2006 Krishnamurthy

CRANFIELD UNIVERSITY
DEFENCE COLLEGE OF MANAGEMENT AND TECHNOLOGY
ENGINEERING SYSTEMS DEPARTMENT
PHD THESIS
Academic Year 2005-2006
Sriram Krishnamurthy
PRESTRESSED ADVANCED FIBRE REINFORCED

COMPOSITES: FABRICATION AND MECHANICAL
PERFORMANCE
Supervisors: Dr RA Badcock and Professor JG Hetherington
February 2006
© Cranfield University 2006. All rights reserved. No part of this publication

may be reproduced without written permission of the copyright owner.
gÉ Åç ctÜxÇàá tÇw ZâÜâá
ABSTRACT
A
DVANCED COMPOSITE MATERIALS have high strength-to-weight ratios, corrosion
resistance and durability and are extensively used in aerospace, energy and defence
industries. This research concentrates on minimising the process-induced residual
stresses, and improving the fibre alignment of composites by employing a fibre prestress
methodology. A novel flat-bed fibre prestress methodology for autoclave processing of
composites was developed. This research investigates the effect of fibre prestress on 1)
residual stresses, 2) fibre alignment, 3) static tensile and compression properties and 4)
fatigue behaviour of composites.
Experimental results show that this prestress methodology, on a 16-ply unidirectional E-
glass/913 epoxy composite, reduces the residual strain of the composite from –600 µε to
approximately zero for a prestress of 108 MPa. The strains measured from optical fibre
sensors were in close agreement with those obtained using strain gauge. The results from
fibre alignment studies showed that fibre prestressing improved the fibre alignment from
20% of fibres aligned to 0° degree in non-prestressed composites to 75% of fibres aligned
to 0° degree in 108 MPa prestressed composites.
Findings have shown that prestressing is beneficial to the static compressive and tensile
performance of composites. The results show that fibre prestressing improves the fatigue
life and resistance to stiffness degradation in the low stress level fatigue region. Also a
change in static and fatigue damage mechanism was observed. The improvement in the
static and fatigue properties is due to the reduction in residual stresses and fibre waviness.
Overall the fibre prestressing methodology enhances the performance of composites by
increasing the resistance to static and fatigue loading. The thesis also suggests that there is
an existence of prestress limits to retain optimal material performance.
Keywords: fibre prestressing, polymer composites, residual strain monitoring, optical fibre
sensors, static tension, static compression, fatigue and composite damage mechanisms.
v
ACKNOWLEDGEMENTS
I would like to thank Professor Gerard Franklyn Fernando, for providing me with the
opportunity to pursue my research in the field of composites and for his invaluable
guidance and support. I would also like to thank Dr Rodney Badcock for his
encouragement and constructive criticism throughout my study.
I would like to thank Professor Brian Ralph for his valuable advice, prompt response and
suggestions relating to this thesis.
I would like to thank Professor John G Hetherington and Dr Mike Iremonger for their
valuable guidance and inspiration in theoretical modelling and also their pastoral care.
I would like to thank Dr Amer Hameed and Mr Mike Teagle for providing technical
assistance in design and development of prestress rig and for their unfailing encouragement
and support.
I would like to thank Dr Paul Robinson and Dr J Hodgkinson at the Department of
Aerospace Engineering, Imperial College, London for giving me permission to use their
facilities in order to undertake compression test of composites.
I would like to acknowledge Cranfield University [DCMT] for sponsoring this project.
Thanks are also due to Mr Brian Duguid, Mr Tony Fallon at the mechanical workshop,
DCMT for manufacturing the prestress rig.
I am grateful to the late Mr John Tetlow, Dr Thomas Butler, Dr G Kister and Dr
Balakrishnan Degamber for providing help in optical fibre sensors.
Thank you to Dr John Rock, Mr Rajanikanth Machavaram, Dr James Talbot and Mr
Samual Proctor for their help and encouragement throughout my period of study.
Thank you to Mr Joseph Meggyesi, Mr Gary Senior and Mr Russell Stracey at the
Department of Aerospace Engineering, Imperial College, London for assisting in specimen
preparation and compression testing.
I would like to thank Ms Maggie Keats, Mrs Ros Gibson, Mrs Alison McFee and Mrs Paula
Bentley for their administrative support and entertainment.
vii
ACKNOWLEDGEMENTS
I also thank the DCMT library staff especially Miss Raichel Daniels, Mr Mike Groves and
Ms Louise Tucker for their assistance in processing the many interlibrary loan requests and
quickly responding to my stack requests.
I am eternally grateful to my father Sri MS Krishnamurthy, my mother Srimathi Sundhari
and my sister Srimathi Yamini, for their endless support, encouragement and prayers.
To my friends Dr Vidyashankar Buravalla, Mr Suresh Jeyaraman, and Miss Darina Fišerová
from whom I drew much inspiration and encouragement; to them I express my deepest
gratitude.
I thank Ms Anna Ward for providing me with a homely environment.
viii
ix
TABLE OF CONTENTS
CHAPTER 1 Introduction .......................................................... 1
1.1 RESEARCH MOTIVATION AND RATIONALE ......................................................................................... 1
1.2 AIMS AND OBJECTIVES ................................................................................................................... 2
1.3 CONTRIBUTIONS............................................................................................................................. 3
1.4 WORK DISSEMINATIONS .................................................................................................................. 4
1.5 ORGANISATION OF THESIS ............................................................................................................... 5
CHAPTER 2 Literature review ................................................... 7
2.1 INTRODUCTION .............................................................................................................................. 7
2.2 PROCESSING OF COMPOSITES ......................................................................................................... 8
2.3 PROCESS-RELATED ISSUES .............................................................................................................. 10
2.4 RESIDUAL STRESSES ....................................................................................................................... 11

2.4.1. Techniques for measuring residual stresses.......................................................... 11
2.4.2. Prediction of residual stresses ................................................................................ 17
2.5 FIBRE WAVINESS ........................................................................................................................... 18

2.5.1. Techniques for measuring fibre alignment .......................................................... 20
2.6 TECHNIQUES TO MINIMISE PROCESS-INDUCED RESIDUAL STRESSES AND FIBRE WAVINESS ....................... 21
2.6.1. Optimisation of the cure cycle.............................................................................. 21
2.6.2. Low temperature cure systems ............................................................................. 22
2.6.3. Electron beam curing............................................................................................. 22
2.6.4. Tailoring interphase properties .............................................................................. 23
2.6.5. Shape memory alloys ............................................................................................. 24
2.6.6. Expanding monomers ............................................................................................ 24
xi
TABLE OF CONTENTS
2.6.7. Fibre prestressing ..................................................................................................... 25
2.7 FIBRE PRESTRESSED COMPOSITES .................................................................................................... 26
2.8 THEORETICAL ANALYSIS OF FIBRE PRESTRESSING ............................................................................... 27
2.9 EVALUATION OF EXISTING FIBRE PRESTRESS METHODOLOGIES ............................................................ 31

2.9.1. Dead-weight method ............................................................................................ 31
2.9.2. V-slot mechanical fastening method................................................................... 32
2.9.3. Filament winding ..................................................................................................... 33
2.9.4. Hydraulic cylinder prestress rig .............................................................................. 34
2.9.5. Horizontal tensile testing machine ........................................................................ 34
2.9.6. Fibre alignment rig .................................................................................................. 35
2.9.7. Biaxial loading frame method............................................................................... 36
2.10 STATIC MECHANICAL PROPERTIES ................................................................................................ 37

2.10.1. Tensile properties..................................................................................................... 37
2.10.2. Impact properties ................................................................................................... 42
2.10.3. Flexural properties ................................................................................................... 43
2.11 FATIGUE OF COMPOSITES............................................................................................................ 44

2.11.1. Fatigue life diagrams .............................................................................................. 45
2.11.2. Factors affecting the fatigue life of GFRP ............................................................ 46
2.11.3. Fatigue of prestressed composites ....................................................................... 47
2.12 DAMAGE MECHANISMS ............................................................................................................. 49

2.12.1. Static tension failure mechanisms......................................................................... 49
2.12.2. Static compression failure mechanisms ............................................................... 50
2.12.3. Fatigue damage mechanics ................................................................................ 52
2.13 SUMMARY ................................................................................................................................. 56
CHAPTER 3 Design and development of the prestress

methodology ........................................................................ 57
3.1 DESIGN CRITERIA ......................................................................................................................... 57

3.1.1. Unidirectional fibre prestressing............................................................................. 58
3.1.2. Monitor and measure the applied pre-load online............................................ 58
3.1.3. Measure and quantify the residual strain development in composites........... 58
3.1.4. Manufacture prepreg-based composites using autoclave.............................. 58
3.1.5. Use of conventional vacuum-bagging method................................................. 59
3.2 FLAT-BED PRESTRESS METHODOLOGY ............................................................................................. 59
3.3 DESIGN CALCULATIONS ................................................................................................................ 62

3.3.1. Load in the locking bolts ........................................................................................ 64
3.3.2. Shear stress in the bolt thread................................................................................ 64
3.3.3. Calculation of the angle of rotation..................................................................... 65
3.3.4. Pre-load calculation ............................................................................................... 66
xii
3.3.5. Rig thermal expansion-induced prestress ............................................................ 68
3.4 CONCLUDING REMARKS............................................................................................................... 70
CHAPTER 4 Experimental procedure ...................................... 71
4.1 MATERIALS .................................................................................................................................. 71
4.2 SENSORS ..................................................................................................................................... 72

4.2.1. Optical fibre sensors................................................................................................ 72
4.2.2. Electrical resistance strain gauges........................................................................ 72
4.2.3. Load Cell .................................................................................................................. 72
4.3 PREPARATION OF COMPOSITE ....................................................................................................... 72

4.3.1. Laminate sequence ............................................................................................... 72
4.3.2. Lay-up....................................................................................................................... 73
4.3.3. Embedding optical fibre sensors........................................................................... 74
4.3.4. Prestress methodology ........................................................................................... 75
4.3.5. Vacuum bagging ................................................................................................... 77
4.3.6. Autoclave processing of composites................................................................... 78
4.3.7. Coding of the composite panels .......................................................................... 79
4.4 EVALUATION OF PRESTRESS METHODOLOGY ................................................................................... 79

4.4.1. Load cell calibration............................................................................................... 80
4.4.2. Prestress rig evaluation ........................................................................................... 80
4.5 EVALUATION OF SENSORS ............................................................................................................. 81

4.5.1. Temperature and pressure tests ............................................................................ 81
4.5.2. Vacuum test ............................................................................................................ 81
4.5.3. Cure cycle test ........................................................................................................ 82
4.6 RESIDUAL STRAIN MONITORING ...................................................................................................... 83
4.7 DIFFERENTIAL SCANNING CALORIMETRY ......................................................................................... 85
4.8 QUALITY CONTROL AND SPECIMEN PREPARATION ........................................................................... 85

4.8.1. Quality control ......................................................................................................... 85
4.8.2. Specimen cutting and polishing ........................................................................... 87
4.8.3. Coding of composite specimens.......................................................................... 88
4.8.4. End-tabbing............................................................................................................. 88
4.9 MECHANICAL TESTING.................................................................................................................. 89

4.9.1. Quasi-static tensile test ........................................................................................... 89
4.9.2. Quasi-static compression test................................................................................ 90
4.9.3. Fatigue test .............................................................................................................. 91
CHAPTER 5 Evaluation of prestress methodology and sensors 93
5.1 RESULTS AND DISCUSSIONS ON PRESTRESS RIG EVALUATION .............................................................. 93
xiii
TABLE OF CONTENTS
5.1.1. Load cell calibration............................................................................................... 93

5.1.2. Prestress rig evaluation ........................................................................................... 95
5.2 PREPREG END-TAB JIG DESIGN AND EVALUATION ............................................................................ 98

5.2.1. End-tab jig design ................................................................................................... 99
5.3 RESULTS AND DISCUSSIONS ON SENSOR EVALUATION ..................................................................... 102

5.3.1. Effect of temperature on sensors ........................................................................ 102
5.3.2. Effect of pressure on sensors................................................................................ 104
5.3.3. Effect of vacuum on sensors................................................................................ 106
5.3.4. Effect of cure cycle on sensors ........................................................................... 107
5.4 CONCLUDING REMARKS............................................................................................................. 108
CHAPTER 6 Residual strain and fibre alignment in the

composites.......................................................................... 111
6.1 RESIDUAL STRAIN DEVELOPMENT IN NON-PRESTRESSED COMPOSITE ................................................. 112

6.1.1. Unidirectional composite ..................................................................................... 112
6.1.2. Cross-ply composite ............................................................................................. 116
6.2 RESIDUAL STRAIN DEVELOPMENT IN PRESTRESSED COMPOSITE .......................................................... 117
6.3 RESULTS OF FIBRE ALIGNMENT ...................................................................................................... 122
CHAPTER 7 Theoretical analysis of prestressed composites .. 127
7.1 INTRODUCTION .......................................................................................................................... 127
7.2 COMPOSITE STRAIN .................................................................................................................... 131
7.3 RESULTS AND DISCUSSIONS .......................................................................................................... 134

7.3.1. Residual strain ........................................................................................................ 134
7.3.2. Strain release ......................................................................................................... 136
CHAPTER 8 Static mechanical properties ............................ 143
8.1 STATIC TENSILE RESULTS AND DISCUSSION ...................................................................................... 143

8.1.1. Strain at deviation from linearity ......................................................................... 143
8.1.2. Tensile strength and modulus .............................................................................. 147
8.2 POST-TENSILE FAILURE ANALYSIS ................................................................................................... 149

8.2.1. Macroscopic examination................................................................................... 149
8.2.2. Microscopic examination .................................................................................... 150
xiv
8.3 STATIC COMPRESSION RESULTS AND DISCUSSION ........................................................................... 155
8.4 COMPRESSIVE STRENGTH COMPARISON ....................................................................................... 158
8.5 POST-COMPRESSIVE FAILURE ANALYSIS ......................................................................................... 158

8.5.1. Macroscopic examination................................................................................... 158
8.5.2. Microscopic examination .................................................................................... 161
CHAPTER 9 Fatigue behaviour .............................................. 169
9.1 TENSION-TENSION FATIGUE .......................................................................................................... 169

9.1.1. Fatigue of non-prestressed composites ............................................................. 169
9.1.2. Fatigue of prestressed composites ..................................................................... 171
9.1.3. Statistical analysis of fatigue data ...................................................................... 175
9.1.4. Fatigue damage monitoring ............................................................................... 181
9.1.5. Post-failure analysis ............................................................................................... 188
9.2 TENSION-COMPRESSION FATIGUE................................................................................................. 195

9.2.1. Fatigue of non-prestressed composite............................................................... 195
9.2.2. Fatigue of prestressed composites ..................................................................... 196
9.2.3. Statistical analysis of fatigue data ...................................................................... 199
9.2.4. Fatigue damage monitoring ............................................................................... 205
9.2.5. Post-failure analysis ............................................................................................... 209

9.3.1. Tension-tension fatigue......................................................................................... 219
9.3.2. Tension-compression fatigue ............................................................................... 220
CHAPTER 10 Conclusions and directions for future research 221
10.1 FIBRE PRESTRESS METHODOLOGY ............................................................................................... 221
10.2 EVALUATION OF PRESTRESS METHODOLOGY AND SENSORS ........................................................... 222
10.3 RESIDUAL STRAIN DEVELOPMENT ................................................................................................ 223
10.4 THEORETICAL ANALYSIS OF PRESTRESSED COMPOSITES.................................................................. 223
10.5 FIBRE ALIGNMENT ..................................................................................................................... 223
10.6 STATIC MECHANICAL PROPERTIES .............................................................................................. 224
10.7 FATIGUE BEHAVIOUR ................................................................................................................ 225
10.8 DIRECTIONS FOR FUTURE RESEARCH ........................................................................................... 227
CHAPTER 11 References ...................................................... 229
xv
TABLE OF CONTENTS
APPENDIX A Bi-axial loading frame method ………………….. 241

APPENDIX B Flat-bed prestress methodology ………………… 245
APPENDIX C Static tensile and compressive results ………… 251
APPENDIX D Fatigue test results …………………………………… 256
xvi
LIST OF FIGURES
Figure 2-1: Schematic representation of the principle of composite processing. .............. 9
Figure 2-2: Typical autoclave cure cycle for prepreg-based composites. a) vacuum
application, b) heating cycle, c) isothermal cure and d) cooling cycle. ................ 10
Figure 2-3: Schematic illustration of the extrinsic Fabry-Pérot interferometric sensor

design. .................................................................................................................................. 13
Figure 2-4: Interference fringe from an EFPI sensor................................................................. 14
Figure 2-5: Principle of operation of an optical-fibre Bragg grating. .................................. 15
Figure 2-6: Fibre waviness in cross-ply laminate...................................................................... 19
Figure 2-7: Ellipse resulting from the plane section of a right circular cylinder. ................. 20
Figure 2-8: Ring opening mechanism of expanding monomers.......................................... 25
Figure 2-9: Schematic illustration of the fibre prestressing principle. 1) prestressing the
fibres in the uncured matrix, 2) matrix curing and formation of fibre/matrix bonding
and development of tensile residual stress in the matrix and 3) release of fibre
prestress at ambient temperature, which induces compressive stress to the matrix.
............................................................................................................................................... 26
Figure 2-10: Dead weight method. 70 ....................................................................................... 31
Figure 2-11: V-slot mechanical fastening method.71 .............................................................. 32
Figure 2-12: Filament winding.78 ................................................................................................. 33
Figure 2-13: Hydraulic cylinder prestress rig.74 ......................................................................... 34
Figure 2-14: Horizontal tensometer prestress rig.76 .................................................................. 35
Figure 2-15: Fibre alignment rig.77 .............................................................................................. 35
Figure 2-16: Biaxial loading frame method. 80 ......................................................................... 36
Figure 2-17: Influence of pre-load on the tensile strength and modulus of glass/polyester
composites70........................................................................................................................ 38
xvii
LIST OF FIGURES
Figure 2-18: Tensile strength and modulus as a function of prestress (Zhao and
Cameron77).......................................................................................................................... 39
Figure 2-19: Cross-sections of Zylon/epoxy composite samples shows the effect of fibre
prestress on the distribution of fibre in epoxy matrix. (a) <70 MPa prestress, (b) 280
MPa prestress, (c) 700 MPa prestress, and (d) 1000 MPa prestress (Huang et al. 83).40
Figure 2-20: Number of initial transverse cracks versus strain, which was obtained by light
microscopy. The mean values of four tests are given here and the data was taken
from the specimen edge (Schulte and Marissen71). ..................................................... 41
Figure 2-21: Matrix cracks induced in [0P/90]T specimens as a function of prestress level
(Tuttle et al .74). .................................................................................................................... 42
Figure 2-22: Influence of prestress on impact strength of composites76. ............................ 43
Figure 2-23: Flexural strength and modulus as a function of fibre prestress (Motahhari
and Cameron79). ................................................................................................................ 44
Figure 2-24: Tension-tension fatigue life diagrams of glass/epoxy composites. ................ 45
Figure 2-25: Fatigue crack propagation of aluminium alloy, CARALL laminates as cured
and prestrained (Lin et al.95) ............................................................................................. 48
Figure 2-26: Schematic Illustration of tensile failure of composites with weak and strong
interfaces. ............................................................................................................................ 49
Figure 2-27: Compression failure modes of composites. ....................................................... 50
Figure 2-28: Schematic of a kink-band formation from initial fibre misalignment. Kink
orientation, α, and the boundary orientation, β, are also indicated. ........................ 51
Figure 2-29: Schematic representation of the development of damage in cross-ply

laminates during the fatigue life.88 ................................................................................... 53
Figure 2-30: Fatigue damage mechanisms in unidirectional composite under tensile

loading parallel to fibres. (a) fibre breakage and interfacial debonding, (b) matrix
cracking, (c) interfacial shear failure.107 ......................................................................... 54
Figure 2-31: Fatigue-life diagram for unidirectional composites under tensile loading
parallel to fibres.107.............................................................................................................. 55
Figure 3-1: Schematic Illustration of flat-bed prestress methodology. ................................ 60
Figure 3-2: Plan view of the flat-bed prestress methodology. .............................................. 60
Figure 3-3: Schematic illustration of pre-fabrication stages involved prior to clamping the
prepreg laminate onto the FBPM..................................................................................... 61
Figure 3-4: Silicone rubber spacer arrangement to enable conventional vacuum

bagging................................................................................................................................ 62
Figure 3-5: Free body diagram of FBPM. .................................................................................. 64
Figure 3-6: Angle of rotation of the load screw as a function of prestress. ........................ 66
Figure 3-7: Free body diagram of FBPM. .................................................................................. 67
xviii
Figure 3-8: Schematic illustration of the deflection (δ) caused by the moment (M) in the
cantilever beam. ................................................................................................................ 67
Figure 3-9: Fibre tensile stress as a function of applied pre-load. ........................................ 68
Figure 3-10: Model representation of prepreg clamped to prestress rig. ........................... 69
Figure 4-1: Vacuum jig used during prepreg lamination....................................................... 73
Figure 4-2: Prepreg lay-up process using vacuum assisted jig. 1) Prepreg ply holding to
the vacuum jig and 2) Prepreg ply stacking using vacuum jig................................... 74
Figure 4-3: Side view of the End-tab Jig. .................................................................................. 76
Figure 4-4: Top view of the End-tab jig. .................................................................................... 76
Figure 4-5: Position and alignment of ERSG sensors with embedded optical fibre sensors.
............................................................................................................................................... 77
Figure 4-6: Detail of vacuum bag lay-up................................................................................. 78
Figure 4-7: The cure cycle for E-glass/913 epoxy prepreg system. ...................................... 79
Figure 4-8: Position of ERSG sensors in steel plate. .................................................................. 80
Figure 4-9: Position of sensors in the laminate. ........................................................................ 82
Figure 4-10: Schematic of the arrangement used to record spectra from the EFPI sensor:
(1) the SLD light source, (2) the EFPI sensor, (3) the 2 × 2 coupler, (4) the broken end
of the fibre used to prevent any back-reflection, (5) CCD spectrometer, and (6)
data acquisition.................................................................................................................. 84
Figure 4-11: Schematic of the arrangement used to record spectra from the FBG sensor:
(1) the 1550 nm Laser diode, (2) the FBG sensor, (3) the 2 × 2 coupler, (4) the
broken end of the fibre used to prevent any back-reflection, (5) the FBG
interrogation unit, and (6) the data acquisition. ........................................................... 84
Figure 4-12: Definition of the parallel cutting angle. .............................................................. 87
Figure 4-13: Illustration of fibre orientation in the cross-section A-A of Figure 4-12............ 87
Figure 4-14: The position of specimens taken from the composite panel.......................... 88
Figure 4-15: Test specimen geometry for static tension and tension-tension fatigue
testing. .................................................................................................................................. 89
Figure 4-16: Illustration of strain at deviation from linearity. .................................................. 90
Figure 4-17: Test specimen geometry for static compression testing.................................. 90
Figure 4-18: Test specimen geometry for tension-compression fatigue testing................. 91
Figure 5-1: Comparison of load from load cell and Instron testing machine. ................... 94
Figure 5-2: Comparison of load measured from the load cell and the applied load
(Instron machine) at different temperatures.................................................................. 94
Figure 5-3: Strain measured from the ERSG as a function of the applied load measured
from the load cell. .............................................................................................................. 95
xix
LIST OF FIGURES
Figure 5-4: Position of ERSGs and the corresponding symmetrical positions in steel plate.
............................................................................................................................................... 96
Figure 5-5: Strain distribution in steel plate at an applied load of 2 kN............................... 96
Figure 5-6: Strain distribution in steel plate at an applied load of 8 kN............................... 97
Figure 5-7: Strain distribution in steel plate at an applied load of 15 kN............................. 97
Figure 5-8: Warped composite. ................................................................................................. 98
Figure 5-9: Microscopic evidence of fibre misalignment near end-tab region................. 99
Figure 5-10: Schematic illustration of End-tab jig design. .................................................... 100
Figure 5-11: Microscopic evidence of improved fibre alignment near end-tab region.101
Figure 5-12: Surface profile of non-prestressed and prestressed unidirectional

composites......................................................................................................................... 101
Figure 5-13: Relationship between optical sensors (embedded and reference) and
temperature. ..................................................................................................................... 102
Figure 5-14: Relationship between reference optical fibre sensors and pressure........... 105
Figure 5-15: Relationship between embedded optical fibre sensors and pressure........ 105
Figure 5-16: Relationship between load measured from the load cell and vacuum. ... 106
Figure 5-17: Relationship between embedded sensors and vacuum. ............................. 107
Figure 5-18: Relationship between reference EFPI sensor and the cure cycle parameters.
............................................................................................................................................. 108
Figure 6-1: Residual strain measured using embedded EFPI sensor in non-prestressed [0]16
laminate during the autoclave curing process. a) before curing, b) heating cycle,
c) isothermal and d) cooling cycle. .............................................................................. 113
Figure 6-2: Differential Scanning Calorimetry (DSC) results for the 913 epoxy resin........ 113
Figure 6-3: Illustration of residual stress development in a unidirectional composite. .... 114
Figure 6-4: Residual strain and autoclave pressure recorded during the cure cycle. ... 115
Figure 6-5: Residual strain development in cross-ply composite. a) before curing, b)

heating cycle, c) isothermal and d) cooling cycle. ................................................... 115
Figure 6-6: Residual strain development throughout the processing of 150 MPa
prestressed composite..................................................................................................... 118
Figure 6-7: The strain release measured from optical fibre sensors and ERSG. This was
recorded during 108 MPa prestress release. ................................................................ 119
Figure 6-8: Sensor configuration, a) EFPI strain sensor and b) FBG strain sensor. PMRI –
periodic modulation of the refractive index of the optical fibre core. ................... 120
Figure 6-9: The final residual strain in the composite for different prestress levels. .......... 121
Figure 6-10: The measured fibre orientation in a composite as a function of applied

prestress.............................................................................................................................. 123
xx
Figure 6-11: A representative micrograph of a transverse section of a non-prestressed (0
MPa) composite. .............................................................................................................. 123
Figure 6-12: A representative micrograph of a transverse section of the 51 MPa




Figure 7-1: Strain release to the composite. .......................................................................... 131
Figure 7-2: Prestress release in element e1. ............................................................................ 132
Figure 7-3: Sensor configuration, a) EFPI strain sensor and b) FBG strain sensor. ............. 136
Figure 7-4: Residual stress in the fibre and matrix with prestress. The theoretical residual
stress at 0 MPa prestress in fibre and matrix was calculated from the Wagner37
model. ................................................................................................................................ 137
Figure 7-5: Residual stress in the fibre and matrix with prestress. The experimentally-
measured residual stress at 0 MPa prestress is included in the theoretical prediction.
............................................................................................................................................. 137
Figure 7-6: Strain released to the composite......................................................................... 139
Figure 7-7: Strain distribution across the width of the composite panel for 29 MPa
prestress release................................................................................................................ 139
Figure 8-1: Stress-strain curve for all prestressed composite panels................................... 144
Figure 8-2: The measured strain at the deviation from linearity in stress-strain curve as a
function of prestress. The error bars represents the standard deviation.................. 146
Figure 8-3: The effect of fibre prestress on composite failure strain................................... 149
Figure 8-4: Photo macrograph of static tensile specimens................................................. 150
Figure 8-5: Micrographs of specimens tested in tension, (i) 0 MPa, (ii) 51 MPa, (iii) 80 MPa,
(iv) 108 MPa and (v) 150 MPa prestressed composites. The failure mechanisms are
denoted thus in each figure: a – fibre fracture/pull-out, b – matrix hackle
formation, c – clean fibre surface, d – matrix plasticity, e – fibre impression and f –
fibre fragment. .................................................................................................................. 153
Figure 8-6: Compression strength of prestressed composites............................................. 155
xxi
LIST OF FIGURES
Figure 8-7: Compression modulus of prestressed composites ............................................ 155
Figure 8-8: A typical compressive stress/strain curve for [0°]16 prestressed composites. 156
Figure 8-9: Comparison of compressive strain measured in front and back of the
composite. ......................................................................................................................... 157
Figure 8-10: Representative compressive failures of 0, 51, 80, 108 and 150 MPa
prestressed composites. .................................................................................................. 159
Figure 8-11: A magnified view of the macroscopic failure of all the prestressed
composites. (i) 0 MPa, (ii) 51 MPa, (iii) 80 MPa, (iv) 108 MPa and (v) 150 MPa....... 160
Figure 8-12: Kink-band failure mechanism observed in non-prestressed (0 MPa)

composite. ......................................................................................................................... 161
Figure 8-13: A magnified view of position 1 in Figure 8-12................................................... 162
Figure 8-14: A magnified view of position 2 in Figure 8-12................................................... 162
Figure 8-15: A Magnified view of the Compressive failure of 0 MPa prestressed

composite presented in Figure 8-11 exhibiting multiple kink-band formation........ 163
Figure 8-16: The typical fibre failure of prestressed composite in compressive loading. (i)
51 MPa, (ii) 80 MPa, (iii) 108 MPa and (iv) 150 MPa prestressed composite. .......... 164
Figure 8-17: The typical matrix failure of prestressed composite in compressive loading.
(i) 51 MPa, (ii) 80 MPa, (iii) 108 MPa and (iv) 150 MPa prestressed composite. ...... 165
Figure 8-18: Schematic representation of imploded failure mechanism in prestressed

composite during compressive loading. ...................................................................... 165
Figure 8-19: Matrix curling fracture in a prestressed composites. (i) 50 MPa and (ii) 108
MPa prestressed composite............................................................................................ 166
Figure 9-1: Comparison of normalised fatigue data of non-prestressed composite

(current work) with similar systems reported in the literature..................................... 170
Figure 9-2: Comparison of the tension-tension fatigue results of the non-prestressed and
prestressed composites. The linear regression fit was calculated including the UTS
at 0.5 cycles. The arrow mark represents the sample run-out................................... 172
prestressed composites. The linear regression fit was calculated without including
the UTS at 0.5 cycles. The arrow mark represents the sample run-out. .................... 172
prestressed composites. A second order fit was calculated without including the
UTS. ...................................................................................................................................... 174
Figure 9-5: Comparison of normalised tension-tension fatigue of non-prestressed and

prestressed composites. The error bars were calculated for 95% confidence level.
............................................................................................................................................. 175
Figure 9-6: Fatigue life as a function of applied prestress in the low stress region (peak
stress of 590 MPa).............................................................................................................. 176
xxii
stress of 524 MPa). The arrow mark represents the specimen run-out. .................... 177
Figure 9-8: Fatigue life as a function of applied prestress in the high stress region (peak
stress of 918 MPa).............................................................................................................. 177
Figure 9-9: Weibull distribution for fatigue life at 524 MPa peak stress level (low stress
level). .................................................................................................................................. 178
Figure 9-10: Weibull Distribution for fatigue life at peak stress level 788 MPa (middle stress
level). .................................................................................................................................. 179
Figure 9-11: Weibull distribution for fatigue life at 1180 MPa peak stress level (high stress
level). .................................................................................................................................. 179
Figure 9-12: The normalised stiffness and surface temperature change as a function of
fatigue life. ......................................................................................................................... 182
Figure 9-13: The macroscopic damage mechanisms observed visually during the
tension-tension fatigue test of a unidirectional composite at 787 MPa peak stress.
a) Normalised life 0, b) Normalised life 0.2, c) Normalised life 0.4, d) Normalised life
0.55, e) Normalised life 0.8 and f) Normalised life 1. ................................................... 184
Figure 9-14: Stiffness reduction of non-prestressed and prestressed composites at 655

MPa peak stress. ............................................................................................................... 186
Figure 9-15: Stiffness reduction of non-prestressed and prestressed composites at 1050

MPa peak stress. ............................................................................................................... 186
Figure 9-16: Surface temperature rise of non-prestressed and prestressed composites at

655 MPa peak stress. ........................................................................................................ 187
Figure 9-17: Surface temperature rise of non-prestressed and prestressed composites at

1050 MPa peak stress. ...................................................................................................... 187
Figure 9-18: Representative macroscopic fatigue failure of non-prestressed and

prestressed composites. a) 1050 MPa peak stress and b) 524 MPa peak stress..... 189
Figure 9-19: T-T fatigue fracture of non-prestressed composite (0 MPa) at a high stress
level (1180 MPa). Features are marked as fibre pull-out (a), matrix hackle formation
(b), clean fibre surface (c) and matrix plasticity (d)................................................... 190
Figure 9-20: T-T fatigue fracture of 51 MPa prestressed composite at a high stress level
(1180 MPa). Features are marked as fibre pull-out (a), matrix hackle formation (b),
clean fibre surface (c) and matrix plasticity (d). ......................................................... 191
Figure 9-21: T-T fatigue fracture of 108 MPa prestressed composite at a high stress level
(1180 MPa). Features are marked as fibre pull-out (a), clean fibre surface (c) and
matrix plasticity (d). .......................................................................................................... 191
Figure 9-22: T-T fatigue fracture of non-prestressed composite (0 MPa) at a low stress
level (590 MPa). Features are marked as fibre pull-out (a), matrix hackle formation
(b), matrix plasticity (d) and fibre impression (e)......................................................... 192
xxiii
LIST OF FIGURES
Figure 9-23: T-T fatigue fracture of 51 MPa prestressed composite at a low stress level
(590 MPa). Features are marked as fibre pull-out (a), matrix hackle formation (b)
and fibre impression (e). .................................................................................................. 192
Figure 9-24: T-T fatigue fracture of 108 MPa prestressed composite at a low stress level
(590 MPa). Features are marked as fibre pull-out (a), matrix hackle formation (b)
and fibre impression (e). .................................................................................................. 193
Figure 9-25: Mechanism of matrix hackle formation138........................................................ 193
Figure 9-26: Matrix hackle formation (b) in a non-prestressed composite. ...................... 194
Figure 9-27: A row of matrix hackle formation (b) and fibre pull-out (a) in a non-
Figure 9-28: Comparison of normalised fatigue data of non-prestressed composite

(current work) with a similar system reported in the literature (Fernando and Al-
Khodari85). .......................................................................................................................... 195
Figure 9-29: Comparison of tension-compression fatigue of the non-prestressed and

prestressed composites. A linear regression fit was calculated................................ 196
Figure 9-30: Comparison of T-T and T-C fatigue behaviour of non-prestressed composite.
............................................................................................................................................. 198
Figure 9-31: Comparison of tension-compression fatigue of the non-prestressed and

prestressed composites. A second order fit was calculated excluding the static
strength (UTS and UCS). ................................................................................................... 199
Figure 9-32: Comparison of normalised tension-compression fatigue of the non-

prestressed and prestressed composites. The error bars were calculated for a 95%
confidence limit. ............................................................................................................... 200
stress of 524 MPa). The arrow mark represents the specimen run-out. .................... 201
Figure 9-34: Fatigue life as a function of applied prestress in the high stress region (peak
stress of 944 MPa).............................................................................................................. 201
Figure 9-35: Fatigue life as a function of applied prestress in the middle stress region
(peak stress of 630 MPa). ................................................................................................. 202
Figure 9-36: Weibull distribution for fatigue life at 524 MPa peak stress (low stress region).
............................................................................................................................................. 203
Figure 9-37: Weibull distribution for fatigue life at 944 MPa peak stress (high stress region).
............................................................................................................................................. 204
Figure 9-38: Weibull distribution for fatigue life at 630 MPa peak stress (middle stress
region). ............................................................................................................................... 204
Figure 9-39: Stiffness reduction in non-prestressed and prestressed composites at 524

MPa peak stress. ............................................................................................................... 206
xxiv
Figure 9-40: Stiffness reduction in non-prestressed and prestressed composites at 840
MPa peak stress. ............................................................................................................... 206
Figure 9-41: Surface temperature rise in non-prestressed and prestressed composites at

524 MPa peak stress. ........................................................................................................ 207
Figure 9-42: Surface temperature rise in non-prestressed and prestressed composites at

840 MPa peak stress. ........................................................................................................ 207
Figure 9-43: Representative macroscopic fatigue failure of non-prestressed and

prestressed composites. a) 945 MPa peak stress and b) 524 MPa peak stress....... 210
Figure 9-44: T-C fatigue fracture of a non-prestressed composite (0 MPa) at a high stress
level (945 MPa peak stress). Features such as tensile fracture (a) and compression
fracture (b) can be seen................................................................................................. 212
Figure 9-45: Compressive failure of a non-prestressed composite at a high stress level

(945 MPa peak stress). Features such as fibre fracture (d) and matrix plasticity (c)
can be seen. ..................................................................................................................... 212
Figure 9-46: T-C fatigue fracture of a 51 MPa prestressed composite at a high stress level
(945 MPa peak stress). Features such as tensile fracture (a) and compression
fracture (b) can be seen................................................................................................. 213
Figure 9-47: A close-up image of the failure of a 51 MPa prestressed composite shown in
Figure 9-46. Features such as tensile fracture (a) and compression fracture (b) can
be seen............................................................................................................................... 213
Figure 9-48: T-C fatigue fracture of a 108 MPa prestressed composite at a high stress
level (945 MPa peak stress). The area marked as a, shows the tensile fracture..... 214
Figure 9-49: Tensile failure of a 108 MPa prestressed composites at a high stress level (945
MPa peak stress). The end face of the fractured fibre (d) represents a tensile
failure. ................................................................................................................................. 214
Figure 9-50: Tensile fibre pull-out (f), compressive fibre buckling (g) and fibre impression
(h) observed in a 108 MPa prestressed composites at a high stress level (945 MPa
peak stress). ....................................................................................................................... 215
Figure 9-51: Fatigue failure involving misaligned fibres under a tension and compression
cycle. a) fibre misalignment in composite, b) fibre/matrix interface debonding, c)
matrix crack initiation and d) matrix crack propagation. The arrow mark shows the
direction of loading.......................................................................................................... 216
Figure 9-52: Ultimate failure of composites by e) Compression fracture and f) tensile

fracture............................................................................................................................... 216
Figure 9-53: T-C fatigue fracture of a non-prestressed composite (0 MPa) at a low stress
level (524 MPa peak stress). Features such as fibre fracture (d) and matrix hackle
formation (e) can be observed. .................................................................................... 217
Figure 9-54: T-C fatigue fracture of a 51 MPa prestressed composite at a low stress level
(524 MPa peak stress). Features such as fibre fracture (d), matrix hackle formation
(e) and fibre impression (h) can be observed............................................................. 217
xxv
LIST OF FIGURES
Figure 9-55: T-C fatigue fracture of a 108 MPa prestressed composite at a low stress level
(524 MPa peak stress). Features such as matrix plasticity (c), fibre fracture (d) and
matrix hackle formation (e) can be observed. ........................................................... 218
Figure 9-56: Compressive failure in a non-prestressed composite at a low stress level (524
MPa peak stress). Features such as compressive fibre fracture (b) and matrix
plasticity (c) can be seen................................................................................................ 218
xxvi
LIST OF TABLES
Table 2-1: Advantages and disadvantages of the various residual stress measurement
techniques. .......................................................................................................................... 12
Table 2-2: The compressive residual strain results of unidirectional and cross-ply
carbon/epoxy laminates reported by Lawrence et al.24............................................. 16
Table 2-3: Literature review of prestressed composites ......................................................... 28
Table 2-4: Comparison of existing prestressing methods....................................................... 37
Table 2-5: Summary of tensile properties reported by Huang et al. 83. ............................... 39
Table 2-6: Mechanical properties of the prestressed and non-prestressed laminate

reported by Schulte and Marissen71. Results are an average of seven tests............ 41
Table 3-1: Properties of E-glass and steel. ................................................................................ 63
Table 3-2: Calculation of angle of rotation. ............................................................................ 65
Table 4-1: Summary of the panels used to record in situ residual strain development,
and the type of sensor and its position in the panel..................................................... 83
Table 5-1: Material properties used. ....................................................................................... 103
Table 5-2: Summary of thermal expansion measured experimentally and calculated

theretically. ........................................................................................................................ 103
Table 6-1: Presents the EFPI strain measured at different stages of the cure cycle in non-
prestressed composites. .................................................................................................. 112
Table 6-2: Residual strain measured from four different prestress levels of the
unidirectional composites at room temperature prior to prestress release. ........... 117
Table 6-3: Strain release recorded from EFPI, FBG and ERSG sensors during prestress
release at room temperature. * Rig thermal expansion induced prestress
calculated from the classical mechanics. ................................................................... 118
xxvii
LIST OF TABLES
Table 6-4: The average strain release recorded using EFPI, FBG and ERSG sensors in a
108 MPa prestressed composite..................................................................................... 121
Table 7-1: Material parameters used...................................................................................... 135
Table 7-2: Residual strain results............................................................................................... 135
Table 8-1: Summary of tensile properties of composites studied in this programme. The
number in parentheses represents the standard deviation. ..................................... 145
Table 8-2: Tensile strength comparison. ................................................................................. 147
Table 8-3: The slope calculated from linear regression analysis for each prestressed
composite presented in Figure 8-7. ............................................................................... 157
Table 8-4: Comparison of UCS of unidirectional controlled specimens with different test
methods. The number in parentheses represents the standard deviation. * l –
gauge length, w – width, t – thickness. ......................................................................... 158
Table 9-1: Summary of composite lay-up sequences and fatigue test parameters used
in the current work and those reported by previous researchers............................. 170
Table 9-2: The individual fatigue life results of non-prestressed and prestressed
composites plotted in Figure 9-2. The arrow mark (⇨) represents that the specimen
ran-out. ............................................................................................................................... 173
Table 9-3: Summary of slope (b) and degradation of tensile strength per decade
(b/σUTS) of non-prestressed and prestressed composites. .......................................... 174
Table 9-4: Plotting formulae for Weibull two-parameter distribution................................. 178
Table 9-5: Summarises the Weibull modulus of non-prestressed and prestressed

composites presented in Figure 9-9 to Figure 9-11. ..................................................... 180
Table 9-6: Summary of the stiffness degradation and surface temperature changes for
the non-prestressed and the prestressed composites at a peak stress of 655 MPa in
T-T fatigue........................................................................................................................... 188
the non-prestressed and the prestressed composites at a peak stress of 1050 MPa
in T-T fatigue....................................................................................................................... 188
Table 9-8: Ssummary of slope (b) and tensile strength degradation per decade (b/σUTS)
for current work and GFAK composites. ....................................................................... 195
Table 9-9: Summerises the individual fatigue life results of non-prestressed and
prestressed composites plotted in Figure 9-29. The arrow mark (⇨) represents the
sample is run-out. .............................................................................................................. 197
Table 9-10: A summary of slope (b) and tensile strength decay per decade (b/σUTS) of
non-prestressed and prestressed composites.............................................................. 197
Table 9-11: Summarises the calculated equivalent σequ from second-order polynomials
for the non-prestressed and prestressed composites without static strength. The
number in parentheses shows the standard deviation. ............................................. 199
xxviii
Table 9-12: Summarises the Weibull modulus of non-prestressed and prestressed
composites presented in Figure 9-36 to Figure 9-38. ................................................... 203
the non-prestressed and the prestressed composites at a peak stress of 524 MPa in
T-C fatigue. ........................................................................................................................ 209
Table 9-14: Table 9-12: Summary of the stiffness degradation and surface temperature
changes for the non-prestressed and the prestressed composites at a peak stress
of 840 MPa in T-C fatigue. ............................................................................................... 209
xxix
ABBREVIATONS
ASTM American society for testing materials

