyof disbanding of the patch system and test temperature (Baker 1993). The boundaryelement method is combined with the method of compatible deformations toanalyses the stress distributions in cracked finite sheets symmetrically reinforced bybonded patches (Young et al., 1992). Cracked aluminium plates repaired withcomposites patch was analysed using Mindlin plate finite theory instead of three-dimensional finite element (Sun et al, 1996).

This problem was analysed using three layer technique, in which two-dimensional Mindlin plate elements with transverse shear deformation capabilitywere used for all three layers: cracked plate, adhesive and composite patch(Naboulsi et al., 1996). The effects of location and dimension of debonding area onstrength recovery were compared, as well as strength of panels with a completelybonded reinforcement and cracked panels without any reinforcement were studied(Denney et al., 1997). The effect of geometric nonlinearity on the damage toleranceof the cracked plate was investigated by computing the stress intensity factor andfatigue growth rate of the crack in the plate (Noboulsi et al, 1998). Quite recentlymany studies have evaluated resistance efficiency of crack growth due to compositepatch by using various techniques of finite element method (Seo et al, 2001 etc).

Composite patch repairing system was little applied for steel structure. CFRPsheets are shown to relive the stress concentration at the of circular holes in steelplates (Okura et al, 2000)

In this paper, durability of a single notched beam repaired with externallybonded composite under fatigue loading was experimentally and theoreticallyinvestigated based on fracture mechanics. In the theoretical study, stress intensityfactor and energy release rate are obtained in closed-form equations. In theexperimental study, it is shown that several fracture modes of test specimenschanges due to characteristic of composite patch under static loading, as

Repairing of structures 17

schematically shown in Figure 1. Under fatigue loading, resistance of crackpropagation is evaluated in each composite patch. Crack propagation anddelamination growth is predicted based on the proposed theoretical analysis.Repairing efficiency and repairing life are examined in terms of material propertiesof composite patch and characteristic bonding strength between base material andcomposite patch. Repairing design is discussed based on a change in fracture modesand ambivalent relation between resistance of crack propagation and repairing life.

Figure 1. Schematical fracture mode of a single notch beam repaired withexternally bonded composite patch

Let the Young's modulus of the composite patch be ER. Assuming that through-the-width debonding area with length 2c extends in both directions between theadhesive interfaces symmetrically with respect to the crack plane, the debondedcomposite patch can be represented as a spring with compliance Ad, to form a two-dimensional mechanical model (Kageyama et al, 1995)

2. Experiment

2.1. Test specimen and test method

Mild steel, SS410, was used as base specimen with a single-edge notch. Thewidth of the specimen was 20 mm, the height was 40 mm, and the distance betweentwo supports was 160 mm. Machined notch length was 16 mm and a fatigue crack of2 mm was introduced at the tip of machined notch, as shown in Figure 5. The size ofthe base specimen was chosen according to ASTM E399-83. Two kinds of CFRPsheets (HS: high strength carbon and HM: high modulus carbon) were used toreinforce the single edge notched specimen: Cl-30 (HS), which was made by TonenCorp., was a high strength CFRP sheet and C8-30 (HM) was a high-modulus CFRPsheet. Three kinds of reinforcing sheet thickness by varying ply number were alsoused: 1-ply, 2-ply and 3-ply. In total, 7 kinds of specimen were prepared.


Under fatigue loading, other factors defining the test included a 6-Hz testfrequency, an R ratio of 0.1 and a maximum load of 13000 N with load control. Thecrack length and debonding length between the CFRP sheet and the base materialwere measured.

Figure 2. Size of single notched steel beam repaired with composite patch

2.2. Result

The relationship between crack growth and DK for each test specimen is shownin Figure 3. When reinforcing sheet is thicker, crack growth is also slower underfatigue loading. However, test specimens repaired with 1-ply HS sheet have littleeffect on resistance to crack growth. Test specimens repaired with HM sheet debondoff the base material before the relationship between da/dN and DK extend toRegion 11.

Figure 3. Relationship between apparent stress intensity and crack growth underfatigue loading


KI was analysed based on the proposed theory. It was observed that therelationship between crack growth and AKI of test specimens repaired with all kindsof sheets was very similar to that of test specimen without repair, as shown inFigure 4.

Figure 4. Relationship between true stress intensity and crack growth underfatigue loading

3. Characteristic of delamination growth between steel and composite patch

3.1. CLS test

Bonding strength was evaluated based on energy release rate used by CLS test asshown in Figure 5. CLS test has the advantage of easy measurement of thedebonding length and single lap joint. However, neutral axis was displaced in thistest specimen because this test specimen is not symmetric. Bending moment wasapplied to this test specimen. A new data reduction method to evaluate bondingstrength based on energy release rate was proposed considering bending moment.

Figure 5. Schematic cracked lap shear test


3.2. Result

Figure 6. Relation between debonding growth rate and A energy release raterange.

The relation between A energy release rate and debonding growth rate wasshown in Figure 6.

Open circles in Figure 6 were average of debonding growth rate. Relationbetween fatigue debonding growth rate and A energy release rate was applied toParis law. Paris law was expressed as :

Linear line could be drawn for relation between debonding growth rate and Aenergy release rate range. mc and Cc in material constant were represented as afollow :

Relation fatigue debonding growth and A energy release rate could be elucidatedusing Paris law.


4. Simulation of crack propagation and delamination growth under fatigueloading based on theoretical analysis

Under fatigue loading relation between crack propagation rate, da/dN, and Astress intensity factor, AK, of steel without composite patch based on Paris low wasexpressed as:

Assuming that material properties of steel and composite patch and size of testspecimen is constant, stress intensity factor and energy release rate was obtainedbased on proposed theoretical analysis to substitute load, initial crack length andinitial delamination length. As follow, crack propagation rate and delaminationgrowth rate were obtained to substitute stress intensity factor and energy release ratefor Paris low. Crack propagation rate and delamination growth rate multiplied bynumerical cycle equal crack propagation length and delamination growth length.New crack length and delamination length equal crack length and delaminationlength added crack propagation length and delamination growth length respectively.This process was continued according to record of relation between numerical cycleand load. The flow of process to simulate crack length and delamination length isshown in Figure 7. Crack length and delamination length of repaired steel withcomposite patch could be predicted under fatigue loading. Simulating result of cracklength and delamination length were compared to experimental result, as shown inFigure 8.

When delamination length was small, simulation was not close to experimentalresult. Because proposed theoretical analysis was over the applicable limitation.However, as for crack propagation rate and delamination growth rate, simulation isclosed to experimental result.


Figure 7. Simulation flow of crack length and delamination length under fatigueloading based on proposed theoretical analysis

Figure 8. Comparison between experiment and simulation by crack length anddelamination length of repaired steel with composite patch


5. Repairing design using composite patch

Under fatigue loading, fracture modes of test specimens repaired with HM andHS are summarized as shown in Table 1.

Table 1. Fracture modes of test specimens repaired with HM and HS under fatigueloading

. ... Type of composite patchLoading type _

HM HSRepairing life Short Long

Resistance of crack ,,. , .High low

propagation

Under fatigue loading, in the case of HM composite patch, crack propagation isfurther suppressed, while, delamination growth occurs rapidly, leading to shorterrepairing life. In those structures repaired using composite patch, some trade-offbetween repairing life and effect of resistance of crack propagation have to beconsidered.

Fracture mode and repairing life might be controlled due to material propertiesof composite patch and bonding strength. Therefore, it may be suggested that if asuitable method is established to control material properties of composite patch andbonding strength, fracture mode and repairing life can be controlled to give a certainrequired repairing life.

6. Conclusion

A single edge notched beam repaired with externally-bonded CFRP sheet wasanalyzed under three-point-bending load based on linear elastic fracture mechanics.Reduction of stress intensity factor at the crack tip was calculated theoretically. Theincrease in static and fatigue strength of test specimens reinforced with variousCFRP sheet patches was confirmed experimentally. Resistance effects of crackpropagation under fatigue loading were also evaluated experimentally. Relationbetween debonding growth rate and energy release rate was elucidated using CLStest. Crack length and delamination length of repaired steel with composite patchunder fatigue loading could be predicted based on proposed theoretical analysis.Prediction was shown to be in a good accordance with the experimental result.Repairing design using composite patch was discussed by suggesting that bondingstrength is a key parameter to control repairing life as well as material properties ofcomposite patch.


References

Baker, A.A., Callinan, R.J., Davis, M.J., Jones, R. and Williams, J.G., "Repair of Mirage IIIaircraft using the BFRP crack-patching technique", Theoretical and Applied FractureMechanics, vol. 2, 1984, p. 1-15.

Baker, A. A., "Repair of cracked or defective metallic aircraft components with advanced fibercomposites - An overview of Australian work", Composite Structure, vol. 2, 1984, p. 153-181.

Baker, A.A., "Repair Efficiency in Fatigue-Cracked Aluminum Composites Reinforced WithBORON/EPOXY Patches", Fatigue and Fracture Engineering Material Structure, vol.16, 1993,p.753-765.

Denny, J.J. & Mall, S., "Characterization of Disbond Effects on Fatigue Crack GrowthBehavior in Aluminum Plate with Bonded Composite Patch", Engineering FractureMechanics, vol. 57, 1997, p.507-525.

Kageyama, K., Kimpara, I., & Esaki, K., "Fracture mechanics study on rehabilitation ofdamaged infrastructures by using composites wraps", ICCM-X, Proceeding of ICCM-10,Gold Coast, 1995, p. III-.597-604.

Naboulsi, S. & Mall, S., "Modeling of a cracked metallic structure with bonded compositepatch using the three-layer technique", Composite Structures, vol. 35, 1996, p.295-308.

Naboulsi, S. & Mall, S., "Nonlinear analysis of bonded composite patch repair of crackedaluminum panels", Composite Structures, vol. 41, 1998, p.303-313.

Okura, I., Fukui, T. & Matsuzaki, T., "Application of CFRP sheets to repair of fatigue cracksin steel plate", JCOM: JSMS COMPSITES-29, Proceeding of JCOM: JSMSCOMPSITES-29, Kusatsu, 2000, p. 133-136

Ratwani, M.M., "A Parametric Study of Fatigue Crack Growth Behavior in AdhesivelyBonded Metallic Structures", Journal Engineering Materials and technology, vol. 100,1977,p.46-51.

Rose, L.R.F., "A cracked plate repaired by bonded reinforcements", International Journal ofFracture, vol. 18, 1982, p. 135-144.

Seo, D. C., Lee, J.J. & Jang, T.S., "Comparison of fatigue crack growth behavior of thin andthick aluminum plate with composite patch repair", ICCM-13, Beijing, 18-22 June 2001.

Sun, T.S., King, J. & Arendt, C., "Analysis of Cracked Aluminum Plates Repaired withBonded Composite Patches", AIAA Journal, vol. 34, 1996, p.369-374.

Young, A. & Rooke, D.P., "Analytical of Patched and Stiffened Cracked Panels Using theBoundary Element Method", International Journal Solids Structures, vol. 29, 1992,p.2201-2216.

RC two-way slabs strengthened withcomposite material

G. Foret, O. Limam, A. Ehrlacher

Ecole Nationale des Ponts et ChausseesLaboratoire Analyse des Materiaux et Identification6 et 8 avenue Blaise Pascal, Cite Descartes - Champs-sur-Marne77455 MARNE LA VALLEE

foret@lami. enpc.fr

limam@lami. enpc.fr

ehrlacher@lami. enpc.fr

ABSTRACT: This paper deals with strengthening of reinforced concrete two-way slabs bymeans of composite material thin plates. The strengthened slab is designed as a three-layered plate, bottom layer is composite material, the middle layer is the steel and the toplayer is the concrete. A simplified laminated plate model is used to describe the behaviour ofthree-layered plate supported in four sides, which is subjected to a load in the centre. Theupper bound theorem of limit analysis is used to approximate the ultimate load capacity andidentify the different collapse mechanisms. Lastly, a parametric study is conducted for a RCtwo-way squared slab strengthened with a squared composite thin plate.

KEY WORDS: Limit analysis, collapse mechanism, composite material, strengthening, RC slab.


1. Introduction

The use of externally bonded composite materials for strengthening bridges andother reinforced concrete structures has received considerable attention in recentyears. This approach is applied to a board range of structural members such asbeams, columns, slabs or masonry walls (Meir 87). Because the composite plates areexternally bonded to concrete structures, it is also realised that the bond at theinterface between concrete and composite reinforcements has significant impact onthe overall performance of strengthened structural member. Experimentalinvestigations conducted by (Erik MA & al, 1995), (Shahaway & al, 1996) and(Teng JG, 2000) demonstrate the advantages of strengthening RC slabs withcomposite material. On the other hand, brittle and sudden failure due todelamination of the bonded composite plates or sheets has also been observed.Experimental investigation conducted by (Garden H.N. & al, 1998) on RC beamsstrengthened with composite material shows that two cases take place, the first iscalled "peeling -off failure" where by the whole thickness of the cover concrete hasbeen removed. This failure mode leaves the internal steel exposed and the coverthickness still bonded to the plate. In the second case, the composite plate is leftexposed with no concrete bonded to it, after failure. Failure can occur in twointerfaces. When applied to multi-layered plates, classical Kirchhoff model fails totake in to account shear stress at the interfaces. Failure of multilayered structuresoften occurs by delamination. As consequence, analysis of separation betweenlayers becomes essential for these structures. We design the strengthened RC slabwith composite material as a three layer plate. The upper bound theorem of limitanalysis is applied with a simplified plate model for multi-layered plate (M4)(Ehrlacher A. & al, 1999) (Hadj-Ahmed R. & al, 2001). It is used to describe thedifferent collapse mechanisms with failure modes in layers and interfaces. Anestimate of the ultimate load then follows from the upper bound theorem of limitanalysis by equating the rate of internal energy dissipation in the velocitydiscontinuities sets to the rate of work done by the applied loading as the slabdeforms in this mechanism.

2. Mechanical model

Lets consider a rectangular RC slab strengthened with composite material with athickness h, length 21, a width 2L (Figure 1). A reinforced concrete slabstrengthened by composite material thin plate is designed as a three-layered plate,bottom layer is composite material, the middle layer is the steel and the top layer is

the compressive concrete zone. The respective ply thickness are e1, e2 and e3

(Figure 2). A z-direction load Q is applied in the centre of the plate. The multi-


layered plate is described as an open cylindrical domain Q of R3, with a base

eoe R2and three layers. ( e x , e y , e z ) is an orthogonal base vector of Q with

(ex,ey)eco.

Figure 1. Three layers plate

2.1. Velocity and stress fields

The multi-layered plate model (M4) gives 2n+l generalised velocity fields.

U (UJj with cce {l,2}) is the average displacement rate in ex and ey direction,

W3 is the overall average displacement rate in ez direction. N (NJxp(x,y) with

a,|3e {l,2}) is the membrane stress tensor in layer i, i' ( TJ;'+1 (x,y) with a € {l,2})

is the inter-laminar shear stress at the interface i,i+l.

The generalised strain velocities are given by; e (£afl(x,y)=—(—-+—-) with2 3x,j dxa

a, P e {l,2}) is the in-surface deformation velocity tensor associated to the


membrane stress tensor at layer i, D (Da''l+1 = (U^1 - U^ + -)) is2 aXfx

the generalised velocity tensor associated to the inter-laminar shear stress at theinterface (U+l)-

Figure 2. RC slab strengthened with composite material

2.2. The upper bound theorem of limit analysis

The upper-bound theorem of limit (Johansen, 1962) and (Sale^on, 1983)involves collapses kinematic fields with discontinuities in velocity fields, denoted

/ in layer i and D' in the interface (i,i+l). Velocity fields are kinematically

admissible (KA) when they occur with boundary limits. Let's define the dissipatefunctions as follows:


Where, the internal energy dissipation is given by:n . n

P d=V |[7CT(DI>1+ )]do>+ V rnN(n,y i)ds and the work done by the appliedi=l fa i=l p.v

loading as the slab deforms is given by Q.q(U) . q(U) is the generalised velocity

associated with Q and T? c co is the set of velocity discontinuities.

When Q £ K the slab decomposes.

3. Application to a three-layered plate

3.1. Boundary conditions and collapse criteria

The boundary conditions are given by;

Uj(x,y) = 0 for x = -L, U2(x,y) = 0 for x=-l and W3(x,y) = 0 for (x,y) in

3d), boundary of ft). Let's considering the next criteria on generalised stress fields;

3.2. Collapse mechanisms

We consider collapse mechanisms which result in a velocity discontinuity inlayers and interfaces. As indicated in figure 3, the field to is divided into 4 open setscoi, cos, C0i' and 0)2'. In the case of layer mechanisms, they are rigid regionsintersections. An infinity of collapse mechanisms are considered by varying the

angle a. The velocity q(U) = W3(0) is related to the load Q.


3.2.1. Layers mechanisms:

In the case of layers mechanisms, we suppose that the velocity generalisedshearing strain rates in interfaces are null: DM+I =0, with ie{l,2}. A and B

respectively in layer 1 and layer 2 represents the velocity discontinuities between o)|and oo,' in x-direction. A' and B ' respectively in layer 1 and 2 represents thevelocity discontinuities between 0)2 and 0)2' in y-direction. The KA velocity fieldsare given by:

Figure 3.Definition ofcoi, 0)2. a)/' and (fy'


Velocity strain rate is q(U)= w3(0,y) = w3(0,y) with -y0 < y < y 0 . A sufficient

condition for collapse is Q.q(U) > Pd (U), which is thus given by:

By considering the velocity discontinuities with a layer mechanisms, we get thesufficient two other sufficient conditions for collapse:

3.2.2. Interface mechanism

In this case of interface mechanism, velocity discontinuities is considered ininterfaces.

A sufficient condition for collapse is:


3.2.3. Mechanisms mixed

In this case of mixed mechanisms, the velocity discontinuities is considered inone layer and one interface. We expose the mixed mechanism case concerninglayer 1 and interface (2,3)- We suppose that the rate of generalised shearing strainbetween layers 1 and 2 is null.

When considering velocity discontinuity with a mixed mechanism, we obtain threeother similar conditions sufficient for collapse.

4. Parametric study

We consider a two-way squared RC slab strengthened by composite material. Asquare slab corresponds to 1 = L with a thickness h =7 cm. Failure can occur inlayers 1, 2 and 3 with steel yielding, concrete crushing and rupture of composite thin

plates. Concrete compressive strength is f,!= 30 Mpa. The tension zone in the

concrete under the neutral axis is neglected. The compressive zone thickness is a. It=3

is as a membrane layer and has a resultant force tensor N , which is applied at a

depth of a / 2. An approximated elastic method is used to calculate a. Thecompressive stress tensor strength in concrete layer is given

by:N? l c=N22c=-0.85af^and Nj2c =-0.085af^. The steel reinforcement is the

same in x and y directions and given by As= 2(j)6/m. Steel strength is

f = 500 Mpa . The compressive and tension strength in the steel layer is given by

Nnt = N22« = As fy>Nnc = N nc = -Asfyand N?2c =O.The "peeling-off failure is

designed as a velocity discontinuity in interface (2,3). The composite thin platedebonding is designed as a velocity discontinuity in interface (1,2). The shear stress

strength at the interface (1,2) is TIcu = t2c''

2 = 5 Mpa, and at the interface (2,3) is

Tlc2'3 = T2c

2'3 = 3 Mpa. We consider that, a = 45° which corresponds to a load

minimisation.


We represent (Figure 4) the maximum supported loads for different types ofeight possible mechanisms as function of L. For layer mechanisms the maximumsupported loads remain constant. For mixed and interface mechanisms it increaseswhile L increases.

Figure 4. Ultimate loads for each collapse mechanisms.

5. Conclusion

According to our simplified model, RC slabs strengthened with compositematerial can fail with a layer mechanism or with an interface mechanism or with amixed mechanism. The parametric study shows that for small slab elongation,interface and mixed slab elongation are dangerous. For streamlined slabs, layermechanisms prove to be significant.

The M4-2n+l plate model doesn't take in account failure due to shear stress.This effect can be depicted independently. The parametric study shows that forstreamlined squared slabs the maximum supported load remain constant while theside length 1 increases.


6. References

Erik M.A., Heffernan PJ., "Reinforced concrete slabs externally strengthened withFRP materials" In Taerwe L, editor. Non-metallic FRP reinforcement forconcrete structures, London: E & FN Spon; 1995. pp. 509-516.

Garden H.N., Quantrill R.J., Hollaway L.C., Thorne A.M., Parke G.A.R., "Anexperimental study of the anchorage length of carbon fibre composite plate usedto strengthen reinforced concrete beams", Construction and building materials,12(1998), pp 203-219.

Hadj-Ahmed R., Foret G., Ehrlacher E., "Stress analysis in adhesive joints with amultiparticle model of multilayered materials (M4)", Int. Journal of Adhesionand Adhesives, Volume 21, Issue 4, 2001, Pages 297-307.

Johansen, K.W., "Yield Line Theory", Cement and concrete Association, London,1962.

Meir U., "Bridge repair with high performance composite material." MaterTechnique, 1987;4: 125-8.

Philippe M., Naciri T. Ehrlacher A., "A tri-particle model of sandwich panels",Composite Science and Technology, 1999, p. 1195-1206.

Salen9on J., « Calcul a la rupture et analyse limite », Presses de I'E.N.P.C. Paris.1983,366pp.

Shahawy M.A., Beitelman T., Arockiasamy M., Sowrirajan R., "Experimentalinvestigation on structural repair and strengthening of damaged prestressedconcrete slabs utilizing externally bonded carbon laminates", Composite B 1996;27(3-4): p. 217-24.

Teng J.G., Lam L., Chan W., Wang J., "Retrofitting of deficient RC cantilever slabsusing GFRP strips.", J. Comp. Constr. L 2000; 4(2): p. 75-84.

Structural Soundness Evaluation of GFRPPedestrian Bridge

Iton Chou* — Keiji Kamada** — Naoki Yamamoto***Shoichi Saeki**** — Kazuo Yamashiro*****

* Technology Planning Department, Research & DevelopmentIshikawajima-Harima Heavy Industries Co, LtdShin-ohtemachi Bldg., 2-2-1, Ohtemachi, Chiyoda-ku, Tokyo 100-8182, JAPAN

iton_chou@ihi. co.jp

** Research & Development DepartmentIshikawajima Inspection & Instrumentation Co., Ltd.

*** Structure & Strength Department, Technical Research LaboratoryIshikawajima-Harima Heavy Industries Co, Ltd.

**** Research Institute, Public Works Research Center

***** Roads & Highways Division, North Region Civil Engineering OfficeOkinawa Prefecture

ABSTRACT: This paper introduces the structure of the GFRP (Glass Fiber Reinforced Plastics)pedestrian bridge, to which GFRP was first applied as the primary structure in Japan,constructed in Okinawa Prefecture in April of 2000. Also described mainly of severalstrength tests in this paper are the static loading and the natural frequency tests performed toevaluate the soundness of the bridge structure. The static loading test evaluated the rigidityof the main girders on the bridge without the pavement, and clarified that the shear rigidityin the web had to be considered in addition to theflexural rigidity in the flange. The naturalfrequency test evaluated the primary frequency of the bridge to be approximately 4.6 Hz, andclarified that the bridge did not cause an uncomfortable feeling in people crossing it.

KEY WORDS: pedestrian bridge, GFRP, structural testing, structural soundness, naturalfrequency


1. Introduction

Japan is an island country with many coastlines. Steel and PC (Pre-stressedConcrete) bridges, therefore, are subject to salt damage and the resulting corrosion.Because the PC slabs of the Shingu Bridge (road bridge) in the Noto Peninsula,Ishikawa Prefecture were damaged by salt, it was decided that a new bridge beconstructed, and FRP (Fiber Reinforced Plastics) was adopted as the material of thisnew bridge (Mutsuyoshi 1992). In 1988 this FRP bridge was constructed. Apedestrian bridge made of FRP alone was also built on an experimental basis by thePublic Works Research Institute (Sasaki 1996); it was a demonstration pedestrianbridge and was built on the premises of the Institute.

In Okinawa Prefecture, a road park was recently constructed on the Ikei-Tairagawa route that runs along the coastline (Nonaka 2000, Sangyo Shizai Shinbun Co.2000, Katayama et al., 2000). Because the road park is exposed to salty windthroughout the year, there is concern over the corrosion of the structures built there.A pedestrian bridge running across the road park, therefore, was made using GFRP(Glass Fiber Reinforced Plastics) because it is superior to steel and reinforcedconcrete in corrosion resistance. This pedestrian bridge was completed in April of2000. It is a two-span continuous girder bridge of 37.76 m in length and 3.5 m ineffective width. Because it was the first pedestrian bridge with its main structuralmembers made of GFRP to be built in Japan, some different types of structuralstrength tests were conducted in the design stage to verify the structural soundness(Chou et al., 2001 a, Chou et al., 2001 b, Yamamoto et al., 2001).

This paper describes two of these structural strength tests conducted to verifythe overall rigidity of the pedestrian bridge. One test was a static loading testconducted in the work yard of Sunamachi's Steel Structure Division before afootpath on the pedestrian bridge was paved. Another test was a natural frequencytest conducted on a temporary bridge built in Okinawa Prefecture after a footpathon the pedestrian bridge was paved. This paper reports the results of these tests.

2. Structure of the FRP pedestrian bridge

The appearance of the GFRP pedestrian bridge is shown in Figure 1. Thegeneral bridge arrangement and the cross section of the FRP pedestrian bridge areshown in Figure 2. In Figure 1, the pedestrian bridge is viewed from the side ofOkinawa's main island toward Ikei Island; the left supporting point is P1, the rightsupporting point is P3, and the central bridge footing is P2. These points correspondto the same points on the general bridge arrangement shown at the left in Figure 2.As is apparent from these figures, the pedestrian bridge is secured at the centralbridge footing.


Concerning the cross-sectional structure of the pedestrian bridge, both maingirders that have a channel cross-section are main structural members, as shown atthe right in Figure 2. A truss structure under the deck is joined to these main girders.The main girder is of a three-part structure. One girder is joined to another girderusing joints at positions 4,650 mm to the right and left of the central bridge footing.

Figure 1. Land view of the GFRP pedestrian bridge

Figure 2. Side (left) and cross section (right) views (unit: mm)


3. Static loading test

3.1. Test method

Before a footpath on the pedestrian bridge was paved, a static loading test wasconducted to verify the rigidity of the main girders. Details of the test setup areshown in Figure 3. The cross section of the main girder as well as how a load wasapplied to the main girder are shown in Figure 4. In the work yard of Sunamachi'sSteel Structure Division, a static loading test was conducted on the pedestrianbridge having no tile pavement on the deck slab.

(Note) F1 : Loading position in P1-P2 side (Load : 46.52kN {4744kgf})F2 : Loading position in P2-P3 side (Load : 45.74kN {4664kgf} )v-l~v-6 : Measurement positions for the deflection of main girder

Figure 3. Schematic view of the static loading test (unit: mm)

The deflection of the main girder was measured at six points (v-1 through v-6 inFigure 3) and the longitudinal strain on the flange below the main girder was alsomeasured at twelve points. To measure deflection, a dial gauge that can measure 30mm maximum was used. To measure strain, a two-axis strain gauge with a gaugelength of 10 mm (KFG-10-120D16-11 L30M3S, made by Kyowa Dengyo Co.,Ltd.) was used.


Figure 4. Static loading test apparatus(left: cross section of main girders, right: loading conditions)

A weight was placed on two H-steels to prevent a load from concentrating onthe deck slab and damaging it, as shown in Figure 4. H-steels were placed on thebrace of the truss structure under the deck. A load was applied to one point (Fl) onthe P1-P2 side and to one point (F2) on the P2-P3 side, as shown in Figure 3. Theload values are also shown in the figure. With a load applied to these two points, thedeflection of the main girder and the longitudinal strain on the flange of the maingirder were measured.

3.2. Results of the static loading test and observation

The results of the static loading test are shown in Table 1. In this table,theoretical values and measured values are shown for comparison regarding thedeflection of the main girder and the longitudinal strain on the flange of the maingirder. A distance from the supporting point P1 at the left of the pedestrian bridge isalso shown (see Figure 3).

In calculating theoretical deflection values, the main girder was regarded as abeam having a channel cross-section, and the deflection caused by the shearing ofthe web was added to the deflection caused by the bending of the main girder. Incalculating the deflection caused by the bending of the main girder, the equivalentmodulus of the overall main girder, E (= 14.8 GPa {1510 kgf/mm2}), wascalculated using the equation shown below since the elastic modulus of the maingirder in the longitudinal direction is different from that of the web in the samedirection.


EF: Elastic modulus in tension when longitudinal strain is applied to the flange ofthe main girder

The measured value is 15.2 GPa {1550 kgf/mm2}

Ew: Elastic modulus in tension when longitudinal strain is applied to the web

The measured value is 13.3 GPa {1360 kgf/mm2}

IF: Moment of inertia of area at the flange 3.34 x 1010 mm4

Iw: Moment of inertia of area of the web 8.95 x 109 mm4

I: Moment of inertia of area of the overall main girder 4.24 x 1010mm4

In calculating the deflection caused by the shearing of the web, the shearmodulus of the web measured during the test (G\v = 2.8 GPa {286 kgf/mm2}) wasused. As a cross-sectional area, a cross section of the web alone was considered.

Theoretical values of longitudinal strain on the flange below the main girderwere calculated based on the bending moment at each point from the left supportingpoint P1, assuming that a distance from a neutral axis of bending to the outsidesurface of the flange is half the main girder's height 1600 mm.

The results shown in Table 1 indicate that not only deflection in bending butalso deflection in shearing must be taken into consideration. As shown in Figure 3,some biased cloth layers were added to stacking sequence to supplement the shearrigidity of the web of the main girder. Judging from the results of a static loadingtest, it is presumed that more biased cloth layers should have been used. Becausethe strength of the flange against longitudinal flexural stress must be considered atthe same time, simply increasing the number of biased cloth layers may not producegood results. We need to make further improvements by making good use of thisexperience.

Concerning the longitudinal strain on the flange below the main girder,theoretical values are nearly equal to measured values at some points whilemeasured values are lower than theoretical values at other points; the results varywidely. Because the main girder was made in the hand lay-up process, the actualflange is thicker than a flange that was designed with a uniform thickness of 35 mm,as shown in Figure 2. The thickness of the actual flange also varies more toward thelongitudinal direction. This is thought to be the cause of the difference betweentheoretical and measured values. Overall, measured values are smaller thantheoretical values and therefore it is concluded that there is no problem with therigidity of the main girders of the pedestrian bridge.

Table 1 Results of the static hading test

Distance from the left pier P1 (mm)

Theoretical deflection value (mm)

Bending

Shear

Total

Measured deflection value (mm)

Longtudinal strain on the lower

flang of main girder (u mm)

Theoretical

Measuted

Distance from the left pier P1 (mm)

Theoretical deflection value (mm)

Bending

Shear

Total

Measured deflection value (mm)

Longtudinal strain on the lower

flange of mail girder (u mm)

Theoretical

Measured

4537

-

-

-

-

78

53

22 827

(Joint)

•

-

-

-

-81

•32

4919

(v-1)

4.02

0.77

4.80

5.70

-

-

23727

(Joint)

-

•

-

-

-46

-3

9838

(v-2)

5.49

1.55

7.04

5.74

-

23 982

(v-4)

1.07

1.54

2.61

2.33

-

-

10527

•

-

-

-

181

88

24 927

(Joint)

-

-

-

-

1

4

13527

(Joint)

•

-

•

-

54

44

25 827

(Joint)

• -

•

-

-

37

32

14 427

(Joint)

•

-

-

•

17

18

28288

(v-5)

2.97

3.09

6.06

3.72

-

•

14757

(v-3)2.78

1.89

4.68

3.81

-

28827

•

-

-

-

154

75

15627

(Joint)

•

-

-

-

-34

-4

32594

(v-6)2.44

0.75

3,19

2.62

-

-

16527

(Joint)

•

-

•

-

-72

-29

33327

-

-

-

-

68

42

(Note) 1 : Plus value of deflection represents the downward one.

2 : Plus and minus values of longtudinal strains represent tensile and compressive strains respectively.


4. Natural frequency test

4.1. Test method

Because it was decided that the FRP pedestrian bridge be first built and then itsfootpath be paved, a natural frequency test was conducted on-site in OkinawaPrefecture to verify the primary natural frequency. Details of the test setup areshown in Figure 5. How a natural frequency test was conducted is shown in Figure6. As acceleration sensors, a servo-type, low-frequency vibroscope (AVL-25A,Akashi Co., Ltd.) and a detector (V401BR, Akashi Co., Ltd.) were used. The settingof these acceleration sensors is shown in Figure 6; after aluminum foils wereaffixed to the tile pavement, the acceleration sensors were set and secured using aninstant adhesive. As shown in Figure 5, the acceleration sensors were set at ninepoints in the longitudinal direction of the pedestrian bridge and vibration wasapplied to six points (@, CD, ®, ©, (2), and ®). Measurement was made and datawas collected at these six points. To cause vibration, a man jumped on each point,as shown in Figure 6, and measurement was made three times at each of these sixpositions.

® ~ ® Accelerometer installation positions

Figure 5. Schematic view of the natural frequency test (unit: mm)


Figure 6. Natural frequency test apparatus(left: loading by the jumping, right: setup of acceleration sensors).

4.2. Test results and observation

The result of spectrum analysis based on data collected at points © is shown inFigure 7. The first peak value appeared at a frequency of 4.60 Hz.

Assuming that the pedestrian bridge is a simple beam having one cross section,primary and secondary natural frequencies can be calculated theoretically (JapaneseSociety of Mechanical Engineers). Providing that the rigidity of a beam is El, themass per unit length is p and the length of a beam is L. The natural frequency fwhen a beam vibrates in a traverse direction can be expressed, using the equation[2]:

Here, A, is a coefficient and a combination of fixed support and simple supporttechniques are used at ends of a beam. In this setup, the primary frequency is A,=3.927 and the secondary frequency is A, =7.069. The length L is 19.677 m on theP1-P2 side and it is 17.223 m on the P2-P3 side.

If El is defined as the design rigidity of a main girder (E =11.8 GPa {1200kgf7mm2}) and p is defined as the actual measured weight, natural frequencies onthe P1-P2 and P2-P3 sides are as follows:

P1-P2 side - primary: 4.58 Hz, secondary: 14.84 Hz

P2-P3 side - primary: 5.98 Hz, secondary: 19.38 Hz

After this result is examined relative to the results shown in Figure 7, theprimary natural frequency of the pedestrian bridge should be about 4.6 Hz. A peakvalue that appeared at 6.45 Hz in Figure 7 is considered to be equivalent to theprimary natural frequency on the P2-P3 side.


To ensure that people feel secure when walking on a pedestrian bridge, it mustbe designed so that its primary natural frequency is controlled well belowapproximately 2 Hz (1.5 to 2.3 Hz) (Japan Society of Roads & Highways 1981). Itis concluded from the results of the static loading test conducted that people can feelsafe and secure when walking on the GFRP pedestrian bridge being discussed inthis paper.

Figure 7. An example of the results on spectrum analysis

5. Conclusions

In developing the GFRP pedestrian bridge, a static loading test was conductedbefore its footpath was paved and a natural frequency measurement test wasconducted after its footpath was paved. We found from the results of the staticloading test that both the flexural rigidity of the flange and the shear rigidity of theweb must be taken into consideration to make proper rigidity design for the maingirder and that the longitudinal strain on the flange below the main girderconstitutes no problem because measured values are mostly smaller than theoreticalvalues. We also verified from the results of a natural frequency measurement testthat the primary natural frequency of the pedestrian bridge is about 4.6 Hz and that4.6 Hz is not the level of frequency that makes people feel unsafe (1.5 to 2.3 Hz)when walking on it.


Acknowledgments

The authors would like to thank to the cooperation and advice of the staff ofAsahi Glass Matex Co,, Ltd for completing this work. We would like to extend oursincere appreciation to Mr. Nayomon Uno, the chief engineer, and Mr. NobuhikoKitayama, the staff engineer, at the Bridge & Road Construction Division, also tothe staff of the Structure & Strength Department at the Research Laboratory, and thestaff at the Instrumentation System Group of Ishikawajima Inspection &Instrumentation Co., Ltd.

References

Chou I., Kamada K., Saeki S., Yamashiro K., "Experimental Evaluation on the Rigidity ofMain Girders and the Natural Vibration Frequency in FRP Pedestrian Bridge", IHIEngineering Review, vol.34, no.4, Oct. 2001 a, p. 101-105.

Chou I., Kamada K., Saeki S., Yamashiro K., "Experimental Evaluation on Joints in FRPPedestrian Bridge", IHI Engineering Review, vol.34, no.4, Oct. 2001 b, p. 110-113.

Japan Society of Mechanical Engineers. Mechanical Engineers' Handbook, A3 Mechanicsand Mechanical Vibrations (in Japanese).

Japan Society of Roads & Highways, Specifications for Pedestrian Bridges, 1981 (inJapanese).

Kitayama N., Saeki S., Yamashiro K., "Schema of FRP Pedestrian Bridge Constructed inOkinawa Prefecture", Proceedings of the 55th Annual Conference of the Japan Society ofCivil Engineers, I-A, no.230, Sept. 2000 (in Japanese).

Mutsuyoshi H., "Application of FRP for Construction Structures", Journal of Japan Societyfor Composite Materials, vol.18, no.3, May 1992, p.95-101 (in Japanese).

Nonaka K., "Zoom Up Bridge - Construction of FRP Pedestrian Bridge in Ikei-TairagawaRoute (Okinawa Prefecture) - The First Application of Plastics for Main Structures",Nikkei Construction April 28th, 2000, p.28-32 (in Japanese).

Sangyo Shizai Shinbun Co., The Engineering Plastic Journal, no.712, June 2000 (inJapanese).

Sasaki I., "Application of FRP for Main Structures of Pedestrian Bridge", Civil EngineeringLetters, vol.38, no. 11, Nov. 1996, p.4-5 (in Japanese).

Yamamoto N., Chou I., Saeki S., Yamashiro K., "Analytical Evaluation on the Joint Structureof the Main Girder in FRP Pedestrian Bridge", IHI Engineering Review, vol.34, no.4,Oct. 2001, p. 106-109.

Analysis of the Efficiency of Composites inImproving Serviceability of DamagedReinforced Concrete Structures

Stephane Avril* — Alain Vautrin* — Patrice Hamelin — YvesSurref**

* SMS/MeM, Ecole Nationale Superieure des Mines de Saint Etienne, 158 CoursFauriel, 42023 Saint Etienne Cedex 2, France.

[email protected]@emse.fr

** L2M, Universite Claude Bernard Lyon I, 43 boulevard du 11 Novembre 1918,69622 Villeurbanne Cedex, France.

hamelin@iutal2m. univ-lyonJ.fr

*** BNM-1NM/CNAM, 292 rue Saint Martin, 75141 Paris, France.

[email protected]

ABSTRACT: The mechanical behaviour of Steel-Reinforced-Concrete beams strengthened withCFRP laminates bonded on the soffit is addressed. The displacement fields over the lateralsurface of the tested beams are measured with a grid method. It is shown that the behaviourat the global scale is well assessed by the beam theory of Bernoulli. It permits to calculate theaverage longitudinal strains in each component just from the curvature and the position of theneutral axis. The displacement fields are also utilized to locate cracks and to measure theirwidths. The method is applied to compare cracking in a damaged concrete beam before andafter bonding a composite laminate. It leads to an interesting characterization of crackbridging induced by the repair and it proves that the serviceability has been enhanced.

KEY WORDS: reinforced concrete, repair with composites, crack bridging, optical method.


1. Introduction

Strengthening or repairing degraded Steel-Reinforced Concrete (RC) structureswith Carbon Fibre Reinforced Polymers (CFRP) is nowadays gaining an increasingsuccess. The technique is well established practically and several commercialprocesses are available all over the world (Ferrier 1999). On the other hand,universal design guidelines are not yet available even if most of the task groups[AFGC 2001] emphasize the need for special requirements to utilize these materialsin the field of civil engineering.

Rehabilitation of concrete can be related either to failure considerations or toserviceability considerations. The latter is addressed here. Under service loads,stresses should be limited to prevent the yielding of steel re-bars. Besides, widecracks may be harmful to internal steel (corrosion). According to several authors(Triantafilou et al., 1992, Raoof et al., 1997), damage mechanisms near the cracks,occurring before yielding of the steel, can also be responsible for the debonding ofthe laminate.

General results on the behaviour of strengthened or repaired beams listed in theliterature (Quantrill et al., 1998, Mukhopadhyaya, 1999) show an increase in thestiffness, a reduction of tensile strains in concrete, a delayed appearance of concretecracks and a narrower crack spacing. However, serviceability analyses are currentlymainly qualitative. Models involving tension stiffening or crack bridging are scarce.Refined experimental studies are still necessary to understand local phenomena andtheir influence onto the global behaviour of the structure.

The present study focuses on this problem. The grid method (Surrel, 1994) isused to obtain global and local information on the mechanical behaviour underservice loads of cracked RC beams repaired with composites.

2. Experimental procedure

2.1. Specimens

The tested specimens are small-scale beams for more convenience and testfacilities. Their design is governed by the similitude theory which leads to thedifferent scale factors to be used with respect to the real-scale reference model.These factors are obtained on the basis of a dimensional study (Ovigne et al., 2000).The basic scale factor for lengths is 1/3.


Steel bars and stirrups dimensions (Figure 1) as well as aggregate size andgranulometry of the micro-concrete are also controlled to match the reference valuesmultiplied by the scale factors.

The real-scale model is a 2000x250x150 mm reinforced concrete beam designedto fail in flexure by steel yielding and concrete crushing.

Figure 1. Details of the specimens and experimental set-up

2.2. Testing program on RC beams

A four-point bending test is carried out on five reinforced concrete beams whoseinternal structure has been described (Figure 1). The main objectives of this first testare:

- to create tensile cracks in order to simulate the degradation,

- to characterize the mechanical behaviour of cracked beams beforestrengthening.

Each test is stopped at 60% of the load corresponding to the rebars yielding.Then, the beams are unloaded. A second bending test is carried out directly up tofailure on one of the precracked beams. This beam is used as the referenceunstrengthened beam.

The other four, out of the five pre-cracked beams, are repaired with a compositelaminate bonded on the bottom surface. The bonded CFRP laminate is made of aunidirectional high modulus carbon fibres taffetas (330 g/m2 reference Hexcel46320) and epoxy resin (Ciba LY 5052). It is directly polymerised on the specimen,the first epoxy resin layer working as the bonding joint. The thickness of thebonding joint is 0.4 mm and the thickness of the composite is 0.4 mm. A tensile testcarried out on such a laminate provides a Young modulus of 55 000 MPa.


After polymerisation, the repaired beams are loaded in flexure up to the steel re-bar yielding load. The bending test and instrumentation are the same as the one usedin the case of unstrengmened beams.

2.3. Instrumentation

Each beam is instrumented with a 145-mm-long Mecanorma Normatex 3135 bi-directional grid on a lateral surface over the constant moment span (Figure 1).

A grid is a set of parallel black lines drawn over a white surface. The process toput it on the surface is very simple : the grids are directly deposited by transfer. Abi-directional grid is then the superposition of two perpendicular unidirectionalgrids. The only requirement to fulfill is that the grid is integral with the specimen.

In-plane displacements of the surface can be deduced from the deformation ofthe grid lines (Surrel 1994). Several papers have already been published on thismethod and the good setting of the parameters of it. Previous studies have conductedto the validation for its application to concrete structures (Avril et al., 2001).

In our experiments, the grid pitch, i.e. the distance between two contiguous lines,is 571 um. We use a numeric BASLER A113 CCD sensor with 1200x1200 pixelsconnected to a PC in order to grab the images. The displacement computation isperformed with an in-house software called Frangyne2000.

The resolution of the measurement, i.e. the smallest displacement we are capableto measure, is about 2 or 3 um, depending on the quality of the grid transfer. Thespatial resolution (Surrel 1999), i.e. the length of an individual sensor, is 1.2 mm.

3. Results

3.1. General aspect

Examples of displacement fields have been plotted in Figure 2a and 2b. Thesedisplacements are similar whether the beam has been repaired or not. In particular,discontinuities of ux field are always linked to the presence of a crack, as it wasshown in a previous study (Avril et al., 2002-1).

It can be noticed that the cracks never propagate beyond a certain height, whichis actually the location of the neutral axis of the beam. Once the neutral axis hasbeen determined, the field can always be divided into two main areas:

- above the neutral axis, the compressive area governed by the mechanics ofcontinuous media.


- below, the tensile area where the material is discontinuous and the mechanicalbehavior is mainly controlled by crack opening mechanisms.

The main effects of the composite are analysed in the following sections,focusing firstly on the global curvature, then on strains and finally on crack growthand opening in the tensile area.

Figure 2a. Example of an experimental ux field.

Figure 2b. Example of an experimental uy field.


Figure 3a. Localisation of pixels where the absolute deviation of ux experimentalfield from the beam model is less than 2 um for an unstrengthened RCbeam.

Figure 3b. Localisation of pixels where the absolute deviation of ux experimentalfield from the beam model is less than 2 um for a strengthened RCbeam.

3.2. Global behaviour

At any step of loading, the actual beam is modelled by an equivalent continuousand homogenous beam verifying the theory of Bernoulli. The displacement fields ofthe modelled beam are only governed by two parameters: the curvature x and theneutral axis position Z. The equations of beams lead to:


where: ux(x,y) is the modelled horizontal displacement field, uy(x,y) is the modelledvertical displacement field, R0 is the local rotation at the origin, ux0 is the horizontaldisplacement at the origin - ux0 = ux(0,0) - and uy0 is the vertical displacement at theorigin - uy0 = uy(0,0).

All the parameters are identified from the experimental fields in the compressivearea. The purpose is to compare experimental and modelled fields. The comparisonis made for ux field by plotting locations where the modelled and the experimentalfield ux are equal (Figure 3a and Figure 3b). The following criterion is used : at onepixel, if the absolute difference between the experimental and the modelleddisplacement is less than 2 um, then the pixel is black, else it is white. A cut offvalue of ±2 um has been chosen because it is the resolution of the optical method.

It can be noticed that most of pixels in the compressive area respect the criterion,both before and after repair (Figure 3). It means that the mechanics of this part of thebeam is not modified drastically by the composite effect and also that it is wellsuited to the identification of the global curvature x. Only the stiffness is slightlyincreased. For example, by investigating the moment / curvature diagram, one cannotice that for the same global curvature, the applied bending moment curvature isincreased (Figure 4).

3.3. Semi-global behaviour

In the tensile area, some rare locations are detected where the modelled and theexperimental field are similar. They are mostly concentrated in narrow strips alignedperpendicularly to the length of the beam (Figure 3). The cross sections located inthese strips are the only ones to remain plane (even if only the surface displacementis measured, it is assumed that the whole cross section remains plane : the internalbehaviour will be discussed further). The trends of ux(x , y=ycs) is linear in thevicinity of the located strips, meaning that only the cross section at the middle ofeach strip can be considered as remaining plane.

In the tensile area, the modelled longitudinal strain has no physical significancefor concrete, because stretching is rather an accumulation of crack widths than a realmaterial straining. Thus it is quite natural that only a few cross sections remain planeafter bending, since deformation modes are really different from the top to thebottom of the beam. However, the existence of several plane sections observedexperimentally shows that the behaviour is globally similar to the one of a classicalbeam.

As a matter of fact, at any cross section where there is no compatibility betweenthe displacement of the cover concrete and the displacement of the reinforcement,sliding induces a shear transfer. Tensile and shear strains result in concrete. It provesthat at any cross section where there is concordance between experimental data andthe model, the sliding of reinforcements must be zero. Accordingly, the average


longitudinal strain, over the distance separating two contiguous cross sectionsamong the only ones to remain plane, is equal to the modelled longitudinal strain atthe same height.

Therefore, the following formula can be used for concrete in compression andsteel:

On the other hand, the average longitudinal strain over the composite is lessbecause of the residual strain when it was bonded:

Figure 4. Moment / curvature diagram for one tested beam


Figure 5. Stages of cracking in a RC beam repaired with composites.

3.4. Local behaviour

The longitudinal displacement field is utilized for crack visualisation andcharacterisation (Avril et al., 2002-2). The results obtained for unrepaired andrepaired beams are quite different (Figure 5). This is the consequence of theoccurrence of two types of new cracks:

- most of them are oblique shear cracks : they do not propagate up to the neutralaxis but they are deflected towards the neighbouring pre-existing crack at the levelof the internal re-bars. They are called tributary cracks.

- a few are vertical and appear halfway between two pre-existing contiguouscracks. They are not deviated in their propagation towards the neutral axis. Theymay result of tensile stresses in the concrete induced by the action of crack bridgingof the composite laminate.


The creation of new cracks, especially tributary ones, is a phenomenon specificto repaired beams. No new cracks are detected when the reference unrepaired beamis loaded up to failure.

Thus, the tensile strain exx(ycs) of the modelled beam corresponds in the tensilearea to the accumulation of two types of crack widths : large ones for pre-existingcracks and smaller ones for the new ones (Figure 5). Crack width can be linked tothe global curvature by the following formula:

where: W(ycs) is the width of a vertical crack at the height ycs, D is the distanceseparating the two localised plane sections which surround the investigated crack(Figure 4), Q(ycs) is homogeneous to a strain: it takes into account either thecontribution of real straining of concrete before the creation of tributary cracks, orthe contribution of the new cracks opening. It results from the crack bridging by thecomposite laminate, phenomenon that is mostly effective near the soffit.

It is worth noting the main difference between unrepaired and repaired beamslies in Q(ycs). For steel-reinforced concrete beams, Q(ycv) is about zero. On the otherhand, for repaired beams, Q(ycs) can represent 20% of the modelled equivalent strainexx(ycs)- However, when ycs is above the position of steel rebars, exx(ycs) is negligible:the contribution of Q(ycs) is mainly concentrated in the cover concrete.

4. Discussion

The objective is here to characterize the range of serviceability improvementinduced by bonding composite laminates. Two points are addressed:

- the stresses in concrete and steel,

- the maximum width of cracks.

The stresses in concrete and steel are derived from the strains multiplied by therespective modulus. The strains in both materials are assessed directly from themodelled beam, because the experimental results have proved that the modelledstrains and the experimental strains are similar in average. Finally, stresses areproportional to the curvature x such as:


where os is the steel stress, oc is the maximum concrete stress, Es the Youngmodulus of steel, Ec the Young modulus of concrete, d is the distance from the soffitup to the steel rebars location and h is the height of the beam.

The maximum width of cracks is given by Equation [5]. It is thus generallyinferior to xDZ for a repaired beam because of tributary cracks. Moreover, D may belower for a repaired beam than for an unrepaired one because of new crackoccurrence. However, for simplicity purpose, we can keep xDZ as an upper limit forcrack widths in a repaired beam. Like the stresses, the crack width upper limit is alsoproportional to the curvature x.

Therefore, a relevant criterion for characterizing serviceability improvementinduced by CFRP reinforcement is the loading increase that the structure can sustainafter repair for a given curvature. The rate is 10% for a curvature of 0.045 m-1 in theexample plotted in Figure 4. This means that if the loading is increased of 10% afterrepair, crack maximum widths will not be affected just thanks to the strengtheningeffect. It is quite important since wide cracks may be harmful with regard topenetration of moisture, salt or oxygen and then induce steel corrosion. Furthermore,the stiffening effect is also significant with regard to stresses and may increase thefatigue strength of the whole structure. Finally, it shows that the durability of a beamcan be increased by bonding a composite plate on its soffit.

However, this study is only a preliminary study and two points should beexamined more carefully:

- the behaviour of the structure is strongly non-linear, because of internal frictionbetween steel and concrete. Moreover, the loading of a real construction includes forthe most part its own weight. Both statement have consequences on the stiffeningeffect of the external reinforcement.

- the mechanical properties of composites and adhesives are time-dependent.Their damage or ageing may reduce the stiffening rate and annihilate the durabilityenhancement (Ferrier 1999).

The former point is addressed here (Figure 4). It can be noticed that the curvaturediminution at a given moment is mainly dependent of the residual curvatureremaining after unloading. The stiffening effect could be improved if the residualcurvature was reduced, provided that the slope after strengthening was not changed.

Finally, the current results highlight the strong dependence of strengthening onthe history of the damaged structure. This dependence is being characterisedpresently in our laboratory in order to supply relevant guidelines for the design offlexural repairs with composites.


5. Conclusion

Five RC beams have been investigated. They have been cracked in order tosimulate service conditions of life of a real structure, and then strengthened with aCFRP laminate bonded on the bottom surface.

Every beam has been equipped with grids over the lateral surface in the midspanarea. An in-house developed optical method, called the grid method, has beenutilized to extract displacement fields from the grid deformation. The analysis ofdisplacements fields has led to the main following conclusions:

- the grid method reveals to be well adapted for the study of cracks. The crackwidth is measured accurately by calculating the height of discontinuities over thefield. A comparison between repaired and unrepaired beams shows that the effect ofrehabilitation by CFRP laminates is a significant reduction of crack widths.

- the detection of plane sections proves that the repair do not modify drasticallythe behaviour of the structure. The parameters of an equivalent homogenous beamcan be identified, meaning that a beam of Bernoulli is still relevant to model themechanical behaviour of the repaired cracked structure.

- the moment/curvature curve of the identified modelled beam is complex. Themain effect of the strengthening is a slight stiffening. However, the effectiveness ofthe stiffening effect strongly depends of the loading history of the damagedstructure.

This study has provided a first insight in composite potentiality for improvingserviceability and durability of constructions and buildings. The objective is now tovalidate the results on full-scale specimens.

Acknowledgement

We are grateful to the "Region RHONE-ALPES" for its financial support to ourresearch work within the framework of the regional project: "rehabilitation of civilengineering structures with composite materials: modelling of repaired crackedbeams".

6. References

AFGC, "Repair and strengthening of concrete structures by means of composite materialswith organic matrix", in: comptes rendus de l'Association Fransaise de Genie Civil,Recommendations of the first task group concerning materials testing and manufacturing.


Avril S., Ferrier E., Hamelin P., Surrel Y., Vautrin A., "Reinforced Concrete Beams byComposite Materials : Optical Method for Evaluation", proceedings of the InternationalConference on FRP Composites in Civil Engineering, CICE 2001, Ed. J.G. Teng,Elsevier, 2001, Vol. 1, p.449-456.

Avril S., Vautrin A., Hamelin P., "Mechanical behaviour of cracked beams strengthened withcomposites: application of a full-field measurement method", Concrete Science andEngineering, submitted January 2002.

Avril S., Vautrin A., Surrel Y., "Grid Method, Application to the characterization of cracks",Experimental Mechanics, submitted March 2002.

Ferrier E., "Composite-concrete interface behaviour under thermo-stimulated creep andfatigue loading. Application to estimated calculation of RC beam durability", Doctoralthesis UCB Lyon I, 1999.

Mukhopadhyaya P., Swamy R.N, "Debonding of carbon-fiber-reinforced polymer plate fromconcrete beams", Proc. Inst. Civ. Engrs., Structs. & Bldgs, vol.134: p.301-317,1999.

Ovigne P.A., Massenzio F., Hamelin P., "Mechanical behavior of small scale reinforcedconcrete beams externally strengthened by CFRP laminates in the static and dynamicdomains", Proceedings of the 3rd International Conference on Advanced CompositeMaterials in Bridges and Structures, Ottawa, 2000.

Quantrill R. J., Hollaway L.C., "The flexural rehabilitation of reinforced concrete beams bythe use of pre-stressed advanced composite plates", Composite Science and Technology,vol.58: p. 1259-1275, 1998.

Raoof M., Zhang S., "An insight into the structural behavior of reinforced concrete beamswith externally bonded plates", Proc. Inst. Civ. Engrs., Structs. & Bldgs, vol.122: p.477-492, 1997.

Surrel Y., "Moire and grid methods in optics : a signal-processing approach", proceedings ofSPIE, vol.2342: p.213-220,1994.

Surrel Y., "Fringe Analysis", in Photomechanics, pp. 57-104, P.K. Rastogi Ed., Springer,1999.

Triantafillou T.C., Plevris N., "Strengthening of RC beams with epoxy-bonded fiber-composite materials", Mater. Struct, vol.25: p.201-211, 1992.

Applications of Retrofit and Repair usingCarbon Fibers

Kohzo Kimura — Hideo Katsumata

OBAYASHl Corporation Technical Research Institute, Tokyo, Japan

[email protected]

KATUMATA @o-net.obayashi.co.jp

ABSTRACT. Oboyashi Corporation has been studying application techniques using carbon fiber

since 1985. In the civil engineering of Japan, fiber reinforced plastics have been used for theretrofit and repair of structures after the Hansin-Awaji earthquake in 1995. In this paper, thesummary of the retrofit techniques developed by Obayashi Corporation, called"Carbon fiberRetrofitting System (CRS)" and "Torayca laminate system", and some applications usingthese techniques are described.

KEY WORDS :: carbon fiber, CFRP laminate, retrofitting, repair, concrete structure

1. Introduction

Research and development of the concrete structures using the reinforcementsconsist of high-strength fibers have been underway since the early of 1980's inJapan. In 1986, the concrete curtain wall, pre-cast concrete outer panel mixedchopped carbon fiber, was installed, and a pre-stressed concrete bridge using carbonfiber reinforced plastic (CFRP) for the pre-stressed strand was constructed inIshikawa prefecture in 1988 (Kimura et al., 2000).

In the civil engineering of Japan, fiber reinforced plastics (FRP) reinforcementsare mainly used for three objects, because of high-strength, light-weight and non-corrosion. The first is on behalf of the conventional reinforcement bar and thestrand. The second is the retrofit material for existing concrete structures. Thedemand of the carbon and the aramid fiber sheets for this use has been increasedyear by year since 1995, after the Hansin-Awaji earthquake. The last is on behalf ofthe steel members such as the steel pipe and the shape steel.

Since 1985, Obayashi Corporation has been studying application techniques ofcarbon fiber (CF), including several cooperative studies with material manufactures(Katsumata et al., 1988, 1996, Kobatake et al., 1993, Hagio et al., 1998). For


retrofitting and repair of existing reinforced concrete structures, we use three typesof carbon fiber products, those are CF strand, CF sheet and carbon fiber reinforcedplastics (CFRP) laminate. The retrofitting and repair techniques using these productsare the following three;

- Shear retrofitting by CF strands winding or CF sheets wrapping (Figure 1)- Flexural retrofitting by CF sheets gluing or CFRP laminates bonding (Figure 2)- Combination of the above two techniques

In this paper, the summary of applications of retrofit and repair using CFdeveloped by Obayashi Corporation are described.

Figure 1. Shear retrofitting by CF strands winding or CF sheets wrapping

gluing of CF sheets bonding of CFRP laminates

Figure 2. Flexural retrofitting by CF sheets gluing or CFRP laminates bonding


2. Seismic retrofitting method of existing concrete structures

2.1. Retrofitting of concrete column (Katsumata et al., 1988, 1996)

Some existing reinforced concrete columns do not have enough shear strengthand ductility against a several earthquake shock. We have developed a new seismicretrofitting method using carbon fiber called "Carbon fiber Retrofitting System(CRS)" in collaboration with Mitsubishi Chemical Corporation (Figure 1). Inprocedure, carbon fiber strands consists of 12,000 monofilaments or carbon fibersheets are wound onto the surface of the existing columns. The carbon fiber strandpasses through resin bath filled epoxy resin and is winding around the concretestructure. And carbon fiber sheet are placed by hand with the adhesive on theconcrete surface in the transverse direction.

This technique improves the earthquake-resistant capacity of the columns asfollows:

- Increase in shear strength- Improvement of ductility- Increase in compressive capacity

This method has the following advantages, comparing with the current methods.

- It is easy to provide required shear and ductile capacities.- Retrofit works do not influence the stiffness of the retrofitted columns.- It is possible to minimize increase in weight accompanied with retrofitting.-There is no need of skillful workers in construction.- It is easy to control the quality of construction.

The winding work of CF strand is carried out using an automatic windingmachine shown in Figure 3 in order to save labor and cost. This machine is alsoapplicable for retrofitting of bride columns.

Figure 3. CF strand winding machine


The carbon fiber winding machine consists of major four parts as shown below.

- Supporting wheel (lower ring) suspended by the suspending-chain and movedup and down

- Rotating wheel (upper ring) coupled with the supporting wheel and rotatedwith the epoxy resin impregnation unit.

- Suspending chain to suspend the supporting wheel from the ceiling.- Epoxy resin impregnation unit to impregnate epoxy resin with the carbon

fiber.

Application: Osaka Castle (Katsumata et al., 2001)

Osaka Castle is one of the most famous historical buildings in Japan (Figure 4).The building age is over 70, so many parts were damaged. The structural evaluationalso revealed that the building was not strong against the considerable maximumearthquakes in future. Thus, the building was retrofitted, including structuralstrengthening.

"The Carbon fiber Retrofitting System" was applied for short columns. CF sheetare placed and glued by hand with impregnating epoxy resin (Figure 5). Cure forFRP fabrication is carried out on site. However, for long columns, CF strandwinding is applied because CF winding is superior on work speed and qualitycontrol and suitable for large-scaled applications. CF winding employs a windingmachine shown Figure 3 and CF strand supplies toward the column, impregnatingepoxy resin and rotating around the column.

Figure 4. Osaka Castle Figure 5. Column reinforced by CF sheets


2.2. Retrofitting of concrete chimney (Kobatake et al., 1993)

Some of existing reinforced concrete chimneys in Japan have often damagedand sometimes broken at the height of 2/3 or more of the total height when a largeearthquake attacked. This is because the previous design regulations did notdemand enough flexural strength in the top part of chimneys. Longitudinalreinforcement should be performed for seismic retrofitting.

In 1987, Obayashi Corporation have developed in collaboration withMitsubishi Chemical Corporation a retrofitting method for increasing flexuralcapacity of existing chimneys. The method employs CF sheets to longitudinallyglue onto the concrete surface in order to provide flexural capacity needed forchimneys. It also employs CF strands to transversely wind on the outside of theglued CF sheets in order to confirm the bond between the CF sheet and theconcrete surface and to prevent concrete from crack by the thermal stress owing tosmoke exhaustion. A special lift scaffold was developed for the retrofit works(Figure.6, Figure 7).

This method overcomes the difficulties arising from the current retrofittingmethods. The technical merits are summarized as follows.

- The operation of the chimney is not disturbed because the outside of the chimneyis retrofitted.

- Increase in weight accompanied with retrofitting is negligibly small because CFsheets, which are very light weight, are glued with epoxy adhesive.

- High retrofitting effect is obtained and the cost of retrofitting is reduced.- The durability of concrete is improved because the CF sheets cover the outside of

the concrete surface and isolate from corrosive gas, acid rain and sea water spray.

In 1991, Japan Building Disaster Prevention Association made a technicalevaluation for this retrofitting technique. The evaluation of this association meansthat the high technical significance of this CF gluing technique is publiclyauthorized. Obayashi Corporation has already retrofitted over 55 chimneys for 10years from 1991 to 2001.

For another application, as shown in Figure 8, a Japanese shrine gate "Torii" wasrepaired using CF sheets.


Figure 6. Scaffold lift for field work Figure 8. Repair of "Torii'

Gluing of CF sheet Winding of CF strand

Figure 7. Sates of the retrofitting on chimney

3. Retrofit and repair method for existing beam and slab (Hagio et al., 1998)

The retrofit and repair method against flexural force using CFRP laminate hasdeveloped by Obayashi, Toray and Sika Japan in 1996. This method is called"Torayca laminate system". "Torayca" is a registered trademark of high performancecarbon fiber manufactured by Toray.


3.1. Material

The CFRP laminate shown in Figure 8 consists of high strength and highmodulus carbon fiber in an epoxy-based thermoset matrix and has 50mm width andthree kinds of thickness, 1.0mm, 1.5mm and 2.0mm. The carbon fibers in thelaminate with 1.0mm thickness are equivalent in 4 or 6 layers of CF sheet used inpractice.

The tensile strength of CFRP laminate is 2.4 kN/mrn2 and the elastic modulus is155 kN/mm2. CFRP laminate is prefabricated by pultrusion process and after curethe contact face with the adhesive is pre-treated with sandind in the factory.

Epoxy resin adhesive of high cohesion is used for gluing onto the concretesurface.

Figure 9. CFRP laminates and Epoxy resin adhesive

3.2. Retrofit and repair method

This system has the following advantages, comparing with the current methods.

- CFRP laminate and CF sheet have the advantage of easy handling and highcorrosion resistance, and there is no change in the sectional dimension of structuralmembers before and after the execution.

- Thanks to the light-weight and the moderate stiffness of CFRP laminate, therepair works are easily at narrow space, such as the repair of the footing beam or theunderside of the lowest floor slab (The left of Figure 10). Usually many equipmentpipes are arranged near the underside of floor slab, this system has made possible torepair without movement of pipes (The right of Figure 10).

- In the case of upward work, due to the use of the high viscosity resin and thelight weight material there is neither need for mechanical equipment for pressing theCFRP laminate onto the substrate nor it is necessary to provide supporting devices tokeep overhead CFRP laminate in place.


Figure 10. The repair of footing beam and underside of floor slabs using this system

The process of this system is following.

At first, the surface of the concrete has to been prepared by sand disk grinder andthen cleaned by vacuum cleaner. And necessary restoration work is carried out withmortar or epoxy moral before application of the adhesive for the CFRP laminate. Nextthe impregnation resin is applied by rubber spatula onto the concrete. Immediatelyafter resin scraped, the epoxy resin is applied in conical shape onto the completelycleaning CFRP laminate by means of a specially developed instrument. The CFRPlaminate has carefully been pressed on by means of a hard rubber roller, squeezing outthe fresh adhesive at the sides. The conical shape of the adhesive layer allowscomplete evacuation of air on both sides during the pressing on by roller. Excessadhesive is carefully removed with spatula and the CFRP laminate surface is cleaned.

3.3. Application

Usually this system is applied for the repair and the retrofit of the concretestructures as shown in Figure 2 and Figure 10, and accordingly the number ofapplication applied this system is over 80 for 5 years from 1996 to 2001. Twospecific applications applied this system, except for concrete structures, aredescribed below.

3.3.1. Kosaka mine office (wooden building; Akitaprefecture) (Onose et al., 2001)

This building, which is three stories wooden structure and has Renaissance styledormer window and balcony, was constructed in 1905 and has been evaluated thearchitectural worth and has been specified the cultural assets of Kosaka-cho in 1997.After repair and restore to its original state, the building has been used for the resortfacility of the town.

The CFRP laminates have been used for the reinforcements of the woodenbeams. For the purpose of the application of this building, the structuralperformances of CFRP laminates glued wooden beam was tested.


Figure 11. The appearance of the building Figure 12. Gluing of CFRP laminates

3.3.2. Shiriya-zaki lighthouse (brick construction; Aomori prefecture)(Kalsumataet. al., 2001)

Shiriya-zaki lighthouse, located in the north end region of Honshu Island, isbeautiful brick tower (Figure 13) and has historical worth. It was designed by Britishengineer R.H. Brunton and constructed in 1877, however the bending strength of thetower against earthquake load was not enough. The upper part from the landing wasdestroyed by a bombing at the second world war, and reconstructed by means ofreinforced concrete after the war. Retrofit was carried out using CFRP laminates.Ten of 86 CFRP laminates arranged around the tower have tensioned and othershave glued onto the surface of bricks. The downside end of the tensioned CFRPlaminate has anchored hi the foundation newly constructed and the other has fixedon the upper bed of the tower landing. The tensioned CFRP laminates have causedcompression to the bricks consequently the bending strength of the tower isincreased.

Figure 13. Appearance of the lighthouse Figure 14. Gluing of CFRP laminates


Prestress Axial Bars Confining Sheet

Figure 15. Retrofitting techniques for lighthouse

4. Conclusions

The repair and retrofitting techniques using carbon fiber products enable changeof sectional dimension of structural elements negligibly small and make easy toexecute in the site due to the superior properties of carbon fiber, light weight andhigh strength. In civil engineering, the application of FRP products will be increasedin the future, as the advancements of material property are higher and higher.

References

Hagio H., Katsumata H., Kimura K and Kobatake Y., "A Study of Existing ReinforcedConcrete Structure Retrofitted by Carbon Fiber", First Asian-Australasian Conference onComposite Materials (ACCM-l), 1998.

Katsumata H., Kobatake Y and Takeda T., "A Study on Strengthening with Carbon Fiber forEarthquake-Resistant Capacity of Existing Reinforced Concrete Columns", Proceedingsof 9WCEE, 1988.

Katsumata H and Kobatake Y, "Seismic Retrofit with Carbon Fibers for Reinforced ConcreteColumns", Proceedings of 11WCEE, 1996.

Katsumata H and Kimura K., "Experience of FRP Strengthening for Historic Structures",Proceedings of 7th Japan International SAMPE Symposium & Exhibition, 2001.

Kobatake Y, Kimura K and Katsumata H., "A Retrofitting Method for Reinforced ConcreteStructures Using Carbon Fiber", Development in Civil Engineering 42, Elsevier, 1993.


Kimura K and Hagio H., "The Application of Fiber Reinforced Plastics (FRP) in theConstruction Field of Japan", The Third Composites Durability Workshop, 2000.

Onose J., Kumagai M., Mizuno T and Yamada S., "The Experimental Study on ReinforcingHistorical Wooden Structure by Carbonfiber Plastic Board", Memories of the TohokuInstitute of Technology, 2001.

Biography

Kohzo Kimura is a researcher of structural engineering, and his work deals with research anddevelopment of new technology using new material.

Hideo Katsumata is a researcher of structural engineering, and his work deals with seismiccapacity evaluation and earthquake resistant construction.

Design and Repairing of Hydraulic Valvesusing composite materials

Nicolas Junker*, **, Alain Thionnet **, Jacques Renard **

* : KSB amri SA, Pare d'activites Remora, 33170, GRADIGNAN, France** : Ecole des Mines de Paris, Centre des Materiaux, 91003, EVRY, France

\. Conception and Design of a butterfly valve made of composite material

A butterfly valve is an industrial structure which has the ability to regulate waterstreams in tubes. It is composed of an obturator, a body, an axis and several joints.The materials mostly used are steel, cast iron and cast steel but, now daysconsidering a weight gain request, composite materials are studied to design newbutterfly valves.

As stratified composite tubes made of vinylester, polyester or epoxy reinforcedglass fibers are commonly used for transportation, composite valves should beuseful.

• First request is a weight gain, particularly for large metallic diameter valves,like 600 mm, which cannot be mounted by a single person.

• Further some applications need a resistance to corrosion which can not bealways achieved with metallic materials : transportation of salted or sulfurettedwater, chemical applications, nautical engines.

One criterion for the choice of composite materials is stress intensity whenworking. Other criterions as price, complexity of the process have to be consideredregarding to the choice of materials. Stress intensity in butterfly valves can be veryhigh (over 200 Mpa in traction or compression, over 100 MPa in transverse loadingat the contact points between axe and obturator).

To satisfy all of these criterions, it is necessary to use different staking of longfiber composite materials. The sandwich conception has to be used for the whole


structure of the obturator to resist to bending. Sometimes metallic parts are neededin transverse directions because of the weakness of composite plies perpendicular tofibers.

Because composite material are heterogeneous, anisotropic and damageable(transverse cracking, dclamination, fiber breaking), numerical techniques have beendcvclopped like homogenization to take heterogeneities into account. Furtherorientation methods are described to model anisotropy of the material. To modeldamage the framework of Damage Mechanics has been used.

The purpose of this paper is to propose different step analysis to solve theseproblems and to use them for designing valve obturator. Dclamination andtransverse cracking arc coupled with calculation to better predict lifetime ofbutterfly valve during cycling.

2. Numerical Methods to calculate layered composite materials andsandwich structures

2.1. Homogenization

The structures we want to calculate are made of laminated unidirectionalcomposite plies composed of long glass fibers wrapped into an cpoxy matrix. Eachply has a given orientation. The stratification has a great number of layers allowingto consider the whole material as an infinite periodic layered material. So thetechniques of periodic homogcnisation can be used [San, 1980].

The purpose of homogenization is to get the characteristics of a virtualhomogeneous material equivalent to the stratified one to calculate global structure.By this way we evaluate the macroscopic stress and deformation fields and then bylocalization procedure, we get the microscopic deformation and stress fields. Themathematical equations involved in the homogenization procedure are explainedbelow in a very shortened way.

If we consider a periodic cell Y constituing a stratification. The physical fieldsdefined on this cell arc :

macroscopic stress and deformationhomogenized elasticity tensor

microscopic stress and deformation: microscopic elasticity tensor

<f>v means the volumic average value of f over the cell


Macroscopic and microscopic fields are mathematicaly related by the volumicaverage calculation : Z=<a>y, E=<e> Y - The homogenization steps are thefollowing:

- Calculation of the six elementary problems :

Calculation of the homogenized elastic characteristics

Finite element calculation of the structure,

Calculation of microscopic fields

- Calculation of Tsai-Hill criterion in each layer to obtain a failure criterion forthe whole stratification.

Following these steps during every FEM calculation, we can give in any partof the structure (i.e. in each layer of the laminated material), the state of failure.

2.2. Transverse cracking

The proposed model [Ren, 1993] simulate the evolution of transverse cracking ineach layer of a laminated structure. The different steps of this model are describedon the figure 1. The results of coupling between calculation and the model can bedisplayed on an example of butterfly valve with sandwich structure and stratifiedcomposite material composed of a periodicity of two layers of unidirectional glassfiber and epoxy matrix. Figure 2 shows the damage rate in the two layers of thestratification.

2.3. Delamination

Our study is focused on delamination between macro components of the butterflyvalve, not delamination between all the layers of the stacking of composite parts.


Then our analysis is first at a macroscopic level. Damage Mechanics has beenused instead of Fracture Mechanics because we need more local information thanglobal energy balance. Our approach consider the interface between tocomponents by using a thin (0.001mm) layer of matrix. This method has beendevelopped by many authors [All, 1992], [Cri, 1998], [Kim, 1998]. Variablesdescribing the behaviour of the interface measure the rate of damage : when theirvalue is 0, the interface is not damaged; when their value is 1, the interface iscompletely delaminated; so the location of delaminated area is known according toevolution of these variables.

Figure 1 : Schematic steps of ply cracking model.


Figure 2 : Ply damage coupling calculation of valve.

Different kind of interface elements have been used (Figure 3). These elementsare degenerated isoparametric volumic elements from which one direction has beenreduced to zero. The thickness of the interface is considered to be a materialcharacteristics of the interface.

Figure 3 : interface elements

Such elements can be used in 2D, pseudo 3D and 3D meshes to separatemacroscopic components. The next paragraph describes the use of such elementsduring calculation of tubes and real industrial butterfly valve applications.


2.4. Orientation of strong anisotropic materials in thick shell with complexshape

Classical thin shells encoutered in air plane design or other industrial design aremostly modelized with shells elements when doing finite elements calculation. Suchelements have the advantage to simply define a normal vector to the surface theymap. Knowing this normal vector, you can easily define the orientation vector fieldsof the heterogeneous and anisotropic material constituing the shell.

Our problem is that the shells constituing the sandwich butterfly valves are muchtoo thick to be described with shell elements; they can only be described by volumicelements and the kinematic of a volumic element of automatic mesh (withtetraedrons for example) does not give simply a normal vector field in every pointof the structure.

The solution we adopted was to perform a pre-fem-calculation giving asa result the normal vector field in a particular simple way. On a thick shellwith complex shape you can define a bottom surface and a top surface. Theresolution of the Laplace equation on the shell with 0 as boundary conditionat the bottom and 1 at the top simply gives a field which gradient naturalydescribes the normal vector flield of the shell.

Figure 4 : Laplace bundary conditions


3. Application to industrial structures

3.1. Application to laminated plates with a circular hole

The studied structure is a four layers laminated holed plate submited to traction.Layers are made of glass-fibers epoxy matrix. Four different stacking sequences arestudied and the damage field at the interface between the first and the second layer isplotted.

The second stratification (30°, -30°, -30°, 30°) is the more susceptible ofdelamination. Results prove the ability of the method to give pertinent evaluation ofdelaminated area inside a stratification (Figure 5).

3.2. Application to a real composite butterfly valve

A real 250 mm diameter composite butterfly valve has been calculated andtested. Both test and calculation give the same location of possible delaminationduring the cyclic life of the valve (between exterior shells and the interior body ofthe valve).

The fourth view shows the location of possible delamination at the interfacebetween exterior shells and the rest of the valve. Every numerical techniqueexplained in this paper has been used for this example.

Nethertheless if this qualitative result is interesting to caracterize thedelamination behaviour of the structure, the load rate at which delamination beginsis overestimated.


Figure 5 : Computation of holed composite plates

4. Conclusion

The design of industrial composite structure using finite element computation ispossible when some numerical tehcniques are developped. These techniques have totake Damage Mechanics into account to refine the calculation wich could be topessimistic if it was only elastic and linear. The strong anisotropy of compositeneeds the development of a special orientation method that is simple and can beeasily used in many different conceptions. The result of the use of all thesedevelopped techniques simultaneously give an interesting evaluation of thebeheviour of an industrial structure giving the ability to optimise the conception interms of dimensions, shapes and material constitution.


Figure 6 : Damage coupled Computation of valve.

S.Bibliography

Allix O., Ladeveze P., 1992 "Interlaminar interface modelling for the prediction ofdelamination" , Comp. Struct. 22, (1992), pp. 235-242.

Crisfield M.A., Mi Y., "Progressive Delamination Using Interface Elements". Journal ofComposite Materials, 32, 1998.


Kimpara I., Kageyama K., Suzuki K., "Finite element stress analysis of interlayer based onselective layerwise higher-order theory", Composites Part A 29A, (1998), pp. 1049-1056.

Renard J, Favre, J.P., Jeggy Th., "Influence of Transverse Cracking on Ply Behaviour :Introduction of a Characteristic Damage Variable". Composite Science andTechnology, 46, 1993, pp. 29-37,

Sanchez-Palencia E., "Nonhomogeneous Media and Vibration Theory", Vol. 127 of LectureNotes in Physics Springer, Berlin, 1980.

lonomer as Toughening and RepairMaterial for CFRP Laminates

M. Hojo* — N. Hirota** — T. Ando*** — S. Matsuda****M. Tanaka* — K. Amundsen*** — S. Ochiai*****A. Murakami****

* Dept. Mechanical Engineering, Kyoto University, Kyoto 606-8501, Japan

hojo@mech. kyoto-u. ac.jp

mototsugu@mech. kyoto-u. ac.jp

** Student, Kyoto University, Kyoto 606-8501, Japan

*** Graduate Student, Kyoto University, Kyoto 606-8501, Japan

****Dept. Chemical Eng., Himeji Institute of Technology, Himeji 671-2201, Japan

[email protected]

[email protected]

***** International Innovation Center, Kyoto University, Kyoto 606-8501, Japan

[email protected]

ABSTRACT: Interlaminar fracture toughness under mode I and II loadings was investigated forunidirectional CF/epoxy laminates with ionomer interleaf. The fracture toughness of ionomerinterleaved CF/epoxy laminates was much higher than that of base CF/epoxy laminates bothunder mode I and II loadings. For mode I loading, the high level of the toughness was keptconstant with the crack growth. Mode I interlaminar toughness initially increased with theincrease of ionomer interleaf thickness, and then leveled off. For mode II loading, thetoughness continuously increased with the ionomer thickness, and reached 9 to 10 kJ/m 2 ,which is one of the highest among already reported results. Using the high bonding propertiesof ionomer, the repairability of delaminated composites was also tried. The delaminatedspecimen was hot-pressed again, and the interlaminar toughness change after repair wasinvestigated. Although hot-pressing without additional ionomer film gave poor results, therepair with ionomer film brought the toughness comparable to the virgin laminates.

KEY WORDS: delamination, fracture toughness, CFRP, interleaf, ionomer, repair


1. Introduction

Although almost twenty years have passed since the importance of delaminationwas recognized (O'Brien 82), interlaminar strength is still one of the design limitingfactors in structural composite laminates. One of the most promising ways toincrease the interlaminar properties is to control the mesoscopic structure byreplacing only the resin layer at the prepreg interface to a tougher system. This wayis often called as "interleaf or "interlayer" method. The original way of this conceptis simply to insert conventional thermoset or thermoplastic interleaves (Sela et al.,89, Aksoy et al., 92). A new commercial product with a heterogeneous interlayerincluding fine thermoplastic particles, T800H/3900-2, has shown excellentcompressive strength after impact (CAI), and has already been applied for primarystructures of Boeing 777 (Odagiri et al., 96). Although this material indicatedexcellent mode II fracture toughness, the mode 1 fracture toughness decreasedgradually with the increment of crack length (Kageyama et al., 95).

The above results suggested that both high ductility and high adhesion strengthare necessary for the interleaf materials to improve the interlaminar fracturetoughness (Hojo et al., 99). Ionomer was introduced as interleaf material because ithas high ductility and good adhesion to epoxy resin. Figure 1 shows the schematicstructure of the transverse section of the ionomer-interleaved carbon fiber(CF)/epoxy laminates (Matsuda et al., 99). There is the interphase region of one- ortwo- carbon fiber thickness between the ionomer interleaf and base lamina, whereepoxy and ionomer are mixed. Since the crack path is often arrested within theinterlayer region by CF, excellent interlaminar properties are expected.

In the present study, the mode I and II interlaminar fracture properties of theionomer-interleaved CFRP were first reviewed. Then, the repairability ofdelaminated composites was investigated using the high bonding properties ofionomer.

Figure 1. Schematic structure of transverse section near ionomer/base laminainterface in ionomer-interleaved CF/epoxy laminates



Laminates used in this study were made from Toho Rayon UT500/111 prepregs.Unidirectional laminates, (0)24, of the nominal thickness of 3 mm were molded in ahot press. The curing temperature was 140°C, holding time was 120 min and thepressure was 1 MPa. Ethylene based ionomer film was inserted at the mid-thicknessduring molding process as interleaf. Here, ethylene methacrylic acid copolymer wasionized partially by zinc iron (Murakami et al., 97). The thickness of ionomer filmwas 12, 25, 100 and 200 mm. The laminates without interleaf were also prepared forcomparison. Starter slits were introduced into the laminates by inserting single 13urn thick polyimide film during molding at midplane. Fracture toughness tests werecarried out both under mode I and II loadings using double cantilever beam (DCB)and end notched flexure (ENF) specimens (JIS K7086).

Repair of laminates was also tried under mode I loading by hot-pressing thedelaminated specimen again with and without reinserting ionomer. After thepreparatory tests, final repair condition was selected as the hot press temperature of130°C, holding time of 130 min and pressure of 2 MPa. The delaminated specimenswith and without ionomer interleaf were hot-pressed again with reinserted ionomerand the same 13 (um thick polyimide film as starter slits. Using this condition, repairwithout reinserting ionomer film was also investigated with 25- and l00um-ionomer-interleaved laminates.

The tests were carried out in a computer-controlled servohydraulic testingsystem (Shimadzu 4880, 9.8kN)(Hojo et al., 94, 97). The cross head speed wascontrolled to be 0.5 to 1.0 mm/min in DCB tests, and the crack shear openingdisplacement speed was controlled to be 0.03 mm/min in ENF tests (JIS K7086).The crack length was computed from the measurement of the compliance by usingthe calibration relation between the compliance and the crack length. The tests werecarried out in laboratory air. The energy release rate under mode I loading wascalculated using modified compliance calibration method. That under mode IIloading was calculated using compliance calibration curves for each specimen(Matsuda et al., 97).

3. Results and discussion

3.1. Mode I and II interlaminar fracture toughness before repair

Since the scatter in the relation between the interlaminar fracture toughness andthe increment of crack length (Aa) is rather large, the average of several specimenswas calculated over subsequent 1 mm increment of the crack length for Aa < 10mmand subsequent 5 mm for Aa > 10 mm. Then, Figure 2 shows the effect of interleaf-film thickness on the R-curve under mode I. Both the initial values, GIc, and thepropagation values, GIR, increased dramatically with the increase of the interleaf


Increment of crack length, Aa (mm)Figure 2. Averaged relation between fracture toughness and increment of cracklength under mode I loading

Increment of crack length, Aa (mm)

Figure 3. Averaged relation between fracture toughness and increment of cracklength under mode I loading


thickness. For the ionomer thickness of 200um, the toughness increased about tentimes from the base laminates. Another important point is that the G]R values kept ahigher plateau value without respect to the crack length. This behavior wascompletely different from that for T800H/3900-2 where the R-curve decreased, andconverged to the base laminate value.

In Figure 3, each GUR data point was calculated as the average value oversubsequent 1 mm increment of crack length in the relation between mode II fracturetoughness and increment of the crack length. The initial values of the fracturetoughness were simply calculated at the maximum load point under mode II loading.Similar to the results under mode I loading, the whole R-curve increased markedlywith the increase of the interleaf thickness. For the ionomer thickness of 200um, thetoughness increased about twenty times from the base laminates. The actualtoughness value of 10 kJ/m2 was also one of the highest among the already reportedresults for CFRP laminates.

Microscopic observation showed that the crack path was arrested by the rigidcarbon fiber at the surface of the base lamina. For conventional interleavedlaminates, there was no toughened resin at the surface of the base lamina, and thiscaused the decrease of the toughness. On the other hand, the crack was still insidethe toughened region for ionomer interleaved laminates. This is responsible for thenon-decrease of the propagation values of the fracture toughness with the incrementof the crack length under mode I loading.

For mode I loading, the permanent deformation of the ionomer was localized inthe vicinity of the crack path. This feature was almost the same without respect tothe ionomer thickness. In this case, the reduced stress intensity factor by theintroduction of the ionomer interleaf is responsible for the toughening mechanism(Tanaka et al., 97), and only the existence (not the thickness) of the interleafcontributes to the increase of the toughness. For mode II loading, the deformationwas expanded to the whole interlayer indicated by large permanent sheardeformation. This means the deformation of the whole interleaf thicknesscontributes to the increase of the toughness, and is related to the linear increase ofthe toughness with the interleaf thickness (Hojo et al., 99).

3.2. Repairability of laminates with ionomer

Figure 4 compares the results of fracture toughness tests after repair withreinserting ionomer. The obtained propagation values, GIR, are comparable to theionomer-interleaved laminates with the same final ionomer thickness. Thus, therepair is quite successful without deterioration. On the other hand, repair withoutreinserting ionomer gave quite poor results as indicted in Figure 5. The toughness isless than 10% of the ionomer interleaved laminates with the same original ionomerthickness. The values are similar to those of base laminates.


Increment of crack length, Aa (mm)Figure 4. Relation between mode I fracture toughness and increment of crack lengthfor 25fJm-ionomer-interleaved and base CFRP repaired with reinsertingionomer

Figure 5. Relation between mode I fracture toughness and increment of crack lengthfor 25um-ionomer-interleaved CFRP repaired without reinserting ionomer


The transverse section of the laminates repaired without reinserting ionomerindicated existence of voids at the interphase region. When the crack path was at theinterphase, the ability of rebonding is possibly rather weak, resulting in voids. Thisis responsible for the poor repairability of laminates without reinserting ionomer.

4. Conclusions

Interlaminar fracture toughness of ionomer-interleaved CF/epoxy laminates wasinvestigated under mode I and II loadings. These laminates indicated dramaticincrease of the toughness from base CF/epoxy laminates both under mode I and IIloadings. The propagation values of the fracture toughness did not decrease from theinitial values with the increment of the crack length under mode I loading.

The delaminated specimen was hot-pressed again, and the interlaminartoughness change after repair was investigated only under mode I loading. Althoughhot-pressing without reinserting ionomer film gave poor results, the repair withreinserted ionomer film brought the toughness comparable to the original ionomer-interleaved laminates.

Acknowledgments

The authors would also like to thank Dr. B. Fiedler of Technical UniversityHamburg-Harburg and Mr. M. Ando of Toho Tenax Co., Ltd. for their helpfuldiscussion.

References

Aksoy, A., Carlsson, L.A., "Interlaminar Shear Fracture of Interleaved Graphite/EpoxyComposites", Composite Science and Technology, Vol.43, 1992, p.55-69.

Hojo, M., Ochiai, S., Gustafson, C-.G., Tanaka, K., "Effect of Matrix Resin on DelaminationFatigue Crack Growth in CFRP Laminates", Engineering Fracture Mechanics, Vol. 49,1994,p.35-47.

Hojo, M., Matsuda, S., Ochiai, S., "Delamination Fatigue Crack Growth in CFRP Laminatesunder Mode I and II Loadings-Effect of Mesoscopic Structure on Fracture Mechanism-",Proc. International Conference on Fatigue of Composites, Paris, 1997, p. 15-26.

Hojo, M., Matsuda, S., Ochiai, S., Murakami, A., Akimoto, H., "The Role of Interleaf/BaseLamina Interphase in Toughening Mechanism of Interleaf-Toughened CFRP", Proc.ICCM12, Paris, 5-9 July, 1999, CD-ROM.


Kageyama, K., Kimpara, T., Ohsawa, I, Hojo, M., Kabashima, S., "Mode I and IIDelamination Growth of Interlayer-Toughened Carbon/Epoxy (T800H/3900-2)Composite System", Composite Materials: Fatigue and Fracture, Fifth Volume, ASTMSTP 1230, Martin, R. H., Ed., ASTM, 1995, pp. 19-37.

Matsuda, S., Hojo, M., Ochiai, S., "Mesoscopic Fracture Mechanism of Mode IIDelamination Fatigue Crack Propagation in Interlayer-Toughened CFRP", JSMEInternationalJournal, Series A, Vol.40, 1997, p.423-429.

Matsuda, S. , Hojo, M., Murakami, A., Akimoto, H., Ando, "Effect of Ionomer Thickness onMode 1 Interlaminar Fracture Toughness for Ionomer Toughened CFRP", Composites,Part A, Vol.30, 1999, p. 1311 -1319.

Murakami, A., Ooki, T., Asami, T., Hojo,, Ochiai, S., Matsuda, S., Moriya, K. "InterlaminarFracture Toughness and Damping Properties of Thermoplastic Ionomer InterleavedComposite", Recent Advancement of Interfacial Materials Science on CompositeMaterials '97, Siguma, Pub., 1997, p.75-79.

JIS K7086-1993, "Testing Methods for Interlaminar Fracture Toughness of Carbon FibreReinforced Plastics", 1993.

O'Brien, T.K., "Characterization of Delamination Onset and Growth in a CompositeLaminate", Damage in Composite Materials, ASTM STP 775, Reifsnider, K.L., Ed.,ASTM, Philadelphia, 1982, p. 140-167.

Odagiri, N., Kishi, H., Yamashita, M., "Development of TORAYCA Prepreg P2302 CarbonFiber Reinforced Plastic for Aircraft Primary Structural Materials", Advanced CompositeMaterials, Vol.5, 1996, p.249-252.

Sela, N., Ishai, O., Banks-Sills, L., "The Effect of Adhesive Thickness on InterlaminarFracture Toughness of Interleaved CFRP Specimens", Composites, Vol. 20, 1989, p. 257-264.

Tanaka, K., Tanaka, H., Kimachi, H., "Boundary Element Analysis of Elastic StressDistribution in Cracked FRP under Mode I Loading", Trans. Japan Society forMechanical Engineers, Vol. 63A, 1997, p. 1894-1901.

Polymer adhesives in civil engineering:Effect of environmental parameters onthermomechanical properties

K. Benzarti* — M. Pastor*—T. Chaussadent*— M.P. Thaveau**

*Laboratoire Central des Fonts et Chaussees (LCPC), Service Physico-chimie desmateriaux, 58 boulevard Lefebwe, 75732 Paris Cedex 15, [email protected]

**Laboratoire Regional des Ponts & Chaussees, BP141, 71405 Autun, France.

ABSTRACT: In this work, aging of two ambient curing thermoset polymers (an epoxy system anda polyester based mortar), commonly used for civil engineering applications, has beeninvestigated. In a first part, microstructural evolutions of the adhesives in a standardenvironment (50% relative humidity, 20°C) were studied. The polymerization kinetics of theepoxy system was monitored by infrared spectroscopy and differential scanning calorimetry(DSC). These experiments showed that the crosslinking process of thermosetting systemsdoesn't go to completion at ambient temperature. DSC analyses also revealed a mechanism ofphysical aging leading to progressive evolution of the polymer network. In the second part,the two materials were immersed in various model solutions (distilled water, salt solution,concrete pore solution). Mass uptake of immersed samples was monitored as a function oftime, and influence of aging treatments on the thermomechanical properties was discussed interms of chemical and microstructural modifications of the polymer network.

KEYWORDS: epoxy, polyester, crosslinking, chemical or physical aging, viscoelastic behavior.


1. Introduction

Polymer adhesives, such as thermoset resins, are commonly used in civilengineering for the repair of damaged concrete structures (bridges, walls, etc...). Agrowing application is the reinforcement of cracked structures with bondedcomposites. Implementation of this technique is based either on the use ofprefabricated composite plates or on wet lay-up process involving carbon fabrics(Karbhari et al. 2000, Toutanji et al. 1997). Polymer adhesives also open up newopportunities for the design of bridges, since parts of the structures could beassembled by gluing in the future. Nevertheless, development of such structuralapplications is still limited, due to an insufficient knowledge of the adhesive bonddurability. In fact, polymer joints are sensitive to environmental parameters such asmoisture, temperature or chemical attacks (Mukhopadhyaya et al, 1998, Nogueira etal. 2001) and the resulting degradations may progressively affect the mechanicalstrength of the adhesive bond. Moreover, polymer adhesives are often in contactwith concrete which is an alkaline and potentially aggressive medium (Chin et al.2001). For all these reasons, there are still serious concerns about the long termbehaviour of repaired structures, and fundamental studies are needed in order toidentify mechanisms involved in the degradation of polymer joints andadhesive/concrete interfaces.

According to the literature, degradation of epoxy joints mainly results frommoisture diffusion into the material. Ingress of water generally induces physico-chemical modifications in the interfacial areas between adhesive and substrate or inthe bulk polymer, such as plasticizing effects (Zanni-Deffarges et al. 1995, Nogueiraet al. 2001). These modifications lead to a progressive loss of mechanical propertieswhich is function of the water content. Pick's model generally provides goodpredictions for the diffusion of liquids in a bulk polymer (Chin et al 1999). For aplane polymer sheet exposed to a diffusing fluid, the change of concentration C ofthe diffusant, at a distance x from the contacting surface, as a function of time t anddiffusion coefficient D, is given by Pick's second law (Cranck et al., 1968):

An approximate solution of equation [1] is:

where m, is the mass uptake of the polymer at time t, moo is the mass uptake atequilibrium, and h is the sample thickness.


Epoxy resins cured with amine hardener are seldom subject to severe chemicaldegradation, such as hydrolysis, since the crosslinked network has a good chemicalstability. However, if the polymerization is not fully achieved, residual monomersmay increase sensitivity of the epoxy network towards chemical attacks.

Polyester resins are much more sensitive to chemical aging than epoxy systems.Indeed, hydrolysis of ester groups can occur in aggressive alkaline environments(saponification) or in acidic media. Examples of hydrolysis in neutral saltenvironments are also reported in the literature (Chin et al. 1999). The base-catalyzed hydrolysis of ester linkages [3] leads to the formation of carboxyl groupswhich can further react with hydroxides, such as KOH or NaOH, to yieldcarboxylate anions COO- via reaction [4]. Such a degradation is irreversible andusually reduces significantly mechanical properties of the polymer.

The objective of this work was to study two thermosetting systems commonlyused for the repair of civil engineering structures: an epoxy adhesive and a polyesterbased mortar.

In a first part, the study focused on microstructural changes of the polymernetworks that can occur in a standard environment (50% relative humidity, ambienttemperature). Experiments were performed by infrared spectroscopy and differentialscanning calorimetry in order to characterize the polymer structure and its eventualevolution.

In a second part, the behaviors of the two systems in aggressive environmentswere investigated: accelerated aging tests were performed by immersing samples inmodel aqueous solutions (distilled water, salt solution and an alkaline solution whichis representative of the concrete medium). The mass uptake of samples wasmonitored as a function of aging time and the viscoelastic behavior of aged samplewas evaluated by dynamic mechanical analysis. Such accelerated tests may not beentirely representative of the actual degradation processes in natural environments,however, they can provide precious information on the sensitivity of the polymernetworks towards external aggressive factors.

2. Experimental

2.1. Materials

Two commercial thermosetting systems that are commonly used in civilapplications were chosen for this study: an epoxy system and a polyester basedmortar.


The epoxy system is a two components adhesive, constituted of a resin and apolyamine based hardener. The resin is a viscous liquid and contains mineral fillers(30 wt %) whereas the hardener is an unfilled paste. This polymer adhesive is usedto paste carbon fabrics on damaged concrete structures, according to the wet lay-upprocess.

The polyester mortar is also made of two components: a polyester resin andmineral fillers containing a small amount of peroxide catalyst (2 wt %). This systemis mainly used for road works but also for the repair of concrete structures.

Table 1 gives the compositions of the two systems and the recommended blendratios. Rectangular specimens (5x5x40 mm) were made by casting the viscousmixtures into silicone moulds. Cure was performed at ambient temperature for thetwo systems.

Table 1. Composition of the two thermoset systems.

Epoxy s

Resin

- Diglycidylether ofbisphenol A (DGEBA)

- CaCO3 fillers (30 wt %)

ystem

Hardener

• Triethylenetetrarame(TETA)

- Alkylethefamme

100 wt part of resin / 40 wt part of hardener

Poly

Resin

- Polyester

- Styrene

ester mortar

Filler and catalyst

- Si02 fillers (98%)

- Peroxide catalyst(2wt%)

1 volume of resin / 1.5 volumeof fillers

2.2. Experimental techniques

2.2.1. Physico-chemical characterizations

Chemical analyses were performed by Fourier transform infrared spectroscopy(FTIR) using a Nicolet IMPACT 410 apparatus equipped with an ATR microscopedevice (attenuated total reflectance). In a first step, this technique gave an evaluationof the polymerization kinetics for the epoxy system: the peak intensity at 915 cm-1

(epoxy rings) was monitored as a function of time, and normalized by rationing theheight of the peak of interest by the height of the aromatic C-H peak at 830 cm-1. Ina second step, surfaces of cured samples that were aged in model solutions, wereanalyzed using the ATR microscope. Comparison with control samples gaveindications on eventual chemical degradations induced by aging treatments.

Experiments were also carried on by differential scanning calorimetry (DSC),using a NETSCH DSC 200 apparatus, in order to evaluate the total heat of reactionand the glass transition temperatures of materials. Analyses were performed in non-


isothermal mode in the range from -40 to 200°C under nitrogen environment, at aheating rate of 10°C/min.

2.2.2. Characterization of the viscoelastic behavior

Viscoelastic properties of the materials were evaluated by dynamic mechanicalanalysis, before and after aging treatments, using a Metravib visoanalyser. Testswere performed on small samples (5x5x40 mm) in tension-compression mode with afixed displacement amplitude of 5 um and a frequency of 5 Hz. The analyzedtemperature range was between 30 and 150°C. This device provided informationabout the storage modulus E' and the loss tangent tan8. The former is representativeof the molecular motion ability of polymer chains.

2.2.3. Accelerated aging treatments in aqueous solutions

Cured specimens were aged for various periods of time in model solutions, atambient temperature (20°C). These treatments were supposed to simulate aging inaggressive environments. Three solutions were chosen: distilled water, a saltsolution representative of seawater (0.58 mol.L-1 NaCl), and an artificial concretepore solution in order to simulate the alkaline environment of cementitious material(0.5 mol.L-1 KOH and 0.1 mol.L-1 NaOH). Periodically, samples were removed fromthe solutions, dried with filter paper, immediately weighed with a Mettler digitalbalance and then returned to their bath. The procedure was repeated until thesamples reached a saturation level. An average of five samples was tested for eachmaterial in each solution.

3. Results and discussions

3.1. Microstructural changes in a « standard » environment

3. 1.I. Structure of the cured epoxy system

In order to investigate the polymerization kinetics of the epoxy system, themixture (blend of resin and hardener) was analyzed by FTIR spectroscopy.

Figure 1 presents the evolution of the normalized peak intensity at 915 cm-1 as afunction of time. The decrease of this intensity is related to the consumption ofepoxy monomers as the crosslinking reaction progresses. In a first stage, the rate ofthe kinetics is very high, due to the reactivity of the aliphatic polyamine hardener.


Figure 1. Evolution of the normalized epoxy peak intensity at 915 cm-1 asfunction of time for the epoxy system (IRTF spectroscopyexperiments).

Gelation occurs very early as the extend of reaction reaches 0.60, typically after fewhours. But in the first days, the kinetics is considerably slowed down and the extendof reaction seems to stabilize around 0.9. The reaction mechanism is then controlledby the slow diffusion of monomers in the polymer network.

DSC experiments were also performed on the liquid epoxy mixture (resin andhardener). They provided values for the total heat of reaction (AH=243 J/g), for theactivation energy (Ea=75 kJ/mol) and the glass transition temperature (50°C).

Figure 2 shows thermograms of two cured epoxy samples which had beenrespectively elaborated 15 days (a) and 10 months (b) before the DSCcharacterization. Both samples were kept at room temperature (20°C) and 50%relative humidity before DSC analyses. On the two curves, exothermic peaks arevisible around 150°C and are related to the cure at high temperature of residualmonomers. Extend of reaction calculated from the residual heat of reaction arerespectively 0.9 and 0.92. These values confirm results from IRTF spectroscopyexperiments: due to the slow diffusion process at 20°C, the maximum rate ofconversion is close to 0.9, and the cure of the epoxy network is never fully achieved.Therefore, about 10% residual monomers still remains trapped in the polymernetwork. Moreover, an endothermic peak can be seen on the thermogram of theolder sample, just above the glass transition temperature. It is a structural relaxationpeak related to the phenomenon of physical aging which will be discussed in thenext section.

3.1.2. Influence of physical aging

Physical aging is a phenomenon common to all amorphous polymers in theglassy state, where the molecular structure is out of thermodynamic equilibrium.


Physical aging is a manifestation of a slow spontaneous evolution of the polymertowards its equilibrium state by time-dependant changes in volume, enthalpy andentropy. This phenomenon is generally accompanied by an evolution of mechanicalproperties, such as increase in stiffness and embrittlement of the material (Struik1978 ). Enthalpy loss during the aging process is recovered during reheating of theaged sample to above Tg (during a DSC experiment for instance). This enthalpyrecovery leads to the apparition of an endothermic peak on DSC thermograms,above the glass transition temperature.

On figure 2, such an endothermic peak is seen for the 10 months old sample. Itmeans that epoxy systems used in civil engineering are subject to physical aging atambient temperature. This can be easily explained, since the glass transitiontemperature of these materials is generally low (about 50°C) and ambient temperatureslie in the range from Tg-30°C to Tg, where fast aging kinetics is observed.

A study is in progress in our laboratory in order to evaluate the influence ofphysical aging on the mechanical properties of these thermoset systems.

Figure 2. DSC thermograms for the epoxy system (a) 15 days after samplepreparation (b) 10 months after sample preparation.

In this first part of the work, two main facts were observed: thermoset resinscured at ambient temperature are not fully polymerized. Indeed, the extend ofreaction is limited and some monomers still remain trapped in the polymer network.Therefore, further variations of temperature can lead to small evolutions of thecrosslink density. Moreover, DSC experiments revealed that a physical agingprocess occurs in these materials at ambient temperature. This phenomenon is themain process susceptible to induce microstructural changes in a standardenvironment (20°C, 50% relative humidity).


3.2. Microstructural changes in aggressive environments

3.2.1. Mass uptake of immersed samples - diffusion phenomenon

Figure 3 shows evolutions of the mass uptake of samples as a function of theimmersion time, for the epoxy system (a) and the polyester mortar (b). Experimentswere performed at 20°C.

As shown by figure 3.a, immersion of epoxy samples in distilled water or in saltsolution led to a rapid mass uptake, resulting from the diffusion of liquid into thematerial. In a second stage, uptake slowed down progressively and reached anequilibrium around 5%. Situation is different in the alkaline solution, where the massuptake at equilibrium is close to 8%. For the three solutions, values of the equilibriummass uptake are elevated and can be explained by the low crosslink density of theepoxy network (low Tg) or by the presence of residual polar groups that can promoteincreased sorption of polar penetrants. Diffusion coefficient derived from Fick'smodel [4] are respectively 7.1xl0-9 cm2.s-', 4.9xl0-9 cm2.s-1 and 8.0xl0-9 cm2.s-1 fordistilled water, salt solution and alkaline solution in the epoxy network.

Figure 3. Mass uptake of immersed samples as a function of time


On figure 3.b, mass variations are globally lower for polyester samples than forepoxy specimens, due to the large mineral filler content of the mortar (about 70 wt%). An interesting feature is the rapid mass loss observed for samples that wereimmersed in the alkaline solution (-1.5%). This phenomenon can be attributed to achemical degradation of the polyester matrix

3.2.2. Analysis of aged samples by ATR-FTIR spectroscopy

In order to verify if aging treatments induced chemical modifications of thematerials, infrared spectroscopy analyses were conducted on the surfaces of agedsamples, using the ATR microscope device.

Immersion of samples in distilled water or in salt solution at 20°C did not modifyFTIR spectra neither for the epoxy system, nor for the polyester mortar. Therefore, itcan be concluded that these two treatments did not induce any significant change ofthe chemical structure of materials, and that diffusion of liquid in the polymernetwork is the main aging process.

The situation is quite different when samples are immersed in the alkalinesolution. Figure 4 shows the FTIR spectra for the surface of the polyester samplesbefore (a) and after (b) immersion in the simulated concrete pore solution. Largemodifications are visible on the spectrum of the aged sample as compared to thecontrol spectrum: peaks related to the organic part of the polyester mortar areremoved from the spectrum of the aged sample (C=O linkages near 1720 cm-1 andC-O linkages at 1250 cm-1). On the other hand, new peaks related to the mineral partof the mortar (silica fillers) appear at 1030, 780 et 694 cm-1. It can be concluded thatthe surface of the aged sample has been degraded during immersion in the alkalinesolution: hydrolysis of the organic part of the mortar (polyester) according to thesaponification process described in [3] and [4] is probably involved. This isconsistent with the mass loss previously observed, since hydrolyzed fragments of thepolymer network can be released in the aqueous medium.

Modifications are also observed on the IRTF spectra of the epoxy sample thatwas immersed in the alkaline solution, suggesting that some degradation of thepolymer network occurred during aging. However this degradation process has notbeen clearly identified and is not accompanied by a mass loss of samples.

Previous results lead to the conclusion that the alkaline solution representative ofa concrete medium is a very aggressive environment, both for polyester and epoxythermoset systems. Of course, this result can not be generalized for a real civilengineering application which is a much more complex situation. However it isprobable that such chemical degradations can also occur in the reality atadhesive/concrete interfaces.


Figure 4. IRTF-ATR spectra for the surface of polyester samples(a) reference (b) aging for110 days in the alkaline solution at 20°C.

3.2.3. Influence of immersion on viscoelastic properties

Viscoelastic properties of the two materials were also evaluated by dynamicmechanical analysis, before and after aging in the various solutions.

Figure 5 shows evolutions of the storage modulus (a) and the loss tangent (b) asa function of temperature for a reference epoxy system and for samples immersed 63days in the three solutions. A significant decrease of the storage modulus is observedfor aged samples at temperatures close to ambient, as compared to the modulus ofthe reference sample. This phenomenon can be attributed to the well knownplasticizing effect of the polymer network by water molecules: the creation ofhydrogen bonds between water molecules and polar hydroxyl groups of the polymerleads to the break of intermolecular linkages (Nogueira et al 2001, Moy et al 1980).This microstructural change is accompanied by a swelling of the polymer networkand by a drop of stiffness and mechanical properties. Moreover, figure 5.b shows anincrease of the loss tangent level at low temperatures for aged epoxy samples, and


suggests that the motion ability of the polymer chains is globally increased byimmersion treatments.