CARALL Carbon fibre reinforced aluminium alloy laminate
CCD Charged coupling device
CDS Characteristic damage state
CFRP Carbon fibre reinforce plastics
CRAG Composite research advisory group
DSC Differential scanning calorimetry
EB Electron beam
EFPI Extrinsic Fabry-Pérot interferometric
ERSG Electrical resistance strain gauges
ETJCET End-tab jig cured end-tabs
FBG Fibre Bragg grating
FBPM Flat-bed prestress methodology
GFRP Glass fibre reinforced plastics
HPCET Hot-press cured end-tabs
ICSTM Imperial college of science, technology and medicine
IFSS Interfacial shear stress
ILSS Interlaminar shear stress
ISO International organisation for standardisation
PEEK Poly (etheretherketone)
PEI Poly (etherimide)
xxxi
LIST OF TABLES
PP Poly (propylene)
PPS Poly (phenylene sulphide)
PTFE Poly (tetrafluroethylene)
RT Room temperature
SEM Scanning electron microscope
SLD Super-luminescence diode
SMA Shape memory alloy
S-N Peak stress verses the logarithmic of number of cycles to failure
T-C Tension – compression
T-T Tension- tension
Tg Glass transition temperature
UCS Ultimate compressive strength
UTS Ultimate tensile strength
VIRALL Vinylon fibre/epoxy-aluminium laminates
xxxii
CHAPTER 1
Introduction
OVERVIEW
Research motivation and rationale is stated.
The research aims are presented.
The contributions of this research to knowledge are given.
Scientific publications generated in the course of PhD are presented.
Organisation of the thesis.
T
HIS PHD THESIS investigates the effect of fibre prestress on, (i) residual stress, (ii)
fibre alignment and (iii) static and fatigue performance of composites. Fibre
prestressing is utilised to minimise the process-induced residual stress and fibre
waviness in composites. This method also improves the static mechanical and fatigue
properties of composites.
The motivation and rationale for this research is outlined in this chapter and the problem
statement is presented as a set of aims and objectives. The contributions from this thesis
are presented, and finally, the organisation of the thesis is explained.
1.1 Research motivation and rationale
During processing of advanced composite materials, residual stresses and fibre waviness
are induced. The residual stresses are developed in composites because of the mismatch in
thermal expansion between fibre and matrix, on the micro-mechanical scale, and between
adjacent plies of different orientation on the macro-mechanical scale1. Fibre waviness is
induced due to fibre movement during processing. These process-induced issues can
CHAPTER 1 Introduction
degrade the structural performance of composites. The presence of residual stresses in a

composite can reduce the strain to micro-cracking and this could expose the fibres to
environmental degradation. The effect of fibre waviness has a detrimental effect on the
overall composite structural integrity by adversely affecting the compressive strength,
fatigue life and tensile strength.
The next generation aircraft structures for civil aviation industry are manufactured using
advanced composite materials. In particular, the primary structures such as wings of the
A380 (AIRBUS Ltd) and fuselage and wings of 787 “Dreamliner” (Boeing Ltd) were
manufactured using composite materials that were processed in an autoclave. These
primary aircraft structures will be subjected to various types of mechanical loading
including fatigue during service. The presence of residual stresses and fibre waviness could
have a detrimental effect on the mechanical properties of composite structures and this
could reduce the service life of aircraft. Therefore, it is of paramount importance to reduce
the process-induced residual stresses and fibre waviness. These process-induced issues play
a vital role in motivating our research.
In recent years fibre prestressing has been proposed as a way to minimise the residual
stresses and to improve mechanical properties of composites2, 3.
Prestressed composites are prepared by applying pre-tension to the fibres and subsequently
curing the composite. Once the composite is cured and cooled down to room temperature,
the pre-tension is released. Now the pre-tensioned fibres will tend to contract, inducing
compressive stresses to the matrix, which will minimise the residual stresses resulting from
the cure cycle. In literature fibre prestressing methodology has been theoretically shown as
a method for reducing residual stresses and experimentally demonstrated to enhance
mechanical properties of polymer composite laminates. In addition, as the fibres are
subjected to a constant load throughout the curing process, fibre misalignment or fibre
waviness in composites can be minimised.
In this research, fibre prestressing is chosen as a method to provide immunity against the
development of process-induced issues in composites.
1.2 Aims and objectives
The aims and objectives of this research are as follows:

• To design and evaluate a bespoke fibre prestressing methodology for autoclave
processing of composites.
• To measure and quantify the residual stress reduction in composites with fibre
prestress.
• To investigate the fibre alignment in composites with fibre prestress.
2
1.3 Contributions
• To model theoretically, the effect of fibre prestressing on residual stresses and

compare the same with experimental results.
• To investigate the effect of fibre prestress on static tensile and compressive
properties.
• To investigate the fatigue behaviour of prestressed composites in both tension-
tension and tension-compression modes and compare with non-prestressed
composites.
• To understand the influence of fibre prestressing on damage of composites
subjected to both static and fatigue loading.
Initially, it was planned to use a prestressing methodology designed by a previous
researcher in this group. However, after initial trials it was found that this system was not
suitable for the current requirements. Hence this research embarked on designing a
prestress rig that satisfies the current requirements.
1.3 Contributions
The contributions of this research are as follows

• Novel design and development of a bespoke prestressing methodology for fibre
reinforced composites. This methodology is successfully used to manufacture
composites in an autoclave using conventional vacuum bagging techniques. The
pre-load applied to the composites is monitored online throughout the curing cycle.
• In-situ monitoring of residual strain development during autoclave processing and
prestress release is demonstrated using optical fibre sensors in real-time. The strain
release recorded using optical fibre sensors and surface-mounted electrical
resistance strain gauge sensors are in good agreement.
• Experimentally, fibre prestress methodology is shown as a method to minimise and
control residual stresses in fibre reinforced composites. The residual strain in
composites with different prestress levels is quantified.
• A theoretical model was developed to predict the strain released to the composite
for a given applied prestress. The experimental results from optical fibre sensors
and strain gauges are in good agreement with the theoretical model. The model was
further developed to predict the Poisson’s effect on the strain variation through the
width of the composite panel.
• Fibre alignment of the composites improved with increasing fibre prestress up to
an optimum limit of 108 MPa prestress. No improvement was observed beyond
this limit.
• From the static tensile results it was found that with fibre prestress, strain at
deviation from linearity increases. This suggests that prestressing enhanced the
3
resistance to initial damage like matrix cracking, weak fibre failure during tensile
loading. However, the tensile strength and modulus are independent of applied
prestress.
• Compressive strength and modulus properties increased with fibre prestress. This
suggests that the improvement in fibre alignment with prestress have enhanced the
compressive properties. The non-prestressed composites fractured by typical fibre
kink-band failure, whereas the prestressed composites fractured by extensive
longitudinal splitting and matrix cracking, resulting in a crushing mode of failure.
This also suggests that the improvement in fibre alignment could have changed the
compression damage mode of composites.
• Improvement in fatigue life and resistance to stiffness degradation was observed in
matrix-dominated fatigue region. This suggests that minimising the matrix tensile
residual stresses by fibre prestressing has enhanced the fatigue properties in the
matrix-dominated fatigue region. The improvement in fatigue life and resistance to
stiffness degradation observed in tension-compression fatigue was greater than in
tension-tension fatigue. This is because of a reduction in matrix residual stress,
improvement in fibre alignment and also the presence of tensile residual stress in
fibre.
• The degree of longitudinal splitting in prestressed composites was lower than the
non-prestressed composites in the low stress fatigue region in both tension-tension
and tension-compression fatigue modes. Also in tension-compression fatigue the
extent of compression damage of fibres was lower in prestressed composites. This
change in the damage mechanism is due to the reduction in matrix tensile residual
stresses and improvement in fibre alignment in composites with fibre prestress.
1.4 Work disseminations
The research findings were presented in the following publications:
• S Krishnamurthy, RA Badcock, GF Fernando and A Hameed, “Novel prestress

methodology for autoclave processing of composites: fabrication and mechanical
performance”, Proceedings of the SAMPE 2005 conference, Long Beach, USA,
May 2005.
• S Krishnamurthy, RA Badcock, GF Fernando and A Hameed, “Design and
evaluation of novel prestress methodology for fibre reinforced composites:
fabrication and in-situ monitoring”, Composite Science and Technology Journal. under
review.
4
1.5 Organisation of thesis
The following publications are generated in collaboration with other researchers during the
course of PhD:
• R.Badcock, S.Krishnamurthy, G.F.Fernando, T,Butler, R.Chen and J.Tetlow,
“Health monitoring of composite structures using a novel fibre optic acoustic
emission sensors”, The 11th European Conference on Composite Materials, 31 May - 3 June
2004 Rhodes , Greece.
• T.Butler, R.Chen, S.Krishnamurthy, R.Badcock and G.F.Fernando, “ A low-cost
fibre optic acoustic emission sensor for damage detection in engineering composite
materials and structures”, The 4th International Workshop on Structural Health Monitoring,
Stanford University, USA, September 15-17, 2003.
1.5 Organisation of thesis
This thesis is structured into nine chapters.
CHAPTER 2
This chapter is intended to set the scene, there are three parts in this chapter. The first part
surveys the current literature covering process-induced residual stress development,
techniques for measuring and predicting residual stresses, factors instigating fibre waviness
during manufacturing process and its effect on mechanical properties of composites. This
is followed by various techniques reported in the literature to minimise process-induced
issues. The second part presents a comprehensive review of the fibre prestress literature.
This includes a critical review of the various prestress methods reported in the literature
and the effect of prestress on mechanical properties of composites. The final part discusses
issues relating to fatigue behaviour and failure modes of unidirectional glass fibre/epoxy
composites.
CHAPTER 3
This chapter presents design criteria for fibre prestress methodology proposed in this
current research. This is followed by a design using flat-bed prestress methodology, which
satisfies all the design criteria. Design calculations are included in this chapter.
CHAPTER 4
The experimental procedures for composite manufacturing, residual strain monitoring,
fibre alignment, quality control, static and fatigue test conditions are presented.
5
The experimental results are described and discussed in chapters five to nine.
CHAPTER 5
The results from the evaluation of prestress methodology and sensors are presented and
discussed.
CHAPTER 6
This chapter is structured into two parts: 1) residual strain development and 2) fibre
alignment. The residual strain development in non-prestressed composites during autoclave
processing is presented. This is followed by the residual strain development and strain
release results of prestressed composites. The reduction in residual stress with fibre
prestress was quantified. The measured fibre alignment in non-prestressed and prestressed
composites are presented and discussed.
CHAPTER 7
In this chapter, a theoretical model is developed to predict the strain released to the
composite for a given applied fibre prestress based on micromechanical theory. This model
is further developed to predict the Poisson’s effect on the strain variation through the
width of the panel during strain release. The experimental results from embedded optical
fibre sensors and surface-mounted strain gauges are compared with theoretical predictions.
CHAPTER 8
The results from static tensile and compression tests, and post-failure analysis of non-
prestressed and prestressed composites are presented and discussed.
CHAPTER 9
This chapter present the tension-tension and tension-compression fatigue test results and
discusses the influence of prestress on fatigue life, stiffness degradation, surface
temperature rise and fracture mechanisms of composites.
CHAPTER 10
The overall conclusions of this study and directions for future research are presented.
6
CHAPTER 2
Literature review
OVERVIEW
The development of process-induced residual stresses and fibre waviness
during composite processing is described;
The methods to mitigate process-induced issues are reviewed;
A critical literature survey on prestressed composites, methodology for fibre
prestressing and the effect of fibre prestress on static mechanical properties is
presented;
A review of composite fatigue behaviour and damage mechanisms, relevant to
this study is presented.
2.1 Introduction
A
COMPOSITE MATERIAL IS a combination of two or more materials, whose
properties are superior to those of the constituent materials acting independently.
Fibre-reinforced polymer composites are usually manufactured by embedding stiff
and strong fibres into a relatively less stiff and compliant, polymeric matrix. The primary
role of the fibres is to provide strength and stiffness to the composite. Typical reinforcing
fibres used are glass, carbon and aramid, with fibre diameters in the range of 6-14 µm.
The role of the matrix resin is to hold the fibres in their position, protect the fibres from
abrasion and the external environment (such as chemicals or moisture) and transfer load
between fibres. Matrices for polymeric composites can be either thermosets or
thermoplastics. Thermoset resins usually consist of a low-molecular weight resin and a
CHAPTER 2 Literature review
compatible curing agent (also called a hardener). When the resin and hardener are mixed
they form a low viscosity liquid that undergoes a chemical reaction to form three-
dimensional cross-linked structures, resulting in an insoluble solid phase that cannot be
reprocessed on reheating. Thermoset polymeric matrices include epoxies, polyesters, vinyl
esters, bismaleimides, cynate esters, polyimides and phenolics. On the other hand,
thermoplastics are fully reacted high-molecular weight polymers that do not cross-link on
heating. On heating to their softening or melting temperature they can be reprocessed a
number of times. Typical thermoplastics used in composites are poly(etheretherketone)
(PEEK), poly(phenylene sulfide) (PPS), Poly(etherimide) (PEI) and Poly(propylene) (PP).
The attainment of good mechanical properties in a composite material depends crucially
upon the efficiency of stress transfer from the matrix to the fibre, which requires
optimisation of the adhesion of the fibre to the matrix. To promote the fibre-matrix
adhesion, different surface treatments/coupling agents are applied to the fibre surfaces to
provide chemical bonding with the polymeric matrix. Silane type coupling agents are
normally applied to glass fibre surfaces during manufacture while carbon fibres are surface
treated by oxidation4 and aramid fibres are usually treated with an epoxy finish5. The
adhesion between fibre and matrix occurs during the processing of composites. The
processing parameters control the properties and therefore the quality of the composites.
This will be discussed in the following section.
2.2 Processing of composites
Composite manufacturing can be mainly categorised into two phases: a) pre-forming and b)
processing. This is schematically represented as a flowchart in Figure 2-1. The two main
raw materials for processing composites are fibres and matrix. In the pre-forming phase the
reinforcing fibres and the accompanying matrix material are placed or shaped into a
structural form. The fibre and matrix may be in a pre-impregnated form or the fibre and
matrix material may be combined for the first time during this step of developing the
structural form. The next step is processing wherein the temperature and pressure are used
to consolidate the structure. For thermoset plastics the chemical cross-linking reaction
solidifies the structure, whereas thermoplastic matrixes become hard after cooling from
their melting temperature.
The end properties of the composite part are determined not only by the function of the
individual properties of the resin matrix and fibre, but also as a function of the
arrangement of the materials themselves into the part, the ratio of the constituent materials
and also the way in which they are processed.
This thesis investigates composites using thermoset matrices. From hereon only theromset
composites are considered. Composites can be manufactured using a wide variety of
processing techniques such as hand lay-up, hot-pressing, vacuum bagging, pressure
bagging, resin transfer moulding, filament winding, pultrusion and autoclave moulding
8
2.2 Processing of composites
techniques. Autoclave-based processing is a widely used method of producing high-quality

composites in the aerospace industry6, and will be utilised in this study.
Fibre architecture Matrix infusion
Pre-forming
Controlling time,
temperature and pressure
Processing
Composite part
Figure 2-1: Schematic representation of the principle of composite

processing.
An autoclave process is a technique that can generate a controlled pressure and

temperature environment. A typical autoclave system consists of a pressure vessel, a gas
compressor, a heater system, a vacuum pump and a control system. Prepreg is widely used,
as a starting material in an autoclave. A prepreg system is a pre-impregnated sheet of a
combination of the reinforcing fibres and the resin matrix (with hardener). In a prepreg,
the resin is advanced to a B-stage condition in which it is a semi-solid (represents a partially
cured state) at room temperature, which melts and flows during the cure cycle. The B-
staged resin normally contains some tack or stickiness to allow it to adhere during the lay-
up operation. With reference to the manufacturing of prepreg-based composites, the
unidirectional prepreg laminas are stacked-up or laminated with the fibres in the required
orientation. This is carried out either manually or by a lay-up machine which places the
individual plies of the prepreg in the required orientation directly onto the tool surface. The
laminated prepreg is then vacuum bagged with release film and breather cloths. The
vacuum bagged prepreg laminate is placed inside the autoclave where a combination of
heat, pressure and vacuum consolidates and cures the laminate.
Figure 2-2 shows a typical autoclave cure cycle for prepreg-based composites. With
reference to the figure, the vacuum is applied before the heating stage of the cure cycle and
maintained throughout the cure cycle. During the heating cycle, initially the viscosity of the
resin decreases, and this allows the entrapped air between the laminas to be removed by the
application of the vacuum. The pressure applied on the prepreg stack will consolidate the
9
laminate. During the isothermal stage of the cure cycle, the temperature and pressure are
kept constant. In this dwell period, the polymerisation of the resin takes place by an
exothermic reaction; this increases the viscosity of the resin rapidly.
Temperature
Pressure
Vacuum
TCure
a b c d
Temperature oC
Pressure = 0.7 MPa
Vacuum = 850 Millibars

TRT
Time
Figure 2-2: Typical autoclave cure cycle for prepreg-based composites. a)

vacuum application, b) heating cycle, c) isothermal cure and
d) cooling cycle.
As the reaction proceeds, the molecular weight of the resin increases by linking several
chains to form a three-dimensional cross-linked network. This transformation from viscous
liquid to elastic gel is known as the gel point. The degree of cure determines the amount of
cross-linking in the resin. This degree of cure of the resin determines the composite’s
performance7. After the dwell period, the cool-down stage of the cure cycle commences by
gradually reducing the temperature and pressure. At the end of the cure cycle, the part is
left to cool down to room temperature. The above steps of converting low molecular
weight viscous resin to a high molecular weight solid by a polymerisation process is also
called curing hence the name cure cycle.
2.3 Process-related issues
Advanced fibre reinforced composites are generally processed at elevated temperatures.

The quality and performance of the composite part is determined by various processing
parameters such as heating rate, dwell time, cooling rate, pressure and tool plate material8, 9,
10
. These processing parameters influence the development of process-induced residual
stresses and fibre waviness, which have been shown to affect the performance of
10
2.4 Residual stresses
composite parts.10,11 The development of process-induced residual stresses and fibre

waviness is discussed in the following sections.
Residual stresses are developed in composites during processing and in-service due to the
anisotropic thermal expansion and elastic properties of their constituents. During
processing of composites at elevated temperatures residual stresses develop due to 1) cure-
induced shrinkage of the polymer matrix and 2) constituents thermal expansion mismatch-
induced volume changes. The cure-induced residual stresses are developed due to the
shrinkage of the resin during the polymerisation reaction, which forms cross-linked
networks.12,13 Thermally-induced stresses are developed because of the mismatch in thermal
expansion between the fibre and matrix, on the micro-mechanical scale, and between
adjacent plies of different orientation on the macro-mechanical scale.
Unlike homogeneous materials, residual stresses in composites cannot be removed by
annealing because the expansion mismatch is between two different materials, not within
the same material. However, some relaxation of residual stresses in time is possible as a
result of the viscoelasticity inherent in most resins.14 The presence of residual stresses may
result in dimensional instability of composite parts. In fact residual stresses can be high
enough to cause cracking within the matrix even before the mechanical loading is applied.
In addition, micro-failures such as matrix cracking, delamination and fibre buckling may be
developed in the composites, which can adversely affect the resistance to crack initiation
and propagation, strength of the material, service limits, and exposes the fibres to chemical
degradation. 10,15
2.4.1. Techniques for measuring residual stresses

In general, residual stress measurement can be divided into two categories, namely
destructive and non-destructive testing methods. The basic drawback of destructive
methods (hole drilling16 and a first ply failure test17) is the obvious disadvantage of
damaging the part to obtain the measurements. The second category, non-destructive
testing methods includes a cure-referencing method18, X-ray diffraction19, photoelasticity20,
optical fluorescence21, Raman spectroscopy22, embedded strain gauge23, and embedded
optical fibre sensors.24,25,26
Table 2-1 shows the advantages and disadvantages of these residual stress measurement
techniques. In this study, extrinsic Fabry-Pérot interferometric (EFPI) and fibre Bragg
grating (FBG) embedded optical fibre sensors are used to monitor the process-induced
residual strain development. In the following sections, a brief introduction and working
principles of EFPI and FBG sensors are given. This is followed by a literature review on
residual strain monitoring of composites using optical-fibre sensors.
11
Table 2-1: Advantages and disadvantages of the various residual stress

measurement techniques.
Technique Advantages Disadvantages

Widely available
Destructive
Hole drilling Quick and simple
Interpretation of data
Wide range of materials
Foreign particle inclusions leads to
Simple method
X-ray diffraction stress concentration
Non-contact
Lab-based systems
Curvature and layer Simple method Limited to simple shapes
removal Wide range of materials Destructive
Simple method Only transparent materials
Photoelasticity
Non-contact Lab-based measurement
High resolution Surface measurements
Raman spectroscopy
In-situ measurement Limited range of materials
Surface measurements
Simple method
Cure referencing method Lab-based measurement
Non-contact
Restricted to flat-surfaces
Difficult to control the alignment
during cure
In-situ measurement
Susceptible to electrical and
Embedded strain gauges Small in size
magnetic interference (EMI)
Simple method
Formation of matrix pocket
leading to stress concentration
In-situ measurement
Mimicking the shape of the
Embedded optical fibre reinforcing fibre Fragile to use before embedding
sensors Immune to EMI Matrix pocket formation
Multiplexing capability
Temperature compensated
Embedded optical fibre sensors

Researchers have demonstrated that optical fibre sensors can be used to monitor a wide
range of parameters including strain, temperature, pressure, humidity, vibration, acoustic-
emission and fracture27,28,29. Recently embedded optical fibre sensors are increasingly used
in advanced composites for process monitoring26,30 and structural health monitoring31,32.
The two main types of embedded optical fibre sensors used are extrinsic Fabry-Pérot
interferometric (EFPI) and fibre Bragg grating (FBG) sensors.
12
(i) Optical-fibre EFPI sensors

Figure 2-3 shows a schematic of an EFPI sensor. EFPI sensors are fabricated by glueing or
fusion-welding two optical fibres into a glass capillary tube and are separated by an air
gap33. The distance between the two optical fibre ends is called the Fabry-Pérot cavity
length. The gauge length is defined as the distance between the two fusion points in the
capillary tube.
When light from a broadband source is launched into the lead-in fibre, a portion of the
light is reflected from the end of the lead-in fibre (R1). The transmitted light travels through
the air-gap, and is partially reflected at the reflector fibre end-face (R2). The light reflected
from the two surfaces is transmitted back through the lead-in fibre. The reflections at
glass/air and air/glass interfaces are called Fresnel reflections. These Fresnel reflections
undergo interference to produce fringes of superposition as shown in Figure 2-4.
This interference fringe is used to calculate the Fabry-Pérot cavity length d using the
following formula:
nλ1λ 2
d= µm (2.1)
2000(λ 2 − λ1 )
where n is the number of fringes between λ1 and λ2 wavelengths in nm as shown in Figure

2-4.
Optical Fibre Capillary

Gap Length
White Light
Source
R1
R2
Lead-in fibre
Reflector
fibre
Fusion Point Gauge Length
Figure 2-3: Schematic illustration of the extrinsic Fabry-Pérot interferometric

sensor design.
13
1.00E-05
λ2
9.00E-06
8.00E-06
7.00E-06
Intensity (Au)
λ1
6.00E-06
5.00E-06
4.00E-06
3.00E-06
2.00E-06
1.00E-06
0.00E+00
1460 1480 1500 1520 1540 1560 1580 1600 1620 1640
Wavelength λ (nm)
Figure 2-4: Interference fringe from an EFPI sensor.
When the sensor is subjected to an axial displacement, there will be a change in cavity
length, which will result in a change in the path length of the Fresnel reflections. The
resulting phase changes produce an interference pattern, which is different from the initial
interference. The change in the cavity length can be calculated from the new interference
pattern. If the gauge length (L) is known, then the strain can be calculated by:
⎛ ∆d ⎞
ε =⎜ ⎟ (2.2)
⎝ L ⎠
where ∆d is the change in Fabry-Pérot cavity length from the strain-free condition.
(ii) Optical-fibre Bragg grating sensors

Fibre Bragg gratings are fabricated by irradiating optical fibres to the interference of ultra-
violet laser beams. This irradiation produces periodic refractive index moduluation on the
core of the optical fibre. When light is coupled into the fibre Bragg grating, the
superposition of Fresnel reflections produces a stronger narrow band reflection. This is
also known as the Bragg peak wavelength. The basic working principle of a fibre Bragg
grating (FBG)-based sensor system is the measurand-induced shift in Bragg peak reflection.
This shift is due to the change in the effective refractive index and the period of the
grating. These sensors are used to measure strain and temperature27, 33. A schematic
representation of the mode of operation of a fibre Bragg grating sensor is illustrated in
Figure 2-5.
The Bragg wavelength λB, is related to the effective core index of refraction, n, and the
grating pitch, Λ, by the relation:
14
λ B = 2 nΛ (2.3)
Input Signal
Reflected Transmitted
Signal Signal
Amplitude (dB)
Amplitude (dB)
Amplitude (dB)
Wavelength λ Wavelength λ Wavelength λ

Input Signal Reflected Transmitted
Signal Signal
Figure 2-5: Principle of operation of an optical-fibre Bragg grating.
Since this Bragg wavelength will shift with changes in either n or Λ, monitoring the
wavelength of this narrow band spectrum will serve to determine the strain or temperature
environment to which the optical fibre is subjected34. Comprehensive reviews of the
fundamentals of Bragg gratings are given elsewhere in the literature.
For a temperature change ∆T, the corresponding wavelength shift ∆λB is given by:
∆λ B = λ B (α + ξ )∆T (2.4)
where α = (1 / Λ )(∂Λ / ∂T ) is the thermal expansion coefficient for the fibre (approximately
0.55 x 10-6 °C-1 for silica). The quantity ξ = (1 / n )(∂n / ∂T ) is called the thermo-optic
coefficient and represents the change in refractive index (optical density) with temperature.
For a germanium-doped silica core fibre, ξ = 8.6 x 10-6 °C-1.
Fibre Bragg gratings are also sensitive to axial, lateral and radial strains. Therefore, for
temperature measurement, the grating has to be isolated from strain and this is achieved by
housing the Bragg grating sealed inside a capillary tube35.
Residual strain monitoring using optical fibre sensors

The reported residual strain monitoring using embedded optical fibre sensors in the
literature is discussed herein.
Many researchers have used optical fibre strain sensors to monitor cure-induced26,36,37 and
thermally-induced24,26 residual strain development during composite processing.
15
The feasibility of measuring the residual strain using optical fibre EFPI sensors was first
demonstrated by Liu et al. in cross-ply carbon/epoxy composite. In their experiments, they
embedded two temperature-compensated EFPI strain sensors between plies number 2 and
3 (sensor 1), and 8 and 9 (sensor 2) in a cross-ply (0, 902, 02, 90, 0, 90)S carbon/epoxy
laminate. They measured the cavity length of EFPI sensors before and after curing of the
composite. From the EFPI cavity length measurements they calculated the residual strain
to be 90 µε and 550 µε in sensor 1 and sensor 2 respectively. The uncertainty in their
residual strain measurements is apparent. The authors explained that the actual orientation
of the sensor in relation to the reinforcing fibre directions was not established, which
resulted in the uncertainty in their measurements. However, the authors suggested that
these results demonstrate that the residual strain in composites can be measured using
EFPI sensors.
Lawrence et al.26 have taken one step forward and showed that EFPI sensors can be used
for in situ monitoring of residual strain development in 20 ply unidirectional and [05/905]S
cross-ply carbon/epoxy laminate during the hot press curing process. In the unidirectional
20 ply laminate two sensors are embedded between plies number 10 and 11 in both 0°and
90° directions. In cross-ply laminates two EFPI sensors are embedded between plies
number 3 and 4, and 10 and 11, in parallel to the reinforcing fibres. From their
experimental results they reported a gradual shrinkage of the laminate during the two
isothermal dwell periods of the cure cycle, which they attributed to cure-induced residual
stresses. The cure-induced shrinkage of unidirectional and cross-ply laminates was 20 µε
and 40 µε respectively. Also during the cool-down cycle they measured a compressive
residual strain, which they attributed to mismatch in thermal expansion properties between
the fibre and matrix. The magnitude of the thermally-induced residual strain was higher
than the cure-induced residual strain. Their experimental results are shown in Table 2-2.
Table 2-2: The compressive residual strain results of unidirectional and cross-
ply carbon/epoxy laminates reported by Lawrence et al. .
Compressive residual strain measurement

Laminate sequence
0° direction 90° direction
Unidirectional 20 ply 65 µε 4600 µε
Cross-ply [05/905]S 450 µε 580 µε
It can be observed from Table 2-2 that there is a significant difference between the residual
strain measurements from sensors embedded in the 0°and 90° directions of a 20 ply
unidirectional laminate. The authors explained that was because of the presence of
reinforcing fibres, the matrix is restrained to shrinking in the 0° direction. This results in a
very small compressive residual strain. On the other hand, in the 90° direction, there are no
16
restrictions for the matrix to shrink, which resulted in a high compressive residual strain.
The magnitude of residual strain measured in the 0° direction is less than the 90° direction
of the cross-ply laminate. They suggested that possible reasons for the difference in the
measurement could be due to 1) misalignment of the EFPI sensor, 2) variation in adhesion
between sensor and composite, 3) error in sensor calibration and 4) strain and temperature
gradients in the panel.
Jinno et al. have also shown that EFPI strain sensors can be used for in-situ monitoring of
residual strain similar to Lawrence et al. but in autoclave processed composites. In addition,
Jinno et al. have shown that EFPI sensors can also be used to measure the resin cure
shrinkage termination point by correlating the EFPI strain to the specific volume change of
the matrix caused by cure shrinkage. This resin cure shrinkage termination point
corresponds to a certain degree of cure of the resin. Thus, by monitoring the EFPI strain
they showed that the degree of cure of the resin could be measured.
However, Jinno et al. also reported that at the start of the isothermal period an exotherm
was observed. This exotherm reaches a peak temperature, which was 20°C higher than the
isothermal temperature. After 20 minutes the exotherm temperature cooled down to the
isothermal temperature. This suggests that measurement of cure-induced residual strain in
Jinno et al. results also include thermal shrinkage of the resin due to cool-down from the
exotherm peak temperature to the isothermal temperature. Also, it is difficult to de-couple
the cure-induced and thermally-induced shrinkage during the curing of the composite. This
suggests that the correlation of cure-induced strain to degree of cure of the resin would
result in error.
From the review of the literature, it was shown that embedded optical fibre sensors can be
used to measure residual strain development during composite processing. It was shown
that the magnitude of thermally-induced residual strain is higher than the cure-induced
residual strain. It also indicated that it is difficult to separate the cure-induced and
thermally-induced residual strains during the curing of the composite. Also, in all the above
reported residual strain measurements using embedded optical fibre sensors, the effect of
strain measured from sensors on the cure cycle parameters such as temperature and
pressure (independently and combined) are not considered. These cure cycle parameters
may have an effect on the measured residual strain.
2.4.2. Prediction of residual stresses

In the literature, numerous micro-mechanical models have been reported in both
mathematical38,39,40 and computational41,42 analysis. Wagner has excellently reviewed the
micro-mechanical models that are relevant for polymer composites. He showed that the
residual stress prediction from one-dimensional model of Tsai and Hahn and three-
dimensional model of Wagner’s are in close agreement for macro composites. These two
models are used in this study to predict the residual stresses in composites.
17
The assumptions of Tsai and Hahn’s model are:

• Both fibre and matrix are isotropic materials;
• The system is a unidirectional composite in a state of plane stress corresponding to
thin laminates, under a thermal strain;
• No mechanical stress is applied;
• There is perfect bonding between fibre and matrix;
• Modulus and thermal expansion coefficients of fibre and matrix are independent of
temperature.
The residual stress components present in the fibres and the matrix in the longitudinal
direction are given by:
σ f = (α m − α f )(T − Tref )
Ef
⎛ φ f ⎞⎛ E f ⎞ (2.5)
1 + ⎜⎜ ⎟⎟⎜⎜ ⎟⎟
⎝ φ m ⎠⎝ E m ⎠
⎛φf ⎞
σ m = −σ f ⎜⎜ ⎟⎟ (2.6)
⎝ φm ⎠
where α and E are the coefficient of thermal expansion and Young’s modulus, respectively,
φ is the volume fraction, with the f and m for the fibre and matrix respectively, and where T
is the test temperature and Tref is the stress-free reference temperature.
Most of the Tsai and Hahn assumptions were considered in the Wagner model except that
Wagner considers a concentric cylindrical model of fibre and matrix having transverse
isotropic properties. Transverse isotropic materials are a special class of orthotropic
materials that have the same properties in one plane (e.g. the x-y plane) and different
properties in the direction normal to this plane (e.g. the z-axis fibre direction). However,
both of these models only predict the thermally-induced residual stresses and not cure-
induced residual stresses.
2.5 Fibre waviness
Fibre waviness or misalignment commonly arises during processing of continuous fibre

reinforced composites. Fibre waviness in composites can be categorised as in-plane and
out-of-plane waviness43. In-plane waviness involves the co-operative movement of fibres
within the plane of the lamina. Out-of-plane waviness, also known as layer waviness or ply
waviness, involves the co-operative motion of multiple plies through the thickness of the
laminate and is prevalent in thick composites. Figure 2-6 shows a cross-section of a cross-
ply [0/90/0/90/90/0/90/0]S laminate. With reference to Figure 2-6 the presence of both
in-plane and out-of-plane fibre waviness in this cross-ply composite can be observed.
18
2.5 Fibre waviness
Figure 2-6: Fibre waviness in cross-ply laminate.
With specific reference to prepreg-based autoclave cured composites, there are a number
of possible reasons that can cause fibre waviness, for example:
• fibre movement during pressurisation and/or during the minimum viscosity phase
of the resin matrix;
• mismatch in thermal expansion between tool plate material and composite;
• excessive resin-bleeding inducing fibre movement;
• over-pressurisation during processing;
• misalignment induced during the prepreg lay-up process;
• entrapped air and foreign body inclusions.
Many researchers have shown that fibre misalignment has a significant influence on
composite properties such as static longitudinal tensile modulus44, static longitudinal
compressive strength45 and modulus46, compression fatigue47 and delamination fracture
toughness. Piggott has comprehensively reviewed the effect of fibre waviness on the static
and fatigue properties of composites. Mrse and Piggott have observed that for a fibre
misalignment angle varying from 1° to 6°, the compression strength reduces from 1.9 to
1.5 GPa. Adams and Hyer have shown that in compression-compression fatigue artificially
induced fibre waviness in composites reduced the fatigue life when compared to a
controlled composite.
19
Chun et al.48 investigated theoretically and experimentally, the effect of fibre waviness on
the nonlinear behaviour of unidirectional composites under tensile and compressive
loadings. They studied the properties in thick composites with fibre waviness in the
sinusoidal form of ratios (amplitude/wavelength (a/λ)) 0.011, 0.034, 0.059. They found
that the fibre waviness significantly affected the tensile and compressive elastic properties
and nonlinear behaviour in composite materials. They showed that composites with fibre
waviness under tensile loading tend to stretch the waved fibres so that they stiffen during
the deformation. However, coupons under compressive loading tend to undulate further
more the waved fibres so that they soften during the deformation.
2.5.1. Techniques for measuring fibre alignment

The method commonly used to measure fibre alignment or misalignment in continuous
fibre composites is the simple sectioning procedure demonstrated by Yurgartis49. This
method is based on the simple observation that a cylinder cut by a plane at an angle θ to its
diameter presents an elliptical shape (see Figure 2-7). The smaller the angle θ between the
cylinder axis (B) and the cutting plane (A), the more elongated the ellipse. The major axis
of the ellipse l, is related to the angle between the sectioning plane and the cylinder axis, θ
by:
d
sin θ = (2.7)
l
where d is the diameter of the cylinder. Therefore, the angle θ can be rewritten as
⎛d⎞
θ = sin −1 ⎜ ⎟ (2.8)
⎝l⎠
A B
Figure 2-7: Ellipse resulting from the plane section of a right circular cylinder.
20
2.6 Techniques to minimise process-induced residual stresses and fibre waviness
If an array of cylinders were cut by the same plane, then measuring the distribution of li will
give the distribution of θi where the subscript indicates the ith cylinder.
To apply these observations to the determination of the fibre angle distribution in a
composite, it must be assumed that all the fibres are cylindrical and have the same
diameter.
It is also possible, by noting the angle between a reference direction and the major axis of
the ellipse, to obtain information about the three-dimensional orientation distribution.
However, this method is extremely cumbersome and can only give information about the
orientation of a relatively small number of fibres at a few locations within the material. The
use of confocal scanning laser microscopy to image material below the surface, can give
some 3D alignment information without the need for serial sectioning was shown by
Clarke et al.50 in glass fibre reinforced composites. However, the maximum useable depth
will depend upon the transparency of the matrix, volume fraction of the fibres and the
distribution of fluorescence in the matrix. For example a maximum depth of about 150 µm
is possible with a fibre volume fraction of less than 30%. For a fibre volume fraction of
50% the maximum possible depth is only 30-40 µm.
2.6 Techniques to minimise process-induced residual stresses

and fibre waviness
The techniques reported in the literature to minimise process-induced residual stresses and
fibre waviness are: 1) optimisation of the cure cycle, 2) using a low temperature cure
prepreg, 3) electron beam curing, 4) modify the interphase properties, 5) use expandable
monomers, 6) use shape memory alloys, 7) use fibre prestressing.
2.6.1. Optimisation of the cure cycle

As discussed in section 2.4 residual stresses develop during both polymer crosslinking and
cool down of a curing cycle. Hahn investigated the effect of residual stresses on the cure
cycle, and has shown that a longer hold at the gel temperature decreases the centre
deflection of an unsymmetrical laminate due to the lower residual stresses in the composite.
The reduction in residual stress could be due to the relaxation of the cure-induced
shrinkage of the matrix. Chen et al. have also studied the development of residual stresses
in composites using a viscoelastic micromechanical model. They have shown that a higher
cooling rate results in higher residual stresses than a lower cooling rate.
White and Hahn51 studied the effect of dwell temperature, dwell time and cool-down rate
on residual stresses using unsymmetrical cross-ply laminates. They showed that by curing at
lower temperatures for longer time or utilising an intermediate low-temperature dwell
between two dwell cure cycles, the curvature of the laminate could be reduced by 25-30%.
21
Also by reducing the cool-down rate of the composite the laminate curvature could be
reduced by 25%.
Madhukar et al. 12 have shown that changing the cure cycle parameters such as dwell
temperature, number of dwells, duration of dwells and heating rate between dwells
influences directly the development of cure-induced residual stresses. In a micro-
composite, using a trial and error method they showed that an optimum cure cycle with
minimum cure-induced residual stresses could be obtained by controlling the cure
parameters. The same authors in subsequent publications showed that using results of a
feedback control system the optimum cure cycle for a composite system that reduces
residual stresses could be obtained.52,53 The feedback control system is based on maximising
the cancellation of stresses due to cross-linking shrinkage with a combination of stress
relaxation and thermal expansion while completing the cure in a short time. They also
showed that slow heating rates allow more time for the polymers to relax and relieve
stresses caused by cure shrinkage of the polymer. Similarly, many other researchers have
shown that optimisation of the cure cycle can be used to minimise process-induced residual
stresses.54
From the above discussion it can be seen that optimisation of cure cycle parameters could
be used as a method to minimise the process-induced residual stresses in composites.
However, this method of controlling the cure cycle parameters varies with different resin
systems and dimensions of the composite parts. In addition, it will be difficult to control
the temperature in thick (>4mm) composites due to greater exothermic temperatures.
2.6.2. Low temperature cure systems

Low temperature cure systems are resins that can be cured at ambient temperatures. The
duration of the cure cycle depends on the type of the resin, type of the hardener and their
mixing ratio. In low temperature resin systems, the residual stresses will develop due to the
cure-induced shrinkage of the resin. However, the thermally-induced stresses will be less
compared to an elevated temperature resin cure system because the temperature difference
between the cure temperature and room temperature will be small.
However, in low temperature cure systems, the glass transition temperature of the resin will
be lower than for elevated resin cure systems; this will reduce the mechanical properties of
the resin55.
2.6.3. Electron beam curing

Electron beam (EB) curing uses high-energy electrons from an accelerator to generate free-
radicals to initiate polymerisation and cross-linking reactions at ambient cure
temperatures.56 Because the temperature changes in the composite during EB cure can be
kept less than for thermal curing systems, the residual stress development in EB cured
22
composites will be lower than for thermally-cured composites. To enable EB curing of

epoxy resins, cationic initiators of required concentrations (generally 1-3%) need to be
added to the polymers57. These cationic initiators react when subjected to irradiation of a
beam of high-energy electrons. EB curing of composites enables fast curing and therefore a
high production rate, resin curing at ambient temperature and therefore less thermal
residual stresses. However, Goodman and Palmese58 showed that the compressive and
interlaminar shear strength (ILSS) of EB-cured composites are lower than those of
thermally-cured composites. The authors suggested that this could be the result of the
brittle nature of EB-cured cationic epoxy systems and also the existence of poor interfacial
shear strength (IFSS) of EB composites. The adhesion between the fibre and matrix has
been optimised for the thermal cure of epoxy-amine systems.
EB curing of composites is propitious in minimising residual stress development,
increasing production rate and offers the possibility of on-site repair of composite
structures. However, the adhesion between the fibre and matrix need to be optimised in
order to improve the mechanical properties such as compression and ILSS of composites.
2.6.4. Tailoring interphase properties

During the manufacturing of reinforcing fibres, sizing or coupling agents are coated on the
surface of the fibres to provide chemical interactions with the polymer matrix and to
protect the fibres during handling and impregnation. For example silane coupling agents
are commercially used on glass fibres to give good adhesion to epoxy matrixes. The net
result of this is the nature of the differing interactions; the physical mixing of matrix and
coupling agents and the nature of chemisorption at the fibre surface gives rise to an
interphase region as opposed to a distinct interface between fibre and matrix59.
Asp et al.60 have shown from a numerical model that the modulus of the interphase region
has a significant effect on the local interfacial thermal residual stresses and transverse
failure stress and strain. If the elastic modulus of the interphase region is lower than the
value of the resin, the local thermal residual stresses are minimised and vice-versa.
However, they also showed that increasing the thickness of the interphase region to greater
than 10% of the fibre radius would reduce the mechanical properties of the composites.
From both computational and experimental results, Hiemstra and Sottos61 have
demonstrated that for a low modulus interphase region, microcracks in the matrix were
minimised as a result of lower thermal residual stresses. Later Sottos et al.62 have also shown
that the glass transition temperature (Tg) of the interphase region influences the thermal
residual stresses and microcracking of the composite. From experimental results they
showed that the presence of a low Tg interphase region increases the resistance to
microcracking. On the other hand, the performance of the composite would be affected at
high temperatures (that is above the Tg of the interphase region).
23
The method of tailoring the interphase properties was shown by researchers to reduce the
residual stress development60,62. By using a low modulus interphase region, the Tg of the
interphase region is reduced, which minimises the temperature difference during the
cooling cycle. Therefore, the thermally-induced residual stresses are minimised. However, it
is not possible to change the interphase region (or coupling agents) in a commercially-
available prepreg laminate. Indeed, the modification has to be done during the fibre
manufacturing process. Also the type and properties of the coupling agents have to be
optimised for different fibre/matrix systems.
2.6.5. Shape memory alloys

Shape memory alloys (SMAs) are materials that change their crystalline structure in
response to a change of temperature or mechanical deformation63. SMAs are available in
the form of wires, strips, and films. In particular, binary Ni-Ti or tertiary Ni-Ti-X (where X
could be copper (Cu), Cobalt (Co) or Ferrous (Fe)) SMA wires are now commercially
available with diameters in the range of 50-200 µm64. The shape memory effect that these
wires exhibit is associated with the reverse transformation from austenite to martensite
during cooling or mechanical deformation.
Psarras et al. have embedded thin SMA wires into aramid/epoxy composites and the stress
distributions in the aramid fibres were obtained by means of laser Raman spectroscopy.
They have shown that by the addition of SMA wires to the composite thermal residual
stresses reduced from –25 MPa to –6 MPa in aramid fibres.
The major drawbacks of using SMA in composite materials are 1) they induce fibre
misalignment65, 2) they create a variation in stress distribution in the composite between
two SMA wires and 3) give stress concentration areas.
2.6.6. Expanding monomers

This technique involves the addition of an expanding monomer to the polymer matrix
resin. An expanding monomer usually consists of a double ring that opens as the monomer
copolymerises with the polymer matrix resin (see Figure 2-8).66 The opening of the double
ring tends to reduce the Tg of the cured resin system, which results in a decrease in the
shrinkage occurring in the cool-down cycle from Tg to room temperature67.
24
O O
R C R
O O
ring-opening
mechanism
O O O O
R C R
Figure 2-8: Ring opening mechanism of expanding monomers.
This mechanism reduces the residual stresses in the cured resin. However, it also reduces
the modulus and glass transition temperature of the matrix. Orso and Vizzini68 showed that
by the addition of expanding monomers, the strain to initiate the transverse cracking in
cross-ply composites could be delayed. However, the transverse crack growth rate was
increased. The authors suggested that this could be due to the increase in the brittleness of
the fibre-matrix interface. They also showed that the longitudinal modulus of the
composite was not affected by the addition of the expanding monomers. However, the
longitudinal strength was reduced by up to 15%.
2.6.7. Fibre prestressing

In the past decade, researchers have demonstrated theoretically that fibre prestressing can
be used as a method to reduce residual stresses in composites. Prestressed composites are
prepared by applying pre-tension to the fibres and subsequently curing the composite.
Once the composite is cured and cooled down to room temperature, the pretension is
released. Now the pretensioned fibres will tend to contract, inducing compressive stresses
to the matrix, which will minimise both the cure-induced and thermally-induced residual
stresses resulting from the cure cycle. Also, it was suggested by Dvorak and Suvorov69 that
as the fibres were prestressed and this was maintained throughout the curing process it was
also possible to minimise the fibre misalignment in composites. However, in order to apply
prestress to the fibres, a suitable prestress methodology is required, which can apply a
uniform fibre prestress and maintain it throughout the curing process.
In the above section various solutions available in the literature to minimise process-
induced issues such as residual stresses and fibre misalignment were discussed. From the
discussion it can be observed that fibre prestressing is the only possible technique, which
can counteract both process-induced residual stresses and fibre waviness. The following
section will explain fibre prestressing in detail.
25
2.7 Fibre prestressed composites
The basic principle of prestressing in composites is adapted from the principle of

production of prestressed concrete. In the Oxford dictionary of English70 the word
prestressed means “previously subjected to stressing; into which stress has been deliberately introduced
during manufacture”. In the literature, the prestressing concept is addressed by different
terminologies like “pre-strained composite” and “pre-tensioned composites”.
Prestressed composites are prepared by applying a pre-load to an n-ply laminate prior to
cure. Since the uncured polymeric matrix possesses negligible stiffness, essentially all the
pre-load applied would be carried by the fibres. The laminate would then be cured
normally, causing the polymeric matrix to gel and solidify. Following cure and cool down
the fibre pre-load is released, and the resulting elastic contraction of fibres relieves the
tensile residual stress in the matrix resulting from the curing cycle by inducing compressive
stresses. This process of prestressing is illustrated in the Figure 2-9.
Prestressed fibres Uncured matrix
Matrix tensile residual stress cured matrix
Fibre/matrix adhesion
Compressive stress
Figure 2-9: Schematic illustration of the fibre prestressing principle. 1)

prestressing the fibres in the uncured matrix, 2) matrix curing and
formation of fibre/matrix bonding and development of tensile
residual stress in the matrix and 3) release of fibre prestress at
ambient temperature, which induces compressive stress to the
matrix.
26
2.8 Theoretical analysis of fibre prestressing
The creation of compressive stresses in the matrix would impede crack propagation
through the matrix, thereby delaying or preventing the formation of matrix cracks in the
composite. This could enhance the matrix-dominated properties of composites. Table 2-3
briefly presents the papers published in the literature on prestressed composites in
chronological order. These papers are critically reviewed in the subsequent sections.
In literature, the effect of fibre prestress on residual stresses and the onset of failure in
continuous fibre composites were mathematically predicted using micro- and macro-
mechanics theory. The prestress theories reported in the literature are discussed herein.
Tuttle proposed a simple micro-mechanics theory based on a rule-of-mixtures to predict
the effect of fibre prestress on the matrix residual stresses in composites. The assumptions
used in his theory were 1) perfect fibre/matrix bonding, 2) both fibre and matrix are
isotropic linear elastic materials and 3) fibre and matrix properties are independent of
temperature. From the theory, Tuttle predicted that the final residual strain in the matrix
(εm) is given by:
− σ fpV f
εm = + ∆T .α 11 (2.9)
E11
where σ fp is the applied fibre prestress, Vf is the fibre volume fraction, E11 is the
composite Young’s modulus in the fibre direction, α11 is the composite thermal expansion
coefficient in the fibre direction and ∆T is the change in temperature.
From the model, Tuttle predicted that in a non-prestressed ( σ fp =0 MPa) carbon/epoxy

composite with 60% Vf, the matrix is subjected to a tensile residual stress of about 24 MPa
in the fibre direction, which is 60% of the ultimate strength of the epoxy resin. By applying
a fibre prestress of 1.4 GPa, he showed that the matrix tensile residual stresses could be
reduced to 0.98 MPa, which is 2.5% of the epoxy ultimate strength. Similar models were
also proposed by other researchers.
Hadi and Ashton have further developed Tuttle’s model to predict the tensile strength of
prestressed composites. From a rule-of-mixtures the tensile strength of the unidirectional
composite ( σ C ) is given by:
σ C = σ f V f + σ mVm (2.10)
where σ f and σ m are the fibre and matrix tensile strength respectively and Vm is the
volume fraction of the matrix.
27
Table 2-3: Literature review of prestressed composites
Prestress
Reference Material Research area Findings
method
Micromechanical He predicted that in non-prestressed composite (Vf = 60%), tensile
Carbon fibre/ epoxy unidirectional modelling of the effect residual stresses in matrix is 24 MPa for a temperature difference of
Tuttle (1988) not applicable
composite. of prestress on residual 150°C. Applying 1.4 GPa prestress, reduces matrix residual stress to
stresses in composites. 0.98 MPa.
Tensile strength and modulus initially increases with prestress, above
E-glass fibre/ polyester unidirectional
Jorge et al.71 Tensile properties in a prestress limit it shows a tendency to stabilise.
composite. Dead weight
(1990) fibre direction Fractographic analysis: No difference between prestressed and non-
Fibre volume fraction: 56 ± 4 %
prestressed composites.
Prestressing increased the average fracture stress and fracture strain
Schulte and Aramid fibre (0°)/ Carbon fibre (90°) / Tensile properties in V-slot
by 2.8% and 3.3% respectively.
Marrisen epoxy hybrid cross-ply composite with fibre direction and Mechanical
Increases the strain to transverse crack initiation by 0.2% strain for a
(1992) (0/90/90/0) laminate sequence. transverse cracking fastening
341 MPa prestress.
Carbon fibre/ epoxy cross-ply Mathematical modelling Prestressing enhanced first ply failure strength in composites. For a
Rose and
composite with (0/90/90/0) laminate and experimental Filament prestress of 690 MPa, first ply failure stress increased by 29%.
Whitney72
sequence. measurement of first winding However, their model does not correlate with the experimental
(1993)
Fibre volume fraction: 70 % ply failure. results.
Vinylon (poly(vinyl alcohol)) fibre
/epoxy-aluminium laminate (VIRALL). Prestressing increases initial modulus, elastic limit strain, yield
Sui et al.73 Tensile properties in
not specified strength and failure strength, together with a decrease in failure
(1995) Fibre volume fraction: 32± 1 % fibre direction
strain.
Aluminium volume fraction: 56.5 ± 1.5 %
Curvature of unsymmetrical laminates decreases by increasing fibre
Carbon fibre/ epoxy unsymmetrical prestress.
Tuttle et al.74 Tensile properties and Hydraulic
cross-ply composite. Transverse cracks reduced with prestress.
(1996) transverse cracking cylinder rig
Fibre volume fraction: 70 % No difference in ultimate tensile strength of the composite with
prestress.
…. continued
Table 2-3: Literature review of prestressed composites … continued
Prestress
Reference Material Research area Findings
method
Motahhari Measurement of applied
E-glass fibre/ epoxy unidirectional Horizontal
and pre-load and Prestressing reduced the residual stresses in the matrix and the
composite. tensile testing
Cameron75 mathematical modelling fibre/matrix interface.
Fibre volume fraction: 62 ± 2 % machine
(1997) of residual stress
33% increase in impact strength was reported for a 60 MPa
Motahhari prestressed composite when compared to non-prestressed
and composites. Above this prestress level a reduction in impact strength
composite. Impact properties tensile testing
Cameron76 was reported.
Fibre volume fraction: 62 ± 2 % machine
(1998) Fractographic analysis: Prestressed composites resulted in more splitting
and delamination compared to non-prestressed composites.
Co-mingled E-glass fibre/
Zhao and Tensile, flexural and Fibre prestressing enhances tensile strength, flexural strength and
poly(propylene) (PP) unidirectional Fibre
Cameron77 Interlaminar shear ILSS of composites by 20%, 21% and 10% respectively.
composite. alignment rig
(1998) strength (ILSS) Above an optimum prestress level the properties stabilises.
Fibre volume fraction: 34.2 %
Hadi and E-glass fibre/ epoxy unidirectional
Tensile properties in Filament
Ashton 78 composite. Improvement in tensile strength and modulus with prestress.
fibre direction winding
(1998) Fibre volume fraction: 30%, 45% and 60 %
Motahhari
and Flexural strength and modulus increases by 33% with prestress.
composite. Flexural properties tensile testing
Cameron79 Above an optimum prestress level the flexural properties decreases.
Fibre volume fraction: 60 ± 2% machine
(1999)
Mathematical modelling
Dvorak and S-glass fibre/ epoxy cross-ply and From the model it was shown that the fibre prestress reduces the
of the effect of residual
Suvorov69 quasi-isotropic laminates. not applicable residual stress in matrix and could increase the resistance to matrix
stresses on prestress
(2000) Fibre volume fraction: 50 % damage.
and failure envelopes.
E-glass fibre/ epoxy cross-ply
Jevons et al.80 laminates with [0/902/02/90/ 0/90]s Biaxial loading They reported a small improvement in low velocity impact properties
Low velocity impact.
(2002) laminate sequence. frame of composites with prestress.
Fibre volume fraction: 56%
29
Hadi and Ashton predicted that the tensile strength of a prestressed unidirectional
composite ( σ Cp ) is given by:
σ Cp = σ f V f + σ mVm + V f σ fp (2.11)
Results from Hadi and Ashton’s model shows that the unidirectional composite tensile
strength increases with an increase in fibre prestress. However, Motahhari and Cameron
have shown that not all the fibre prestress (applied prior to matrix cure) is released to the
cured matrix, some part of the prestress remains in the fibre as tensile residual stresses.
They also showed that this residual tensile stress in the fibre increases with fibre prestress.
It is well known that in the fibre direction of the composite, the main load-bearing member
is the fibre. If the fibres are initially subjected to a tensile residual stress due to the prestress
curing process then subsequent mechanical loading of the composite will result in a
reduction in the tensile strength.
Dvorak and Suvorov have predicted the effect of fibre prestressing on symmetrical elastic
laminates using laminate plate theory. From the stress analysis calculations they have shown
that the overall contribution of fibre prestress to the symmetrical elastic laminate is
equivalent to applying compressive mechanical stresses to each ply, when removing the
prestress forces originally applied to the respective plies. This reduces the matrix residual
stresses in the composite. From their prediction of ply failure using a maximum stress
criterion, they showed that fibre prestressing could increase the resistance to first ply failure
(90° laminates) by reducing the tensile residual stresses in the matrix. Also they suggested
that fibre prestressing could minimise the fibre waviness in composites, because the fibres
are subjected to prestress during the curing process, which will minimise the movement of
the fibres. This could result in an improvement of the compressive strength of composites.
From their stress analysis calculations they have shown that fibre prestress can be
optimised to control the overall deformation in symmetric laminates.
When a compressive load is applied to the prestressed composite, at first the tensile
residual stress in the fibres has to go to zero and is then subjected to compression. This
could also results in improvement in compressive strength of the composites.
From the Tuttle and the Dvorak and Suvorov models for prestressed composites, it was
shown that fibre prestressing minimises the residual stresses in the matrix. By reducing the
residual stresses Dvorak and Suvorov showed that the first ply failure strength of the
composites could be improved. This also suggests that the matrix-dominated properties
such as impact, flexural and low stress level fatigue of prestressed composites could be
improved.
2.9 Evaluation of existing fibre prestress methodologies
The manufacture of a prestressed composite entails the need for a rig to impart a
predetermined prestress during the cure cycle. To this effect, designing the right rig for this
purpose is critical for this research. Before implementing a design, a study was carried out
to ascertain whether existing approaches to solving this problem could be used for this
study. The following section briefly discusses the fibre prestress methodologies reported in
the literature to verify their suitability for the current design criteria.
2.9.1. Dead-weight method

Jorge et al. showed that prestressed composites can be prepared by applying a suitable load
to the fibre ends by attaching and tightening the fibres to steel pins with a spacing of 2 mm
to provide a uniform distribution of prestress, as shown in Figure 2-10. They prepared E-
glass/polyester resin composites using a wet lay-up process. The E-glass fibre rovings were
applied through two combs like sets of steel pins (see Figure 2-10), on parallel sides of the
lower glass plate, producing a separation of approximately 2 mm between the rovings. The
pre-load was applied to the ends of the fibres by tying it to a dead-weight. The resin was
applied to the fibres. The fibres were impregnated with polyester resin manually. The
composite was cured at room temperature for 24 hours, followed by a 3 hours post-cure at
80°C. The applied pre-load was in the range of 0-100 N.
Top View
Gla s s P la te Ma trix R es in
P
2 mm
E -g la s s F ibre S teel P ins
S ide View
Figure 2-10: Dead-weight method. 71
31
Using this methodology it is difficult to obtain a uniform fibre distribution through the
width of the composite. The bending of the fibres in the steel pin region may cause fibre
fracture. Conventional prepreg tape material cannot be used and is difficult to process in
autoclave.
2.9.2. V-slot mechanical fastening method

Using aluminium plates with two V-shaped slots machined at the ends, Schulte and
Marissen have shown a method of prestressing composites. They prepared hybrid (0°-
Kevlar/90°-Carbon epoxy) cross-ply composites with a [0/90/90/0] laminate sequence.
The prepreg laminates were laid-up between two aluminium plates. At both ends of the
lower plate the 0° plies were mechanically fastened with a V-shaped bar into these slots as
shown in Figure 2-11. The total pre-strain achieved was 1.1%. The composite was cured in
a hot-press at 125°C. As the composite was cured at 125°C, pre-strain will also be applied
by the thermal expansion of the aluminium plate. I was unable to gather from Schulte and
Marissen article whether the thermally-induced pre-strain was included in the total-pre-
strain calculation.
P P
Bar
Top Plate Prepreg Laminate V-notch
Base Plate
Figure 2-11: V-slot mechanical fastening method.
This method can be used to apply only one prestress level. Brittle fibres like glass and
carbon cannot be prestressed using this technique because the kinking of fibres in the V-
slots will result in numerous fibre fractures leading to an uneven stress distribution. In
addition, as the thickness of the composite increases, there is a possibility of variation in
applied prestress through the thickness. In other words, the top ply of the prepreg will
experience more pre-strain than the bottom ply of the laminate, which is due to the V-slot
clamp design.
32
2.9.3. Filament winding

In filament winding, a continuous band or tape of resin-impregnated fibres is laid onto a
revolving mandrel to produce a composite structure. As shown in Figure 2-12, a large
number of fibre rovings are pulled with a known pre-load, collimated into a band through
the use of textile thread board or a stainless steel comb, passed into a liquid resin bath and
a doctor blade, and then warped over the mandrel. Hadi and Ashton showed that by
applying different pre-load levels, composites with various prestress levels could be
prepared.
kg
R es in Impreg na ted
F ibre F ibre
R ollers R es in B a th
Ma ndrel
F ibre R eel
kg
Figure 2-12: Filament winding.
Hadi and Ashton prepared uni-directional prestressed composites using this method in a
square flat sided mandrel. Rose and Whitney prepared cross-ply laminates by constructing
a square mandrel, which could be rotated appropriately during the winding process. After
prestressing, the composites were cured in an autoclave.
The inherent drawback of using this method is that it is very difficult to monitor the pre-
load during the curing process and before releasing. Therefore it will be difficult to
determine the prestress released to the composite. In addition, with fibre prestress, the
fibre volume fraction of the composite will increase because of the increase in compaction
of fibres laid onto the mandrel (see Section 2.10.1). Therefore it will be difficult to
manufacture composites with various levels of prestress and with uniform fibre volume
fractions. Care should be taken in selecting the mandrel material because if the composite
is processed at elevated temperatures the mandrel thermal expansion-induced prestress
should also be included.
33
2.9.4. Hydraulic cylinder prestress rig

Tuttle et al. have shown that by using a hydraulic cylinder and a hot press, composites with
different levels of prestress can be manufactured (see Figure 2-13). The cylinder was
mounted on a horizontal frame, such that the prestressed plies remain parallel to the heated
platens of the hot press. This technique enables prepreg laminates to be manufactured. The
prepreg plies to be prestressed are wrapped around both the movable loading rod and a
fixed loading rod mounted on the far side of the hot-press. They reported that in the initial
experiments, slippage occurred between the ply and loading rod surface at high fibre
prestress levels, which was latter eliminated by knurling the surface of the loading rods.
Mova ble
F ixed Loa ding R od
Loa ding R od Hydra ulic
Cylinder
Hea ted P res tres s ed

P la tens La mina te
Figure 2-13: Hydraulic cylinder prestress rig.
The load applied to the fibres was monitored from the pressure gauge fixed in the cylinder.
Nevertheless, this technique is limited to a hot-press manufacturing process.
2.9.5. Horizontal tensile testing machine

Motahhari and Cameron showed that by using a horizontal tensile testing machine, fibre
prestress could be achieved (as shown in Figure 2-14). The electric motor provides the
mechanical energy and rotates the drum, which pulls the fibres in one direction. The load
applied was monitored by a load cell. A U-shaped mould was used to prepare composites.
A similar method was also used by Lee81 and Ali82 to prestress prepreg-based composites.
This prestress methodology is very simple and can measure the pre-load applied during the
curing process. However, this method restricts itself to room temperature and hot-press
cure composite systems.
34
Loa d Cell
F ibres
Motor
Control
B ench P a nel
Figure 2-14: Horizontal tensometer prestress rig.
2.9.6. Fibre alignment rig

Zhao and Cameron have designed a fibre alignment device as shown in Figure 2-15 for
manufacturing composites with different levels of prestress.
The fibres were first wound onto a steel frame, which was then transferred to the fibre
loading device. The frame was attached to a tensile testing machine to be stretched to the
required level, the lock nut of the locking system was then turned down to lock the bars at
the required position. The composite was cured by a compression moulding process.
Locking bolt
Fibre Spring
Figure 2-15: Fibre alignment rig.
35
Using this prestressing technique, it will be difficult to monitor the fibre prestress during
cure and prestress release and therefore difficult to determine the prestress applied. Also
the load applied by the tensile testing machine may reduce after locking the load due to
bending of the frame. In Zhao and Cameron’s work they use co-mingled fibres, which
consist of E-glass fibres and poly (propylene) (PP) fibres in 34.2:65.8 ratio. During
prestressing, initially both E-glass and PP fibres will be carrying the applied pre-load, but
during processing as the PP fibres melt and wet the E-glass fibres, the pre-load carried by
the PP fibres will be transferred to the E-glass fibres. Therefore, the prestress applied to
the E-glass fibres increases, which has to be recalculated to find the actual prestress in the
fibres. This was not discussed in their paper. Also it is difficult to manually align the fibre
to 0° while winding onto a steel frame.
2.9.7. Biaxial loading frame method

Jevons et al. 80 have shown a method for prestressing cross-ply laminates and processing in
an autoclave. Their design consisted of C-channel section frame with four clamps linked to
the frame by bolts as shown in the Figure 2-16. Prestressed composites were manufactured
by laying up cross-ply prepreg laminate, and curing the prepreg ends with aluminium end
tabs. After curing the end tabs and cooling down to room temperature, seven holes of 10
mm diameter were drilled. These end tabs were secured into the clamps by tightening the
bolts. Prestress was applied by tensioning the locking bolts between the clamps and the
frame, by a pre-determined amount. The frame was vacuum bagged and processed in an
autoclave.
Locking bolts Composite
Clamps
Figure 2-16: Biaxial loading frame method.
36
2.10 Static mechanical properties
This is the first method published in the literature to enable prestressing in two directions
and curing in autoclave. However, in this method, the tension applied to the fibres may
reduce because of the bending of the channel frame (39.4 nm per unit force), which was
calculated as shown in Appendix A. Consequently, determining the applied pre-load to the
composite panel is cumbersome. It will be difficult to control the alignment of the clamps,
which are crucial in order to apply uniform prestress in both directions. In addition to that,
as the rig has may sharp corners and complicated shapes, conventional vacuum bagging is
difficult. This raises a need for a special vacuum bagging system to prepare composites in
autoclave.
Table 2-4 shows a comparison of all the prestress methods discussed above.
Table 2-4: Comparison of existing prestressing methods.
Measurement of Measurement
Prestress Processing
prestress during of composite
methodology technique
composite processing residual stress
Dead-weight Room Temperature (RT) Yes No
V-slot mechanical Hot-press No No
fastening
Filament winding72, 78 Oven and Autoclave No No
Hydraulic cylinder Hot-press Yes No
Tensile testing RT and Hot-plate Yes No
machine75, 81, 82
Fibre alignment rig Compression moulding No No
Biaxial loading frame Autoclave No No
Researchers have investigated the effect of fibre prestress on tensile, impact and flexural
properties of composites. In this section, the above mentioned static mechanical properties
are reviewed.
2.10.1. Tensile properties

The tensile properties in unidirectional prestressed composites are studied by Jorge et al.,
Zhao and Cameron, Hadi and Ashton, and Huang et al.83.
Jorge et al. studied the effect of prestress on tensile properties in unidirectional
glass/polyester composites. They used a dead-weight prestressing method to prepare
composites with different prestress levels. Figure 2-17 shows the tensile strength and
modulus results reported from their study. They interpreted that there is an improvement
in tensile strength and modulus with fibre prestress. However, from the statistical analysis
37
using 95% confidence level it was found that their tensile strength results for all the
prestressed composites are within the standard deviation of the measurements.
550 40
35
500
30
Tensile Strength (MPa)
Tensile Modulus (GPa)

450
25
400 20
15
350
10
300
Tensile strength 5
Tensile Modulus
250 0
0 20 40 60 80 100 120
Pre-load (N)
Figure 2-17: Influence of pre-load on the tensile strength and modulus of

glass/polyester composites .
Zhao and Cameron have studied the effect of fibre prestress on co-mingled
poly(propylene) (PP)/glass fibre. They reported that as the fibre prestress increased, the
tensile strength and modulus increased up to a critical limit of applied prestress (85 MPa)
and above this limit the properties decreased. Figure 2-18 shows a trend of improvement in
tensile strength and modulus with fibre prestressing. From their results, it can be observed
that there is no significant difference between the tensile strength and modulus of all the
prestressed composites except 0 MPa prestress level. This raises a question about the
quality of the 0 MPa prestressed composite. They have not reported the method of
manufacturing 0 MPa prestressed composites and the extent of fibre waviness in the 0 MPa
composite. Therefore, their interpretation of “improvement in tensile strength and
modulus” is questionable.
38
60 1000
900
50
Tensile Strength (MPa)

Tensile Modulus (GPa)
800
40 700
600
30
Tensile Modulus
500
Tensile Strength
20 400
-20 0 20 40 60 80 100 120 140 160 180 200
Prestress (MPa)
Figure 2-18: Tensile strength and modulus as a function of prestress (Zhao

and Cameron ).
Huang et al. 83 and Hadi and Ashton studied the tensile properties of composites
prestressed using a filament winding technique. Huang et al. have reported that prestressing
increased the fibre volume fraction of the composites by increasing the compaction of
fibre bundles. Optical micrographs of various levels of prestressed composites are shown
in Figure 2-19. At low levels of prestress, there is a gap between the Zylon fibre bundles.
As the prestress increases, the fibre volume fraction of Zylon fibres increased. As a result,
the ultimate tensile strength (UTS) and modulus of the composites also increased (see
Table 2-5). Also Hadi and Ashton found that fibre prestressing enhanced the tensile
strength and modulus of the composites. However, Hadi and Ashton did not report the
effect of prestress on fibre volume fraction in their composite systems.
Table 2-5: Summary of tensile properties reported by Huang et al. 83.
Prestress (MPa) UTS (GPa) Modulus (GPa) Fibre Volume fraction Vf (%)
0 2.9 170 67.3
700 3.3 205 77.5
39
100 µm 100 µm
a b
100 µm 100 µm
c d
Figure 2-19: Cross-sections of Zylon/epoxy composite samples show the

effect of fibre prestress on the distribution of fibre in epoxy
matrix. (a) <70 MPa prestress, (b) 280 MPa prestress, (c) 700 MPa
prestress, and (d) 1000 MPa prestress (Huang et al. 83).
Schulte and Marissen and Tuttle et al. have studied the tensile behaviour of 0° direction
prestressed cross-ply composites. Schulte and Marissen studied the effect of prestress on
tensile properties and transverse cracking of hybrid fibre (Kevlar fibre in 0° and carbon
fibre 90°) epoxy cross-ply composite with a [0/90/90/0] laminate sequence. They showed
that prestressing the 0° Kevlar fibres increased the strain to initiate the transverse ply
cracking and the resistance to transverse crack propagation during tensile loading of
composites (see Figure 2-20). They suggested that the delaying of transverse cracks could
be due to the reduction in tensile residual stress in the 90° plies as a result of 0°
prestressing. They also showed that there was an improvement in the average tensile
strength and fracture strain of the composites. From Table 2-6 reproduced from their
paper, it can be observed that the average tensile strength and fracture strain shows an
improvement. However, the standard deviation of their measurements shows that the
properties of prestressed composites are within the limits of non-prestressed composites.
40
Table 2-6: Mechanical properties of the prestressed and non-prestressed

laminate reported by Schulte and Marissen. Results are an
average of seven tests.
Hybrid Kevlar/Carbon fibre epoxy cross-ply composite

Tensile Properties
Non-prestressed Prestressed
Fracture strain ε (%) 1.75 (±0.15) 1.81 (±0.1)
Tensile Strength σ (MPa) 555 (±24) 571 (±40)
Modulus E (GPa) 31 (±1) 31 (±1.1)
80
70
Number of initial transverse cracks
on 20mm specimen length
60
50
40
30
20
Non-Prestressed
10
Prestressed
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Strain ε (%)
Figure 2-20: Number of initial transverse cracks versus strain, which was
obtained by light microscopy. The mean values of four tests are
given here and the data was taken from the specimen edge
(Schulte and Marissen).
Tuttle et al. have shown a method to fabricate unsymmetrical laminates with a range of
prestress applied to 0° plies. They reported that fibre prestressing significantly reduced the
number of matrix cracks induced within the 90° ply of a [0p/90] T specimen (the superscript
‘p’ represents prestress) as shown in Figure 2-21. This was also observed by Schulte and
Marissen. They also suggested that this could be due to the reduction in tensile residual
stresses within the 90° ply as a result of 0° prestressing. However, unlike Schulte and
Marissen, they have shown that there is no correlation between ultimate tensile strength
and fibre prestress.
41
1.4
25% of Ultimate
1.2
Matrix cracks per millimetre 50% of Ultimate
1.0 75% of Ultimate
0.8
0.6
0.4
0.2
0.0
-0.2
-100 0 100 200 300 400 500 600 700
Prestress Level (MPa)
Figure 2-21: Matrix cracks induced in [0P/90]T specimens as a function of

prestress level (Tuttle et al .).
2.10.2. Impact properties

Motahhari and Cameron reported that fibre prestressing increased the impact strength of
composites up to an optimum prestress level of 60 MPa prestress and above this level, it
decreased (see Figure 2-22). The authors’ explanation was that as the fibre prestress
increased the debonding fracture overcame the transverse fracture, resulting in the
formation of new large surface areas, which absorbed more energy. Above a critical limit of
prestress (60 MPa), the impact properties declined as the ability of the interface to resist
shear debonding decreased.
The author’s suggested that the increase in the resistance to transverse fracture of the
matrix could be due to the reduction in matrix residual stresses. As a result, the fracture
mode changes from transverse to interfacial debonding, leading to more energy absorption.
Fancey84 studied the effect of viscoelastically-induced prestress on impact properties of
composites. In his work, Nylon 6,6 fibres were stretched under a load, which was released
prior to matrix impregnation. On solidification of the matrix, the viscoelastically-strained
fibres impart compressive stresses. From their results of charpy impact test they have
shown that prestressed composites absorb 25% more impact energy than non-prestressed
composites.
42
10000
9000
Impact Strength (J.m-1)
8000
7000
6000
5000
4000
-10 0 10 20 30 40 50 60 70 80 90
Prestress (MPa)
Figure 2-22: Influence of prestress on impact strength of composites.
2.10.3. Flexural properties

Motahhari and Cameron investigated the effect of fibre prestressing on the flexural
properties of glass fibre/epoxy composites, and showed that prestressing increases the
flexural strength and modulus of composites (see Figure 2-23). Zhao and Cameron also
observed this effect in PP/glass fibre composites.
Motahhari and Cameron presented two reasons for the improvement in flexural properties
1) a tensile residual stress in the fibres and 2) a reduction in matrix and interface residual
stresses. The same authors in another publication showed that as the fibre prestress
increases the tensile residual stresses in fibres increase. The tensile residual stress in fibres
produces a component normal to the direction of the fibres. This force component acts
against the bending force during the flexural test. As a result, compared to the non-
prestressed composite, more bending force is required to cause a certain amount of
deflection. Therefore, the flexural properties increase.
Also, as the applied fibre prestress increases, the compressive residual stress in the matrix
increase and consequently more resistance against crack initiation and propagation.
Therefore, more bending force was required to provide the crack with enough energy to
grow. Above an optimum level of prestressing there was a drop in strength, which they
attributed to the fibre-matrix debonding. At higher prestress levels, the residual shear stress
at the interface increases. This, upon raising the applied load, progressively results in
debonding of the fibres and matrix and reduces the overall strength of the composite.
43
Thus, they have shown that there exists an optimum fibre prestress limit to obtain the
maximum possible flexural strength and modulus, and therefore resistance to crack
initiation and opening. Above this optimum prestress limit the flexural properties are
reduced by the rapid fibre-matrix debonding damage. 79
35 800
700
30
600
Flexural Strength (MPa)

Flexural Modulus (GPa)
500
25
400
20
300
200
15
F lexura l Modulus
100
F lexura l S treng th
10 0
-10 0 10 20 30 40 50 60
Pres tres s Level (MPa)
Figure 2-23: Flexural strength and modulus as a function of fibre prestress

(Motahhari and Cameron ).
2.11 Fatigue of composites
Almost all engineering components are subjected to varying loads throughout their
working life, and fatigue is easily the most common failure in service. Fatigue in materials is
their response to cyclic loading or, in other words, fatigue is a process whereby mechanical
damage caused by repetitive or fluctuating stresses results in a material failure at lower
stress levels than would be required under static loading. The purpose of this review is to
understand the fatigue life diagrams, damage mechanisms and various factors, which
influence the fatigue properties of Glass Fibre Reinforced Plastics (GFRP). This is
followed by review of literature on the effect of fibre prestress on the fatigue performance
of composites.
44
2.11.1. Fatigue life diagrams

Traditionally the fatigue data is represented by an S-N curve (peak stress verses the
logarithm of the number of cycles to failure). A typical tension-tension fatigue (R = +0.1)
S-N curve of a unidirectional glass/epoxy composite85 is presented in Figure 2-24. With
reference to this figure, it can be observed that the fatigue behaviour of the composite
appears to show a linear relationship. Many researchers previously observed this linear
behaviour.85,86
950
850
Peak Stress (MPa)
750
650
550
450
350
2 3 4 5 6 7
Log N
Figure 2-24: Tension-tension fatigue life diagrams of glass/epoxy composites.
Mandell conducted a series of experiments on different types of glass-fibre reinforced

plastics (GFRP). He showed that when the number of cycles to failure lay between 102 and
106, the S-N curve could be approximated by a straight line of the form:
σ max = σ UTS + b log N (2.12)
where σmax is the maximum applied stress, σUTS is the ultimate tensile strength, N is the
number of cycles to failure and b is a constant. The ratio of b/σUTS was defined by Mandell
as the fractional loss in tensile strength per decade of cycles.
Daniel and Charewicz87 showed that the fatigue data of unidirectional and cross-ply
graphite/epoxy composites could also be fitted by a second-order polynomial curve:
σ max = σ equ + x log N + y (log N )2 (2.13)
45
where σmax is the maximum applied stress, σequ is the equivalent UTS, N is the number of
cycles to failure and x and y are constants. Using this equation Daniel and Charewicz
showed that the fatigue life of a composite could also be predicted.
The arrow mark in Figure 2-24 shows the run-out of the glass/epoxy composite that is the
composite has reached its matrix fatigue limit. Fatigue damage mechanisms are discussed in
Section 2.12.3.
2.11.2. Factors affecting the fatigue life of GFRP

Fatigue life can be influenced by frequency, stress concentration, stress levels, temperature
and environment.
(i) Frequency and temperature

The monotonic strengths of GFRP are markedly loading-rate dependent, by virtue of the
viscoelastic character of the matrix and because the strength of the glass fibres is rate
dependent, the fatigue response will vary with test frequency.88,89 Unlike CFRP, GFRP
composites cannot readily dissipate heat and they gradually heat up during cycling. The
structure and properties of glass reinforcement will not be influenced by this temperature
difference, but the matrix properties can be substantially altered. Mainly the residual stress
state will be altered due to the superposition of thermally-induced stresses, which in turn
will reduce the strength properties of the matrix.90 The matrix strength decrease will
adversely affect the transverse strength of unidirectional plies so that the cross-ply will
begin to fail. Thus, fatigue failure is hastened. Generally laminates dominated by mainly
continuous fibres in the test direction show lower strains and little hysteresis heating at test
frequencies around 10 Hz.
(ii) Stress Levels

The mean tensile stress has been shown to have an important effect on the fatigue
properties of fibre reinforced composites. For 0° and ±15° E-glass/epoxy laminates, the
stress amplitude at a constant life tends to decrease with increasing tensile mean stress. In
GFRP there is a significant rate effect; the greater the rate of testing the greater the
strength. Therefore, in GFRP static and fatigue tests are carried out at the same rate.
(iii) Environment
In GFRP, the moisture content has a detrimental effect mainly due to the stress corrosion
of fibres91. In matrix resins, the moisture may act as a plasticiser and will lead to higher
failure strains. This may tend to increase fatigue resistance by inhibiting local crack growth.
However, water also weakens the reinforcing glass fibres so that the net result is more likely
46
to be a reduction in fatigue resistance. 92 The water ingress may be stopped by efficient fibre
matrix adhesion and by minimising debonding.
2.11.3. Fatigue of prestressed composites

Harris93 comprehensively reviewed the fatigue and accumulation of damage in fibre
reinforced composites. He said, “there is clear evidence from fatigue tests on many reinforced plastics
that the matrix and interface are the weak links as far as fatigue resistance is concerned. In consequence,
any materials treatment or processing that can improve resistance to crack propagation of the matrix or
strengthen the adhesion at the fibre/matrix interface is likely to improve the fatigue properties”.
Tuttle from his mathematical model showed that tensile residual stresses in the matrix
could be minimised or even subjected to compressive stress by applying a controlled
prestress level to the laminates. The reduction of residual stresses in the matrix could
increase the resistance to matrix crack propagation. Many researchers3,76,77 have also shown
that fibre prestressing enhances the static mechanical properties of composites as a result
of a reduction in residual stresses (see Section 2.10). Tuttle suggested that “the fatigue life
and/or the ultimate strength of a properly prestressed laminate will be improved to that of an otherwise
identical but non-prestressed laminate”. In the following, the effect of prestress on fatigue
properties of composites is reviewed.
Sui et al.94 investigated the fatigue properties of prestressed VIRALL (Vinylon (poly(vinyl
alcohol)) fibre/epoxy-aluminium laminates). VIRALL are produced by laminating
vinylon/epoxy composite layers and aluminium alloy layers alternately. During the
manufacturing process because of the thermal mismatch, a tensile stress was induced in the
aluminium layers and compressive stresses were induced in the vinylon/epoxy layers. This
residual stress development might reduce the fatigue properties of VIRALL laminates. Sui
et al. have prestressed the vinylon fibre prior to cure and bonded this to aluminium
laminates using epoxy adhesive. The prestress was held throughout the curing process at
about 120°C and released after cooling to ambient temperatures. They have shown that the
fatigue life of the prestressed VIRALL laminates increased by eight times at a low peak
stress level of 100 MPa. On the other hand at high peak stress levels there is no significant
difference between the non-prestressed and prestressed composites.
Sui et al. have reasoned that with increasing prestress, the compressive stress induced to the
aluminium layers increases and this results in increasingly lower stress in the aluminium
layer under subsequent fatigue loading of the VIRALL. However, they suggested that there
was a prestress limit, beyond which the interlaminar shear strength will not be able to
sustain the stresses developed and the laminate will debond.
Lin et al.95 have studied the effect of thermal residual stresses on the fatigue properties of
CARALL (Carbon fibre reinforced aluminium alloy laminate). From their prediction of
thermal residual stresses, there exists a tensile residual stress in the aluminium layer and
compressive stresses in the carbon/epoxy layer. From their study, they have predicted that
47
thermal mismatch between the aluminium alloy sheet and the carbon/epoxy layer can be
reduced by placing a glass/epoxy layer between the aluminium alloy sheet and
carbon/epoxy layer. They have also shown that post-cure stretching of the carbon/epoxy
plies in the CARALL laminate after curing or prior to mechanical testing can further
reduce the tensile residual stress in the aluminium alloy sheet. This induces a compressive
stress in the aluminium layer.
20
18 2024-T3 Aluminium alloy
16 C4 (as cured)
14
Crack length, a (mm)
C4 (0.235% Prestrained)
12
10
0
3 3.5 4 4.5 5 5.5 6 6.5 7
Log (Fatigue life)
Figure 2-25: Fatigue crack propagation of aluminium alloy, CARALL

laminates as cured and prestrained (Lin et al.)
From Figure 2-25 it can be observed that the composite with 0.235% prestrain showed
greater resistance to fatigue crack growth. They suggested that the increase in resistance to
fatigue crack growth could be due to the reduction in the residual stresses in aluminium
layers of the CARALL laminate.
The experimental studies on the effect of fibre prestress on fatigue properties are so far
carried out on metal-composite laminates. As polymer composites are widely used in
aerospace structures, the demand for materials with higher resistance to fatigue increases.
To the best of the author’s knowledge so far the effect of prestress on fatigue of fibre
reinforced polymer composites has not been studied. From the above review, it can be
observed that the prestress effect on metal-composite laminates was shown to improve the
resistance to crack propagation. From the above reviewed literature, it can be observed that
prestressing the laminates could enhance the fatigue performance of the composites.
48
2.12 Damage mechanisms
This section presents a brief literature review of damage mechanisms in unidirectional

composites tested in quasi-static tension, compression and fatigue.
2.12.1. Static tension failure mechanisms