Figure 5. Evolution of the storage modulus (a) and the loss tangent (b) as a functionof temperature for a reference epoxy and for samples immersed 63 days inthe various solutions

Figure 6 shows the evolutions of the storage modulus and the loss tangent as afunction of the temperature for the reference polyester and for samples aged 115days in the various solutions. As it was noticed for the epoxy system, there is a dropof the storage modulus of aged samples at temperatures close to ambient, due to theplasticizing effect of the network by water molecules. Observed variations are lessimportant than they were for epoxy, since the organic content of the polyester mortaris small. The level of the loss tangent at low temperature is also higher in agedsamples than in the reference material, which can be attributed to an increasedmolecular motion ability.

Figure 6. Evolution of the storage modulus (a) and the loss tangent (b) as a functionof temperature for a reference polyester and for samples immersed 115days in the various solutions.


4. Conclusions

The aim of this work was to study aging of two thermoset polymers, an epoxysystem and a polyester mortar, in standard and in aggressive environments.

The first part of the study focused on the evolution of these materials in astandard environment (20°C, 50 relative humidity). Polymerization kinetic of theepoxy system was studied by 1RTF spectroscopy. These experiments showed thatthe extend of reaction at 20°C is limited to 0.9 and that some monomers still remaintrapped in the polymer network. Therefore, further variations of the temperature canlead to small evolutions of the crosslink density. DSC experiments also revealed thata physical aging phenomenon can occur at ambient temperature, leading to adecrease of the material enthalpy and volume. Further studies are needed in order toevaluate the influence of physical aging on the mechanical properties of thermosets.

In the second part, samples of the two materials were immersed in variousaqueous solutions (distilled water, salt water and simulated concrete pore solution)in order to simulate the effect of aggressive environments.

The mass uptake of samples was first monitored as a function of immersion time.

For the epoxy system, mass uptake is related to the diffusion of liquid into thematerial and seemed to follow a Fickian behavior. For the polyester mortar, aninteresting feature was the mass loss resulting from immersion in the alkalinesolution, which was attributed to chemical degradations of the polymer network.

Surface analyses of the aged samples were then performed by FT1R-ATRspectroscopy. Experiments showed that the chemical structure of the two materialsis not affected by immersion in distilled water or in the salt solution. However,immersion in the alkaline solution induced saponification (ester hydrolysis) of thepolyester network.

Finally, the viscoelastic behavior of aged samples was investigated by dynamicmechanical analysis. Plasticizing effects accompanied by a significant decrease ofthe storage modulus at ambient temperature were observed for all samples immersedin any of the three model solutions.

The authors would like to thank F. Farcas, P. Bartolomeo and E. Massieu(LCPC) for their contribution to this work.

5. Bibliography

Chin J.W., Aouadi K., Haight M.R., Hugues W.L., Nguyen T., "Effects of water, salt solutionand simulated concrete pore solution on the properties of composite matrix resins used incivil engineering applications", Polymer Composites, vol. 22, 2001, p. 282.


Chin J.W., Nguyen T., Aouadi K., 1999, "Sorption of water, salt water and concrete poresolution in composite matrices", Journal of Applied Polymer Science, vol. 71, 1999, p.483-492

Cranck J., Park G.S., Diffusion in polymers, New-York, Academic Press, 1968.

Karbhari V.M. And Zhao L., "Use of composites for 21st century civil infrastructure",Computer Methods in Applied Mechanics and Engineering, vol. 185,2000, p. 433.

Moy P., Karasz F.E., Polymer Engineering and Science, vol. 20,1980, p. 315.

Mukhopadhyaya P., Swamy R.N., Lynsdale C.J., "Influence of aggressive exposureconditions on the behavior of adhesive bonded concrete-GFRP joints", Construction andBuilding Materials, vol 12, 1998, p. 427-446.

Nogueira P., Ramirez C., Torres A., Abad M.J., Cano J., Lopez J., Lopez Bueno I., Barral L.,"Effect of water sorption on the structure and mechanical properties of an epoxy resinsystem", Journal of Aplied Polymer Science, vol. 80, 2001, p. 71-80.

Struik L.C.E., Physical ageing of amorphous polymers and other materials, Amsterdam,Elsevier, 1978.

Toutanji A., Gomez W., "Durability characteristics of concrete beams externally bonded withFRP Composite Sheets", Cement and Concrete Composites, vol. 19, 1997, p. 351.

Zanni-Deffarges M.P., Shanahan M.E.R., "Diffusion of water into an epoxy adhesive :comparison between bulk behavior and adhesive joints", Int. Journal of Adhesion andAdhesives,\o\. 15, 1995, p.137-142.

Overwrapped Structures : A ModernApproach ?

M J Hinton*, J Cook**, A Groves**, R Hayman** and A Howard'

* Future Systems Technology Division (FST), QinetiQ, Fort Halstead, Sevenoaks,Kent, TN14 7BP, UK.

** Structures and Materials and Centre, FST, QinetiQ, Farnborough, Hampshire,GU14 OLX, UK.

©QinetiQ Ltd 2002E-mail to mi [email protected]

ABSTRACT. The concept of overcropping a pressure vessel with high strength material in theform of wires or hoops has a history going back at least as far as the 13' century. In recentyears, the availability of reinforcing fibres with very high strength to weight ratios has giventhis ancient concept a new lease of life. This paper starts from the early history of the subject,setting in context the opportunities that are now possible with new high performancematerials. Particular attention is given to the concept of tensioned overwrapping where thetheory is presented for both thick and thin walled pressure vessels. Finally, examples oflightweight, tension-overwrapped structures are presented to illustrate the current state of theart.


1. Introduction

Although the overwinding of pressure vessels is a very old concept, theavailability in recent years of high strength, low density, fibres together with theintroduction of some novel manufacturing techniques, is leading to some excitingtechnical developments and a range of new applications.

All of the early applications of overwinding were to vessels that were broadlycylindrical in shape, typically guns or gas storage tanks. Originally, the idea ofoverwinding was based on the observation that many materials (usually metals) canbe made with a much higher tensile strength when they are in the form of wires orfilaments than they can in bulk. For example, drawn steel wire may be considerablystronger than a casting of similar composition. This arises partly by virtue of thecontrolled amount of cold work involved in the drawing process, partly by virtue ofthe better control of the heat treatment when in finely divided form and partly by theavoidance of the large defects that can occur, particularly in castings.

The other driver behind overwinding is that in a cylindrical vessel subjected tointernal pressure loading the circumferential, or hoop, membrane load exceeds thelongitudinal load by a significant factor. For a closed cylindrical vessel thehoop/longitudinal load ratio is approximately 2:1. For open-ended or partially open-ended vessels it is higher. The load ratios occurring in various types of pressurevessel are illustrated in Fig. 1. In an overwound cylinder the load is partitioned sothat the fibres take at least half the hoop load, leaving the metal in a balanced bi-axial stress state (i.e. approximately 1:1) in which it acts at close to maximumefficiency. This is illustrated in Fig. 2. The weight saving is achieved from the factthat the overwind has a higher strength to weight ratio than the bulk material itreplaces.

If the strain to failure of the fibre exceeds that of the bulk material by a sizeablemargin then, for a pressure vessel, additional benefit can be obtained by applying theoverwind under tension. The effect of this is to drive the bulk metal intocircumferential and radial compression. The idea is illustrated in Fig. 3, whichshows a stress-strain curve for a typical metal liner material. Without pre-stressing,the metal would start at a state represented by point A and then move under theeffects of the pressure loading to point B. With a tensioned overwrap, it is possibleto start at point C and move to point B. The effective extension of the elastic rangeis obvious. However, the overwind starts in a state of tension and then experiencesthe same incremental strain as the liner during pressurisation. It follows that for pre-tensioning to be viable, the overwinding fibre must have an appreciably greaterbreaking strain than the liner material, typically by a factor of at least two. It is alsohelpful, although not essential, if its modulus is also at least comparable with that ofthe liner material.


The overall result of applying a tensioned overwind to a pressure vessel is thatwhen the vessel is subsequently pressurised, it reaches its yield point or its ultimatetensile strength at a higher value of internal pressure than would otherwise be thecase. This effect can be used to increase the burst pressure, increase the fatigue life,give further reductions in weight or achieve some combination of all of these.

The modern fibres that are currently available have now made overwinding aneven more attractive proposition than it ever was. These fibres can be made intocomposites with unidirectional strength to weight ratios exceeding those of bulkmetals by factors up to about 10 (Fig. 4). This has enabled spectacular weightsavings to be achieved on overwound structures comparatively easily, often by asimple extension of the existing manufacturing method.

The principles of overwinding are also applicable to pressure vessels of non-cylindrical shape, and one striking example of this, namely toroidal overwinding, isalso discussed in section 7.

2. History of Overwinding

The technology for producing large monolithic metal structures started todevelop from the 15th century, and then only in a very imperfect form. Overwindingwas first introduced as means of circumventing this difficulty by allowing largepressurised structures to be built up from moderately sized components. Later,when it became possible to cast or forge large metallic pressure vessels of acceptablequality, overwinding was retained and used instead as a means of improving theirstructural performance, a trend that continues to this day.

Since the thirteenth century, and perhaps earlier, it has been appreciated thatwrapping a strong reinforcing material in a hoopwise manner around the outside of astructure increases its ability to withstand internal pressure. Early barrels for thestorage of foodstuffs employed metal hoops that held together an assembly ofwooden bars or staves. Exactly the same technique was used to produce the earliestcannons (which are also pressure vessels) in the early fourteenth century. Closelyfitting staves would be placed around a wooden mandrel and temporarily fixed inplace. Initially these staves were of wood and later of iron. Hot iron rings would beslipped onto the assembly and as they cooled would shrink and thereby press thestaves tightly together. This is shown schematically in Fig. 5. In the case of ironbarrels, the staves were then welded by raising to a white heat and the woodenmandrel subsequently removed or burnt out. It can be seen that weaknesses werebound to occur by this method of manufacture, and in the latter part of the fourteenth


century, when the casting of iron had sufficiently progressed, smaller barrels werecast in one piece. However the practice of manufacture using iron staves andreinforcing hoops was retained for larger barrels, of which the most famous exampleis the 'Mons Meg' cannon. This barrel, which is 14 feet long and of 20 inch calibre,was produced in 1453, and may still be seen on public display in Edinburgh.Frequently these built-up barrels were wrapped in leather and wound with rope toprotect the structure from damage and corrosion. This method of gun barrelconstruction remained unchanged until the introduction of wrought iron, which hadsuperior strength and reliability to cast iron. Wrought iron was used to make theinner tube of the barrel, but it was still reinforced externally with iron rings for extrastrength.

In the seventeenth century, the first lightweight gun barrels were designed andproduced in Sweden. These were fabricated from hardened leather with iron orbrass reinforcing hoops and lasted 5-10 shots. A later barrel design of this typeconsisted of a thin copper tube lashed with rope and covered with leather. Thebarrel screwed onto a brass breech, itself strengthened with strips of iron. The useof this type of construction was widespread in Europe, notably in Scotland andSwitzerland. The enhanced portability made possible by the comparatively lowweight was the principle attraction.

In time, steel was introduced and used to produce the inner tube of the barrel.Wrought iron was still used for the hoops, which were shrunk on and varied inthickness to provide the requisite strength. Thus thicker hoops were used over thechamber section to contain the highest pressures and thinner hoops towards themuzzle end of the barrel where the internal pressure is lower. By the late nineteenthcentury these hoops were also being produced from steel. The higher strengthmaterial allowed thinner sections and lighter barrels to be made. However theintegrity of these hoops had to be taken on trust. Imperfections in the structure wereonly discovered when the gun was fired. It was after a number of serious incidentsinvolving bursting guns that the need to carry out a proof pressure test prior to usebecame recognised.

By the mid 1850s, as gun sizes and gun power dramatically increased, the idea ofusing highly drawn wire instead of hoops had been mooted, but it was not until the1880s that this was implemented. After the basic tube had been produced it wasrotated in suitable machinery and drawn steel was wound on under tension.Inspection of the wire during winding, and the fact that the tensioning process itselftested the strength of the wire, increased confidence in the integrity of the finishedbarrel. The tension also resulted in the inner tube being compressed, similarly to thebarrels with shrunk-on hoops, and being able to withstand higher firing pressures asa result. This method of manufacture also resulted in lighter barrels, the first ofwhich was of 9.2inch calibre produced at the Royal Gun Factory in 1884. Wireoverwound construction then became the standard construction for British guns for


the next thirty years, encompassing naval and artillery pieces ranging from 3" to 15"calibre. This manufacturing approach is shown in Fig. 6, where the basic gunconstruction and quantities of wire used (several hundred miles per gun!) are clearlyillustrated.

The one drawback of wire wound guns, and the reason why their use was notmore widespread, was that the wire wrapping provided no longitudinal stiffness tothe barrel. This meant that gun barrels of this type were prone to droop under theirown weight and to 'whip' on firing the shot, and both of these led to increasedprojectile dispersion. For the early guns the inaccuracy resulting from this wasinsignificant compared with all the other sources of error, but as gun designs becamemore advanced the effect became noticeable. Wire winding was eventually replacedby 1/24-inch strip steel, which in turn was followed by shrunk-fit compoundcylinders and finally over the last thirty to forty years by monobloc forgingsmachined to final dimensions.

In addition, a technique known as autofrettage is now widely used. This consistsof applying internal pressure to the barrel to take it beyond yield. On removal of thepressure loading, the barrel bore is then left in a state of circumferential compressivepre-stress in a similar manner to that brought about by tensioned overwrapping (Fig.3). The main purpose of autofrettage is to aid fatigue life. It can be used on non-overwrapped thick wall tubes. For overwound vessels it can be employed as analternative to the use of winding tension.

More details on the history of the use of overwinding on guns are given inreferences 1 to 7.

3. Theory of Overwinding

Given that the idea of an overwrapped cylinder dates back to the 13th century, itis not surprising that numerous theories for modelling overwrapped and multi-cylinder pressure vessels have been developed. However, accurate methods ofanalysis emerged only towards the end of 19th century when the classical theories of'Elasticity' emerged based on advanced calculus techniques. In essence suchmethods of analysis arose from the need for Victorian engineers to enhance theirunderstanding of structures following the rapid industrialisation in the UK andelsewhere during the 19th century.

The universally accepted and definitive design equations for pressure vessels canbe ascribed to Lame 8 who solved the elastic equations of state for the type ofcylindrical vessel shown in Fig. 7 for both the circumferential and radial stresses toobtain:


where r is the radius through the vessel and A and B are constants of integration.The values of A and B values are derived from application of the boundaryconditions. For an internally pressurised cylinder the constants are simply derivedby setting the radial stress to zero at the outside radius and a value equal, butopposite in sign,8 to the applied internal pressure Pi at the inner radius. Equations (1)and (2) can be suitably modified to account for material anisotropy, that is tomaterials whose elastic properties are different in the radial (r) and circumferential(0) directions, as shown in Fig. 7. Details of this more complex analysis are given inreference 9.

While these equations are valid for all cylindrical pressure vessels, it can beshown that for very thin-walled pressure vessels, they can be greatly simplified. Forcases where the internal pressure exceeds the external pressure, the respectivestandard thin-walled circumferential and radial stress equations reduce simply to:

where P0 is the external pressure, R is the mean radius and t is the wall thickness.Such equations are significantly easier to use than the quadratic type equationsdeveloped by Lame. As a result, for constructions involving isotropic materials, thethin-walled cylinder equations can be used with little error when the ratio of R/t isten or greater. Rocket motor cases fall into this category. However, for gun barrels,where the R/t ratio can approach unity, it is necessary to revert to the Lameequations.

Where materials are highly anisotropic, as is the case with fibre reinforcedpolymer composites, then the above guideline is no longer valid. For materials ofthis kind the radial modulus (£R) will be significantly lower than the circumferentialmodulus (EH), possibly some forty times lower. When internally pressurised, thereis a tendency for the tube wall to contract radially, i.e. effectively squash, whichleads to difficulties in transferring load into the outermost rings of fibres. Fig. 8shows the hoop stress distribution as a function of radius for internally pressurisedthick walled tubes (of R/t = 3) having varying degrees of anisotropy. It is evidentthat for a EH/ER ratio in excess of 10, the non-uniformity of fibre loading becomesappreciable. For this effect to disappear the R/t value would have to be 30 or more


for a typical carbon fibre composite. It follows that considerable caution must beapplied when designing with advanced composites for pressure vessel applications.This is equally true for composite overwraps on metallic liners.

Methods for analysing overwrapped cylinders consisting of steel wire or layeredsteel strip had been developed by the start of the 20th century. These arose partlyfrom the need to model gun barrels and other high performance pressure vessels. Inthese theoretical developments simple compound cylinders were modelled viaequations (1) and (2) for each layer in turn. A succession of simultaneous equationswas then built up and then solved for the resulting constants of integration. Wherediffering materials were used, use was made of the Hookian equations8 relatingstress to strain. At the same time continuity of radial displacement was maintainedacross material boundaries.

For more complex situations, where pre-stressing is imposed by thermalcontraction of an outer cylinder, equations (1) and (2) are used in combination withthe Hookian stress/strain equations, but with an additional thermal expansion termccAT. Here a is the thermal expansion coefficient and AT the shrink fit temperature.A series of simultaneous equations is again developed to obtain the integrationconstants.

Pre-stressing was quickly recognised as a method of:

• Inducing a compressive pre-stress in the liner to increase the effectiveelastic range of the material and thereby increase the operating pressure, asillustrated earlier in Fig. 3;

• Offsetting thermal mis-match problems between dissimilar materials;

• Rigidly clamping the cylinder components together.

For the tension winding process, the theory has been developed whereby thetension overwrap is mathematically represented by a pre-tensioned 'elastic' bandapplied around the liner and previously applied layers. The equations of state areagain those developed by Lame suitably modified to Include material anisotropy asappropriate. However, to determine the level of contraction, conservation of energyis applied whereby the sum of the forces through the tensioning layer, previouslyapplied layers and liner is integrated to zero. The resulting compressive stresschange is then added to the stress state in all previously applied layers and lineraccording to the principle of superposition. For the case where the layers are verythin, e.g. composite overwraps which are typically O.lmm thick, the resultingsummation can be represented by an integral expression to reduce numericalcomputation times.


The effect of temperature and internal pressure combined can be easilyaccommodated by a simple extension of this approach.

4. Materials Selection

Table 1 outlines the mechanical and physical properties of a selection of somecurrently available reinforcing fibres as well as traditional reinforcements such aspiano wire, leather and cast iron. The large difference between the two classes ofmaterial explains why overwinding has received a new lease of life in recent years.

5. Dry Overwinding

For fibre reinforced composite laminates in general, a matrix is essential fortransferring loads from ply to ply. Without this mechanism, it would be impossibleto stress a muti-layer composite as intended. In an overwind, where the composite isessentially unidirectional, this inter-ply load transfer mechanism is not needed.Nevertheless, the matrix still performs two further important functions. Firstly it actsas a lubricant during the forming operation, be this filament winding, pressing ormoulding, thus preventing fibre damage. It also protects the fibres from frettingagainst each other during service. Secondly it allows more strength to be realisedfrom the fibres by virtue of the length-strength effect. The essence of this effect isthe observation that for all types of fibre the average measured strength decreases asthe length under test (the gauge length) increases. For glass and carbon fibres, themagnitude of this is of the order of 10% strength reduction every time the gaugelength is doubled. This is a direct consequence of the strength being dominated bythe presence of flaws within the fibres and the higher probability of a critical flawexisting in a long fibre than in a short one. This raises the question of what theeffective fibre gauge length is in a unidirectional fibre composite. Where a fibrebreaks, the load it was carrying is transferred into neighbouring fibres through thematrix and back in again at the far side of the break. The length over which thisoccurs (i.e. the effective gauge length) depends on the shear modulus of the matrixand the interfacial shear strength. For carbon or glass fibre reinforced plastics it isof the order of a millimetre. Without the matrix being present, the effective gaugelength would be much greater, as the only load transfer mechanism available isfriction between fibres. A simple estimate suggests that the effective gauge lengthmight be of the order of the tube diameter. From the above figures, it is evident thatthis would seriously degrade the realisable strength.

However, the strength of drawn wire and of ropes and leather strips is largelyindependent of gauge length. Moreover none of these materials is particularly


sensitive to abrasion. Consequently these considerations hardly apply, whichexplains why they could be applied very effectively as overwraps without anymatrix to bind them together. More recently, since the mid-1970s, a class of fibresknown as aramids (current trade names Kevlar or Twaron) has becomecommercially available. These fibres have strength to weight ratios similar to thoseof carbon fibres but have some useful additional characteristics. Firstly, whencoated with an appropriate size, the aramid behaves as a textile fibre and needs nofurther lubrication. For similar reasons it is not prone to fretting. Secondly, withthese fibres the length-strength effect is so small it is difficult to measure.Consequently, these fibres also offer the prospect of dispensing with a resin matrix.

These considerations came to the fore in the late 1970s, when a requirementarose in the UK for a rocket motor case for a weapon known as LAW 80 (Fig. 9).This was to be designed as a cheap man-portable unguided anti-tank weapon thatwas to be manufactured in considerable numbers. The central feature of LAW 80was a projectile consisting of a warhead launched by a rocket motor. A conventionalsolid propellant rocket motor is essentially a cylindrical pressure vessel containingthe propellant charge. The propellant generates gas as it burns and this gas exits therear of the motor through a relatively small aperture (nozzle) thereby creating thrust.Structurally, a rocket motor case can normally be treated, to a good approximation,as a closed cylindrical vessel.

For LAW 80, the requirements for the case were:

• Low, but not absolutely minimum weight;

• Of low cost, implying rapid production;

• To be manufactured in an ordnance factory with limited experience of non-me tallies.

An exceptional feature of LAW 80 was that it had an extremely large throat byrocket motor standards (Fig. 10). This, in turn meant that the membrane loads in thecylindrical wall were in the ratio of 4:1 rather than 2:1 as would be the case in aclosed-ended cylindrical vessel. This fact renders the LAW 80 case a primecandidate for overwinding as, in principle, approximately three-quarters of the metalcan be replaced by a lightweight overwind. While this would deliver the requiredweight savings, there was concern that conventional wet winding would beunacceptabiy slow for a mass produced item of this kind. In view of this, a decisionwas made to pursue the dry overwinding route, and this resulted in suitable windingmachines being installed in ordnance factories within 18 months of the start of theprogramme. The technique is now established as a standard UK method for rocketmotor case construction.


There were two particularly important issues that had to be resolved before thedesign of LAW 80 could achieve safety clearance, both connected with the time-dependent properties of aramid fibres, namely stress relaxation and stress-rupture.In the context of a rocket motor case there has to be sufficient initial winding tensionto ensure that after many years of hot storage followed by firing cold (worst case),the overwind will not relax to such an extent that it slips along the motor tube underthe very high acceleration loads. At the same time, the locked in stress must not beso high that it results in failure under prolonged loading at elevated temperature (i.e.stress-rupture). An intensive programme of research was needed to establish thatthere is a viable 'window' of winding tensions that would avoid these two pitfallsand guarantee a safe design. Some of the techniques used to give this assurance aredescribed in the following section.

6. Associated Test Techniques

The main experimental technique used to establish the magnitude of therelaxation and stress-rupture effects in aramid fibres is the 'split ring' test.

With carbon fibres both relaxation and stress-rupture effects are very muchsmaller and occur at higher temperatures. The main technique appropriate tomeasuring these, the 'dog bone1 test, is also described.

The 'Split Ring' Test:

This test was developed in-house specifically to qualify aramid fibres for theLAW 80 programme. The test rig, shown in Fig. 11, consists of an eccentricallybored ring, split in the axial direction at the thinnest cross-section and bent inwardseach side of the gap. This bending prevents the fibres under test coming into contactwith either a sharp edge or a small radius that might introduce high through-thickness compressive stresses. Each ring is calibrated through suitable loadingpins. In use, lubricating tape is wound on the area of the ring that comes intocontact with the fibre and end clamps applied to pre-compress the ring by to aknown extent. The fibres are then wound on and the clamps released, leaving thefibres under a known state of stress. Subsequent opening of the gap can be relatedto the rate of relaxation. If the ring is set such that the fibre is at a sufficiently highstress then a stress-rupture failure will eventually result. The design and use of thistest rig is fully described in references 10 and 11, and other techniques used formeasuring the short-term strength of aramids in reference 12.


The advantage of the split ring is that the test specimen is compact and robust, sothat it can be readily inserted in an oven or other chamber to allow stress-rupture orrelaxation measurements to be conducted in a variety of adverse environments.

A full characterisation of the stress-rupture behaviour of any type of fibrerequires a large number of measurements of time-to-failure at various stress levelsand temperatures. This requires a large number of split rings. The plot of time-to-failure versus stress constitutes the stress-rupture curve for that temperature. Foraramids it is then possible to superpose these stress-rupture curves to a single mastercurve using temperature-time superposition. For high temperature measurements,split rings have been manufactured from maraging steel for thermal stability. Theserings are suitable for other fibre types, in particular carbon fibre and carbon prepregtows.

The 'Dog bone'Test

The split ring technique, described above, was designed for stress-rupture testingof single tows in a range of environments. An alternative technique, known as the'dog bone' because of its shape, has been devised to test resin multi-layeredimpregnated carbon fibre over-wraps at high service temperatures. Fig. 12 showsthis test specimen, which comprises a short thin-walled steel cylinder, over-wrappedwith a number of layers of tensioned prepreg tow. This test has been used tomonitor the progressive relaxation of the overwrap material at elevated temperatureby measuring the changes in the internal bore. Because carbon fibres are so stable,this test in effect measures the relaxation effects in the resin.

The 'dog bone' test piece can also be used to determine the residual strength ofthe overwind by internally pressurising the cylinders to failure after a period ofexposure.

7. Toroidal Pressure Vessels

The overwinding of toroidal vessels is a direct extension of the dry windingtechnique used for rocket motor cases. In studying the use of 'Breathing Apparatus'by fire brigades and divers, it became apparent that there would be considerableergonomic advantages to be derived from containing the compressed air supply in atorus shaped vessel rather than in the conventional cylindrical geometry. Forexample, it would protrude far less from the back, be far more comfortable to wearand the pressure regulator could be sited in a protected position in the central hole.As a structure, a toroidal pressure vessel has a similar efficiency to that of acylinder. However, over the years, the mass of cylindrical vessels has been


progressively reduced by the use of filament wound construction on top of a thinmetallic liner. To achieve a similar result with a toroidal vessel is not so simple, byvirtue of its topology.

Toroidal winding machines are widely used in the electrical industry for windingtransformers and other items of equipment but not for filament winding. It wasevident that such a machine could overcome the problem of feeding the filamentsthrough the central hole, and a small machine of this type had previously been usedin QINETIQ to investigate the feasibility of winding carbon and aramid fibres oncomponents with a central hole. For the breathing apparatus application, thequestion was the extent to which a fully filament wound solution was feasible. Wetfilament winding on to a torus is extremely difficult by virtue of the complexity ofthe machine and the fact that it would need to be gripped by rollers that would needto contact the uncured resin. Rapid indexing of the torus during winding to produce'helical1 patterns presents further theoretical and practical difficulties that render fullfilament wound solutions unattractive. Dry overwinding represents the bestcompromise, and a vessel made in this way is approximately half the weight of theall-metal equivalent. While this is not as light as a composite cylinder of the samevolume, with the toroidal shape it is possible to eliminate the mass of some of thestructure needed to mount the vessel on the body, and this broadly compensates forthe additional mass of the vessel itself. The advantage of the torus then manifestsitself in all the ergonomic advantages discussed previously. An overwound toroidalpressure vessel complete with a pressure inlet is shown in Fig. 13. More details onthe design and construction of this vessel are given in reference 13.

8. Concluding Remarks

• The dry overwrap technique is now well established in the UK, and is now afavoured method of construction for rocket motor cases. There may also be somescope for the use of carbon fibre overwinds as a means of achieving similar benefitswithin a smaller volume.

• The application of modern fibre materials to the overwinding of guns is a veryattractive option, and although not discussed in detail in this paper, is an area whereQINETIQ is actively researching at present. Some of this work is reviewed inreference 14.

• There are potentially very large markets for overwound toroidal vessels in bothbreathing apparatus and vehicle applications. The technology is still far frommature, but the design problems are well on the way to being solved, as are thewinding issues.


• It is worth noting that for both guns and toroidal vessels, overwinding isperceived to be the only viable method of achieving weight reductions. For guns,fully composite solutions are ruled out on grounds of wear, erosion and temperaturecapability. For toroidal vessels they are likely to be ruled out on grounds ofmanufacturing complexity.

• One recent and rapidly growing market for lightweight composite pressurevessels is in offshore oil and gas. Several initiatives are underway to developflexible risers and pressurised valve assemblies, where it is believed that tension-overwrapped structures may offer an attractive alternative.


9. References

Hogg I., Batchelor J., "Naval Gun", 1978, Blandford Press.

V. Ian., I. Hogg., "British and American Artillery of World War 2", 1978, Anns and ArmourPress.

Carman W.Y., "A History of Firearms", Routledge and Kegan Paul, 4ed., 1970.

Gardine R., "The Eclipse of the Big Gun: The Warship 1906-45", 1992.

The Handbook of Artillery Weapons', RCMS, Shrivenham, 1988.

Hodges P., "The Big Gun - Battleship Main Armament, 1860-1945", 1981.

H. Melvin., Jackson H., "Eighteenth Century Gun-Founding", 1973.

Timoshenkol S., Goodier J.N., "Theory of Elasticity"' 3ed., 1970, McGraw-Hill BookCompany, New York.

Groves A., Margetson A.J., in Proceedings of the IMechE, Design in Composite Materials, 7-9 March 1989. 'A Design Assessment for Metallic Pressure Vessels CircumferentiallyReinforced with a Pre-tensioned High-specific Strength Anisotropic CompositeOverwind'.

Cook J., Howard A.,: in RISO Conference (Denmark) pp 187-192, 1982. "A Compact HoopTest for Determining the Creep and Static Fatigue of Nominally Elastic Fibres andRings".

Cook J., Howard A., Parratt N.J., Potter K.D., : in RISO Conference (Denmark) pp 192-197,1982. "Creep and Static Fatigue of Aromatic Polyamide Fibres".

Cook J., : in TEQC 1983, University of Surrey, publ. Butterworths, 1983. "Tensile StrengthTesting and Quality Control Procedures for Aromatic Polyamide Yarns".

Cook J., Chambers J. K., Richardsl B.J., : in European SAMPE Conference, Paris, April1998. "Toroidal Pressure Vessels for Breathing Apparatus".

Groves, Hinton M. J., Howard A., : in Proceedings of 17th International Symposium onBallistics, Midrand, South Africa, 1998. "A Review of the DERA Composite ReinforcedGun Barrel Programme".

Fibre type Young'smodulus

GPa

Tensilestrength

GPa

Densitykgm3

xl03

Specific modulusGPa^kg'1

xl03

Specific strengthGPa-m^kg1

xl03

Fibrediameter

MmVarious fibre types

Carbon fibresHigh strength - PAN-basedInter modulus - PAN-basedHigh modulus - PAN-basedUltra high modulus - Pitch-based

Aramid fibresKevlar 49Twaron

Glass fibresE-glassS-glass

224 - 235294 - 303380 - 436588 - 827

117-130115

7390

3.53 -4.05.3 - 5.641.9 -4.212.2 - 2.37

2.7 - 2.92.8

3.44.7

1.75-1.791.77-1.91.84-1.91.94-2.18

1.451.45

2.602.49

128 -131166 -159206 -229303 -379

80.7 -89.679.3

28.136.1

2.02 -2.232.99 -2.971.03 -2.221.13 -1.09

1.86 -2.001.93

1.311.89

755

10

1112

1510

Various resin systemsEpoxy

Bismaleimide

2.6 -3.8

3.2 -5.0

0.06 -0.085

0.048-0.110

1.1 -1.2

1.2 -1.32

2.36-3.17

2.67 - 3.79

0.054-0.071

0,040 - 0.083

-

-

Other materials for comparison

Designated as Piano wire1

Designated as Pianoforte hard rawn2

Cast iron - grey

Cast iron - whiteLeather belt

210-

110152

-

3.01.86 -2.33

10023030 -50

7.8-

7.157.70

-

26.9-

15.4

19.7

-

0.38

0.23 -0.303

14.0

29.9

-

----

-1 Science Data Book Kaye and Laby 3 Using the Kaye and Laby density value

Table 1: Mechanical properties of a selection of fibres and resins


a) Closed cylindrical pressure vesselb) Open-ended cylindrical pressure vessel

c) Intermediate case - rocket motor with large throat

Figure 1 : The membrane load ratios in various types of cylindrical pressure vessel


Figure 2 : The principle of overwinding is that approximately half the thickness ofthe metal can be replaced by overwind still leaving sufficient metal to carry the

longitudinal load.

Note 1: In practice, the overwind ends and vessel end closures require carefuldesign.

Note 2: When the tri-axial stress state in the metal is taken into account, rather lessthan half the metal (typically 43%) can be substituted in this way.

a) Section through a closed cylindrical vessel of monolithic metal,b) Section through an overwrapped closed cylindrical vessel.


Figure 3 : A schematic stress-strain curve for an overwound metal liner.Without pre-tension in the overwind, the stress-strain state moves from point

A to point B as internal pressure is applied. With pre-tension in the overwind(=pre-compression in the liner) the liner can be made to operate over a larger

range of strain, from C to B.


Figure 4: Strength to weight ratios for a number of high strength metals andunidirectional polymer-composite materials


Figure 5 : A diagram showing how early barrels were built up around a woodenmandrel. Iron staves are temporarily held around the wood while heated iron ringsare pushed over them. The rings shrink as they cool and hold the staves tightlytogether. Finally, the entire structure is raised to a "white heat", welding the stavestogether and burning out the wooden mandrel

Figure 6: Wire winding construction technique. The wire windings can be seenclearly 1 n the dissected barrel.


Figure 7 : Circumferential and radial stresses in a thick walled cylinder underinternal pressure loading.

Figure 8 : Hoop stress distribution for thick orthotropic cylinders.


Figure 9 : A LA W80 man-portable anti-tank weapon.

Figure 10: A LAW80 rocket motor case showing the membrane loads in thecylindrical section in the ratio 4 (circumferential) : 1 (axial). In principle, threequarters of the metal in the wall can be replaced by a lightweight overwrap and

there is still sufficient to take the axial loads.


Figure 11 : The split ring test piece used for the measurement of stress relaxationand stress-rupture in aramid and other fibres.


Figure 12 : The 'dog-bone' cylindrical specimen used for the measurement ofrelaxation and loss of residual strength on thick CFRP overwraps at elevatedtemperature.


Figure 13 : A 9-litre toroidal pressure vessel overwrapped with aramid fibre.

Development of Scarf Joint AnalysisCustomized System (SJACS)

A Guide for Standard Analysis of Composite BondedRepairs

Toru Itoh* — Tadashi Tanizawa** — Shyunjiro Saoka**

* Kawasaki Heavy Industries, Ltd. 1 Kawasaki-cho, Kakamigahara City, Gifu Japan

itoh_toru@khi. co.jp

** Kawaju Techno Service Corporation, 1 Kawasaki-cho 3-chome, Akashi, HyogoJapan

[email protected]

ABSTRACT: Automated Finite Element (FE) analysis system was developed as a useful tool forthe analysis of composite bonded repairs. This system, Scarf Joint Analysis CustomizedSystem (SJACS), will guide those who have little knowledge of FE analysis and help thembuild a reliable FE model of bonded repairs and obtain reasonable results easily. The systemutilizes a commercial FE analysis code and customizes it so that FE models are generatedautomatically based on the simple input data of geometry, materials and loads. Shear andpeel stresses of adhesive layer as well as stresses of the parent structure and repair patch canbe displayed on the screen of a personal computer. The system was developed in conjunctionwith the activities of Analytical Technique Task Group of Commercial Aircraft CompositeRepair Committee.

KEYWORDS, composite repairs, scarf joints, finite element analysis, standardization, CACRC,tensile tests


1. Introduction

Composite materials have been applied to aircraft structures since more than afew decades ago. Application of composite materials reduces weight of structuresand saves fuel consumption. As the application of composite materials expanded,airlines began to realize the inconvenience of the repair of composite structures. Themain problem of composite repairs is that each Original Equipment Manufacturer(OEM) requests airlines to apply their own repair materials and repair processesaccording to their Structural Repair Manual (SRM). If airlines operate aircraftmanufactured by multiple OEM's, they should store a variety of repair materials ofdifferent material specifications and apply different repair process specificationseven though composite parts themselves look quite similar.

In 1991, Commercial Aircraft Composite Repair Committee (CACRC) wasestablished under the sponsorship of Air Transport Association (ATA), InternationalAir Transport Association (IATA), and Society of Automotive Engineers (SAE) todevelop and improve maintenance, inspection and repair of commercial aircraftcomposite structure and components as it is written in the charter of CACRC.Members of CACRC are regulatory agencies, OEM's, Airlines, TrainingOrganizations, Material Suppliers, Repair Station, and others who are interested inthe activities. Through the ten years activity, ten Aerospace Material Specifications(AMS), four Aerospace Information Reports (AIR), and five AerospaceRecommended Practices (ARP) were published. There are seven Task Groups inCACRC, i.e., Repair Materials, Repair Techniques, Design, Inspection, Training,Airline Inspection & Repair Conditions, and Analytical Repair Techniques.Members are cooperatively working to establish standard documents.

As for the standardization of analysis for composite repairs, AnalyticalTechnique Task Group (ATTG) was organized in 1999. As it is written in its charter,the purpose of this activity is to develop a guide of generally accepted stress analysismethods used for the design and substantiation of composite repairs. After two yearsdiscussion, ATTG has almost finished drafting the standard guide for the analysis ofcomposite repairs.

In 1999, New Energy and Industrial Technology Development Organization(NEDO) granted three years research on standardization of analytical technique ofcomposite repairs to Society of Japanese Aerospace Companies (SJAC) based on thesubsidy from Ministry of Economy, Trade and Industry (METI). SJAC has selectedKawasaki Heavy Industries, Ltd. (KHI) as a contractor to perform the research.SJAC and KHI have participated in ATTG of CACRC since 1999 and involved inthe activities to develop analytical standard for composite repairs.


2. Repair and assessment of composite structures

2.1. Damage and repair of composite structures

Composite parts of aircraft incur various damages during operation. The mainsources of damages are lightning strike, tool drop, service vehicle collision, andimpact by hail, runway debris, and birds. Repair methods for aircraft compositestructures are prescribed in detail in SRM. SRM is the proprietary of OEM's and isnot open to public. However, if open literatures with regard to composite repairs areinvestigated, repair methods utilized in airlines or repair stations will be made clearto some extent (Armstrong et al. 1997), (Hart-Smith et al.1986), (Niu, 1992). Repairmethods are dependent upon the type of structures, location of damages, size andtype of damages, and so forth. Figure 1 depicts damages and repairs of compositestructures.

Figure 1. Classification of damages and repair methods

Among the various repair methods, scarf bonded repair has been widely adoptedfor the repair of composite structures. Scarf bonded joint is able to transfer loadsefficiently with minimum stress concentration of adhesive layer as well as the parentstructure and repair patch at the periphery of the repair patch. Typical process of this


repair is shown in Figure 2, where a sandwich panel composed of composite skinsand a rigid form core is repaired with composite repair patch.

Figure 2. Typical bonded repair

Repairing of structures 13 5

2.2. Assessment of composite repairs

When damages are within the scope of SRM, airlines repair the damagesaccording to SRM. However, if damages found are larger than those prescribed inSRM and affect the flight safety of aircraft, airlines will ask OEM's how to repairthe parts. Since airlines do not want to ground aircraft for a long time, repair methodshould be determined in a short time. Airlines may propose repair methods toOEM's to make use of their experience and repair materials in stock. Variousanalytical methods have been proposed to evaluate the strength of bonded joints inthe past. Hart-Smith proposed a useful analytical method with computer codes in1970's, which has been widely used to evaluate the strength of bonded joints (Hart-Smith 1973). Parameters which affect the strength of bonded joints are the taperratio, the stiffness of the parent structure and repair patch, and the materialproperties of the adhesive layer. The analytical methods should take into accountthese factors. Finite Element (FE) Analysis method is also a powerful tool to analysethe bonded joints in detail especially for the complex configuration.

3. Development of SJACS

3.1. Advantages and disadvantages of FE analysis

Although Hart-Smith method is a useful tool to evaluate the strength of bondedjoint, it gives results based on the assumption introduced in the derivation of theequations. If detail analysis is necessary to evaluate the composite repairs, FEAnalysis is adequate means for the purpose. It is able to solve problems of complexcontoured parts as well as 2-dimensional repairs. While FE analysis has anoutstanding advantages as mentioned above, it usually takes weeks to make a soundFE model and obtain reasonable results. Pre- and Post processors provided bysuppliers of FE analysis codes have been improved greatly in the past decades toassist stress engineers. However, experts of FE analysis are still necessary toperform such FE analysis. In general, airlines do not have such experts of FEanalysis, or sufficient time for the evaluation of bonded repairs.

3.2. SJA CS

In order to overcome the aforementioned disadvantages of FE analysis,automated analysis scheme was developed to provide a useful tool for the analysisof composite bonded repairs. This system, Scarf Joint Analysis Customized System(SJACS), will guide those who have little knowledge of FE analysis and help them


build a reliable FE model of bonded repairs and obtain reasonable results in a shorttime. The system utilizes a commercial FE analysis code, MSC visual Nastran forWindows 2001 (vN4W), and customizes it so that FE models are generatedautomatically based on the simple input data of geometry, materials and loads. Shearand peel stresses of adhesive layer as well as stresses of the parent structure andrepair patch can be displayed on the screen of a personal compute.

This system can analyse both 1-D and 2-D repairs. Figure 3 and 4 show inputdata windows of 1-D and 2-D repairs, respectively. For 1-D repair analysis, theparent structure and the repair patch are modelled with Bar elements. Adhesive layeris modelled by combination of two non-linear rod elements aligned tangential andnormal to the adhesive layer, for vN4W does not have non-linear spring elements.The simplified modelling scheme (Loss et al. 1984) is employed in this 1-D analysissystem.

Figure 3. Input data for 1-D scarf joint analysis

As for the 2-D repair analysis, the parent (base) structure and repair patch aremodelled by Shell elements, and adhesive layer is modelled by non-linear Solidelements. A cover ply, which is very common in the actual repair, is included in the

Repairing of structures 13 7

repair patch. By changing geometric parameters of repair patch, circular patch aswell as rectangular patch with corner radius can be modelled without difficulty.

Figure 4. Input data for 2-D scarf joint analysis

4. Verification by test results

4.1. Test specimen and test conditions

Scarf joint coupon tests and thick adherend lap joint tests were performed in1999 to obtain test data to evaluate the adequacy of the bonded joint analysis. TorayFabric FF6273H-24 was used for the composite adherends, and FM-300K was usedas an adhesive material. A composite laminate was cured first and taper sanded toyield three taper ratios: 1:10, 1:15, and 1:20. Then, the same composite material waslaid up with adhesive FM-300K. Tensile tests were conducted in Low TemperatureDry (LTD), Room Temperature Dry (RTD), and Hot Temperature Dry (HTD)conditions.


4.2. Comparison of test data and output of SJACS

SJACS was used to analyse the test specimens of 1-D repair as described above.Input data for this analysis is shown in Figure 3. Shear stress and strain relation ofadhesive was taken from MIL-HDBK-17-1E. Figure 5 shows the shear stressdistribution along adhesive bond line for three load levels in RTD condition. Figure6 shows the result of the specimen with taper ratio 1:15, where maximum shearstresses at the edge of scarf joint are plotted against applied loads with solid points.Non-linear behaviour of adhesive was accounted for in the analysis. It is clear thatextrapolation of the analysis results in the prediction very close to the test results.

Figure 5. Adhesive shear stress distribution along bond line


Figure 6. Comparison of analysis and test result5. Conclusions

5. Conclusions

To make use of the advantages of FE analysis, Scarf Joint Analysis CustomizedSystem (SJACS) was developed, which will guide engineers who have littleknowledge of FE analysis and help them build a reliable FE model of bondedrepairs. Since FE model can be generated easily, this SJACS enables engineers toperform parametric study for the bonded joints to determine the adequate repairconfiguration. Results obtained by SJACS were compared with test results andshowed reasonable coincidence.

Acknowledgements

The authors would like to thank NEDO and Japanese Standard Association(JSA) for providing adequate guidance for this study. Our appreciation extends toMr. Kazuhiko Inoue of SJAC for encouragement and various supports in the courseof this research. The authors express appreciation for Mr. Yoshio Noguchi ofNational Aerospace Laboratory (NAL) for obtaining valuable test data. Variouscomments and suggestions for this research provided by Project Committeemembers are gratefully acknowledged.


References

Armstrong, K.B and Barrett, R.T., "Care and Repair of Advanced Composites," SAE, 1998.

Hart-Smith, L. J., "Design Details for Adhesively Bonded Repairs of Fibrous CompositeStructures," Douglas Paper 7637, 1986.

Hart-Smith, L. J., "Adhesive-Bonded Scarf and Stepped-Lap Joints," NASA CR 112237,1973.

Loss, K.R. and Kedward, K.T., "Modelling and Analysis of Peel and Shear Stress inAdhesively Bonded Joints," AIAA Paper, 84-0913.

Niu, M.C.Y., "Composite Airframe Structures," Conmilit Press Ltd., 1992.

Facing the Progress of Composite Materialsin the Maintenance of Aircraft

Claude Bathias

CNAM/ITMA2 Rue Conte - 75003 PARIS - Francebathias(a),cnam.fr

I. Introduction

It is universally quoted that 80% of airline accidents and incidents are a result ofhuman error. Such error includes the actions of pilots, air traffic controllers,engineers and others. However, improper maintenance followed as the secondhighest cause of aircrafts fatalities during the 90th. While better engines, airframe,navigation systems have improved the safety of aviation over the past decades, thereare still opportunities to improve the performances of maintenance.

CarrierAmerican Airlines DC- 10Eastern Airlines L-101 1JAL 747Aloha AirlinesBM AirTours 737United Airlines DC- 10Continental ExpressNorthwest Airlines

LocationChicagoBahamasJapanHawaiiManchesterIowaTexasNorita

Initiating FailureEngine separationO-ringsBulkheadFuselage failureBurner CanFan disk failureDeicing bootEngine separation

Date5/25/795/05/838/12/854/28/881/08/897/19/899/11/913/01/94

Figure 1. Examples of maintenance error (from FAA)

The figure 1 given by the FAA, lists several accidents where the probable causewas maintenance related. In all those cases, only metallic components was involved.The figure 1 shows the importance of maintenance in the past and at the present timewhere the age of the commercial jet fleet is higher and higher. According theinventory of the Douglas company (figure 2), of the active 2863 aircrafts on 1995,


over 1167 have exceed the original 20 years design objective. Some aircrafts haveexceed thirty years of service. It means that the maintenance program must bedeveloped beyond the initial standards.

Figure 2. Inventory of Douglas commercial jet fleet (from Douglas Company)

To the manufacture, weight reductions, structural requirements,manufacturability and production costs have long been obviously priority. Onlyrecently, maintainability and repairability have been added to this list, associatedwith composite structures.

Composite material usage has increased to typically represent about 20% of allstructural weight in current aircraft design.

For the operator, this now represents a significant percentage of structurerequiring a new range of engineering skills, materials, and equipment to maintain. Ithas also necessitated the adaptation of existing inspection methods and thedevelopment of new inspection techniques to ensure the continued integrity of thesestructures. The importance of these facts has been focused in the last few years bythe number of Airworthiness Directives which have been issued on such structures.

2. Services experiences

For an historical point of view it is interesting to notice a report of BritishAirways given recently about the supersonic Concorde aircraft for which of fewcomponents were made in carbon fiber composite material. The primary flight


control surfaces are composite structures which have operated about six thousandsof flights at supersonic speed in conditions of heat and ultrasonic vibration notnormally encountered by such structures on conventional aircraft. When the first in-flight damage to Concorde rudder occurred on 1990 and with no retrieved failedparts to examine, an assumption was made that some form of impact damageinstigated a rapid failure. However, trailing edge disbond was suspected as a resultof paint stripper entering the bond line and non-destructive testing (NDT) ultrasonicinspection was introduced at the trailing edge. Following additional problems, arepeat four flight inspection of the remaining area was introduced. Realizing thatthis regime could not continue, all rudders were removed and sent to a specialistcenter for immersion C-scan inspections which, being a more sensitive techniquedetected many more areas end potential areas disbond. This caused considerabledisruption to the operations of Concorde as the repair of the structure was complexand time consuming. To enable the operation to continue and because under suchconditions so little was known about the aging effects and disbond propagation rateson the structure, that a damage limit of one square inch was set with a repeatmonitor inspection of three flights only. It does not take much imagination to realizethe resources required to continuously inspect for a square inch defect and less stillto appreciate that the probability of missing such a defect would be relatively high.Inevitably the only acceptable long term answer was to build a complex set of newsurfaces at considerable cost.

For a general point of view, the ACEE program conducted by NASA Langley isthe best documentation to illustrate in service experience about different compositecomponents (figure 3). The discussion that follows summarizes some typicalexamples:

- L-1011 Kevlar 49-Epoxy FairingsThe L-1011 fairings were fabricated with Kevlar 49 fibers (in fabric form), F-

155 and F-161 epoxy resins, and Nomex. During the ten year service evaluationperiod, the Kevlar 49-epoxy fairings installed on L-1011 aircraft were inspectedannually. Minor impact damage from equipment and foreign objects was noted onseveral fairings, primarily the honeycomb sandwich wing-to-body fairings. Surfacecracks and indentations were repaired with filler epoxy and, in general, the crack didnot propagate in service.

- B-737 Graphite Epoxy Spoilers.The B-737 spoilers used three different graphite-epoxy unidirectional, tape

systems: T300-5209, T300-2544, and AS-3501. the spoilers were fabricated withupper and lower graphite-epoxy skin, aluminium fittings, spar and honeycomb core,and fibreglass-epoxy ribs. During the 13 year-service evaluation period, severaltypes of damage were encountered, with over 75% of the damage incidents beingrelated to design details. Damage was most often due to actuator rod interferencewith the graphite-epoxy skin, which was resolved by redesigning the actuator rodends. The second most frequent cause of damage was moisture intrusion andcorrosion at the sparto-center hinge fitting splice. Miscellaneous cuts and dents


related to airline use were also encountered. Damage from hailstones, bird strikes,and ground handling equipment occurred on several spoilers.

- DC-10-Graphite-Epoxy-RuddersThe graphite Epoxy T300-5208 rudders were installed on DC-10 aircraft since

1976. There were seven incidents that required rudder repairs, including three minordisbands, rib damage due to ground handling, and damage due to lightning. Minorlightning strike damage to the trailing edge of a rudder and rib damage occurredwhile the rudder was off the aircraft for other maintenance. The lightning strikedamage was limited to the outher four layers of graphite-epoxy, and a room-temperature repair was performed in accordance with procedures established whenthe rudders were certified by the FAA. The rib damage was more extensive, and aportion of a rib was removed and rebuilt.

Components in serviceAircraft

L-1011

B-737

C-130DC- 10

B-727L-1011B-737S-76

206LCH-53TOTAL

Component

Fairing panelsAileronSpoilersHorizontal stabilizerCenter wing boxAft pylon skinUpper rudderVertical stabilizerElevatorAileronHorizontal stabilizerTail rotors and horizontalstabilizerFairing, doors, and vertical finCargo ramp skin

Originally

188

1081023151

1081014

1601

350

As of June1991

158

338221018880

511

139

Start ofservice

January 1973

July 1973

October 1974August 1975April 1976January 1987March 1980<March 1982March 1984February1979March 1981May 1981

Figure 3. NASA ACEE composite structures flight service summary

To conclude this short review, a study performed be British Aerospace and UnitedAirlines in 1998 is summarized. This study concerning Airbus A320, is based on639 records and 53 airframe annual visits (figure 4). It is said that 61% of routinemaintenance actions are devoted to composite materials.


Figure 4. Distribution of maintenance record write-ups (639 records total) bymaterial type for A320 wings and stabilizer (from British Aerospace)

3. Source of Defects and Damage

According to the services sources, many factors can influence the maintainabilityof composite components. Among them, the most important are listed below:

- Conceptual design: damage resistance, hole effect more important than fatigueresistance

- Manufacturing defects: voids, delamination, surface impacts

- In service defects: penetration damage, erosion, delaminations, moisture,temperature, lightning.

4. Inspections

Compared to the relative simplicity of conventional metallic structures, compositematerials present more complexities for maintenance.


Historically the secondary structure inspections have generally been visual. But,for primary structures the operator is increasingly having to employ more reliableways of detecting damage to ensure continued integrity. These methods include X-rays, ultrasonics, thermography, and C-scan techniques. They all need specializedtechnical engineers to accomplish and inevitably have additional requests. Theinspection effort is directed towards disbond and delamination, the main agents aremoisture, followed impacts manufacturing, processing problems, and corrosion ofaluminium honey comb cored structures. X-rays and thermography will successfuldetect moisture. Experience (usually very costly) can however dictate that apredetermined level of detectable moisture is cause for removal and repair.

The ultrasonic (single side or through transmission technique) and C-scaninspections detect disbond and determination but can be affected by skin thicknessand skin-to-core bond-line irregularities. C-scan requires the part to be removedfrom the aircraft.

Three recommendations for NDT to be successful in detecting the extend ofdamage in composite structures are:

- Suitable NDT methods and facilities including safety

- Excellent operators of NDI equipment to ensure accurate and reliable results

- Available data bases because NDT methods are comparative in nature.