Researchers have shown that the tensile failure process in polymer composites begins with
fracture of weak fibres at an early stage of tensile load application along with matrix
cracks96,97. The fibre fracture will occur at its weakest point determined by internal or
surface flaws. The subsequent fracture will depend upon the fibre/matrix interface
adhesive strength. A schematic illustration of the tensile failure of composites with weak
and strong interfaces is shown in Figure 2-26.
Weak Interface
Matrix
cracking
Fibre
fracture
Matrix Interface
Fibre debonding
Unloaded
Sample
Strong Interface
70 -80% of UTS 80 -90% of UTS Failure
Figure 2-26: Schematic Illustration of tensile failure of composites with weak

and strong interfaces.
If the fibre/matrix interface is weak, the failure is expected to occur along the interface (see
Figure 2-26). This will lead to extensive longitudinal splitting combined with fibre fracture
giving rise to the classical “brush-like” failure. This type of failure is typically observed in
glass fibre/epoxy composites85,86. On the other hand, if the adhesion between the fibre and
49
matrix is strong, the failure may occur in the form of matrix cracks away from the interface
region with greater fibre fracture density (see Figure 2-26). This will lead to separation of
the sample into two or more pieces. This type of failure is typically observed in carbon
fibre/epoxy composites.
2.12.2. Static compression failure mechanisms

The compressive strength of unidirectional fibre reinforced composites is typically 60 – 80
% of their tensile strength as a consequence of 0° fibre microbuckling,98,99 which is of
major concern since there are many applications in which such materials are subjected to
high compressive stress. An understanding of the phenomena involved in compressive
failure is crucial to the development of improved composite materials.
Modes of failure
Jelf and Fleck100 identified elastic microbuckling, fibre crushing, matrix failure and plastic
microbuckling as failure from experiments on model composite systems designed to
exhibit one particular mode. A schematic illustration of the compression failure modes is
presented in Figure 2-27.
Shear mode Extensional mode

Elastic buckling
Fibre crushing
Failure
on
these
planes
F
Matrix failure
Plastic buckling
Figure 2-27: Compression failure modes of composites.
Generally, three macroscopic failure modes are observed in unidirectional composites. The
first mode of failure is associated with poor fibre-matrix adhesion, which will lead to
interfacial failure and longitudinal splitting. As most of the engineering composite systems
involve fairly strong interfaces, this usually is not the dominant mode of failure. The
50
second mode of failure is associated with fibre microbuckling and kink-band formation. A
significant number of previous experimental results have revealed that fibre microbuckling
and kink-band failure along the fibre direction of a unidirectional composite are the
initiating mechanisms of compressive failure that lead to global instability. The third mode
of failure is fibre “crushing” which is associated with pure compressive failure of the fibres
themselves. This occurs when the axial strain within the composite attains a value equal to
the critical crushing strain of the fibres. This higher compressive strength failure mode is
usually prevented by the occurrence of fibre microbuckling at lower stress levels due to
fibre misalignment.
Kink-band formation
Previous researchers100,101,102 have shown that fibre microbuckling or kink-band formation
are the usual compressive failure mechanisms of unidirectional fibre reinforced composites.
A schematic representation of kink-band formation is shown in Figure 2-28.
σ σ
Kink
α boundaries
σ σ
Figure 2-28: Schematic of a kink-band formation from initial fibre
misalignment. Kink orientation, α, and the boundary orientation,
β, are also indicated.
51
Published experimental evidence suggests that kink bands in unidirectional composites

loaded in compression form in three steps: elastic kinking and plastic kinking followed by
final collapse as a result of fibre fracture at the kink boundaries.99,103,104 The elastic kinking is
initiated by defects such as fibre misalignment , residual stresses105, voids and poor
bonding106. In the plastic kinking stage, the matrix within the kink band deforms plastically
in shear under the applied compressive load. Berbinau et al. analysed the shear strain
developed in the matrix due to fibre microbuckling failure and showed that the criterion of
fibre failure is initiated on the compression side at points of maximum fibre curvature and
not on the tensile stress on the convex side of the fibre curvature. This leads to the final
collapse as a result of fibre fracture.
2.12.3. Fatigue damage mechanics

Fatigue damage can be defined as the cyclic-dependent degradation of internal integrity.
The degree of influence of these changes in integrity on the mechanical response of
materials may vary greatly. This suggests that one will only be concerned with those
damages, which are directly related to strength, stiffness, or service life of the component.
Composite materials exhibit very complex failure mechanisms under static and fatigue
loading due to the anisotropic characteristics of their strength and stiffness. They
accumulate damage in a random rather than a localised fashion, and failure does not always
occur by propagation of a single macroscopic crack. The basic microstructural mechanisms
of damage accumulation in composites are fibre-matrix debonding, matrix cracking,
delamination and fibre breakage. Any combination of these could cause fatigue damage
that will result in reduced strength and stiffness. Both the type and degree of damage vary
widely, depending on material properties, laminations and type of fatigue loading.
For any composite, in the early stages of fatigue life, the weak fibres will begin to break in a
random fashion throughout the loading region as shown in Figure 2-29. Locally, the matrix
will transfer the loads shed by the broken fibres into neighbouring fibres. At reasonable
working strains, an unbroken fibre reinforcement network carries the majority of the load
in the composite.
When the stress amplitude is above the elastic levels of the matrix, viscoelastic flow occurs
preferentially between closely spaced fibres, because of the high strain magnification in
these regions. This flow will not be fully reversible when the stress is reduced and during
cyclic loading additional stress and strain concentrations develop which lead to the
initiation of debonding at applied stresses below those observed in monotonic loading. The
debonding cracks grow during cyclic loading because some flow occurs at the crack tip
during the loading half of the cycle, which is not fully reversed on unloading. As in uni-
axial tensile tests the cracks nucleate and propagate in regions of closely spaced fibres by
the growth and coalescence of individual fibre debonds. When the fibres are widely spaced,
the growth of the crack from one fibre to the next depends on the fatigue crack growth
52
resistance of the matrix. As the laminate is further stressed, matrix cracking will
subsequently occur from the weakest (transverse ply) to the strongest layers (unidirectional
ply). The majority of the fatigue life of a laminate is spent in the crack multiplication stage
where the density of cracks increases.
Damage modes during fatigue life
1-Matrix cracking, 3-Delamination, 5-Fracture

Fibre breaking Fibre breaking
0° 90° 0° 0° 90° 0° 0° 90° 0°
Damage
0° 90° 0° 0° 90° 0°
CDS
2-Crack coupling, Interfacial 4-Delamination growth,

debonding, Fibre breaking Fibre breaking
0 Percentage of life 100
Figure 2-29: Schematic representation of the development of damage in

cross-ply laminates during the fatigue life.88
As the crack density increases, cracks begin to grow into each other, forming larger cracks.
When the laminate is still further loaded, the crack density will be increased to a limiting
value where stress redistribution would limit the initiation of new cracks. This state of
damage is known as the Characteristic Damage State (CDS) and is a laminate property,
which is achieved near the first stage as shown in Figure 2-29. At this point, matrix
cracking becomes a macroscopic form of damage that could dictate the initiation of other
damage mechanisms. The subsequent damage is usually delamination and then fibre
breakage as shown in Figure 2-29.
53
The overall deterioration process is manifested as a general loss of both stiffness and
residual strength as cycling proceeds. These changes are often monitored as indicators of
fatigue damage. The point at which the residual strength of the composite has fallen to the
level of the peak cyclic stress determines the effective life of the material. In practice,
GFRP composites have a great deal of variability in the strength, and this again is provoked
by the stochastic nature of damage accumulation during fatigue. Thus, the residual strength
curve always exhibits scatter and, as fatigue regimens in service are usually variable, it
becomes a statistical problem to define the effective life of glass fibre reinforced
composites.
Damage mechanics in unidirectional composites

Talreja107 suggested that the damage mechanisms in unidirectional composites depend
strongly on the relationship between the static failure strain of composites, which reflects
the stiffness and failure strain of fibre, and matrix fatigue strain limits. The damage
mechanisms in unidirectional composites under tensile fatigue can be divided into three
regions namely fibre breakage, matrix cracking, and interfacial shear failure.
Fibre Matrix
a b c
Figure 2-30: Fatigue damage mechanisms in unidirectional composite

under tensile loading parallel to fibres. (a) fibre breakage and
interfacial debonding, (b) matrix cracking, (c) interfacial shear
failure.
Referring to Figure 2-30, fibre breakage (a) occurs at stresses exceeding the strength of the
weakest fibre in the composite. An isolated fibre break causes shear-stress concentration at
the interface close to the tip of the broken fibre. The interface may then fail, leading to
54
debonding of the fibre from the surrounding matrix. The debonding length depends on the
interfacial shear strength of the fibre-matrix interface. The magnified tensile stress
concentration at a debonded area may exceed the fracture stress of the matrix leading to a
transverse crack in the matrix (b). A matrix crack would stop at the interface at low strains,
while at high strains the stress at the crack tip might exceed the fracture stress of fibres,
leading to fibre failure. The matrix crack may now propagate under fatigue as a macrocrack
until it hits an interface, where the shear stresses may cause its propagation to a progressive
failure of the interface (c).
Fibre breakage, interfacial debonding
εc
Matrix cracking,
interfacial shear failure
Fatigue limit of matrix
εm
Log (Number of cycles to failure)
Figure 2-31: Fatigue-life diagram for unidirectional composites under tensile

loading parallel to fibres.
Based on the above mechanisms, a fatigue-life diagram was proposed and demonstrated by
Talreja. A schematic fatigue-life diagram with co-ordinates of maximum strain in the y-axis
and the logarithmic number of cycles to failure in the x-axis is shown in Figure 2-31. Strain
instead of stress was chosen as an independent variable, as both fibres and matrix would be
subjected to the same strain while stresses in the two phases would differ depending on the
volume fraction and the elastic moduli of the two phases. The diagram shows three
different bands:
a) the horizontal band centred about the composite static failure strain, εc,
corresponds to the fibre breakage and the resulting interfacial debonding. The
failure in this band was said to be catastrophic damage;
b) the sloping band corresponds to matrix cracking and interfacial shear failure which
is a progressive damage;
55
c) the horizontal line below the sloping band represents the fatigue strain limit of the
matrix, εm.
Several sets of experimental data for different fibre stiffness were shown to agree with the
basic pattern presented in Figure 2-31.
2.13 Summary
This chapter firstly presented a review of process-induced residual stresses and fibre
waviness in composites. From the review, it was shown that residual stresses and fibre
waviness could substantially degrade the structural performance of composite materials.
Also various techniques to measure the residual stresses in composites were reviewed. A
more detailed review was presented for embedded optical fibre sensor-based residual strain
monitoring in composites.
Secondly, the various techniques for minimising the process-induced residual stresses and
fibre waviness in composites were reviewed. From the discussion, it can be concluded that
fibre prestressing is the only technique, which can counteract both process-induced
residual stresses and fibre waviness.
Thirdly, a comprehensive review of the literature on fibre prestressing of composites was
presented. In this review a detailed analysis of various fibre prestressing methods was
presented. Of all the methods reviewed only a few methods monitored the applied pre-load
to the fibres throughout the curing process. However, from the analysis it was found that
none of the methods presented in the literature measured the residual strain in composites
with fibre prestress.
The reported literature on the effect of prestress on static mechanical properties of
composites was reviewed. From the literature, it was shown that by applying optimum fibre
prestress, static mechanical properties such as flexural and impact strength of composites
can be enhanced. It was shown previously in this Chapter that fibre waviness in composites
could significantly reduce the compressive properties of composites. The fibre alignment in
composites could be improved by fibre prestressing. This suggests that the compressive
properties of the composites could be improved.
From the reported literature on the fatigue properties of metal-composite laminates it was
shown that fibre prestressing enhanced the resistance to fatigue crack propagation and
fatigue life. This suggests that fatigue resistance in polymer composites could also be
improved by fibre prestressing.
In the research described in this thesis it was decided to monitor and measure both the
applied fibre pre-load and residual stress development in composites during autoclave
56
CHAPTER 3
Design and development of

the prestress methodology
OVERVIEW
In chapter 2, a review of the prestress methodologies reported in the literature
was presented. From the study it was shown that fibre prestressing could be
achieved by using many mechanical arrangements. The study also compared all the
prestress methodologies and presented these in Table 2-4. From the review it
was shown that none of the methods satisfied all the current design criteria. In
this chapter, a new approach to fibre prestressing using flat-bed methodology is
presented. This is presented in three sections.
The current design criteria are presented and discussed.
The working principle of the flat-bed prestress methodology is given.
The design calculations pertinent to the prestress rig are presented with its
limitations.
3.1 Design criteria
I
N THIS RESEARCH, it was required to design a prestress methodology satisfying the
criteria: (a) unidirectional fibre prestressing, (b) monitor the applied pre-load online,
(c) measure the residual strain development, (d) manufacture prepreg-based
composites using an autoclave and (e) vacuum-bagging using conventional methods.
CHAPTER 3 Design and development of the prestress methodology
3.1.1. Unidirectional fibre prestressing

In this research, it was required to investigate the effect of fibre prestress on process-
induced residual stresses and fibre waviness, and the mechanical performance of
unidirectional laminates. Because this will enable us to understand the fundamental
phenomena of the effect of prestress in unidirectional laminates, which could then be used
to predict the performance of multidirectional laminates using classical laminate plate
theories. Therefore, the prestress methodology should be capable of applying a
unidirectional fibre prestress.
3.1.2. Monitor and measure the applied pre-load online

In manufacturing prestressed composites, a pre-defined load is applied to the fibres prior
to cure. This load is maintained throughout the curing process and released after the
composite has been cooled to room temperature. During this manufacturing process due
to the change in temperature, pressure and creep in the clamps, the load applied to the
fibres prior to cure may not be the same as the load after the cure cycle. Most of the
prestress methods presented in chapter 2 did not monitor or measure the load applied to
the fibres during the curing process, or after the load was released. As a result, the load
released to the composite is difficult to calculate. Therefore, it was required to measure the
load applied to the fibres throughout the manufacturing process, in order to measure the
final load released to the composite. Also a suitable load sensor which is temperature
compensated (up to the cure temperature of the composite) was required.
3.1.3. Measure and quantify the residual strain development in

composites
In this study, it was required to measure the residual stresses in composites in order to
quantify the effect of fibre prestress on residual stresses. This was achieved by using
embedded optical fibre EFPI and FBG strain sensors (see Chapter 4). These sensors were
monitored throughout the curing process and during load release.
3.1.4. Manufacture prepreg-based composites using autoclave

Prepreg-based composites are used in aerospace industries because it is easy to control the
fibre orientation, fibre content and easy to fabricate into prepreg laminates prior to curing.
In this study, it was required to prepare high-quality prepreg-based composites to
understand the fibre prestress effect. Autoclave processing is a widely used method of
producing high-quality composites in the aerospace industry. Therefore, in this work,
autoclave processing was used to prepare the prestressed composites.
58
3.2 Flat-bed prestress methodology
3.1.5. Use of conventional vacuum-bagging method

In an autoclave curing process, it is required to apply a vacuum to the composite in order
to release the entrapped air in the prepreg laminates during the curing process and also for
consolidation of the laminates. This is achieved by using a conventional vacuum bagging
method, which uses a plastic sheet as a vacuum bag, tacky tape, breather and bleeder cloth
and porous film. One of the most important requirements for conventional vacuum
bagging is that the composite part should have no sharp corners/protrusions, because this
may burst the vacuum bag. Alternatively, a special vacuum bag made of a high peel
resistant material such as rubber (e.g., silicone rubber) could be used. However, most
aircraft products are manufactured using conventional vacuum bagging methods.
Therefore, it was decided to use a conventional vacuum bagging method for the prestress
rig. This dictated that an important criterion for the prestress rig design is that it should
have no sharp corners/ protrusions, which would burst the vacuum bag.
In the following section a new approach to prestressing the fibres using a flat-bed
methodology is presented. This method is evaluated to determine whether it satisfies all the
above design criteria.
The flat-bed prestress methodology (FBPM) is a basic principle of using a load screw to
apply the prestress. A schematic illustration of the FBPM is presented in Figure 3-1. In this
design, two blocks are used; of which one is fixed to a base plate and the other is movable.
The moving block slides on guide bars, which are connected to the fixed block as shown in
Figure 3-2. These guide bars are used to align the clamps of the moving and fixed blocks.
The load screw is positioned in the moving block such that the screw end sits on a load cell
positioned in the fixed block as shown in Figure 3-1. The locking bolts are used to lock the
movement of the moving block by clamping it to the base plate after loading. An
engineering two-dimensional drawing and three-dimensional solid model of FBPM are
presented in Appendix B.
59
7 8 9 10
F
View A
1 2 3 4 5
5
1 Fixed block 8 Spacer block
6 2 Load cell 9 uncured Prepreg
3 Base plate Laminate
4 Moving block 10 cured composite
end-tab
5 Load screw
11 Guide bars
6 Locking bolts
11
7 Clamp
View A
Figure 3-1: Schematic Illustration of flat-bed prestress methodology.
6
1 2 11 4
Figure 3-2: Plan view of the flat-bed prestress methodology.
60
Prestressed composites were manufactured by laying up a unidirectional prepreg laminate,

and pre-curing the ends with aluminium tabs attached as shown in Figure 3-3. The ends of
the prepreg were cured in order to clamp the prepreg onto the prestress rig and to transfer
the applied load to the fibres. After curing the end tabs, holes were drilled (see Figure 3-3)
for the clamp bolts. The end-tabbed prepreg was attached using serrated-faced clamp
plates, which in turn were bolted to the prestress rig using M12 bolts. The load was applied
by tightening the load screw in a clock-wise direction as shown in view-A in Figure 3-1. By
tightening the load screw, the movable block moved in the tensile loading direction and as
the prepreg is clamped to the moving block, the fibres were loaded. The applied load was
directly measured from the load cell connected to the end of the load screw (see Figure
3-2). Once the required load was achieved, the locking bolts were tightened. This clamped
the moving block to the base plate.
Prepreg laminate
Positioning the end tabs

Aluminium end tab
End tab curing
Cured composite
end tab
End tab hole drilling
12 mm Drilled hole
Figure 3-3: Schematic illustration of pre-fabrication stages involved prior to

clamping the prepreg laminate onto the FBPM.
Once the prestress was applied to the fibres, silicone rubber spacers were used to cover the
sharp corners of the FBPM as shown in Figure 3-4. This enabled conventional vacuum
bagging of the prepreg laminate clamped to the FBPM. A photograph of a vacuum bagged
prepreg laminate clamped to the FBPM in the autoclave is shown in Appendix B.
61
Side view
Silicone rubber spacers
Top view
Figure 3-4: Silicone rubber spacer arrangement to enable conventional

vacuum bagging.
3.3 Design calculations
This section will discuss the design calculations for determining the maximum load capacity
of the FBPM and its limitations.
In the literature, for the unidirectional E-glass/epoxy composite systems, Mottahari and
Cameron76,79 showed that the mechanical properties of composites improve up to an
optimum limit of fibre prestress and above this optimum limit the mechanical properties
indicates a detrimental effect (see Chapter 2). These optimum limits for impact and flexural
properties were shown to be 60 MPa and 20 MPa respectively. Therefore, the current
FBPM should be capable of applying the optimum limit found by Mottahari and Cameron.
In addition, it was also required to study the detrimental effect on mechanical properties
above the optimum prestress limit. For this research, it was decided to apply a maximum
prestress of 100 MPa to the fibres using FBPM.
62
Prepreg cross-sectional area (Ac)= 200 mm (width) × 2 mm (thickness) = 400 mm2.

Fibre volume fraction (Vf ) = 60%.
Fibre cross-sectional area (Af) = Ac × Vf = 400 mm2 × 0.6 = 240 mm2.
The required force (F) to apply 100 MPa prestress to the fibres is given by108:
F = σ prestress
f
× A f = 100 × 240 = 24kN (3.1)
where σ prestress
f
is the prestress applied to the fibres. In order to achieve a prestress of 100
MPa, it is required to obtain a final force of 24 kN in the fibres of 240 mm2 cross-sectional
area. The corresponding strain in the fibre (εf) is expressed by :
σ prestress
εf = (3.2)
Ef
where Ef is the Young’s modulus of fibres. The value of the fibre Young’s modulus Ef is
presented in Table 3-1.
From equation 3.2 the strain in the fibre for an applied prestress of 100 MPa is calculated
to be εf = 1315.78 µε. For a fibre length of L = 290 mm, the displacement is expressed as:
∆L = L × εf = 0.3815 mm (3.3)
This suggested that the displacement of the moving block would be 0.3815 mm for a
maximum prestress of 100 MPa. To allow for this, the moving block was designed to have
a free movement of 5 mm.
After applying the pre-determined prestress, the locking bolts are tightened to the base
plate. This applies a tensile load to the locking bolts. It is required to calculate the tensile
load applied to the locking bolts in order to determine whether the tensile load in the
locking bolt is within the maximum limit of the bolt strength.
Table 3-1: Properties of E-glass and steel.
Young’s Modulus E Thermal expansion

Material
(GPa) coefficient (α) (/°C)
E-glass 76 4.9 × 10-6
Steel 200 11.7 × 10-6
63
3.3.1. Load in the locking bolts

Figure 3-5 shows a free-body diagram of the FBPM.
Force applied (F) = 24 kN.
Moment:
M = F1 x1 - F2 x2 + F3 x3 (3.4)
x1 = 25 mm, x2 = 71.1 mm, and x3 = 2 × 82 ( as there are two locking bolts in the rig, one
in the front and the other in rear side).
M = 24 × 25 – 24 × 71.1 + (2 × 82) F3
F2
M
X2
F1
X1
F3
X3
Figure 3-5: Free body diagram of FBPM.
In a static loading, Moment will be zero.

0 = 24 × 25 – 24 × 71.1 + 164 F3
F3 = 6.74 kN
The load in the locking bolt is 6.74 kN for an applied load of 24 kN in the fibres.
3.3.2. Shear stress in the bolt thread

The average shear stress on a screw-thread is obtained by assuming that the load is
uniformly distributed over the nut height h, and that the threads would shear on the
contact position (that is between minor and major diameter of the bolts) 109. This gives:
2F
τ= (3.5)
πd r h
64
Shear stress in locking bolts

The locking bolt major diameter (dm) = 10 mm, pitch = 1.5 mm and its height h = 20 mm.
The minor diameter of the locking bolt is expressed by:
Minor diameter dr = dm – 1.226869 (pitch of the bolt)
dr = 10 – 1.226869 (1.5) = 8.1596 mm
Case Shear Stress τ (MPa)

F = 6.74 kN 26.31
F = 15 kN 58.51
Shear stress in load screw

The load screw major diameter (dm)= 20 mm, pitch = 1.5 mm and its height h = 50 mm.
Minor diameter dr = 20 – 1.226869 (1.5) = 18.1596 mm
Case Shear Stress τ (MPa)

F= 24 kN 16.82
F= 50 kN 35.05
The calculated shear stresses in the locking bolts and loading screw are 26.31 MPa and
16.82 MPa respectively. This is smaller than the yield strength of the bolts (220 MPa)
respectively.
3.3.3. Calculation of the angle of rotation

In this work, the pre-load applied to the fibres using the load screw was directly measured
from the load cell. This can also be theoretically predicted by calculating the load screw
angle of rotation. The following Table 3-2 shows the calculation of the angle of rotation θ
of the loading screw to apply a pre-determined load.
Table 3-2: Calculation of angle of rotation.
Input Calculation Output

Pitch = 1.5 mm, ∆L = 0.3815 mm 0.3815/1.5 = 0.2543
91.56°
a = ∆L/ Pitch, θ = a × 360 θ = 0.2543 * 360°
65
250
200
Angle of Rotation θ ( )
o
150
100
50
0
0 50 100 150 200 250
Prestress (MPa)
Figure 3-6: Angle of rotation of the load screw as a function of prestress.
The degree of rotation of the load screw for various loads is shown Figure 3-6. With
reference to Figure 3-6 it can be observed that to apply a prestress of 100 MPa to the
fibres, the load screw has to be rotated to an angle (θ) of 91.56°.
3.3.4. Pre-load calculation

When a load is applied to the fibres by tightening the load screw, due to the resulting
moment created in the moving block (as shown in Figure 3-5) there will be a reduction in
the applied load. This is calculated in this section by considering two cases:
• Case I: Considering the deflection caused by the moment at the moving block M1
(as shown in Figure 3-7); and
• Case II: Considering the deflection caused by the moment at the base plate on the
composite M2 (as a worst case).
Consider the moving block and base plate as a cantilever beam as shown in Figure 3-8.
When a moment M is applied to the end of the cantilever beam as a result the beam will
deflect. This deflection of the beam would cause a change in the prestress applied to the
fibres connected to the FBPM. This deflection of the beam caused by moment is expressed
by:
ML2
δ = (3.6)
2 EI
66
where L is the length of the beam, E is the Young’s modulus, and I is the second moment
of inertia.
M1
M2
Figure 3-7: Free body diagram of FBPM.
Figure 3-8: Schematic illustration of the deflection (δ) caused by the

moment (M) in the cantilever beam.
67
120
Tensile Stress (MPa) 100
80
60
40
20
Initial Stress
Resulting Stress
0
0 5000 10000 15000 20000 25000 30000
Force (N)
Figure 3-9: Fibre tensile stress as a function of applied pre-load.
The total deflection of the prestress rig per unit force was calculated to be 0.33 nm (Case I)
and 0.63 nm for (Case II). This is negligible when compared to other rig systems reviewed
in this work (see Section 2.9.7, Chapter 2). From Figure 3-9 it can also be observed that the
reduction in the tensile stress in the fibres due to FBPM bending from case II is 4%.
3.3.5. Rig thermal expansion-induced prestress

The E-glass/epoxy prepreg system used in this study is cured at a temperature of 120°C.
The flat-bed prestress rig was manufactured using mild steel material. When the prepreg is
clamped to the flat-bed rig at room temperature (20°C) and subjected to the cure
temperature of 120°C, thermal stresses are developed due to mismatch in the thermal
expansion coefficient between E-glass fibre and steel prestress rig. The epoxy resin will not
be subjected to any thermal stresses during heating cycle, because it has negligible stiffness.
This will subject E-glass fibres to tension and steel prestress rig to compression as shown
in Figure 3-10. This tensile stress applied to the E-glass fibres due to the change in
temperature is referred to here as the thermally-induced prestress. This was calculated by
using the well known composite bar model proposed by Timoshenko110. The assumptions
used in this model were:
• Perfect bonding exist between the two materials under consideration;
• Linear elastic isotropic materials;
• Temperature-independent material properties;
• Uniform temperature distribution;
68
• No external force applied.
E-glass fibre
Steel
+∆T
Figure 3-10: Model representation of prepreg clamped to prestress rig.
Consider two rectangular plates made from E-glass and steel, which are fastened together
at their ends at a temperature T°C as shown in Figure 3-10. When there is a change in
temperature ∆T, thermal stresses will be induced into the plates.
In equilibrium:
F E-glass + F Steel = 0, (3.7)
i.e, σ E-glass A E-glass +σ Steel A Steel = 0.
Where F is the force, σ is the stress and A is the cross-sectional area. The stress-strain
relationships for each material are:
σ E − glass
ε E − glass = + α E − glass ∆T (3.8)
E E − glass
σ Steel
ε Steel = + α Steel ∆T (3.9)
E Steel
where α, E and A represent the coefficient of thermal expansion, modulus and cross-
sectional area respectively.
Combining the above stress-strain relationship equations gives :
E E − glass E Steel AE − glass (α Steel − α E − glass )∆T

σ Steel = + (3.10)
E E − glass AE − glass + E Steel ASteel
69
E E − glass E Steel ASteel (α Steel − α E − glass )∆T

σ E − glass = − (3.11)
E E − glass AE − glass + E Steel ASteel
By substituting the materials properties from Table 3-1 in equation 3.10 and 3.11 the
thermal residual stresses in the E-glass (Vf = 60%) and steel (for ∆T = 100o C) were
calculated to be 51.21 MPa and –1.23 MPa respectively. In other words, the thermally-
induced prestress in the as-clamped E-glass fibres was 51.21 MPa. Therefore, the total
prestress will include both mechanical prestress (applied via the load screw at room
temperature prior to prepreg cure) and thermal prestress (applied during the curing cycle
due to the change in temperature).
3.4 Concluding remarks
In this chapter, a new approach to fibre prestressing was presented using a flat-bed
prestress methodology for manufacturing prestressed composites. The FBPM was designed
to apply a unidirectional prestress to the fibres. The design criteria for the current study
were: 1) unidirectional fibre prestressing, 2) monitor and measure the applied load online,
3) measure and quantify the residual strain in composites, 4) manufacture prepreg-based
composites in an autoclave and 5) use a conventional vacuum bagging method. All the
current design criteria were satisfied by the FBPM design. From the design calculations, it
was shown that the rig is capable of applying 100 MPa prestress to the fibres. At the
maximum prestress of 100 MPa, the shear stress in the load screw and locking bolts were
smaller than the yield strength of the bolts respectively. The reduction in tensile stress in
the fibres due to prestress rig bending was shown to be small (4% reduction for case II).
The FBPM is designed to monitor and measure the applied load to the fibres via the load
cell. From classical mechanics, the rig thermal expansion-induced prestress to the fibres
was calculated to be 51 MPa prestress for 100°C temperature change. The advantages of
the flat-bed prestress methodology are 1) one parameter to control the applied load (load
screw), 2) monitoring the applied load online, 3) enables conventional vacuum bagging and
4) allows processing prepreg-based composites in an autoclave.
70
CHAPTER 4
Experimental procedures
OVERVIEW
This chapter presents the materials and equipment used in this research.
The experimental procedure for manufacturing non-prestressed and
prestressed composites is described.
The experimental methodology for residual strain monitoring, static tensile and
compression, and fatigue tests of composites are presented.
4.1 Materials
T
HE PREPREG MATERIAL USED in this study was Fibredux 913G-E-5-30% (E-glass
fibre/ epoxy resin) supplied by Hexcel Composites Ltd, Duxford, Cambridge,
UK. This material was chosen because it is used in aerospace structural
components and helicopter blades111, and also this material has been used in many of the
previous research studies carried out in the Sensors and Composites Group, Cranfield
University, Shrivenham Campus, UK. The prepreg was supplied as unidirectional reels with
a 300 mm width and 0.125 mm ply thickness. The prepreg was stored in sealed
polyethylene bags at -18°C in a freezer according to the manufacturer’s recommendations.
Silica gel was placed within the storage container to absorb any moisture. The prepreg
storage container was removed from the freezer and left to thaw to room temperature
prior to the removal of the prepreg from the container.
CHAPTER 4 Experimental procedures
4.2 Sensors
4.2.1. Optical fibre sensors

Optical fibre sensors were used to measure the residual strain and the magnitude of pre-
strain release in composites during processing. The optical fibre sensors used in this study
were (i) extrinsic Fabry-Pérot interferometric (EFPI) strain sensors and (ii) Fibre Bragg
grating (FBG) strain and temperature sensors. These sensors were manufactured in-house
by Mr J Tetlow (EFPI) and Dr G Kister (FBG). The single mode optical fibres working at
800 nm and 1550 nm were supplied by Fibrecore Ltd, UK. The EFPI sensors were
manufactured in single mode 800 nm optical fibres. In this study EFPI sensors with a gap
length from 50 to 100 µm were used. The FBGs were manufactured using single mode
1550 nm optical fibres. Bragg grating wavelengths from 1530 to 1560 nm were used in this
study.
4.2.2. Electrical resistance strain gauges

The electrical resistance strain gauges (ERSG) used in this study were supplied by TML
Ltd. These were 120 Ω electrical resistance gauges with a 2 mm gauge length (used in static
compression testing) and 5 mm (used in pre-strain release). ERSGs were bonded to the
surface of the specimen after curing. They were used to compare the prestrain release
measured from optical fibre sensors.
4.2.3. Load Cell

A compressive load cell was used to monitor and measure the applied pre-load to the
reinforcing fibres during composite processing. The load cell (Model No: 060-4771-01-08)
was supplied by RDP Electronics Ltd, UK. The capacity of the load cell used in this study
was 44.5 kN in compression. The load cell is temperature compensated up to 121°C.
4.3 Preparation of composite
4.3.1. Laminate sequence

The laminate sequences used in this study are as follows:
• [0]16
• [0/90/0/90/90/0/90/0]S
• [0/90/0/90/90/0/90/0]4S
72
The [0]16 unidirectional laminate sequence was used to prepare non-prestressed and
prestressed composites. The static and dynamic mechanical properties were determined for
this unidirectional composite. The cross-ply laminate sequences ([0/90/0/90/90/0/90/0]S
and [0/90/0/90/90/0/90/0]4S) were used only to compare the residual strain development
with those in unidirectional composites.
4.3.2. Lay-up
The day before prepreg lay-up, the prepreg reel was taken out of the freezer and kept at
room temperature for 24hrs. This was carried out to prevent condensation onto and
moisture-ingress into the prepreg and thermally equilibrated before use. The prepreg was
cut to 300 × 300 mm (for non-prestressed composite) and 350 × 200 mm (for prestressed
composite) using a rotary cutter. The composites manufactured in this study consisted of
16 plies of prepreg to obtain 2 mm thickness of laminate. The prepreg plies were stacked
according to the laminate sequence using a vacuum assisted jig as shown in Figure 4-1. The
vacuum assisted jig was used to aid in the alignment of the prepreg during lamination (see
Figure 4-2). A roller was used to consolidate the prepregs during lamination and to expel
the entrapped air in between the prepreg plies.
2 mm Holes
Air gap To Vacuum Pump
Figure 4-1: Vacuum jig used during prepreg lamination.
73
Air gap To Vacuum Pump
Prepreg
To Vacuum Pump
Prepreg Laminate
Figure 4-2: Prepreg lay-up process using vacuum assisted jig. 1) Prepreg ply
holding to the vacuum jig and 2) Prepreg ply stacking using
vacuum jig.
4.3.3. Embedding optical fibre sensors

The EFPI and FBG sensors were embedded in the 0° fibre direction in all the laminate
sequences used in this study. The sensors were embedded in the mid-plane of the laminate
to avoid any strain caused by bending. The positions of the embedded optical fibre sensors
are as follows:
• between 8th and 9th plies of [0]16 laminate.
• between 8th and 9th ply of [0/90/0/90/90/0/90/0]S laminate.
• between 32nd and 33rd plies of [0/90/0/90/90/0/90/0]4S laminate.
The prepreg layers were laminated into two parts, below and above the sensor. This was
followed by careful placement of the optical fibre sensors on the surface of the bottom
laminate and maintaining the alignment of the optical fibre with the fibre orientation in the
laminate. Then the top laminate was sandwiched to the bottom laminate with careful
alignment. Because of the fragile nature of the optical fibre sensors, due care and attention
was essential during embedding of the sensors and processing in the autoclave. In order to
avoid any optical fibre kink in the entry and exit point of the laminate during autoclave
processing, poly (tetrafluroethylene) (PTFE) tubing was used to protect the optical fibre
lead in and lead out. Also silicone rubber resin was used to seal both the ends of the PTFE
to avoid resin getting inside the tube, which could make the ends brittle and fragile.
74
4.3.4. Prestress methodology

A flat-bed fibre prestress methodology was designed (as shown in Chapter 3) and
developed in-house to manufacture fibre prestressed composites. In this study, fibre
prestressing was applied in two ways: (i) by thermal expansion of the prestress rig, and, (ii)
mechanical loading to a predetermined level using a load screw. The prestress induced by
thermal expansion of the rig was calculated to be 51 MPa for the cure cycle of E-glass/913
epoxy resin (see Section 3.3.5, Chapter 3). Three different mechanical loads (7 kN, 14 kN
and 24 kN) were applied. The total prestress applied to the fibres is the addition of the rig
thermal expansion-induced prestress and the mechanically-induced prestress. In total four
different prestressed composites (51 MPa, 80 MPa, 108 MPa and 150 MPa) were prepared
apart from the non-prestressed (0 MPa or reference) composite.
In order to clamp the prepreg in the prestress rig (see Figure 3-1), aluminium tabs were
bonded to both ends of the prepreg over a 30 mm length. The prepreg end-tabbing was
carried out using an in-house designed end-tab jig (as shown in Figure 4-3 and Figure 4-4).
This was to avoid any fibre misalignment and composite curing away from the end-tab
region.
Once the prepreg laminate had been prepared using the procedure described in Section
4.3.2, it was then end-tabbed. Aluminium alloy (5251-H2 supplied by Metalfast Ltd,
Swindon, UK) end-tabs were cut to size (30 (l) × 200 (w) × 1.5 (t) mm) and abraded on
both sides, then degreased using acetone. The end-tabs were aligned with the ends of the
prepreg laminate and cured (at 120°C for 1 hour) using the end-tab jig. Once the end-tabs
were cured onto the prepreg ends, 12 mm diameter holes were drilled to enable clamping
of the prepreg onto the prestress rig (see Figure 3-3). After drilling, the holes were cleaned
using a de-burring tool. Prior to clamping the end-tabbed prepreg to the prestress rig, the
rig was cleaned and a silicone release agent (Rocol, PR spray supplied by RS Consumables
Ltd, UK) was applied. A torque of 20 N m was applied using a torque wrench to each of
the clamp bolts to ensure uniform pressure distribution on the end-tab.
Mechanical prestress was applied to the reinforcing fibres (in the prepreg) to a
predetermined load by means of the load screw. The load applied was monitored by use of
the load cell. Once the required prestress was applied, the locking bolts were tightened to
20 Nm. The composite was then vacuum bagged using conventional procedures (see
Section 4.3.5) and cured in the autoclave (see Section 4.3.6).
During composite curing the residual strain development was monitored in-situ using
optical fibre sensors (EFPI and FBG sensors) (as explained in Section 4.6). After curing
and being cooled to room temperature, the vacuum bag was removed. The ERSG sensors
were positioned and aligned to the optical fibre sensors embedded in the composite as
shown in Figure 4-5. These were then bonded onto the composite surface using
Cyanoacrylate adhesive (supplied by TML Ltd, UK). The prestress was released by
slackening the load screw, locking bolts and clamp bolts. The strain transferred to the
75
composite was monitored and recorded using optical fibre sensors (EFPI and Bragg
grating) and ERSG sensors.
Uniform Pressure
Cartridge Heaters Aluminium End tab
30 mm
Water Cooling Heat insulating
plastic Aluminium Block
unit
350 mm
Figure 4-3: Side view of the End-tab Jig.
Heater supply
Water supply
Figure 4-4: Top view of the End-tab jig.
76
100 mm
2 mm
ERSG
175 mm
Optical fibre
sensors
Figure 4-5: Position and alignment of ERSG sensors with embedded optical
fibre sensors.
4.3.5. Vacuum bagging

The prepreg laminate was vacuum bagged using conventional techniques to enable the
application of a vacuum during autoclave processing. The vacuum bag assembly used in
this study is schematically represented in Figure 4-6. The vacuum bag materials used in this
study were supplied by Aerovac Ltd, UK.
Peel ply: The peel ply is placed in direct contact with the composite surface. It is a Teflon
coated porous Nylon fabric (product code: B100). The Teflon coating prevents the resin
from adhering to the bagging materials.
Perforated release film: The perforated release film (product code: A2000) separates the
laminate from breather/bleeder cloth. They are perforated in order to remove air,
volatiles or excess resin out of the laminate.
Bleeder and breather cloth: These are made of non-woven polyester fabrics (product code:
Airbleed 10). The cloth serves two purposes in composites processing. Firstly, it allows
air and volatiles to be removed from within the vacuum bag throughout the cure cycle.
Secondly, they may also be used to absorb excess resin present in some composite lay-
ups.
77
Caul Plate: A caul plate must be the same size and shape of the prepreg lay-up. This was
used in the lay-up process to transmit normal pressure throughout the laminate. The
application of a uniform pressure on both sides of the laminate compacts the fibres, and
maintains a uniform fibre-to-resin ratio throughout the composite. A smooth
aluminium caul plate free from surface defects was used (supplied by Metalfast Ltd,
Swindon, UK).
Teflon film: This is porous Teflon coated glass fabric (produce code: 200TFP-I). Teflon is a
polymer of a tetrafluoroethylene monomer. This Teflon film ensures easy removal of
the laminate from the tool plate. This film was used as a release film on tool plates.
Vacuum Bag Sealant: The vacuum bag sealant tapes are rubber-based adhesives that are used
to form a tight seal between the vacuum bag and the tool plate (product code: LTS90B).
Vacuum Bag Film: This is a film used to contain the vacuum during the cure process. A
Nylon-based vacuum bag film was used in this work (product code: Capran 512).
Bag Vacuum Line

Vacuum Bag Caul Plate Breather cloth
PTFE Film
Sealant Perforated Peel Ply Prepreg Tool Plate

Release Film Laminate
Figure 4-6: Detail of vacuum bag lay-up.
4.3.6. Autoclave processing of composites

In this study composites were prepared using an autoclave. The autoclave (Serial No: 1519)
was supplied by Aeroform Ltd, UK. After vacuum bagging the composite, it was placed
inside the autoclave bed and a vacuum (-0.084MPa (-850 millibars)) was applied. The
manufacturer’s recommended cure cycle for the E-glass/913 epoxy prepreg system112 is 1
hour at 120°C, reached at a rate of ~2 K⋅minute-1, with a pressure of 0.69 MPa (101 PSI) as
shown in Figure 4-7.
78
4.4 Evaluation of prestress methodology
140 0.8
Temperature
120 Pressure
0.6
Vacuum
100
Vacuum (MPa) Pressure (MPa)

Temperature (oC)
0.4
80
60
0.2
40
0
20
0 -0.2
0 1 2 3 204
Time (hours)
Figure 4-7: The cure cycle for E-glass/913 epoxy prepreg system.
4.3.7. Coding of the composite panels

In order to identify the composites with different levels of prestress from each other, a
code system was used. This is described as follows
URX where U = unidirectional, R = reference (non-prestressed), X = panel number.
For example, UR6 represents unidirectional reference panel number 6.
UPTX where P = prestressed and T= prestress rig thermal expansion-induced prestress.
UPX_FkN where P = prestressed, FkN = pre-load in kN
For example, UP6_14kN represents unidirectional prestressed panel number 6 with 14 kN
pre-load.
4.4 Evaluation of prestress methodology
In this section the procedure for evaluation of prestress rig is presented. The evaluation of
prestress rig involves (i) calibration of load cell at room temperature and elevated
temperatures and (ii) evaluation of strain distribution in the prestress rig.
79
4.4.1. Load cell calibration

The load cell was calibrated by applying a static compression load using an Instron testing
machine, model 8501, and comparing the load from the load cell and the Instron testing
machine. The test was conducted at room temperature. The test rate was 0.125 kN/second.
The static compression test was also conducted at 70, 100 and 121°C, to verify the
temperature compensation of the load cell. This temperature compensation test of the load
cell was carried out using an Instron model 1195 equipped with an oven.
4.4.2. Prestress rig evaluation