5. Repair

Repairing even relatively minor damage in composite components requiresspecific materials, highly experienced technicians, special tooling, interpretation oforiginal drawings. Furthermore, a controlled temperature and moisture is mandatoryduring repair.

It is becoming apparent to the operator that the material and the time-consumingpreparatory work for the repair of composite structures are factors not taken intoaccount at the design stage, but it is important for the operational economics of theaircraft.

6. In - service lessons from Airbus fleet

In - service lessons from maintenance, inspection and repair of the Airbus fleet arevery interesting because composite structures were extensively introduced for morethan 20 years.


Also, investigators find themselves facing the possibility that, for the first timeever, mechanical failure of a composite part may have played a role in the crash forAmerican Flight 587 on November 2001.

Following Aviation Weeks, this accident was the 15th fatal accident involving anAirbus airliner (excluding acts of war terrorist) in the 20th years since the Europeanplane-maker's initial production aircraft, an A300B, first flew. In reports on each ofthe 14 mishaps, investigators concluded that mechanical or structural failure did notcause or contribute to any of the accidents.

In 13 of the 14 previous crashes, crew error was identified as the main factor.Wind shear was pegged as the cause of the 14th previous fatal mishap, summaries ofthe accidents show. Weather played a role in seven of the 14 previous accidents,including the wind shear occurrence.

Models involved in the 15 fatal Airbus accidents were three A300Bs, three A300-600s, four A310s, four A320s and one A330. Two of the crashes came during non-revenue flights (an Airbus test flight and a training flight by a carrier).

Investigators are a long way from determining precisely why Flight 587, anAirbus A300-600, went down shortly after takeoff, but so far, all indications are thatthe separation of the tail played a key role. Six attachment fittings that hold the tailto the fuselage apparently came free, meaning either something caused the pins thatsecure the fittings to break, or the fittings themselves failed. Nothing hit the tail,investigators said, meaning it broke away for some other reason.

Several different types of failure are seen on the composite fin attach lugs. The sixfin lugs attach steel double lugs, or clevises, on the fuselage. Three fin lugs failed atthe lug hole itself-both forward attachments and the aft right attachment to havefailed in net tension because the break line is essentially perpendicular to the pullforce.

The other three lugs (both center points and the aft left lug) failed away from thelug hole, because the clevises are still holding parts at the fin. The experts said thatthe aft left attachment appears to have failed in skin delamination. The center leftlug failed in a nearly straight line parallel to bolts added to try to stop a delamination(figure 5).


Figure 5. Repair and fracture of the center left lug of AA 587 vertical stabilizer(from www.ntsb.gov/Events/2001)

It is clear from the figure 5, that the fracture of the center left lug occurred outsidethe area which was repaired. For the moment, there is no evidence that one orseveral of the fittings was damaged before Airbus 587 crashed. At the contrary,Aviation Week had published a calculation showing the large rudder motions onAirbus A300-600 R, can create forces exceeding ultimate load on the vertical carbonfiber composite stabilizer.


7. Conclusions

In conclusion of this short review based on in-service lessons from aircraftmaintenance, it is shown that maintenance of composite parts is a new problem withseveral facets:

- new NDT methods

- education and training of operators

- design for maintainability

- new standardization

8.References

1- http://www.aviationnow.com

2- FAA-NASA - International Conferences on the Continued Airworthiness of Aircraft.

Possibility of Inverse-ManufacturingTechnology for Scrapped Wood usingWrapping Effect in Prepreg Sheet

Kiyoshi Kemmochi* — Hiroshi Takayanagi**Toshiaki Natsuki** — Hiroshi Tsuda**

* Faculty of Textile Science and Technology, Shinshu University3-15-1, Tokida.Ueda, Nagano 386-8567, Japan

[email protected]. ac.jp

** Smart Structure Research CenterNational Institute of Advanced Industrial Science and TechnologyTsukuba AIST Central 2, Tsukuba, 305-8568, Japan

h. [email protected]

[email protected]

[email protected]

ABSTRACT: This is a study of wood composites produced by combining unidirectional carbonfiber-reinforced plastic and wood. The mechanical properties and strength reliability of woodcomposites could be largely improved by using only a small amount of carbon fiber-reinforced plastic. The tensile and bending rigidities of wood composites were investigatedbased on the laminated plate theory and rule-of-mixtures. In analyses of bending deflectionand strain, the largest analytical errors were between the two procedures. The reason forthese differences is that the laminated plate theory deals with the off-axis stress-strainrelation of a unidirectional layer, whereas the rule-of-mixtures does not.

KEY WORDS: laminated plate theory, rule-of-mixtures, shear deformation, off-axis stress-strainrelation, anisotropy


1. Introduction

Global warming and desertification attributed to mass consumption and thewaste of energy and products have become serious problems. The preservation ofthe earth's environment and natural resources is pertinent to the survival of allhuman beings. Recently, studies on environmentally conscious composite materialshave received considerable attention. With the development of productiontechnology and improved production methods, wood composites can bemanufactured by simple processes. Various wood composites, such as thosereinforced with fiber and plastics, are currently being studied(Kawai 97).

As new materials and products are developed, it is very important to investigateand predict their mechanical properties. In this study, a composite structurecomposed of a small amount of unidirectional (UD) carbon fiber-reinforced plasticand wood was manufactured in order to improve the performance of wood andutilize the used wood, andmechanical properties wereevaluated with the use oftensile and bending tests.The effect of the plyorientation and thicknessratio of a UD layer on therigidity of wood compositeswere investigated with theuse of the laminated platetheory and rule-of-mixtures.

Figure 1. A schematic diagram of a tensile test

2. Materials and methods

Vertically sawn western hemlock (Tsuga heterophyJJa Sarg., specific gravity0.43 in dry air) is used for wood.

Prepreg,P2053-15 (carbonfiber T800H, epoxy resin2500, weight percent 30 ofresin) produced by TorayCo., Ltd., Japan, was used.It adheres to a 31 cm x31cm wooden board at atemperature of 170 , apressure of 0.5 MPa, and aholding time of 90 min.Specimens were cut outfrom board by using a

Figure 2. A schematic diagram of a three-pointbending test


numerically controlled router. The tensile specimens shown in Fig.l were processedaccording to the Japanese Industrial Standard Z2112 scaled down to 290mm from390mm. Table 1 lists the specimen dimensions under the tensile and bending tests.Specimens of T and B, shown in Table 1, are the specimens without the carbonfiber-reinforced plastic. The number of wood composite specimens was five,whereas the numbers of Specimens T and B were 15 and 10, respectively. Tensileand three-point bending tests were carried out with an Instron testing machine. Forthe three-point bending test shown in Fig. 2, the distance between supporting noseswas 240mm, and specimens were loaded at a loading rate of l0mm/min. The elasticconstants of the UD layer and wood are shown in Table 2. The shear modulus of theUD layer was calculated from Hayashi's equation for anisotropic plates. Thebending modulus of wood was obtained from the deflection caused by the bendingmoment obtained by reducing shear deformation. Because the transverse Young'smodulus and the shear modulus of wood were not dominant, they are assumed to beone-twentieth (Sawada 70) of the longitudinal Young's modulus. Poisson's ratio ofwood was assumed to be 0.4 (Sawada 70).

Table 1. Specimen dimensions of wood composites

Kinds of

loading

Tensile

Bending

Specimen

No.

T

TP1

TP2

TP3

B

BP1

BP2

BP3

Number

of plies

0

1

3

5

0

1

3

5

UD layer

thickness

(mm)

-

0.137

0.412

0.686

-

0.134

0.424

0.695

Thickness

(mm)

15

15

17

17

Width

(mm)

5

5

17

17

Table 2. Elastic constants of UP layer and woodUP layer Wood

PropertiesTensile

162.0

8.81

4.57

0.332

Tensile

11.4

0.57

0.57

0.40

Bending

12.5

0.69

0.690.40



3.1. Calculation of tensile and bending rigidities of wood composites

Consider wood composites of a UD layer thickness hs, wood thickness hc, andwidth b, as illustrated in Figs.l and 2. Wood composites were subjected to tensileand three-point bending load as shown in Figs.l and 2, respectively. The coordinatesystem is also shown in Figs.l and 2. The tensile and bending specimens weresymmetrical with respect to the x-axis. In the analysis, the principal materialdirections of the wood coincided with the x- and y-axes and were fixed, whereas theprincipal material directions of UD layer varied. The effects of the ply orientationand thickness ratio of the UD layer on the rigidity of wood composites wereinvestigated with the use of the laminated plate theory and the rule-of-mixtures.

3.1.1. Laminated plate theory (Tsai et al., 1980)

Based on Hooke's law, the force tensor {N} and the moment tensor {M} of awood composite can be written as

For a symmetrical wood composite, the matrix elements of Aij, By, and Dij.respectively, are expressed as follows

where Is and IC are the geometrical moments of inertia of the UD layers and wood,

respectively. are the off-axis modulus components for the UD layersand

and wood, respectively. is the mean modulus component.

Considering that the wood composites are subjected to tensile load as shown inFig.l, Young's modulus E,.(0) in the .x-axis direction can be obtained by

For a beam under three-point bending with a span of L as shown in Fig. 2, thedeflection and strain at the center of a beam can be given by


where

3.1.2. Rule-of-mixtures.

The tensile modulus is given by

where Es is the modulus of UD layer, written as

and EC is the modulus of wood in the principal material direction.For the wood composite beam, the deflection and strain at the center of the beam

can be given by

3.2. Relation between the observed deflection and strain and the calculated ones.

Table 3 shows the mean experimental Young's modulus and the calculated onebased on the laminated plate theory under tensile load. The coefficients of varianceare also shown in Table 3. For tensile tests, Young's moduli of the wood compositesincreased when the thickness of the UD layers was increased. The moduli andstrength of reliability of the wood composites could be largely improved using onlya little of the fiber-reinforced plastic. For the tensile property, Young's modulusincreased by 56% when the wood was substituted for a UD layer of only 1.8 %.

Because the span depth ratio in the bending test was below 15, the sheardeformation (Sawada et al., 1968) had to be considered and was calculated based onthe energy method. Shear deformation was reduced from the observed totaldeflection. Table 4 shows the mean experimental deflections caused by the bendingmoment and the calculated ones based on the laminated plate theory under thebending load. Table 5 shows the mean experimental strains and calculated onesbased on the laminated plate theory under bending load. The coefficients of varianceare also shown in Tables 4 and 5.

The bending rigidity and strength reliability could be largely increased whenthree-ply UD layers were used. It can be shown that the bending rigidity will slowlydecrease as the UD layer thickness increases.


Table 3. Comparison of the experimental Young'smodulus with the calculated one

Specimen ^exp ^ f^ca] £exp/

No. Mean C-V- (GPa) £'(GPa) (%) v cal

T 11.4 19.5

TP1 17.8 18.1 14.2 1.25

TP2 23.3 12.6 21.7 1.07

TP3 26.6 9.6 25.2 1.06

* C.V.: Coefficient of variance.

Table 4. Experimental deflections caused by thebending moment at a bending load oflkN comparedwith the calculated ones

£ C

Specimen exp <)ca, expN o . Mean C.V." * . / ?

(mm) (%) (mm) °aAB 3.86 7.5

BP1 2.61 3.2 2.73 0.96BP2 1.63 2.1 1.59 1.03BP3 L25 3.0 1.25 1.00


Table 5. Experimental strains at a bending load of IkNcompared with calculated ones

c pSpecimen exp £*ca| exp

No. Mea" C.V.* 3 £(IP'3) (%) (IU ' ^1

B 5.80 9.7BP1 3.64 3.4 4.07 0.90BP2 2.31 4.5 2.22 1.04BP3 U61 4.9 1.64 0.98



Figures 3 and 4 contain theanalytical results for the bendingtest. The wood composites withvarious ply orientations and thethickness ratio (2h, Ih) of the UDlayer were analyzed based on thelaminated plate theory and therule-of-mixtures. Figure 3 showsthe properties of normalizeddeflection and strain with respectto ply orientation. The normalizeddeflection and strain increasedwhen the ply orientation wasincreased. Normalized deflectionand strain were lowest at 0° andhighest at 90°, remaining almostconstant at an angle greater than45°. It can be shown from thestress analysis of the UD layersand wood that the stressespredicted by the laminationtheory vary more slowly with theply orientations than thosepredicted by the rule-of-mixtures.It has been shown that theanalytical error between thelaminated plate theory and therule-of-mixtures was larger inSpecimen BP1 than in SpecimenBP3. The analytical errorbetween the laminated platetheory and the rule-of-mixtures was highest at a plyorientation of 15° to 18°. Thereason for this difference is thatthe laminated plate theorydescribes the off-axis stress-strainrelation of the UD layer, whereasthe rule-of-mixtures does not.

Figure 4 shows the variation of normalized deflection and strain with respect tothickness ratio. The calculated values of normalized deflection based on thelaminated plate theory coincided with the calculated values of normalized strain.The calculated values of normalized deflection based on the rule-of-mixtures,however, were larger than the calculated values of normalized strain.

Figure 4. Effect of thickness ratio onnormalized (a) deflection and (b) strain

Figure 3. Effect of ply orientation onnormalized deflection or strain


4. Conclusions

Tensile and bending rigidities and strength reliability can increase drasticallywhen a small UD layer is used. Experimental results show that the tensile propertycan be significantly improved by using one-ply UD layer, and the bending propertycan be significantly improved with a three-ply UD layer.

Tensile and bending properties were calculated based on the laminated platetheory and rule-of-mixtures. The effect of the ply orientation and thickness ratio of aUD layer on the rigidity of the wood composites was investigated. In analyses ofbending deflection and strain, the largest analytical errors were between the twoprocedures within a ply-orientation range of 15° to 18° and at a thickness ratio of2.5%. The reason of these differences is that the laminated plate theory involves theoff-axis stress-strain relation of the UD layer, whereas the rule-of-mixtures does not.

References

Kawai S., "Current trends in research and development on wood composite products",Mokuzai Gakkaishi., vol.43, 1997, p.617-622.

Sawada M., "Strength properties of wooden sheet materils", Mokuzai Gakkaishi, vol.16,1970, p.251-256.

Sawada M and Yamamoto H., "Studies on wooden composite beams: Deflectioncharacteristics within proportional limit of wooden composite beams", Research Bulletinsof the College Experiment Forests, College of Agriculture, Hokkaido University., vol.26,1968, p. 11-44.

Tsai S.W and Hahn H.T., "Introduction to Composite Materials'", Technomic, Connecticut,1980, p.217-276.

High temperature behavior of ceramicmatrix composites with a self healing matrix

P. Forio and J. Lamon

Laboratory of Thermostructural CompositesUMR 5801 (CNRS-SNECMA-CEA-Universite Bordeaux 1)3 allee de la Boetie33600 PessacFrancelamon@lcts. u-bordeaux.fr

A BSTRACT: The fatigue behavior of a SiC/Si-B-C composite with a self-healing multilayeredmatrix via chemical vapour infiltration (CVI), is investigated at high temperatures in air. Theinfluence of glass healing on damage and lifetime is detemined. Contribution of variousphenomena including oxidation-, loading- and temperature-related mechanisms is evaluatedon basis of tangent modulus degradations. In afirt step, features of the mechanical behaviorand damage under monotonic loading at room temperature are established.

KEYWORDS : Ceramic matrix composite, fatigue behavior, high temperature, glass healing.


Introduction

Ceramic matrix composites reinforced with long fibers are potential candidates foruse in aerospace industry, under severe conditions of temperatures and environment.For instance, the SiC/SiC composites consisting of a SiC matrix reinforced usingSiC fibers display some favorable characteristics such as high mechanical propertiesand a good resistance to high temperatures.

It is well acknowledged that the properties of fiber/matrix interfaces determine themechanical behavior of brittle-matrix composites (Evans et al., 1989, Kerans et al.,1988). Furthermore pyrocarbon (PyC) has proven to be a tremendously efficientinterphase to control fiber/matrix interactions and the composite mechanicalbehavior (Naslain, 1993, Droillard et al, 1996). But pyrocarbon is sensitive tooxidation at temperatures above 450°C. In order to protect the PyC interphaseagainst oxidation, multilayered composites and matrices have been developed(Lamouroux et al., 1995), and composites with multilayered interphases or matriceshave been investigated (Carrere, 1996, Forio, 2000). Such multilayered matricescontain phases which produce sealants at high temperatures causing healing of thecracks and preventing oxygen from reaching the interphase and fibers (Forio, 2000).

Damage is influenced by composite structure. SiC/SiC composites made usingchemical vapor infiltration (C VI) of a woven fiber preform display a heterogeneousstructure consisting of infiltrated tows, large pores (referred to as macropores), and auniform layer of matrix over the fiber preform (the intertow matrix). Figure 1 showsan example of composite structure. Damage under monotonic loading results frommatrix cracking, first in the intertow matrix, then in the transverse infiltrated towsand finally in the longitudinal tows (Guillaumat et al., 1993).

This paper investigates the damage and lifetime of a textile SiC/Si-B-C compositewith a self-healing multilayered matrix. Tangent modulus has been used for damagecharacterization during fatigue at high temperatures..

1. Material and experimental procedure

The SiC/Si-B-C composite was produced via Chemical Vapor Infiltration bySNECMA (France). It consists of a woven preform of tows of treated (proprietarytreatment, SNECMA) SiC fibers (Nicalon, Nippon Carbon Co., Japan), coated witha thin layer of pyrocarbon (interphase) and a multilayered matrix which containsphases of the Si-B-C ternary system (Fig 2). Fiber volume fraction was about 40%,and residual porosity was about 10-13%. Dog bone shaped test specimens with thefollowing dimensions were prepared : 200 mm x 16 mm x 4.5 mm.


Tensile tests were performed at room temperature for determination of referencemechanical behavior and associated damage. Cross-head displacement rate was 0.05mm/min. The polished surface of specimens was inspected during the tests, using anoptical microscope. Images were recorded using a digital camera under increasingstrains : 0.08%, 0.10%, 0.15%, 0.20%, 0.25% 0.30%, 0.40%. Then they were storedon disks using a PC.

Figure 1 : Microstructure of a textile SiC/SiC composite

Figure 2 : Microstructure of the multilayered matrix of a SiC/Si-B-C composite

Deformations were measured using an extensometer (gauge length = 25 mm).Unloading-reloading cycles were carried out, in order to estimate tangent modulus.Tangent modulus is derived from the slope of the stress-strain curve on reloading(Guillaumat et al, 1993, Forio et a/., 2001).

Cyclic and static fatigue tests were performed using an Instron testing machine, inair at 600°C and 1100°C under load-controlled conditions (Table 2). Cycling


frequency was 0.25 Hz and stress ratio R = amin/amax = 0.1, where omin and amax arerespectively the minimum and the maximum applied stresses. omax as well as thestress applied during pre-cracking were selected with respect to the induced damage,on the basis of the mechanical behavior at room temperature. Deformations weremeasured using an extensometer with A12O3 rods (25 mm gauge length). Unloading-reloading cycles were carried out, at a rate of 400 MPa/min (R = 0), in order toestimate tangent modulus.

After ultimate failure, test specimens were examined using scanning electronmicroscopy and optical microscopy.


2.1. Mechanical behavior at room temperature

Tensile stress-strain curves (a-e) show the typical features of non-brittle compositebehavior (figure 3), including a non-linear domain beyond the proportional limitreflecting damage tolerance.

Table 1 : Damage in SiC/Si-B-C composite at room temperature duringmonotonic loading

Young'sModulusE0(GPa)

191

Stresses (MPa)

0

70

150

220

365

Strains(%)

0

0.37

0.1

0.25

0.8

0.86

Relativetangent

modulus E/E0

1

1

0.75

0.45

0.25= 0.5VfE,/E0

«

Damage

Elasticdeformations

Cracking in theintertow matrix

Matrixcracking in thetransverse tows

Matrixcracking in the

longitudinaltows

Saturation

Ultimatefailure

Stress induced damage is reported in table 1. It is comparable to that observed inconventional 2D SiC/SiC (Guillaumat et al., 1993). It can be also noticed from table 1,


that the tangent modulus decreases steeply during the first two stages of damage, andthen much gently during cracking in longitudinal tows. The terminal value of relativetangent modulus is equivalent to the minimum value E/Eo = 0.5 Vf Ef/E0, indicatingthat the load is carried solely by the fibres (Forio et a/., 2000). Individual fiber breaksoccur under high stresses near the ultimate failure (Forio et al., 2000).

It is worth pointing out that a significant amount of damage involving the intertowand in the intratow matrix was generated during pre-cracking (strain = 0.25%). Thecorresponding value of initial tangent modulus is E/Eo « 0.45.

04 0.6Deformatton(%)

Figure 3 : Example of tensile stress-strain behavior for a SiC/Si-B-C compositeunder monotonic loading at room temperature

2.2. Lifetime and damage during fatigue

The lifetime data (Table 2), show that there is no significant influence of fatigueconditions (static or cyclic fatigue) nor of precracking. The lifetime drops when theapplied load is 220 MPa. Much longer lifetimes were obtained at 1100°C.

It is worth pointing out that the lifetime of a conventional SiC/SiC composite ismuch shorter under comparable conditions (< 1 hour under a smaller constant stress(100 MPa), at 700°C (Carrere, 1996)).

The magnitude of tangent modulus during the fatigue tests is determined by theamount of initial damage and it depends on temperature (Fig. 4 and Fig. 5). Theinitial damage was induced either by the applied load during the first cycle or by thepre-cracking load. It is indicated by the initial value of relative modulus E(t0)/E0).


Table 2 : Testing conditions and lifetimes

Specimens Temperature omax Precracking Frequency Lifetime Lifetime(°C) (MPa) (Hz) (hour) (Cycles)

12345678

6006006006001100110011001100

150150150220150150150220

NoNoYesNoNoNoYesNo

00.250.250.2500.250.250.25

12h03minI5h20min13h23min4h02min4h55min52h32min49hl4min2hlOmin

***1346111680***

330246239437881675*

failure by thermal shock

At 600°C, when E(t0)/E0) > 0.47, tangent modulus decreases slowly (figure 4, testspecimen 2). The initial damage consists of two families of cracks located in theintertow matrix and in the transverse tows (table 1). When E(t0)/E0) = 0.47, tangentmodulus remains constant during 2000 cycles and then decreases gently (testspecimen 3, figure 4). When E(t0)/E0) < 0.47, the modulus decrease is moresignificant than previously (amax = 220 MPa, test specimen 4, figure 4). The initialdamage is more severe and it involves cracks in the longitudinal tows (those parallelto the loading direction) (table 1).

At 1100°C (Fig. 5), similar trends are observed, but the modulus decreases are lesssignificant: tangent modulus decreases when E(t0)/E0) > 0.47 (test specimens 6 and8), and remains constant when E(t0)/E0) = 0.47 (test specimen 7).

The modulus decreases reflect an environment-activated damage, which may beattributed to extension of debond cracks as a result of oxidation of pyrocarboninterphases. They are observed at 600°C essentially, but also at 1100°C when initialdamage involves cracks in longitudinal tows.

2.3. Failure and damage observations

The fracture surfaces of those specimens that were tested at 600°C were generallyflat, with limited fiber pull-out (Fig. 6). Examination of polished longitudinalsections revealed the presence of cracks in the intertow matrix, in the transverse towmatrix and also in the longitudinal tow matrix. For specimens 1 and 2, E(t0)/E0) >0.47 : therefore the matrix cracks in the longitudinal tows were not created duringloading (table 1). The applied stress (amax = 150 MPa) was insufficient to generatethem according to data reported in table 1. They probably appeared during thefatigue tests, as a result of oxidation. For specimens 3 and 4, E(to)/E0) < 0.47 : it


seems logical to attribute the presence of such cracks to the load applied duringfatigue (220 MPa , specimen 4) or during pre-cracking (specimen 3), according totable 1. Limited healing features were identified on those specimens tested at 600°C.

Figure 4 : Evolution of relative elastic modulusduring cyclic fatigue at 600°C in air

Figure 5 : Evolution of relative elastic modulusduring cyclic fatigue at 1100°C in air

During the tests at 1100°C, failure occurred in those regions of specimenssubjected to lower temperatures (500°C - 600°C), as a result of the temperaturegradient associated to the cold grip testing method. Polished longitudinal sections ofthose regions at the temperature of 1100°C, were inspected using opticalmicroscopy. No crack was detected in the longitudinal tows of specimens 5 and 6 :E(to)/E0) > 0.47. In specimens 7 and 8, E(to)/E0) < 0.47, matrix cracks were found inlongitudinal tows. Figure 6 shows evidence of crack healing. The cracks appear tobe filled by a glass which may consist of fused silica.


Figure 6 : Micrograph showing a fracture surface after fatigue at 600°C

Figure 7 : Glass healing of cracks near a macropore at 1100°C

2.4. Discussion

The longest lifetimes were obtained on those specimens tested at 1100°C. This canbe attributed to the contribution of crack healing, which was observed essentially atthis temperature. Extension of the cracks initiated in the intertow matrix and in thetransverse tow was detected only after the tests at 600°C. It can be related tooxidation of pyrocarbon interphases when crack healing is not effective. Thesecracks then reach the periphery of longitudinal tows. The degradation of pyrocarboninterphases at the periphery of longitudinal tows influences load sharing, leading tooverloading of the longitudinal tows and further matrix cracking. Under larger loadsthe initial cracks reach the longitudinal tows (pre-cracking load or 220 MPa).


Damage may be attributed to debonding induced by degradation of interphaseswithin the longitudinal tows and to associated matrix cracking.

The fatigue behavior is well illustrated by the plots of relative tangent modulusversus strain (referred to as E(e)/E0 in the following) shown in figures 8 and 9.Pertinent strains are those at the beginning of the unloading cycles dedicated totangent modulus measurement. It is interesting to compare the E(e)/E0 curvesobtained in fatigue with the reference curves determined during monotonic tensiletests performed in argon respectively at 600°C and at 1100°C, that reflect compositeresponse to stress induced damage (Forio, 2000).

Figure 8 : Relative modulus versus deformation for specimens tested at 600°C.

Figure 9 : Relative modulus versus deformation for specimens tested at 1100°C


Figure 8 shows that the E(e)/E0 curves at 600°C are well predicted by equation (1)which describes tangent modulus dependence on elastic damage (Forio et al., 2001)

The E(e)/E0 curves are always located below the reference curve.

Those curves determined under 150 MPa are identical. They are independent ofthe fatigue loading conditions (static or cyclic) and initial damage. Terminalvalues of relative tangent modulus are generally E(t0)/E0) < 0.47. Under 150 MPa,a strain of 0.25% was reached. Under 220 MPa, strains are larger than 0.3%. Allthese data are consistent with the presence of matrix cracks within the longitudinaltows (table 1).

At 1100°C, the E(e)/E0 curves obtained under 150 MPa do not coincide with thehyperbola predicted by Eq. 1 (figure 9). They are located above or below thereference curve, depending on the values of initial tangent modulus E(t0) and strain.If the strain increase resulted from damage only (elastic deformation), the E(e)/E0

curves would be located below the master curve and would be predicted by equation1. When crack healing operates alone, E(t) and e (t) remain constant and the E(e)/E0

curves would amount to a single data point (E (t) = E (t0), e (t) = e (t0)) whoselocation is dictated by E (t0). The E(£/E0 curves indicate a slight modulus decrease(when E(t0)/E0 > 0.47) or a constant modulus (when E(to)/ E 0 < 0.47) but asignificant strain increase in both cases (e (t) > e (t0)). Therefore, the E(£)/E0 curvescan be logically attributed to a combination of crack healing (limiting oxidation-induced damage) and creep of fibers (Bodet et al., 1996), responsible for thesignificant deformation increases in the absence of damage when E(to)/ E 0 < 0.47(the load is carried only by the longitudinal bundles) or with a slight damage whenE(t0)/E0>0.47.

The E(e)/E0 curve obtained under 220 MPa coincides with the hyperbolaspredicted by Eq. 1 (Figure 9) and it is identical to that determined at 600°C. Thisindicates that the composite experienced damage during fatigue as previously at600°C and that crack healing was not effective, because the crack openingdisplacement was too large.

3. Conclusion

A SiC/Si-B-C composite with a self-healing multilayered matrix was investigatedunder static and cyclic fatigue loading conditions at 600°C and 1100°C in air.


Fatigue behavior depends on temperature, loading history (i.e. initial damage createdduring pre-cracking or loading), and magnitude of applied load. Lifetime depends ontemperature and applied load. A significant influence of crack healing on damageand lifetime was determined from trends in tangent modulus and from scanningelectron microscopy. Crack healing, as it reduces or stops the amount of oxygen thatmigrates towards the pyrocarbon interphases, limits fatigue damage and leads tosubstantial lifetime improvements. Crack healing by production of a glass fromoxidation of the multilayered matrix was particularly effective at 1100°C. At1100°C, crack healing, limited damage and creep were evidenced. At 600°C,contribution of crack healing was limited. Damage during fatigue involvedextension of initial cracks and debonding within the longitudinal tows, as a result ofoxidation of interphases.

4. Acknowledgements

This work was supported by Snecma and CNRS through a grant given to P.F. Theauthors wish to thank E. Pestourie for valuable discussions, SNECMA for theproduction of samples, B. Humez for assistance with mechanical tests, J. Forget andC. Dupouy for manuscript preparation.

5. References

Evans A.G. and Marshall D.B., «The mechanical behavior of ceramic matrix composites »,Ada Metit vol. 37, n° 10,1989, p. 2567-2583.

Kerans R.J., Hay R.S., Pagano N.J., Parthasarathy T.A. « The role of the fiber-matrixinterface in ceramic matrix composites », Am. Ceram. Soc. Bull., vol. 68, n° 2, 1988, p. 429-442.

Naslain R. "Fiber-matrix interphases and interfaces in ceramic matrix composites processedby CVI", Composite Interfaces, 1993, p. 253.

Droillard C. and Lamon J., "Fracture toughness of 2-D woven SiC/SiC CVI-composites withmultilayered interphases", J. Am. Ceram. Soc., vol. 79, n° 4, 1996, p. 849-858.

Lamouroux F., Pailler R., Naslain R., Cataldi M, French Patent n°95 14843, 1995.

Carrere P.,« Thermostructural behavior of a SiC/SiC composite », Ph.D Thesis, n° 1592,1996, University of Bordeaux 1.


Forio P., « Thermostructural behavior and lifetime of a 2D woven SiC/Si-B-C composite witha self-healing matrix », Ph.D Thesis, n° 2171, 2000, University of Bordeaux 1.

Guillaumat L., Lamon J., « Multicracking of SiC/SiC composites », Revue des Composites etdes Materiaux Avances, vol. 3, n° hors serie, 1993, p. 159-171 (in French).

Forio P., Lamon J., « Fatigue behavior at high temperatures in air of a 2D SiC/Si-B-Ccomposite with a self-healing multilayered matrix », Advances in Ceramic MatrixComposites VII, Ceramic Transactions, vol. 128, p. 129-140, 2001.

Bodet R., Lamon J., Jia N., Tressler R.E., « Microstructural stability and creep behavior of Si-C-O (Nicalon) fibers in carbon monoxide and argon environments », J. Am. Ceram. Soc.,vol. 79, n°10,1996, p.2673-2686.

Part II:

Development and use ofsmart techniques for strainmeasurement or damagemonitoring

Piezoelectric Fiber Composites for vibrationcontrol applications

Development — modelling - characterization

Y. Vigier* - C. Richard** - A. Agbossou* - D. Guyomar**

* Laboratoire Materiaux Composites (LaMaCo)ESIGEC-Universite de Savoie -73376 Le Bourget du Lac Cedex - France

yves. [email protected], [email protected]

Laboratoire de Genie Electrique et Ferroelectricite (LGEF)INS A de Lyon. Bat. G. Ferrie - 69621 Villeurbanne Cedex - France

crichard@ge-serveur. [email protected]

ABSTRACT: The fabrication of a planar piezoelectric composite transducer made withcommercial PZT fibers is presented. A method for the PZT volume fraction control isdescribed and a set of resonators are made and characterized to derive the fiber properties.Two fabrication methods are proposed for the integration of a piezocomposite actuator to anepoxy cantilever beam. Coupling coefficient of these actuators are measured and compared toa bulk PZT type one. Vibration damping capabilities are derived showing the advantage ofusing piezocomposite. The optimisation of the proposed structure given showing thepossibility of using short fibers.

KEY WORD : piezoelectricity, PZT fiber, composites, vibration damping, modelling


Figure 1: The variation of the PZT fiber volume fraction with the coating layerthickness control, r is the fiber radius andx the epoxy layer thickness.

1. Introduction

Following the development of composite materials for structural applications,the idea of dispersing piezoelectric ceramic elements in a lighter matrix has beenexperimented in the late 70's for improving material robustness and decreasingdensity (Newnham 1980). One of the major realizations is known as the 1-3connectivity piezocomposite, consisting in piezoelectric rods aligned in parallel andembedded in a polymer matrix. This material exhibits interesting properties forhydrostatic sensors or ultrasonic transducers. Most recent efforts led to thedevelopment of commercial PZT fibres with diameter ranging from l0um to 250um (Yoshikawa 1995). Sheets or plies of 1-3 piezocomposites used for vibrationcontrol or cantilever actuation are made with one or more layers of long fibres withinterdigitated electrodes allowing the poling and activation of the transducers with areasonable voltage (Bent 1997). It is the purpose of the present paper to describe anoriginal method for the fabrication of composites made with commercial PZT fibers,and allowing control of the PZT volume fraction. PZT fiber properties are derivedfrom the evaluation of composite properties on a large volume fraction distribution.Coupling coefficient and damping performances obtained on cantilever beamsactuated with these piezocomposites are described showing the advantage of suchadaptable material. Finally, the description of a novel piezocomposite fabricationprocess points out the possibility of using short fibers without much degrading thecomposite performances.

2. PZT fiber properties evaluation

In order to derive the PZT fiber properties and to demonstrate the possibility forPZT volume fraction control, a set of samples with different volume fractions wasmade and evaluated using a resonance method (IEEE Standart ANSI STd 176-1987). From the various master curves giving the composite properties versus the

Development and use of smart techniques 175

PZT volume fraction and using a simple homogenisation model, the fiber propertieswere derived.

2.1. Sample fabrication

The PZT fibers used were PZT5A manufactured by Ceranova Corp. The averagefiber diameter is 138um. The epoxy used was a standard composite processingepoxy LY5052 + HY5052 hardener from Ciba Specialty Chemicals. In order tocontrol the volume fraction, it was assumed that the PZT fibers could be closely andnaturally packed in a mould and that the volume fraction could be controlled by theintroduction of an epoxy layer previously deposited on each fiber as shown onFigure 1. The volume fraction is therefore a function of the epoxy coating thicknessx. In order to be able to reach easily various coating thickness, a spacer consisting invoided glass microspheres was added during the coating process. The followingroute was used: with a usual dip-coating technique, an epoxy resin layer was firstdeposited and gelified at 60°C. Then glass powder was projected on the fibers. Afterfinal curing, a last epoxy layer was finally added. The global coatings weresufficiently regular and the thickness compatible with the desired volume fractions.In this process the final epoxy coating thickness is a function of the gelation timewhich allows the control of the first coating adhesive force (Lee 2000).

The coated fibers were cut to proper size in length (18mm), naturally packed in amould (4x4x20mm) and epoxy was finally poured under vacuum to fill-in theremaining voids.

2.2. Sample characterization

For measurements, various samples were cut: length expander bar (4x4x10mm)for 3.3 mode characterization, thickness expander plate (4x4x1mm) for thicknessmode and lateral expander bar (4x2x1mm) for 1.3 mode characterization. A classicconductive ink (Du-Pont E5007) was used for electroding both composite ends.Poling of the whole batch was conducted in a 80°C oil bath with a 2KV/mm electricfield for 1 minute. This conditions were found to be a good trade-off betweenremnant polarization, coercive field and a large parasitic electric leakage current andelectric breakdowns for temperature above 100°C (Lee 2000).


Figure 2: dielectric permittivity and charge coefficient versus the PZT volumefraction for the PZT 5A/Epoxy piezocomposites

Property I PZT 5A Fiber I Bulk PZT 5Ae33

T /£o 1064 1700e33

s /£0 742 830R3| -0.19 -0.34k33 0.55 0.70kt 0.46 0.48d33 220 pC/N 374 pC/Nd3, -75pC/N -171 pC/Ns,i^ I.55E-I1 Pa'1 I .64E-1I Pa'1

s33^ ~ 1.70E-llPa'' ~ 1.88E-l lPa 's,," 1.49E-I1 Pa'1 1.44 E-11 Pa'1

s33p " 1.19E-llPa' 0.946 E-ll Pa'1

p I 7000-7300 SI I 7750 SI

Table I: fitted PZT 5A fiber properties compared to bulk PZT material

Figure 2 shows e33T and d33 as a function of the PZT volume fraction. The results

show classical behaviours for composites and it is interesting to remark that a 30 %PZT volume fraction composite gives a k33 coupling coefficient close to 50% with ad33 close to 200 pC/N. Finally The main discrepancy was that the compositeproperties extrapolation to 100% PZT was generally lower than the bulk PZT 5Adata. These values were then modified (while keeping coherence of the data set) andthe homogenisation modelling was iterated in order to get a good fit between thetheoretical and experimental values. The fitting was done considering k33, kt, d33, d31,£33T> s33

E and SHE as functions of the PZT volume fraction

The final set of data for the PZT 5A fiber given in Table 1 allows a good globalfit. It is interesting to remark that most of the fiber properties like permittivity,charge coefficient, coupling coefficient are slightly lower (20%) than that of bulk


PZT. The results also depends on the PZT fiber batch used. This extensive study wasdone on the first batch supplied, more recent but only partial results showed highercoupling coefficients, but still 10% lower than bulk (for k33).

Figure 3: insertion of planar PZT fiber composite segments in the epoxy bea.

3. Integration of a composite transducer in a cantilever beam

In order to make the evaluation of the composite performances for a vibrationdamping application, the piezocomposite previously described was inserted in acantilever structure, clamped on its extremity and vibrating on its first bendingmode. 16 composite inserts (with an average 15% PZT volume fraction), 30mmwide, 10mm long (along polar axis) and 1.5mm thick, were inserted following ananti-parallel arrangement on the upper and lower faces of an epoxy beam ( 400mmlong, 30mm wide and 5mm thick) as on Figure 3. From impedance measurements,the coupling coefficient of the first bending mode was derived using the usualrelation:

where cos and 0}, are respectively series and parallel (or short-circuited and open-circuited) resonance angular frequencies. The beam was further driven with anelectromagnet at constant force around these resonance frequencies. The tipdisplacements were monitored as a function of the frequency under open circuit andshort circuit conditions, and then when the inserts are shunted with an adaptedresistor R (Hagood 1991) given by :

Figure 4a shows the various plots obtained in these 3 configurations, it allows toquantify the transducer performance for a piezoelectric passive resistive damping


device. A 28% vibration amplitude reduction is obtained with an overall 21.7%effective coupling coefficient.

Figure 4: vibration damping performances of the piezocomposite actuated (a) andbulk PZT actuated (b) epoxy beam. Open circuit, short circuit and matched resistiveshunt resonance behaviours are compared.

Figure 5: piezocomposite coupling and Young Modulus variations compared to thebeam bending coupling coefficient. The PZT is PZT5A (nominal data).

As a matter of comparison, the same measurement obtained with a similar beam(Richard and al. 2000) equipped with bulk PI94 (Saint-Gobain Quartz) PZT platesworking in the 3.1 mode exhibited a 11% effective coupling coefficient with a 5%


vibration reduction amplitude (Figure 4b). Moreover in the later case, the active PZTmaterial quantity was twice the total PZT content of the composite actuated beam.

Two arguments have to be pointed out: first, in the composite, the ceramic isworking in the 3.3 mode, leading to a better coupling coefficient, and secondly theactuator elasticity is better adapted to the elasticity of the structure resulting inoptimised coupling. This point is illustrated on Figure 5 comparing for various PZTvolume fraction the piezocomposite k33 coupling coefficient and elastic modulus tothe beam effective coupling coefficient. Properties of the composite are obtainedwith an homogenisation approach (Vigier 2001) and the global beam modelling ismade with the ANSYS FEM code. For low PZT volume fractions, there is not astrong mismatch between the beam and actuator elastic constants, the coupling ofthe composite material is quite optimal resulting in optimum beam response near a10% volume fraction. This optimum depends on the considered vibratingmechanical structure stiffness which could benefit of a tailored actuating material.

4. Composite beam with short fibers: UD-segmented technique

In order to simplify the fabrication procedure and to increase the globalcapacitance of the final transducer, a second fabrication procedure wasexperimented leading to much less manipulation stages and using interdigitatedelectrodes. Figure 6 illustrates the method which gives the UD-segmentedpiezocomposite (UD for Uni-Directional) opposed to the previous UD-insertedtechnique.

First (Fig 6a) an epoxy beam structure is moulded (150mm long, 15mm wide and5mm thick) and a cavity (45mm long, 10mm wide and 1.2mm deep) is hollowed oneach face. Pre-coated PZT5A fibers (40 mm long) are then inserted in each cavity(Fig 6b). The fiber coating was set to get a 15% volume fraction as in the previouscase. Three layers are necessary to fill the cavities, they are afterward embeddedwith polymer degassed and reticulated. Then grooves (0.3mm thick and 1.2mmdeep) perpendicular to the fibers are made using a diamond cut-off blade used forthe "dice-and-fill" processing technique (Fig 6c). Both fiber inserts are thensegmented into smaller sections (2mm long for the proposed experiment). Thegrooves internal surface are electroded with a silver ink (Fig. 6d) defining compositeelements 2mm long. These elements are wired in parallel using a silver ink wiringpattern with driving lines running along the beam edges (Fig 6e). After the groovesre-impregnation, the piezocomposite elements are then poled in-situ (same polingconditions than previously described) and are then arranged with anti-parallel polingdirections.


Figure 6: fabrication steps of the UD segmented piezocomposite beam.

Figure 7: Fiber aspect ratio optimisation of the UD segmented piezocomposite(a) and damping performance of the experimental beam (b).


After poling the wiring is modified to get opposed strains on the upper and lowerfaces of the transducer thus allowing piezoelectric coupling to flexure motion. Themain advantage of this procedure, except less fiber manipulation, is the much highertotal electric capacitance of the actuator. For comparison, the UD-segmentedtransducer capacitance is 440pF while the larger UD-inserted one was only 120pF.Increasing the transducer capacitance means decreasing the resistance of thematched damping resistor or allowing easier transistor switching capabilities whenimplementing the Synchronised Switch Damping (SSD) technique (Richard 2000).The beam was then characterized in terms of coupling coefficient and vibrationdamping performance whith the transducer shunted with a 3.8 MQ matched resistoraccording to equation [2]. Figure 7a shows a 28% vibration amplitude reduction andan overall 18% effective coupling coefficient.

An important point is that in this last case the composite is composed of shortfibers (2mm instead of 10mm long) and very few degradation of the couplingcoefficient is observed. It is therefore interesting to derive what is the limit lengthfor which there is a notable decrease of the transducer performance. Modelling ofthe stress transfer was conducted using the ANSYS FEM code with periodicityconditions. Using different loading conditions, homogenised composite sectionproperties were obtained taking into account the fiber length or more precisely thefiber aspect ratio (length to diameter ratio). Then the global flexure mode effectivecoupling coefficient was derived as a function of the fiber aspect ratio (Vigier 2001).The result is shown on Figure 7b. This points out that there is a slight decrease of thecoupling coefficient down to an aspect ratio of 5 (a 0.75mm length for a fiberdiameter equal 150um), then a sharp decrease between 5 and zero. This means thatpiezocomposite with short fibers are still effective and that this fabricationprocedure could be extended down to a 1mm groove spacing, allowing a fourfoldincrease of the capacitance without affecting much the coupling coefficient

5. Conclusion

Piezoelectric Fiber composites were made from commercial PZT 5A fibers withan original technique allowing PZT volume fraction control over a range extendingapproximately from 10% to 70%. Characterization of a batch of resonators gave thecomposite properties variations versus the volume fraction. It allowed the PZT fiberproperties derivation through the use of a simple homogenisation model. Fibercoupling coefficients were found to be 10% to 20% lower than bulk PZT. Thesecomposite were used to perform vibration damping experiments. An epoxy beamwas equipped. Results obtained were better than with a bulk PZT actuator plate. Thepiezocomposite is working in 33 mode and its elasticity is much well matched to avibrating polymer structure. An alternative method for the composite fabricationwas experimented leading to short piezocomposite segments. It pointed out thepossibility of operation with short fibers. The critical fiber aspect ratio resulting innotable composite properties degradation was found to be close to 5, thus allowing


the use of short 1mm segments. Use of short segments results in higher actuatorcapacitance and lower voltage facilitating vibration damping conditions especiallywith the Synchronised Switch Damping technique (SSD).

Acknowledgements

The authors wish to thank the Region Rhone-Alpes for partial support.

References

Bent A. A. - Active fiber composites for structural actuation- , Ph.D Thesis, TheMassachusetts Institute of Technology, January 1997.

Hagood N.W., Von Flotow A., "Damping of structural vibrations with piezoelectric materialsand passive networks", Journal of Sound and Vibrations - Vol. 146, no.2, 1991

Lee H.S., Richard C. and al.. "Fabrication and Evaluation of 1.3 PZT Fiber / EpoxyComposites", Proceedings of the 2000 IEEE-ISAF Symposium, Honolulu, August 2000.

Newnham R.E., Bowen L.J., Klicker K.A. and Cross L.E."Composite PiezoelectricTransducers" Materials in Engineering, Vol. 2, pp. 93-106, 1980.

Richard C., Guyomar D. and al. "Enhanced semi-passive damping using continuous switchingof a piezoelectric device", Proceedings of the 7th SPIE-ICSSM Symposium, March 2000

Vigier Y. "Materiaux Composites a fibres piezoelectriques pour applications en controle devibration ", Doctoral Thesis, Universite de Savoie, no. 2001CHAMS019, October 2001.

Yoshikawa S.Y., Selvaraj U., Moses P. and al.. " Pb(ZrTi)O3 [PZT] fibers : Fabrication andMeasurement methods " Journal of Intelligent Materials and Structures, Vol. 6, 1995

Health Monitoring System For CFRP ByPZT

Ja-Ho Koo — Toshiaki Natsuki — Hiroshi TsudaJunji Takatsubo

Smart Structure Research CenterNational Institute of Advanced Industrial Science and TechnologyTsukuba AIST Central 2, Tsukuba, 305-8568, Japan

[email protected]

ABSTRACT: In this work, we manufactured the piezoelectric ceramics transducers embeddedCFRP and its more exact source location method on microcracking was investigated.Especially, we studied the way to determine the arrival time when high level noise is includedand to use wavelet transformation when the amplitude of symmetric mode is so small that

searching the arrival time is difficult. The transducers were able to detect the signals withoutany amplifier well. Control of oscilloscope by personal computer made real-time healthmonitoring possible. When a signal included a large noise in front of the real response,backward searching method (BSM) was useful to eliminate it. Wavelet transformation (WT)method was useful to determine the arrival time of the symmetric mode Lamb wave as well asthat of anti-symmetric mode. On the other hand, we introduced a theory of plate to calculate

the more exact wave velocity in any case of laminates including non-symmetric laminates.

KEY WORDS: smart structure, acoustic emission, health monitoring, piezoelectric ceramics,source location, lamb wave


1. Introduntion

A large portion of the recent studies on smart materials and structures areconcentrated on CFRP (Tang et al., 1998, Prosser et al., 1999, Seale et al., 2000,Seydel et al., 2001). CFRP has so high specific strength and rigidity that it is used atimportant parts in aeronautic and astronautic field. Therefore, if it fails the loss isalso so large. In order to prevent such failure, real-time health monitoring onmicrocracking like matrix cracking, debonding, delamination, transverse crackingand fiber breakage is required.

If the microcracking takes place, the released energy propagates as elastic wave.The elastic wave consists of symmetric mode Lamb wave and anti-symmetric modeLamb wave. In the case of the study to identify the source of external shock similarto vertical shock, to deal with anti-symmetric mode is useful because out-of-planecomponent is dominant. Contrarily in the case of identifying a microcrack,symmetric mode is available as it has in-plane component.

So, in this work, we manufactured the piezoelectric ceramics transducersembedded CFRP and its more exact source location method on microcracking wasinvestigated. Especially, we studied the way to determine the arrival time when highlevel noise is included and to use wavelet transformation when the amplitude ofsymmetric mode is so small that searching the arrival time is difficult.

2. Theories

2.1. General solution of wave velocity in angle ply laminates

In the case of arbitrarily laminated plates, for example [10/20/30/.../180], couplingstiffness and coupling normal-rotary inertia coefficient should be considered toevaluate the non-symmetric components.

We used the same assumption for the displacement field as that of Yang, Norrisand Stavsky (Yang et al., 1966). That is as follows.

where u, v, w are the displacement components in the x, y and z directions, u0 andv0 are the midplane displacement components, and wx and wy are the rotationcomponents along the x and y axes, respectively. The stress-strain relations for anylayer are given by


where Qij for I, j = 1, 2, 6 are plane-stress reduced stiffnesses, and Qy for I, j - 4, 5are transverse shear stiffnesses. Defining the force and moment resultants per unitlength as

where h is the thickness of the plate, we have

where the laminates stiffnesses are given by

The shear correction factors KJ and KJ are included to account for the fact that thetransverse shear strain distributions are not uniform across the thickness of theplate. Neglecting body forces, the equations of motion are


where and p is the mass density.

For wave propagation, we consider plane waves of the type

where k is the wave number, m and n are the direction cosines of the wave vector inthe x and y directions, respectively, w is the circular frequency, and

and are the amplitudes of the plane harmonic waves. Substituting Eq. [4] andEq.[7] into Eq.[6], we can obtain Eq.[8].

where MJJ are as follows


Then, the characteristic equation for symmetric and anti-symmetric wave mode isexpressed as follows.

The phase velocity (to/k) and the group velocity (deo/dk) can be obtained fromabove equation.


2.2. Arrival time determination

2.2.1. Backward searching method

Because the detected AE signals include some noise in front of main signal, theconventional threshold method sometimes makes mistake on determination ofarrival time. Therefore we use Backward Searching Method (BSM) which is notinfluenced by the front noise. In BSM, as shown in Fig. 1, when some points groupthat proceeds backward from maximum peak enters into a limited range, the lastpoint of the group becomes the first arrived point.

Figure 1. Backward searching method

2.2.2. Wavelet transformation method

The definition of the continuous wavelet transformation (WT) of a function f(t) isas follows (Kishimoto et a/., 1995):

where a > 0 and the overbar means the complex conjugate. From WT we can get theinformation of behavior of a particular frequency component in time domain. Thecalculation can be carried out at high speed by FFT. The mother wavelet used in thiswork is Gabor function (Eq. [12]). Its Fourier transform is expressed as Eq. [13].Here, (wo is the center frequency and y is positive constant.

By WT method, we can determine the difference of arrival times for arbitrary frequency


component because the peak on the ((0,t) plane means the arrival time of the group velocity atthe frequency. It can be applied to both of symmetric mode and anti-symmetric mode.

2.3. Source location

Suppose that Tj and t; are the true and measured arrival times of /-th transducerrespectively. The true arrival time is expressed as follows:

where (x, y) is a source position, (Xj, y^ is a transducer position, is a radius oftransducer, and Vj is the velocity in the direction. If fj is defined as Eq. [15], we canfind the (x,y) that satisfies Eq. [16] by nonlinear least-square method.

Here E is convergence limit.

3. Experimental setup

Figure 2. Photograph of the polyimide sheet with piezoelectric ceramics andcircuit


We manufactured six kinds of CFRP specimen ([0/90]2s, [0/±45/90]s, [0/30/-60/90]s, [04/904], [02/452/-452/902], [02,302/-602/902]). The properties of a layer areas follows : Ex = 119.35 GPa, Ey = 9.16 GPa, vx = 0.355, p = 1.51g/cm3. Polyimidesheet with four embedded piezoelectric ceramics (Fig. 2) was inserted to eachcenter layer. The dimension of the specimen is 145x200x1.8mm. The thickness anddiameter of the piezoelectric ceramics is 200fim and 5mm.

For source location test of out-of-plane AE source, pencil lead break test wascarried out at the position of x=40, y=90mm. The pencil lead is 0.5mm 2H type.Any amplifier was not used. The block diagram is shown in Fig. 3.

Figure 3. Block diagram of experimental setup


Fig. 4 shows the case of [0/90]2s. There are four signals detected at each channelin Fig. 4(a). The signals were modified with 0-point correction and noise filteringFig. 4(b). In order to avoid the influence of the residual large noise, arrival timeswere searched with BSM Fig. 4(c). In this case, the source location error is verysmall, about 1 mm.

On the other hand, we also tried to test the WT method on the same signals. Asshown in Fig. 5(a), the center frequency of the symmetric mode of the detectedsignal (eg., 1 ch.) is 474 kHz. Fig. 5(b) shows the WT coefficients of 474 kHzcomponent in time domain. We let the first peak arrival time. The source locationerror is also very small, about 1 mm.

In the case of the other five specimens, source location analysis gave good results.


Figure 4. (a) detected signal, (b) filtered signals and (c) arrival times by BSM in[0/90]2s

Figure 5. (a) WT of channel I and (b) WTat 474 kHz

5. Conclusions

Some piezoelectric ceramics were embedded into CFRP thin plate for sensing thesimulated AE signal in this work. The transducers were able to detect the signalswithout any amplifier well. Control of oscilloscope by personal computer madereal-time health monitoring possible. When a signal included a large noise in frontof the real response, backward searching method was useful to eliminate it. Wavelettransformation method was useful to determine the arrival time of the symmetricmode Lamb wave as well as that of anti-symmetric mode. On the other hand, wecan calculate the more exact wave velocity with the 5x5 matrix of M in any case oflaminates including non-symmetric laminates.

References

Tang B., Henneke II E. G, Stiffler R. C., Acousto-Ultrasonics: Theory and Application, NewYork, Plenum, 1988.


Prosser W., Gorman M., Humes D., Journal of Acoustic Emission, vol. 7, 1999, p.29.

Prosser W., Scale M., Smith B., J. Acout. Soc. Am., vol. 105, 1999, p.2669.

Scale M., Madaras E., J. Compo. Mater., vol. 34, 2000, p.27.

Seydel R., Chang F. K., Smart Mater. Struct., vol.10, 2001, p.354.

Yang P. C, Norris C., Stavsky Y, Ml J. of Solids and Struct., vol. 2, 1966, p.665.

Kishimoto K., Inoue, H., Hamada M., Shibuya T., J. Appl. Mech., vol. 62, 1995, p.841.

Characterisation of Fibres and Compositesby Raman Microspectrometry

Ph. Colomban1

LADIR-UMR7075 CNRS & UPMC,

2 rue Henry Dunant 94320 Thiais, France

[email protected]

ABSTRACT: Raman spectrometry is a unique technique providing information on thestructure, short-range order and stress of solid through the intensity, polarization,wavenumber and bandwidth of the Raman peaks. The paper provides a comprehensive studyon Raman spectroscopy versatility as a fast and non-destructive tool for the understandingand imaging of phase organisation as well for the prediction of the mechanical properties(tensile strength, the Young's modulus (E)) of fibres. Selection of the laser exciting•wavelength gives micron lateral resolution and reduces the in-depth penetration to ~<100nm,allowing the analysis of fibre surface, coatings and interphases. Stress-induced Raman shiftscan be used to determine the stress/strain in any phase a few micron in scale. Quantitativeresults follow from wavenumber calibrations.

KEYWORDS: Raman, Imaging, Stress, Strain, Fibres, Interphases

also Consultant at ONERA 92322 Chatillon France


1. Introduction

As heterogeneous materials, composites need to be analysed at different levels: thefibre (including coatings) nanostructure, the matrix microstructure, the fibre-matrixinter-phase (micron scale or less), bundles and lamina/fabrics (tens of microns tomillimetre scale) and finally the structure level. Modelling fibre stress mathematicallywould be difficult, especially in the case of coated fibres. Indeed, inter-phase materialspromote stress relaxation, due to higher compliance, micro-cracking or thermalexpansion mismatch. The deficiency of micro-mechanical models to predict thecomposite strength was attributed to the random nature of the failure and the need touse statistical methods, the variety of failure modes, and the very local nature of failureinitiation. Another deficiency was the lack of an experimental technique capable ofmeasuring stress distribution from the fibre scale (a few micrometers) to the laminascale (a few millimetres) and finally to the part level. The recent development ofmicro-Raman spectroscopy as a micro-mechanical experimental technique hasprofound consequences on the understanding of solid mechanics in general andheterogeneous material micro-mechanics in particular. Micro-Raman spectroscopy isthe only technique capable of measuring local stress in a wide range of materials witha spatial resolution of ca. Ium. In recent years, a considerable number of instrumentaldevelopments were made. Microscopes allow wide solid angle collection of thescattered light, with improved geometrical resolution (confocal setting). Yet, series-imaging (also called mapping) of an area can be achieved by a step-by-step scanningof the sample, with a finely focused laser beam. Raman spectra fitting procedures thenallow the reconstruction of various maps.

2. The Fibre Nanostructure Level

The Raman effect results from the modulation of the (laser) light by opticalvibrations of the atoms/molecules. If the energy of the laser approaches those of thevarious electronic states of the material (in other words if the material is coloured),then near-resonant/resonant Raman scattering occurs and the penetration depth canbe reduced to a few tens of nanometers. This makes Raman microscopy a method ofsurface analysis (Colomban 00). Raman spectroscopy is sensitive to the chemicalbonds as well as to their relative organisation and allows analysis whatever the stateof polymorphism or crystallinity of the compound. The use of various exciting laserlines allows a specific, topological or chemical analysis.

2.1. Correlation between Raman Spectra and Particle Size

In "large" crystals, the phonons propagate "to infinity" and the first order Ramanspectrum only consists of "q=0" Brillouin zone centre phonon modes (momentumselection rule). However, since impurities or lattice disorder, including the surface


where atoms environment is singular, destroy crystalline perfection, the pnononfunction of many crystals is spatially confined. This results in band broadening andwavenumber shifts and can enhance the intensity of symmetry forbidden modes.This was first observed for semi-conductors but also exists in most materials issuedfrom liquid or polymeric routes (Suzuki et al, 01). This phenomenon becomesdominant in nano-sized grains because the number of atoms at or near the surfacebecomes equal to those in the bulk. Thus, important information regarding the latticedisorder can be obtained from simple shape analysis of Raman bands, which can bemade using the spatial correlation model. Figure 1 presents the Raman spectra of aSiC fibre and its fit according to the spatial correlation model. See Colomban et al.(Colomban et al, 02) for a comparison of SiC grain size in various fibres.

Band assignments in disordered carbons are still being discussed. Pure diamondand graphite having sharp peaks at 1331 and 1581 cm'1, respectively, the firsttemptation was to assign the main two bands of amorphous carbons to diamond-likeand graphite-like entities, the reason why the bands were named D and G (Fig. 1).There is actually no doubt that G band ensues from the stretching mode of Csp - Csp

bonds (E2g symmetry in graphite crystals). Resonance excites the bonds and makes theusual "structural approach" (group theory assignment) inappropriate. Csp

2-Csp3 bonds

must concentrate at carbon crystallite grain boundaries, in contact with the favourablysp3-hybridized Si and C atoms of the SiC fibre network. On account of the small sizeof carbon moieties, their contribution will be large. The density of these bonds isproportional to Lg

2, where Lg represents the mean size of graphitic moieties, whileCsp

2-Csp2 density should obey Lg

3 dependence. As a matter of fact, the intensity ratioID/IG is proportional to Lg"' (Gouadec et al., 01; ib. 02; Tuinstra et al., 70).

Figure 1: Centre, example of Raman spectrum of a Hi-S Nicalon™ SiC fibreshowing both the SiC and C fingerprints (A, = 5145nm, see the text for the labelexplanation); left, detail on the Si-C modes region (TO and LO modes) fitted withthe phonon confinement model. L is the calculated "grain" size and q the wavevector position within the Brillouin zone; right, modification of the C-C bondstretching modes for 12 different exciting wavelengths in the 450-680 nm range.


2.2. Correlation between Raman Spectra and Strength

Each point on Fig. 2 (left) corresponds to a given annealing temperature. There isobviously a linear correlation between the ultimate strength in SiC fibres and RamanD1350 band parameter. Its wavenumber and width shift respectively by 10 cm"1 and15 cm"1 every GPa (Gouadec et a/., 01; Colomban et al., 02). The linearity suggestsmacroscopic (strength measurement) and microscopic (Raman spectrum) responsesto stress obey the same phenomenon. The average failure strength can be consideredas the summation of the local response (seen in Raman) of the chemical bonds tomicro-stress. Other correlation, between Raman spectra and micro-hardness, havebeen evidenced and discussed (Amer et al., 99; Gouadec et at., 01). In carbon-richfibres, the Se extensometry parameter (see further) is proportional toE-0.5 (Gouadec etal. 02).

Figure 2: Correlation between Raman parameters (D band: left y-axis, thewavelength; right y-axis, the bandwidth "L") of NLM and Hi NicalonIM SiC fibresand the mechanical strength measured at different temperatures (x-axis). The plot ofthe full-width-at-half-height of carbon peak shows the diffusion of carbon during thesynthesis of the composite (black dots, pristine fibre; open dots, embedded in Tialloy).

3. The Composite Micrestructure Level

3.1. Phase Analyse of Coatings and Interfacial Regions

It is well established that the nature of the fibre-matrix interfacial region is veryimportant for the thermo-mechanical behaviour of composites. Information aboutthe change of the fibre surface can be obtained from the examination of extractedfibres or by in situ analysis of the fibre surface, periphery and core on compositessections polished nearly parallel to the fibre axis (Gouadec et al, 01). Coating andsurrounding matrix can be analysed in the same way. For instance Figure 3 shows aspectral map-scan (2|xm step) recorded on a perpendicular section of a SCS-6


Textron™ fibre embedded in a Ti6242 alloy (Gouadec et al, 00, Colomban 00).Fives zones are straightforward. From the fibre periphery: i) the pure carbonovercoating, the pure SiC outer zone with the broad SiC fingerprint characteristic ofnanosized, heavily faulted SiC, iii) the zone containing highly disordered graphiticcarbon and various types of SiC polytypes, iv) a carbon interphase and v) thegraphitic carbon core fibre. Comparison between the spectra of a pristine SCS-6fibre and that embedded in the Ti6242 alloy (Fig. 2) evidences the physical andchemical changes induced by the processing, the carbon diffusion from the core tothe first fibre periphery.

Figure. 3: Map-scan along the white rectangle on the optical micrograph (2/Mnstep) showing the different Raman spectra imaging the carbon content of a (140jUmdiameter) SCS-6 Textron™ fibre embedded in a Ti 6242 alloy. Carbon spectra ofHi-Nicalon™ fibres embedded in a celsian matrix in "micros-configuration(examination of a single fibre) and "macro"-configuration (simultaneousexamination of more than 1500 fibres) with A= 514.5 nm.

Similar analyse has been made on Hi Nicalon™ fibres reinforced monocliniccelsian prepared at the NASA Glenn Research Center (Cleveland, USA) (Gouadecetal.,01).

4. In situ Stress and Strain Measurement

Usually the harmonic oscillator approximation is used to describe the atomicmotions. Within this approximation, solid lattice spacings and Raman wavenumbersshould be independent of the temperature. When anharmonicity is taken intoaccount, the vibrational energy level of the oscillator are not equally spaced and thepotential is anymore symmetrical: any stress- (Ae), pressure-(Ap), temperature-(AT)induced interatomic distance alteration should change the interatomic force


constants and results in atomic vibrations wavenumber shift. (Av), e.g. for the strainAe and the stress Aa (Colomban 00):

This is the principle of Raman extensometry.

Following the pioneer works of Anastassakis and of Gardiner on the study ofstressed silicon and oxide films, Galiotis (Galiotis 93) and Young (Young 94) werethe first to demonstrate that the stress-induced-Raman shift could be used to followthe deformation of aramid and carbon fibres in polymer-matrix model composites.The results, obtained through this method on carbon fibres-reinforced (model)polymer matrix composites, have been extensively discussed by Schadler & GaliotisSchadler et al., 95), Beyerlein et al. (Beyerlein et al, 98), Kawagoe et al,(Kawagoe et al, 99), Amer & Schadler (Amer et al, 99) and Galiotis et al. (Galiotiset al, 99). Most studies over the past ten years concerned carbon and aramid fibresand their reinforced model composites, but some data on SiC/C fibres embedded indifferent matrices are now available (Yang et al, 94, ib 96, Gouadec et al, 98;Chollon et al, 98, Colomban 00, Colomban et al, 02). The study of Ceramic(CMCs) and Metal Matrix (MMCs) Composites is more complex. It is mandatory tocheck that the wavenumber shift provoked by the local laser-induced heatingremains lower than the wavenumber determination accuracy. The calibrationprocedure on single fibres loaded by controlling the applied strain is described byGouadec et al., 98. Fibres extracted from composites or thermally treated, in order tomimic the surface evolution during the composite synthesis need to be used to

obtain more reliable data. Se , is close to -7 and -10 cm'V% for carbon fibres,

-4 cm'V% and -2.7 cm-'/% for the NLM-Nicalon™ and Hi-Nicalon™ SiC fibres.Isolated carbon precipitates have typical -2/-3 cm"V%. Se increases with Raman bandorder, i.e. by using an harmonics or a combination band as a probe: Se is -28.9 cm"V% for the second order 4290 cm'1 combination (2xD+G) of the FT700™ carbon

fibre (Tonen, Japan) when first order Sp is only - 9.2 cm"V% (Gouadec et al. 02).

4.2. Limitations of the Technique and Corrections

The major limitations of the techniques are: i) a poor transparency of the matrixfor many "real" composites, ii) the rather small wavenumber shift which makes awell-defined procedure mandatory and iii) the fact that thermally induced Ramanshifts depend on the illuminated and the adjacent phases (thermostat). Anotherlimitation is that the fibre strain is calibrated only in the axial direction. Although,the translucency of ceramic/polymer matrices is sometimes sufficient to performanalysis of embedded fibres (up to 20-50 urn below the surface (Karlin et al. 97, ib98, Wu et al., 97) analysis is usually performed on polished sections. This ismandatory for metal (MMCs) and ceramic (CMCs) matrix composites. However,


given the difference in matrix stiffness with respect to organic matrices, the transferlength reduces to a few microns or less (to be compared with hundreds of micronsfor polymer matrix composites. Hence, valuable data can be measured on MMC orCMC polished sections (Wu et al, 97; Gouadec et al., 00; ib 01).

Examples of in situ results are sketched in Figure 3 & 4. The Referencewavenumber is obtained on the non-embedded, bare, coated or extracted fibres. Anylocalised heating induced by the laser impact will lead to an overestimate of tensilestresses and an underestimation of compressive ones (Gouadec et al., 01).

Figure 4: Schematic of the error when thermal effects are neglected: solid line,measurement without any correction; dashed line: corrected wavenumber aftersubtraction of the thermal induced down shift.

Not only the recording conditions but also the statistical dispersion between thefibres (batch, diameter, coating, environment...) must be taken into account toascertain the effect of chemical degradation or stress concentration on the Ramanspectra. The best method to obtain a statistical view is macro Raman examination(Fig. 3). With the ca. 2-3 mm diameter of the laser (macro) spot, thousands of fibrescan be examined simultaneously. The recorded spectrum integrates the contributionof all the fibres. Such a study using macro-configuration is very promising todetermine the mean properties of composites. However good spectra are onlyobtained if the fibre spectrum dominates those of the matrix and coatings.

5. Perspectives

The improved sensitivity of the most recent spectrometers decreases therecording time requested to map relevant parameters and, hence, facilitates theimaging of the physical, structural and chemical state, at the micron scale.Unpublished results show Se depends on the laser wavelength, which is correlatedwith laser penetration, for carbon bands. In-depth probing might be considered.Laser polarisation might also improve the accuracy of the method and could makespossible the discrimination between axial and radial components.


The author wishes to thank Drs. S. Karlin, J. Wu, G. Gouadec and M. G. Sagon,for their contributions. Special thanks are due to Drs N.P. Bansal (NASA) and M.Parlier (ONERA) for the samples they provided us with.

6. References

Amer M.S., Schadler L.S., "The Effect of Interphase Toughness on Fibre/Fibre Interaction inGraphite/Epoxy Composites: An Experimental and Modelling Study," Journal of RamanSpectroscopy, vol.30, no. 10, 1999, p.919-28.

Amer M. S., Busbee J., Leclair S.R., Maguire J. F., Johns J., Voevodin A., "Non-destructive,In situ, Measurements of Diamond-like-Carbon Film Hardness using Raman and RayleighScattering", Journal of Raman. Spectroscopy, vol 30, no. 10, 1999, p. 947-50.

Amer M.S., Schadler L.S., "The Effect of Interphase Toughness on Fibre/Fibre Interaction inGraphite/Epoxy Composites: An Experimental and Modelling Study," Journal of RamanSpectroscopy, vol.30, no. 10, 1999, p.919-28.

Beyerlein I.J., M.S. Amer, L.S. Schadler, S.L. Phoenix, "New Methodology for Determiningin-situ Fibre, Matrix and Interfaces Stresses in Damaged Multifiber Composites", Scienceand Engineering Composites Materials, vol. 7, no. 1-2, 1998, p.151-204.

Chollon G., Takahashi J., "La microscopic Raman appliquée aux composites Carbon/Carbon",Actes des I Icmes Journees Nationales sur les Composites -JNC //, Arcachon, 18-20Novembre 1998, vol. 2, AMAC, Paris, p. 777-85

Colomban Ph., "Raman Micro-spectrometry and Imaging of Ceramic Fibers in CMCs andMMCs"; in Advances in Ceramic Matrix Composites V; Ceramic. Transactions, vol. 103,2000, p.517-540.

Colomban Ph., "Raman Micro-spectrometry and Imaging of Ceramic Fibers in CMCs andMMCs", Ceramic. Engineering Science Proceedings, vol. 21, no. 3, 2000, p. 143-53.

Colomban Ph., Gouadec G., " Non-destructive Mechanical Characterization of (nano-sized)Ceramic Fibers", Actes 7"' Conference & Exhibition of the European Ceramic society -Euro Ceramics VII, Brugge, 9-13 September 2001, Key Engineering Materials vols. 206-213, 2002, p. 677-80.

Galiotis C., "Laser Raman Spectroscopy, a new Stress/strain Measurement Technique for theRemote and on-line Non-destructive Inspection of Fiber Reinforced PolymerComposites", Materials Technology, vol. 8, no.9-10, 1993, p.203-9.

Galiotis C., Paipetis A., Marston C., "Unification of Fibre/Matrix Interfacial Measurementswith Raman Microscopy," Journal of Raman. Spectroscopy, vol. 30, no. 10, 1999, p. 899-912.

Gouadec G., Karlin S., Colomban Ph., "Raman Extensometry Study of NLM202 and Hi-Nicalon SiC Fibres," Composites Part B, vol. 29B, 1998, p. 251-61.


Gouadec G., Colomban Ph., "De 1'analyse micro/nanostructurale et micromecanique a1'imagerie des fibres de renfort d'un composite a matrice metallique", Journal de PhysiqueIV France, vol 10, 2000, p. Pr4-69-PR4-70.

Gouadec G., Colomban Ph., " Non-Destructive mechanical characterization of SiC fibers byRaman spectroscopy", Journal of The European Ceramic Society, vol.21, 2001, p. 1249-59.

Gouadec G., Colomban Ph., Bansal N. P., "Raman study of Hi-Nicalon-Fiber-ReinforcedCelsian Composites: I, Distribution and Nanostructure of Different Phases", Journal ofThe American Ceramic Society, vol 84 no.5, 2001, p.l 129-35.

Gouadec G., Colomban Ph., Bansal N. P., "Raman study of Hi-Nicalon-Fiber-ReinforcedCelsian Composites: II, Residual Stress in Fibers", Journal of The American CeramicSociety, vol. 84, no. 5, 2001, p.l136-42.

G. Gouadec, Ph. Colomban, "Measurement of the residual Stress of Matrix-Embedded Fibersby Raman Spectrometry: State of the Art and Perspectives", Actes 7th Conference &Exhibition of the European Ceramic Society - Euro Ceramics VII, Brugge, 9-13September 2001, Key Engineering Materials, vols. 206-213, 2002, p. 617-20.

Gouadec G., Forgerit J.P., Colomban Ph., " Choice of the working conditions for Ramanextensometry of carbon and SiC fibers by 2D correlation", Composites Sciences &Technology, 2002.

Karlin S., Colomban Ph., "Raman Study of the Chemical and Thermal Degradation of As-Received and Sol-Gel Embedded Nicalon and Hi-Nicalon SiC Fibres Used in CeramicMatrix Composites," Journal of Raman Spectroscopy, vol. 28, 1997, p.219-28.

Karlin S., Colomban Ph., "Micro Raman study of SiC-oxide matrix reaction," CompositesPart B, vol. 29B, 1998, p. 41-50.

Kawagoe M., Hashimoto S., Nomiya, M. Morita M., Qiu J., Mizuno W., Kitano H.," Effectof Water Absorption and Desorption on the Interfacial Degradation in a Model Compositeof an Aramid Fibre and Unsaturated Polyester Evaluated by Raman and FT Infra-redMicrospectroscopy", Journal of Raman Spectroscopy, vol. 30, no. 10, 1999, p. 913-18.

Schadler L.S., Galiotis C, "Fundamentals and Applications of Micro Raman Spectroscopy toStrain Measurements in Fibre-Reinforced Composites," International Material Review,vol. 40, no. 3, 1995, p. 116-34.

Suzuki T., Kosacki I., Anderson H., Colomban Ph., "Electrical Conductivity and latticedefects in Nanocrystalline Cerium oxide thin films", Journal of The American CeramicSociety vol. 84 no. 9,2001, p. 2007-14.

Tuinstra F., Koenig J.L., "Characterization of Graphite Fiber Surfaces with RamanSpectroscopy," Composites Materials, vol. 4, 1970, p. 492-99.

Wu J., Colomban Ph., "Raman Spectroscopy Study on the Stress Distribution in theContinuous Fibre-Reinforced CMC," Journal of Raman Spectroscopy, vol. 28, 1997, p.523-29.

Yang X., Young R.J., "Fibre Deformation and Residual in Silicon Carbide Fibre ReinforcedGlass Composites", British Ceramic Transactions, vol. 93, no. 1, 1994, p. 1-10.


Yang X., D.J. Bannister, R.J. Young, "Analysis of the Single-Fiber Pullout Test Using RamanSpectroscopy: Part III, Pullout of Nicalon Fibers from a Pyrex Matrix", Journal of TheAmerican Ceramic Society, vol. 79, 1996, p. 1868-74.

Young R.J., " Raman Spectroscopy and Mechanical Properties", in Characterization of SolidPolymers, S.J. Spells Ed., p. 224-75, London, Chapman & Hall, 1994.

Demonstrator Program in Japanese SmartMaterial and Structures System Project

Tateo Sakurai* — Naoyuki Tajima* — Nobuo Takeda**Teruo Kishi***

* R&D Institute of Metals and Composites for Future Industries3-25-2 Toranomon, Minato-ku, Tokyo 105-0001, Japan

[email protected]

[email protected]

** Graduate School of Frontier Sciences, The University of Tokyo

*** National Institute for Materials Science

ABSTRACT: The Japanese Smart Material and Structure System Project has started in 1998 andhas been developing several key sensor and actuator elements. This project consists of fourresearch groups such as structural health monitoring, smart manufacturing, active/adaptivestructures, and actuator materials/devices. In order to integrate the developed sensor andactuator elements into a smart structure system and show the validity of the system, twodemonstrator programs have been established. Both demonstrators are CFRP stiffenedcylindrical structures with 1.5m in diameter and 3m in length. A Damage Detection andDamage Suppression function is to be demonstrated by the first one, and the second onedemonstrate a suppression of vibration and acoustic noise generated in the compositecylindrical structure. The present status of the demonstrator program is presented.

KEY WORDS: smart materials and structures, composite structures, damage detection, damagesuppression, noise and vibration reduction


1. Introduction

The "R&D for Smart Materials and Structures System" project has proceededsince late 1998 as the five-year program, being supported by NEDO (New Energyand Industrial Technology Development Organization), Japan. The project is one ofthe Academic Institutions Centered Programs, namely, collaborated research anddevelopment among universities, industries and national laboratories. At first, itconsisted of four sub-themes which were 1) Health Monitoring, 2) Active andAdaptive Structures, 3) Smart manufacturing, and 4) Actuator Materials andDevices. In early 2000, the Concept Demonstrator Program was added to theproject. It is aimed at evaluating what extent research and development items of sub-themes have attained their targets and establishing common basic technologies for afuture " Smart Structure System".

The Concept Demonstrator is focused on an aircraft fuselage of the compositestructures and designed to integrate several research and development results into it.Two demonstrators are being manufactured. The one is aimed at Damage Detectionand Damage Suppression, and the other is at Noise and Vibration Reduction.

The NEDO "R&D for Smart Materials and Structures" project in Japan is now thefirst runner of the Academic Institution Centered Programs in Japan, where thecollaborated research and development among universities, industries and nationallaboratories are conducted. Seven universities, seventeen companies and onenational laboratory take part in the project. RIMCOF (R&D Institute of Metals andComposites for Future Industries) is the management office of the project. Theproject includes the above four sub-themes and the Concept Demonstrator Program.Four sub-themes are mainly basic element level research and development and theConcept Demonstrator is actual application-oriented one. The organization of theproject is shown in Figure 1.

The Concept Demonstratoris designed so as to integrateresearch and developmentresults of four sub-themes. Ofcourse, we could not use anyresults at the start point of theproject. Therefore we startedthe preliminary design of theConcept Demonstrator twoyears later after the researchand development of four sub-themes started.

Figure 1. Organization of theproject


2. Selection of demonstration themes

Before starting the preliminary design, we discussed what themes wereappropriate for the purpose of the Concept Demonstrator Program. At first, we askedparticipating members for submission of demonstration theme proposal. Overthirteen proposals were submitted, but it was difficult to include all of them into oneor two demonstrators. So, we selected demonstration themes in accordance with thefollowing criteria, namely;

1) Is the theme an advanced technology?

2) Do users need the theme for future fuselage structures of an aircraft?

3) Is it possible to show the results of the developed research and developmenton the demonstrator?

4) Is it appropriate to schedule of the project?

Finally, seven themes were selected and classified into the following categories asshown in Table 1.

They are also divided into groups for two Demonstrators respectively, that is;

(1) Damage Detection and Damage Suppression: Theme #1 through #6

(2) Noise and Vibration Reduction: Theme #7

Table 1. Demonstration themes

#

1

2

3

4

5

6

7

DemonstrationCategories

Real Time Detection ofImpact Damage

Damage Detection

Damage Suppression

Smart Manufacturing

Noise and VibrationReduction

Demonstration Themes

Optical Fiber Sensors Embedded into CFRPLaminated Structures

Integrated Acoustic Emission Sensor NetworkSystems

Strain Distribution Measurement in Wide AreaUsing Distributed BOTDR*1 sensors

Damage Detection by electric conductivity changein Smart Patch (Carbon fiber composite sheets)

Damage Suppression System Using EmbeddedSMA (Shape Memory Alloy) Foils

Smart Manufacturing of Low Cost IntegratedPanel by RTM (Resin Transfer Molding)

Noise and Vibration Reduction Technology inAircraft Internal Cabin

* 1 Brillouin Optical Time Domain Reflectometer


3. Concepts of demonstrators

As mentioned earlier, the demonstrator is focused on an aircraft fuselage.Although it is desirable to test full size fuselage in the view of actual demonstration,it is expensive and takes long time in design and manufacture. Moreover, it needswide space and a lot of test and measurement facilities. On the other hand, it isdifficult for a small demonstrator to simulate primary physical parameters of thefull-scale fuselage due to minimum gauge of materials and standard parts (bolts,nuts, rivets and so on). Stress and strain are key parameters for demonstration theme#l-#6 of Table 1 (Damage Detection and Damage Suppression) and naturalfrequencies for #7 (Noise and Vibration Reduction). As a result of trade-off study,the diameter of 1.5m (approximately 1/3-scaled size of a small class jetliner) isdetermined.

It is impossible for the 1/3-scaled demonstrator to simulate both parameters ofstress/strain and natural frequencies simultaneously and, moreover, it is difficult foronly one demonstrator to perform all tests within the period of limited schedule.Consequently we decided to prepare two demonstrators.

Structures of the demonstrators are mainly made of composites, but some partsthat are not influential for physical parameters are made of metals due todevelopment cost reduction. Because of simulating an aircraft fuselage, innerpressure and external bending moment are to be loaded for the Damage Detectionand Damage Suppression Demonstrator. On the other hand, speakers and/or shakersexcite externally the Noise and Vibration Reduction Demonstrator without bendingmoment and inner pressure. Images of both demonstrators are shown in Figure 2 andFigure 3 respectively.

Figure 2. "Damage detection and damage suppression demonstrator" left

Figure 3. "Noise and vibration reduction demonstrator" right


4. Damage detection & damage suppression demonstrator

4.1. Structure of test article

Preliminary design of test article for the Damage Detection and DamageSuppression Demonstrator is summarized below, and the outline of the test article isshown in Figure 4.

-Test article: consists of composite materials, simulating an aircraft fuselage with alength of 3m and diameter of 1.5m.

-Structure: a build-up structure with composite skin-stringer panels and aluminumalloy frames. The panels are divided into four along the circle, and also the supportand the loading jigs at both ends are also divided into four corresponding to thepanels. The bulkhead panel can be freely removed/mounted, allowing a fastenerjoint to be connected to the loading jig section.

-Arrangement: The frame and stringer have a pitch of about 500mm and 150-200mm, respectively. The test article has a floor inside of the fuselage for testpreparations.

-Material: The skin-stringer panels are carbon fiber reinforcement composite. Theframes are made of aluminum alloy such as 2024 and 7075. The support and loadingjigs at both ends are made of steel.

Figure 4. Demonstration test article


(1) Upper panel

The upper panel is divided into three lengthwise at STA1000 and STA2000. Theskin in the range of STA1000-2000 is integrated with small-diameter optical fibernewly development in the present project for impact damage detection. The skin andstringer of upper panels are co-cured. Connection on the horizontal axis is madewith butt joints.

(2) Side panels

The side panels are not divided in the 3m lengthwise directions, and the skin-stringer panels are co-cured. The external panels have optical fibers embedded underthe layers of the skin and stringer between STA500 through 2500 as BOTDR sensorsfor widerange strain distribution measurement.

(3) Bottom panel

The bottom panel is divided into three lengthwise at STA 1500 and STA2500.Shape memory alloy foils are embedded in the external panel on the STA 1500 -2500 starboard for damage suppression. To reduce the production risk, the skin andstringer are assembled with the secondary bonding. Likely with the top panel, theconnection on the length of the panels are made with butt joints, and the joints withthe side panels are made with lap joints allowing it to pull out the electric heatingterminals.

(4) Bulkhead panel

A part of the pressure bulkhead at the load side has a removable structure, wherethe RTM formed panel is attached for the pressure test.

4.2. Test

(1) Test fixture

The Demonstrator is mounted to the test frame on the cantilever mode, and thefuselage bending load and internal pressure are applied.DFigure5 illustrates thedemonstrator test setup.

(2) Test loadO

Shear load (approx. 20 tons at max) is applied to the free ends of the test article asa bending load. Internal pressure (0.75atm at maximum) is applied to the test article.Various levels of impact loads (approx. 50 joules at maximum) are applied to theupper panel.

(3) Test sequence

The test is performed in the order of load-unload test, static test, pressure test andimpact test. In the load-unload test, the bending load is gradually increased in a


quasi-static condition. Before and after each test, visual and ultra-sonic inspectionsare performed. The test sequence is shown in Figure 6.

Figure 5. Demonstrator test setup

Figure 6. Test sequence

4.3. Demonstrator theme verification

The engineering contents to be verified in each demonstrator theme shown inTable 1 are outlined below. Verification positions in test articles are indicated inFigure 4.

(1) Real Time Detection of Impact Damage using Optical Fiber Sensors embeddedinto CFRP Laminated Structures

Using small-diameter optical fiber sensors embedded in the upper panel, detectionof any impact damage and identification of its location are demonstrated. They areverified in the impact test phase.

(2) Real Time Detection of Impact Damage using Integrated Acoustic EmissionSensor Network Systems

Using the AE sensor mounted on the side panel, time of occurrence, location andmagnitude of the impact load are identified. They are verified in the impact testphase.


(3) Strain Distribution Measurement in Wide Area Using Distributed BOTDRsensors

Using optical fibers embedded in the side panels and externally installed to theoverall test article, damage location and its magnitude are identified from the widerange strain distribution that is measured. Performed in the static test phase.

(4) Damage Detection by electric conductivity change in Smart Patch (Carbon fibercomposite laminate)

Two types of smart patches, carbon fiber fracture type and conductive particledispersion type, will be applied at the bottom panel of the demonstrator both in load-unload and static test phases to demonstrate the smart patches memorize the appliedmaximum strain.

(5) Damage Suppression using Embedded SMA (Shape Memory Alloy) Foils

Aims to verify that the shape memory alloy foils embedded in the bottom panel cansuppress the occurrence and growth of damages. In the load-unload test phase, theevaluation is performed by comparing the occurrence times of damage depending onwhether or not the shape memory alloy foil is present or not.

(6) Verification of Smart Manufacturing of Low Cost Integrated Panel by RTM(Resin Transfer Molding)

An optical fiber sensor, used for monitoring the manufacturing on the bulkheadpanel with RTM process, is verified in order to measure strains in the pressure testphase.

4.4. Test schedule

The test schedule of Damage Detection & Damage Suppression Demonstrator isshown in Table 2.

Table 2. Test schedule of damage detection & suppression demonstrator

II 2001 I 2002A M J U A S I O N D I J F M A M J I J A S l O N D l J F M

DesignManufactureTest PreparationTestEvaluation


5. Noise and vibration reduction demonstrator

Acoustic absorption materials have good noise reduction features in highfrequency range. But, in low frequency range, they need thick absorption layers inorder to reduce noise level significantly. Therefore, it is a practical solution to useactive noise control in low frequency range and acoustic absorption materials in highfrequency range. In Noise and Vibration Reduction Demonstrator, the targetfrequency range is below 500Hz.

The demonstrator is of the same size as the Damage Detection and DamageSuppression Demonstrator as mentioned in "Concepts of Demonstrators". Skinpanel thickness, dimension of stringers and frames and spaces between the stringersas well as the frames of the Noise and Vibration Reduction Demonstrator aredifferent from the Damage Detection and Damage Suppression Demonstrator due todifferences of key parameters to be simulated. In this demonstrator, naturalfrequencies are key parameters to be simulated. All the natural frequencies of thedemonstrator cannot meet those of the assumed jet liner. Therefore our policies toplacement of natural frequencies are the followings.

(1) To meet approximate natural frequencies of panel one bay enclosed by stringersand frames

(2) To meet the order of structural vibration natural frequencies and acousticvibration natural frequencies

In accordance with the above policies, dimensions of stringers and frames, spaceof them and shape of end caps are designed. High performance PZT actuators are tobe used which the "Actuator Materials and Devices" group developed.

The conventional way to reduce the noise in the internal cabin of the aircraft is tomount the sound absorption material. Sound absorption material is effective in noisereduction in a high frequency range, however not in a lower frequency range.Therefore, noise reduction with structure vibration control has been studied inresearch organizations worldwide. For the time being, however, verified noisereduction methods only apply to the specific frequencies and narrow band frequencyzone. In this test, therefore, we manufacture a test article assuming an aircraftfuselage with a size 1/3 of that of a small size passenger aircraft. Applying theinternal-cabin noise reduction technologies developed in the "active/adaptivestructure technology development" to the above test article, we plan to demonstratethe noise vibration reduction in a wide range of low frequencies for theactive/adaptive structure.

The outlines of Noise and vibration Reduction Demonstrator are described below.


5.1. Test objectives

There are two objectives in this test, which are:

- Increase the attenuation factor by 20 percent or more, and

- Decrease the noise level by 3 dB or more.

5.2. Test article

The test article is illustrated in Figure 7. At present, using software such asNASTRAN/MATLAB for the test article, acoustic vibration analysis and controlsimulation are performed to determine the number and arrangement of the PZTactuators that are optimal for noise and vibration reduction. We are also designingthe applicable control rules.

5.3. Test contents

The outline of the test is illustrated in Figure 8. This test consists of three itemsas listed below.

(1) Test for obtaining vibration characteristics data

The vibration characteristics (natural frequency/vibration mode/attenuationfactor) of the test article will be derived from the vibration force and vibrationacceleration data obtained from the test article structure by applying a vibration loadto the test article with a vibration exciter.

The vibration characteristics of the test article derived above are used toverify/review the PZT actuator arrangement based on the existing control rule designand to tune such control rules.

(2) Vibration control test

The vibration load is applied to the test article with the vibration exciter bothwhen the noise/vibration control system is operating and when it is not operating.Then, the vibration characteristics of the test article (natural frequency/vibrationmode/attenuation factor) are derived from the obtained vibration force and vibrationacceleration data on the test article structure.

It is verified that, by comparing the attenuation factors of the control system whenoperating and when not operating, the attenuation factor is improved by about 20percent or more.

(3) Noise control test

A noise load is applied to the test article from an external speaker in the anechoicroom to obtain the sound pressure level data inside of the test fuselage both whenthe noise/vibration control system is operating and when it is not operating.


It is verified that, by comparing the internal sound level of the test fuselageobtained when the control system is operating and when it is not operating, the

Shape and basic size: Cylinder, 1.5mDIAx3.0mL(exclude the bulkheads)Structural arrangement:Material of structures:

Sensor:Actuator:

Skin/Stringer/FrameSkin; C/EP FRP (P3060B-12)Stringer/Frame/Bulkhead; Al AlloyPZT/Accelerometer/Microphone/Strain gaugePZT

sound pressure level decreases by 3dB or more.

Figure 7. Test article of noise and vibration reduction demonstrator

Figure 8. Test configurations


5.4. Test schedule

The test schedule is shown in Table 3.

Table.3. Test schedule of noise and vibration reduction demonstrator

II 2001 I 2002A M J |J A SIO NDIJ F M A M J J J A S |O N D|J F M

Design of test articleManufacture of test articleVibration charactaristics acquisitionPZT installationV bration control testNoise control testEvaluation

6. Conclusions

The "R&D for Smart Materials and Structures" project has just finished the fourthyear of the five-year program. From the third year, the project has focused on twodemonstrators such as 1) Damage Detection and Damage Suppression and 2) Noiseand Vibration Reduction. Now, the detail design of both demonstrators hascompleted and some components are being fabricated and assembled. In the nextfiscal year (FY2002, April to March), the final assembly will be conducted and thetest is scheduled in the autumn. The test results will become available in the nextfiscal year.

Acknowledgement

This research is being conducted as a part of the "R&D for Smart Material andStructures System" project within the Academic Institutions Centered Programsponsored by METI entrusted to RIMCOF through NEDO (New Energy andIndustrial Technology Development Organization) in Japan.

We, herewith, gratefully acknowledge the support of METI, NEDO and all of theresearchers from industries, universities and national Institutes who have beenparticipating in this project.

Real-Time Damage Detection in CompositeLaminates with Embedded Small-DiameterFiber Bragg Grating Sensors

Nobuo Takeda — Yoji Okabe — Shigeki YashiroShin-ichi Takeda — Tadahito Mizutani — Ryohei Tsuji

Graduate School of Frontier Sciences, The University of Tokyoc/o Komaba Open Laboratories, The University of Tokyo4-6-1 Komaba, Meguro-ku, Tokyo 153-8904, Japan

Takeda@compmat. rcast. u-tokyo. ac.jp

ABSTRACT: Newly developed small-diameter fiber Bragg grating (FBG) sensors, whose outsidediameter was 52 mm, were applied for the damage detection in CFRP laminates. The FBGsensors are very sensitive to non-uniform strain distribution along the entire length of thegratings. Thus reflection spectra from the embedded FBG sensors deformed because of thestrain concentrations at tips of transverse cracks or the change in the strain distribution dueto a delamination. These deformations of the spectra could be reproduced by theoreticalcalculations. From these results, it was found that the small-diameter FBG sensors coulddetect the occurrence of the transverse cracks and the delamination quantitatively in realtime.

KEY WORDS: CFRP, fiber Bragg grating sensor, transverse crack, delamination, healthmonitoring, reflection spectrum


1. Introduction

CFRP composites are used in various fields owing to their high specific strengthand specific modulus. The failure process of CFRP laminates involves uniquemicroscopic damages, such as transverse cracks and delaminations (Takeda et al.,1994). The detection of these damages in real time is important in order to makepractical use of the CFRP laminates effectively and reliably.

A candidate for the sensing device of the microscopic damages is a fiber Bragggrating (FBG) sensor. FBG sensors are very sensitive to non-uniform straindistribution along the entire length of the gratings (Huang et al., 1994). The straindistribution deforms the reflection spectrum from the FBG sensors. Takingadvantage of the sensitivity, the authors applied FBG sensors for detectingtransverse cracks that caused non-uniform strain distribution in CFRP laminates(Okabe et al.,2000).

However, the cladding of common optical fibers is 125 mm in diameter, which isalmost the same as the normal thickness of one ply in CFRP laminates andapproximately 20 times larger than the diameter of carbon fibers. Thus, when thenormal FBG sensors are embedded into CFRP composites, there is a possibility thatthe optical fibers might deteriorate the mechanical properties of the laminates. Inorder to prevent the deterioration, small-diameter FBG sensors have recently beendeveloped by the authors and Hitachi Cable Ltd. (Satori et al., 2001). The outsidediameter of polyimide coating is 52 mm, and the cladding is 40 mm in diameter.

The small-diameter FBG sensors could also detect transverse cracks in CFRPcross-ply laminates sensitively (Okabe et al., 2002). In this research, the authorsattempted to detect transverse cracks in CFRP quasi-isotropic laminates using thesame method. Furthermore, the small-diameter FBG sensors were applied for thedetection of the delamination in CFRP laminates.

2. Detection of transverse cracks in a quasi-isotropic laminate

2.1. Experimental procedure

Bragg gratings were fabricated to have periodic refractive index changes in thecores of the small-diameter optical fibers. The outside diameters of the polyimidecoating, the cladding, and the core are 52 mm, 40 mm, and 6.5 mm, respectively. Thegrating length is 10 mm, and the grating period is about 0.53 mm. The profile of therefractive index modulation was controlled as a cosine function to suppress theside-lobe of the reflection spectrum (Satori et al, 2001).

These FBG sensors were embedded in CFRP T800H/3631 (Toray Industries,Inc.). The laminate configuration was quasi-isotropic: [45/0/-45/90]s. The FBGsensor was located in -45° ply on the border of 90° ply. Since the optical fiber wasembedded to be parallel to carbon fibers in -45° ply, it was hardly broken by the


stress concentrations due to transverse cracks that run through the thickness andwidth of the 90° ply.

Quasi-static tensile load was applied to the specimen by a material testing system(Instron Corporation, Load Frame 5582) at room temperature. The loading speedwas 0.25 mm/min. Tensile strain was measured with a strain gage attached on asurface of the specimen, and the tensile load was measured with a load cell. Theoptical fiber was illuminated by an amplified spontaneous emission (ASE) lightsource unit (Ando Electric Co., Ltd., AQ6310 (155)). The reflection spectrum wasobtained under tensile loading by using an optical spectrum analyzer (Ando ElectricCo., Ltd., AQ6317), and the specimen was unloaded after the spectrum measurement.Then, a polished edge surface of the specimen was replicated on a cellulose acetatefilm with methyl acetate as a solvent. From the replica film, the positions andnumbers of transverse cracks in 90° ply were observed. This loading/unloadingprocedure was repeated as the maximum strain was increased, until the specimenfractured completely.

2.2. Experimental results

Figure 1 shows the crack density p measured through the loading/unloading testfor the quasi-isotropic laminate with the embedded small-diameter FBG sensor as afunction of the tensile strain e. The crack density was defined as the number oftransverse cracks per unit length along the loading direction in 90° ply.

In Figure 2, the reflection spectra measured at various strain levels are shown.They correspond to the data (A) - (E) in Figure 1. While there was no transversecrack, the spectrum kept its shape and the center wavelength shifted correspondingto the strain. After transverse cracks appeared, the reflection spectrum deformed andbecame broad with an increase in the crack density p.

Figure 1. Crack density pas a function of strain e measured for the quasi-isotropiclaminate with the embedded small-diameter FBG sensor


Figure 2. Reflection spectra measured at various values of tensile strain £ Thesecorrespond to the data (A) - (E) in Figure 1

2.3. Analysis

For confirmation that the change in the form of the spectrum was caused bytransverse cracks, the spectrum was calculated theoretically. In the calculation, itwas assumed that the FBG sensor was affected only by the axial strain distribution,and the optical fiber adhered perfectly to the matrix of the -45° ply.

At first, the non-uniform strain distribution in the FBG sensor was calculatedusing FEM analysis with ABAQUS code. The CFRP laminate was analyzed by a3-D model that included transverse cracks in 90° ply and the optical fiber in ^45° ply.The positions where transverse cracks occurred were determined from theobservation of the replica films. The axial strain in the core was obtained along the

Figure 3. Calculated reflection spectra, which correspond to the measured spectrain Figure 2


entire length of the FBG sensor. Next, the distributions of the grating period and theaverage refractive index of the FBG were calculated from the axial straindistribution (Van Steenkiste et al., 1997). Then the reflection spectrum wassimulated from the distributions using the software "IFO_Gratings" developed byOptiwave Corp. This program can calculate the spectrum by solving the couplemode equations using transfer matrix method (Kashyap 1999).

The calculated results of reflection spectra are shown in Figure 3. These spectracorrespond to those in Figure 2. The change in the form of the calculated spectrum issimilar to that of the measured one. These results show that the change in thespectrum is caused by the non-uniform strain distribution due to the occurrence ofthe transverse cracks. Thus, the transverse cracks in quasi-isotropic laminates canalso be detected from the deformation of the reflection spectrum.

2.4. Dependence of spectrum width on crack density

With increase in the crack density, the width of the reflection spectrum changedin both the experimental result and the theoretical calculation. Thus, the spectrumwidth and crack density were plotted as a function of the tensile strain in Figure 4.The spectrum width was defined as full width at quarter maximum (FWQM) andnormalized by the value before loading. The FWQM obtained from the experimenthas the same tendency of an increase as the crack density. On the other hand, thecalculated FWQM increases drastically at the early stage of the crack accumulation,and the values are much larger than the experimental results. This is because thecalculated spectrum has many peaks over the broad range whose wavelength islonger than that of the maximum peak. This difference between the measured andcalculated results may be due to the inaccuracy of the strain distribution calculatedby FEM analysis and inexact optical parameters of the small-diameter FBG sensorused for the theoretical calculation. However, the calculation result agrees

Figure 4. Crack density and spectrum -widths as a function of tensile strain


Figure 5. Embedment position of small-diameter FBG sensor

qualitatively with the experimental result. Thus, the crack density in quasi-isotropiclaminates can be evaluated quantitatively by the spectrum width.

3. Detection of detainination

3.1. Experimental procedure

The above technique was also applied to the detection of delamination. Thespecimen was CFRP composite T800H/3631 and the laminate configuration wascross-ply [90]10/04/9010]. As shown in Figure 5, the FBG sensor was embedded in 0°ply to be parallel to carbon fibers and in contact with 90° ply. A strip typedelamination was grown along a 0°/90° interface by four-point bending test. For thedelamination onset from the tip of a transverse crack, a vertical notch wasintroduced at the mid-span of the specimen. The transverse crack occurred from theroot of the notch and reached the 0°/90° interface. An end of the FBG sensor was seton the tip of the transverse crack in order to propagate the delamination in onedirection within the region of the grating.

Quasi-static bending load was applied to the specimen with a four-point bendingdevice at room temperature. The optical fiber was illuminated by the ASE lightsource, and the reflection spectrum was obtained with the optical spectrum analyzerafter unloading. The length of the delamination was measured from soft X-rayphotograph. The total delamination length was expressed by d and divided into theleft part d\ and the right part dr.

3.2. Experimental results

Figure 6 shows the reflection spectra measured at various steps or thedelamination progress. These spectra were normalized by the intensity of the highest

various steps of the


Figure 6. Reflection spectra measured at various steps of delamination progress:(a) d = 0.0 mm, dr = 0.0 mm; (b) d = 5.4 mm, dr = 2.8 mm; (c) d = 8.4 mm, dr =4.2 mm; (d) d= 13.0 mm, dr = 6.4 mm

component. When there was only a transverse crack before the occurrence of thedelamination, the reflection spectrum had only one sharp narrow peak as shown inFigure 6(a). After the delamination was initiated from the crack tip, another peakappeared at longer wavelength. The intensity of the longer wavelength peakincreased relatively with an increase of the delamination length.

3.3. Analysis

The reflection spectra were also simulated theoretically. In this case, the laminateincluding the delamination was analyzed using 2-D plane strain model. From thecalculation, the strain distribution at 25 mm above from the 0°/90° interface, wherethe center axis of the embedded optical fiber was positioned, was obtained. Then,the longitudinal strain distribution in the 0° ply was assumed to be the same as theaxial strain distribution in the FBG sensor, and the reflection spectrum wassimulated from the strain distribution.

Figure 7 shows the calculated results, which are also normalized by the intensityof the highest component. These spectra reproduce the measured spectra shown inFigure 6 very well. The strain distribution that was obtained by FEM analysis andused for the calculation of the spectrum in Figure 7(c) is plotted in Figure 8. Thisstrain distribution mainly consists of two strain levels: level I and II. Thus, thereflection spectrum was calculated on the assumption that the FBG sensor wassubjected to the uniform strain of level I or II. As a result, it was found that thelonger and shorter wavelength peaks of the spectrum in Figure 7 corresponded to thestrain of level I and II, respectively. The level I and II are related to the strain at thedelaminated area and that at the bonded area, respectively. Hence, as the


Figure 7. Calculated Reflection spectra, which correspond to the measuredspectra in Figure 6

delamination length increases, the region of the uniform strain at the level I willenlarge, so that the intensity of longer wavelength peak in the spectrum will increaseconsistently.

For the quantitative evaluation of the delamination length, the intensity ratio ofthe two peaks IL/Is is defined, where /IL, and IS are the intensities of longer and shorterwavelength peaks, respectively. Figure 9 shows the logarithmic plot of the intensityratio against the delamination length along the FBG sensor, which is expressed by dr

in Figure 5. During the dr is less than 4.2 mm, the IL/IS obtained from the experimentand that from the theoretical calculation are almost the same. However, when the dr

is over 4.2 mm, the increase of the calculated IL/IS becomes larger than that of themeasured IL/IS. This difference was caused by the error of delamination lengthmeasurement using soft X-ray radiography. Since intralaminar delaminations were

Figure 8. Strain distribution along the FBG sensor at d = 8.4 mm and dr = 4.2 mm


Figure 9. Intensity ratio of the two peaks against delamination length along theFBG sensor

found in 0° ply around the tip of the interlaminar delamination by the observation ata polished edge surface using optical microscope, the dr measured from the softX-ray photograph might be larger than the actual length of the interlaminardelamination due to the existence of the intralaminar delaminations. Thus, the peakintensity at the longer wavelength in the calculated spectrum was estimated to behigher than that in the measured spectrum. However, since the intensity ratiosobtained from both experiment and calculation increase in monotone with anincrease of the delamination length, the intensity ratio of the two peaks can be aneffective indicator for quantitative evaluation of the delamination length.

4. Conclusions

In this research, newly developed small-diameter FBG sensors, whose claddingand polyimide coating diameters are 40 mm and 52 mm, respectively, were applied todetect the transverse cracks and the delamination in CFRP laminates.

First, for the detection of the transverse cracks, the FBG sensor was embedded in-45° ply of a CFRP quasi-isotropic laminate [45/0/-45/90]s. When a tensile load wasapplied to the specimen, the form of the reflection spectrum from the FBG sensorwas distorted sensitively, as the crack density in 90° ply increased. Then thereflection spectrum corresponding to the measured one was calculated theoretically.The calculated spectrum reproduced the change in the form of the spectrum verywell. From this agreement, it was confirmed that the change in the spectrum wascaused by the non-uniform strain distribution, which was induced by the transversecracks. Hence, the transverse cracks in quasi-isotropic laminates could also be


detected from the deformation of the reflection spectrum. Furthermore, the crackdensity could be evaluated quantitatively by the spectrum width.

Secondly, the delamination originating form a tip of a transverse crack in across-ply laminate [9010/04/9010] was detected using a similar technique. After theFBG sensor was embedded in 0° ply on the border of 90° ply, the delamination wasgrown along a 0°/90° interface by four-point bending test. When the delaminationappeared, the reflection spectrum had two peaks, and those intensities changeddepending on the delamination length. From theoretical calculation, it wasconfirmed that the two peaks corresponded to the uniform strain at the delaminatedarea and that at the bonded area. Hence, the intensity ratio of the two peaks wasfound to be an effective indicator for the prediction of the delamination length.

Acknowledgements

This research was conducted as a part of the "R&D for Smart Materials StructureSystem" project within the Academic Institutions Centered Program supported byNEDO (New Energy and Industrial Technology Development Organization), Japan.

References

Huang S., Ohn M.M., LeBlanc M., Measures R.M., "Continuous arbitrary strain profilemeasurements with fiber Bragg gratings," Smart Mater. Struct., vol. 7 no. 2, 1998, p.248-256.

Kashyap R., Fiber Bragg gratings, San Diego, Academic Press, 1999.

Okabe Y., Mizutani T., Yashiro S., Takeda N., "Detection of microscopic damages incomposite laminates with embedded small-diameter fiber Bragg grating sensors," Compo.Sci. Technol., 2002, (accepted for publication).

Okabe Y, Yashiro S., Kosaka T., Takeda N., "Detection of transverse cracks in CFRPcomposites using embedded fiber Bragg grating sensors," Smart Mater. Struct., vol. 9 no.6, 2000, p. 832-838.

Satori K., Fukuchi K., Kurosawa K., Hongo A., Takeda N., "Polyimide-coated small-diameteroptical fiber sensors for embedding in composite laminate structures," Proc. SPIE, vol.4328, 2001, p. 285-294.

Takeda N., Ogihara S., "In situ observation and probabilistic prediction of microscopic failureprocesses in CFRP cross-ply laminates," Compo. Sci. Technol., vol. 52 no. 2, 1994, p.183-195.

Van Steenkiste R.J., Springer G.S., Strain and temperature measurement with fiber opticsensors, Lancaster, Technomic, 1997.

Measuring the non-linear viscoelastic,viscoplastic strain behaviour of CFREusing the electronic speckle patterninterferometry technique

Pascal, J.-P. Bouquet - Albert, H. Cardon*

Department Mechanics of Materials and Constructions (MEMC),Vrije Vniversiteit Brussel (VUB),Pleinlaan 2,B-1050 Brussels- Belgium

Pascal.Bouquet@,vub. ac. be* mbourlau@vub. ac.be

ABSTRACT: Polymer matrix composites behave as viscoelastic-viscoplastic anisotropiccontinua. Various models, based on the viscoelastic behaviour, propose an acceleratedcharacterisation procedure for composites that would allow the prediction of long termproperties from short-term experiments including time-stress-superposition procedures andnon-linear viscoelastic behaviour under creep conditions. Creep measurements of testspecimen provided with strain gages and/or extensometers are not conclusive on the lifetimeprediction of these carbon-fibre reinforced epoxy matrix composites. The question arises ofthe failure initiation in the test specimen and the ability to measure the mechanical responseof the unidirectional composite material. Digital imaging methods like the Electronic SpecklePattern Interferometry are full field techniques to determine in situ properties at a local scalecommensurate with the continuum modelling procedure.