The flat-bed prestress rig designed and developed in this study (as shown in Chapter 3) was
evaluated by applying a tensile load to a steel plate (clamped in the prestress rig) and
measuring the strain distribution through the width of the plate. The steel plate was
instrumented with ERSGs in the positions as shown in Figure 4-8. The static tensile test
was conducted at room temperature. In order to enable clamping of the steel plate, holes
(12 mm diameter) were drilled. The steel plate was secured into the clamps by tightening
the bolts to 20 Nm torque. Pre-tension was applied to the steel plate by tightening the load
screw while monitoring the load from the load cell. The strain from ERSGs bonded in
different positions on the steel plate was recorded during the tensile test.
200 mm
1 2 3
350 mm
125 mm
4 5 6
130 mm
40 mm
Figure 4-8: Position of ERSG sensors in steel plate.
80
4.5 Evaluation of sensors
4.5 Evaluation of sensors
Optical fibre sensors and load cell were used during processing of composites to monitor
and measure the residual strain development and applied pre-load respectively. Because the
composites are processed in an autoclave at elevated temperature (120°C) with air pressure
and a vacuum, the effect of these process parameters on the sensors needed to be
investigated prior to using these sensors to monitor residual strain development. This
section gives the experimental procedure to evaluate these sensors.
4.5.1. Temperature and pressure tests

The test was conducted on a cured composite panel (290 mm × 200 mm × 2 mm). The
sensors used in this test and their positions are described in the following:
Embedded Sensors: EFPI strain sensor and FBG (strain and temperature) sensors are
embedded between the 8th and 9th ply of a 16 ply unidirectional composite and placed in
the position as shown in the Figure 4-9.
Reference sensors: The reference EFPI strain and FBG (strain and temperature) sensors were
placed on top of the composite in the same position as the embedded sensors. They
were secured using high temperature polyimide tape.
Load Cell: The load cell was placed at the end of the loading screw in the flat-bed prestress
rig as shown in Figure 3-1. The load cell is used in pressure test. No prestress is applied.
The Load cell was not in contact with the load screw.
Thermocouple: A thermocouple (K type) was surface mounted on top of the composite and
placed in the position shown in Figure 4-9, using high temperature polyimide tape.
The test conditions were the same as the composite cure cycle (Temperature - 120°C for 1
hour with a heating rate of 2 K. minute-1, Pressure - 0.69 MPa for 1 hour with a pressure
rate of 0.015 MPa per minute). The temperature and pressure tests were carried out for two
repetitions.
4.5.2. Vacuum test

The test was conducted on a prepreg (un-cured) laminate. Before starting the curing
process vacuum is applied to the composite in order to remove the entrapped air in the
prepreg laminate. In order to understand the effect of a vacuum on the optical fibre
sensors during this process, the vacuum test was conducted on the prepreg. An FBG strain
sensor was embedded between the 8th and 9th ply of 16 ply prepreg, and placed in the
position as shown in the Figure 4-9. The vacuum test conditions are -850 millibars (-0.084
MPa) for 1 hour. The load cell was in contact with the end of the loading screw in the flat-
bed prestress rig as shown in Figure 3-1.
81
Thermocouple Optical fibre

strain sensor
Optical fibre
temperature
350 mm sensor
175 mm
Laminate
200 mm
Figure 4-9: Position of sensors in the laminate.
4.5.3. Cure cycle test

The test was conducted on a cured composite panel (290 mm × 200 mm × 2 mm).
Embedded Sensors: EFPI strain sensor was embedded between 8th and 9th ply of 16 ply
laminate and placed in the position as shown in the Figure 4-9.
Reference sensors: The reference EFPI strain sensors were placed on top of the vacuum bag in
the same position as the embedded sensors. They were secured using high temperature
polyimide tape.
Load Cell: The load cell was placed at the end of the loading screw in the flat-bed prestress
rig as shown in Figure 3-1.
Thermocouple: A thermocouple was surface mounted on top of the vacuum bag and placed
in the position as shown in Figure 4-9, using high temperature polyimide tape.
The test conditions were the same as the composite cure cycle (Temperature - 120°C for 1
hour with a heating rate of 2 K. minute-1, Pressure - 0.69 MPa for 1 hour with a pressure
rate of 0.015 MPa per minute and vacuum -850 millibars (-0.084 MPa) maintained
throughout the test).
82
4.6 Residual strain monitoring
4.6 Residual strain monitoring
The process-induced residual strain in composites was monitored in-situ using EFPI and
FBG grating sensors. The residual strain development was monitored in unidirectional [0]16
and cross-ply [0/90/0/90/90/0/90/0]S, [0/90/0/90/90/0/90/0]4S laminates. Table 4-1
shows the number of panels used to record in-situ residual strain development, and the
sensor type and position in each panel.
The EFPI and FBG interrogation systems used to monitor the gap length and Bragg
wavelength shift respectively are shown in Figure 4-10 and Figure 4-11. Interference
spectra from the EFPI sensors were recorded using a charged-coupling-device (CCD)
spectrometer (supplied by Ocean Optics Inc., The Netherlands) and was powered by an
super-luminescence diode (SLD) light source with a centre wavelength of 850 nm. The
wavelength of the reflected narrow band spectrum from the FBG was monitored using a
FBG interrogation system (supplied by Fiberpro Ltd., South Korea) and was powered by a
class 3A Laser diode. The 2 × 2 coupler was manufactured in-house by Dr R Chen. A
custom-written LabVIEW™ (version 6.1) program was used to record the data from the
interrogation unit.
Table 4-1: Summary of the panels used to record in-situ residual strain
development, and the type of sensor and its position in the
panel.
Panel Code Middle 40 mm from

the edge
UR1 EFPI
UR8 EFPI
UR9 EFPI
UPT8 EFPI
UPT11 EFPI
UP1_7kN EFPI
UP2_7kN EFPI
UP2_14kN EFPI EFPI
UP3_14kN EFPI EFPI
UP4_14kN EFPI
UP5_14kN EFPI
UP6_14kN EFPI
UP7_14kN EFPI
UP9_14kN EFPI + FBG
UP12_14kN FBG
UP1_24 kN EFPI
83
(1) (2)
(5)
(3)
(4)
(6)
Figure 4-10: Schematic of the arrangement used to record spectra from the
EFPI sensor: (1) the SLD light source, (2) the EFPI sensor, (3) the 2 ×
2 coupler, (4) anti-reflection terminal (broken fibre covered with
index matching gel) used to prevent any back-reflection, (5)
CCD spectrometer, and (6) data acquisition.
(1) (2)
(5)
(3)
(4)
(6)
Figure 4-11: Schematic of the arrangement used to record spectra from the
FBG sensor: (1) the 1550 nm Laser diode, (2) the FBG sensor, (3)
the 2 × 2 coupler, (4) anti-reflection terminal (broken fibre
covered with index matching gel) used to prevent any back-
reflection, (5) the FBG interrogation unit, and (6) the data
acquisition.
84
4.7 Differential scanning calorimetry
4.7 Differential scanning calorimetry
A differential scanning calorimeter (Diamond DSC supplied by Perkin Elmer Ltd, UK) was
used to study the 913 epoxy resin cure characteristics. The DSC was calibrated using high
purity indium and tin. A 5 mg sample was cured in a conventional aluminium pan, using
the 913 epoxy resin cure temperature cycle (see Section 4.3.6). The heating rate used was 2
K.minute-1.
4.8 Quality control and specimen preparation
After preparation of the composite panel, the quality of the panel was checked, prior to
specimen cutting, polishing and end-tabbing. This section explains the procedures used for
panel quality control and specimen preparation.
4.8.1. Quality control

The quality of the composite panels was examined by employing both non-destructive
(ultrasonic C-Scanning) and destructive (optical microscopy, scanning electron microscopy
and resin burn-off) methods. This is described in the following section.
Non-destructive testing
An ultrasonic C-Scan was used to examine the extent of void and delamination defects in
composite materials. The C-Scan equipment (Model No: UPK-II/D) was supplied by
Physical Acoustics Corporation, UK. The composite panel was placed in a water tank and
an ultrasonic transducer working at 5 MHz was used to scan the panel. Distilled water was
used as a coupling medium between the ultrasonic transducer and the composite panel.
When the signal reaches a void or a delamination, the signal is attenuated. This information
is processed to produce a two-dimensional image of the panel showing the void and defect
positions.
Destructive testing
Destructive testing methods were used to measure the fibre volume fraction, void volume
fraction and fibre alignment of composite panels. Optical microscopy, scanning electron
microscopy (SEM) and resin burn-off (pyrolysis) methods were used to destructively
examine the composite panels. For microscopy examination, samples of 15 × 15 mm
dimensions were sectioned from the composite panel and mounted in a polyester resin
supplied by Struers Ltd, UK. The specimens were polished using different grades of SiC
abrasive paper from 220 to 2400 grit. The final polishing stage was carried out using water-
85
based diamond suspensions, starting from 6 µm and going down to 0.25 µm. The polishing
equipment (Rotopol-21 and RotoForce-4) was supplied by Struers Ltd, UK.
Once the samples were polished they were dried and stored in laboratory conditions. The
polished samples were then examined using optical microscopy and SEM. The optical
microscope used in this study was a Leica DMLM set with a camera and image acquisition
software Leica Qwin (version 2.3) supplied by Leica Ltd, UK. The SEM used in this study
was a LEO 435VP supplied by Carl Zeiss SMT Ltd, UK.
(i) Fibre volume fraction (Vf)

The fibre volume fractions of manufactured composites were determined using two
techniques namely, (i) burn-off and (ii) image analysis.
In the burn-off method, a piece of composite (15 (l) × 15 (w) × 2 (t) mm), conditioned at
ambient laboratory conditions, was weighed accurately using an electronic balance with a
measurement accuracy of 0.0001 gm. This was placed in a clean heat-resistance crucible in
the furnace. The furnace temperature was increased from room temperature to 650°C and
held for 6 hours until all the resin was decomposed. The residue, which consisted of glass
fibres, was weighed and the resin weight was calculated. The fibre volume fraction of glass
(Vf) was determined using the following equation:113
⎛Wf ⎞
⎜ ⎟
⎜ρ ⎟
Vf = ⎝ f ⎠
(4.1)
⎛ W f ⎞ ⎛ Wm ⎞
⎜ ⎟+⎜ ⎟
⎜ρ ⎟ ⎜ρ ⎟
⎝ f ⎠ ⎝ m⎠
where Wf, Wm, ρf and ρm are the weight fraction and densities of the fibre and matrix resin,
respectively. The average fibre volume fraction was calculated from ten individual
specimens.
In the image analysis method, a transverse cross-section of the unidirectional composite
taken from the scanning electron microscope (SEM) was examined using Leica image
analysis software. The average fibre volume fraction was calculated from a minimum of ten
micro-frames (540 × 360 µm) for each prestress level.
(ii) Void volume fraction (Vv)

Microscopic images from the SEM were used to calculate the void volume fraction using
the Leica image analysis software. The average void volume fraction was calculated from a
minimum of ten micro-frames (540 × 360 µm) for each prestress level.
86
4.8 Quality control and specimen preparation
(iii) Measurement of fibre alignment

In this work, the Yurgartis method for measuring the fibre alignment in composites was
used. The principle of Yurgartis method was presented in Section 2.5.1, Chapter 2. The
specimens were prepared by cutting samples from the unidirectional laminate parallel to the
fibres (see cross-section A-A in Figure 4-12). The samples were then potted in a room
temperature curable polyester resin (supplied by Struers Ltd) with the fibres aligned parallel
to the polishing direction as shown in Figure 4-13. This specimen was then polished using
the procedure described above. This polished section (A-A) was then examined under the
scanning electron microscope. The major axis length l was measured manually and
substituted in equation 2.8 to determine the fibre orientation. The fibre, which is spanning
throughout the length of the micro-frame, is assumed to have aligned in the 0° direction.
The average fibre alignment was calculated from a minimum of six micro-frames (876 ×
660 µm) for each prestress level.
Figure 4-12: Definition of the parallel Figure 4-13: Illustration of fibre orientation
cutting angle. in the cross-section A-A of
Figure 4-12.
4.8.2. Specimen cutting and polishing

The laminates were cut to size for testing with a water-cooled diamond saw, avoiding plate
edges and areas indicated as defective by the ultrasonic C-scan technique. Both sides of
each specimen were polished to 2400 grit SiC abrasive paper to avoid any edge defects that
could initiate premature failure.
87
4.8.3. Coding of composite specimens

In order to identify the composite specimen position from the composite panel, a coding
system of numbers was used. The specimen numbers in a composite panel are shown in
Figure 4-14. This number was added to the composite panel code as shown below.
5 mm
1 2 3 4 5 6 7 290 mm
250 mm
20 mm 15 mm
Figure 4-14: The position of specimens taken from the composite panel.
URX_Y where U = unidirectional, R = reference (non-prestressed), X = panel number and

Y= specimen number. For example, UR6_5 represents specimen number 5 taken from
unidirectional reference panel number 6.
UPTX_Y where U = unidirectional, P = prestressed, T = prestress rig thermal expansion-
induced prestress and Y = specimen number.
UPX_FkN_Y where P = prestressed, FkN = pre-load in kN and Y= specimen number.
For example, UP6_14kN_5 represents specimen number 5 taken from unidirectional
prestressed panel number 6 with 14 kN pre-load.
4.8.4. End-tabbing
The static tension and fatigue test specimens were end-tabbed with aluminium alloy tabs
and for the static compression test GFRP (woven-fabric) end-tabs were used. Conventional
88
4.9 Mechanical testing
end-tabbing procedures were used. The adhesive, 3M Scotch-weld epoxy-based system was
used to bond end-tabs on the composite specimen. The end-tabs were cured at room
temperature for 24 hours, followed by post-curing at 65°C for 2 hours. Specimens were
stored in a desiccator at ambient laboratory conditions until required.
All tests were carried out in ambient laboratory atmospheric conditions.
4.9.1. Quasi-static tensile test

The static tensile properties were obtained using a servo-hydraulic Instron testing machine
(model 8501). The static tensile tests were conducted according to the Composite Research
Advisory Group 114 (CRAG) standard. The dimensions of the static tensile test specimens
were 250 mm (l) × 20 mm (w) × 2 mm (t) with a gauge length of 150 mm as shown in
Figure 4-15. The tensile tests were conducted under position control with a crosshead
speed of 5mm.min-1. The strain was measured with an extensometer over a 50 mm gauge
length. The load, strain, position and time were recorded using custom-written
LabVIEW™ software (version 6.1). Six specimens were tested with the aforementioned
conditions to measure the tensile properties of each prestressed composite.
50 mm 150 mm
20 mm
1.5 mm 2 mm
Figure 4-15: Test specimen geometry for static tension and tension-tension
fatigue testing.
The properties measured from the tensile test were: 1) secant modulus, 2) ultimate tensile
strength, 3) failure strain and 4) strain at deviation from linearity. The strain at deviation
from linearity (εSDL) is defined as the strain at which the onset of non-linearity occurs in the
stress-strain curve during a tensile test of a composite specimen. This is illustrated in Figure
4-16.
89
Straight line
Linear region Stress-Strain

Curve
σ MPa
Non-linear region
εSDL
ε%
Figure 4-16: Illustration of strain at deviation from linearity.
4.9.2. Quasi-static compression test

The compressive test was carried out in a Zwick machine model 1488 at the Department of
Aerospace Engineering, Imperial College, London. The compression test was conducted
using the Imperial College of science, technology and medicine115 (ICSTM) fixture. The
dimensions of the compressive test specimens were 90 mm (l) × 10 mm (w) × 2 mm (t)
with a gauge length of 10 mm as shown in Figure 4-17.
10 mm
10 mm
90 mm
1.5 mm 2 mm
Figure 4-17: Test specimen geometry for static compression testing.
Both ends of each specimen were ground using a surface grinding machine to avoid any
edge defects that could initiate premature failure. The specimens were instrumented with
two ERSG sensors of 2mm gauge length (one in the front and the other in rear). The
compression tests were conducted under position control with a crosshead speed of
1mm/minute. The load and strain were recorded during the test. A minimum of five
specimens were tested to measure the compressive properties for each prestress level. The
90
properties studied in a compression test were 1) ultimate compressive strength, 2) modulus

of elasticity.
4.9.3. Fatigue test

The fatigue tests were carried out in a servo-hydraulic Instron testing machine (model
8501). The specimen dimensions for tension-tension fatigue were the same as the static
tensile specimens (see Figure 4-15) and for tension-compression fatigue the dimensions
used were 200 mm (l) × 20 mm (w) × 2 mm (t) as shown in Figure 4-18.
50 mm 100 mm
20 mm
1.5 mm 2 mm
Figure 4-18: Test specimen geometry for tension-compression fatigue

testing.
The fatigue tests were conducted according to CRAG and ISO 13003116. Stress ratios (R)
of +0.1 and –0.3 were employed in this investigation. The fatigue tests were carried out
under constant amplitude loading with a sinusoidal waveform, and a loading rate of 250 kN
s-1. Seven different stress levels (from 35% to 90% of the ultimate strength of the
composite) were used to determine the fatigue life diagram. At each stress level, three
specimens were tested. A custom-written LabVIEW™ (version 6.1) program was used to
monitor stiffness, amplitude stress and strain, and surface temperature changes as a
function of fatigue life in real-time.
91
CHAPTER 5
Evaluation of prestress
methodology and sensors
OVERVIEW
Results from the evaluation of strain distribution in the flat-bed prestress rig at
different tensile loads to a rectangular steel plate instrumented with strain
gauges are discussed.
The design and development of an end-tab jig to circumvent (i) fibre
misalignment in and near the end-tab region and (ii) pre-cure away from end-
tab region are described.
The effect of autoclave process parameters such as temperature, pressure and
vacuum on the load cell and optical fibre (EFPI and FBG) sensors are
presented.
5.1 Results and discussions on prestress rig evaluation
5.1.1. Load cell calibration

The load cell calibration was carried-out according to the experimental procedure given in
Section 4.4.1, Chapter 4. Figure 5-1 shows the results from the load cell calibration test.
From Figure 5-1 it can be observed that the load recorded from the load cell is in good
agreement with the load measured from the Instron machine.
CHAPTER 5 Evaluation of prestress methodology and sensors
-30
-25
-20
Load Cell (kN)
-15
y = 0.9782x + 0.0147
R2 = 1
-10
-5
0
0 -5 -10 -15 -20 -25 -30
Machine Load (kN)
Figure 5-1: Comparison of load from load cell and Instron testing machine.
16
o 2
At T = 27 C R =1
14 o 2
At T = 70 C R = 0.9997
o 2
At T = 99 C R = 0.9996
12 o
At T = 121 C
2
R = 0.9989
Measured Load (kN)
10
4 27 C Temperature
2 70 C Temperature
99 C Temperature
0
121 C Temperature
-2
0 2 4 6 8 10 12 14 16
Applied Load (kN)
Figure 5-2: Comparison of load measured from the load cell and the
applied load (Instron machine) at different temperatures.
The temperature effect on the response of the load cell was also studied according to the
procedure given in Section 4.4.1. Figure 5-2 presents the results from the temperature
94
response of the load cell. With reference to the figure it can be observed that the load cell
responds linearly during the static tests conducted at different temperatures (27°C, 70°C,
99°C and 121°C). It can be observed that there is a change in load of about -0.8 kN for a
94°C temperature difference. This shows that for a 94°C temperature difference the load
measured will change by -0.8 kN.
5.1.2. Prestress rig evaluation

The calibrated load cell was used in the prestress rig to apply a pre-defined static tensile
load to a steel plate (clamped to the prestress rig) to study the strain distrubution in the
steel plate. Figure 5-3 shows the strain measured at different positions (see Figure 4-8) in
the steel plate during the static tensile test in the prestress rig. With reference to Figure 5-3
it can be observed that the strain measured at different positions on the steel plate for an
applied load are in good agreement. This shows that the strain distribution through the
width of the steel plate is uniform. The load-strain curve is linear because the applied loads
were within the elastic region of the steel plate.
250
ERSG1
ERSG2
ERSG3
200 ERSG4
ERSG5
ERSG6
Linear Regression
150
Strain (µε)
R2 = 0.99
100
50
0
0 2 4 6 8 10 12 14
Load (kN)
Figure 5-3: Strain measured from the ERSG as a function of the applied load
measured from the load cell.
In order to visualise the strain distribution through the plate area two-dimensional plots
were drawn at different load levels as shown in Figure 5-5, Figure 5-6 and Figure 5-7. It
was assumed that the strain distribution is symmetric throughout the plate. Figure 5-4
95
shows the positions of the ERSGs and the corresponding symmetrical positions (marked
with superscript-1).
200 mm
41 61 51 61 41
1 31 2 3 11
350 mm
125 mm
4 61 5 6 41
130 mm
40 mm
Figure 5-4: Position of ERSGs and the corresponding symmetrical positions in

steel plate.
Load = 2 kN
300
250
250-300
200-250 200
150-200
100-150 150 Strain (µε)
50-100
100
0-50
50
70 0
100 270
Width (mm)
145
130 Length (mm)
20
Figure 5-5: Strain distribution in steel plate at an applied load of 2 kN.
96
Load = 8 kN
300
250-300 250
200-250
150-200 200
100-150
50-100 150 Strain (µε)
0-50
100
50
70 0
100 270
Width (mm)
145
130 Length (mm)
20
Load = 15 kN
300
250
250-300
200-250
200
150-200
100-150 150 Strain (µε)
50-100
0-50 100
50
70 0
100 270
Width (mm)
145
130 Length (mm)
20
From Figure 5-5 to Figure 5-7 it can be observed that the strain distribution in the plate is
uniform with a variation of about 10% in strain through the length and 4% variation
through the width of the plate. This small variation in the strain distribution in the steel
97
plate could be due to Poisson’s effect and small misalignment of the strain gauge surface-
mounted on the steel plate.
5.2 Prepreg end-tab jig design and evaluation
In this study initially, prepreg end-tab curing was carried out using the hot-press method.
During this curing process only the end-tab region was placed in the hot-press. The rest of
the prepreg laminate was supported using an aluminium plate with PTFE spacers (to
prevent the curing of the prepreg away from the end-tab region). Once the end-tab was
cured, the prepreg laminate was clamped onto the flat-prestress rig and cured in the
autoclave. The first prestressed composite panel was prepared after designing and
evaluating the flat-bed prestress rig. The composite panel manufactured was warped as
shown in Figure 5-8. This section investigates the cause of composite warping and presents
a design and development of a flat-bed end-tab jig to eliminate warping in composites.
The major reason for the composite bending during manufacturing could be due to the
eccentric release of prestress caused by fibre misalignment near the end-tab regions. In
order to identify the cause of the bending, at first samples were cut in the end-tab region
and examined in an optical microscope. From the examination (see Figure 5-9), it was
found that fibre misalignment exists near the end-tab region. This was observed
throughout the width of the composite near the end-tab region.
Figure 5-8: Warped composite.
98
Figure 5-9: Microscopic evidence of fibre misalignment near end-tab

region.
From Figure 5-9 it is also apparent that there exists a difference in the thickness of the
composite panel between the hot-press cured and the autoclave cured regions. This is due
to high pressure being applied in the end-tab region during the hot-press cure. Also during
the hot-press cure, the heat is transferred away from the end-tab region. This leads to resin
melting, which instigates fibre movement, and subsequently the resin near the end-tab
region cures. This freezes the fibre misalignment near the end-tab region. When a
pretension is applied there may be a possibility of reducing the fibre misalignment near the
end-tab region but not completely, because the composite near the end-tab region is pre-
cured. Once the composite is processed in autoclave and cooled-down to room
temperature, the prestress is released. This induces compressive stress to the composite in
the direction of the fibres. Because the fibres near the end-tab region are misaligned, the
prestress is released eccentrically, which caused composite bending.
5.2.1. End-tab jig design

In order to circumvent the fibre misalignment near the end-tab region and composite pre-
curing away from the end-tab region, a custom-designed end-tab jig was manufactured as
shown in Figure 5-10. The jig consists of a heating and cooling unit in both the ends of the
jig. The heating element is connected to a temperature controller and the cooling system is
connected to a temperature-controlled water cooler. After laminating the prepreg the
aluminium end tabs were carefully attached to the ends of the prepreg. This was then
placed inside the end-tab jig such that the end-tabs sits between the heating blocks and the
99
prepreg near the end-tab region sits between the cooling blocks. The end-tab is then cured
according to the manufacturer’s recommended cure cycle. The end-tabbed prepreg was
then clamped in the prestress rig and the prepreg laminate was then cured in an autoclave.
Uniform Pressure
Cartridge Heaters Aluminium End tab
30 mm
Water Cooling Heat insulating
plastic Aluminium Block
unit
350 mm
Figure 5-10: Schematic illustration of End-tab jig design.
Figure 5-11 shows a cross-section of the end-tab cured using the end-tab jig. From Figure
5-11 it can be observed that the fibre misalignment was eliminated near the end tab region.
The surface profile of composite panels was measured using a dial gauge for composites
manufactured with and without using the end-tab jig. The results from the surface profile
measurements are shown in Figure 5-12. In Figure 5-12 the prestressed composites with
hot-press cured end-tabs are coded as HPCET and end-tab jig cured end-tab are coded as
ETJCET. The non-prestressed composite in Figure 5-12 refers to the composite cured out-
of-prestress rig and without any end-tab. The measured surface profile of non-prestressed
composites shows the variation in thickness of the composite through the length. From
Figure 5-12 it is apparent that composites prepared using the end-tab jig show a
significantly reduced bending of the composite.
100
Figure 5-11: Microscopic evidence of improved fibre alignment near end-

tab region.
2
Non-Prestressed Composite
1.8 HPCET Prestressed Composite
1.6 ETJCET Prestressed Composite
1.4
Surface Profile (mm)
1.2
0.8
0.6
0.4
0.2
0
0 50 100 150 200 250 300
Length (mm)
Figure 5-12: Surface profile of non-prestressed and prestressed unidirectional

composites.
101
5.3 Results and discussions on sensor evaluation
The optical fibre sensors and load cell were used during processing of composites to
monitor and measure the residual strain development and applied pretension respectively.
Because the composites are processed in an autoclave at elevated temperature (120°C) with
air pressure and vacuum, the effect of these process parameters on the sensors needed to
be investigated. This section discusses the preliminary results from the effect of autoclave
process parameters on sensors. The results from the sensor behaviour during temperature,
pressure, vacuum and cure cycles are discussed in the following sections.
5.3.1. Effect of temperature on sensors
Optical fibre sensors

The effect of temperature on optical fibre sensors was studied according to the
experimental procedure given in Section 4.5.1 of Chapter 4. Figure 5-13 shows the
response of embedded and reference sensors during autoclave temperature tests. With
reference to Figure 5-13 it can be observed that the temperature measured from the
embedded FBG temperature sensor is in good agreement with the temperature measured
from the thermocouple in the autoclave.
120
Theoretical Prediction
900 Embedded EFPI Strain
100
Embedded FBG Strain
Reference EFPI
700 80
Reference FBG Strain
Temperature rise (oC)
Autoclave Temperature
Strain (µε)
60
500 Embedded FBG Temperature
40
300
20
100
0
-100 -20
0 1 2 3 4 5 6 7 8 9
Time (hours)
Figure 5-13: Relationship between optical sensors (embedded and

reference) and temperature.
102
From Figure 5-13 it can be observed that there is no significant change in strain measured
from the reference sensor during the temperature test. A small variation in measured strain
of about ± 25 µε and ± 50 µε in EFPI and FBG sensors were recorded respectively. This
variation in strain measured from the reference sensors could be due to (i) the influence of
the polyimide tape bonding-resin on the reference sensors, (ii) the transverse sensitivity of
the FBG sensor and (iii) the error in the measurement. This small variation is negligible and
therefore it can be concluded that the temperature does not influence the response of
optical fibre strain sensors.
With reference to Figure 5-13 it can be seen that the embedded strain sensors measures the
thermal expansion of the cured unidirectional composite. Theoretically, the thermal
expansion of a unidirectional composite (αC), in the fibre direction is given by the
following equation:
α mν m E m + α f ν f E f
αC =
E mν m + E f ν f
(K −1 ) (5.1)
where α, ν and E are the thermal expansion coefficient, Poisson’s ratio and modulus
respectively. The subscript m and f stand for matrix and fibre respectively.
Table 5-1: Material properties used.
Properties E-glass Epoxy Resin

Young's Modulus [MPa] 72500 4000
Poisson's ratio [-] 0.22 0.4
Thermal expansion coefficient [°C-1] 5.00×10-06 60.0×10-06
Volume fraction 0.6 0.4
Table 5-2: Summary of thermal expansion measured experimentally and

calculated theoretically.
Method Thermal expansion (×10-06 K-1)

Theoretical prediction (equation 5.1) 6.38
Experimental measurement (EFPI and FBG) 6.45
Equation 5.1 is based on a micro-mechanics theory of composites. The material properties

from Table 5-1 are used to predict the thermal expansion of the composite (using equation
5.1) for the temperature rise shown in Figure 5-13. The theoretical prediction of thermal
expansion (strain) of the composite is included in Figure 5-13. A summary of the thermal
expansion predicted from equation 5.1 and measured from experiments is presented in
Table 5-2. It can be observed that the theoretically-predicted thermal expansion of the
103
composite during the heating and isothermal cycles is in good agreement with the
measured EFPI and FBG strain sensors. Also from Table 5-2 it can be seen that the
experimental measurement of thermal expansion of composite is in good agreement with
theoretical prediction. During the heating and cooling cycle, perturbations are observed in
the strain measured from both the EFPI and FBG strain sensors. The perturbation
observed in the EFPI sensor response may be due to the time-delay for the optical fibres
inside the capillary to respond to the thermal strain induced by the composite to the sensor.
However, the measured strain from optical fibre sensors was within ± 50 µε of the
theoretical prediction.
5.3.2. Effect of pressure on sensors
a) Load cell
The effect of pressure on the load cell was studied. From the pressure test it was found
that a change of 0.04 kN was recorded for an increase in pressure of 0.69 MPa. During this
test the load cell was positioned at the end of the load screw, but the load screw was not
touching the load cell. This shows that there is no significant effect of pressure on the load
cell.
b) Optical fibre sensors

The effect of pressure on optical fibre sensors was studied according to the experimental
procedure given in Section 4.5.1, Chapter 4. Figure 5-14 shows the effect of pressure on
the reference optical fibre EFPI and FBG sensors. From Figure 5-14 it can be observed
that there is no significant effect of pressure on the EFPI and FBG sensors. The recorded
strain of ± 20 µε in both the reference optical fibre sensors was within the accuracy of the
instrumentation used.
Figure 5-15 shows the response of embedded FBG temperature and strain sensors on the
effect of autoclave pressure. From figure it can be observed that with increase in pressure
there is a change in strain of about ± 20 µε was observed. This is very similar to the
reference optical fibre sensors. It can also be observed that with increase in pressure an
increase in temperature of about 4.8°C was observed. The increase in temperature may also
have influenced the change in strain measured in the embedded sensor. The temperature
measured from thermocouple and FBG temperature sensors are in good agreement. The
observed change in the measured strain from embedded and reference optical fibre sensors
on the effect of pressure are negligible.
104
400 1
Reference EFPI Strain
350 Reference FBG Strain 0.9
300 Pressure 0.8
250 0.7
Pressure (MPa)
200 0.6
Strain (µε)
150 0.5
100 0.4
50 0.3
0 0.2
-50 0.1
-100 0
0 0.5 1 1.5 2 2.5 3 3.5
Time (hours)
Figure 5-14: Relationship between reference optical fibre sensors and

pressure.
200 1
Autoclave Temperature Embedded FBG Temperature
0.9
150 Embedded FBG Strain Pressure
0.8
Strain (µε)
0.7
100
0.6 Pressure (MPa)
Temperature rise ( C)
50 0.5
o
0.4
0
0.3
0.2
-50
0.1
-100 0
0 0.5 1 1.5 2 2.5 3 3.5
Time (hours)
Figure 5-15: Relationship between embedded optical fibre sensors and

pressure.
105
5.3.3. Effect of vacuum on sensors
a) Load cell
Figure 5-16 presents the relationship between the load measured from the load cell and the
applied vacuum. From Figure 5-16 it can be observed that a linear relationship exists
between the load and the vacuum.
In this vacuum test the sample was not clamped to the prestress rig. The load cell was
positioned such that the load screw was just in contact with the sensor face of the load cell.
Initially no load was applied to the load cell. When vacuum was applied, a uniformly
distributed compressive load was induced to the vacuum bagged flat-bed rig. The load cell
through the contact of the load screw recorded the compressive load applied by the
vacuum.
0.06
0.04
0.02
Load (kN)
-0.02
R2 = 0.9778
-0.04
-0.06
0 -100 -200 -300 -400 -500 -600 -700 -800 -900
Vacuum (millibars)
Figure 5-16: Relationship between load measured from the load cell and
vacuum.

Figure 5-17 presents the observed relationship between the optical fibre sensors and the
vacuum. From Figure 5-17 it can be observed that the optical fibre sensor experienced a
compressive strain of about 80 µε for an applied vacuum of 826 millibars (-0.084 MPa).
106
When the vacuum is removed not all the compressive strain induced in the optical fibre
sensor during the application of the vacuum is recovered. This is because prior to the
solidification of the resin, it is very difficult to control the relative orientation of the FBG
sensor. This could be the reason why the FBG sensor did not return to its initial value (at
the start of the vacuum test). No repetitions were carried out on this test. Only FBG
sensors were used. Therefore further investigations are required to confirm the
reproducibility of the results.
10 100
0 0
-10 -100
-20 -200
Vacuum (millibars)
-30 -300
Strain (µε)
-40 -400
-50 -500
-60 -600
-70 -700
Embedded FBG strain
-80 -800
Vacuum
-90 -900
0 20 40 60 80 100 120 140
Time (minutes)
Figure 5-17: Relationship between embedded sensors and vacuum.
5.3.4. Effect of cure cycle on sensors
a) Load cell
The effect of cure cycle on the load cell was studied according to the procedure described
in Section 4.5.3, Chapter 4. From the cure cycle test it was found that a change in load of
about 0.4 kN was observed. This shows that the cure cycle does not have a significant
influence on the load recorded from the load cell.
107

Figure 5-18 shows the relationship between the cure cycle parameters on the reference
EFPI and FBG sensors. From Figure 5-18 it can be observed that a variation in strain of
about ± 25 µε and ± 50 µε was recorded in EFPI and FBG sensors respectively. This
variation was also observed in the temperature and pressure tests. This could be due to the
effect of the polyimide bonding tape on the sensors during the cure cycle. In the embedded
sensor the thermal expansion of the composite was measured to be 7.30 ×10-6 K-1. This
measured thermal expansion is 1 ×10-6 K-1 greater than the thermal expansion measured
from embedded sensors during temperature test. This increase in thermal expansion could
be the effect of pressure and vacuum combined with temperature.
200 0.8
Temperature
0.7
Reference EFPI Strain
150 Reference FBG Strain 0.6
Temperature (oC)
Pressure
Vacuum (millibars) = -830 0.5
Pressure (MPa)
100
0.4
0.3
50
0.2
Strain (µε)
0.1
0
-50 -0.1
0 2 4 6 8 10
Time (hours)
Figure 5-18: Relationship between reference EFPI sensor and the cure cycle
parameters.
The evaluation of the prestress methodology was carried out in three stages. Firstly, the
load cell calibration (at different temperatures) and strain distribution in a steel plate during
an application of tensile load was investigated. Secondly, the end-tab jig design to
circumvent fibre misalignment near the end-tab jig and resin pre-cure away from the end-
tab region was presented. Finally, the effect of the autoclave process parameters on sensors
108
used in this work (load cell and optical fibre sensors) was studied. From the above study
the following conclusions can be made.
Load cell calibration

From the static compression calibration test it was shown that the load recorded from the
load cell is in good agreement with the applied machine load. A linear relationship was
observed from the load cell during the static tests conducted at different temperatures
(27°C, 70°C, 99°C and 121°C). A small change in load of about -0.8 kN for a 94°C
temperature difference was found.
Prestress rig evaluation:

From the static tensile prestress rig evaluation test, it can be concluded that the strain
distribution in the steel plate is uniform with a variation of 4% strain through the width of
the panel. This variation in the strain could be due to (i) the Poisson’s effect, (ii) any
misalignment in the ERSGs during bonding and (iii) measurement errors. A linear
relationship was found between the strain measured from the steel plate and the applied
load using the load screw in a prestress rig. This is because the load applied is within the
elastic region of the steel plate. Furthermore, this shows that this prestress rig system
responds linearly.
Prepreg end-tab jig design and evaluation:

Evidence of fibre misalignment near the end-tab region was found to be the cause of
composite bending. It was shown that as a result of fibre misalignment near the end-tab
region, the compressive load was induced eccentrically to the composite during prestress
release. This results in composite bending.
A flat-bed end-tab jig was designed to circumvent the fibre misalignment near the end-tab
region. Using this end-tab jig it was shown that fibre misalignment near the end-tab region
was eliminated.
Results from the surface profile of the composite shows that the surface profile of the
prestressed composite manufactured with the aid of an end-tab jig is in agreement with the
surface profile of the non-prestressed composite.
Sensor evaluation:
From the preliminary investigation of the effect of cure cycle parameters on the optical
fibre sensors, the following conclusions can be made.
109
Effect of temperature:
No significant change in the measured strain from the reference optical fibre sensors
during the temperature test. Therefore it can be concluded that the response of the optical
fibre strain sensors are not influenced by temperature.
Embedded optical fibre sensors measured the thermal expansion of the composite. The
measured thermal expansions from optical fibre sensors are in good agreement with the
theoretically-predicted thermal expansion from the micro-mechanical model.
The temperature measured from the embedded FBG temperature sensor is in good
agreement with the temperature measured from the thermocouple in the autoclave.
Effect of pressure:
During the pressure test a small change of about 0.04 kN was measured from the load cell.
This shows that there is no significant effect of pressure on the load cell.
It was found that there is no significant change in the measured strain from the reference
and embedded optical fibre sensors during the pressure test. This shows that the responses
of the optical fibre sensors are not influenced by the pressure.
Effect of vacuum:
The load cell showed a linear relationship with the application of a vacuum. An increase of
about 0.09 kN load in compression was measured for an increase in vacuum to -830
milibars (-0.084 MPa). This is because the uniform distribution of compressive load applied
by application of vacuum to the prestress rig was measured through the contact of the load
screw to the load cell.
An embedded optical fibre sensor in the prepreg recorded a compressive strain of about 80
µε for -830 milibars (-0.084 MPa) vacuum. However, when the vacuum was released not all
the compressive strain in the optical fibre was removed. This is because prior to the
solidification of the resin it is difficult to control the relative orientation of the FBG sensor.
Effect of cure cycle:
The load cell showed an increase of 0.4 kN during the cure cycle. This variation in load is
due to the combined effect of temperature, pressure and vacuum.
No significant variation was observed in the strain measured from the reference optical
fibre sensor throughout the cure cycle. A small variation of about ± 25 µε from EFPI and
± 50 µεfrom FBG was recorded. This could be due to the effect of the bonding tape and
optical fibre sensor during the cure cycle.
110
CHAPTER 6
Residual strain and fibre

alignment in the composites
OVERVIEW
This chapter presents the residual strain and fibre alignment results on the effect
of prestress.
The first part of the chapter presents the residual strain development in non-
prestressed and prestressed composites. The development of residual strain
during autoclave processing of a non-prestressed composite (unidirectional and
cross-ply) is discussed in four stages.
This is followed by the discussion of the residual strain results on the effect of
four different prestress levels in unidirectional composites. The results from
the strain released to the composite recorded from optical fibre sensors and
ERSG are discussed and compared.
The second part of this chapter investigates the effect of fibre prestress on the
alignment of fibres in the composite. The measured fibre alignment and
micrographs are presented for all the prestressed composites and discussed.
Finally, the conclusions were drawn from the residual strain and fibre
alignment results.
CHAPTER 6 Residual strain and fibre alignment in the composites
6.1 Residual strain development in non-prestressed composite
6.1.1. Unidirectional composite