KEYWORDS: Creep, Electronic Speckle Pattern Interferometry, viscoelasticity, viscoplasticity


1. Introduction

The matrix of a polymer-based composite is time dependent and is sensitive tothe environmental conditions. Unidirectional reinforced polymer matrix compositesbehave as viscoelastic-viscoplastic anisotropic continua, which concerns not onlythe stiffness but also the strength characteristics.

Various models for the lifetime prediction consider the changes in stiffnessproperties as an expression of damage superposed to a viscoelastic-viscoplasticmodel. These models based on the viscoelastic behaviour propose an acceleratedcharacterisation procedure for composites that would allow the prediction of longterm properties from short term experiments including time-temperature-stress-superposition procedures and non-linear viscoelastic behaviour as well as modelsto predict delayed failures such as creep ruptures.

Creep measurements obtained from different load levels of test specimenprovided with strain gages and/or extensometers were not conclusive. The questionarises of the failure initiation in the test specimen and the ability to measure the non-homogenisation in mechanical response of the unidirectional composite material.Digital imaging methods like the Electronic Speckle Pattern Interferometry aretechniques to determine in situ properties at a local scale commensurate with thecontinuum modelling procedure. The resolution of the method combined with thearea of inspection drastically improves the monitoring of the strains on the outersurface.

It is emphasised that the measurement doesn't allow in depth measurement likee.g. ultrasonic inspection, however it presents some promising features, especially ifthe in-depth events can be related to the surface behaviour. Testspecimen aresubjected to artificially introduced defects, a hole, as to simulate mechanicalbehaviour in the presence of non-homogeneities or damage.

2. Non linear viscoelastic-viscoplastic analysis

The method to describe the viscoelastic behaviour used is based on thegeneralised time-temperature-stress superposition principle as developed bySchapery. In a uniaxial stress situation the equation describing the strain is:


where dY' = — is the reduced time and S0 the instantaneous compliance, AS is the

transient compliance and g0, g1, g2 and as are the non-linearising functions.

Creep and creep recovery tests have shown that plasticity has to be included inthe analysis. The Zapas-Crissman functional was proposed.

Summarising all the sources of deformation we obtain the strain as a function ofelastic, viscoelastic, viscoplastic and damage behaviour.

3. Damage analysis

Typically modulus degradation in measured stress-strain behaviour together withpermanent deformation is used as a basis of the extent of damage in a polymermatrix composite. This assumption has to be used with much care since this is onlyvalid as long as the measurement technique is to obtain strain values commensuratewith the size of the inferred damage region or a "representative volume element"(RVE). One might question if the strain measured by an electrical strain gage orextensometer is truly representative of the strains within the damaged regionsespecially when failure occurs outside the strain measurement device range. Modulidetermined by this classical measurements may not be representative of localconstitutive behaviour and analytical models based upon global observations. Onecan imagine that under constant creep conditions, stresses are not uniform but differon the whole test area resulting in a non-homogenuous strain field. Carbon fibrereinforced epoxy resin test samples were subjected to incremental loading. Tenlinear load cycles proportional to a tenth of the ultimate stress value were performedand the strains were measured with strain gauges. The respective stiffness to eachload cycle was calculated (Figure 1).


Figure 1: stiffness evolution for |90°|10 laminates

From these measurements the evolution of the stiffness as a function of itsloading history is hardly noticeable. One would assume that the specimen would beaffected mechanically when loaded at stresses commensurate to the rupture stress.Modulus calculations from cyclic mechanical loading of testspecimen with straingages were not conclusive.

4. Tensile creep lifetime analysis

The original purpose of the research was to obtain a prediction model for thelifetime of a long-fibre thermoset matrix composite based on the knowledge andexperimental experience of the Schapery model for non-linear viscoelasticbehaviour. Lifetime prediction, as discussed by Hiel, was based on the free energyaccumulation during creep and was based on a chain of mechanical Kelvin modelsapproximated with the Power Law. Experiments until failure for various creeplevels and different off-axis laminates were performed but gave non-satisfactoryresults, e.g. Figure 2.


Figure 2: experimental creep strain curves at an equal elevated stress level

5. Electronic Speckle Pattern Interferometry

The technology is based on the scattered reflection of incident light on a roughsurface. By applying monochromatic laserlight on the surface of the testspecimenthe scattered light will have a characteristic granular appearance the so-calledspeckle pattern (Figure 3). Each point from the reflection surface scatters theemitted wave. The path lengths travelled by these waves, from source to objectpoint to the receiving point, can differ from zero to multiples of wavelengths,depending on surface roughness and the geometry of the system. Interference of thede-phased but coherent waves arriving at the receiving point will cause the resultantirradiance to be anything from dark to fully bright. The resultant of the wavesarriving at a neighbouring point will probably give a quite different brightness. Thisvariation in resultant irradiance from one receiving point to another is the cause oflaser speckle. This type of speckle is known as the objective speckle.


Figure 3: Speckle image of a carbon-epoxy specimen with a hole

When using two identical waves symmetrical incident on an object surface; acamera aligned perpendicular to the reflecting surface will visualise an interferencialimage due to the combination of the two speckle patterns. The camera video signalcorresponding to the interferometer image plane speckle pattern of the undisplacedobject is stored electronically, whereas the live video image of the displaced object,detected by the camera, is subtracted from the stored picture electronically.

Figure 4: the interference fringe image of the specimen of Figure 3 uponloading.

The output is then high-pass filtered, rectified and displayed on a monitor wherethe correlation fringes are observed in real time. In order to understand theformation of fringes, consider the intensities of the beams Ibefore, before displacementand Igfterthe intensity after displacement in each point of the image.


where f is the phase difference between the reference beam and the object beam

before the displacement, Df is the phase change caused by the displacement

Assumed that the camera output signals Vbefore and Vafier are proportional to the input

image intensities, the subtracted signal V, is then given by

Figure 5: set-up of the ESPI camera in a tensile test


Figure 6: Block diagram of the electronic speckle pattern interferometer set-up

6. Crack area monitoring on a CFRE beam with a central hole

It has been observed that the area of increased strain at the initiating failure ishardly detectable with the strain gage technique due to its limited measurement baseand the averaging of the measurement over it's measurement area.

A hole was drilled centrally in a composite specimen as to monitor the strains inthe vicinity of the inhomogeneity. The [+/-450, 90°3]s specimen was subjected to atensile test till rupture. From these pictures, Figure 7 and Figure 8, the non-uniformstrain field is visualised, especially the initiation of the rupture is noticeable at thecircumference of the hole at the left and right side for both x and y deformations.

Figure 7: Deformation x-direction [mm]


Figure 8: Deformation y-direction [mm]

In Figure 9 the broken specimen is shown. It is noticed that the crackpropagation is different in its location for each layer of the laminate, but started atthe location of the strain disturbances at the left and right side of the circumferenceof the hole.

Figure 9: ruptured test beam [+/-450,90°3]s


7. Conclusion

From the experiments we obtain information with a high resolution for thelocal strain response on the surface of a composite of carbon fibre reinforcedepoxy resin.

The aim of the analysis of the non homogeneous strain around the hole is tounderstand the mechanical behaviour of materials and especially of fibre reinforcedpolymer composites in the vicinity of discontinuities like cracks or voids. Thefurther research consists in the analysis of crack induced CFRE material.

Microscopic discontinuities are hard to measure with classical strain gagetechniques.

This technique shows some attractive features in the analysis of complex mechanicalsystems like composite materials.

Acknowledgements

This research was made possible by financial support from the Science fund ofthe Flemish region (FWO-Vlaanderen) and the Research Council of the VrijeUniversiteit Brussel (OZR-VUB).

8. References

Schapery, R.A., 1967, "Stress analysis of viscoelastic composite materials", Journal ofComposite Materials, 1: pp. 153-192.

Hiel, C., 1983, "The nonlinear viscoelastic response of resin matrix composites", PhD-thesisVUB -Virginia Polytechnic Institute.

Cardon, A.H., Bouquet, P., Van Vossole, Chr., "Structural integrity, durability and reliabilityof polymer based composite systems - recent developments (What do we need? What isavailable?), Proc. of the International Conference on Composite Science and Technology(ICCST-3), Durban, South Africa, pp. 217-222.

Brinson, H.F., "Matrix dominated Time dependent failure predictions in polymer matrixcomposites", Composite Structures 47 (1999): pp.445-456.

Ennos A. E., Speckle Interferometry; Dainty J.C. (ed.), Laser speckle and relatedphenomena, pp. 203-253; ISBN 0 387 07498 8.


Proceedings International -workshop: "Video-Controlled Materials testing and in situmicrostructural characterization", 1999, Ecole des Mines de Nancy (France).

Wattrisse B., Chrysochoos A., Muracciole J.-M. and Nemoz-Gaillard M, "Analysis of strainlocalization during tensile tests by digital Image correlation", Experimental Mechanics(SEM), Vol.41, n°. 1, March 2001.

Mechanical Property and Application ofInnovative Composites Based onShape Memory Polymer

Qing-Qing Ni — Takeru Ohki — Masaharu Iwamoto

Kyoto Institute of TechnologyDivision of Advanced Fibro-Science in Graduate SchoolMatsugasaki, Sakyo-ku, Kyoto 606-8585, Japan

nqq@ipc. kit. ac.jp

b622071 l@ipc. kit. ac.jp

iwamoto@ipc. kit. ac.jp

ABSTRACT: Recently, shape memory polymer as one of functional materials has received muchattention and its mechanical properties have been investigated. Shape memory polymer ofpolyurethane series has the glass transition temperature (Tg) around the room temperature.Based on the large change in modulus of elasticity above and below Tg, the material hasexcellent shape memory effect. In this study, the glass fiber reinforced shape memory polymerwas developed for wide applications in the fields of industry, medical treatment, welfare anddaily life. The specimens with different fiber weight fractions were fabricated and theirmechanical behavior was mainly investigated experimentally. Then, the influence of fiberweight fraction on the shape memory effect was evaluated. It was confirmed that static andcyclic behavior of the shape memory polymer was improved by the reinforcement of fibers andthe shape memory effect was measurably kept in the developed composites.

KEY WORDS, shape memory polymer, stress-strain-temperature relation, mechanical property,fiber weight fraction


1. Introduction

Recently, shape memory materials have received much attention in industries andthe other fields, particularly for Shape Memory Alloys (SMAs) that are a group ofmetallic alloy and exhibit a shape memory effect. On the other hand, shape memorypolymers (SMPs) are mentioned as one of such shape memory materials. AlthoughSMPs indicate a phenomenon that the deformed shape returns to the original shape byheating, the mechanism of shape memory effect and the change of mechanicalproperties in the SMPs were different from those in the SMAs. Compared with SMAs,SMPs have the advantages, such as lightweight, large recovery ability, superiorprocessability and lower cost. Most of SMPs has the glass transition temperature (Tg)around the room temperature. Based on a consequence of the thermo-elastic phasetransformation and its reversal at the temperatures above and below Tg, SMPs haveexcellent shape memory effect. This means that SMPs may also be used as atemperature sensor or an actuator.

In the SMPs, the polyurethane series has following advantages: the formingprocesses for other thermoplastic polymer can still be used; the shape recoverytemperature can be set at any value within ±50K around the room temperature; thereexist the large differences of the mechanical properties (Tobushi et. al, 1991, 1992 &1998), the optical property and the water vapor permeability at the temperaturesabove and below Tg. Based on these advantages, the SMP of polyurethane series areexpected to have wide applications in the field of industry, medical treatment, welfareand daily life. However, the use of these materials was quite limited due to lowstrength of the polyurethane SMP bulk.

In this study, fiber reinforced composites based on SMP were developed in orderto overcome the low strength of SMP bulk. The materials developed were the glassfiber reinforced SMP of the polyurethane series with different fiber weight fractions(Ni et. al, 1999 & 2000). For the practical use of developed composites, it isimportant to clarify the fundamental mechanical and cyclic properties to meet thereliability requirement. Additionally, the thermo-mechanical behaviors, such asstress-strain-temperature relations with the influence of thermal factors, are alsoimportant. Thus, the mechanical properties of the developed composites withdifferent fiber weight fractions (SMP bulk, 10wt%, 20wt%, 30wt%) and testingtemperatures (Tg-20K, Tg, Tg+20K) were evaluated in static tensile test. Cyclic testsin two conditions, i.e., constant strain and constant stress, were performed at roomtemperature (Tg-20K) with different fiber weight fractions. Then, thermo-mechanicalcycle tests with consideration of both mechanical and thermal factors were carried outand the influence of fiber weight fraction and the thermal condition on shape memoryeffects was investigated.



2.1. Fabrication of specimens

As the matrix of developed composites, the shape memory polymer (DIARY,MM4510: MITSUBISHI HEAVY INDUSTRIES Co., Ltd.) was used with Tg about318K. As the reinforcement, the chopped strand glass fibers with fiber length of 3mm(03MA411J,ASAHI FIBER GLASS Co., Ltd.) were used.

The matrix and reinforcements were compounded by a twin screw extruder(LABOTEX-300, produced by JAPAN STEEL WORKS Co., Ltd) at the cylindertemperature of 483K and the screw rotation of 200rpm. The fiber weight fractionswere SMP bulk, 10wt%, 20wt% and 30wt%, respectively. Dumbbell type specimens(JIS K7113 Typel) were fabricated by an inline screw type of injection moldingmachine (Plaster Ti-30F6, produced by TOYO MACHINERY and METAL Co., Ltd.)after enough drying of compounded materials at 353K. The fabricated specimens arenon-weld. The cylinder temperature, mold temperature and injection speed were483K, 303K and 27.4 cm3/sec, respectively. Figure 1 illustrates the geometry of aspecimen.

Figure 1. Geometric shape and size of specimen

2.2. Experimental equipment

The experimental equipment used in this study was an Instron Universal TestingInstrument (Type 4466) with a temperature-controlled chamber. Heating or coolingfor specimens was controlled by compressed and heated or cooled air in theatmosphere condition and the temperature was measured by a thermocouple near thespecimen. The tip of the thermocouple was put between two 1.5mm thickness plateswith the same material as the specimen to make the same temperature conditionwithin the specimen.


2.3. Static tensile test

The static tensile test was performed with cross head speed of 5mm/min. withinthe temperature-controlled chamber at the testing temperatures of 298K(Tg-20K),318K(Tg) and 338K(Tg+20K), respectively. The strain was calculated by the ratios ofthe elongation obtained by the crosshead displacement to the span length (60mm)with a maximum of 300% due to the limit of the chamber.

2.4. Mechanical cycle test

Mechanical cycle tests were performed at room temperature (298K=Tg-20K). Forthe testing condition of constant stress, the upper limit stress was set to be 50% valueof the maximum stress in a static tensile test. For the testing condition of constantstrain, the upper limit strain value was set to be 50% value of the strain at themaximum stress in a static tensile test. Both cyclic tests were performed at thecrosshead speed of 5 mm/min. until the cycle numbers of 20,40 and 60, respectively,to observe the influence of cycle number and fiber weight fraction on mechanicalbehavior.

2.5. Thermo-mechanical cycle test

Figure 2. The schematic of thermo-mechanical cycle test

Thermo-mechanical cycle tests were performed to investigate the strain recoveryafter different number of cyclic loading. Figure 2 shows a schematic of stress-strain


curves in a thermo-mechanical cycle. The specimen was loaded to the strain em at aconstant crosshead speed of 5 mm/min. at the temperature Th (Process 1). Then, it wascooled to the temperature T| by keeping the same strain em (Process 2). After fiveminutes at the temperature T1, the load on the specimen was taken off (Process 3), andthen the specimen was heated from T1 to Th during ten minutes under no-loading(Process 4). This forms one thermo-mechanical cycle and then the test was repeated toN cycles. The conditions for the thermo-mechanical cycle test were as follows: Em=100%, Th=338K, T,=298K, the crosshead speed of 5mm/min., and N=5. The strainwas measured as done in the static tensile test.

Figure 3. The stress-strain curves in static tensile test at T=298 K


3.1. Static tensile property

In calculation of the experimental data, the engineering stress and strain were used.Figure 3 shows the stress-strain curves at the testing temperature of 298K(Tg-20K).The figures for 318K(Tg) and 338(Tg+20K) were omitted. When the temperature wasat T=298K(Tg-20K), the 10wt%, 20wt% and 30wt% specimens had small fracturestrain, while the bulk specimen was of a upper yielding point and had no fracturewithin the strain range of 300%. However, the yielding phenomenon was observed forall different fiber weight fraction specimens due to occurrence and growth of localnecking during testing. For the stress-strain curves at T=318K(Tg), 20wt% and30wt% specimens ruptured at the strain of 120% and 220%, respectively. However,the bulk and 10wt% specimens had no fracture within the testing limit strain of 300%.At higher temperature T=338K (Tg+20K), the specimens indicated a lower stress andthe final fracture did not occur within the strain range of 300% except the 30wt%specimen.


Figure 4. The Relationship between Temperature and Young's modulus

Figure 4 shows the relationship between temperature and Young's modulus. Thelarge change of Young's modulus above and below Tg was observed for all ofspecimens, which is a key point to utilize and control the shape memory effects ofSMP based materials. In other words, the result in Fig.4 means that the developedmaterials may have shape memory effect.

The results in the static tensile tests can be remarked briefly as follows: a obviousincrement of the strength of the developed materials when fiber weight fractionincreased; a high Young's modulus and high yield stress at low temperature; a largechange in Young's modulus above and below Tg for all materials.

3.2. Mechanical cycle property

Figure 5 shows the stress-strain curves in constant strain cycle tests. For thespecimens with different fiber weight fractions, a large hysteresis loop was observedat first cycle and there was no obvious difference in the loop shape except the slope ofthe loop, which corresponded to the Young's modulus. It is considered that the largehysteresis loop at first cycle is contributed by matrix deformation and failures aroundfibers. However, the loops following the first cycle showed almost no hysteresis dueto the characteristics of SMP with a training effect. Figure 6 shows the total residualstrain after prescribed cycle numbers of 20,40, and 60. The total residual strain in thebulk and 10wt% specimens increased between 20 cycles and 40 cycles, and tended tobe stable after 40 cycles. But it seems to be unchanged in 20wt% and 30wt%specimens even the cycle number was larger. This indicates that reinforcement fibersmixed in the SMP will reduce stress decrement and stabilize the cyclic behavior ofdeveloped materials.


Figure 5. The stress-strain curves in mechanical cycle test

Figure 6. The Residual strain for each fiber weight fractionafter cyclic loading


3.3. Thermo-mechanical cycle property

Figure 7 shows the stress-strain curves in a thermo-mechanical cycle test. Themaximum stresses and Young's modulus in each cycle increased with the increment ofthe cycle number N. This may be caused by the strain hardening of the materials. Here,let us look at the strain e r (see Fig.2), i.e., the strain recovered when the materialswere heated from T1 to Th without loading (Process4).

Figure 7. The stress-strain curve for 10wt% in thermo-mechanical cycle test

Figure 8. The relationship between strain recovery ratios and the number of cycle


The relationship between the strain recovery ratio and the number of cycle areshown in Fig.8. The strain recovery ratio is defined by the value of er / em. Inrecovery ratio at first cycle, considerable difference appeared for the specimens withdifferent fiber weight fractions. It is clear that the strain er in the specimens with fiberweight fractions of 10, 20 and 30 indicated greatly lower value than that in bulkspecimen. But, the strain er after second cycle was almost unchanged.

However, the parameters, such as the recovery time and temperature, may controlstrain recovery ratio. In the case of changing recovery temperature to 358K (Tg+40K)by keeping recovery time (10 minutes) same, the relationships between the strainrecovery ratio and the number of cycle are shown in Fig.9. Recovery ratio was high inall specimens in comparison with the case of the recovery temperature 338K (Fig.8)and varied greatly with different fiber weight fractions. With these results, therecovery temperature may be a dominant parameter as compared with the recoverytime in the shape recovery effect.

Figure 9. The strain recovery ratios at 358 K

4. Conclusions

In this study, the composites based on the SMP were developed and their cyclicbehavior and shape memory effects were investigated by the experimental approach.The results obtained are remarked as follows.

1. The tensile strength of the developed materials became higher with theincrement of fiber weight fraction under each temperature condition.

2. The resistance to cycle loading for the composites with SMP was clearlyimproved due to reinforcement fiber.


3. It is predicted that there exists an optimum fiber weight fraction between 10wt%and 20wt% to have an extremely low residual strain during cyclic loading.

4. The temperature was a dominant parameter as compared with recovery time forthe shape recovery effect, and this will be a useful opinion in the practical use.

5. It was confirmed that developed composites measurably keep the shape memoryeffect.

References

Ni Q., Ohsako N., Sakaguchi M, Kurashiki K and Iwamoto M., "Mechanical Properties ofSmart Composites Based on Shape Memory Polymer", The 24th Composites Symposium ofJapan Society for Composites Materials, Japan, 1999 p. 17 (in Japanese).

Ni Q., Ohsako N., Sakaguchi M, Kurashiki K and Iwamoto M., "Mechanical Properties ofSmart Composites Based on Shape Memory Polymer", JCOM: JSMS COMPOSITES-29 ofthe society of Material Science, Japan, 2000, p. 293 (in Japanese).

Tobushi H., Hayashi S. and Kojima S., "Mechanical Properties of Shape Memory Polymer ofPolyurethane Series", Transactions of the Japan Society of Mechanical Engineers, A, 57,1991, p. 146 (in Japanese).

Tobushi H., Hayashi S. and Kojima S., "Cycle Deformation Properties of Shape MemoryPolymer of Polyurethane Series'", Transactions of the Japan Society of MechanicalEngineers, A, 58, 1992, p. 139 (in Japanese).

Tobushi H., Hayashi S. and Kojima S., "Constitutive Modeling for Thermo-mechanicalProperties in Shape Memory Polymer of Polyurethane Series", Transactions of the JapanSociety of Mechanical Engineers, A, 64, 1998, p. 186 (in Japanese).

Piezoelectric Fibers and Composites forSmart StructuresAndreas Schonecker — Lutz Seffner — Sylvia Gebhardt—Wieland Beckert

Fraunhofer IKTSWinterbergstr. 28D-01277 Dresden, Germany

Andreas. [email protected]. deLutz.Seffner@ ikts.flig.deSylvia.Gebhardt@ ikts.flig.deWieland.Beckert@ ikts.flig.de

ABSTRACT: This paper describes advanced and cost-efficient manufacturing ofpiezoceramic fibers, piezoelectric composite materials made thereof and intendedapplications in the field of smart structures, health monitoring and diagnostics.

KEY WORDS: piezoceramic fibers, piezoelectric composites, ultrasound transducer, actuator•wrap

248 Repairing Structures using Composites Wraps

1. Introduction

Lightweight design has become very important in a multitude of industrialapplications mainly to reduce the effects of accelerated mass. However, lightweightstructures often suffer from vibrational sensitivity, tendency to buckling, and lowdamage tolerance. These issues beckon a need for adaptive mechanical propertiescoupled with the ability to monitor structural integrity and conduct diagnostics toensure safety and reliability.

A promising way to solve these problems is the use of multifunctional materials(Hagood et.all.,1993; Schmidt et.all., 1992). Critical deformations, accelerations orother physical quantities can be detected and measured by integrated sensors. Incombination with suitable real-time controllers these impacts can be reduced oreliminated throughout the use of structurally conformable, embedded or appliedactuators (Crawley et.all., 1989). The most well known and promising of thesesensing/actuating materials are piezoceramics. Much progress in the field of smartstructure technology is expected by using piezoceramic fibers and composites.

This paper gives a summary of our work on piezoceramic fibers (chapter 2),piezoelectric fiber composites made thereof (chapter 3) and the perspective on theirapplications (chapter 4).

2. Piezoceramic Fibers

Various methods of preparing piezoceramic fibers have been documented,among them the sol-gel process (Yoshikawa, et.all.,1994; Glaubitt, et. all., 1997),suspension extrusion (CeraNova, 2000) and suspension spinning process (Cass,1991;Taeger et.al, 1998).

In our case, the fiber production is based on a cellulose forming, suspension andspinning process, known as the LYOCELL-process (Taeger et. al, 1998). See Fig. 1.Essentially, commercial PZT powders are dispersed in a mixture consisting of acellulose - NMMO (N-methylmorpholin-N-oxid-monohydrate) solution. Thedispersion is pressed through a nozzle defining the fiber cross-section. The NMMOorganic is replaced by water in the coagulation bath accompanied by networkformation of the cellulose binder. As result, filaments of ceramic green fibers areobtained. Subsequently, the green fibers are dried and fired.

The LYOCELL-process allows for the production of a variety of advancedceramics on an industrial scale. Different filament shapes, hollow fibers as well asbi-component fibers have been manufactured successfully. Fig. 1 (right) shows anexample of sintered PZT fibers consisting of 250 mm in diameter and 150 mm inlength. Such fibers have been commercialized by the Smart Material Corp. (Florida,USA, www.smart-material.com). They are made from Type II and Type VI (U.S.Navy designation standards) piezoceramics and are offered in the diameter rangebetween 100-800 mm.


Figure 1. Schematic of green fiber preparation according to the LYOCELL process/courtesy TITK Rudolstadt, see (Teager et.al. 1998)/. After sintering, the fibers are straightand ready for composite fabrication.

The level of functional properties is 55-60 % of that of monolithic ceramics,which is attributed to the unusual high surface to volume ratio of fibers. Table 1shows typical fiber data. Improvements are expected by compositionalmodifications.

Table 1. Properties of Piezoceramic Fibers. (The relative values are deduced byrelating the fiber data to the monolithic ceramic data of the same composition.)

Piezoceramic

Navy Type VI

Navy Type II

Navy Type II

Fiber-0urn300

250

140

e33.f

T/e0 Relative e-value d33.f Relative d- value

% pC/N %2525

1300

1063

66

62

63

690

470

374

55

51

59


3. Piezoelectric Composites

3.1. Composites vs. Monolithic Piezoceramics

Composites of parallel aligned piezoelectric rods embedded in a passivepolymer matrix (see Fig. 3 left) show superior properties for ultrasonic transducerapplications as compared to monolithic piezoceramics plates of the same geometry.They combine a high coupling coefficient, low acoustic impedance, low mechanicalquality, minimized lateral mode coupling and an intermediate dielectric constant(Smith et .all., 1989). The quasistatic as well as dynamic properties are anisotropic,which allows for decoupling of the in-plane properties from those in the normaldirection.

Composite laminates that utilize piezoelectric fibers for structural control (seeFig. 3, right) have been under rapid development in the recent past as they offermany advantages over traditional piezoceramic actuators (Williams, 2000):

They are more robust than brittle monolithic piezoelectric materials.They can be made to conform to the curved surfaces of realistic applications.They can be added along with conventional fiber-reinforced laminate.They exhibit in-plane actuation anisotropy, which affords them the ability toapply both bending moments and twisting motions.They exhibit in-plane sensing anisotropy.The in-plane arrangement of fibers and the use of interdigitated electrodes allowmuch higher forces or displacements to develop by capitalizing on the strongerlongitudinal (d33 constant) piezoelectric effect.

Figure 3. Schematic of 1-3 piezoelectric composites as used for acoustic transduction (left)and for structural control (right). The ceramic fibers are embedded in a polymer matrix (notshown). The acoustic transducers are terminated by layer electrodes (left, not shown), whilstinterdigital electrodes (IDE) are applied on the surface of the composite patches (right).


3.2 Composites by "Arrange & Fill"

Piezoceramic-polymer 1-3 composites (using our methodology) are prepared byepoxy infiltration of fiber bundles and dicing of the cube-shaped blocksperpendicular as well as parallel to the direction of the fibers. In the first case,strain-stress sensing piezoelectric sheets are obtained, usable in the quasistatic aswell as ultrasound frequency domain. The sensitivity is primarily in the normaldirection of the sheet, thus decoupling of transversal mechanical stimulation isachieved. Sensing elements for Health Monitoring & Diagnostics as well as Non-Destructive Testing are seen to be the primary applications.

In the second case, flexible piezoelectric sheets with in-plane sensing oractuation anisotropy are obtained. In this case, interdigitated electrodes (IDE) areapplied on the sheet surface. This approach allows for the preparation of large sizeflexible wraps, serving as the actuator and/or sensor part in smart structures.

As shown by optical microscopy at low fiber contents < 25 vol % the distributionis statistical, whereas a more regular arrangement occurs at higher phase volumefractions, e.g. > 50 vol. %.

Figure 4. Cross section of 1-3 piezoflber composites with 25 vol% (left), 50 vol% (middle)and 65 vol% (righ) . The PZT fibers are spaced randomly.

4. Intended Applications

4.1 Ultrasonic Transduction

The properties of 1-3 composites with random element spacing, as preparedusing our technology, correspond to those expected theoretically. The thicknessresonance frequency is defined by the frequency constant of the material of about


1550 Hz m and the thickness of the sample. So far, samples in the wide resonancefrequency from 50 kHz to 2 MHz have been prepared and tested by diversecustomers. As seen in Fig. 5, spurious modes around the thickness vibration arecompletely suppressed.

Figure 5. Impedance /Z/ and phase angle theta as function of frequency measured on 1-3fiber composites with random element spacing (fibers : PZT Navy Type II, 2 50 mm, composite:65 vol% , sample size 20mm x 20mm) showing only the thickness vibration mode at 1,5 MHz.No spurious modes occur. (Fiber composites with various characteristics are commerciallyavailable by the Smart Material Corp., Florida, USA)

4.2 Sensor Patches: Coupling in Normal Direction

The sensing capability of thin patches of 1 -3 composites, as sketched in Fig. 3(left) with thickness of 200 - 300 mm have been investigated using a testingmachine. See Fig. 6.

Figure 6. Plot of testing machine after 108 cycles applied on 5 mm x 5 mm samples at 35°C1 - Charge yield from piezo-composite; 2 - Charge yield from PVDF, amplification 8 x3 - Applied stress, amplitude 10 Mpa; 4- Strain measured by a Laser system/courtesy Dr. Brunner, Fraunhofer - ISC/


The yield of charge was found to be 8 times the value of that for a conventionalPVDF sensor. The signal turned out to be very stable under the test conditions.Flexible sensor patches of large size are available, see Fig 7.

Figure 7. Flexible sensor patch with normal load sensitivity fixed on a glass tube.

4.3 Sensor / Actuator Wraps: In-plane Coupling

There is a general interest in fiber composites for actuation (Janas et.all,1998,Wilkie et. all, 2000, Schonecker et. all, 2000). We succeeded in preparing flexibleactuating/sensing components by slicing 1-3 composite blocks into thin layers. Thestructure of which corresponds to that sketched in Fig. 3 (right) with a tolerablemisalignment of the single fibers. IDEs serve for field coupling. If the IDEs areapplied on one side only, the component works like a bending actuator. See fig. 8.

Full characterization and improvement of the structural design is still underinvestigation. The scope of design is determined by geometrical factors such as fiberdiameter, sample thickness, straightness of fibers, finger electrode width/spacing(Beckert et. all., 2001), and selection of the constituent phases.

The fiber composites are expected to show improved robustness, flexibility,damage tolerance and handling capability.

Figure 8. Bending of the fiber composite along the middle axis depending on the drivingvoltage (parameter).


5. Conclusion

Piezoceramic fibers allow for a unique and cost-efficient piezo-transducertechnology. Ultrasound transduction materials with suppressed spurious modes canbe prepared for working frequencies between 50 kHz - 4 MHz. The acousticimpedance can be adjusted to the needs of sonar applications, non-destructive testingand biomedical diagnosis. Flexible piezoelectric components with sensing andactuating anisotropy have been developed. They are expected to find widespreadapplications in smart structures. R&D at Fraunhofer-IKTS is still ongoing. Productsare being commercialized in co-operation with Smart Material Corp., Florida, USA.Prototype samples are available for evaluation, (www.smart-material.com).

Acknowledgements

The authors would like to thank Thomas Daue, John Wright, Fumio Aikawa,Dieter Vorbach and Giinter Helke for valuable discussions and assistance.

12. Bibliography/References

Hagood N.W., Bent A.A., "Development of Piezoelectric Fiber Composites for StructuralActuation", Proc. 43th AIAA ASME, Adaptive Structures Forum, April 19-22, 1993, LaJolla, CA

Schmidt W., Boiler C.,"Smart Structures - A Technology for Next Generation Aircraft", 15 thMeeting AGARD - Structure and Materials Panel, Lindau, 5.-7.10.1992

Crawley E.F., Anderson E.H., "Detailed Models of Piezoceramic Actuation of Beams", 1989,AIAA Journal

Yoshikawa S., Selvaraj U., Moses P., Jiang Q., Shrout T.. "Pb(Zr,Ti)O3 (PZT) Fibers-Fabrication and Properties"', Ferroelectrics 154 (1994) 325-330.

Glaubitt W., Watzka W., Scholz H., Sporn D., "Sol-gel processing of functional and structuralceramic oxide fibers"; J. Sol-Gel Sci. Technol. 8(1997) 29-33.

CeraNova Commercial Brochure, 2000 CeraNova Corp.

Cass. R. B., "Fabrication of Continuous Ceramic fiber by the Viscous Suspension SpinningProcess": Am. Ceram. Soc. Bull. 70(1991) 3,424-29.

Teager E., Berghof K., Maron R., Meister F., Michels Ch., Vorbach D., " Lyocell productswith build-in functional properties", Chem. Fibers Int., vol. 48, 1998, p. 32-35.


Smith W.A., Shaulov A., Auld B., "Design of Piezocomposites for Ultrasonic Transducers",Ferroelectrics, 91 (1989), pp. 155-162

Williams R. Brett, "An Introduction to Composite Materials with Active PiezoelectricFibers", Lecture Virginia Tech, 2000

Janos, B. Z. and Hagood, N. W., "Overview of Active Fiber Composites Technologies,"Proceedings of the 6th International Conference on New Actuators - ACTUATOR 98,June 98, Bremen, Germany.

Wilkie, W. K., Bryant, G. R., High, J. W. et al., "Low-Cost Piezocomposite Actuator forStructural Control Applications," Proceedings, SPIE 7th Annual International Symposiumon Smart Structures and Materials, Newport Beach, CA, March 5-9, 2000.

Schonecker A., Sporn D., Watzka W., Seffner L., Wierach P., Pannkoke K., "High-Performance Piezoelectric Thin Fibers and Sheets as Functional Components for SmartMaterials", Proceedings, SPIE 7th Annual International Symposium on Smart Structuresand Materials, Newport Beach, CA, March 5-9, 2000.

Beckert W., Kreher W. S., "Modelling Piezoelectric Modules with Interdigitated Structures"Proceedings of 11th International Workshop for Computational Mechanics and ComputerAided Design of Materials (IWCMM 11), Freiberg (Germany), September 2001, to bepublished in Computational Materials Science

Application of Metal Core-PiezoelectricFiber

Embedment in CFRP

Hiroshi Sato — Yoshiro Shimojo — Tadashi Sekiya


[email protected]

ABSTRACT: Research on piezoelectric fibers was started in the Active Materials and StructuresLaboratory at MIT in 1992. Now, these fibers are used in commercial products, such as skiboards and tennis rackets for vibration suppression. However, these fibers have somedisadvantages. For example, interdigitated electrodes are necessary for the use as sensorsand actuators. Furthermore, they are fragile because of the ceramics. These problems weresolved using metal core piezoelectric fibers manufactured by a hydrothermal method. Thefibers obtained are difficult to be broken and require no electrodes. Using the novel fiber anew smart board was developed.

KEY WORDS: metal core piezoelectric fiber, CFRP, smart board, sensor, actuator


1. Introduction

Piezoelectric material has been used for sensor and actuator. Recently muchattention is being paid to the application of piezoelectric material for structurehealth monitoring and vibration control on embedding into composite materialssuch as CFRP and GFRP. In the composite including piezoelectric material, it isimportant to minimize the harm to the mechanical performance of composite. Asone solution, the Active Materials and Structures Laboratory at MIT proposed touse piezoelectric material in fiber shape (Bent et al, 1993). They say that theirfiber is strong, conformable, and therefore can be used to some commercialproducts, such as ski board and tennis racket for suppressing the vibration.However, their fiber has disadvantages, as interdigitated electrodes are necessaryfor the use as sensor and actuator and the fragility is not completely solved. Inorder to solve these problems, we propose piezoelectric fiber with metal core,which is fabricated by the hydrothermal method. The advantages of ourpiezoelectric fiber are as follows:

(1) No need of electrodes.

Generally, the piezoelectric material needs one couple of electrodes in using assensor and actuator. However, in our piezoelectricity fiber, the electrode is notrequired, since the metal core in the fiber can be used as one electrode and CFRPitself becomes ground electrode because of the high electric conductivity of thecarbon fiber.

(2) Difficult to be broken

Although piezoelectric ceramics such as PZT are fragile, the fragility can beovercome by the metal core.

(3) High resistance to the noise from the outside.

The sensitivity of the sensor is evaluated by S/N ratio. Therefore, it is importanthow to increase an output signal from the sensor and how to decrease a noise fromthe surroundings. Our fiber is embedded in CFRP composite with high electricalconductivity. Therefore, the CFRP composite easily cuts off the noise from theoutside, and it is possible to enhance the signal from the sensor.

(4) Decrease of the thermal stress

Sol gel method and extrusion method are considered as the other ways to producethe piezoelectric fiber including metal core. However, it is necessary to sinter at hightemperatures as high as 1000°C to obtain the final product. At that time, ceramicsmay be broken, because of the difference in the thermal expansion coefficientbetween metal core and piezoelectric ceramics. Using the hydrothermal method, theinfluence of the thermal expansion can be reduced, since the hydrothermaltemperature is 150°C or less. Furthermore, the polarization processing isunnecessary.


(5) Low cost

Manufacturing cost is a problem. The hydrothermal method enables to producefibers in a large number at one time. Then, it is possible to utilize for sensor andactuator only by embedding in the CFRP composite.

In this paper, we make piezoelectric fibers with metal core using a hydrothermalmethod and develop the fiber-embedded CFRP smart board. In addition, it is shownthat this board can generate the vibration and detect the vibration.

2. Metal core-piezoelectric fiber

Piezoelectric PZT fibers with metal core were fabricated by a hydrothermalmethod same as reported by Shimomura in Tokyo Institute Technology (Shimomuraet al., 1991). And now, micro ultrasonic motor, excitation type tactile sensor andgyroscope are developed as the application example (Kurosawa et al., 1999; Sato etal, 1999). This method has many advantages further than Sol-Gel, sputtering andCVD techniques as follows:

(1) PZT thin film (about 5 to 50 mm) can be fabricated on the three-dimensionaltitanium structure.

(2) The crystalline film is deposited at temperatures as below as 1500.

(3) The resultant film needs no polarization process.

(4) The thickness of PZT layer can be controlled by repeating the crystal growthprocess.

In the hydrothermal process, PZT precipitates according to the following reaction,

This method consists of two processes, that is, nucleation process and crystalgrowth process. In the nucleation process, titanium substrate was hydrothermal-treated in the mixed solution of zirconium oxychloride, lead nitrate and potassiumhydroxide in an autoclave. The reaction condition is 140°C for about 24h. Ions Pb2+

and Zr4+ are supplied from the solution and titanium substrate itself is Ti4+ source.Thus, PZT nuclei are formed on the titanium substrate surface.

After the nucleation process, the titanium substrate was subjected to the crystalgrowth process in order to increase the thickness of PZT layer. In this process,titanium tetrachloride was added to the above solution as further Ti4+ source, andreaction was made at 120°C for about 24h. Then PZT crystals are subsequentlygrown on the nuclei. Figure 5 shows a SEM image after the crystal growth process.It can be seen that PZT crystal grains of about 5 to 10 um in size are grown on thetitanium substrate.


Figure 1. SEM image of PZT thin film

3. Application to the smart board

By embedding sensor and actuator in the composite structure, and as a result bygiving health monitoring and vibration suppression functions, it becomes possiblethat the structure of reliability is increased and the span of life is extended. Thattime, it is necessary to consider the shape of sensor and actuator so as to minimizeharmful influence on the mechanical performance of the composite material. Wereduce the influence by embedding the piezoelectric fiber in the CFRP compositealong the direction same as that of the carbon fibers. We made a cantilever structurewith piezoelectric fibers embedded on CFRP composite, as shown in Figure 2.Piezoelectric fibers are put on the six layers-stacking of CFRP prepreg. Then,prepreg are pressed under 0.3MPa at for 135D for 2 hours by using a hot press, andthe CFRP composite[02 / 902 / 02 ] in which the piezoelectric fibers were embeddedwas produced. This cantilever is 70mm in length, 30mm in width and 0.7mm inthickness.

3.1. Use as actuator

The piezoelectric material needs two electrodes (upper electrode and lowerelectrode), when used as sensor and actuator. However, in our piezoelectricity fiber,the electrode is not required. The metal core in the fiber can be used as oneelectrode, and CFRP composite plays role of ground electrode because of the highelectric conductivity of the carbon fiber.


In this experiment, six piezoelectric fibers were embedded in the cantileverstructure. 50V AC voltage was applied between six titanium cores and CFRPcomposite, then the piezoelectric fibers were elongated or shrank due to theconverse piezoelectric effect. Finally, CFRP board was bent by deformation of thepiezoelectric fibers. We measured this bending displacement of the beam tip using alaser displacement meter as shown in Figure 3.

Figure 4 shows relationship between input frequency and vibration displacementof the beam end. It can be seen from this figure that the cantilever vibrates in therange of about l0nm to lmm having a resonant point at about 180Hz.

Figure 2. Fabrication process of the smart board

Figure 3. Block diagram of experimental systemfor examination of actuator function


Figure 4. Relationship between applied frequency andvibration displacement of the beam end

3.2. Use as sensor

Next we applied this board as a vibration sensor. In this experiment,electromagnetic vibrator was put on the tip of the cantilever to make referencevibration. Piezoelectric fibers on the CFRP board are shrank or elongated as theboard is bent. Then an electric charge was generated from the piezoelectric fiber bythe direct piezoelectric effect. This electric charge was detected by using Lock inamplifier, as shown in Figure 5.

Figure 6 shows relationship between applied vibration and output voltage as afunction of frequency. The solid line indicates the displacement of the tip ofcantilever measured by laser displacement meter and the dotted line means an outputvoltage came from our piezoelectric fiber. From this figure, it is proved that theoutput voltage from the fiber is almost proportional to the magnitude of thereference vibration.


Figure 5. Block diagram of experimental systemfor examination of sensor function

Figure 6. Relationship between reference vibrationand output voltage of the piezoelectric fiber

4.Conclusions

In this paper, we developed piezoelectric fiber with metal core wire andproposed new smart board incorporated this piezoelectric fiber on the surface of theCFRP composite. It was shown that these complex fibers could be used as sensorand actuator in the CFRP board. As further smart application of this piezoelectricfiber, it is expected to extend to construct self-sensing, health monitoring andvibration control systems. In the near future, it may be possible to produce linearsensor network using this fiber.


References

Bent, A., Hagood N and Rodgers J., "Anisotropic Actuation with Piezoelectric FiberComposites", Proceedings of the DGLR Conference, Germany, 1993.

Kurosawa K. and Higuchi T., "A Cylindrical Shaped Micro Ultrasonic Motor Utilizing PZTThin Film", Proceedings of the 10th International Conference on Solid-State Sensors andActuators (Transducers'99), 1999, p. 1744-1747.

Sato H., Fukuda T., Arai F and Itoigawa K, "Parallel Beam Gyroscope", Proceedings of the10th International Conference on Sol id-State Sensors and Actuators(Transducers'99),1999, p. 1586-1589.

Shimomura K., Tsurumi T., Ohba, Y and Daimon M., "Preparation of Lead ZirconateTitanate Thin Film by Hydrothermal Method", JpnJ.AppI.phys., Vol. 30, 1991, p. 2174-2177.

Part III:

Process Improvement

Cure monitoring of composites usingmultidetection technique

Emmanuel Chailleux* — Michelle Salvia* — Nicole Jaffrezic-Renault* — Yves Jayet** — Abderrahim Maazouz*** — GerardSeytre**** — Ivan Kasik*****

*IFOS, UMR CNRS 5621, Ecole Centrale de Lyon36 avenue Guy de Collongue, 69131 Ecully, Francee-mail: [email protected]

**GEMPPM, UMR CNRS 5510, INSA deLyon20 avenue A.Einstein, 69621 Villeurbanne, France

***LMM, UMR CNRS 5627, INSA deLyon20 avenue A.Einstein, 69621 Villeurbanne, France

****LMPB, UMR CNRS 5627, Universite Claude Bernard,43 boulevard du 11 novembre, 69622 Villeurbanne, France

*****IREE, Academy of Sciences of the Czech RepublicChaberska 57182 51 Prague, Czech Republic

ABSTRACT : Since the last decade, fibre reinforced plastics have been increasingly used ascomponents in engineering structures. Ageing, load-transfer, and off-axis behaviour ofcomposites are directly dominated by the viscoelastic matrix properties linked to the cureprocess. So there is a growing need for sensors, which provide real-time, in situ monitoringof the manufacturing process. This study proposes to follow the cure mechanism of an epoxy-amine resin simultaneously using three sensors embedded in the material: a fibre-opticsensor (refractive index), a piezoelectric element and a dielectric sensor.KEY WORDS: cure monitoring, optical fibre, dielectric, ultrasound, thermoset.


1. Introduction

High performance composites have been used extensively in high-tech areas,such as aerospace and automobile industries etc. Numerous primary structural partsare made with these materials. In particular, epoxy resin reinforced with continuousglass fibre is a system with good mechanical properties and low density. Thereinforcing fibre dominates largely the mechanical behaviour when the compositesare loaded in the fibre direction. However, ageing, load-transfer, and consequentlycreep and off-axis loading are directly dominated by the viscoelastic matrixproperties (epoxy resin) linked to the cure process.

Three sensors, good candidates to provide in situ evaluation of the thermosetmatrix cure process, have been developed in previous work: fibre-optic sensors(Chailleux et al, 2001), piezoelectric sensors (Jayet et al., 1998) and microdielectricmeasurements (Pichaud et al., 1999).

This study proposes to monitor the cure of an epoxy-mine system, using thesesensors simultaneously on the same sample. The multidetection monitoring will beperformed in terms of refractive index, viscoelastic properties, and conductivity.This multidetection technique allows these parameters to be determined in the sameexperimental conditions. This point is particularly important because the epoxy-amine reaction is exothermic, so kinetic parameters depend strongly on the samplegeometry and quantity. Comparing the results should enable us to understand theinformation provided by the in situ sensors for each step of the epoxy-amine curemechanism. Particular attention will be given to the changing physical properties,from the liquid to the solid state.

2. Theoretical part

2.1. Cure of epoxy-amine system

The amine-cured epoxy system gives a three-dimensional macromolecularnetwork synthesised by the polyaddition of polyfunctional molecules. The finalmorphology of this three-dimensional network, which determines the properties ofthe material, depends on this transformation. During the thermoset resin cure, thereis an interaction between the chemical kinetics and the changing physical properties,which may involve an incomplete degree of conversion of the system. Thisphenomenon is particularly important because the glass transition temperature is afunction of the degree of conversion. Di Benedetto's approach (Di Benedetto 87)assumes that this relation is independent of the cure temperature:

Process improvement 269

where Tg0 ,Tgoo are the glass transition temperatures of, respectively, the unreactedresin mixture and the fully cured resin (l, is an adjustable parameter). This relationhas been compared with success to experimental data for an epoxy-amine system(Pichaud et al, 1999). The chemical transformation involves first the epoxy groupswith the primary amine to give secondary amine. The secondary amine reacts withthe epoxy group to give tertiary amine. These two reactions are competitive.Moreover, two phases may appear during the reaction according to the curetemperature: gelation and/or vitrification. Gelation is the liquid to rubber transition,which occurs when the system reaches a certain degree of conversion. This degreeof conversion corresponds to the time when an infinite network is formed. The gelpoint can be determined with fraction gel experiments or dynamic mechanicalspectrometry. This transition is not frequency dependent. Vitrification is rubber toglass transition, which occurs when the glass transition increases to the temperatureof cure. This transition is frequency dependent. The occurrences of these transitionsaccording to the cure temperature have been reported by Enns and Gillham innumerous works (Enns and Gillham, 1983,1983b).

2.2. Refractive index

The refractive index measurement is carried out using an embedded fibre opticsensor (Figure 1). The principle of this sensor is based on measurement of angulardistribution of light transmitted through the optical fibre (Figure 2). The differencebetween the cladding and core refractive indices is directly responsible for the lightguiding properties of optical fibres. So, by partially removing the cladding andimmersing the stripped region in an external medium it is possible to monitor itsrefractive index variation. However, the refractive index of the new medium has toobey the relation: ncore>nmedium>ncladding in such a way that guiding conditions andexternal medium sensitivity will hold. The optical fibre has to be selected inaccordance with the tested material. A theoretical model allows the refractive indexof the surrounding medium to be determined by fitting the angular distribution ofthe transmitted light power data. The model is based on the following parameters:refractive indices of the core, claddings, and external media, core and claddinglength, then diameter of the core. The coating media (epoxy resin and silicone inthis work) are considered to be imperfect dielectrics, so their refractive indices haveimaginary parts related to optical loss. The silica core is considered to be lossless.Moreover, due to a relatively large core diameter (about 300 mm) it is possible touse theories of geometrical optics. To monitor the dynamic reaction ofpolymerisation a fixed angle of incidence is chosen. The sensitivity and the abilityof this optical sensor have been reported hi a previous work (Chailleux et al, 2001).


Figure 1. Schematic of the fibre sensor detection system

Erreur! Signet non defini.

Figure 2. Angular distribution of the transmitted light power for the fibre-opticsensor immersed in cured and uncured epoxy resin

In order to understand the optical response of the epoxy system during thereaction, it is necessary to study how the chemical and physical structure contributeto the refractive index. The Lorentz-Lorentz formula links the refractive index (n) tothe molecular weight (M), the molar refractivity (R) and the density (p):

The molar refractivity is independent of temperature or physical state and, forlarge number of compounds it is additive for the bonds present in the molecule(Bauer et al., 1960). Knowing that the three-dimensional network is synthesised bythe polyaddition of polyfunctional molecules and, assuming that chemicaltransformation during the reaction is insignificant in terms of molar refractivity andmolecular weight, the Lorentz-Lorentz formula during cure can be written asfollows:


where k(t) is a kinetic dependent parameter. This assumption implies the refractiveindex variation is only due to density during cure. Moreover, it has been verifiedthat the relation between n and p can be considered linear in the refractive index anddensity range of epoxy-mine during cure. Numerous works report contradictoryresults about density and degree of cure relationships for epoxy-amine systems.Cizmecioglu et al (1986) assumes the increase in density (measured at roomtemperature) with conversion is due to the cross-linking points, which reduce thefree volume of the resin system. Cizmecioglu finds a linear relation between densityand conversion, independent of cure temperature. It is to be noted that the epoxysystem (TGDDM-DDS), used in this case, was in a non-stoichiometric ratio([epoxy]/[amine]=2). On the other hand, Enns et al. (1983b) show that densitydecreases as conversion extends (whereas glass transition increases) hi the case of astoichiometric mixture of Epon828 cured with DDS. This result is explained interms of the non-equilibrium nature of the glassy state. From these observations, itseems the relationships between density (and so refractive index) and extent ofreaction may not be easy to predict. Experimentally, Afromowitz and Lam (1990)measured the refractive index according to the extent of reaction for an Epon828cured with 14 phr-m-phenylenediamine. They find that the refractive index growslinearly with the extent of reaction until the system reaches a critical degree ofconversion for Tcure = 90°C and 130°C and a perfect linear relation for Tcure = 60°C.

2.3.yiscoelasticproperties by ultrasound

The study of ultrasonic wave propagation was used for long time to monitor thecure of thermoset resin (Sofer and Hauser, 1952). The technique used in this work isbased on the measurement of the electrical impedance of piezoelectric ceramic. Inthis work, the electrical impedance is measured in the frequency range of theceramic thickness vibration mode (2.2 MHz). A one-dimensional approach issufficient to model the electrical impedance according to the frequency in relationwith the axial vibration mode. An analytical expression is obtained by consideringthe fundamental relations of piezoelectricity and the wave propagation equation fora harmonic longitudinal excitation in a viscoelastic material. The geometry of theone-dimensional problem is shown in Figure 3. The validity of this model has beenpresented in previous works (Perrissin-Faber and Jayet, 1994, Jayet et al, 1998).


Figure 3. Geometry of the one-dimensional model, p: density, V: longitudinal soundvelocity, Att: ultrasonic attenuation, h33: piezoelectric constant in the ZZ' direction,b33: dielectric constant.

The unknown ceramic parameters are determined by analysing the response ofthe electrical impedance when the element is immersed in a medium of knownultrasonic properties. The experiment is then performed on the epoxy system fromthe initial liquid mixture to the solid state. Figure 4 shows the electrical impedance,in the frequency range of the ceramic thickness mode of vibration, at the end of theepoxy cure. The model (continuous line) allows the longitudinal sound velocity (V|)and the attenuation (a1) to be determined. An optimisation algorithm, based on asimplex optimisation method, is used to fit the experimental data (circle).

The relations between ultrasonic wave propagation and mechanical properties inviscoelastic medium are well known. The wave equations for an harmoniclongitudinal excitation in such a material give the following relationships (withreasonable approximation in the ultrasonic frequency range):

where M' and M" are respectively the storage and loss longitudinal modulus, p is thedensity and w is the radian frequency. The complex modulus (M*) determined fromvelocity and attenuation is the linear combination of the bulk and shear modulus:M* = K* + 4/3 G*. During cure, the increase of the molecular weight involves anincrease of the mechanical properties from the liquid to the glassy state. Thecomplex modulus is well known as an interesting parameter to study the elasticproperties as well as the relaxation spectra of thermoset resins during cure.Numerous works report dynamic mechanical experiments to determine the gelationand vitrification transitions. Nevertheless, the viscoelastic response must beexplained carefully due to the high frequency used (2.2 MHz). Morel (Morel et al,1989) assumes that (3 transition, usually determined under 0°C at low frequency, ishigher than room temperature at ultrasonic frequency.