Figure 6-1 presents the residual strain development measured using embedded EFPI
sensor in [0º]16 laminate during the autoclave curing process. Table 6-1 summarises the
residual strain measured using the embedded EFPI sensors in unidirectional non-
prestressed composites at different stages of the cure cycle. The residual strain
measurement using EFPI sensors during the autoclave curing process could be considered
in four stages: a) before curing, b) heating cycle, c) isothermal and d) cooling cycle.
Prior to embedding the EFPI sensors into the prepreg, the strain recorded was zero (start
of the cure cycle in Figure 6-1). After embedding the EFPI sensor in the prepreg laminate,
a compressive residual strain was measured as shown in the Figure 6-1 and Table 2-1.
During embedding sensors, rollers were used to expel the air between the laminates. The
back and forth movement of the roller during the prepreg laying processes could have
caused the compressive strains in the EFPI sensor. From Table 6-1 it can also be noticed
that the compressive strain recorded from the EFPI sensor after embedding changes at
different times in the same prepreg. This is because of the manual lay-up and rolling
process. During the heating cycle, the viscosity of the resin initially decreases. The
compressive strain in the sensor decreases and above 100ºC the sensor is subjected to a
tensile strain. This suggests that the reduction in the resin viscosity could have released the
compressive strain induced in the sensor during the embedding process. During the heating
cycle, the resin and fibre also expands. The tensile strain recorded from the sensors could
be due to the friction between the sensor and the constituent materials in the laminate.
Table 6-1: Presents the EFPI strain measured at different stages of the cure
cycle in non-prestressed composites.
Before After
Reference at 120oC Final residual strain at RT
embedding embedding
panels (µε) (µε)
(µε) (µε)
UR1 0 -71 40 -587
UR8 0 -63 42 -640
UR9 0 -101 24 -586
In the isothermal stage, the exothermic reaction of the resin begins, and this increases the
resin viscosity. The exothermic reaction increases the temperature of the composite from
120ºC (cure temperature) to 127ºC. Figure 6-2 shows the result from differential scanning
calorimetry for the 913 epoxy resin system. From Figure 6-2 it can be seen that when the
resin temperature reaches 120°C an exothermic peak is observed. In this stage, the resin
112
reacts to form a polymeric three dimensional cross-linked network structure. These cross-
linking reactions induce shrinkage of the resin, which induces a compressive residual strain
as shown in Figure 6-1. This was also observed by many researchers12,13,36. This is called
cure-induced shrinkage.
200 140
EFPI Embedded
120
0
Temperature
Residual Strain ( µε )
100
-200
Temperature ( oC )
Residual Strain = -640 µε 80
-400
60
a b c d
-600
40
-800
20
-1000 0
0 1 2 3 4 5 20
6
Time (hours)
Figure 6-1: Residual strain measured using embedded EFPI sensor in non-
prestressed [0]16 laminate during the autoclave curing process.
a) before curing, b) heating cycle, c) isothermal and d) cooling
cycle.
5 140
Heat flow
Temp
0 120
Heat Flow Endo Up (mW)
-5 100
Temperature ( C)
o
-10 80
-15 60
-20 40
-25 20
-30 0
0 20 40 60 80 100 120 140 160
Time (minutes)
Figure 6-2: Differential Scanning Calorimetry (DSC) results for the 913 epoxy
resin.
113
Once the temperature in the composite is reduced from the exotherm to the cure
temperature, the residual strain stabilises (see Figure 6-1). At the end of the isothermal
stage, the cooling cycle of the process commences. From Figure 6-1 it can be observed that
the compressive residual strain measured by the EFPI sensors increases. Because of the
mismatch in thermal expansion between the fibre and matrix, as the matrix shrinks, the
fibres in the matrix are compressed as illustrated in Figure 6-3. The arrow marks show the
direction of the residual stress. Because of the presence of rigid fibres, the matrix is
restrained from shrinking, which induces tensile strains in the matrix. This is called
thermally-induced residual strain. It can be seen that the thermally-induced residual strain is
significantly higher than the cure-induced residual strain. Previous researchers,36,37 have also
reported this phenomena.
Fibre
Matrix
Figure 6-3: Illustration of residual stress development in a unidirectional

composite.
When the temperature cooled down to 55°C, a slight decrease in the measured
compressive strain was observed, this coincides with the decrease in pressure (see Figure
6-4). Once the pressure reaches zero the residual strain stabilises. The final average residual
strain measured from three [0]16 non-prestressed composite was –604 µε at room
temperature (RT).
The results from EFPI sensors show that the reinforcing fibres in the composite are
subjected to a compressive strain during the curing process. In Chapter 7 the experimental
results from optical fibre sensors are compared with theoretical models from the literature
(for predicting residual strain in composites).
114
300 0.8
0.7
0.6
Residual Strain ( µε )
0
0.5
EFPI Embedded
Pressure (MPa)
0.4
Pressure
-300 0.3
Residual Strain = -640 µε
a b c d 0.2
0.1
-600
0
-0.1
-900 -0.2
0 1 2 3 4 5 20
6
Time (hours)
Figure 6-4: Residual strain and autoclave pressure recorded during the cure
cycle.
200 180
0 Residual strain 160

Temperature
-200
140
-400
Residual strain (µε)
120
Temperature( C)
o
-600
100
-800
a b c d 80
-1000 Final strain = -1240 µε
60
-1200
40
-1400
-1600 20
-1800 0
0 1 2 3 4 5 6 207
Time (hours)
Figure 6-5: Residual strain development in cross-ply composite. a) before

curing, b) heating cycle, c) isothermal and d) cooling cycle.
115
6.1.2. Cross-ply composite

The residual strain development was studied in a cross-ply laminate to compare the
magnitude with the unidirectional laminates. Figure 6-5 presents the residual strain from
the EFPI sensors for a [0/90/0/90/90/0/90/0]4S glass/epoxy with the corresponding
processing temperature. This is a thick laminate of 64-ply thickness (8 mm). The shape of
the residual strain development curve for the cross-ply composite is very similar to the
unidirectional composite presented in Figure 6-1. However, the magnitude of the exotherm
temperature (stage c) and the residual strain development recorded in the cross-ply
composite is twice the unidirectional composite. This is expected because it is a cross-ply
laminate and a thick composite system. Many researchers also observed this difference in
residual strain between unidirectional and cross-ply composites. Because the 90º ply have a
much greater stiffness than the resin matrix, the 0º ply contraction during cooling cycle will
be restrained by the 90º ply and vise versa.
It can be observed that when the temperature reached 120°C (which is the cure
temperature of the resin) there was a sudden increase in the mould temperature to 154°C,
due to the exothermic reaction of the resin. This increase of 34°C in composite
temperature is possible because of the 64-ply thick composite. This cross-linking reaction
of the epoxy resin induces compressive strain to the composite, which is due to the cure-
induced shrinkage. With reference to Figure 6-5, it can be seen that a compressive strain of
about 400 µε was recorded. Jinno et al. and Kang et al. have shown that this compressive
strain is due to cure-induced shrinkage of the epoxy resin and they have shown that this
could be related to the degree of cure of the resin. However, looking at Figure 6-5 again, it
can be seen that this is not only due to cure-induced shrinkage but also due to thermally-
induced shrinkage. Because the temperature after reaching the exotherm peak cools down
to the cure temperature (the temperature difference is 34°C), which induces a thermal
shrinkage of the fibre and matrix (which may also be bonded to each other due to the start
of the curing reaction). Therefore, it will be difficult to separate the cure-induced and
thermally-induced residual strain in the composite.
During the cooling cycle (stage d), a compressive strain was recorded from the EFPI
sensors. This compressive strain is due to 1) mismatch in the thermal expansion between
fibre and matrix at a micro-scale and 2) mismatch in the thermal expansion between the
adjacent plies of different orientations (0° and 90°) in the ply-scale. The final residual strain
measured in the cross-ply non-prestressed composite was –1240 µε at room temperature.
The residual strain was also measured in a [0/90/0/90/90/0/90/0]S laminate. The residual
strain development was similar to the [0/90/0/90/90/0/90/0]4S laminate shown in Figure
6-5. The final measured residual strain for the [0/90/0/90/90/0/90/0]S laminate was –
1277 µε at room temperature.
116
6.2 Residual strain development in prestressed composite
This study concentrates on unidirectional prestressed composites. The fibre prestressing of

cross-ply laminates was not studied in this work. In the unidirectional [0º]16 prestressed
composite, residual strains were measured using optical fibre (EFPI and FBG) sensors.
After composite curing and cool down to room temperature, ERSGs were bonded onto
the surface of the composite, in the same place as optical fibre sensors (as shown in Figure
4-5). The magnitude of the strain released to the composite when it was removed from the
prestress rig was recorded and quantified by using optical fibre sensors and surface-
mounted ERSG sensors. In this study composites were manufactured with four different
prestress levels. The residual strain and strain release measurements were carried out at all
the prestress levels and are presented in Table 6-2 and Table 6-3 respectively. The FBG
sensor was used (only in two panels with 108 MPa prestress) to compare the results with
EFPI sensors. Figure 6-6 shows the recorded pre-load from the load cell, residual strain
from EFPI sensors and strain release from ERSG sensors during the processing of
unidirectional composite prestressed to 150 MPa.
Table 6-2: Residual strain measured from four different prestress levels of the
unidirectional composites at room temperature prior to prestress
release.
Residual strain
Panel Code EFPI FBG EFPI 40 mm from
Middle Middle the edge
UPT8 -566
UPT11 -581
UP1_7kN -586
UP2_7kN -574
UP2_14kN -564 -643
UP3_14kN -503 -653
UP4_14kN -573
UP5_14kN -553
UP6_14kN -520
UP7_14kN -605
UP9_14kN -529 -742
UP12_14kN -714
UP1_24 kN -592
117
Table 6-3: Strain release recorded from EFPI, FBG and ERSG sensors during
prestress release at room temperature. * Rig thermal expansion
induced prestress calculated from the classical mechanics.
Prestress Strain released (µε)

released
Panel code EFPI FBG EFPI 40 mm ERSG ERSG 40 mm
(MPa)
middle middle from the edge middle from the edge
UPT8 51.21* - -109 -160
UPT11 51.21* -108 -194 -221
UP1_7kN 79.95 -184
UP2_7kN 80.36 -190 -223
UP2_14kN 108.50 -630 -656 -696
UP3_14kN 107.22 -637 -685 -602 -763
UP4_14kN 108.62 -579 -628
UP5_14kN 108.41 -537 -548
UP6_14kN 112.89 -552 -613
UP7_14kN 108.45 -600 -639
UP9_14kN 113.57 -551 -684 -677
UP12_14kN 107.61 - -608 -600
UP1_24 kN 150.07 -944 -958 -1113
600
Residual strain
400 120
Strain gauge
Prestress
200 Temperature releasing
Temperature ( C)
Pre-load 100
0 o
-200
Fibre 80
prestressing
-400
-600 60
-800
40
Pre-load (kN)
-1000
-1200
20
-1400
-1600 0
0 5 10 15 20 25
Time (Hours)
Figure 6-6: Residual strain development throughout the processing of 150

MPa prestressed composite.
118
From Figure 6-6 it can be seen that the recorded strain from EFPI sensors during the
application of pre-load is small because the EFPI sensors are not bonded to the prepreg. It
can be observed from Figure 6-6 that the applied pre-load was maintained throughout the
curing process to within 1 kN. During the heating cycle, the load cell did not record the
prestress induced in the fibres due to thermal expansion of the rig. This is because in the
current set-up there is no frame of reference for the load cell to record the thermally-
induced prestress to the fibres. When there is a temperature rise the whole prestress rig set-
up expands, this will induce prestress to the glass fibres. However, the load applied to the
glass fibres due to thermal expansion of the rig will not be recorded by the load cell.
The residual strain development curve and its magnitude in unidirectional prestressed
composites (see Figure 6-6 and Table 6-2) are very similar to the non-prestressed
composites as shown in Figure 6-1 and Table 6-1. After curing and cool-down to room
temperature the prestress is released. On the release of the prestress, a compressive strain is
induced into the composite. Figure 6-7 shows the strain release recorded from the EFPI,
FBG and ERSG sensors when a prestress of 108 MPa was released (Panel code
UP9_14kN). The pre-load was released in steps.
100
EFPI Embedded
0
FBG Embedded
-100 ERSG Surface-mounted
Strain released (µε)
-200
-300
-400
-500
-600
-700
-800
0 10 20 30 40 50 60
Time (minutes)
Figure 6-7: The strain release measured from optical fibre sensors and ERSG.
This was recorded during 108 MPa prestress release.
From the Figure 6-7 it can be seen that initially the strain measured from EFPI, FBG and
ERSG are in agreement with each other. However, as the magnitude of the pre-load release
increases the EFPI sensor records slightly less compressive strain when compared to FBG
and ERSG sensors. On the other hand the FBG and ERSG sensors are in very good
119
agreement. Also from Table 6-2 it could be observed that the compressive residual strains
recorded during the autoclave curing cycle from FBG sensors are slightly higher than the
EFPI results. This difference in the strain recorded from EFPI and FBG sensors could be
due to one or more of the following reasons.
a) The difference in configuration of the optical fibre sensors (as shown in Figure 6-8)
could have an effect on the measured residual strain.
b) Difficult in accurately measuring the gauge length of EFPI sensors. The change in
measurement of the gauge length in an EFPI sensor will lead to a change in the
strain measured. In order for the effective gauge length to be equal to L as shown
in Figure 6-8, there would be need to satisfy the following. a) perfect bond at fusion
points, b) void-free fusion points (the presence of voids could transport foregin
material into the cavity) and c) friction-free surfaces between fibre and capillary. If
all the above conditions are not satisfied, the effective gauge length will change, and
this will lead to a change in the strain measured by the EFPI sensor.
L Fusion Point
(a)
PMRI
(b)
Figure 6-8: Sensor configuration, a) EFPI strain sensor and b) FBG strain
sensor. PMRI – periodic modulation of the refractive index of the
optical fibre core.
Also from Figure 6-7 it could be observed that the data recorded from the FBG sensor did
not follow the stepwise pre-load release similar to EFPI and ERSG sensors. This is because
the rate of data acquisition in FBG sensors was very low (2 measurements per minute)
when compared to the EFPI and ERSG sensor data acquisition rate (1 measurement per
second). Table 6-4 presents the average strain release measured from EFPI, FBG and
ERSG sensors in 108 MPa prestressed composite. From Table 6-4 it can be observed that
the average strain release measured from FBG is slightly higher than from the EFPI sensor.
However, by considering the standard deviation and the number of samples it can be seen
that the strain recorded from the optical fibre (EFPI and FBG) and ERSG sensors are all
in good agreement.
120
Table 6-4: The average strain release recorded using EFPI, FBG and ERSG
sensors in a 108 MPa prestressed composite.
Sensors Number of Average strain Standard deviation

samples release (µε) (µε)
EFPI 6 -592.8 56
FBG 2 -646.0 54
ERSG 7 -616.2 35
After releasing the prestress, the residual strain of the composite changes. Therefore, the
final residual strain εfinal in the fibre is expressed as:
ε final = ε r − ε p (6.1)
where, εr is the residual strain measured before releasing (that is cure-induced and
thermally-induced residual stresses) and εp is the strain measured during pre-load release.
Figure 6-9 shows the final residual compressive strain in the fibres calculated for all the
prestressed composites. From this figure it can be seen that the compressive strain in the
fibre reduces as the applied prestress increases. It can also be observed that with the
application of a fibre prestress of about 108 MPa, fibres with approximately zero final
residual strain can be achieved. Above 108 MPa prestress, the final residual strain becomes
tensile. This demonstrates that fibre prestressing enables the control of final residual strain
in the composite.
600
EFPI Middle
400 EFPI 40 mm from edge

ERSG Middle
200 ERSG 40 mm from edge

FBG Middle
-200
-400
-600
-800
-20 0 20 40 60 80 100 120 140 160
Prestress (MPa)
Figure 6-9: The final residual strain in the composite for different prestress
levels.
121
From Figure 6-9 and Table 6-3 it can be observed that the strain gradient across the panel
(middle and 40 mm from edge) increases with an increase in prestress. This variation in
strain reaches a maximum of 14% at 150 MPa prestress. We conjecture that this difference
in the measured strain release is due to an edge effect caused by the variation in Poisson’s
contraction at an unconstrained edge. When the pre-load is released, a compressive force is
induced in the matrix. This results in a transverse expansion of the composite. At the edges
uniaxial compressive stresses exist. However, in the centre, some transverse stress will exist
due to the transverse stiffness of the composite. These transverse compressive stresses
produce tensile longitudinal strains, which reduce the measured value of the strain released
in the centre. This is mathematically modelled and presented in Chapter 7.
6.3 Results of fibre alignment
It was shown in Chapter 2 that fibre waviness or misalignments in composites are caused
by many factors during the processing of composites. In manufacturing prestressed
composites, prior to composite curing, the reinforcing fibres are mechanically prestressed.
This firstly straightens the wavy fibres present in the prepreg and secondly avoids
movement of fibres during the composite curing process. As a result, the reinforcing fibres
would be more aligned to the direction of the applied load.
The fibre alignment in all the prestressed composites were measured according to the
procedure described in Section 4.8.1 (iii), Chapter 4. Figure 6-10 shows the fibre
orientation angle measured in all the composites. From Figure 6-10 it can be observed that
in non-prestressed composite the fibres are aligned between 0° to 6°. Yurgartis117 also
reported a similar degree of fibre misalignment in prepreg systems. A representative
micrograph of a non-prestressed composite, taken from the transverse direction is shown
in Figure 6-11.
From Figure 6-10 it can also be observed that with an increase in prestress the percentage
of fibres accurately aligned increases. Figure 6-12 to Figure 6-15 show representative
micrographs taken from the transverse direction of prestressed composites. From Figure
6-10 and Figure 6-12 to Figure 6-15 it can be clearly seen that fibre prestressing improved
the fibre alignment in composites. It can also be observed that the 0° fibre alignment
reaches a maximum of about 75% in a composite with a prestress of 108 MPa, above
which no further improvement was observed. This is because of the original fibre
alignment in the prepreg. A certain percentage of fibres are so misaligned that no matter
what prestress is applied they will not be oriented parallel to the loading direction. For this
composite, the results show that 25% of fibres fall into this category.
122
90
0 MPa
80 51 MPa
80 MPa
70 108
150 MPa
60
Counts (%)
50
40
30
20
10
0
0 1 2 3 4 5
0
Fibre orientation angle ( )
Figure 6-10: The measured fibre orientation in a composite as a function of

applied prestress.
Figure 6-11: A representative micrograph of a transverse section of a non-

prestressed (0 MPa) composite.
123
Figure 6-12: A representative micrograph of a transverse section of the 51


124


125
The present study was aimed at an investigation of the effect of fibre prestress on residual
strain and fibre alignment in glass/epoxy composites. From the results presented the
following conclusions can be made.
Residual strain
Residual strain development in two different lay-ups (unidirectional and cross-ply) of glass
fibre/epoxy composite panels was measured using optical fibre EFPI sensors. From the
residual strain data, it was shown that two types of residual strains are developed (i) due to
shrinkage of the resin during cure and (ii) due to the thermal mismatch between fibre and
matrix. The residual strain development curve was similar for both the laminate types.
However, the magnitude of the measured residual strain for a cross-ply composite panel
was twice the residual strain of a unidirectional composite. This is because the 90º plies
have a much greater stiffness than the resin matrix.
Results from the residual strain measurement using the EFPI sensors in non-prestressed
and prestressed unidirectional [0º16] composite panels show that fibre prestressing
methodology can be used to reduce or control the residual strain development in
composites. For example, by applying about 108 MPa prestress, a composite panel with
approximately zero final residual strain can be prepared.
Experimentally, it was also shown that the measured strain release from EFPI, FBG and
ERSG sensors were in good agreement with each other. The difference in strain measured
through the width of the composite is conjectured that to be due to Poisson’s ratio and
edge effects and this is discussed in Chapter 7. It was also shown that the strain release
measured from embedded EFPI and FBG sensors in the composite are in good agreement.
Fibre alignment
From the experimental results it was shown that fibre prestressing improved the alignment
of fibres in composites. This improvement in fibre alignment resulted in an increase from
20% for non-prestressed to 75% of fibres oriented in the 0° direction for a 108 MPa
prestress. Above this prestress level no further improvement in the fibre alignment was
observed.
126
CHAPTER 7
Theoretical analysis of
prestressed composites
OVERVIEW
This chapter concentrates on (i) a theoretical analysis of the effect of fibre
prestress on residual stresses in composites and (ii) compares the theoretical
prediction with experimental results.
The theory predicts the final residual stress in fibre and matrix after releasing
the prestress. The theory is further extended to predict the strain distribution
through the width of the composite panel.
Experimental results from extrinsic Fabry-Pérot interferometric (EFPI) strain
sensors, Fibre Bragg Grating (FBG) strain sensors and electrical resistance
strain gauges (ERSG) are in good agreement with theoretical predictions.
A small variation of 1% in the strain distribution across the width of the
composite panel is predicted.
7.1 Introduction
P
RESTRESSED COMPOSITES ARE PREPARED by applying a pre-load to the reinforcing
fibres prior to cure. Since the uncured polymeric matrix possesses negligible
stiffness, essentially all the pre-load applied would be carried by the fibres. The
composite would then be cured, causing the polymeric matrix to gel and solidify. Following
cure and cool down the fibre pre-load is released, and the resulting elastic contraction of
CHAPTER 7 Theoretical analysis of prestressed composites
fibres relieves the tensile residual stress in the matrix resulting from the cure cycle by
inducing compressive stresses.
The theoretical prediction is described in three stages namely,
1. Application of pre-load to the fibres
2. Process-induced residual stress development during elevated temperature
3. Fibre pre-load release
The following model assumes that perfect bonding exists between fibre and matrix.
The above three steps of fibre prestressed composite preparation can be mathematically
described as follows
Step 1:
Pre-load is applied to the reinforcing fibres. This applied pre-load is maintained throughout
the curing process. Prior to curing, the stress in the fibre and matrix are:
σ f =σ f
prestress
W (7.1)
where σ f
prestress =
Af
σm =0 (7.2)
where σ f and σ m are the stresses in the fibre and matrix, σ fprestress is the applied prestress to
the fibres and W and Af are the pre-load applied to the fibre and cross-sectional area of the
fibre.
Step 2:
During composite processing, residual stresses are developed. This is due to resin cure-
induced volume changes and thermal expansion mismatch (between fibre and matrix)
induced volume changes. The stresses in the fibre and matrix after curing can be written as:
(σ )f
ac =σ f
prestress +σ f
R (7.3)
(σ ) m
ac = σ mR (7.4)
where (σf)ac and (σm)ac are the values of σf and σm after curing; and σfR and σmR are the
residual stresses in the fibre and matrix. The residual strain development in the fibre and
matrix was calculated using the one-dimensional Tsai and Hahn model and the three-
dimensional concentric cylinder Wagner model.
The Tsai and Hahn model assumptions were:
128
7.1 Introduction
• Both fibre and matrix are isotropic materials.

• The system is a unidirectional composite in a state of plane stress corresponding to
thin laminates, under a thermal strain.
• No mechanical stress is applied.
• There is perfect bonding between fibre and matrix.
• The modulus and thermal expansion coefficients of fibre and matrix are
independent of temperature.
Most of the Tsai and Hahn assumptions were considered in the Wagner model except that
the Wagner model considers that the fibre and matrix are transversely isotropic.
Transversely isotropic materials are a special class of orthotropic materials that have the
same properties in one plane (e.g. the x-y plane) and different properties in the direction
normal to this plane (e.g. the z-axis fibre direction). Both these models predict only the
thermally-induced residual stresses and not the cure-induced residual stresses.
The residual strain predicted from these models for E-glass/epoxy composite is presented
in section 7.4.1.
Step 3:
After composite curing and cool-down to room temperature the applied pretension in the
fibres is removed. This induces compressive stresses to the matrix. During pretension
release, not all the fibre prestress is released to the matrix. Some of the prestress remains in
the fibre as tensile residual stress (σf )P. Hence, the resulting stresses in the fibre and matrix
can be written as:
(σ ) f
apr =σ f
R + (σ f )P
(7.5)
(σ ) m
apr = σ m R + ∆σ (7.6)
σ f
prestress = σf( ) P
+ ∆σ (7.7)
where (σf)apr and (σm)apr are the values of σf and σm after pretension release; and ∆σ is the
stress released to the matrix.
The tensile residual stress in the fibre resulting from pretension release is calculated as
follows. In the following analysis the process-induced residual stresses will be omitted
initially for simplicity and added later (see equations 7.18 and 7.19).
The change in strain due to pretension release is expressed as ∆ε. The strain released in the
fibre (εf) and matrix (εm) will be equal.
129
εf =
(σ ) f
P
=
W
− ∆ε (7.8)
Ef Af E f
where Ef is the modulus of fibre. Equation 7.8 can be rewritten as:
(σ ) P
W
+ ∆ε =
f
(7.9)
Ef Af E f
(σ m )P
εm = = ∆ε (7.10)
Em
As there are no external forces applied to the composite, at equilibrium:
(σ ) φ + (σ )
f
P
f m
P
φm = 0 (7.11)
where φf and φm are the volume fraction of fibre and matrix present in the composite that is
φf
(σ m )P = −(σ f ) P
(7.12)
φm
Substituting equation 7.12 in 7.10 gives:
(σ m )P (σ ) P
φf
∆ε = =− ×
f
(7.13)
Em Em φm
Substituting equation 7.13 in 7.9 gives:
(σ ) f
P
−
(σ ) f
P
×
φf
=
W
(7.14)
Ef Em φm A f E f
⎛ 1 φf ⎞
(σ ) P
⎜
⎜E
− ⎟= W
⎟ (7.15)
⎝ f E mφ m ⎠ A f E f
f
⎛ E mφ m − E f φ f ⎞
(σ )
f
P
⎜
⎜ E E φ
⎟= W
⎟ A E (7.16)
⎝ f m m ⎠ f f
(σ )
f
P
=
W
×⎜
⎛ E mφ m ⎞
⎟ (7.17)
Ef A f E f ⎜⎝ E mφ m − E f φ f ⎟
⎠
From equation 7.17, the tensile residual strain in the fibre resulting from pretension release
can be calculated. Now adding together the process-induced residual strain and the strain
resulting from pretension release, the strain in the fibre and matrix after pretension release
become:
130
7.2 Composite strain
⎛ E mφ m ⎞
(ε )
f
apr =ε f
R +
W
×⎜
⎜
A f E f ⎝ E mφ m − E f φ f
⎟
⎟ (7.18)
⎠
(ε )m
apr = ε m R + ∆ε (7.19)
where εfR and εmR are the process-induced residual strain in the fibre and matrix
respectively.
The pretension is released in the longitudinal direction (direction 1 in Figure 7-1) of the
composite, so the resulting strain of the composite (εc1) in the longitudinal direction is
given by:
ε C 1 = ∆ε (7.20)
σ1
End-tab
Composite
e1 e2
σ1
Figure 7-1: Strain release to the composite.
The dimensions of the composite panels prepared in this study are 290 mm (l) × 200 mm
(w) × 2 mm (t) and so the panels can be considered to be thin and the stress system to be
plane stress. When pretension is released to such a system, there will be a variation in the
residual strain at various locations across the width of the composite due to Poisson’s
effect.
131
Consider the element e1 and e2 in Figure 7-1. When a prestress σ1 is released to the
composite, the elements e1 and e2 will undergo compression. Also due to the Poisson’s
effect there will be an expansion in the transverse direction (direction 2), shown as dotted
lines in Figure 7-1. Because of the free-edge, the element e2 can easily expand in the
transverse direction (it does not have any constraint against expansion). However, the
element e1 will be constrained in the transverse direction by the transverse stiffness of the
composite. As an approximation it is assumed that element e1 is prevented from expanding
in the transverse direction (that is ε2 = 0). The material adjacent to element e1 imposes a
transverse compressive stress, σ2, as shown in the Figure 7-2. This in turn will reduce the
strain released in the longitudinal direction.
σ1
σ2 σ2 ε1
1 ε2 = 0
1
2 σ1 2
Stress Strain
Figure 7-2: Prestress release in element e1.
Assuming that the composite is an orthotropic material the stress-strain relationship can be
written as:
⎛ 1 − ν 21 ⎞
⎜ 0 ⎟
⎛ ε 1 ⎞ ⎜ E1 E2 ⎟ ⎛σ1 ⎞
⎜ ⎟ ⎜ − ν 12 1 ⎟ ⎜ ⎟
⎜ ε2 ⎟ = ⎜ 0 ⎟ × ⎜σ 2 ⎟ (7.21)
⎜ γ ⎟ ⎜ E1 E2 ⎟ ⎜ ⎟
⎝ 12 ⎠ ⎜ 1 ⎟ ⎝τ 12 ⎠
⎜ 0 0
G12 ⎟⎠
⎝
where σ, τ are the direct and shear stresses. ε and γ are the direct and shear strains, E and
G are the direct and shear moduli. Subscripts 1 and 2 denote the longitudinal and
transverse directions respectively. ν12 and ν21 are the major and minor Poisson’s ratios of
the composite.
132
From equation 7.21, the strain in the longitudinal direction resulting from pretension
release can be expressed as:
σ1 ν 21
ε1 = − σ2 (7.22)
E1 E2
therefore the strain in the element e1 is given by:
ν 21
ε C 1 = ∆ε − σ2 (7.23)
E2
where σ2 is the stress induced in the lateral direction because of the Poisson’s effect (due to
the greater width of the specimen) and E2 is the transverse stiffness of the composite.
Calculation of σ2
Consider the element e1 at the centre. If the strip were infinitely wide then no transverse
strain would be possible. The element would be 100% laterally constrained (ε2 = 0).
From equation 7.21, the strain in the transverse direction is given by:
ν 12 1
ε2 = − σ1 + σ2 (7.24)
E1 E2
Since ε2 = 0, :
⎛E ⎞
σ 2 = ⎜⎜ 2 ⎟⎟ν 12σ 1 (7.25)
⎝ E1 ⎠
So, returning to equation 7.8 and 7.10, but applying it to the element in the centre (e1):
(σ ) P
νf W
εf = − σ2 = − ∆ε
f
(7.26)
Ef Ef Af E f
(σ m )P ν m
εm = − σ 2 = ∆ε (7.27)
Em Em
(σ ) P
νf W (σ ) ν P
− σ2 = − m + m σ2
f
(7.28)
Ef Ef Af E f Em Em
Also the strain in the matrix in the transverse direction is zero (εm = 0):
σ2 (σ m )P
−ν m =0 (7.29)
Em Em
σ 2 = ν m (σ m )P (7.30)
133
From equation 7.11 (σf)P can be expressed as:

φm
(σ )
f
P
= −(σ m )
P
φf
(7.31)
Substituting equation 7.30 and 7.31 in equation 7.28 we get:
⎛ 1 φ ⎛ ν f ν m ⎞⎞
(σ m )P ⎜⎜ − m −ν m ⎜
⎜
+ ⎟⎟ = W
⎟⎟ (7.32)
⎝ Em E f φ f ⎝ E f Em ⎠ ⎠ A f E f
⎛ Em E f φ f ⎞
(σ m )P =
W ⎜
⎜ E φ − E φ − ν φ (ν E − ν E
⎟ (7.33)
Af E f ⎝ f f m m m f f m m f )⎟⎠
The value of (σm)P calculated from equation 7.33 is substituted into equation 7.30 to
determine the transverse stress (σ2) induced to the element e1 due to lateral constraint.
The strain released to the free-edge of the composite panel (element e2) and centre of the
composite panel (element e1) can be calculated from equations 7.20 and 7.23 respectively.
7.3 Results and discussions
7.3.1. Residual strain

The residual strain predicted from the Tsai and Hahn one-dimensional model and the
Wagner three-dimensional model using the materials parameters from Table 7-1 are
presented along with the experimentally-measured residual strain (using EFPI and FBG
strain sensors) in Table 7-2.
From Table 7-2, it is apparent that the experimentally measured residual strain is higher
than the predicted residual strain from the Tsai and Hahn and Wagner models. The reason
for this difference between models and experimental results are as follows.
• The Tsai and Hahn and Wagner residual strain models include only the thermal
mismatch effects on residual strains, whereas the experimental results include both
thermal strains and cure-induced (resin shrinkage) residual strains. The magnitude
of the cure-induced residual strain development varies greatly between different
polymer matrix materials. In general, the higher the cross-link density, the higher
the strain development118. Researchers have shown that cure-induced residual strain
can range from 1% to as much as 30% of the total residual strain and may be even
higher.
• The interaction between the neighbouring fibres and the host fibre on the residual
strain measurement were not included in the models (because they are concentrated
on a micro-mechanical scale).
134
• The variation of fibre and matrix modulus and thermal expansion with temperature
has been ignored.
• Any variation in fibre volume fraction within the composite panel is ignored.
Table 7-1: Material parameters used
Properties E-glass Epoxy Resin

Young's Modulus [MPa] 72500 4000
Poisson's ratio [-] 0.22 0.4
Thermal expansion coefficient [°C-1] 5.00×10-06 60.0×10-06
Table 7-2: Residual strain results.
Models Residual strain in fibre (µε)

Tsai and Hahn -195
Wagner -280
Experimental EFPI results -600 ± 58
Experimental FBG results -737 ± 35
From Table 7-2, it can also be observed that there is a small difference in the residual strain
measured from EFPI and FBG strain sensors. The possible reasons for the differences in
residual strain measurements are:
a) The difficulty to measure accurately the gauge length of EFPI sensors. The change
in measurement of the gauge length of an EFPI sensor will lead to a change in the
strain measured. In order for the effective gauge length to be equal to A as shown
in the Figure 7-3, there would need to be perfect bond at fusion points and
friction-free surfaces between fibre and capillary. If both of these conditions are
not satisfied, the effective gauge length will be less, so the strain measured should
be greater.
b) The possibility of resin-wicking in to the EFPI capillary, which could reduce the
gauge length.
c) The configuration of optical sensors (as shown in Figure 7-3) could have an effect
on the measured residual strain.
135
Fusion Point
A
(a)
(b)
Figure 7-3: Sensor configuration, a) EFPI strain sensor and b) FBG strain
sensor.
7.3.2. Strain release

The prestress is released to the composite after curing and cool-down to room
temperature. The release of prestress induces compressive stresses to the polymer matrix.
The effect of prestress on the residual stress in fibre and matrix are shown in Figure 7-4.
The theoretical residual stress in the fibre and matrix was calculated from the Wagner
model. From Figure 7-4, it can be observed that with an increase in prestress, the tensile
residual stress in the matrix reduces and above 21 MPa prestress the matrix is subjected to
compressive stresses. This change in matrix residual stress with prestress could increase the
resistance to crack initiation and propagation during mechanical loading. On the other
hand, the compressive residual stress in the fibre reduces with prestress and above 21 MPa
prestress, the fibre is subjected to a tensile residual stress. The experimentally-measured
fibre residual strain from EFPI and FBG strain sensors is converted to stress and plotted
along with the theoretical prediction in Figure 7-4. From Figure 7-4, it can be seen that
there is a constant difference between the experimental results and the theoretical
predictions of fibre residual stress. This could be due to the cure-induced residual stress
build-up during resin shrinkage. The Wagner model predicts only the thermally-induced
residual stress.
The experimentally-measured fibre residual strain using EFPI and FBG strain sensor
includes both cure-induced and thermally-induced residual strains. When the
experimentally-measured fibre residual strain is included in the theoretical prediction at 0
MPa prestress, it can be observed that the experimental results are in agreement with the
theoretical prediction as shown in Figure 7-5. Also it can be observed that the zero-residual
stress in fibre and matrix occurs at 45 MPa prestress.
136
200
150
100
Residual Stress (MPa)
50 Fibre Stress
Matrix Stress
0
EFPI Results
-50 FBG Results
ERSG Results
-100
-150
-200
-250
0 20 40 60 80 100 120 140 160 180
Prestress (MPa)
Figure 7-4: Residual stress in the fibre and matrix with prestress. The
theoretical residual stress at 0 MPa prestress in fibre and matrix
was calculated from the Wagner model.
150
100
50
Residual Stress (MPa)
Fibre Stress
0 Matrix Stress
EFPI Results
FBG Results
-50
ERSG Results
-100
-150
-200
0 20 40 60 80 100 120 140 160 180
Prestress (MPa)
Figure 7-5: Residual stress in the fibre and matrix with prestress. The
experimentally-measured residual stress at 0 MPa prestress is
included in the theoretical prediction.
137
The small difference between the theoretical predictions and experimental results of the
residual stress in the fibre with prestress could be due to 1) a small misalignment in the
strain sensors (EFPI and FBG) during embedding and 2) error in the measurements.
The release of prestress to the composite, is very similar to applying a compressive stress to
the composite. Before releasing the prestress, ERSG sensors were surface-mounted on the
composite to compare the strain released from the optical sensors (EFPI and FBG) and
ERSG.
Due to the greater width of the composite there will be a small variation in the strain
distribution across the width of the composite panel due to Poisson’s effect. This is
explained in Setion 7.2. The strain induced in the free-edge and middle of the composite
width are calculated from equations 7.20 and 7.23 respectively. The calculated strain
released to the composite panel with applied prestress is plotted along with the measured
strain release from EFPI, FBG and ERSG sensors in Figure 7-6. The experimental results
are plotted with prestress measured before releasing.
In this research, prestress is applied to the fibres not only by mechnical load but also by the
thermal expansion of the prestress rig. Using classical mechanics, the flat-bed prestress rig
thermal-expansion-induced-prestress to the fibres was calculated to be 51 MPa (see Section
3.3.5, Chapter 3). However, the thermal-expansion-induced-prestress calculated above is
the stress applied to the fibres at 120°C (which is the curing temperature of the composite).
During the cool-down cycle, as the temperature decreases to room-temperature, the rig
thermal-expansion-induced-prestress is released. When the composite is released from the
prestress rig, a small strain release of about 110 µε is recorded. This corresponds to 9 MPa
prestress from the theoretical prediction as shown in Figure 7-6.
The predicted strain distribution across the width of the composite is plotted at three
different prestress levels (29, 60 and 100 MPa) as shown in Figure 7-7, Figure 7-8 and
Figure 7-9. It can be observed that there is a variation of about 1% strain release between
the free-edge to the centre of the composite panels at all prestress levels.
138
2000
Equation 7.20
Equation 7.23
1600 Thermal Strain
EFPI Results
Strain released (µε)
FBG Results
1200
ERSG Results
800
400
0
0 20 40 60 80 100 120 140
Prestress (MPa)
Figure 7-6: Strain released to the composite.
Prestress Released = 29 MPa
500
450-500
450 400-450
350-400
300-350
Strain Released (µε) 400
350
250
300
0 125
Length
100 (mm)
0
Width (mm) 200
Figure 7-7: Strain distribution across the width of the composite panel for 29
MPa prestress release.
139
900
850-900
850 800-850
750-800
700-750
750
250
700
0 125
Length
100 (mm)
0
Width (mm) 200
1400
1350-1400
1300-1350
1250-1300
1350 1200-1250
1250
250
1200
0 125
Length
100 (mm)
0
Width (mm) 200
140
The present theoretical analysis is aimed at predicting (i) the effect of prestress on the
residual stress in the fibre and matrix and (ii) the distribution of the strain released across
the width of the composite panel during prestress release. Experimental results from
residual strain measurements and strain release measured using strain sensors are compared
with the theoretical analysis. From the theoretical analysis and experimental results the
• Fibre prestress methodology can be used to minimise the process-induced residual
stress in fibre and matrix present in the composite.
• Experimental residual strain results measured using EFPI and FBG strain sensors
are greater than those predicted from the Tsai and Hahn and Wagner models. This
is because these models predicts only the thermal stresses and on a micro-
mechanical scale. On the other hand the experimental results measured using
optical fibre sensors include both the cure-induced and thermally-induced residual
stresses.
• The theoretical prediction of strain release to the composite is in good agreement
with the experimental results.
• During prestress release, due to Poisson’s effect, a small variation in strain
distribution of about 1% across the width of the composite panel is predicted.
141
CHAPTER 8
Static mechanical properties
OVERVIEW
This chapter discusses the results of the effect of fibre prestress on static tensile
and compression properties of composites.
The tensile properties such as strength, modulus, and strain at deviation from
linearity are discussed.
The effect of prestress on macroscopic and microscopic tensile damage is
discussed.
The compressive properties such as strength and modulus are discussed.
The influence of prestress on macroscopic and microscopic compressive
damage of composites is discussed.
The conclusions are drawn from the observed results for the effect of fibre
prestress on tensile and compressive performance of composites.
8.1 Static tensile results and discussion
A
SELECTION OF REPRESENTATIVE STRESS-STRAIN curves for all prestressed
unidirectional [0°]16 composite panels are shown in Figure 8-1. A summary of the
tensile test results is presented in Table 8-1. The individual sample tensile test
results for all the prestressed composites are presented in Appendix C.
8.1.1. Strain at deviation from linearity

With reference to Figure 8-1, a degree of non-linearity can be observed for all the
prestressed composite panels studied. The average strain at which the non-linearity was
CHAPTER 8 Static mechanical properties
observed is summarised in Table 8-1 and plotted in Figure 8-2. From Table 8-1 and Figure
8-2 it can be observed that the average εSDL increases with increase in fibre prestress up to
108 MPa prestress, above which there is an indication of a reduction. However, from the
statistical analysis student t-test it was found that there is no significant difference in εSDL
measured between 108 MPa and 150 MPa.
1600
1400
1200
1000
Stress (MPa)
800
0 MPa Prestress
600
51 MPa Prestress
400 80 MPa Prestress
108 MPa Prestress

200
150 MPa Prestress
0
0 0.5 1 1.5 2 2.5 3 3.5
Strain (%)
Figure 8-1: Stress-strain curve for all prestressed composite panels.
This deviation from linearity in the stress-strain curve during tensile loading may be
attributed to the following reasons: a) non-linear behaviour of the matrix and b) fibre
waviness.
(i) Non-linear behaviour of the matrix

It is possible to expect the non-linear behaviour of the epoxy matrix (because of its
viscoelastic behaviour) to influence the stress-strain curve of the unidirectional composite.
Goutianos et al.119 studied the linear and non-linear behaviour in single fibre carbon/epoxy
composites in tensile and compressive loadings. From their investigation, they showed that
during tensile loading, the deviation from linearity is due to the on-set of matrix plasticity
and the reduction in shear modulus.
This strain to non-linear behaviour of the matrix could be influenced by the process-
induced residual stresses. The possible reason for the observed improvement in εSDL (see
Table 8-1) could be attributed to a reduction in the tensile residual stress of the matrix with
fibre prestress. Motahhari and Cameron showed that the reduction in matrix and interface
144
residual stresses with prestressing enhances the mechanical properties. They suggested that
reducing matrix and interface residual stresses could prevent the opening of micro-cracks
in the matrix and debonding at the interface.
On the other hand, above an optimum prestress level residual stress at the interface
increases, which makes the interface vulnerable and causes fibre-matrix debonding. From
Table 8-1 it can be observed that above 108 MPa prestress, the average εSDL indicates a
reduction. This suggests that there is an optimum prestress level for the improvement in
εSDL. Previous researchers76,77,79 have shown that there is an optimum prestress level for the
improvement in flexural and impact behaviour of composites.
Table 8-1: Summary of tensile properties of composites studied in this

programme. The number in parentheses represents the
standard deviation.
Ultimate Secant Strain at Fibre Void

Failure
Prestress tensile modulus at deviation from volume volume
strain
(MPa) strength ε = 0.25% linearity εSDL fraction fraction
(%)
(MPa) (GPa) (%) (%) (%)
1311 43.37 3.18 1.49 58 0.34
0
(±80) (±2.64) (±0.10) (±0.05) (±1) (±0.09)
1244 41.69 3.05 1.73 58 1.13
51
(±110) (±3.44) (±0.15) (±0.06) (±1) (±0.40)
1295 43.69 3.11 1.78 58 1.13
80
(±65) (±2.48) (±0.10) (±0.04) (±1) (±0.37)
1285 43.20 3.10 1.92 58 0.93
108
(±138) (±2.62) (±0.17) (±0.15) (±2) (±0.18)
1273 42.60 3.08 1.80 59 0.78
150
(±66) (±1.52) (±0.13) (±0.05) (±2) (±0.32)
(ii) Fibre misalignment