Figure 4. Module of the ceramic electrical impedance when the element isimmersed in cured epoxy (circle). Continuous line is drawn using the theoreticalmodel with the following parameters: V=2400m/s, Att=8.10-6Np.m-1.Hz-1

2.4.Dielectric behaviour

When dielectric material is put into an alternating electric field, conduction andpolarisation phenomena take place in the material. The knowledge of the phase anglebetween input voltage and current delivered through the material and of the currentamplitude allows the sample complex permittivity (e* = E' + j e") to be determined.The parameter chosen for study is conductivity (a). Conductivity is deduced from thedielectric loss factor (e") and the frequency of the measurement (w).

with EO being the permittivity of the free space. During cure, conductivity variationsare firstly due to ionic mobility and, secondly to dipolar motion. The best conditionsto measure conductivity due to ionic transport are low frequency as well as lowviscosity. On the other hand, the dipolar relaxation times are responsible for theconductivity when the measurement is made at high frequency and when theviscosity of the system reaches a critical level. It is to be noted that dipolar responseis frequency dependent whereas ionic response is frequency independent. During


the epoxy cure, the ionic impurities are responsible for the ionic conductivity untilthe viscosity reaches certain value (Pichaud et al., 1999).

The dielectric response will be then due to dipolar motion in close relation withviscosity. In this study, dielectric measurement is performed using Micrometeumetric system III apparatus. This device generates a sinusoidal signal that istransmitted to sensor electrodes. The electrode configuration is an interdigited combpattern. The software linked to this device provides complex permittivity andconductivity according to the frequency. Thus, it is possible to obtain the dielectricrelaxation spectra in relation to the dipolar motion.

3. Experimental part

3.1. Experimental setup

The epoxy resin is commercial DGEBA (LY 556 resin from Ciba) cured withIPD (IsoPhorone Diamine from Aldrich). Resin and hardener are mixed instoichiometric ratio. Glass transitions of the initial and fully cured resin are:Tg0 = -37°C and Tgoo = 155°C. The gelation limiting temperature is Tgel =32°C.

In order to have gelation and vitrification successively, the measurement shouldbe performed between 32°C and 155°C .The refractive index of the initial mixturemeasured at room temperature with an Abbe refractometer is 1.555. In order tomonitor the reaction simultaneously with the three sensors the resin is cured in aninstrument-equipped mould (Figure 5). The mould enables the insertion of anoptical fibre and the immersion of the dielectric sensor as well as the piezoelectricceramic in the resin. The mould is pre-heated to the test temperature. The mixture isthen poured while the responses of the sensors are recorded (Figure 6).

3.2. Temperature effect

hi the first stage, the output signals of the sensors reflect the competition betweentemperature and reaction. The mechanical characteristics and refractive index fallcontinuously while the conductivity increases. These phenomena must be attributed tothe density and viscosity decrease as resin temperature increase, In this polymericliquid state M' is equivalent to the bulk modulus K' since G' << K' (Ferry 1990) and soM' depends on the free volume. On the other hand, G" is not necessarily negligiblecompared to K". Thus, the loss modulus M" depends on the resin viscosity. Ferrygives the following relation: M"=w(hv+4/3h) where hv and h' are the bulk (orvolume) viscosity and standard viscosity. According to equation [3], density isresponsible for the refractive index decrease. Also, the conductivity variation can beexplained, in this liquid state, since ionic mobility rises while viscosity decreases


(Pichaud et al, 1999). It is to be noted that the small variation of density (2%calculated from the refractive index) induced by the temperature increase, leads to anon-negligible effect on the bulk modulus (24%) as well as on conductivity (36 %).This effect is particularly important because it depends on the storage conditions (Rathet al, 2000) (temperature, moisture), since softening of the unreacted resin is linked tothe macromolecular initial state: molecular weight, cross-linked density andconsequently initial glass transition temperature. The earlier change in conductivity,compared to the output signals of the other sensors, can be explained both by thematerial zone analysed and by the different sensor locations. As previously mentioned,the optical fibre is embedded in the middle of the sample and both dielectric andultrasonic sensors are located at the mould/resin interface where the temperaturethreshold is reached earlier than in the bulk of the material. The signal output of thedielectric sensor is due to phenomena occurring at the sensor/resin interface (like theoptical fibre) while the US response is linked to bulk properties.

Figure 5.. Schematic of the instrument-equipped mould


Figure 6. Multidetection monitoring -TCJirc =90°C

3.3. Cross-link effect

Before the resin reaches the test temperature, the output signals of the sensorsbegin to reflect the cross-linking reaction. The mechanical characteristics andrefractive index rise, while conductivity decreases. Cross-linked node formationleads to a decrease in free volume and an increase in viscosity. The refractive index,which is representative of this free volume, shows sigmoidal variation. At first, M1

and tan8 rise slowly, then increase dramatically, respectively when n reaches 20%and 50% of its asymptotic value. Then, tan 8 presents two peaks while M1 showssigmoid variation. This M' variation denotes the occurrence of the shear elasticresponse of the resin (G'). At the same time, the decrease of conductivity, firstlinked to the formation of microgel resulting in a decrease of the ionic mobility,begins to be the consequence of dipolar relaxation times. In fact, conductivitycurves pass through maxima in dependence on frequency tests. These maxima arelinked to vitrification (Wang et al., 1994). Taking into account the frequencydependence of this relaxation phenomenon, vitrification times at 2.2 MHz aredetermined from the dielectric measurements (Figure 7).


Figure 7. Vitrification time at 2.2 MHz extrapolated from dielectric measurement

From this result, the vitrification time measured at 2.2 MHz seems to be close tothe first tan5 maximum. In order to estimate gelation time, viscosity h' has beenmeasured with a dynamic mechanical analyser (DMA) between 1.58 rad/s and 100rad/s. Figure 8 shows tand compared to h' during cure for two temperatures: 70°Cand 90°C.

The h' dramatic increase, which appears after the first tans maximum, showsthat the system is close to gelation. Vitrification time, measured at 2.2 MHz, andgelation, appear in the time range for Tcure =70°C and 90°C. From this result, thesecond peak on tan 6 curves could be explained by an interaction between gelationand vitrification. This point must be verified in future works. It is to be noted thatboth conductivity and refractive index are not linked to any particular event duringthe dramatic increase in viscosity, and so to the gelation phenomenon. At the end ofpolymerisation, the variations of the dielectric, mechanical and optical parametersreflect the low rate of the reaction since the cure mechanism is now controlled bythe molecular diffusivity. It is interesting to note that M' is more sensitive at the endof the reaction than the refractive index and conductivity (Figure 6).


Figure 8. Loss factor and Viscosity versus cure time for Tcure =70°C (continuousline) and 90°C (dotted line)

4. Conclusion

An amine-epoxy cure process was evaluated in terms of refractive index,viscoelastic properties and conductivity. These properties were measuredsimultaneously in the same experimental conditions, using three sensors able toprovide in situ monitoring the cure process of composites. The difference in kineticscould be attributed to the different sensing location. In fact, the dielectric sensor andoptical fibre provide information at the sensor/resin interface while the UStechnique informs about bulk behaviour. The temperature effect, on the introductionof the resin, allows the relationships between density (from refractive index),viscoelastic properties and conductivity of the unreacted resin to be determined, andcould be linked to the initial mixture quality. During the isothermal cross-linkedreaction, an increase in density is measured, linked to the chemical kinetics. Theelastic modulus determined from the US measurement rises firstly due to the bulkmodulus and, secondly increases dramatically when the shear modulus becomesnon-negligible. The occurrence of the shear elastic response is linked to the firstrelaxation phenomenon on the loss coefficient (tans). This relaxation is attributed tovitrification from the dipolar relaxation observed on the conductivity curves usingthe frequency-time dependence criteria. It appears that the vitrification, determinedfrom US measurements at 2.2 MHz, occurs close to the gelation transition for Tcure

=70°C and 90°C. These results show that multidetection monitoring provides a


powerful tool for understanding the changes in physical properties of thermosetduring cure.

Acknowledgements

The authors would like to thank Vlastimil Matejec, (Institute of RadioEngineering and Electronics, Academy of Sciences of the Czech Republic), Jean-Michel Vemet (IFOS laboratory, Ecole Centrale de Lyon), Lucien Deville(GEMPPM, INSA de Lyon) and Susan Goodacre (Ecole Centrale de Lyon) for theirassistance.

References

Afromowitz M.A., Lam K.Y., "The optical properties of curing epoxies and applications tothe fiber-optic epoxy cure sensor", Sensors and Actuators, A21-A23, 1990, p. 135-139.

Bauer N., Fajans K., Lewin S., Physical methods of organic chemistry vol.1 -Part II,ch.XVIII, Interscience publishers, 1960, p.l 162-1169.

Chailleux E., Salvia M., Jaffrezic-Renault N., Matejec V., Kasic I., "In situ study of theepoxy cure process using a fiber optic sensor", Smart Materials and Structures, vol.10,2001,ppl-9.

Cizmecioglu M., Gupta A., Fedors F., "Influence of cure conditions on glass transitiontemperature and density of an epoxy resin", Journal of applied polymer science, vol.32,1986, p. 6177-6190.

Di Benedetto T., "Prediction of the glass temperature of polymers: a model based on theprinciple of corresponding state", Journal of Polymer Science, vol. 25, 1987, p. 1949-1969.

Enns J., Gillham J., "Effect of the extent of cure on the modulus, glass transition.waterabsorption, and density of an amine-cured epoxy", Journal of applied polymer science,vol.28, 1983, p. 2831-2846.

Enns J., Gillham J., "Time-temperature-transformation (ttt) cure diagram: modeling the curebehaviour of thermoset", Journal of applied polymer science, vol. 28, 1983, p. 2567-2591.

Ferry J.D., Viscoelastic properties of polymer ch.18, John Wiley and Sons, 1980, p.562-568.

Jayet Y., Baboux J., Guy P., "The piezoelectric implant method:implementation and practicalapplication, Proceedings of 4th ESSM and 2nd MMR Conference, Harrogate IOPPublishing, 1998, p. 505-510.

Morel E., Bellenger V., Bocquet M., Verdu J., "Structure-properties relationships for denselycross-linked epoxide-amine systems based on epoxide or amine mixtures", Journal ofMaterials Science, Vol. 24, 1989, p. 69-75.


Perrissin-Faber I., Jayet Y., "Simulated and experimental study of the electric impedance of apiezoelectric element in a viscoelastic medium", Ultrasonic, vol. 32, 1994, p. 107-112.

Pichaud S., Deuteutre X., Fit A., Stephan F., Maazouz A., Pascault J.P., "Chemorheologicaland dielecric study of epoxy-mine for processing control", Polymer international vol. 48,1999,p 1205-1218.

Rath M., Doring J., Stark W., Hinrichsen G., "Process monitoring of moulding compounds byultrasonic measurements in compression mould", NDT and E International, 33 ,2000, p.123-130.

Sofer G., Hauser E., "A new tool for determination of the stage of polymerisation ofthermosetting polymers", Journal of Polymer Science, 8 , 1952, p.611-620.

Wang Y., Argiriadi M., Limburg W., Mahoney S., Kranbuehl D. D., Kranbuehl D. E.,"Monitoring polymerization and associated physical properties using frequencydependent sensing" Polym. Mat. Sci. and Eng, 70, 1994, p. 279-80.

Mechanical behavior simulation of glassfiber reinforced polypropylene foamlaminates

Tsuyoshi Nishiwaki* — Akihiko Goto**

* ASICS Corporation, R. & D. Dept.6-2-1, Takatsukadai, Nishi-ku, KOBE, 651-2271, Japan

[email protected]. asics. co.jp

** Osaka Sangyo Univ., Dept. of Information Systems Eng.3-1-1, Nakakakiuchi, Daito, OSAKA, 574-8530, Japan

[email protected]

ABSTRACT: A glass fiber reinforced PP foam (GF/PP foam) can produce the contraryrequirement properties, for example high specific modulus and high damping. The GF/PPforam is a heterogeneous material with some designing parameters, fiber volume fraction,foaming ratio. In this study a simplified numerical model of GF/PP foam is proposed. In casethat mechanical behaviors of the heterogeneous plates are predicted.consideration of theheterogeneity is an important key. In thisproposed method, GF and matrix foam arerepresented by orthotropic shell and beam elements, respectively. The reduction raio in thecross-sectional area of beam elements corresponds to the foaming ratio. In order to check thevalidity of the proposed model, 3-point bending and eigenvibration analyses are performedand compared with experimental results.

KEY WORDS: GF/PP foam, foaming raio, heterogeneous numerical model, bending analysis,eigenvibration analysis


1. Introduction

In order to fabricate fiber reinforced plastics (FRP) with the more excellentproperties, various reinforcements and matrices have been proposed. In theconventional designing, mainly the requirement properties have been static or quasi-static strength and modulus. Nowadays dynamic properties such as eigenfrequenciesand damping are focused. In the homogeneous materials, the above static propertiesare contrary to the above dynamic properties. To put it the other way round, highdamping material cannot produce the high strength and stiffness. However FRP havea possibility to produce both the excellent static and dynamic properties. This isbecause FRP is a heterogeneous material. The static and dynamic properties areaffected by reinforcement fiber and matrix, respectively. In case that the optimizedfiber, matrix and reinforced shape are selected, the FRP can produce the highstrength, high modulus and high damping at the same time. Hybrid compositematerials (Goto et al., 1996) and FRP with flexible interphase (Nishiwaki et al.,2002) have been fabricated in order to propose a new FRP with high static anddynamic performances. Moreover a new FRP based on the foaming technique, fiberreinforced plastic foam has been proposed. For the FRP foam, there are variousdesigning parameters, fiber volume fraction, foaming ratio and fiber orientation.Therefore FRP foam can produce the more extensive mechanical parameters, ascompared with the conventional FRP.

In the application of FRP foam to the actual products, the prediction of themechanical properties is very important. In this prediction, the finite element methodis a very powerful and convenient tool. However the direct application of theconventional homogeneous model to the numerical simulation of FRP foam causesvarious issues. In the other words, numerical modeling considering the heterogeneityof FRP foam is required.

In this paper, numerical modeling method of the FRP foam composed of shortglass fiber and polypropylene(PP) foam is proposed. Two types of simulations, staticbending and eigenvibration analyses by using the proposed modeling method arecarried out and the validity of the modeling method is checked by the comparisonwith the experimental method. Finally the eigenvibration behaviors of the laminatedFRP foam are also discussed.

2. Test specimens

Powder typed PP and chopped glass fiber (GF, length 10-20mm, diameter 10-13mm) are used. Figure l(a) shows the typical fabrication process of the glass fiberreinforced polypropylene foam (GF/PP foam). First of all, PP powder and GF aremixed with water surface active agent foam. Secondly the mixture is sheeted due tosome drying processes. Thirdly the sheet are thermally expanded and molded.Figure l(b) shows the photomicrograph of GF/PP foam finished. In this study 6


types of GF/PP foam plates with various densities are used. The fiber weightfraction and nominal thickness are constantly 54% and 5mm, respectively. Table 1shows the list of specimens, here Type-1 denotes the lightest GF/PP foam and Type-6 denotes the GF/PP foam with the highest density, 1.04 as a convenience.

Figure 1. GF/PP foam

Table 1. Type number list and properties

Type

123456

Density [ g/cm3 ]

0.270.400.540.680.771.04

GF fraction| [ wt% 1 [ Vf% ]

54.0 5.8054.0 5.6454.0 11.6754.0 14.6954.0 16.6054.0 22.47

3. Modeling method

In case that the mechanical behaviors of the heterogeneous composite structuresare predicted, the application of homogeneous model with the equivalent stiffnesscauses various problems. As already mentioned, the stiffness and damping aremainly affected by GF and PP foam, respectively. This indicates that the numericalmodeling considered the heterogeneity is required. The application of thehomogeneous model with the equivalent stiffness must not give us directinformation for the designing. In our previous studies, the simplified heterogeneousnumerical model called as quasi-three-dimensional model has been proposed and thevalidity has been also checked for various simulations (Nishiwaki et at., 1993, 1995,1996, Tanimoto at al., 2001 ). In this modeling, the composite laminated structure isdefined as the stacking structure with fiber plates and interlaminar matrix. The


quasi-three-dimensional model is constructed by shell and beam elements, whichcorrespond to reinforcement fiber plate and interlaminar matrix, respectively. Theapplication of the quasi-three-dimensional modeling method to GF/PP foam plate isproposed.

Figure 2 shows the modeling concept proposed in this paper. At first, the plateis divided into n layers in the thickness direction. This division produces theinterlaminar matrix. Figure 2(a) indicates the example with n=2. GF is gathered onthe neutral surface in each layer, as shown in Figure 2(b). Then each layer isdivided into 3 layers, PP foam layer / GF layer / PP foam layer. The GF and PPfoam layers are represented by orthotropic shell and beam elements. In the GFlayer, the fiber volume fraction is 0.907, that is the theoretical maximum value(Hull, 1992 ). Shell elements are connected by beam elements corresponding to PPfoam in the thickness direction. Here, the outmost PP foam layers are ignored. Inthe previous quasi-three-dimensional modeling, beam elements have dottedrectangular cross-section as shown in Figure 2(c). In this case, beam elementshave the smaller cross-section in order to represent the void. Table 2 shows theparameters of all Types. These values can be easily calculated by densities of GFand PP, those are 2.5g/cm3 and 0.9g/cm3.

Figure 2. Modeling concept for GF/PP foam

Table 2. List of geometric parameter and contents of constituents in Type-1 to -6

Type

123456

size [ mm ]

270*270*4.8

Totalweight [g]

94140189238269364

GFw t [ % 1

54.00

V f [ % ]5.808.64

11.6714.6916.6022.47

Matrix volume [ cm3 ]PP foam

329.61319.68309.10298.51291.82268.29

PP48.0471.5696.60

121.64137.49186.04

Void281.57248.12212.50176.87154.3382.25


The thickness of shell element, t1 can be calculated by :

where n is division number in the thickness direction. Vf denotes the fiber volumefraction listed in Table 2. As already mentioned, the target has a constant thicknessof 5mm. Then the dimensions of beam element cross-section, L1 and W| can beobtained from:

where LO and WO denote the dimensions in the conventional quasi-three-dimensional model corresponding to the conventional FRP. r denotes a void co-efficient, which can be calculated by :

Vppfoam denotes the volume of PP foam listed in Table 2. VPP is obtained from VpPfoam

minus void volume, VVoid.

By using the above procedure, the proposed model has the constant materialproperties listed in Table 3. The differences in the proposed models corresponding toType-1 to Type-6 are the shell element thickness and cross-sectional dimensions ofbeam element.

Figure 3. Dimensions in beam element with void

Table 3. Material properties applied to the proposed model

shell element

EL=19.6 GPa, ET=12.7 GPa, Ez=7.84 GPa, NULT=0.03,

GLT=6.37 GPa, GLZ=3.92 GPa, GTZ=3.14 GPa, Den=2.27g/cm3

beam element

E=0.12 GPa, NU=0.4, Den=0.90 g/cm3


4. Analytical procedure

In order to check the validity of the proposed modeling, two types of analyseswere carried out. In these analyses, n is set to 2, this indicates the minimum elementnumber.

4.1. Three-point bending analysis

The object has 100mm length, 15mm width and 5mm thickness. The spanlength is 80mm. Figure 4 shows the numerical model with n=2. This model has 40shell elements and 33 beam elements. The fabrication process of GF/PP foam platehas the rolling process as shown in Figure l(a). This indicates that the GF/PP foamplate has an orthotropy. Two types of bending analyses were performed. One is thatthe rolling direction (L) is set to be x direction in Figure 4. The other is that thewidth direction (T) is set to be x direction. From these simulations, flexural moduliin Type-1 to -6 were predicted.

Figure 4. Numerical model with n=2 used in the 3-point bending simulation

4.2. Eigenvibration analysis

The object has 270mm length, 270mm width and 5mm thickness. In thisanalysis, 1st, 2nd and 3rd eigenfrequencies and vibration modes are predicted underthe free boundary condition. The model with n=2 has 50 shell and 36 beamelements.

5. Analytical results

In order to check the validity of the analytical results, experimentalmeasurements were carried out. In the three-point bending test, INSTRON testingmachine was used. In the eigenvibration test, modal analysis system ( AD3542, A &D Co.Ltd.) was used. In this modal analysis, GF/PP foam plate was suspended by afine string. This is equivalent with the analytical free boundary condition.


Figure 5 shows the flexural modulus plotted against density obtained fromanalytical and experimental results. Judging from these figures, it was confirmedthat both the relationships had similar tendencies in both the directions. On thewhole, predicted moduli are little smaller than experimental ones. This is derivedfrom the small division number, n=2 in the thickness direction. In this modeling, asalready mentioned, the outmost PP foam layers are ignored. Therefore the error inanalytical and experimental moduli can be reduced with increasing n.

Figure 5. Comparison of density dependency in analytical and experimental results

Table 4 shows the comparison between analytical and experimental results. Inthe eigenvibration analyses, 1st, 2nd and 3rd eigenvibration modes were 1st torsionalmode, 1st bending mode in T direction and 1st bending mode in L direction,respectively. This order was supported by the experimental results. The order cannotbe affected by GF/PP foam density. All the differences in both the eigenfrequenciesare smaller than 12%. This indicates that not only stiffness matrices but also massmatrices in the proposed model are valid.


Table 4. Comparison in eigenvibration modes1st ( 1st torsional) 2nd ( 1 st trans.bending) 3rd( 1st longt.bending)

Type123456

FEM exp error88.2888.4388.3683.39

78.8779.91

79.5878.81

87.58 78.6788.06 ! 84.91

0.120.110.11

0.060.110.04

FEM119.90120.20120-l0119.90119.40119.40

exp error123.80124.52122.43124.00126.82133.87

-0.03-0.03

-0.02

-0.03-0.06-0.11

FEM exp error148.90 144.03 i 0.03149.00 153.74 -0.03149.00 152.06 -0.02148.90 160.00 -0.07148.30 159.15 -0.07148.60 169.34 -0.12

error = (FEM-exp)/exp

6. Discussions

In the previous chapter, the proposed model with n=2 could predict both thestatic flexural moduli and eigenfrequencies of GF/PP foams with various densities.In this chapter, n dependency on the flexural modulus is first discussed. Secondaryeigenvibration properties of GF/PP foam laminates with various stacking arediscussed.

Figure 6. Comparison of density dependency in analytical and experimental results


The PP foam volume ignored depends on n. In the other words, the volume ofthe PP foam ignored is decreasing with increasing n. Figure 6 shows the 3-pointbending analytical results with n=2, 3 and 4. For all analytical results, it wasconfirmed that the flexural modulus was proportional to the density. Flexuralmodulus obtained from n=4 is larger than that from n=2. This is derived fromreduction in PP foam volume ignored. As shown in Figure 6, three lines obtainedfrom the proposed model are within the dispersion of experimental results. It wasconcluded that the simplest numerical model with n=2 was most effective for theprediction of the flexural modulus on GF/PP foam considering the total solutiontime. The total solution time with n=2 is 7 sec. on PC ( HP Limited, Pentium II,450MHz).

The eigenvibration properties of three layered GF/PP foam laminates withvarious density distributions are also predicted. The specimens have 250mm length,200mm width and 14.4mm thickness. Here length and width directions are set toL(rolling direction in Figure l(a)) and T directions, respectively. The specimensused are Type-242, -323, -545, -616 and -646. Type-242 denotes the skin layers areType-2 and core layer is Type-4. Each layer has the constant thickness of 4.8mm.For the adhesive between GF/PP foams epoxy resin was used. The epoxy resin wasnot modeled, because the adhesive has much higher stiffness and smaller thicknessas compared with each layer thickness. Figure 7 shows the modeling example ofType-616 with n=2. In this modeling, cross-section of beam element changes at theinterlamina between Type-6 and Type-1. This interlamina was modeled by using twotypes of beam elements with different cross-sections as shown in Figure 7(b). Byusing the model, 1st, 2nd and 3rd eigenfrequencies and vibration modes werepredicted under the free boundary condition. Table 5 shows the comparison betweenanalytical and experimental results. From this table, the order of the modes excitedis same as that of single layer as shown in Table 4. For all the laminates, it wasconcluded that the proposed was very effective for the eigenvibration properties ofGF/PP foam laminates.

Figure 7. Modeling example of Type-616 laminate


Table 5. Comparison between analytical and experimental results in eigenvibrationmodes for GF/PPfoam laminates

1st ( 1st torsional) 2nd (1st T.bending) 3rd (1st L.bending)

Type

242

323

545

616

646

FEM exp error

333.40 1 319.60

407.20 363.00

410.00 387.21

374.40 388.42

413.30 ! 418.65

0.04

0.12

0.06

-0.04

-0.01

FEM exp error

532.90

562.90

631.30

607.00

677.20

524.46 0.02

585.40 -0.04

627.79 0.01

601.89 0.01

680.78 -0.01

FEM exp error

643.50 624.26 0.03

681.80 725.98 -0.06

687.40 764.08; -0.10

729.50 753.64 -0.03

814.00 816.34 0.00error = (FEM-exp)/exp

7. Conclusions

For the prediction of the flexural modulus and eigenvibration properties ofGF/PP foam plate, the simplified heterogeneous numerical model was proposed. Thevalidity was checked by comparison with experimental results. The advantage of theproposed model is the independent consideration of components in GF/PP foam.GF/PP foam has three components, GF, PP and void. By using the proposed model,influences of these components on the whole GF/PP foam plate can be individuallypredicted. Therefore it was confirmed that the proposed model was very effectivetool for the actual designing.

References

Goto A and Maekawa Z., "Analysis of vibration damping properties of hybrid composite withflexible matrix resin", Material Science Research International, vol.2, no.3, 1996, p. 160-165.

Nishiwaki T and Tange A., " Static and dynamic properties of unidirectional CFRP laminateswith flexible interphase", Composite Interface, 2002, in press.

Nishiwaki T and Yokoyama A., "A simplified tensile damage analysis method for compositelaminates using a quasi-three-dimensional model", Composite Structures, vol.25, 1993,P.61-67.

Nishiwaki T and Yokoyama A., "A quasi-three-dimensional elastic wave propagation analysisfor laminated composites", Composite Structures, vol.32,1995, P.635-640.


Nishiwaki T and Yokoyama A., " A q uasi-three-dimensional strength analysis method forlaminated composite materials.", Proceedings of American Society for Composites,vol.11, 1996, p.150-158.

Tanimoto Y and Nishiwaki T., "A numerical modeling for eigenvibration analysis ofhoneycomb sandwich panels", Composite Interface, vol.8, no.6,2001, p.393-402.

Hull D., An introduction to composite materials, Cambridge, Cambridge University Press,1992.

Short-fibre-reinforced thermoplastic forsemi structural parts: process-properties

Eric HARAMBURU*' ** — Francis COLLOMBET*Bernard FERRET* — Jean-Stephane VIGNES**Pierre DEVOS*** — Christophe LEVAILLANT****Fabrice SCHMIDT****

* Laboratoire de Genie Mecanique de ToulouseInstitut Universitaire de Technologie Paul Sabatier133 avenue de Rangueil, 31077 Toulouse cedex 4, [email protected] - [email protected]@gmp. iut-tlse3.fr

** MICROTURBO Groupe SNECMA8 chemin du Pont de Rupe - BP. 2089 - 31019 Toulouse cedex2, France

[email protected]

* * * DRIRE Midi-Pyrenees12 rue Michel Labrousse - BP. 1345 - 31019 Toulouse cedex2, France

pierre. [email protected]

**** Ecole des Mines d'Albi Carmaux / CROMEPCampus Jarlard - Route de Teillet - 81013 Albi Cedex 09, France

[email protected] - [email protected]

ABSTRACT. This work concerns the high pressure injection of semi structural polymer -shortfibre reinforced-parts. A research group composed of three aeronautic firms of Toulouse andtwo academic laboratories of the French Midi-Pyrenees Region, intends to deal with theseglobal problems involving the manufacturing process to the structural analysis of the injectedparts. More precisely, this -work concerns the interfacing of the industrial computation toolsof fibres orientation and mechanical response of structures. Relying on orientation data, thecomputation and the localization of the distributions of homogeneous elastic properties areperformed for three industrial parts, which are respectively a body of aircraft pressure valve(Liebherr Aerospace), a first stage stator (Microturbo) and a fan wheel (Technofan).

KEY WORDS: Short-fibre-reinforced composites; Fibre orientation; Injection molding;Computational structure mechanics.


1. Introduction

The presented work lies within the scope of a research project which relates tothe high pressure injection moulding of semi structural technical polymer -shortfibre reinforced- parts. From a collective reflection carried out by equipmentsuppliers of the aeronautical sector of the Midi-Pyrenees Region emerged anincreasing need related to the replacement of metallic materials by polymericcomposites. Competition, in the sector of aeronautics, justifies the interest inthermoplastic short fibre reinforced composites. Indeed those allow reducing thecurrent production cost of certain parts concerned with high mechanicalcharacteristics and precise geometry.

In a more detailed way, there is an interest in lightening the on board systems, inincreasing the production series and the productivity, matching the functionintegration requirements. However, a certain number of obstacles can be noted suchas strong heterogeneity of materials, dependence with the manufacturing processboth the final mechanical characteristics and geometrical forms as well as aninexperience in the field of process and non destructive control techniques. Althoughthe aeronautic requirements have to be respected, the project is not reduced to asimple substitution of metallic materials by advanced technical materials but mustalso be linked to a new design of the part.

The industrial part of the project is to develop, to produce and to validate thefollowing composite parts (Figure 1), a body of aircraft pressure valve, a fan wheeland a first stage stator for a gas turbine based on the industrial competences of theMidi-Pyrenees.

Body of aircraft pressure valve First stage stator with an assembly Fan wheel(Liebherr Aerospace) of composite blades (Microturbo) (Tcchnofan)

Figure 1. Industrial goals (in UItem 2300)

The objective of the research consists in characterizing the coupling between thedesign and the manufacture of semi-structural composite parts injected with shortfibres, in order to carry out a stress analysis representative of heterogeneity of theinjected item. It is well-known that the reality of the composite material only existsin the achieved part and strongly depends on the manufacturing process. Howeverthere is no optimization method of the process, no design tool available to theindustrial engineering and design departments being able to deal with this reality. It


represents an impediment of the economic interest of the composite solutionregarding the traditional metal solutions.

As far as injected short fibre composites are concerned, it is of major importanceto be able to determine either the local orientation of the fibres or the presence ofwelded joints induced by the filling mode of the moulds or the voids. Indeed todesign the mould, the plastic moulder has to know the distributions of the fibresorientation in order to control the shrinkages while cooling. In the same way, beforedesigning the mould, the mechanical engineer has also to be sure of the fibresorientation to realise a reliable mechanical design according to the heterogeneity ofthe material. Thus, the orientation mechanisms during the filling are the commondenominator between the plastic moulder and the mechanical engineer's works.

The research group thus intends to deal with the global problems represented bythe manufacturing process to the structural analysis and the final control of thedefects within the injected parts, in respect to the experimental and numericalaspects. Together with the industrial firms, the academic partners are the researchteam PRO2COM of the LGMT and the CROMeP of the Engineering School of theMines in Albi-Carmaux (EMAC). Three PhD. works are supported. This paper moreparticularly features the assessment about the research team PRO2COM of theLGMT (PhD.1 in progress by E. Haramburu, jointly supervised withMICROTURBO company).

Within the research strategy (Figure 2), a main part of this work is the interfacingof the industrial computational tools of fibres orientation and mechanical responseof structures. The mathematical representation of the fibres orientation followsAdvani's method (Advani and Tucker, 87). It is integrated within the calculationmethods of the mechanical properties, based on homogenization techniques.

Figure 2. Research strategy

From a practical point of view, this interface allows to gather the results of anumerical injection simulation in terms of fibre orientation data (components of theAdvani's tensor), one finite element after the other and through out the thickness ofthe piece. An estimation of the homogenized mechanical properties is carried outlocally thanks to the Mori and Tanaka's method. The interface then builds thenumerical pattern in relation to a code by associating the calculated local properties

1 This work features interactions with two other PhD. in progress (M. Wesselmann and G.Saint-Martin) supervised by CROMeP and LIEBHERR and TECHNOFAN firms.


with the finite elements and a chosen mesh (from a same CAD pattern for the twoanalysis types).

The corresponding results obtained at the various stages of the interface programexecution are available for the design department engineer in order to check itsrelevance. The quantitative validation of this approach known as "wholly numerical"is underway to be achieved thanks to some tests on industrial structures.

2. Elastic properties estimation method

The heterogeneous nature of the injected short fibre composites is well adaptedto the application of micromechanical modelling from homogenization techniques.Indeed, these methods consist in determining the elastic properties of an equivalenthomogeneous material according to the properties of the various components.Moreover, they can be extended to the nonlinear behaviour and damage problems(Collombet et al, 97), (Dunn and Ledbetter, 97), (Wang and Weng, 92).

The homogenisation is obtained on average over an elementary representativevolume (ERV) of the material provided that is chosen an ellipsoidal geometricalrepresentation of heterogeneities. The ERV is the volume of the material containingall the heterogeneities (microscopic scale) that influences the mechanical response(macroscopic scale). In a classical way, a first stage called "the representation stage"requires to choose the various heterogeneities types and their geometricaldimensions associated with their ellipsoidal representation. This choice ofteninduces a numerical fitting with experimental results on elementary specimens andmastered conditions of injection.

The "localization" stage consists in defining average constitutive laws betweenthe micro and macro-scale (Hill, 63). Thanks to strain and stress concentrationtensors A and B as ratios between the average heterogeneity strain (or stress) and thecorresponding average in the composite, stiffness and compliance tensors C and S ofcomposite are given by:

where superscript i indicates quantities associated with the N heterogeneities of theERV, and superscript 0 denotes a matrix quantity. Symbol f represents the volumefraction for matrix or heterogeneity phases.

Equation [1] gives dual generic expressions for stiffness and compliance tensorsin terms of strain and stress concentration tensor A and B. Then, the differentmicromechanical approaches in the literature provide different ways to approximateAorB.


2.1. Eshelby's model

Eshelby (Eshelby, 57 and 61) calculates the disturbance of the strain field in theERV because of the heterogeneity of the given elastic properties (principle ofequivalent inclusion). Eshelby obtains the strain-concentration tensor (in theprincipal local directions of heterogeneity i) such as:

where symbol I represents the fourth-order unit tensor and Ei denotes the Eshelby'stensor in accordance to the shape of the ellipsoidal heterogeneity by its aspect ratior=l/d (with 1 and d, respectively the length and the diameter) and of the elasticproperties of the isotropic matrix by the Poisson's ratio (Mura, 82).

Moreover, if we consider an orientation of the heterogeneity i in the globaldirections of the ERV, one obtains (Pettermann et al., 97):

where T(q,<j> ') is the transformation tensor for fourth-order tensors in terms of Eulerangles qi and qi which performs the rotation from the local system of heterogeneity iRLOC to the global ERV system RERV (Figure 3).

Figure 3. Local directions RLOC in the heterogeneity i in the global directions RERV

However, Eshelby's model represents the elementary situation of an isolatedheterogeneity. It does not consider the interactions between the ERV phases. Fromthe model given by Eshelby, Mori and Tanaka have determined a disturbance onaverage in the macroscopic fields due to the each heterogeneity presence in theERV.


2.2. Mori & Tanaka's method

The aim of Mori & Tanaka's method (Mori and Tanaka, 73) is to take intoaccount the influence of the local interactions between the phases on the stress andstrain fields of the ERV. According to Benveniste formulation (Benveniste, 87), theobtained strain-concentration tensor is as follows:

From [1] and [2] the ERV equivalent stiffness is given by:

3. Coupling Advani's tensor and Mori and Tanaka's method

A mathematical representation of fibres orientation distribution (FOD) ininjected composite material with a symmetric rank 2 tensor is proposed by Advani(Advani and Tucker, 87). As much as this quantity is commonly used to providefibre orientation results from simulation or measurements, Advani's tensor does notallow a direct estimation of the overall set of elastic properties.

The use of the Advani's tensor in an estimation of the elastic properties of anequivalent homogeneous material has to be performed in accordance to thegeometrical meaning of the components ay. In particular, the components an, a22 anda33 show the probability of presence of a fibres population in the primary globalbasis associated with the given Advani's tensor. The off axis components (a12, a13

and a23) are essential to the three-dimensional geometrical representation of FOD butit is difficult to evaluate the physical reality of their contributions.

Thus, the eigendirections of the Advani's tensor corresponding to ERV vectorsbasis represent always a fair situation in order to use FOD into Mori & Tanaka'smethod. By means of the corresponding eigenvalues ai as probability of presence ofthree heterogeneities, the fibres volume fraction f in the ERV is spread in theeigendirections with a'.f. The relation [3] becomes:

In the above relation, we note that Eshelby's strain-concentration tensors A'Edepends on the transformation tensor in terms of Euler angles which performs therotation from local system of each heterogeneity in each eigendirection to principalbasis of Advani's tensor.


The elastic properties thus calculated are expressed in the principal directions of theAdvani's tensor. Finally, the shift from the principal directions to the primarydirections of Advani's tensor is performed thanks to the transformation tensor builtwith the normalised eigenvectors.

To sum up, the probability of presence of fibres in a given direction isrepresented by the diagonal Advani's tensor. Its principal directions define the greataxis of the three heterogeneities for the corresponding fibres groups. They representthe directions of the local axis of the heterogeneities whose respective fibre contentsare a weighting of the total fibre content by the eigenvalues of the Advani's tensor.Mori and Tanaka's estimation of the stiffness matrix of the equivalent homogeneousmaterial is carried out in these axes. The last stage consists in coming back to theprimary basis.

4. Interfacing the simulation tools

Calculating the elastic properties from the orientation tensors is used to interfacethe tools of the injection simulation through the orientation prediction and thecomputer codes of structure.

4.1. Interfacing strategy

For the design of a part, the engineer has to fulfil the feasibility and mechanicalbehaviour requirements. FOD is the bridge between results of the injection conditionsimulation and a stress analysis, starting from determining the homogeneousmechanical properties. This tool has to be usable by the design department. For thatthe designer must have means of graphically representing the intermediate resultsfor a non stop analysis from the process to the structure design.

MTD (Mori-Tanaka-DOS) is the generic name of the program allowing theprocessing of orientation measures coming from the analysis of MEB images orfrom a simulation thanks to Moldflow® software for example.

In this way, the series of operations leading to a stress analysis starts with a CADmodel of the part (Figure 4). The part is then meshed and imported in Moldflow®injection simulation software used in this study. According to the adjustments of thespecified injection parameters (pressure, temperature, gate(s) localization and so on),the calculation of the orientation prediction is made. As output data, the orientationresults are given in the form of one or several Advani's tensors per finite element(several Advani's tensors for the given finite element represent the FOD through thethickness).

Starting from the orientation data, the calculation and the localization of thedistributions of homogeneous elastic properties are performed along the injectedparts. A first possibility is to calculate mechanical characteristics to the finite


elements defined for a Moldflow® simulation. The cell used for the injection model,can be use for a stress analysis. Apart from the Moldflow® meshing, the designercan juxtapose a mesh devoted to the stress analysis (with other types of finiteelements, thinness, and so on). It should be emphasized that this assignment ofproperties is done by means of a proximity criterion. This criterion gets a sense onlyif the two meshes (Moldflow® and stress analysis) have been created from the sameCAD model of the studied part and with some obvious conditions of common sense.

Figure 4. Interfacing strategy and operation sequence

4.2. MTD interface patterns

Interface MTD is a program written in FORTRAN 77 allowing a wide use underMS-DOS and UNIX with various blocks of routines representing more than 10000lines of code. The user interface allows the acquisition of input data via a commandfile with a MTD specific syntax. The designer has some "user" information currentlyoperating at his disposal such as the data card reports or warnings and errors duringMTD execution. A graphic interface called XMTD can be the user's assistant for theedition of the command files, the executions of calculations and the visualization ofthe intermediate results provided by MTD.

5. Numerical results

Several illustrations of the various stages of the calculation sequence arepresented for the three industrial parts.

5.1. The orientation prediction

The fibres orientation data are obtained at the end of the filling simulation of thepart volume (performed by Moldflow®) with the injection parameters defined by themoulder. The study is limited to the mere filling phase. The runners, the global cycleof injection including the packing and cooling times have not been modelled. These


simplifications are justified because only the FOD are required. In this case, wesuppose that only the filling phase has an effect on the FOD (Figure 5).

Figure 5. Filling of industrial parts at 0.5 second (Moldflow® software)

5.2. Computation and assignment of elastic properties

The orientation tensors are the input data calculation, in each finite element(ERV), of the elastic properties. They are used in a stress analysis via the MTDinterface. The visualization of the properties is not possible with the commercialcomputer codes such as for example the codes used by the project partners(Samcef®, Ansys®, I-deas® and Nastran®). Visualization options can be activatedin MTD interface for real noting the distributions of some Young's moduli of thecomposite parts (Figure 6). Figure 6 shows for example with grey levels thedistribution of the Young's modulus in the direction of the paddle height of theMicroturbo stator blade.

Figure 6. Distribution of the mechanical characteristics (Young's moduli in GPa)


These computations have been done by giving the following values to thecharacteristics of the matrix and the short fibres (GE Plastics Ultem®2300):

- Ultem®1000 matrix: E = 3.2 Gpa and v = 0.38- Glass fibres E: E = 70 GPa, v = 0.20, r = 15 and f = 17.5%

The aspect ratio r equal to 15 has been obtained after a numerical fitting by MTDthanks to simple standard traction specimens (Figure 6). This result recovers anexperimental characterisation of the length distribution of fibres carried out by M.Wesselmann and G. Saint-Martin supervised by CROMeP, Liebherr Aerospace andTechnofan firms. In more detailed way, the injection and filling conditions impose asingle direction of the fibres along the great axis of the specimen. The experimentalvalue of the Young's modulus in this direction is of 9.8 GPa. Figure 6 shows astrong sensitivity of the Young's modulus versus the aspect ratio. Indeed, acorresponding value for r = 10 is about 9 GPa. In this situation, it is easy to find theaccurate aspect ratio value thanks to homogenization.

Figure 7. MTD distribution of the Young's modulus (in Pa) values after injection inthe main axis of the specimen

5.3 Stress analysis

The stiffness matrices calculated on each finite element of the Moldflow® meshcan give birth to the following alternative. They can be used to export mechanicalcharacteristics on the same elements and finally to edit a data file for a computercode containing the coordinates of the nodes, connections of the finite elements andthe elastic moduli of the composite (which implies, in our example, to re-use thetriangles of the Moldflow® model). The stiffness values can be assigned ongeometrical points thus making possible to associate elastic moduli with the finiteelements of another mesh type which would come to be superimposed instead of theMoldflow® finite element model. In this case, MTD writes a data file for a


computer code containing only the characteristics of the materials assigned to eachfinite element of the mesh chosen for the stress analysis.

Finally, the output data of MTD interface enables to start a stress analysis for anymodel. The nodes, the finite elements and the properties of materials being defined,it remains then to provide the modelling elements relating to the boundaryconditions. With confidential industrial specifications of the modelling stages, anumerical simulation of the stator blade with quadratic tetrahedral finite elements isperformed; body valve and fan wheel as well (Figure 8).

Body valve with 20000 TFE Stator blade with 8000 TFE Fan wheel with 20000 TFE(Liebherr Aerospace) (Microturbo) (Technofan)

Figure 8. Von Mises stress map for the three parts during operation by Samcej©software after the backing of the Mold/low® orientation predictions (scale and unitsare not provided)

7. Conclusion

In the field of short fibres composite structure analysis, the major problem is thelack of information concerning the heterogeneity in the composite parts. Theheterogeneous nature depends on both the process and the design phase. It is animpediment to the industrial solution as far as cost is concerned. The cost reductionof composite parts depends on the development of an advanced numerical toolcapable to loop from the manufacturing conditions to the mechanical response of apart. The MTD interface is actually used by the industrial partners with their owncodes of structure. This interface allows the designer to consider the influence of theinjection process on the distribution of the mechanical properties for a new industrialdesign approach. On various scales, from the specimens to the industrial parts inservice conditions, the experimental validation campaign is progressing thanks toindustrial partners.


Acknowledgements

The authors would like to thank Claude Rossignol and Matthias Wesselmann (LiebherrAerospace), Olivier Darnis and Gilles Saint-Martin (Technofan) for their helpful technicalcollaboration. This work was done with the financial support of the Midi-Pyrenees RegionalCouncil, the European Union and the French Agency of Technical Research (ANRT).

8. Bibliography

Advani S.G., Tucker III C.L., "The Use of Tensors to Describe and Predict Fiber Orientationin Short Fiber Composites", J. of Rheology, vol.31, 1987, p. 751.

Benveniste Y., "A new approach to the application of Mori-Tanaka's theory in compositematerials", Mechanics Materials, vol. 6, 1987, p. 147-157.

Collornbet F., Bonnan S., Hereil P.L., "A mesomechanical modelling of porous aluminiumunder dynamic loading: comparison experiment - calculation", International conferenceon mechanical and physical behaviour of materials under dynamic loading, Eurodymat97, Toledo (Spain), 22-26 September 1997, Journal de Physique IV, Colloque C3, LesEditions de Physique, 1997, p. 643-648.

Dunn M.L., Ledbetter H., "Elastic-Plastic behavior of textured short-fiber composites", ActaMetallurgica, Vol. 45, n°8, 1997, p. 3327-3340.

Eshelby J.D., "The determination of elastic field of an ellipsoidal inclusion and relatedproblems", Proceedings of the Royal Society, London, vol. A241, 1957, p. 376-396.

Eshelby J.D., "Elastic inclusions and inhomogeneities", Sneddon IN, Hill R. editors, Progressin Solid Mechanics, vol.2, 1961, p. 89-140.

Haramburu E., "Etude des couplages entre la conception et la fabrication de piecescomposites semi-structurales injectees avec fibres courtes, en vue de 1'obtention deproprietes mecaniques optimales", Rapport d'activites de lerc Annee de These,LGMT/PRO2COM, 2001.

Hill R., "Elastic properties of reinforced solids: Some theoretical principles", J. Mech. Phys.Solids, vol. 11, 1963, p. 357-372.

Mori T., Tanaka K., "Average stress in matrix and average elastic energy of materials withmisfitting inclusions", Acta Metallurgica, vol. 21, 1973, p. 571-574.

Mura T., "Micromechanics of defects in solids", Martinus Nijhoff Editor, The Hague, 1982.

Pettermann H.E., Bohm H.J., Rammerstorfer F.G., "Some direction-dependent properties ofmatrix-inclusion type composites with given reinforcement orientation distributions",Composites Part B: engineering, vol. 28B, 1997, p. 253-265.

Wang, Weng, "The influence of inclusion shape on the overall viscoelastic behavior ofcomposites", ASME, Vol. 59, 1992, p. 510-518.

Guidelines for a quality control procedureto ensure sound strengthening andrehabilitation of concrete structures usingFRP

J.L. Esteves, A.T. Marques

INEGI/DEMEGI/FEVP, R, do Barroco 174- 214, 4465 - 431 Leca do Balio,Portugal

[email protected]@fe.up.pt

ABSTRACT: In this paper, a discussion will be made regarding the procedures to be followedfor the quality control of the strengthening and rehabilitation of concrete structures usingcarbon fibre/epoxy composites. Bearing in mind the different relevant parameters, which mayconsider short and long - term behaviour of the application of these materials, the procedurewill give information concerning: specifications/quality assurance; quality control of the re-inforcement system; quality control of the adhesive; quality control of the surface; monitoring

The work presented follows research work carried out in Portugal, together with collectedand treated information from material suppliers.

KEYWORDS: Composites, quality control, concrete structures, mechanical behaviour.

1. Introduction

The strengthening and rehabilitation techniques of concrete structures have beenmoving, recently, for the use of CFRP - Carbon Fibre Reinforced Plastic (epoxyresin) either as a laminate in a strip shape or as semi-product (prepreg like) to becured in-situ. In the first case, the strips are bonded to the concrete structure, with orwithout pre-stress. In the second situation, after the application of a resin rich adher-ence coating to the concrete, the semi-product will be impregnated with an epoxyresin promoting an exothermic reaction that will end up with the cure of all system.


Today, there are quite a lot of different solutions for the above purpose (wet lay-up systems and systems based on prefabricated elements) corresponding to severalmanufacturers and suppliers, based on different configurations, types of fibres, ad-hesives, etc ..., and there is a need for an efficient Quality Assurance and Control toavoid costly surprises and to produce a sound rehabilitation or strengthening of thestructure. An interesting approach can be seen in Machida (1997).

2. Quality Assurance and Specifications

It is essential to define clearly what are the requirements for the reinforcementsystem. Hence, the project of the structure must be available and the actual condi-tions of the concrete must be evaluated using non-destructive or very little intrusivetests. Particular conditions, such as fire resistance, have to be considered.

As the reinforcement system has a polymeric matrix, there is a need to identifyclearly the environmental conditions, particularly temperature and temperaturefluctuations, as this may affect the short and, even more, the long-term behaviour ofthe system. For the same reason, although in a small part, it is necessary to definethe type of loads in respect to the possible place where the reinforcement will be ap-plied, as well as their frequency and the likelihood of having vibrations and theirpossible magnitude. Moreover, the design concepts and safety must be based in theEUROCODE 1 and EUROCODE 2, following the philosophy of limiting states.

In order to have Quality Assurance, it is necessary the integration of procedure toverify the conformity at four levels, Juvandes (1999):

• Certification of reinforcing materials;• Qualification of suppliers and applicators;• Control of the application procedure: inspection of local conditions, inspec-

tion of surface preparation, inspection of primer and adhesive application, in-spection of CFRP composite, inspections of the bonding;

• Inspection in service and maintenance

2. Quality control of the application procedure

2.1. Inspection of local conditions

Typically, one can do the following:Detection and measurement of the recovering of the internal armaturesEsclerometric tests

a Ultra-sonic tests, by the indirect method to evaluate the depth of the cracksPull-off tests, to determine the tensile strength of the superficial layer of theconcrete


2.2. Quality control of CFRP system

Unless the materials are of proven quality and performance the following testsaccording to standard test methods have to be made:

Tensile and bending tests

DMTA-Dynamic Mechanical Thermal Analysis to evaluate the influence oftemperature in the modulus and to determine Tg

Fatigue and creep tests

Coefficient of thermal expansion

Nominal mass density

Fibre content

Moisture absorption and chemical stability

Some of the above tests must also be conducted after accelerated ageing.

In Table 1, it can be seen, to illustrate the importance of some parameters, thevariation of tensile strength, strain at rupture and tensile modulus as function ofageing conditions for two particular systems of reinforcement, Bravo (1999).

The samples were subjected to 30 cycles of one day in the following conditions:'Winter' 14 h at -5°C, 10 h at 15°C'Summer' l0h at 20°C, 14 h at 50°C.

Sample

Non-Aged

Aged(Winter)

Aged(Summer)

A

B

A

BA

B

Tensile strength(MPa)

331

1300334

1 3103401310

Strain at rup-ture %

1.1

0.81.2

0.91.5

0.9

Modulus(Gpa)

26.7

15224.7

15124.6150

Table 1 - Mechanical properties as function of ageing conditions for twosystems

(A - Replark CF-sheet, Mitsubishi Chemical corporation)

(B - INEGI CFRP-laminate strip 50x1.4)


3.3. Primer and adhesive

The adhesive is a key element in the reinforcing system. Hence the followingcharacteristics have to be known:

Glass transition temperaturea Shrinkage

Bond strengtha Shear strengtha Static modulus

Creep modulusa Coefficient of thermal expansion

The results of some tests made by Gonsalves (1998) to characterize two adhe-sives are presented in figure 1 and 2 in order to illustrate typical behaviours.

Figure 1. Bending modulus versus temperature for Epotherm like adhesive

Figure 2. Bending modulus versus temperature for CEMENT like adhesive

For the case of adhesives, the stress/strain curves must be obtained consideringequilibrium conditions in respect of temperature and relative humidity, Esteves (1991).


3.4. Bonding inspection

The bonding of the composite system to the concrete has to be inspected to de-tect any problem. The method to be used must be able to detect voids, displacementsor delaminations. Hence, thermographic and ultrasound methods, together with 'taptest type' method can be used.

The bond performance can be evaluated by means of direct pull-off tensile test-ing of the CFRP/bonding agent/concrete substrate combination. Test specimens areobtained by taking cores from the applicability test specimen. Tests are performed at7 days and 14 days under the specified curing conditions.

4. Application

To guarantee a sound reinforcement, the following procedures may be followed.

4.1 Surface preparation

In order to have an adequate bonding, the surface should be roughened and madelaitance and contamination free. This must be cleaned by means of blasting (sand,grit, water jet blasting) or grinding. The surface must be dry and free of any oil,grease or foreign matter likely to impair bonding.

4.2 Anchorages and couplers

Bearing in mind that the mechanical properties of the reinforcement system maybe, significantly, affected by anchorages and couplers. Hence, unless they are placedexactly according to the design specifications, there is a strong possibility of pre-mature failure. They must be corrosion-proofed to avoid reduced durability. A thor-ough inspection has to be made to these materials, and a specific control techniqueon the anchoring work must be followed.

4.3 A pplication

The application of the CFRP reinforcement system should be performed byqualified and experienced workers, in accordance with any special specificationsgiven by the manufacturers of adhesives and CFRP reinforcement, provided thatthey are not at variance with these specifications unless backed up by adequate re-search data.


Care must be taken to avoid excessive bending or impact during placement ofCFRP, as well as excessive temperatures, chemicals, welding sparks and over-tightening.