Fibre prestressing prior to cure and maintaining this throughout the curing process could
reduce fibre misalignment (both in-plane and out-of-plane) and this could have also
contributed to the increase in εSDL. In non-prestressed composites, during processing, fibre
misalignment can be instigated due a number of reasons. For example: (i) fibre movement
during pressurisation and/or during the minimum viscosity phase of the resin matrix, (ii)
differential thermal expansion of various constituent materials, (iii) low fibre volume
fraction, (iv) excessive resin-bleeding inducing fibre movement and (v) over-pressurisation
during processing.
145
Fernando and Al-Khodairi studied the εSDL for the same material used in this study. They
reported that the average εSDL for E-glass/913 epoxy was 1.37 ± 0.08%. The average εSDL
reported by Fernando and Al-Khodairi is smaller when compared to the current work (1.49
± 0.05%). The possible reason could be different processing methods. In the current study
the composites are prepared using an autoclave process, which is accepted as a process for
manufacturing high quality composites and also mostly used by aerospace industries.
However, in Fernando and Al-Khodairi’s work, composites were prepared using the hot-
press moulding process. Fernando and Al-Khodairi reported that during the hot-press
moulding process it was not possible to control the leakage of matrix resin on a consistent
basis. It is generally recognised that the hot-press moulding process is prone to induce
more fibre waviness and voids in composites. This could have reduced the εSDL in
Fernando and Al-Khodairi’s work. Also no details were given regarding the extent of fibre
waviness in their composites to compare with the current work.
2.2
Strain at deviation from linearity (%)
1.8
1.6
1.4
1.2
1
-20 0 20 40 60 80 100 120 140 160
Prestress level (MPa)
Figure 8-2: The measured strain at the deviation from linearity in stress-strain
curve as a function of prestress. The error bars represents the
standard deviation.
The fibre alignment of non-prestressed and prestressed composites are presented in Figure
6-10. It was shown that the number of fibres aligned to 0° direction of the composite
increases with fibre prestress. Also the strain at deviation from linearity shows an
improvement with fibre prestress (see Figure 8-2). This suggests that the fibre alignment
could influence the non-linear behaviour of the composite. With reference to Figure 6-10,
146
it can also be observed that the fibres aligned to 0° direction of the composite reaches a
maximum of about 75% in 108 MPa prestressed composite and above which no further
improvement was observed. This also correlates with the observed optimum prestress limit
to attain higher strain to on-set of non-linearity.
8.1.2. Tensile strength and modulus

The measured tensile strength for the 0 MPa prestressed E-glass/913 epoxy material in the
current work is compared with the published literature results for the same material in
Table 8-2.
Table 8-2: Tensile strength comparison.
Researchers Tensile Strength (MPa) Vf (%)

Current Work 1311 (80) 58.2 (1.5)
Fernando and Al-Khodairi 1160 (40) 62.2 (1.6)
Patridge et al.120 1337 (63) 69.0 (1)
Dickson et al. 121 1400 (-) ~ 60.0 (-)
Broughton and Lodeiro122 1210 (-) 51 (-)
From Table 8-1 it can be observed that the average tensile strength and secant modulus are
independent of applied prestress. In the literature there is a contradiction in tensile
properties with fibre prestress. Zhao and Cameron studied the effect of fibre prestress on
tensile properties in co-mingled PP/glass fibre composite. They showed that with fibre
prestressing, the tensile strength and modulus increases. They suggested that this
improvement is due to the reduction in residual stresses.
Schulte and Marissen studied the effect of fibre prestress on hybrid (Kevlar/Carbon) epoxy
cross-ply composites. They studied two prestress levels 0 MPa and 341 MPa (1.1%
prestrained) in composites. From their study, they have shown that fibre prestressing has
minimised the transverse matrix cracking during tensile loading and improved the tensile
fracture strength and strain of composites. From Table 2-6 reproduced from their paper, it
can be observed that the average strength and modulus show an improvement. However,
the standard deviation of their measurements shows that the properties of prestressed
composites are within the limits of non-prestressed composites.
Tuttle et al. showed that in unsymmetrical carbon fibre/epoxy laminate, fibre prestressing
minimised the transverse matrix cracks during tensile loading. This correlates with Schulte
and Marrissen’s results. However, they showed that no correlation was observed in tensile
properties with an increase in fibre prestress.
147
Lee and Ali studied the effect of prestress on tensile properties of Kevlar49/epoxy and
Twaron/epoxy composites respectively. They both showed that tensile modulus and
fracture strain decreases with an increase in fibre prestress.
It was shown in Chapter 7 that after composite curing and releasing the prestress, the
residual strains in the fibre is given by:
ε f =ε f R +ε fT (8.1)
where εfR is the residual strain developed due to curing process and εfT is the tensile residual
strain remaining in the fibre due to release of prestress, which can be expressed as:
W ⎛ E mφ m ⎞
ε fT = ×⎜ ⎟ (8.2)
A f E f ⎜⎝ E mφ m − E f φ f ⎟
⎠
Therefore, the tensile failure strain of the prestressed composite to the first approximation
can be given by:
ε PC
F
= ε NPC
F
−ε f (8.3)
where, εFNPC is the average tensile failure strain of the non-prestressed (0 MPa) composite.
From the experimental results presented in Chapter 6, it was also confirmed that as the
applied fibre prestress increases, the residual tensile stress in the fibre increases; this was
also observed by Motahhari and Cameron. Figure 8-3 shows the theoretical prediction
given by equation 8.3 of the effect of the prestress on the tensile failure strain along with
the experimentally-measured results. From Figure 8-3 it can be observed that from the
theoretical prediction there is a reduction in the composite failure strain with increasing
fibre prestress. The experimental results are presented with standard deviations (99%
confidence limits). Considering the greater standard deviation in the experimental results
for all the prestressed composites, it can be concluded that no significant difference in the
failure strain was observed with prestress. This correlates with the results of Tuttle et al..
148
8.2 Post-tensile failure analysis
3.4
3.2
3
Failure strain ε C PC (%)
2.8
2.6
2.4
Theoretical Prediction
2.2
Experimental Results
2
-20 0 20 40 60 80 100 120 140 160 180
Prestress (MPa)
Figure 8-3: The effect of fibre prestress on composite failure strain.
The tensile fracture of the composites was examined from the macroscopic (by visual
inspection) to the microscopic (scanning electron microscopy) scale. The results from the
examination are discussed as follows.
8.2.1. Macroscopic examination

A representative tensile failure from each prestressed composite investigated in this study is
presented in Figure 8-4. Macroscopically all the prestressed composites fractured by
longitudinal splitting, which is a typical fracture of unidirectional E-glass/epoxy composite,
also known as “brush-like” failure; this type of failure has been reported by many
researchers85,122. From Figure 8-4, it can be observed that macroscopically there is no
difference in the tensile fracture of composite samples with fibre prestress.
149
Figure 8-4: Photo macrograph of static tensile specimens.
8.2.2. Microscopic examination

Samples for microscopic examination were taken from the representative specimens from
each prestressed composite. They were examined using a scanning electron microscope.
Figure 8-5 shows the micrographs from all the prestressed composites.
150
(i) – 0 MPa
(ii) – 51 MPa
151
(iii) – 80 MPa
(iv) – 108 MPa
152
(v) - 150 MPa
Figure 8-5: Micrographs of specimens tested in tension, (i) 0 MPa, (ii) 51 MPa,
(iii) 80 MPa, (iv) 108 MPa and (v) 150 MPa prestressed
composites. The failure mechanisms are denoted thus in each
figure: a – fibre fracture/pull-out, b – matrix hackle formation, c –
clean fibre surface, d – matrix plasticity, e – fibre impression and
f – fibre fragment.
From Figure 8-5 it can be observed that in general in all the prestressed composite
specimens tensile failure mechanisms like fibre/matrix debonding, fibre fracture/pull-out,
matrix hackle formation and matrix plastic deformation are present. The mechanism of
hackle formation is presented in Chapter 9. A clean fibre surface indicates a weak
interfacial bonding between fibre and matrix, which is a typical behaviour of E-glass/epoxy
composites. These tensile failure modes were also reported by many researchers. There is a
difference in the fibre damage at high prestress levels (108 and 150 MPa) when compared
to low prestress levels (0, 51 and 80 MPa). At a high prestress level a greater number of
fibre fragments can be observed (as shown in Figure 8-5 (iv) and (v)).
The possible reasons for the observed greater number of fibre fragments at high prestress
levels are: (i) a reduction in the number of weak fibres during prestressing and (ii) some
tensile residual strain in the fibre.
153
(i) Reduction in weak fibres

In a non-prestressed composite both weak and strong fibres are present. This difference in
the strength of glass fibres is due to the presence of surface flaws. This strength
distribution is also present along the length of a fibre. Therefore, fibres may fail at their
weakest point, at stresses below the fracture stress of the bulk composite. When a non-
prestressed composite is subjected to a tensile load, fibre failure first occurs in the weakest
fibres. This may be simultaneous with matrix cracking and interfacial debonding.
Prior to composite curing, when a prestress is applied to the fibres, the weak fibres will
fracture leaving predominantly strong fibres. The higher the prestress level applied the
more weak fibres break. When a prestressed composite is subjected to a tensile load, the
first failure of the fibres will occur at a higher applied tensile stress level, compared to the
fibre failure in a non-prestressed composite. The fracture of the fibre will release strain
energy and this is dissipated in various ways, depending on the strength of the fibre/matrix
adhesion. In this study, all the prestressed composites were manufactured from the same
prepreg material. Therefore, there is no chemical change in the fibre/matrix adhesion. The
energy released by the fibre fracture in a prestressed composite will be greater than the
energy released by the fibre fracture in a non-prestressed composite. This release of energy
will increase with an increase in fibre prestress. At a very high-energy release rate, possibly a
“shock wave” effect would be induced in the neighbouring fibres and also along the
fractured fibre length, which would lead to an increase in the number of fibre fragments
present in the fractured surface.
(ii) Tensile residual strain in fibre

In Chapter 7 it was shown that the residual tensile stress in fibres increases with applied
prestress. This was also observed by Motahhari and Cameron. Scherf and Wagner123 have
studied the effect of fibre prestress on interfacial shear strength of single fibre micro-
composites using a fibre fragmentation method. They have shown that at a higher applied
fibre prestress, the number of fibre fragments increased. Their measurement of interfacial
shear stress increases with an increase in fibre prestress. They explained that “the apparent
increase in the value of the interfacial shear strength does not mean that stronger interfaces
are created by fibre pre-loading; rather, the larger value obtained is an artefact arising from
the increased number of fibre breaks obtained”. They showed that with the applied
prestress the residual tensile stress increases. This is stored as additional strain energy in the
fibres, which might induce dynamic recoil effects upon fibre breakage. This could induce
supplementary fibre breaks (possibly by a “shock wave” effect) and therefore an increased
number of fragments. This means that composites with a higher applied prestress will have
greater strain energy stored in the fibres.
154
8.3 Static compression results and discussion
Static compression tests are conducted in prestressed composites according to the

procedure explained in section 4.9.2. Figure 8-6 and Figure 8-7 show the compressive
strength and modulus results of prestressed composites. The individual sample
compression test results are presented in Appendix C. Typical axial compressive stress-
strain responses of all the prestressed composites are shown in Figure 8-8.
1400
Compressive Strength (MPa)
1300
1200
1100
1000
900
-20 0 20 40 60 80 100 120 140 160
Prestress Level [MPa]
Figure 8-6: Compression strength of prestressed composites.
48
46
Compressive Modulus (GPa)
44
42
40
38
36
-20 0 20 40 60 80 100 120 140 160
Prestress Level [MPa]
Figure 8-7: Compression modulus of prestressed composites
155
From Figure 8-6 and Figure 8-7 it can be observed that there is an increase of 9% in
average ultimate compressive strength (UCS) and 9% in average compressive modulus of
composites up to 80 MPa prestress. This increase in the compressive properties with
prestress could be due to (i) an improvement in the fibre alignment and (ii) a reduction in
the tensile matrix residual stresses. Previous researchers have shown that fibre
misalignment can significantly reduce the compressive properties of unidirectional
composites. Above an optimum prestress level (108 MPa for compressive strength and 80
MPa for compressive modulus) the compressive properties decrease. This reduction in
compressive properties could be due to the increase in the interfacial shear residual stresses
in the composite, which arise due to increased tensile residual stresses in the fibre and
compressive residual stresses in the matrix. This increase in interfacial shear residual
stresses could reduce the resistance to interfacial crack initiation and propagation. This
could lead to the onset of fibre-matrix cracking at a very early stage of the loading. This will
result in a reduction in the compressive properties of the composite. This reduction in
compressive properties suggests that an optimum limit of prestress exists for the
improvement. Motahhari and Cameron76, 79, and Zhao and Cameron observed this
phenomenon of optimum prestress limit in tensile, flexural, impact and inter-laminar shear
strength.
550
500
Stress (MPa)
900
450
800
400
700
350
9000 9500 10000 10500 11000 11500 12000
600
Stress (MPa)
Strain (µε)
500
0 MPa Prestress
400
51 MPa Prestress
300
80 MPa Prestress
200
108 MPa Prestress
0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
Strain (µε)
Figure 8-8: A typical compressive stress/strain curve for [0°]16 prestressed

composites.
From Figure 8-8 and Table 8-3 it can be observed that there is an increase in slope for all
prestressed composite when compared to 0 MPa prestressed composite. The slope of the
156
stress-strain curve increased up to 80 MPa prestressed composite. This shows that there is
an optimum limit for the increase in the slope. Above this optimum limit, the slope of the
composite decreased.
Table 8-3: The slope calculated from linear regression analysis for each
prestressed composite presented in Figure 8-8.
Prestress (MPa) Slope (m)

0 0.0407
51 0.0453
80 0.046
108 0.0454
150 0.0429
In this study the compressive strain was measured using electrical resistance strain gauges
on both sides of the specimen. The strain measured from both sides of the specimen
showed a variation of 5% and is the standard variation found in the testing. A comparison
of front and rear strain measurement during a compressive test is shown in Figure 8-9.
800
700
600
Stress (MPa)
500
400
300
200 Strain gauge front

Strain gauge rear
100
0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
Strain (µε)
Figure 8-9: Comparison of compressive strain measured in front and back of

the composite.
157
8.4 Compressive strength comparison
Compressive strength measured in this study and the values available in the literature124 for
the same material are summarised in Table 8-4. From Table 8-4 it can be observed that the
compressive strength values measured in the current study is similar to the CRAG
standard, but higher than the ASTM methods. It can also be noticed that the deviation of
UCS measured in the current study (using the ICSTM rig) is small (4%) when compared to
the CRAG value (15%). The large discrepancy between the ASTM and the current work
could be due to the specimen misalignment during testing, and manufacturing problems
like fibre waviness. Mrse and Piggott have shown that fibre waviness can significantly
reduce the compressive strength of composites. In their experiments on carbon/PEEK
composites with four different prepreg types, deviations ranging from 1º to ~6º in fibre
waviness were obtained, with corresponding strengths from 1.9 GPa to 1.5 GPa. Matthews
studied the effect of different compression test methods and specimen configurations on
the compressive strength of carbonXAS/914 epoxy composite. He showed that out of the
12 different test methods investigated, the ICSTM test method and specimen configuration
gave the higher mean strength, together with low scatter. He concluded that the
discrepancy in the compression strength results is due to specimen misalignment, stability
problems and operator inexperience.
Table 8-4: Comparison of UCS of unidirectional control specimens with

different test methods . The number in parentheses represents
the standard deviation. * l – gauge length, w – width, t –
thickness.
Specimen dimension Test speed UCS

Test method
(l × w × t)* mm [mm/min] [MPa]
ASTM D3410 12.7 x 6.35 x 3 1.26 642 (46)
ASTM D695 (mod) 80 x 19/12 x 2 1 592 (45)
CRAG 400 (Celanese) 10 x 10 x 2 3.25 1230 (188)
Current Work (ICSTM) 10 x 10 x 2 1 1177 (47)
8.5 Post-compressive failure analysis
8.5.1. Macroscopic examination

Figure 8-10, shows representative macroscopic compressive failures of all prestressed
composites. It can be observed that all the prestressed composites show longitudinal
splitting and fibre fracture except the 0 MPa prestressed composite. Under close
examination it was found that the 0 MPa prestressed composites have failed in the fibre
158
microbuckling mode as shown in Figure 8-11, whereas all the rest of the prestressed
composites show extensive longitudinal fibre splitting and minimal global buckling (see
Figure 8-11). The sudden catastrophic nature of the failure made detection of any failure
initiation event extremely difficult.
Figure 8-10: Representative compressive failures of 0, 51, 80, 108 and 150
MPa prestressed composites.
159
Figure 8-11: A magnified view of the macroscopic failure of all the

prestressed composites. (i) 0 MPa, (ii) 51 MPa, (iii) 80 MPa, (iv)
108 MPa and (v) 150 MPa.
The possible reasons for the observed change in the damage mechanism could be due to
the improvement in fibre alignment and reduction in matrix residual stresses.
Fibre misalignment or waviness can occur during the manufacturing process (as explained
in section 2.5). When a compressive load is applied, the misaligned or wavy fibres start to
buckle. Hahn and Williams and Lankford125 have shown that in a unidirectional composite
160
the in-plane buckling of the fibres places the matrix pre-dominantly in shear. This leads to
a shear mode of fibre “microbuckling or kink-band” failure, which is described in section
8.5.2.
During manufacturing of prestressed composites, the fibres are restrained under a constant
load and this prevents fibre movement. Thus, fibre misalignment or waviness is reduced in
prestressed composites. From Figure 8-11 it can be observed that prestressed composites
show an absence of the microbuckling mode failure. This indicates that prestressing could
have reduced fibre misalignment. It can also be observed that prestressed composites show
more longitudinal splitting and fibre fractures when compared to the 0 MPa prestressed
composite. This change in macroscopic failure mechanism also supports the observed
improvement in the compressive strength and modulus of the prestressed composites.
8.5.2. Microscopic examination
Non-prestressed (0 MPa) composite:

Figure 8-12 illustrates a fibre “kink-band” formation observed in a non-prestressed (0
MPa) composite. This kink-band failure was observed through thickness (2 mm) of the
composite specimen in the end-tab region. An arrow mark in Figure 8-12 shows the
propagation direction of the kink-band failure. From Figure 8-12 it can be observed that
the kink-band has initiated near the free-edge of the composite.
Figure 8-12: Kink-band failure mechanism observed in non-prestressed (0

MPa) composite.
161
Figure 8-13 and Figure 8-14 shows the positions 1 and 2 in Figure 8-12. With reference to
Figure 8-13 it can be observed that initially the failure started as two bands of fibre-kinking,
which then changed to one kink-band (see Figure 8-14). As the failure propagates the
failure mode changed from kink-band to fibre buckling (position 3) as seen in Figure 8-12.
The kink boundary orientation β was calculated to be 25° for the kink-band shown in
Figure 8-14. The calculated kink-band orientation is within the typical rage (10° –30°)
reported by Budiansky and Fleck.
Figure 8-13: A magnified view of position 1 in Figure 8-12.
Figure 8-14: A magnified view of position 2 in Figure 8-12.
162
A close-in view of the Figure 8-11 (i) (compressive failure of 0 MPa prestressed composite)
is presented in Figure 8-15. From Figure 8-15 it can be observed that multiple kink-band
failure is apparent. From the microscopic examination it can be concluded that the 0 MPa
prestressed composite fractures by a kink-band mechanism under compressive loading.
Figure 8-15: A magnified view of the Compressive failure of 0 MPa

prestressed composite presented in Figure 8-11 exhibiting
multiple kink-band formation.
Prestressed composite:
Prestressed composites showed extensive longitudinal splitting and fibre fracture as a result
of compressive loading. Typical micrographs of prestressed composite compressive
fractures are shown in Figure 8-16 and Figure 8-17.
From microscopic examination, it can be observed that prestressed composites exhibited
fibre-matrix debonding (A), matrix deformation (B), hackle formation (C), fibre fracture
(D) and fibre imprint (E) as shown in Figure 8-16 and Figure 8-17. In contrast to the non-
prestressed composite, no fibre kink-band failure was observed.
In prestressed composite manufacture, as the fibres are prestressed, there is no possibility
for the fibre to move and this reduces the fibre misalignment. When a compressive load is
experienced by the prestressed composite, the brittle glass fibres fracture by crushing or
collapsing damage, resulting in an imploded failure mode as represented in Figure 8-18.
163
Figure 8-16: The typical fibre failure of prestressed composite in compressive

loading. (i) 51 MPa, (ii) 80 MPa, (iii) 108 MPa and (iv) 150 MPa
prestressed composite.
The fibre crushing mode of failure is apparent from the micrograph presented in Figure
8-16. In all prestressed composites more matrix hackle formation can be observed (see
Figure 8-17), whereas in the 0 MPa prestressed sample no matrix hackle formation was
observed. The possible reason for the observation of matrix hackle formation in
prestressed composites is an improvement in fibre alignment and a change in the residual
stresses in the matrix and interface.
164
Figure 8-17: The typical matrix failure of prestressed composite in

compressive loading. (i) 51 MPa, (ii) 80 MPa, (iii) 108 MPa and
(iv) 150 MPa prestressed composite.
Figure 8-18: Schematic representation of imploded failure mechanism in

prestressed composite during compressive loading.
165
Figure 8-19 shows a matrix “curling” fracture in a prestressed composite tested in

compression. This type of matrix fracture in compressive loading of a unidirectional
composite was not reported in the literature. In prestressed composites, by applying
compressive stresses, matrix tensile residual stresses are reduced. With increasing prestress
the compressive residual stress in the matrix will increase. When prestressed composites are
subjected to compression, because of the improved fibre alignment and increase in matrix
compressive stresses, the matrix could fracture normal to the fibre direction. When the
fibre collapses the interfacial crack will propagate leading to relaxation of the residual stress
in the interface and matrix, which could lead to curling of the matrix.
Figure 8-19: Matrix curling fracture in a prestressed composites. (i) 50 MPa

and (ii) 108 MPa prestressed composite.
166
This chapter presented the results from static tensile and compressive properties of non-
prestressed and prestressed composites. From the results presented the following
conclusions can be made.
Static tensile properties

Fibre prestressed composites show an improvement in strain at deviation from linearity in
their stress-strain curves. This suggests that prestressing could have increased the resistance
to the onset of damage (matrix cracking, fibre-matrix debonding and weak fibre failure).
This improvement in εSDL was observed up to 108 MPa prestress, above which it indicates
a reduction. This suggests that an optimum prestress limit for the improvement in strain at
deviation from linearity exists.
A theoretical prediction for the effect of prestress on the failure strain of composites was
presented. This theory predicts that the fracture strain of the composite decreases with an
increase in prestress. This is because the residual tensile stress in the fibre increases with
prestress. This reduces the tensile fracture strain of the composite. However, from the
experimental results no significant change in the tensile fracture strain was observed. This
could be due to greater scatter in the measurements. It was also shown that the tensile
strength and modulus of composites are independent of fibre prestressing.
From the macroscopic examination it can be concluded that there is no change in the
composite failure mechanism with fibre prestress. However, from the microscopic
examination it was found that at higher prestress levels more fibre fragmentation was
observed. This could be due to (i) breaking of the weak fibres under the prestressing load
leaving predominantly strong fibres in the composite and (ii) an increase in the tensile
residual stress in the fibre with prestress. This results in an increased strain energy stored in
the fibres.
Static compressive properties

Fibre prestressing gives an improvement in the average compressive strength and modulus
by 9%. However, above an optimum prestress limit (108 MPa for compressive strength
and 80 MPa for compressive modulus) the property reduces. This could be because of the
increase in micro-residual stresses at the interface with fibre prestress, above the optimum
limit the fibre-matrix interface is prone to debonding, which could lead to a reduction in
compressive properties.
Fibre prestressing gives a change in compression damage mechanism. The non-prestressed
(0 MPa) composites fractured by fibre kink-band formation, whereas prestressed
composites fractured by extensive longitudinal splitting and matrix deformation. This
167
change in damage mechanism suggests that prestressing could have caused a reduction in
the fibre waviness in composites.
From microscopic examinations, a curling mode of matrix failure was found in prestressed
composites. This type of fracture mode has not been reported in the literature previously.
The possible mechanisms for matrix curling could be due to the relaxation of the residual
stress in the interface and matrix during fibre-matrix debonding.
168
CHAPTER 9
Fatigue behaviour
OVERVIEW
This chapter discusses the results from tension-tension (T-T) and tension-
compression (T-C) fatigue behaviour of non-prestressed and prestressed
unidirectional composites investigated in this study. Fatigue performance in both
T-T and T-C modes are separately presented in four stages:
The fatigue behaviour of non-prestressed composites is presented and
compared with the published literature for the same material.
The fatigue properties of prestressed composites are discussed by comparing
the fatigue life in terms of S-N curves, normalisation techniques and Weibull
statistics.
The stiffness degradation and surface temperature rise during fatigue was
studied to enable an understanding of the influence of prestress on damage
development.
Macroscopic and microscopic post-fatigue failure analysis was studied in order
to assess the relation between the failure mode and prestress.
9.1 Tension-tension fatigue
9.1.1. Fatigue of non-prestressed composites

Figure 9-1 shows the normalised S-N curve for the non-prestressed unidirectional
composite from the current work along with two similar systems published in the literature
(Fernando and Al-Khodairi (GFAK), Broughton and Lodeiro (GBL)). The vertical axis in
Figure 9-1 represents the peak stress normalised to the respective ultimate tensile strength
CHAPTER 9 Fatigue behaviour
(UTS) for each of the composite systems. A summary of the lay-up sequence, ultimate
tensile strength (UTS), fibre volume fraction and fatigue test parameters for the composites
are presented in Table 9-1.
From Figure 9-1 and Table 9-1, it can be observed that firstly the relative slopes for the
current work, GFAK and GBL data sets are very similar. Secondly, the scatter in the
fatigue results for the current work, compared with GFAK and GBL data is less. Finally, a
variation in fibre volume fraction of 51% to 62% does not seem to have an effect on
fatigue properties.
1
Current Work
0.9 GFAK
GBL
Normalised Stress (σmax/σUTS)
0.8 Linear (Current Work)

Linear (GFAK)
0.7 Linear (GBL)
0.6
0.5
0.4
0.3
0.2
2 3 4 5 6 7
Log N
Figure 9-1: Comparison of normalised fatigue data of non-prestressed

composite (current work) with similar systems reported in the
literature.
Table 9-1: Summary of composite lay-up sequences and fatigue test

parameters used in the current work and those reported by
previous researchers.
Rate of stress
Material Stress ratio UTS Vf
Lay-up application
(E-glass/913 epoxy) (R) (GPa) (%)
(kN s-1)
Current work [0]16 +0.1 1.31 58.2 250
GFAK [0]16 +0.1 1.16 62.2 500
GBLError! Bookmark not
defined.
[0]8 +0.1 1.21 51.0 Fixed frequency 5 Hz
170

Figure 9-2 represents the S-N curves for non-prestressed and prestressed composites with
the UTS included in a linear-regression fit. The quasi-static UTS measured at 0.6 kN s-1 was
included at 0.5 cycles (log 0.5 = -0.3) in the S-N curves. Table 9-2 shows the individual
fatigue life results of non-prestressed and prestressed composites plotted in Figure 9-2. In
the S-N curves, the issue of including the UTS at 0.5 cycles is questionable, because the
UTS used at 0.5 cycles is measured from quasi-static test and not fatigue. Also the quasi-
static test rate is different to the fatigue test rate. Therefore, S-N curves were also plotted
without including the UTS at 0.5 cycles as shown in Figure 9-3.
Comparing Figure 9-2 and Figure 9-3 it can be observed that the S-N curves without
including UTS show a linear behaviour in both low and high cycle fatigue regions. On the
other hand, the S-N curves including the UTS show that the fatigue data at the low cycle
region did not fit well to the linear behaviour observed at high cycle region. Both Figure
9-2 and Figure 9-3 shows a linear behaviour in the high cycle region S-N curves, this was
previously observed by Mandell. Mandell conducted a series of experiments on different
types of glass-fibre reinforced plastics (GFRP). He showed that when the number of cycles
to failure lay between 102 and 106, the S-N curve can be approximated by a straight line of
the form:
σ max = σ UTS + b log N (9.1)
where σmax is the maximum applied stress, σUTS is the ultimate tensile strength, N is the
number of cycles to failure and b is a constant. The ratio of b/σUTS was defined by Mandell
as the fractional loss in tensile strength per decade of cycles. The values of the slope (b)
from linear-regression data analysis and the tensile strength decay per decade of cycle
(b/σUTS) are summarised in Table 9-3.
From Table 9-3, it can be observed that fibre prestressing has minimised the tensile
strength degradation per decade (b/σUTS) in composites when compared to non-prestressed
composites. There is a reduction of 1% in the tensile strength degradation per decade with
51 MPa prestress.
171
1600
0 MPa Prestress
51 MPa Prestress
1400
108 MPa Prestress
Linear (0 MPa Prestress)
1200
Peak stress (MPa)

1000
800
600
Unidirectional E-glass/ 913 epoxy composite
Tension-tension fatigue
400 Stress ratio R = +0.1
Loading rate = 250 kN s-1
200
-1 0 1 2 3 4 5 6 7
Log N
Figure 9-2: Comparison of the tension-tension fatigue results of the non-

prestressed and prestressed composites. The linear regression fit
was calculated including the UTS at 0.5 cycles. The arrow mark
represents the sample run-out.
1400
0 MPa Prestress
51 MPa Prestress
1000 Linear (51 MPa Prestress)
Peak stress (MPa)
800
600

400 Tension-tension fatigue
Stress ratio R = +0.1
200
1 2 3 4 5 6 7
Log N

prestressed and prestressed composites. The linear regression fit
was calculated without including the UTS at 0.5 cycles. The
arrow mark represents the sample run-out.
172
Table 9-2: The individual fatigue life results of non-prestressed and

prestressed composites plotted in Figure 9-2. The arrow mark (⇨)
represents that the specimen ran-out.
0 MPa 51 MPa 108 MPa

Peak stress Number of Number of Number of
Log N Log N Log N
MPa cycles cycles cycles
1180.76 151 2.178 52 1.176 141 2.149
1180.76 334 2.523 68 1.832 240 2.380
1180.76 423 2.626 277 2.442 413 2.615
1049.56 641 2.806 146 2.164 1,122 3.049
1049.56 496 2.695 410 2.612 593 2.773
1049.56 574 2.758 430 2.633 585 2.767
918.36 1,841 3.265 3,026 3.480 1,702 3.230
918.36 2,138 3.330 3,262 3.513 1,910 3.281
918.36 2,887 3.460 1,167 3.067 1,977 3.296
918.36 - - 1,057 3.024 - -
787.17 9,206 3.964 33,229 4.521 4,537 3.656
787.17 9,702 3.986 8,195 3.913 8,978 3.953
787.17 9,140 3.960 7,670 3.884 8,834 3.946
655.97 46,302 4.665 39,175 4.593 20,586 4.313
655.97 52,027 4.716 69,275 4.840 17,479 4.242
655.97 60,321 4.780 147,790 5.169 26,249 4.419
590.38 94,131 4.973 577,556 5.761 86,772 4.938
590.38 262,964 5.419 455,483 5.658 66,216 4.820
590.38 230,227 5.362 317,964 5.502 94,015 4.973
524.78 446,633 5.649 900,042 5.954 851,335 5.930
524.78 550,935 5.741 942,689 5.974 673,667 5.828
524.78 614,721 5.788 1,130,147 ⇨ 6.053 1,152,913 ⇨ 6.061
524.78 - - 1,058,439 ⇨ 6.024 - -
459.184 2,180,001⇨ 6.338
459.184 1,231,979 ⇨ 6.090
459.184 1,535,627 ⇨ 6.186
173
Table 9-3: Summary of slope (b) and degradation of tensile strength per
decade (b/σUTS) of non-prestressed and prestressed composites.
Prestress
Slope (b) b/σUTS %
(MPa)
0 -161.0 11.1
51 -137.8 10.1
108 -156.0 10.9
However, from Figure 9-3, it can be seen that the 108 MPa prestressed composite fatigue
data does not show a linear behaviour. Figure 9-4 presents a second order polynomial fit to
the fatigue results without including the UTS. The polynomial fit was previously used by
other researchers85,87 to fit the fatigue data of unidirectional glass fibre/913 epoxy and
cross-ply carbon/epoxy composites. From Figure 9-4, it can be seen that the low and high
cycle fatigue regions of all the prestressed composites fit the polynomial curve. The
polynomial curve fit could be used to estimate the number of cycles to failure at any given
stress level between the UTS and the fatigue limit of the matrix (that is the run-out sample
peak stress level).
1300
0 MPa Prestress
51 MPa Prestress
108 MPa Prestress
1100
Polynomial fit (0 MPa Prestress)
Peak stress (MPa)
900 Polynomial fit (108 MPa Prestress)
700
500 Unidirectional E-glass/ 913 epoxy composite

300
1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5
Log N

prestressed and prestressed composites. A second order fit was
calculated without including the UTS.
174
9.1.3. Statistical analysis of fatigue data

Fatigue data from each set of non-prestressed and prestressed composites were subjected
to statistical analysis to determine the effect of prestress on the fatigue life. Three types of
statistical analysis were carried-out a) normalisation, b) Student t-test and c) a two-
parameter Weibull distribution.
In order to compare the fatigue life of non-prestressed and prestressed composites, the
fatigue peak stress (σmax) was normalised with respect to the UTS as shown in Figure 9-5.
The standard deviation presented in Figure 9-5 was calculated for the 95% confidence
level. Harris et al.126 have shown that by normalising the peak stress with respect to the
UTS, the effect of different fibre strengths, fibre volume fractions and lay-up sequences
can be eliminated.
From Figure 9-5, it can be observed that prestressing E-glass/913 composite resulted in an
improvement in fatigue life in the low stress fatigue region. At a normalised stress of 0.4
(524 MPa peak stress), the improvement achieved in the average fatigue life for the 51 and
108 MPa prestressed composites were 1.75 and 1.55 times greater than that of the 0 MPa
(non-prestressed) composite. However, at a normalised stress above 0.6, no significant
variation in fatigue life was observed with prestress.
1
0 MPa Prestress
51 MPa Prestress
0.9
108 MPa Prestress
0.8

0.7
0.6
0.5

0.3
1 2 3 4 5 6
Log N
Figure 9-5: Comparison of normalised tension-tension fatigue of non-

prestressed and prestressed composites. The error bars were
calculated for 95% confidence level.
175
Figure 9-6 and Figure 9-7 show the fatigue life as a function of applied prestress in the low
stress fatigue region (590 MPa and 524 MPa peak stress). At 590 MPa peak stress level
three specimens with 80 MPa prestress were also tested to identify the optimum limit for
the fatigue life improvement. With reference to Figure 9-6, it can be observed that the
maximum fatigue life for the E-glass/913 epoxy composite system is at 51 MPa prestress.
Above this prestress level, there is an indication of a reduction in fatigue life. This indicates
that there is an optimum prestress limit to achieve maximum fatigue life. A student t-test
was used to test the significance of the observed difference between the results of non-
prestressed and prestressed composites at 524 MPa peak stress. The test showed that the
non-prestressed and 51 MPa prestressed composites are significantly different at a 99%
confidence level. From Figure 9-8, it can be observed that the fatigue life of non-
prestressed and prestressed composites are all the same and shows that there is no
influence of prestress in the high stress fatigue region.
600000
Peak Stress = 590 MPa
400000
Cycles to failure N
300000
200000
100000
0
-20 0 20 40 60 80 100 120
Prestress (MPa)
Figure 9-6: Fatigue life as a function of applied prestress in the low stress
region (peak stress of 590 MPa).
176
1300000
1100000 Loading rate = 250 kN s-1
Cycles to failure N
900000
700000
500000
300000
-20 0 20 40 60 80 100 120
Prestress (MPa)
region (peak stress of 524 MPa). The arrow mark represents the
specimen run-out.
4000
3500
3000
Cycles to failure N
2500
2000
1500
1000 Unidirectional E-glass/ 913 epoxy composite

0
-20 0 20 40 60 80 100 120
Prestress (MPa)
Figure 9-8: Fatigue life as a function of applied prestress in the high stress
177
Researchers127,128 have shown that a two-parameter Weibull distribution function can be

used to determine the degree of varibility of the fatigue life of the composite. A two-
parameter Weibull distribution was calculated for the non-prestressed and prestressed
composite fatigue life data. The two-parameter Weibull cumulative distribution function
(Cdf) is given in Table 9-4 .
Table 9-4: Plotting formulae for Weibull two-parameter distribution.
Distribution Cdf F(x) Horizontal axis Vertical axis

m
⎛ ⎛ x ⎞⎞
ln ( xi ) ln (− ln (1 − pi ))
Weibull
two-parameter 1 − exp⎜⎜ − ⎜ ⎟ ⎟⎟
⎝ ⎝θ ⎠⎠
Here xi is the cycles to failure; pi is the probability plotting position given by the formula (i-
0.5)/n; n is the total number of samples; θ is the characteristic value or scale parameter; m is
the shape parameter usually called the Weibull modulus and reflects the degree of
variability in the fatigue life. For example, the higher is the Weibull modulus, the less
variable is the fatigue life.
0.5
0
Ln(-Ln(1-Pi))
-0.5
0 MPa Prestress
-1
51 MPa Prestress
-1.5 108 MPa Prestress
Linear 0 MPa Prestress

-2 Linear 51 MPa Prestress

-2.5
12.7 12.9 13.1 13.3 13.5 13.7 13.9 14.1 14.3 14.5 14.7
Ln Xi (Cycles to Failure)
Figure 9-9: Weibull distribution for fatigue life at 524 MPa peak stress level
(low stress level).
178
1
0 MPa Prestress
0.5 51 MPa Prestress

108 MPa Prestress
0 Linear 0 MPa Prestress
Ln(-Ln(1-Pi))
-0.5 Linear 108 MPa Prestress
-1
-1.5
-2
-2.5
7 7.5 8 8.5 9 9.5 10 10.5 11 11.5 12
Figure 9-10: Weibull Distribution for fatigue life at peak stress level 788 MPa
(middle stress level).
0.5
0
Ln(-Ln(1-Pi))
-0.5
-1 0 MPa Prestress
51 MPa Prestress
-1.5 108 MPa Prestress
-2 Linear 51 MPa Prestress
-2.5
3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8
Figure 9-11: Weibull distribution for fatigue life at 1180 MPa peak stress level
(high stress level).
Table 9-5 presents the Weibull modulus for the prestressed composites presented in Figure
9-9 to Figure 9-11. It can be seen that the Weibull modulus of 51 MPa prestressed
composites is higher than the non-prestressed composite for low stress fatigue (524 MPa).
This shows that the degree of variability in fatigue life of 51 MPa prestressed composites is
179
smaller than for the non-prestressed composite. This also supports the concept that the
improvement in fatigue life observed at 51 MPa prestressed composites is statistically
significant. However, from Table 9-5 it can be observed that the Weibull modulus is small
in all the composites at high stress fatigue (1180 MPa).
Table 9-5: Summarises the Weibull modulus of non-prestressed and

prestressed composites presented in Figure 9-9 to Figure 9-11.
Weibull modulus (m)

Prestress
MPa Peak stress Peak stress Peak stress
524 MPa 788 MPa 1180 MPa
0 6.4197 30.757 2.0826
51 10.781 3.4143 1.1388
108 4.199 2.7084 2.1251
Low stress level fatigue

From the normalisation (see Figure 9-5), student t-test (see Figure 9-6 to Figure 9-8) and
Weibull distribution (see Figure 9-9 to Figure 9-11 and Table 9-5), it can be seen that in the
low stress fatigue region there is an improvement in fatigue life with prestress. Talreja
showed that in the low stress region, matrix cracking and fibre/matrix interfacial
debonding are the predominant damage mechanisms. Therefore, low stress level fatigue
performance of composite materials depends on the matrix and interfacial properties or in
other words matrix and interface resistance to crack initiation and crack growth.
During composite processing, tensile residual stresses are developed in the matrix and
corresponding compressive residual stresses are developed in the fibres due to anisotropic
thermal expansion of the composite constituents (fibre and matrix). In extreme cases the
tensile stress component in the matrix can be as high as the matrix yield strength1,129. When
the composite is subjected to an external mechanical load, the first damage can prematurely
occur in the matrix due to these tensile residual stresses. On the other hand in a prestressed
composite, the matrix is subjected to a compressive stress due to the release of fibre
prestress. This in turn reduces the process-induced tensile residual stresses in the matrix.
This reduction in residual tensile stresses in the matrix would increase the strain to crack
initiation and/or crack propagation in the matrix. This could have resulted in the
improvement in fatigue life of the composite and suggests that prestressed composites
have improved resistance to matrix damage.
From Figure 9-6, it can also be observed that there is a reduction in fatigue life above the
optimum limit (51 MPa prestress). Mottahari and Cameron have also shown that there is
an optimum prestress limit above which the residual stresses at the fibre/matrix interface
180
increase. This may reduce the resistance to interfacial crack initiation and/or crack growth.
This suggests that the observed reduction in fatigue life above the optimum prestress level
may be attributed to the weak interfacial crack resistance.
Middle and high stress level fatigue

From Figure 9-5, Figure 9-8, Figure 9-10 and Figure 9-11, it can be observed that above
0.55 normalised peak stress, regardless of prestressing, the fatigue life of composites are the
same. This may be because the strain induced in the composite during fatigue is well above
the matrix cracking strain. Also other researchers92,107,133 have shown that in high stress
fatigue, the fatigue damage is a combination of matrix cracking, interfacial debonding,
delamination and fibre fracture. Above 0.8 normalised peak stress the 51 MPa prestressed
composite indicated a reduction in the fatigue life. However, a statistical analysis of the data
using the student t-test showed that there is no significant difference between the 0 MPa
and 51 MPa prestressed composites.
9.1.4. Fatigue damage monitoring