During application, the working area must be clean and having the adequate am-bient conditions to promote cure of the polymeric systems. If necessary, an externalsource of heat must be used to get complete cure.

4.4 H andling and storage

Bending beyond the limits, shocks, dragging during transport; temperature, hu-midity, dampness or direct sunlight during storage; welding sparks and chemicalsmay affect the reinforcing system prior to start the work. Hence, CFRP must be han-dled and stored carefully to prevent any damage caused by the referred factors.

Anchorages and couplers and any material used for this purpose must be care-fully stored to have them clean and undamaged.

5. Monitoring

Composite materials are particularly prone to become smart materials and to makesmart structures. The laminate is made in such way that gives the possibility to incor-porate fibre optical sensors and to perform remote monitoring. The application of thistechnique has been described by Frazao (2000) and is illustrated in figure 3 and 4.

Figure 3. CFRP reinforced concrete plate containing FBG (Fiber Bragg Grat-ings) sensors.


Figure 4. Test results of the strain and temperature evolution of the a) non-reinforced concrete plate and b) reinforced concrete plate containing 3 FBG sen-sors, Frazao (2000)

Traditionaly the monitoring of the CFRP reinforcement system can be made bythe use of electrical strain gauges illustrated in figure 5.


Figure 5. Monitoring of the CFRP reinforcement system applied on "NossaSenhora da Guia Bridge ", Ponte de Lima, Portugal.

5. Conclusions

A sound strengthening and rehabilitation of concrete structures can be obtainedproviding that an adequate design methodology is followed, good surface prepara-tion is done, application conditions are correctly executed and a quality controlmethodology is applied to avoid catastrophic surprises.

A smart monitoring system can be used, in order to have a close eye to the evolu-tion of the reinforcing system.

References

Juvandes, L., 'Reforco e reabilitacao de estruturas de betao usando materials compositesde CFRP', PhD Thesis, FEUP, 1999.

Bravo, S et al 'Avaliacao do comportamento a traccao, apos tratamento termico de duassolu96es de reparacao', LNEC, Report 132/99, Lisboa, 1999.

Esteves, J L 'Estudo do comportamento de adesivos estruturais, Tese de Mestrado,FEUP, 1991.

Frazao, O. et al. 'Optical fibre embedded in a composite laminate with applications tosensing', BIANISOTROPICS 2000, Lisboa, Portugal, 27 - 29/9/2000.

Gon9alves, F. A., 'Etude de materiaux composites dans le cadre d'un projet derenforcement de ponts en beton', Final Year Undergraduate Project, ENSAM/FEUP,1998.

Machida, A. (ed.), 'Recommendation for design and construction of concrete structuresusing continuous fibre reinforcing materials' Concrete Engineering Series 23, JSCE, 1997.


'Externally bonded FRP reinforcement for RC strutures', Fib Task Group 9.3 FRP (FibreReinforced Polymer) reinforcement for concrete structures), CEB - FIB, Federation Interna-tionale du Beton, July, 2001.

Normative references

LNEC E 226 - 'Betao. Ensaio de compressao', Lisboa, 1968LNEC E 227 - 'Betao. Ensaio de flexao', Lisboa, 1968LNEC E 397 - ' Betoes. Determinacao do modulo de elasticidade em

compressao', Lisboa, 1993pr EN 1542 - 'Products and systems for the protection and repair of concrete

structures - Test methods - Measurement of bond strength by pull-off, November1998

pr EN 1766 - 'Products and systems for the protection and repair of concretestructures - Test methods - Reference concrete for testing', 1999

pr EN 13687-2 - 'Products and systems for the protection and repair of concretestructures - Test methods - Determination of thermal compatibility - Part 2: Thun-der-shower cycling (thermal shock)', 1999

pr EN 13687-3 - 'Products and systems for the protection and repair of concretestructures - Test methods - Determination of thermal compatibility - Part 3: Ther-mal cycling without de-icing salt impact', 1999

pr EN 13706-1 - Reinforced plastics composites - Specifications for pultrudedprofiles: Designation

pr EN 13706-2 - Reinforced plastics composites - Specifications for pultrudedprofiles: test methods and general requirements

ISO 527: Plastics - Determination of tensile properties, 1993JSCE-E 131-1995 'Quality specifications for continuous fibre reinforcing mate-

rial'JSCE-E 531-1995 'Test method for tensile properties of continuous fibre rein-

forcing material'JSCE-E 532-1995 'Test method for flexural tensile properties of continuous fibre

reinforcing material'JSCE-E 533-1995 'Test method for creep failure of continuous fibre reinforcing

material'JSCE-E 534-1995 'Test method for long-term relaxation of continuous fibre re-

inforcing material'JSCE-E 535-1995 'Test method for tensile fatigue of continuous fibre reinforc-

ing material'JSCE-E 536-1995 'Test method for coefficient of thermal expansion of continu-

ous fibre reinforcing material'JSCE-E 537-1995 'Test method for performance of anchorages and couplers in

pre-stressed concrete using continuous fibre reinforcing material'


JSCE-E 538-1995 'Test method for alkali resistance of continuous fibre rein-forcing material'

JSCE-E 539-1995 'Test method for bond strength of continuous fibre reinforcingmaterial by pull-out testing'

JSCE-E 540-1995 'Test method for shear properties of continuous fibre rein-forcing material'

Numerical simulation of reinforcementsforming: the missing link for theimprovement of composite parts virtualprototyping

Patrick de Luca, Yanik BENOIT

ESI Software (ESI Group)99, Rue des Solets, SILIC11394538 Rungis CedexFrancepdl@esi-group. comybe@esi-group. com

ABSTRACT: Composites draping simulation is introduced. There are basically two kinds ofmethod: geometric approach and mechanical approach. The possible results that can beobtained using these methods are illustrated by an example. This type of simulation can beused not only to optimize the fabrication process but also to improve the mechanicalperformance calculations and more generally speaking the composite parts design. Forexample, the influence of the preforming operation on resin injection for processes like resinTransfer Molding (RTM) is demonstrated on a numerical example.

KEY WORDS: numerical simulation, composites, RTM, draping


1. Introduction

The numerical simulation is nowadays fully integrated in the design process ofindustry for metallic parts. In spite of the progress made in the modelling, this levelof maturity has not been reached so far for composite parts. One of the identifiedreason is that the strong effects of manufacturing on the mechanical performancecould not be taken into account.

One reports here the status of the simulation of the forming operations that wasdeveloped the last ten years (section 2). These methods are useful to assist theprocess engineer in the optimisation of the fabrication and even more importantlyprovides information to subsequent analysis. As an example, the use of the drapingresults to improve the injection in resin transfer molding process is reported insection 3. This is presented as a first step toward the development of a full numericaltool that will enable to perform mechanical performance analysis based on adescription of the composite part as it is built.

2. Reinforcement Forming Simulation

There are essentially two numerical methods available to simulate the formingoperations: the geometric method and the mechanical method. One describes each ofthese methods.

The geometric method uses only geometrical information: the part geometry.The numerical method used is known as the 'fisher-net algorithm' (Rudd et al.,1997) . This method is very rapid, the simulation time being of the order of onesecond. The results are made of the shear angle and possibly of the flat pattern.Because only the geometry of the part is used in a simulation, the reinforcementarchitecture is not taken into account, nor the process or the process variant used.The common use of this method is to identify the areas with large shear and tocompare the results with the maximum shear angle that can sustain the fabrics ('thelocking angle'). That allows for a rough estimation of the part feasibility.

The second method is the mechanical approach and leads to the use of the finiteelement method. Most of the examples reported as today are based on an explicittime integration. This method is very popular for dynamic problems like car crashsimulations and was extended successfully to forming problems (Pickett, 1995).This method can handle easily the various non linearities encountered in formingsimulations: large displacements, rotations and strains, non linear mechanicalbehaviour and non linearities induced by extensive contact.

For unidirectional reinforcement or for woven fabrics, non linear elasticbehaviour is assumed. A special treatment is done regarding the modelling of theyarns bending and shearing (Cartwright, 1999); this is dictated by the discrete natureof the reinforcement, as opposed to a continuum medium. If the reinforcement is not


dry but pre-impregnated, a viscous modelling of the matrix is introduced. In thiscase it is also necessary to use a mixed dry and viscous modelling of the sliding.When relevant, thermal modelling is included in the analysis; the temperaturemodifies the viscosity of the resin and the friction coefficients. Additional detailscan be found in (Pickett et al., 1996). As can be guessed from this short description,significant material characterization is necessary before conducting a simulation(Clifford ef al., 2001).

Thanks to this accurate modelling, all the details of the process like blankholderforces, tools and laminates temperatures, holding systems can be considered (deLuca et al., 1998). Also, different materials give different results. This method isclearly a tool to optimise a forming process.

One reports here an example where the geometric method fails and themechanical method reproduces successfully the reality. The part is a section of aprototype helicopter blade. An unidirectional prepreg is draped by hand over a tool.Regardless of the way to drape it, a wrinkle consistently appears at the sameconstant location and a lack of adherence is noticed on a part of a radius. The figure1 depicts a view of a ply after draping as computed by the finite element method. Awrinkle is clearly visible on the right hand side. The figure 2 shows sections alongthe length of the tool. One can see the tool sections (red lines) and the ply sections(green lines). There is a zone with a lack of adherence that is shown by thecalculation that is visible exactly at the location where it happens in the reality.

figure I. Prototype blade Helicopter: Figure 2. Sections view. WrinklingWrinkling. and lack of adherence.

The hand lay-up process is modelled using a static pressure with a valuecomparable to the pressure that can be exerted by the hands of a worker. Otherexamples have been reported elsewhere for a wide range of processes: matchedmetal tools, diaphragm forming, rubber pad forming and roll forming.


To be exhaustive on this topic, one has to mention the development ofintermediate method that combine a geometric approach and a mechanical approachthrough the minimization of the deformation energy (Long et al., 2001). This termusually includes only a limited number of terms (for instance only shear energy).This is useful to take into account the nature of the reinforcement but still theprocess is not taken into account.

3. Use of forming results in RTM injection simulations

Briefly the RTM (Resin Transfer Molding) unfolds in two steps. In a first step, apreform is placed in a mold which is closed. Then, resin is injected and flowsthrough the reinforcement. After curing, one obtains the composite part.

The works around the simulation of the RTM injection simulation started aboutthe same time as the development of the preforming simulation (Trochu et al., 1993)and have reached now a good level of maturity. The critical material parameters thatdrives the filling of the mold is the so-called permeability K that appears in theDarcy's law which is used to model the flow through the reinforcement.

where P is the hydraulic pressure, n is the fluid viscosity, and V is the velocity fieldTo perform a simulation, an experimental measure of the reinforcement permeabilityis necessary. Most of the time the permeability values used are the one of theundeformed reinforcement. Though it is known that the deformation modifiessignificantly the permeability, not only the numerical values but also the principaldirections of the permeability tensors as observed by (Louis et al., 2001). Thisappeals for calculation of the permeability field prior the beginning of an injectionsimulation. One reports thereafter an example of such an influence.

The geometry studied is a bath tube (figure 3). The filling time contour using aconstant permeability can be seen on figure 4. To study the effects of the draping, afilling simulation is done using a permeability field based on a preliminarygeometric calculation of the fiber reorientation that occurs during the draping(figures 5 and 6). The shearing angle reaches 52°. The permeability is computedusing the Kozeny-Carman model (Rudd et al., 1997). The results are shown onfigure 6: the shape of the flow front is different. From a practical point of view, itmeans that the vents should be located at different points. Other examples dealingwith a bonnet geometry shows a variation of the filling time of 20% (de Luca et al.,2002).


Figure 3. Bath tube geometry Figure 4. Filling time contour

Figure 5. Fiber reorientation Figure 6. Shear angle

Figure 7. Filling time contour


4. Conclusion and perspectives

A state of the art of the numerical simulation of reinforcements of compositeparts was presented. These new types of numerical tools enable not only to optimizethe draping process but also make available the manufacturing information (fiberreorientation) to further analysis. An example regarding the influence of draping onRTM injection was reported.

The current research tackles the modelling of new types of reinforcementarchitectures: multi-axial fabrics, knitted and braided reinforcements. To take fullyadvantage of the draping results, it becomes necessary to develop appropriatepermeability models for all of these reinforcements both in undeformed anddeformed state.

Finally, this work in the RTM field is only the first step of the development ofcomprehensive simulation tools for the design of a composite parts. The next stepsinclude using the draping information in mechanical analysis and in impact orcrashworthiness studies.

Acknowledgements

The author would like to thank GKN Westland for the results of the section 2.

5. Bibliography/References

Cartwright B.K., de Luca P., Wang J., Stellbrink K., Paton R., "Some Proposed Experimentaltests for use in Finite Element Simulation of Composites Forming", Proceedings of 12th

International conference on composites materials, 5th -9th July 1999, Paris.

Clifford M.J., Long A.C., de Luca P., Proceedings of The Minerals, Metals & MaterialsSociety (TMS) 2001 Annual Meeting, 11-15 February 2001, New Orleans, USA

de Luca P., Pickett A.K., Lefebure P. , "Numerical and Experimental Investigation of SomePress Forming Parameters of two Fibre reinforced Thermoplastics: APC2-AS4 and PEI-CETEX", Composites part A, vol. 29A, 1998, p. 201-110..

de Luca P., Benoit Y., Trochon J., Morisot O., Pickett A.K., "Coupled Preforming/InjectionSimulation of Liquid Composites Molding Processes", Proceedings of the SAMPE 2002Conference, May 12-16,2002 Long Beach, USA.

Long A.C., Souter B.J., Robitaflte F., "A fabrics mechanics Approach for Draping of Wovenand Non Crimp reinforcements ", Proceedings Of the American Society for Composites,15th Technical Conference, College Station, September 25-27 2000 USA, TechnomicPublishing Co. Inc., paper 176.


Louis M., Huber U., Maier M., "Harzinjektionssimulation unter Beriicksichtung desEinflusses der Drapierung aud die permeabilitat", Proceedings of the German SAMPE,2001.

Picket! A.K., Cunningham J.E., Johnson A.F., Lefebure P., de Luca P., Mallon P., SunderkmdP., O'Bradaigh C., Vodermayer A.M., Werner W, "Numerical techniques for the pre-heating and Forming Simulation of Continuous Fibre Reinforced Thermoplastics",proceedings of SAMPE Europe Conference and Exhibition , Basel, 28-30 may 1996.

Pickett A.K., Queckborner T., de Luca P., Haug E., "An explicit Finite Element Solution forthe Forming prediction of Continuous Fibre reinforced Thermoplastic Sheets",Composites manufacturing , vol. 6 no. 3-4,1995, p. 237-244.

C. Rudd, A.C. Long, K. Kendall and C. Mangin, Liquid Composite Molding Technologies,Woodhead Publishing Ltd., Cambridge, 1997.

Trochu F., Gauvin R. Gao D.M. "Numerical Analysis of the Resin transfer Molding Processby the Finite Element Method", Advances in Polymer Technology, vol. 12 no. 4, 1993, p.329-342.

*

Patrick de Luca is responsible at ESI Software of the development of theComposites Unified Solution. He worked the last ten years in the simulation ofcomposites forming. He holds a PhD. In Applied Mathematics from BordeauxUniversity in 1989.

Yanik Benoit was a main developer of the LCMFLOT software for RTM injectionsimulation the last five years. He develops now the Composites Unified Solution atESI Software. He graduated at Ecole Polytechnique de Montreal, Applied SciencesMaster, 1996.

Monitoring of Resin Flow and Cure UsingElectrical Time Domain Reflectometry

Kei Urabe — Tomonaga Okabe — Hiroshi Tsuda


[email protected]

ABSTRACT: This paper presents the potentiality of using responses to electromagnetic signalfrom a transmission line constructed inside a structure or material, as a new tool for in-situcure monitoring in the manufacturing process of resin composites. Experimentalinvestigations on the time domain response to a sharp step input signal from a modeltransmission line, where epoxy resin fills the gap between a pair of metal conductors of amicrostrip line were carried out. The results demonstrated that the time domain response cansuccessfully provide clear information on resin flow, poor impregnation and discontinuity ofthe cure stage, including information on the position along the line. Next, we propose the useof carbon fiber for conductive elements constructing the transmission line so as to usematerial reinforcements (i.e., carbon fiber) as sensing probes. The results demonstrated thepossibility of carbon fiber as transmission line elements.

KEY WORDS: cure monitoring, time domain reflectometry, electromagnetic wave, transmissionline, epoxy resin, carbon fiber


1. Introduction

The quality of resin composites strongly depends on the conditions of themanufacturing process. Hence, it is important to monitor the manufacturing processof the resin composites and to properly control the manufacturing tool according tothe monitored signal (Ciriscioli et al., 1991, Kenny 1994). As sensing methods forthe manufacturing process, dielectric monitoring (Mijovic et al, 1993, Shepard etal., 1995, Yamaguchi et al., 1999), piezoelectric devices (Ohshima et al., 2001),optical fibers (Chen et al., 1999, Osaka et al., 2001) and ultrasonic monitoring(Chen et al., 1999) have been investigated. Among these, dielectric monitoring hasbeen widely investigated and used as an in-situ monitoring method in themanufacturing process. In the manufacture of large structures or the resin transfermolding process, it becomes important to monitor distribution of the properties,discontinuities and/or resin flow. However, in the general dielectric monitoringmethod, a signal with a relatively low frequency (<1 MHz), of which the wavelengthis much longer than the size of the sensor or of the material to be monitored, is used.Hence, the obtained information is point data, or data integrated over the whole areaof the sensor (Shwab et al., 1996, Motogi et al., 1999).

In contrast, when a transmission line of electromagnetic wave is constructedinside the material or structure, an electromagnetic signal with a high frequency(>100 MHz) propagates as a wave with a wavelength comparable to the typical sizeof materials or structures. The propagating signal is affected by electrical propertiesof the material between the conductors of the line. The signal is then expected toprovide information on the properties of the material or structure, including theirdistribution or discontinuity (Banks et al., 1996). Therefore, we have recentlyproposed a new cure monitoring technique with a high-frequency electromagneticwave transmission line, and have carried out some experimental and theoreticalstudies on the frequency characteristics of reflectance using model transmissionlines filled with epoxy resin (Urabe et al., 2000). The results suggested the potentialof the technique as a tool for in-situ monitoring of curing and other properties,including implicitly information on local distributions or discontinuities.

The study presented in this paper seeks to obtain more explicit and clearinformation on discontinuity or distribution in the "transmission line method". Wetherefore investigated on the use of the time domain response to a step input signalfrom an electromagnetic wave transmission line, which is generally called "TimeDomain Reflectometry (TDR)" (Freeman 1996). We present and discuss theexperimental results of flow and cure monitoring of epoxy resin in a modeltransmission line, constructed with metal, using TDR. We also present experimentalresults when the transmission lines were constructed with carbon fiber cloth andcarbon fiber strands, which are typical material elements of advanced composites,expecting to avoid a deterioration of material property and an increase inmanufacturing cost caused by embedding of sensors.


2. Experimental setup and theoretical background

The experimental setup, for the experiments using a model transmission lineconstructed with metal, is shown in Figure 1. A microstrip line was constructedwith a straight brass line, 2 mm in diameter, and the bottom plate (35 x 25 cm) of analuminum box. The distance between the line conductor (the brass line) and theground conductor (the bottom plate) was set at 5 mm. Each end of the line wasconnected to the inner conductor of the coaxial receptacle screwed on each side ofthe aluminum box. The receptacle on the input side was connected to a digitizingoscilloscope with time domain plug-in (Agilent Technologies Model 54750 withModel 54754), through a coaxial cable with a characteristic impedance of 50W. A 50 W broad-band termination was connected to the receptacle on the terminalside.

As for the experiments of using carbon fiber as conductive elements, a PAN typecarbon cloth (Toray Industries, Inc., Torayca® Type 615IB, where carbon fiberstrand type T300B-1000 is woven in a density of 17.5 strands/25mm in twodirections perpendicular to each other.) was used as the ground conductor, and acarbon strand pulled out from that type of cloth was used as the line conductor.Experiments using a copper wire of diameter 0.1 mm as the line conductor were alsocarried out. The carbon cloth, 8 cm in width and about 35 cm in length, was placedon the bottom of an acrylic frame. The flange of a coaxial receptacle was glued tothe center of one end of the cloth using conductive epoxy. The center conductor ofthe receptacle was connected with one end of the carbon strand, also usingconductive epoxy. The distance between the line conductor and the ground

Figure 1. Experimental setup of a model transmission line constructed with metal


conductor was set at 2.5 mm. The receptacle was connected to the digitizingoscilloscope above mentioned. Spacers made of Teflon plates and silicone rubbersheets were set at each end of the line, to maintain the gap between the pair ofconductors and to prevent resin from flowing out. The terminal was short-circuitedby connecting the line conductor to the carbon cloth with a conductive adhesive tapeat the terminal.

The oscilloscope generated a sharp step voltage of 200 mV with 30 ps rise time,and voltage change with time at a fixed point inside the oscilloscope was measured.The measured time domain data were shown on the display of the oscilloscope, andrecorded by a computer. The full scale of the time axis, consisting of 1024 points,was set at 10 ns, which means that time resolution was about 10 ps.

When there are discontinuities in the characteristic impedance of thetransmission line, a reflection of the voltage signal occurs at each of the boundariesbased on the boundary condition of the electromagnetic field. The reflectance of thevoltage at the boundary of line 1 with characteristic impedance Z\ and line 2 withcharacteristic impedance Z2, Rn, is expressed as (Freeman 1996),

When the change in the voltage signal as a function of time (i.e., the "timedomain response) is monitored at a fixed point inside the oscilloscope, the reflectedvoltage is added to the measured voltage after a time delay corresponding to thetravelling time for the signal transmitting from the fixed point to the boundary andback (Freeman 1996). If there are other boundaries, the reflected signals at each ofthe boundaries are added one after another with time delay corresponding to thedistance to the boundary.

Because reflectance at a boundary depends on the values of the characteristicimpedance, Z, of the transmission lines of both sides of the boundary as indicated inEquation [1], the time domain response shows a stepwise rise or drop at thecorresponding time depending on the change in characteristic impedance at theboundary. Z is inversely proportional to the square root of the relative permittivity,£, of the material between the pair of conductors of the line (Sucher et al, 1963).Thus, the time domain response to a step voltage input signal provides informationon the discontinuity of dielectric properties of the materials between the pair ofconductors, such as resin flow or variation in cure stage, including information onthe position along the line. In practice, there is a transmission loss which originatesin the dielectric loss factor of the material between the conductors of the line andohmic loss of the conductors. Such loss gives rise to exponential relaxation of thestepwise rise or drop in the time domain response (Freeman 1996).

Bisphenol-A type epoxy resin (Epikote®828, Yuka-Shell Epoxy, Inc.), mixedwith an equivalent amount of diethylenetriamine as curing agent was used as thesample. It was put in and around the line and cured slowly at room temperature


without controlling the resin temperature. Teflon plates (1.0 mm thickness) wereused as barriers for the experiments of monitoring resin flow, existence of air andvariation in cure stage.


Figure 2 shows time domain responses to a step input signal at various curestages of the resin that filled the whole line. In the response before the line wasfilled with resin (denoted as "E " in the figure), the rise at point "a" is caused byreflection at the input, and the drop at point "b" is caused by reflection at theterminal. The impedance of the line, about 150 W., is higher than that of thereceptacle, cable or termination, which is 50 W. Therefore, voltage rose at the inputand dropped at the terminal, as can be expected from Equation [1]. When the linewas filled with resin before cure, the response caused by reflection at the inputbecame low and time between the rise and the drop became long, because thedielectric permittivity of the resin is higher than that of air. That is, the highpermittivity results in the reduction of impedance of the line, and the velocity of theelectromagnetic wave being transmitted along the line decreased as an inverselyproportional function of the real part of the square root of the complex permittivity(Sucher et al., 1963). As curing of the resin progressed, the time between the riseand the drop gradually decreased, and the response level gradually became high.Taking the level at 4 ns as an index, change in the level with time is plottedin Figure 3 together with the change in resin temperature. The peak increase

of the response level occurred beforethe peak temperature. This is probablybecause the peak temperature correspondedto almost the end of the main bridgingreaction of the epoxy. Similar results

Figure 2. Time domain response, toa step signal, from the microstripline filled with resin at various curestages.E: empty (no sample),1: 8 min after mixing,2: 68 min after (before cure),3: 138 min after (near the peaktemperature),4: 258 min after (after cure)

Figure 3. Change in amplitude of theresponse at 4 ns with time, during acuring process of the resin, for the samesample as Figure 2


had been obtained forfrequency characteristics(Urabe et al., 2000), andthe time domain responsemonitors the curing statein a similar way. Theovershoot of the responseat the input is caused byair between the receptacleand resin end. The timeconstant of voltagereduction after thisovershoot decreased ascuring progressed. Thesechanges with the progressof curing are caused by decreases inboth the permittivity and the lossfactor of the resin with the progress ofcuring (Urabe et al., 2000).

Figure 4 shows the results whenthe line was gradually filled withresin from the terminal side of theline. The drop in the responsecorresponds to the resin front.Therefore, the time between the riseand the drop in the response, which isindicated as "Air" for the case of "5cm filled" in the figure, correspondsto the time required for a round tripfrom input to the resin front, and thelength can be evaluated using thevelocity of light.

Figure 4. Time domain response from the microstripline gradually filled with resin from the terminal side

Figure 5 shows the response whenthere was air in the resin, a model ofpoor impregnation of resin. Becausethe characteristic impedance of theline became higher at the air part, itwas detected as a local peak at thecorresponding position of theresponse. A 1 cm long air part was clearly detected. The peak was sharper aftercure than before cure, because of the decrease of the loss factor as curingprogressed. When there were two air parts (Figure 5 (b)), they were separatelydetected at corresponding positions on the time axis.

Figure 5. Time domain response fromthe microstrip line filled with resinhaving air part(s)


Figure 6 shows the responsewhen there was a variation in the curestage. When there was a variation inthe cure stage, it was clearly detected,as can be seen in the solid line of thefigure where the drop indicated witharrow corresponds to the boundary.The variation in the cure stage isclearly detected, includinginformation on the position of theboundary. After all the resin hadcome to the end of cure reaction, thedrop disappeared showing uniformproperty of the resin.

Figure 6. Time domain response from themicrostrip line filled with resin having avariation in cure stage

results

(a) Line conductor: CF strandxl (Line length=27cm)

All thepresented above wereobtained using the modeltransmission lineconstructed with metal.Next, we show resultswhen carbon fiber wasused as conductiveelements. Figure 7 showstime domain responses fortwo different conditionsof line conductor, acarbon fiber strand and acopper wire of 0.1 mmdiameter, of the microstripline gradually filled withresin from the terminalside. Carbon cloth wasused as ground conductorin both of the cases.When a copper wire wasused as the line conductor(Figure 7 (b)), a similar

response as in Figure 4was obtained. This meansthat the carbon clothworks as a groundconductor of themicrostrip line in the sameway as a metal plate. In contrast, when a carbon fiber strand was used as the line

(b) Line conductor: 0.1 mm(f copper (Line

Figure 7. Time domain response from the microstripline, where carbon cloth was used as groundconductor, gradually filled with resin from theterminal side


conductor (Figure 7 (a)), the response showed an exponential increase after the riseat the input. This change in response can be attributed to the higher electricalresistance of the carbon fiber strand compared to the copper wire. Although therewas such a difference in the response, the resin flow front was clearly detected evenwhen a carbon fiber strand was used as the line conductor of the .microstrip line.However, to obtain a clearer response, it is desirable to use a thin metal wire as theline conductor. Progress of cure, an existence of an air part in the resin and avariation in cure stage could also be monitored when a carbon fiber was used as theconductive element of a transmission line, although the responses were less clearthan those obtained when the line conductor was metal.

4. Conclusion

To develop a new monitoring tool of resin flow and cure in the manufacturingprocess of composites, the use of time domain response to a step input signal froman electromagnetic wave transmission line filled with resin was investigated for amodel transmission line. Resin flow, progress of cure, existence of air and variationin cure stage were successfully detected together with explicit and direct informationon the position along the line. We also proposed and experimentally investigatedthe use of carbon fiber as conductive elements, and found that a carbon cloth workedthe same way as metal for the ground conductor. When a carbon fiber strand wasused as the line conductor, the response was somewhat affected by the higherelectric resistance of the carbon fiber.

Further investigations are necessary into the situations in which the materialproperty changes gradually and has no clear boundaries, the effects of bends andchanges in the separation of the conductors of the line, and the use of other types oftransmission lines, such as parallel-wire, parallel plates or coplanar. Quantitativeanalysis of the relationships between the response and material properties betweenthe conductors is also important to extract more useful and detailed information.The methodology proposed in the present paper is also applicable to healthmonitoring utilizing electric conductivity (Schulte 2001), and can give informationon the position of damage.

References

Banks W. M, Dumolin F., Hayward D., Pethrick R. A and Li Z. C., "Non-destructiveexamination of composite joint structures: a correlation of water absorption and high-frequency dielectric propagation", Journal of Physics D., Vol.29, 1996, p.233-239.

Chen J. Y., Hoa S. V., Jen C. K and Wang H., "Fiber-optic and ultrasonic measurements forin-siru cure monitoring of graphite/epoxy composites," Journal of Composite Materials,Vol.33, 1999, p. 1860-1881.


Ciriscioli P. R and Springer G. S., "An expert system for autoclave curing of composites",Journal of Composite Materials, Vol.25, 1991, p. 1542-1587.

Freeman J. C., "Fundamentals of Micro-wave Transmisison lines," New York, John Wiley &Sons, 1996.

Kenny J. M., "Application of modeling to the control and optimization of compositesprocessing", Composite Structures, Vol.27, 1994, p. 129-139.

Mijovic J., Kenney J. M., Maffezzoli A., Trivisano A., Bellucci F and Nicolais L., "Theprinciples of dielectric measurements for in situ monitoring of composite processing",Composites Science and Technology, Vol.49, 1993, p.77-90.

Motogi S., Itoh T and Fukuda T., "Multi-functional sensor properties and 2-dim flowdetection for RTM", Proceedings of the 6th Japan International SAMPE Symposium,1999, p. 1033-1036

Ohshima N., Aoki K., Motogi S and Fukuda T., "Cure monitoring of fiber reinforced plasticsby piezoelectric ceramics", Materials Science Research International, Vol.SPT-2, 2001,p.89-94.

Osaka K., Kosaka T., Asano Y and Fukuda T., "Off-axis strain monitoring of FRP laminatesin autoclave molding", Materials Science Research International, Vol.SPT-2, 2001,p.105-109.

Schulte K., "Electrical properties of polymer composites", Composites Science andTechnology, Vol.61,2001, p.799.

Shepard D. D., Day D. R and Craven K. J., "Application of dielectric analysis for curemonitoring and control in the polyester SMC/BMC molding industry", Journal ofReinforced Plastics and Composites, Vol.14, 1995, p.297-308.

Shwab S. D., Levy R. L and Glover G.G., "Sensor system for monitoring impregnation andcure during resin transfer molding", Polymer Composites, Vol.17,1996, p.312-316.

Sucher M., Fox J., Handbook of microwave measurements, New York, Polytechnic Press,1963.

Urabe K., Takahashi J., Tsuda H and Kemmochi K., "Cure monitoring of matrix resin withhigh-frequency electromagnetic wave transmission line", Journal of Reinforced Plasticsand Composites, Vol.19, 2000, p. 1235-1250.

Yamaguchi Y., Yoshida M., Jinno M., Sakai S., Osaka K and Fukuda T., "Autoclave curemonitoring of CFRP laminates by embedded sensors comparing with cure prediction bykinetics", Proceedings of the 6th Japan International SAMPE Symposium, 1999, p. 1037-1040.

Effects of orientation errors on stiffnessproperties of composite laminates

A. Vincenti, P. Vannucci, G. Verchery, F. Belaid

LRMA-ISAT49, rue Mademoiselle Bourgeois58027NeversFrance

[email protected]. [email protected]@u-bowgogne.fr

ABSTRACT: In this paper, we present a study on the effects of layer orientation defects on theproperty of quasi-homogeneity for composite laminates: we suggest a measure of thedeviation from quasi-homogeneity, introducing the concept of degree of quasi-homogeneity.

We then present the results of a wide numerical analysis in the case of orientation errorsrandomly distributed on the stacking sequence.

We developed the theoretical and numerical calculations thanks to the polar method ofrepresentation of fourth order tensors introduced by Verchery.

KEY WORDS: defects; thermo-mechanical properties; laminates; quasi-homogeneity;uncoupling; stiffness properties.


1. Introduction

The general equations of the Classical Laminated Plate Theory (CLPT),describing the thermo-mechanical behaviour of a composite laminate, are (Jones,1975):

where N and M are the tensors of in-plane forces and bending moments, e0 thetensor of in-plane strains in the middle plane, % the tensor of curvatures, TO thedifference of temperature of the middle plane with respect to a non-strain condition,Dt the difference of temperature between the upper and lower face and h thethickness of the plate. A and D are the tensors describing the in- and out-of-planerigidity behaviours of the plate, while B represents the coupling between these twobehaviours. U, V and W have the same meaning as A, B and D, with respect to theefforts produced by thermal strains.

A composite laminate is said to be quasi-homogeneous when it is uncoupled andit has the same in-plane and bending behaviour. Using the symbols of the CLPT, wecan express the property of quasi-homogeneity of a composite laminated plate for itselastic behaviour:

Only the exact matching of stacking sequences to the theoretical solutionsassures the desired property for the laminate. Nevertheless, in the practice somedefects may affect the production of a composite laminate, so that the real laminatehas different characteristics than the designed one.

In this paper we deal with orientation defects, which are common laminateimperfections, and we investigate how they affect the property of quasi-homogeneity of a laminate. A study already exists on the effect of orientation errorson uncoupling of composite laminates (Belaid et al, 2001; Vannucci, 2002).

We consider the most general case of laminates composed by identical plies. Wepropose a measure of the deviation from quasi-homogeneity by introducing theconcept of degree of quasi-homogeneity, and we develop a wide numerical analysisin the case of randomly distributed orientation errors in the stacking sequence. Westudy also the influence of characteristic parameters (kind of material, number of


plies, etc.) on the deviation from the designed property. We propose an empiricalfunction to describe the dependence of the degree of quasi-homogeneity from theseparameters.

We developed the study and the numerical analysis using the polar method ofproposed by Verchery (Verchery, 1979), and successfully used by him and co-workers in different researches (Verchery et al., 1979 to 2002).

2. The degree of quasi-homogeneity

Let L be any one of the preceding tensors. In our study we used the followingquantity L, which is an invariant for L and has the same properties as the norm of atensor (Kandil et al., 1988). We will refer to it as the norm of the tensor L:

In Eq. [3], T and R are the so-called polar components of a second rank tensor,while TO, T], R0 and R1 are those of a fourth rank tensor.

The condition for elastic quasi-homogeneity is to have B and C equal to zero,that is equivalent to have their norms B and C equal to zero. On the contrary, whenB and C are non-zero, we can estimate the deviation from quasi-homogeneity if wemeasure the pair of values (B, Q. In fact, B and C have the same properties as thenorm and when we compare these quantities for two different laminates, we can saywhich one is more uncoupled and which one has more similar in-and out-of-planebehaviours.

Nevertheless, B and C are not homogeneous quantities: we can not directlycompare their values and say that the deviation from quasi-homogeneity for alaminate depends more on its uncoupling than on the difference between the in-plane and bending behaviours, or viceversa.

Hence we decided to use the pair (b, in place of (B, C), where b and l are theratios:

bmax and Cmax are the maximum values of the norms B and C for a givennumber of plies and for a given material of the elementary ply.


Quantities b and l are non-dimensional, so that we can compare pairs of values(b, l) for laminates with different numbers of layers and composed by differentmaterials, b and l are also homogeneous and we can compare their values for asingle laminate to establish the prevailing influence of uncoupling or of thedifference between in- and out-of-plane behaviours on the deviation from quasi-homogeneity. Moreover, band l are normalized quantities and their value belong tothe range [0, 1]:

Hence, we can classify laminates on a scale of quasi-homogeneity with thevariation of b and l.

If we consider the pair (b ,g) as the representation of a vector in a plane, we canuse the norm and the orientation tgq of this vector as a degree of quasi-homogeneity:

Analogous considerations can be made for Kand Z, (Vincenti et al., 2002).The stacking sequences, for a given number of layers and kind of material, that

give Bmax or Cmax have been theoretically determined (Vannucci, 2002, Vincenti etal., 2001 and 2002), together with formulas for their computation. Equations givingthe degree of quasi-homogeneity in the case of only one layer affected by anorientation error have also been found, and the reader is addressed to the referencesabove. Here, we consider the more realistic case when a random error affects theorientation of each layer of a laminate. In this case, an analytical formula giving thedegree of quasi-homogeneity cannot be found. So, we made a wide numericalinvestigation in order to study the variation of the degree of quasi-homogeneity withthe orientation defects and to assess the influence of the different parameters:number of plies of the laminate, material of the elementary layer, orientations of thelayers, magnitude of the errors. The results of this analysis are presented in thefollowing section.


3. Numerical analysis

We made each calculation considering a vector E of angular defects: its genericcomponent ek represents the error on the orientation of the k-th ply of the laminate.All the components of E are statistically independent. The error ek is normallydistributed around the theoretical angle for each ply. We introduce the characteristicangle Y in place of the standard deviation a£ to describe the distribution: each ek

belongs to range [-Y, Y] with a probability of 95%. Hence it is 0E =1.96Y.

As E is a random error vector, we calculated each value for b and l as a mean ona population of np tests: we chose np = 10000 to have a good stability of results.

We made tests on quasi-homogeneous sequences belonging to the set of quasi-trivial solutions found by Vannucci and Verchery (2001). To have the most generalresults, we chose both non-symmetrical and symmetrical sequences. We consideredthe case of laminates with two theoretical orientation angles, 0° and oc, the numberof plies for each angle is generally not the same.

We describe the influence of the variation of the parameters on thisphenomenon: orientation angle a, characteristic error angle Y, number of layers n,ratio P=R0/R1 for the elementary layer; p is the only parameter needed to describethe material properties (Vannucci, 2002).

3.1 Influence of the characteristic angle Y

We studied the influence of the characteristic angle Y in the case of variousquasi-homogeneous stacking sequences, chosen in the set of quasi-trivial solutions,with different number n of plies and composed by elementary materials withdifferent characteristic ratios p. We made tests with various values of the orientationangle q too.

We found that the variation of is linear with Y, and the slope of the curvedecreases with n. In Fig. 1 we illustrate the variation of and tgq with Y in the caseof 8- and 20-ply laminates with a = 30° and p - 0.01.

3.2 Influence of the orientation angle a

We studied the influence of the orientation angle a in the case of laminates withdifferent number n of layers and for elementary materials with different p. We fixedY = 5°, while the variation of a is between 0° and 90°.

We found that and tgq are completely independent of a. For this reason, wemade all successive tests with a fixed value for the orientation angle and we chose anon-standard orientation, a - 30°.


Figure 1. Variation of andtg(0) with Y(a= 30°, p = 0.01).

3.3 Influence of the number of layers n

We studied the influence of the number of layers n in the case of theoreticalquasi-homogeneous laminates composed by materials with different p. We fixedY = 5° and a = 30°.

In Fig. 2 we show the results for and tgq in the case of laminates withp = 0.01. We found that the dependence of is described by two different curvesfor even or odd n. We remark that on logarithmic axes b , l and are linear with n,both for even n and for odd n.

Figure 2. Variation of ln( ) and ln(tg(l) with ln(n) (Y= 5°; p = 0.001, a= 30°).


3.4 Influence of the characteristic ratio p of the elementary material

We studied the influence of the material characteristic ratio p = R0/R1 forlaminates with different number of layers, n. We fixed Y= 5° and a = 30°.

In Fig. 3 we show the results for and tgq We notice that on logarithmic axesthe curve representing has a step variation towards zero, the upper value of thestep being about double than the lower value. Hence, the influence of orientationerrors on the deviation from quasi-homogeneity is much more important forlaminates composed by materials with p > 1. It is maximum when p tends toinfinity, which is the case of plies reinforced by balanced fabrics, that have R1=0.

Figure 4. Variation of ln( and ln(tg q) with ln(p) (Y= 5°, a =30°).

4. Overall description of the results

As the degree of quasi-homogeneity is completely independent on the theoreticalorientation angles of the stacking sequence, we suggest to represent the variation of£ with p and n by the function f(p,n) for a given value of Y :

Curves in Fig. 1 show that and tgq have a linear dependence from Y . Hence,we can represent the variation of with all the three parameters Y, p, n by theempiric function:


We determined numerically the coefficients ai, of Eq. [7] for the representationof . In Table 1 we show their values in both the cases of even n and odd n. Fig. 4shows the empirical function in the case of even n and Y = 5°.

even nodd n

a1

-0.38-0.34

a2

-0.06-0.07

a3

-0.49-0.50

a4

0.850.78

a5

-0.10-0.11

Table 1. Coefficients a, of the empiric function ln f(n,p).

Figure 5. The function f(n,p) (Y= 5° even n).

5. Conclusion

In this paper we describe the influence of orientation errors on compositelaminates designed to be quasi-homogeneous. First, we introduce the concept ofdegree of quasi-homogeneity; then, we show the results of a wide numerical analysisin the case of orientation errors randomly distributed over all the layers of alaminates. We notice that the theoretical stacking sequence is not a relevantparameter for the deviation from quasi-homogeneity. On the contrary, othersparameters affect the deviation from quasi-homogeneity. In fact, there is a lineardependence on the amplitude of orientation errors, described by the characteristic


angle Y . The dependence on the number of layers n is still linear in a logarithmicscale. In a logarithmic scale there is a step variation of the degree of quasi-homogeneity with p. materials with higher values of p, such as plies reinforced bybalanced fabrics, are more sensible to orientation errors. Finally, we propose asynthetic description of the results by mean of an empiric function, which describesthe dependence of the degree of quasi-homogeneity upon all the parameters Y ,p,n.

6. References

Belaid F., Vannucci P., Verchery G., "Numerical investigation of the influence of orientationdefects on bending-tension coupling of laminates", Proceedings of ICCM 13, Beijing,June 2001, paper 1406.

Jones R. M., Mechanics of Composite Materials, USA, Taylor & Francis, 1975.

Kandil N., Verchery G.,"New methods of design for stacking sequences of laminates",Proceedings of CADCOMP 88, C. A. Brebbia, W. P. De Wilde and W. R. Blain eds.,Computational Mechanics Publications and Springer Verlag, Southampton, 1988, p. 243-257.

Vannucci P., Verchery G., "A special class of uncoupled and quasi-homogeneous laminates",Composites Sciences and Technology, vol. 61,2001, p. 1465-1473.

Vannucci P., Verchery G., "Stiffness design of laminates using the polar method",International Journal of Solids and Structures, vol. 38,2001, p. 9281-9294.

Vannucci P., "On bending-tension coupling of laminates", Journal of Elasticity, 2002 (toappear).

Verchery G., "Les invariants des tenseurs d'ordre quatre du type de I'&asticite", ProceedingsofEuromech Collegium 115, Villard-de Lans, 1979, Paris, CNRS Editions 1982, p. 93-104 (in French).

Verchery G., "Designing with anisotropy. Part 1: Methods and general results for laminates",Proceedings of ICCM 12, Paris, 1999, paper 734.

Vincenti A., Vannucci P., Verchery G., "Anisotropy and symmetry for elastic properties oflaminates reinforced by balanced fabrics", Composites Part A, vol. 32, 2001, p. 1525-1532.

Vincenti A., Vannucci P., Verchery G., Belaid F., "Effetti degli errori di orientatzione sullaquasi-omogeneita dei laminati in composito", Proceedings of AIMETA XV(15th Congressof Theoretical and Applied Mechanics), Taormina, Italy, 2001 (in Italian).

Vincenti A., Vannucci P., Verchery G., "Influence of orientation errors on quasi-homogeneityof composite laminates", Composite Science and Technology, 2002 (submitted).

Mechanical properties of pultruded GFRPsmade of knitted fabrics

Hiroshi Fukuda* — Hirokatsu Wakabayashi**Koshiro Hayashi*** — Gen Ohshima***

* Department of Materials Science and TechnologyTokyo University of Science2641 Yamazaki, Noda, Chiba 278-8510, Japan.

fukuda@rs. noda. tus. ac.jp

** Former Graduate Student at Tokyo University of SciencePresent address: TOSTEM Co Ltd.

*** Asahi Glass Matex Co. Ltd.1-2-27 Miyashita, Sagamihara, Kanagawa 229-1112, Japan.

ABSTRACT: Pultrusion is a promising fabrication method of FRPs for civil engineering use. Toconstruct large-scale structure like a bridge, lateral strength and modulus are also requestedin addition to the longitudinal properties. To this end, knitted fabrics are expected asreinforcements for pultrusion. However, data of knit-fabric pultruded composites are very fewbecause this combination is relatively new. This paper reports test data of the knit-fabricpultruded plate. The superiority of the knitted fabric to conventional woven cloth isdemonstrated. Effects of matrix resins are also discussed.

KEY WORDS: knit fabric, pultruded plate, tensile properties, direction dependency, unsaturatedpolyester resin, vinylester resin


1. Introduction

Fiber reinforced plastics (FRP) are nowadays used in various engineering fields.Among them, the field of civil engineering is one of the most promising areas toapply FRPs because of their corrosion resistance and long term durability (Liao etal. 98), possibility of constructing large-scale structures by joining same-sizeelements, their light weight, and so on.

Pultrusion is an attractive fabrication process where endless products of uniformand arbitrary cross sectional shape can be made with uniform quality (Roux et al.98). Therefore, this pultruded material is suitable for architectures of civilengineering and there are not a few examples of bridges made of this material.

In the pultrusion process, glass fibers are mainly oriented to the axial direction,which leads highly anisotropic products. Due to high anisotropy, the lateralproperties of pultruded materials are relatively inferior and to increase the lateralproperties, a combination of glass mat and glass roving is commonly used. Glassroving cloth is also tried as a constitutive material for pultrusion although the rovingcloth is not necessarily appropriate as will be described later.

Recently, another raw material, knitted fabric (DeWalt et al. 94) have beendeveloped and this material is expected as a promising candidate for pultrusion. Theconfiguration of the knitted fabric is schematically shown in Fig.l(a) whereasFig.l(b) is a typical roving cloth. The advantage of the knitted fabric is fullydiscussed by DeWalt et al.

In the present paper, tensile properties of pultruded plates where knitted fabric isused as reinforcements are examined. As a reference, unidirectionally reinforcedplates as well as plates made of roving cloth are tested and the superiority of theknit-fabric pultruded material will be demonstrated. As for matrix materials, bothunsaturated polyester resin and vinylester resin are examined.

2. Experiment

2.1. Materials

Reinforcements used here are unidirectional glass roving (denoted by "U"), 0/90crossply of knit fabric ("C"), quasi-isotropic layout of knit fabric ("Q"), and plainwoven roving cloth ("RC"). The quasi-isotropic plates were fabricated by stacking0/90 layer and ±45 layer in turn. As for the matrix resin, both unsaturatedpolyester ("UP") and vinylester ("VE") are employed. Using these raw materials,


total 8 types of pultruded plates of the width of 200mm and the thickness of 2.5-3mm were fabricated by Asahi Glass Matex Co. Ltd. The details of these plates arelisted in Table 1. Each orientation rate was measured by burning out the resin andweighing glass fibers of each direction. Unfortunately, the "Q" plates with UP resinwere not quasi-isotropic, that is, the orientation rate in the 0 and 90 digree directionwas quite small; this is probably due to some mistake during fabrication.

(a) knitted fabric (b) roving cloth

Figure 1. Knitted fabric and roving cloth

Table 1. Details of pultruded panels

thicknessresin stacking sequence

VE unidirectional (U)

crossply (C)

roving cloth (RC)

quasi-isotropic (Q)

UP unidirectional (U)

crossply (C)

roving cloth (RC)

quasi-isotropic (Q)

0 roving olny

(0, 90)5

(plain woven #600)8

(± 45/0, 90)3/(± 45)

0 roving olny

(0, 90)5

(plain woven #600)g

(± 45/0, 90)3/(± 45)

t(mm)

2.5

2.53.2

32

2.5

3.2

3.2

32

f(/°)

60.7

61.7

58.8

58.8

58.8

orientation rate

0(%)

100

51.3

50.0

26.6

100

52.1

50.0

17.7

± 45 (%)

0

0

0

47.9

0

0

0

63.6

90 (%)

0

48.7

50.0

25.5

0

47.9

50.0

18.7

2.2. Tensile test

From the above 8 types of plates, tensile test coupons were cut in accordancewith JIS K 7054-1995. To examine the effect of anisotropy, test coupons were madein the 0, 45, and 90 degree directions from the machine direction. The specimen


size was 10mm in width and 200mm in length. A pair of GFRP tabs was glued onboth sides of both ends of each specimen and actual gage length of the specimenwas 100mm.

Figure 2. Young's modulus (matrix : VE)

A pair of two-axis strain gages (gage length = 2mm) was glued on both surfacesat the center of each specimen. Tensile tests were conducted using an Instron-typetesting machine at the crosshead speed of Imm/min. Applied load and strains weresaved in a personal computer using a data logging system at an interval of 1 s.


3.1. Young's modulus

Figure 2 summarizes Young's modulus of each material at each direction wherevinylester resin is used as matrix material. Unidirectional composites (U) havestrong anisotropy whereas Young's modulus of quasi-isotropic composites (Q) isalmost the same in three directions tested. As far as Young's modulus is concerned,the crossply composites (C) and the roving cloth composites (RC) exhibited similarvalues. These results are all reasonable and therefore, of little interest. Data ofunsaturated polyester resin showed similar tendency, although they are a little bitinferior to VE resin composites.


3.2. Strength

Figure 3 summarizes the strength of (a) VE-matrix composites and (b) UP-matrix composites. The overall tendency is the same as Young's modulus.

Figure 3. Tensile strength

Figure 4. Comparison of tensile strength betweencrossply (C) and roving cloth (RC) composites


Among data of Fig.3, data of knit fabric crossply composites and roving clothcomposites are picked up in Fig.4. These should be essentially the same because thefiber directions are 0 and 90 degrees and the amount of fibers is also the same.However, we can clearly see that there exists big difference between crossply androving cloth reinforced composites. That is, the transverse (90 degrees direction)strength of the roving cloth composites is very low comparing to the strength oflongitudinal direction nevertheless the fiber content is the same in both directions.This may be understood as follows: During the pultrusion process, fairly largetensile load is applied to the longitudinal roving and fibers of this direction tend tobecome straight. On the other hand, fibers in the transverse (width) direction sufferfrom no tensile load and the waviness of these fibers is more severe. If exaggerationis allowed, transverse fibers of the roving cloth has little role of reinforcement.

In the case of knit fabric, the longitudinal and transverse fiber bundles areessentially straight; they are merely knitted by thin polyester threads. Thus thetensile strength in the width direction remains so so, although some amount ofdecrease compared with the strength in the longitudinal direction is recognized. Onereason of this slight decrease of the strength is attributed to a little bitfewer fibercontents in the transverse direction (see Table 1, crossply). Another reason might bethe faint waving of the transverse rovings during the pultrusion process. Figure 5 isits evidence, which were taken after evaporating the matrix resin in a Muffle furnaceat 625C, 10 hours. Anyway, the characteristics that the transverse mechanicalproperty remains for knit fabric pultruded composites are very important for large-scale structures where the relatively large transverse strength is required.

Figure 5. Surface view after evaporating

3.3. Shearing modulus

In some practical cases, high shearing modulus or strength becomes important.For example, if these pultruded elements are connected with a bolt, large shearingstress takes place around the hole and the joint may collapse by "shear out" if the


shearing strength is small. The quasi-isotropic plate is designed for this purpose.Figure 6 depicts the shearing modulus, G, of each type where G was calculated fromthe tensile test of 45 degree coupons applying the following equation (Carlsson andPipes 87):

Figure 6. Shearing modulus of each type ofpultruded composites

where

and

From Fig.6, it is clear that the shearing modulus increases by inserting ± 45degree layers, although it is too primitive to describe. Again the superiority ofvinylester resin to unsaturated polyester resin is demonstrated, although we must beaware that the fiber volume fraction of each test panel is a little different each other.


4. Conclusions

During a series of experiments, data of knit-fabric pultruded composites wereaccumulated. Woven roving may not be suitable as a constitutive material forpultrusion because the transverse properties are pretty inferior. On the other hand,knitted fabric is a promising candidate where the decrease of the mechanicalproperties in the transverse direction is not so serious. Mechanical properties ofvinylester matrix composites were found to be superior to those of unsaturatedpolyester composites.

Acknowledgements

The authors thank Mr. Hiroshi Igarashi, Dr. Masaaki Itabashi and Dr. AtsushiWada for their assistance in experiments and preparing manuscripts.

References

Carlsson L. A. and Pipes R. B., "Experimental Characterization of Advanced CompositeMaterials" Prentice-Hall, 1987.

DeWalt P. L. and Reichard R. P., "Just How Good are Knitted Fabrics," J. ReinforcedPlastics and Composites, vol.13, 1994, p.908-917.

Liao K., Schultheisz C. R., Hunson D. L. and L. Brinson C., "Long-term Durability of Fiber-Reinforced Polymer-Matrix Composite Materials for Infrastructure Applications: AReview," J. Advanced Materials (SAMPE), vol.30, No.4, 1998, p.3-40.

Roux J. A., Vaughan J. G. and Shanku R., "Comparison of Measurements and Modeling forPultrusion of a Fiberglass/Epoxy 1-Beam," J. Reinforced Plastics and Composites, vol.17,1998, p. 1557-1578.

[Claude Bathias, Hiroshi Fukuda, Kyoshi Kemmoshi, (BookZZ.org)

Documents

Transcript of [Claude Bathias, Hiroshi Fukuda, Kyoshi Kemmoshi, (BookZZ.org)