In this study, the damage induced during fatigue loading was studied in non-prestressed
and prestressed composites using two methods (1) real-time monitoring of stiffness
degradation88,130 and (2) surface temperature rise131,132. Visual inspection of specimens was
also carried out. Reifsnider and other researchers107,85 have shown that with a cyclic load,
the reduction in stiffness happens continuously over a period of time and this is an
excellent indicator of the damage development in composite materials. It has also been
shown131,132 that the surface temperature rise can be used as an indicator of damage
development since much of the energy released during fatigue is converted into heat. The
temperature rise is generally measured using an infrared thermographic technique.
However, it was not possible to use this technique as it was not available and an alternative
technique using a surface-bonded thermocouple was used. In this study all the fatigue test
specimens were instrumented with a surface-mounted extensometer (to measure the strain)
and a surface-bonded thermocouple (to measure the change in surface temperature of the
specimen). The strain and corresponding load measurements were used to calculate the
change in stiffness.
Non-prestressed composite
Figure 9-12 shows the normalised stiffness and surface temperature change as a function of
fatigue life for a non-prestressed composite during fatigue testing at 655 MPa peak stress.
In general, the stiffness degradation during fatigue exhibited a three-stage behaviour. This
was also reported by other researchers88, 107,85. Stage I is characterised by an initial decrease
in stiffness and usually corresponds to 10% of the fatigue life. Stage II is an intermediate
181
but long period of approximately 85 – 90% of the fatigue life. During this period, stiffness
reduction was observed to be small. Stage III is very short-lived and is characterised by a
rapid decrease in stiffness as a result of catastrophic failure of the specimen. This
catastrophic “sudden-death” behaviour of glass /epoxy composites was also reported by
Dharan133 and Curtis . The rapid decrease in stiffness values in stage III correlates with
extensive splitting parallel to the fibres and delamination of the specimen, leading to a
brush-like failure.
Steif134 showed that fibre breakage, combined with interfacial debonding of the
fibre/matrix interface, would lead to stiffness reductions. Furthermore, Fernando and Al-
Khodairi have shown that the extent of the interfacial debonding was higher for samples
tested at higher cyclic stress. Curtis135 has shown that in unidirectional glass/epoxy
composites, matrix cracking and interfacial debonding causes longitudinal splitting of
fibres. He further showed that the rate of degradation processes in the matrix and at the
interface is a function of the bulk strain in the resin as well as the nature of the matrix.
1.2 25
Normalised Stiffness III
Temperature Rise
1
20
Normalised Stiffness (E/E0)
Temperature Rise oC
0.8
15
II
0.6
10
0.4
I
5
0.2
0 0
0 0.2 0.4 0.6 0.8 1 1.2
Normalised life (-)
Figure 9-12: The normalised stiffness and surface temperature change as a

function of fatigue life.
From Figure 9-12, it can also be observed that the surface temperature change showed a
three-stage behaviour similar to the stiffness decay. Stage I shows a rapid increase in
temperature. Stage II shows a small increase in the temperature rise. Stage III shows a rapid
increase in temperature when approaching failure. One of the limitations of the surface-
bonded thermocouple technique is the location of the thermocouple with respect to the
fracture zone, which can greatly influence the measured surface temperature. In other
182
words, if a thermocouple were located in the vicinity of a longitudinal split or internal

delamination, then a high surface temperature would be recorded. Also delamination of the
surface plies in some experiments resulted in the thermocouple becoming ineffective.
During fatigue testing damage development in the composite was also monitored by visual
examination of the specimens. Figure 9-13 shows the macroscopic damage development in
a unidirectional composite during fatigue testing at 787 MPa peak stress. In stage I of the
fatigue life, no macroscopic damage was observed. However, acoustic emissions were
noticed. Bhat and Murthy136 have shown that during fatigue of unidirectional glass
fibre/epoxy composites, acoustic emissions during stage I correspond to the occurrence of
a number of matrix cracks initiated and distributed over the specimen length. This could
explain the possible reason for the initial reduction in stiffness and increase in surface
temperature.
In stage II, at 0.2 normalised life, initiation of longitudinal splits in the surface ply of the
specimen was observed (see Figure 9-13 (b)). The longitudinal splits were seen to propagate
along the whole gauge length of the specimen. These longitudinal splits initiate and grow
through the width (see Figure 9-13 (c and d)) and thickness (see Figure 9-13 (e)) of the
composite resulting in the delamination of the inner plies. This makes the specimen act as
several samples gripped together at the end-tab regions. Therefore, the load transfer
between these longitudinal splits becomes limited.
When nearing the end of life (stage III), extensive longitudinal splitting resulting in
catastrophic fracture was observed. This leads to the classic brush-like failure of the
specimen (see Figure 9-13 (f)). In stage III it was also observed that the width of the
longitudinal split reduces. In addition, the catastrophic fracture of fibres at the very end of
the life emits tiny particles of fibres and matrix (see Figure 9-13 (f)).
183
Figure 9-13: The macroscopic damage mechanisms observed visually

during the tension-tension fatigue test of a unidirectional
composite at 787 MPa peak stress. a) Normalised life 0, b)
Normalised life 0.2, c) Normalised life 0.4, d) Normalised life 0.55,
e) Normalised life 0.8 and f) Normalised life 1.
184
Prestressed composite
Figure 9-14 and Figure 9-15 show the stiffness reduction of non-prestressed (0 MPa) and
prestressed composites during fatigue testing at peak stresses of 655 MPa and 1050 MPa,
respectively. Table 9-6 and Table 9-7 summarise the stiffness degradation at selected
normalised life of non-prestressed and prestressed composites. Appendix D presents the
stiffness degradation at selected normalised life for all the individual samples tested at 655
MPa and 1050 MPa peak stress. In general, prestressed composites showed a three-stage
behaviour for the stiffness reduction and surface temperature rise similar to that observed
for a non-prestressed composite. With reference to Figure 9-14, it can be observed that in
the low stress fatigue region (655 MPa), prestressed composites show an increase in
resistance to stiffness degradation during stage II of the fatigue life. This suggests that the
micro-crack initiation has been attenuated, and/or crack growth is reduced as a
consequence of reduction in the tensile residual stresses in the matrix. The stiffness
degradation in the low stress fatigue region correlates with the results from fatigue life.
On the other hand in high stress level fatigue (1050 MPa), prestressed composites show
rapid stiffness degradation in both stage I and II (see Figure 9-15). The decrease in the
resistance to stiffness degradation in prestressed composites at high stress level regions
could be attributed to the increase in tensile residual stresses in fibres with prestress.
The surface temperature measurements from the thermocouple at low and high stress
fatigue regions are presented in Figure 9-16 and Figure 9-17. In general, it can be seen that
the surface temperature change measured in a prestressed composite shows a three-stage
behaviour similar to a non-prestressed composite. It can be seen from Figure 9-16 and
Table 9-7 that in a low stress region the surface temperature rise in prestressed composites
during stage I of the fatigue life was slightly less when compared to non-prestressed
composite. In stage II of the fatigue life a slightly higher temperature rise was recorded in
prestressed composites. However, the temperature rise recorded from the prestressed
composite was lower than the non-prestressed composite in stage III (see Figure 9-16).
The surface temperature rise in a high stress region (1050 MPa) shows a rapid linear
increase with normalised fatigue life in non-prestressed and prestressed composites (see
Figure 9-17). It can be observed that the surface temperature rise in prestressed composites
is higher when compared to the non-prestressed composite. This could be due to
accelerated damage because of the presence of the greater magnitude of micro-residual
stresses in the interface and tensile residual stress in fibres.
185
800
0 MPa Prestress
700
51 MPa Prestress

Stiffness degradation (MPa)
500
400
300
200
100
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalised fatigue life (-)
Figure 9-14: Stiffness reduction of non-prestressed and prestressed

composites at 655 MPa peak stress.
240
0 MPa Prestressed
51 MPa Prestressed
200 108 MPa Prestressed
160
120
80
40
0
0 0.2 0.4 0.6 0.8 1 1.2
Normalised life (-)
Figure 9-15: Stiffness reduction of non-prestressed and prestressed

186
20
0 MPa Prestress
18
51 MPa Prestress
16
108 MPa Prestress
14
o
12
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Figure 9-16: Surface temperature rise of non-prestressed and prestressed

10
0 MPa Prestress
51 MPa Prestress
8
108 MPa Prestress
o
0
0 0.2 0.4 0.6 0.8 1 1.2
Figure 9-17: Surface temperature rise of non-prestressed and prestressed

187
Table 9-6: Summary of the stiffness degradation and surface temperature

changes for the non-prestressed and the prestressed
composites at a peak stress of 655 MPa in T-T fatigue.
Stiffness degradation (MPa) Temperature rise (°C)

Normalised
Prestress (MPa) Prestress (MPa)
life
0 51 108 0 51 108
0.1 110 80 80 3.7 2.1 2.8
0.3 170 100 90 3.6 3.3 4.5
0.5 200 110 90 4.4 5.0 5.3
0.7 680 110 90 4.7 5.7 6.2
0.9 710 120 100 5.6 6.2 6.0

composites at a peak stress of 1050 MPa in T-T fatigue.

Normalised
life
0 51 108 0 51 108
0.1 30 30 110 0.3 0.1 1.3
0.3 30 40 130 2.0 2.4 2.6
0.5 40 70 150 3.3 3.9 3.5
0.7 120 70 190 4.4 5.5 4.9
0.8 3110 1680 18150 4.7 5.3 6.3
9.1.5. Post-failure analysis

This section will concentrate on the effect of prestress on macroscopic and microscopic
damage in the composite materials investigated in this study.
Macroscopic post-fatigue (T-T) fracture

Figure 9-18 shows the macroscopic fatigue failure of non-prestressed and prestressed
composites at a 1050 and 524 MPa peak stress level respectively. At high stress levels, see
Figure 9-18 (a), it can be observed that both the non-prestressed and prestressed
composites show a brush-like failure. This type of macroscopic fatigue failure was also
observed by many researchers89,85,122,132,133,135.
It can be observed from Figure 9-18 (b) that the samples in low stress level tests (524 MPa)
show a significant reduction in the number of longitudinal splits. In general, it was
188
observed that the width and thickness of the longitudinal splits were greater at the low
stress level. In low stress level fatigue the density of the damage in prestressed composites
is less than in the non-prestressed composite (see Figure 9-18 (b)). The extent of
longitudinal splitting and delamination was significantly reduced in prestressed composites.
This could be attributed to the reduction in the tensile residual stresses in the matrix, which
could have increased the resistance to longitudinal split initiation and split growth. It can be
seen from Figure 9-18 (a) that in high stress level fatigue there is no difference between the
fracture of non-prestressed and prestressed composite.
Figure 9-18: Representative macroscopic fatigue failure of non-prestressed

and prestressed composites. a) 1050 MPa peak stress and b) 524
MPa peak stress.
Microscopic post-fatigue (T-T) fracture

Figure 9-19 to Figure 9-21 show the fracture surfaces of non-prestressed and prestressed
composites at higher stress levels. Features such as fibre pull-out (a), matrix hackle
formation (b), clean fibre surface (c), matrix plasticity (d) and fibre impression (e) can be
observed.
Figure 9-22 to Figure 9-24 show the low stress fatigue damage in non-prestressed and
prestressed composites. Two of the major features that characterise the fracture are hackle
formation and fibre impression.
The shear failure of the matrix between two fibres or laminates resulting from the inter-
fibre or inter-laminate shear stresses are referred to in the literature as hackles, serrations,
189
lacerations, chevrons and cusps137, 138, 139. A mechanism of hackle formation was proposed
by Purslow and this is explained here.
When a band of epoxy matrix is subjected to a shear stress, failure will commence as a
series of micro-tensile fractures in the plane of the shear band. This is generally observed to
be about 45° degrees as shown in Figure 9-25. With increasing shear stress the number of
tensile cracks will increase, and individually they will elongate (see Figure 9-25 (b)), and
curve over as they approach the limits of the shear band (Figure 9-25 (c)). Ultimate shear
failure occurs along the line of cracks when these tensile failures simultaneously coalesce,
see Figure 9-25 (d). Figure 9-26 and Figure 9-27 show examples of this fractography of
shear bands or hackle formation in non-prestressed composite.
From Figure 9-22 to Figure 9-24, it can be seen that the extent of hackle formation in low
stress level fatigue is greater than in the high stress level fatigue (see Figure 9-19 to Figure
9-21). This indicates that the low stress level fatigue involves more interlaminar shear
fracture and that the matrix dominates the damage. It was observed from the fractures in
the low stress level fatigue tests that the non-prestressed and prestressed composites
exhibit similar fracture features.
From the high stress level fatigue fracture (see Figure 9-19 to Figure 9-21) it was observed
that there is no difference in fracture morphology between non-prestressed and prestressed
composites.
Figure 9-19: T-T fatigue fracture of non-prestressed composite (0 MPa) at a

high stress level (1180 MPa). Features are marked as fibre pull-
out (a), matrix hackle formation (b), clean fibre surface (c) and
matrix plasticity (d).
190
Figure 9-20: T-T fatigue fracture of 51 MPa prestressed composite at a high

stress level (1180 MPa). Features are marked as fibre pull-out (a),
matrix hackle formation (b), clean fibre surface (c) and matrix
plasticity (d).
Figure 9-21: T-T fatigue fracture of 108 MPa prestressed composite at a high
clean fibre surface (c) and matrix plasticity (d).
191
Figure 9-22: T-T fatigue fracture of non-prestressed composite (0 MPa) at a

low stress level (590 MPa). Features are marked as fibre pull-out
(a), matrix hackle formation (b), matrix plasticity (d) and fibre
impression (e).
Figure 9-23: T-T fatigue fracture of 51 MPa prestressed composite at a low

matrix hackle formation (b) and fibre impression (e).
192
Figure 9-24: T-T fatigue fracture of 108 MPa prestressed composite at a low
matrix hackle formation (b) and fibre impression (e).
45°
d
Figure 9-25: Mechanism of matrix hackle formation .
193
Figure 9-26: Matrix hackle formation (b) in a non-prestressed composite.
Figure 9-27: A row of matrix hackle formation (b) and fibre pull-out (a) in a
non-prestressed composite.
194
9.2 Tension-compression fatigue
9.2.1. Fatigue of non-prestressed composite

In this study tension-compression (T-C) fatigue was conducted at a stress ratio (σmin/σmax)
R= -0.3 with a loading rate of 250 kN s-1. Fernando and Al-Khodari (GFAK) have
investigated the T-C fatigue in the E-glass/913 epoxy composite using the same fatigue
parameters used in this study. Figure 9-28 shows the normalised S-N curve for a non-
prestressed composite (current work) along with GFAK results. With reference to Figure
9-28, it can be observed that the slopes in the current work and GFAK data set are similar.
From equation 9-1 the loss in strength per decade of cycles was calculated for both the
current work and GFAK results and summarised in Table 9-8. It can be seen that the
calculated tensile strength losses per decade, for the current work and GFAK composites
are equal.
1
Current Work
GFAK
0.9
Linear (Current Work)
Linear (GFAK)
0.8
0.7
0.6
0.5
0.4
0.3
2 2.5 3 3.5 4 4.5 5 5.5 6 6.5
Log N
Figure 9-28: Comparison of normalised fatigue data of non-prestressed

composite (current work) with a similar system reported in the
literature (Fernando and Al-Khodari).
Table 9-8: Summary of slope (b) and tensile strength degradation per
decade (b/σUTS) for current work and GFAK composites.
Material Strength loss per decade

Slope (b)
(E-glass/913 epoxy) (b/σUTS) %
Current work 0.138 10.6
GFAK 0.129 10.6
195

Figure 9-29 presents the S-N curves for non-prestressed and prestressed composites along
with a linear-regression fit. Table 9-9 shows the individual fatigue life results of non-
prestressed and prestressed composites plotted in Figure 9-29. It can be seen that the
fatigue data shows a linear S-N behaviour. The values of the slope (b) from linear-
regression data analysis and tensile strength decay per decade of fatigue cycles (b/σUTS) are
summarised in
Table 9-10. From
Table 9-10, it can be observed that prestressed composites show a 1% reduction in the
strength decay during fatigue when compared to the non-prestressed composite.
1000
0 MPa Prestress
51 MPa Prestress
900
108 MPa Prestress

800
Peak Stress (MPa)

700
600

Tension-compression fatigue
500
Stress ratio R = -0.3
400
2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7
Log N
Figure 9-29: Comparison of tension-compression fatigue of the non-

prestressed and prestressed composites. A linear regression fit
was calculated.
196
Table 9-9: Summarises the individual fatigue life results of non-prestressed

and prestressed composites plotted in Figure 9-29. The arrow
mark (⇨) represents the sample is run-out.
0 MPa 51 MPa 108 MPa

Peak stress
MPa Number of Number of Number of
Log N Log N Log N
cycles cycles cycles
944.60 668 2.824 1,215 3.084 1,146 3.059
944.60 610 2.785 1,380 3.139 710 2.851
944.60 756 2.878 1,357 3.132 1,307 3.116
839.65 3,201 3.505 1,951 3.290 2,011 3.303
839.65 1,184 3.073 5,906 3.771 2,505 3.398
839.65 2,564 3.408 2,885 3.460 2,370 3.374
734.69 3,467 3.539 6,865 3.836 20,220 4.305
734.69 10,957 4.039 4,930 3.692 18,440 4.265
734.69 14,975 4.175 12,173 4.085 8,825 3.945
734.69 2,685 3.428 - - - -
629.73 102,690 5.011 109,200 5.038 89,590 4.952
629.73 109,811 5.040 148,440 5.171 124,759 5.096
629.73 94,640 4.976 103,920 5.016 150,000 5.176
524.78 314,556 5.497 1,053,835 ⇨ 6.022 1,063,872 ⇨ 6.026
524.78 285,032 5.454 815,660 5.911 856,060 5.932
524.78 322,500 5.508 1,064,911 ⇨ 6.027 872,800 5.940
524.78 - - 826,291 5.917
472.30 1,013,352 ⇨ 6.005
472.30 1,050,184 ⇨ 6.021
472.30 1,051,543 ⇨ 6.021
Table 9-10: A summary of slope (b) and tensile strength decay per decade
(b/σUTS) of non-prestressed and prestressed composites.
Prestress
Slope (b) (b/σUTS) %
(MPa)
0 -135.5 10.6
51 -121.86 9.6
108 -121.6 9.6
197
Figure 9-30 presents the comparison of T-T and T-C fatigue behaviour of non-prestressed
composite. With reference to Figure 9-30 it can be observed that the T-C fatigue
endurance of E-glass/913 epoxy composite is less when compared to the T-T fatigue
endurance. The S-N curves of T-T and T-C fatigue data merge after 105 cycles. This could
be because of the matrix fatigue limit. This shows that the T-C fatigue cycle of non-
prestressed composites is more sensitive to damage than the T-T fatigue cycle (as
expected).
1400
Fatigue T-T
1200 Fatigue T-C
Linear (Fatigue T-T)

1000
Peak Stress MPa
Linear (Fatigue T-C)

800
600
400
Tension-tension fatigue (R = +0.1)
200 Tension-compression fatigue (R = -0.3)
0
1.5 2.5 3.5 4.5 5.5 6.5
Log N
Figure 9-30: Comparison of T-T and T-C fatigue behaviour of non-prestressed

composite.
The T-C fatigue data of non-prestressed and prestressed composites was also plotted with
a polynomial fit (see Figure 9-31). From Figure 9-31 the calculated equivalent strength σequ
is summarised along with experimentally-measured UTS and UCS as shown in Table 9-11.
From the table it is interesting to observe that the calculated σequ is higher than the
experimentally-measured UTS and UCS. This is because the fatigue test was conducted at
250 kN s-1 whereas the static tensile tests were conducted at 0.6 kN s-1 and the static
compression test was conducted at 0.2 kN s-1. It is well known that the strength properties
of glass fibre/epoxy composites are loading-rate sensitive140 and this appears to be the
reason for the difference in calculated σequ and measured UTS and UCS.
198
1000
0 MPa Prestress
51 MPa Prestress
800 Polynomial fit (51 MPa Prestress)
Peak Stress (MPa)
700
600

500
400
2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7
Log N
Figure 9-31: Comparison of tension-compression fatigue of the non-

prestressed and prestressed composites. A second order fit was
calculated excluding the static strength (UTS and UCS).
Table 9-11: Summarises the calculated equivalent σequ from second-order

polynomials for the non-prestressed and prestressed composites
without static strength. The number in parentheses shows the
standard deviation.
Measured Measured
Prestress (MPa) σequ
σUTS σUCS
0 1311 (80) 1177 (47) 1436
51 1244 (110) 1294 (68) 1846
108 1285 (138) 1280 (67) 1598
9.2.3. Statistical analysis of fatigue data

The effect of prestress on T-C fatigue of composites is apparent when the peak stress is
normalised with respect to the UTS as shown in Figure 9-32. The standard deviation
presented in Figure 9-32 was calculated for a 95% confidence level.
199
0.8
0 MPa Prestress
51 MPa Prestress
0.7 108 MPa Prestress

Normalised Stress (σ max / σ UTS)
0.6 Polynomial fit (108 MPa Prestress)
0.5

0.3
2.5 3 3.5 4 4.5 5 5.5 6 6.5
Log N
Figure 9-32: Comparison of normalised tension-compression fatigue of the

non-prestressed and prestressed composites. The error bars were
calculated for a 95% confidence limit.
With reference to Figure 9-32, it can be observed that the prestressed composites show an
improvement in fatigue life in low and high stress fatigue regions. The individual fatigue
results from non-prestressed and prestressed composites in the low stress region (524 MPa
peak stress), the high stress region (944 MPa peak stress) and the middle stress regions (630
MPa peak stress) are presented in Figure 9-33, Figure 9-34 and Figure 9-35 respectively.
The arrow mark in Figure 9-33 shows that the specimen run-out (the test was stopped
because no macroscopic failure was observed in the specimen even after one million
cycles).
From Figure 9-33 and Figure 9-34, it can be observed that with an increase in prestress the
fatigue life of composites improved significantly. At 524 MPa peak stress, the improvement
in the average fatigue life for the 51 and 108 MPa prestressed composite, were 3.05 and
3.02 times than that of the 0 MPa composite. Also at 944 MPa peak stress, the average
fatigue life of the 51 MPa and 108 MPa prestressed composites was 1.94 and 1.50 times
that of the 0 MPa composite.
200
1200000
1000000
800000
Cycles to failure N
600000
400000
Peak stress = 524 MPa
200000
0
-20 0 20 40 60 80 100 120
Prestress (MPa)
region (peak stress of 524 MPa). The arrow mark represents the
specimen run-out.
1500
1250
1000
Cycles to failure N
750
500
250 Peak stress = 944 MPa
0
-20 0 20 40 60 80 100 120
Prestress (MPa)
Figure 9-34: Fatigue life as a function of applied prestress in the high stress
201
200000
160000
Cycles to failure N
120000
80000

40000
Peak stress = 630 MPa
0
-20 0 20 40 60 80 100 120
Prestress (MPa)
Figure 9-35: Fatigue life as a function of applied prestress in the middle stress
To check that the difference between the non-prestressed and prestressed composites was
significant, a student t-test was performed. It was found that at 1% significance level for
524 MPa peak stress level fatigue, the non-prestressed and prestressed composites are from
different populations. Also at a 944 MPa peak stress it was found that the 0 MPa and 51
MPa prestressed composites are from different populations with 1% significant level.
However, in the middle stress region it was observed that the fatigue life of the composite
was independent of prestress.
The two-parameter Weibull distribution was calculated for the fatigue results from non-
prestressed and prestressed composites at 524 MPa peak stress (low stress region) (see
Figure 9-36), 944 MPa peak stress (high stress region) (see Figure 9-37) and 630 MPa peak
stress (middle stress region) (see Figure 9-38). Table 9-12 summarises the Weibull modulus
calculated for non-prestressed and prestressed composites presented in Figure 9-36 to
Figure 9-38. From these Figures and the Table, it could be observed that the improvement
observed in prestressed composites at the 524 MPa peak stress level is statistically
significant. Also at a 944 MPa peak stress level the fatigue life of the 51 MPa prestressed
composite is greater than for the non-prestressed composite.
202
Table 9-12: Summarises the Weibull modulus of non-prestressed and

prestressed composites presented in Figure 9-36 to Figure 9-38.
Weibull modulus (m)

Prestress
MPa Peak stress Peak stress Peak stress
524 MPa 944 MPa 630 MPa
0 17.099 10.475 15.405
51 6.9981 15.83 5.2383
108 8.1859 3.4881 4.3862
1
0 MPa Prestress
51 MPa Prestress
0.5
108 MPa Prestress

Ln(-Ln(1-Pi))
-0.5 Linear (108 MPa Prestress)
-1
-1.5
-2
-2.5
12 12.4 12.8 13.2 13.6 14
Ln Xi (Cycles to failure)
Figure 9-36: Weibull distribution for fatigue life at 524 MPa peak stress (low
stress region).
203
1
0 MPa Prestress
51 MPa Prestress
0.5
108 MPa Prestress

0
Ln(-Ln(1-Pi))

-0.5
-1
-1.5
-2
6 6.3 6.6 6.9 7.2 7.5
Figure 9-37: Weibull distribution for fatigue life at 944 MPa peak stress (high
stress region).
1
0 MPa Prestress
51 MPa Prestress
0.5
108 MPa Prestress

Ln(-Ln(1-Pi))
-0.5 Linear (108 MPa Prestress)
-1
-1.5
-2
11 11.2 11.4 11.6 11.8 12
Figure 9-38: Weibull distribution for fatigue life at 630 MPa peak stress
(middle stress region).
204
Low stress fatigue region

From the results of the statistical analysis it was shown that prestressed composites show a
significant improvement in fatigue life compared to non-prestressed composites. The
possible reasons for the improvement are (i) reduction in matrix residual stresses and (ii)
improvement in the fibre alignment.
A reduction in process-induced residual strain in the matrix resulted in an increased strain
level to matrix cracking. This could increase the critical crack length to fracture. The
presence of fibre misalignment will induce a shear stress in the matrix during the fatigue
cycle, which will accelerate the matrix cracking and debonding. Adams and Hyer have
shown that in compression-compression fatigue, artificially-induced fibre waviness in
composites reduced the fatigue life when compared to the controlled composite. It was
shown in Section 6.3, Chapter 6 that fibre prestressing improved the number of fibres
aligned to the 0° degree of the composite. This improvement in fibre alignment in a
prestressed composite will reduced the shear stress-induced damage in the matrix and
interface. This in turn would be expected to improve the fatigue life of prestressed
composites.
High stress level region

At 944 MPa peak stress the 51 MPa prestressed composite showed an increase in fatigue
life compared to a non-prestressed composite. The possible reasons for the increase in
fatigue life are (i) an improvement in the fibre alignment and (ii) an increase in tensile
residual stress in the fibres.
It is well known that the composite failure stress in compression is less than in tension. It
has already been shown that fibre alignment in composites improved with fibre
prestressing (see Chapter 6). This will increase the compressive failure strength of the
composite (see Chapter 8). Also in Chapter 6 and 7 it was shown that the tensile residual
stress in fibres increases with fibre prestress. The resulting composite has tensile residual
stresses in fibres, which improves the compressive properties (as discussed in chapter 8).
This may have increased the fatigue life of the composites in the 944 MPa peak stress
region.
9.2.4. Fatigue damage monitoring

Figure 9-39 and Figure 9-40 present the stiffness degradation of prestressed composite and
non-prestressed composites during fatigue at 524 and 840 MPa peak stress respectively.
The corresponding surface temperature rises are given in Figure 9-41 and Figure 9-42. It
can be seen that both the stiffness degradation and surface temperature rise show a three-
stage behaviour, which was also observed in T-T fatigue. At 524 MPa peak stress it can be
observed that the stiffness degradation in the non-prestressed composite is rapid during
205
stage II of the fatigue life when compared to the prestressed composite. This also
correlates with the higher surface temperature recorded from the non-prestressed
composite, whilst for the prestressed composite the recorded surface temperature was
comparatively low.
25000
0 MPa Prestressed
51 MPa Prestressed
20000
108 MPa Prestressed
15000
10000
5000
0
0 0.2 0.4 0.6 0.8 1 1.2
Normalised Life (-)
Figure 9-39: Stiffness reduction in non-prestressed and prestressed

10000
0 MPa Prestress
9000
51 MPa Prestress
8000
108 MPa Prestress
7000
6000
5000
4000
3000
2000
1000
0
0 0.2 0.4 0.6 0.8 1 1.2
Normalised life (-)
Figure 9-40: Stiffness reduction in non-prestressed and prestressed

206
16
0 MPa Prestressed
51 MPa Prestressed
108 MPa Prestressed
12
Temperature Rise ( C)
o
0
0 0.2 0.4 0.6 0.8 1 1.2
Normalised Life (-)
Figure 9-41: Surface temperature rise in non-prestressed and prestressed

12
0 MPa Prestress
51 MPa Prestress
10
108 MPa Prestress
Temperature Rise ( C)
8
o
0
0 0.2 0.4 0.6 0.8 1 1.2
Normalised life (-)
Figure 9-42: Surface temperature rise in non-prestressed and prestressed

Also the surface temperature recorded in prestressed composites in the low stress fatigue
region shows a reduction and increase during the fatigue test (see Figure 9-41). This is
because each prestressed sample endured for two and a half days. During the nighttime the
207
laboratory temperature slightly decreases. This is the reason for the observed decrease in
temperature. On the other hand the non-prestressed specimen lasted only for 285, 032
cycles, which lasted only for 18 hours.
Table 9-13 and
Table 9-14 summarise the stiffness degradation and the corresponding surface temperature
rise for non-prestressed and prestressed composites at 524 and 840 MPa peak stress
fatigue. Appendix D presents the stiffness degradation and surface temperature rise at
selected normalised life for all the individual samples tested at 524 MPa and 840 MPa peak
stress.
At 840 MPa peak stress level the stiffness degradation during the stage I of the fatigue life
shows a rapid stiffness decay in non-prestressed composites. On the other hand the
stiffness decay in prestressed composites is low. In stage II of the fatigue life, the
prestressed composites showed relatively lower stiffness degradation. A marginally higher
surface temperature rise was recorded in non-prestressed composites compared to
prestressed ones (see Figure 9-42).
At a low stress level (524 MPa peak stress) the observed rapid stiffness degradation in non-
prestressed composites could be attributed to the tensile residual stresses in the matrix and
fibre waviness in composites. The process-induced residual tensile stresses in the matrix
lead to crack initiation and/or crack growth during the tensile phase of the fatigue cycle.
The presence of fibre waviness in composites will induce shear stresses in the matrix,
which could accelerate the damage in the matrix and fibre/matrix interface debonding. The
combination of tensile and compressive stress excursions is likely to hasten the damage
initiation and growth process. This in turn results in a rapid stiffness degradation and
surface temperature rise during the fatigue cycle. On the other hand in prestressed
composites, by inducing compressive stresses the matrix tensile residual stresses are
reduced. As a result the prestressed composite could have a higher resistance to matrix
cracking. Also by applying prestress the alignment of the fibres to the longitudinal
directional (0° degree) in composites is improved. This minimises the shear stress applied
to the matrix through fibre misalignment.
In high stress level (840 MPa peak stress) fatigue the prestressed composites showed a
greater resistance to stiffness degradation when compared to the non-prestressed
composites. In non-prestressed composites during processing, the fibres are subjected to
compressive residual stresses due to the cure-induced and thermal-induced residual
stresses. There is also the possibility of fibre movement due to matrix flow during the
curing process; this would be expected to increase the degree of fibre misalignment in non-
prestressed composites. As a result the compressive strength of non-prestressed
composites could be affected. However, in prestressed composites, the movement of fibres
is restricted at any stage of the curing cycle by the applied pre-load. In addition, with an
increase in prestress the fibres are subjected to tensile residual stresses.
208

composites at a peak stress of 524 MPa in T-C fatigue.

Normalised
life
0 51 108 0 51 108
0.1 2630 2190 1530 6.5 5.5 4.9
0.3 6980 2200 1740 9.9 3.4 2.6
0.5 11060 2200 1790 10.7 3.5 4.2
0.7 12270 2220 1960 10.7 2.2 4.7
0.9 15130 2810 1840 9.6 2.8 4.1
Table 9-14: Table 9-12: Summary of the stiffness degradation and surface
temperature changes for the non-prestressed and the
prestressed composites at a peak stress of 840 MPa in T-C
fatigue.

Normalised
life
0 51 108 0 51 108
0.1 1410 780 390 3.1 3.0 2.4
0.3 1540 1090 780 4.2 3.3 4.1
0.5 1950 1210 1260 5.3 4.3 4.6
0.7 2190 1270 1700 6.3 3.6 5.5
0.9 2220 1330 2330 7.3 3.6 4.8
9.2.5. Post-failure analysis

This section will concentrate on the effect of prestress on macroscopic and microscopic
damage of composite materials investigated in this study.
Macroscopic post-fatigue (T-C) fracture

Figure 9-43 (a) and (b) shows the macroscopic fatigue fractures of non-prestressed and
prestressed composites at high (945 MPa peak stress) and low stress (524 MPa peak stress)
level fatigue. The macroscopic T-C fatigue damage of composites at high stress level (945
MPa peak stress) shows a change in failure mode (see Figure 9-43) when compared to the
T-T fatigue (see Figure 9-18) due to the compressive stress excursion during fatigue
cycling. It is clear that the concentration of longitudinal splits and delamination are less in
209
T-C fatigue when compared to T-T fatigue at high stress levels. From Figure 9-43 (a) it can
be seen that in high stress level fatigue no significant difference in the damage was
observed between the non-prestressed and prestressed composites.
In low stress level fatigue (524 MPa peak stress), the extent of longitudinal splitting in
prestressed composites is less than in non-prestressed composites (see Figure 9-43 (b)).
This may be due to a reduction in matrix residual stresses and improvement in fibre
alignment.
Figure 9-43: Representative macroscopic fatigue failure of non-prestressed

and prestressed composites. a) 945 MPa peak stress and b) 524
MPa peak stress.
Microscopic post-fatigue (T-C) fracture

Figure 9-44 to Figure 9-50 show the microscopic failure of non-prestressed and prestressed
composites tested in T-C fatigue in the high stress region (945 MPa peak stress). In these
figures a number of failure mechanisms such as tensile failure (a), compressive failure (b),
matrix plasticity (c), fibre fracture (d), matrix hackle formation (e), fibre pull-out (f), fibre
buckling (g) and fibre imprint (h) can be seen. It was observed that the non-prestressed (see
Figure 9-44) and 51 MPa (see Figure 9-46) prestressed composites show a combination of
compressive and tensile fracture. However, the 108 MPa prestressed composites showed a
tensile failure (see Figure 9-48). The tensile failure is associated with fibre pullout from the
210
matrix, whereas, compressive failure is generally attributed to the shear failure of the fibres
and matrix. It can also be seen that in compression fracture, the matrix is holding the
fractured fibres in place. More matrix debris was observed in the compressive failure region
(see Figure 9-45). Tai et al.141 has also shown that during T-C fatigue of carbon fibre/PEEK
composite both tensile and compressive failures occurred.
The compressive failure in a non-prestressed composite appears very similar to the static
compression failure of non-prestressed composites discussed in Chapter 8. The
compressive failure of fibres is attributed to the fibre misalignment in composites. Piggott
has shown a mechanism of tension-tension fatigue failure for misaligned fibres. A
schematic of the mechanism is presented in Figure 9-51. When a tensile load is applied,
fibres alternatively straighten and relax. Because of this process, initially the interfacial
debonding occurs (see Figure 9-51 (b)) and this leads to matrix crack initiation. The cyclic
loading propagates these matrix cracks and ultimate failure will occur by fibre pull-out from
the matrix.
On the other hand in a tension-compression fatigue, the misaligned fibres are not only
straightened in the tensile cycle but they are also subjected to a micro-buckling/bending
process during the compression cycle. During the compression cycle, the matrix is
subjected to a transverse tensile stress due to the Poisson’s ratio difference between the
matrix and fibre, and stress concentrations caused by voids can initiate fracture in the
fibre/matrix interface and matrix. The presence of a misaligned fibre induces shear stresses
in the matrix, which also accelerates the crack growth within the matrix. If a fibre buckles,
fracture occurs at both the fibre/matrix interface and within the matrix and leads to
ultimate failure.
It is conjectured that failure in compression initiated by a misaligned fibre (see Figure 9-52
(e)) during the compression cycle will initiate catastrophic failure in the next tensile cycle by
fibre pull-out from the matrix (see Figure 9-52 (f)), if the load transferred to the adjacent
fibres is greater than their ultimate tensile strength. This phenomenon is observed in high
stress level fatigue. The difference in the failure mechanism between T-T and T-C fatigue
may explain why the composites show more sensitive behaviour to T-C fatigue than in the
T-T fatigue.
In Chapter 6 it was shown that prestressing improved the fibre alignment in composites
and that a maximum number of fibres aligned to 0° degree (~75%) was achieved at 108
MPa prestress. It can be seen that the extent of compressive failure in 51 MPa prestressed
composites is less when compared to non-prestressed composites. In 108 MPa prestressed
composites, predominantly tensile failure was observed (see Figure 9-48 and Figure 9-49).
In Figure 9-49, the radial fracture patterns indicate a tensile failure of the unidirectional
composite. A small amount of fibre buckling was also observed in 108 MPa prestressed
composites (see Figure 9-50).
211
Figure 9-44: T-C fatigue fracture of a non-prestressed composite (0 MPa) at

a high stress level (945 MPa peak stress). Features such as tensile
fracture (a) and compression fracture (b) can be seen.
Figure 9-45: Compressive failure of a non-prestressed composite at a high

stress level (945 MPa peak stress). Features such as fibre fracture
(d) and matrix plasticity (c) can be seen.
212
Figure 9-46: T-C fatigue fracture of a 51 MPa prestressed composite at a

high stress level (945 MPa peak stress). Features such as tensile
fracture (a) and compression fracture (b) can be seen.
Figure 9-47: A close-up image of the failure of a 51 MPa prestressed

composite shown in Figure 9-46. Features such as tensile fracture
(a) and compression fracture (b) can be seen.
213

high stress level (945 MPa peak stress). The area marked as a,
shows the tensile fracture.
Figure 9-49: Tensile failure of a 108 MPa prestressed composites at a high

stress level (945 MPa peak stress). The end face of the fractured
fibre (d) represents a tensile failure.
214
Figure 9-50: Tensile fibre pull-out (f), compressive fibre buckling (g) and fibre
impression (h) observed in a 108 MPa prestressed composites at
a high stress level (945 MPa peak stress).
Figure 9-53 to Figure 9-56 show the T-C fatigue fracture of non-prestressed and
prestressed composites tested at a low stress level (524 MPa peak stress). Features such as
matrix plasticity (c), fibre fracture (d), matrix hackle formation (e) and fibre imprints (h) can
be seen on the fracture surface. It can be seen that the extent of fibre fracture in non-
prestressed composites is high (see Figure 9-53 and Figure 9-56) when compared to
prestressed composites. The fractures in prestressed composites are predominantly matrix
and interface damage.
215
a b c d
Figure 9-51: Fatigue failure involving misaligned fibres under a tension and
compression cycle. a) fibre misalignment in composite, b)
fibre/matrix interface debonding, c) matrix crack initiation and
d) matrix crack propagation. The arrow mark shows the
direction of loading.
e f
Figure 9-52: Ultimate failure of composites by e) Compression fracture and

f) tensile fracture.
216
Figure 9-53: T-C fatigue fracture of a non-prestressed composite (0 MPa) at

a low stress level (524 MPa peak stress). Features such as fibre
fracture (d) and matrix hackle formation (e) can be observed.
Figure 9-54: T-C fatigue fracture of a 51 MPa prestressed composite at a low

stress level (524 MPa peak stress). Features such as fibre fracture
(d), matrix hackle formation (e) and fibre impression (h) can be
observed.
217

low stress level (524 MPa peak stress). Features such as matrix
plasticity (c), fibre fracture (d) and matrix hackle formation (e)
can be observed.
Figure 9-56: Compressive failure in a non-prestressed composite at a low

stress level (524 MPa peak stress). Features such as compressive
fibre fracture (b) and matrix plasticity (c) can be seen.
218
This chapter presented the results from tension-tension and tension-compression fatigue
tests of non-prestressed and prestressed composites. From the results presented the
9.3.1. Tension-tension fatigue

Normalised tension-tension fatigue data showed that the fatigue life of E-glass/913
composites at low stress levels was improved by prestressing. The fatigue life improvement
achieved at a 524 MPa peak stress level for the 51 MPa and 108 MPa prestressed
composites were 75% and 55% greater than that of the non-prestressed composites. In low
stress level fatigue the damage is concentrated in the matrix leading to matrix cracking and
interfacial debonding failure. The reduction in tensile residual stress in the matrix by
prestressing will delay the matrix damage, which will increase the fatigue endurance limit of
the composite. However, in the high stress level region no significant difference was
observed between the non-prestressed and prestressed composites. This is because in high
stress level fatigue the damage is dominated by the fibre properties of the composite. As
the fibres are subjected to tensile residual stresses by prestressing, the applied fatigue load
would be higher and this would reduce the fatigue life of the composites.
The stiffness degradation and surface temperature rise in non-prestressed and prestressed
composites showed a three-stage behaviour. It was found that the prestressed composites
were more effective in delaying the stiffness degradation in low stress level fatigue. The
surface temperature rise in non-prestressed composites was slightly higher than the
prestressed composites in stage I. This also suggests that the micro-damage in prestressed
composites could be delayed. The enhanced resistance to stiffness degradation could be
due to the reduction in residual stress in the matrix, which in turn reduces the micro-
damage in the matrix. Whereas, in high stress level fatigue the stiffness degradation in
prestressed composites was rapid and this could be due to the increase in tensile residual
stresses in fibres and micro-residual stresses at the interface. This in turn reduces the
resistance to interfacial crack initiation and crack growth of the composite. Also the surface
temperature measurements recorded a slightly higher surface temperature in prestressed
composites compared to non-prestressed composites.
From the macroscopic failure examination it is concluded that for low stress level fatigue
the extent of longitudinal splitting was significantly reduced in prestressed composites.
However, for high stress level fatigue both the non-prestressed and prestressed composite
showed a similar brush-like failure. From the microscopic failure analysis it was found that
in both low and high stress levels there is no difference in failure between non-prestressed
and prestressed composites.
219
9.3.2. Tension-compression fatigue

A significant improvement in the fatigue life of prestressed composites was observed in the
low stress level region of normalised fatigue data. The fatigue life improvements achieved
at a 524 MPa peak stress for the 51 MPa and 108 MPa prestressed composites were over
200% greater than that of the non-prestressed (0 MPa) composite. The reduction in tensile
residual stress in the matrix would delay the damage initiation and crack growth in the
matrix, which would have increased the fatigue life of composites at low stress levels. Also
for high stress level fatigue, the 51 MPa and 108 MPa prestressed composites showed an
improvement in fatigue life of 94% and 50% respectively when compared with the non-
prestressed composite. The improvement in fatigue life could be due to increased tensile
residual stresses in fibres, which would have reduced the compressive damage in
composites. However, at intermediate stress levels there appeared to be no influence of
prestress on fatigue life. This could be due to the dominance of fibre, matrix and interfacial
properties acting together.
The stiffness degradation in prestressed composites was significantly less when compared
to non-prestressed composites in both low and high stress level fatigue cycles. The reason
for the delay in stiffness degradation in low stress level fatigue could be due to the
minimisation of tensile residual stresses in the matrix and improvement in the fibre
alignment. This will increase the resistance to matrix and interfacial fatigue damage. Also in
high stress level fatigue, the increase in the tensile residual stress in the fibre will increase
the resistance to compressive damage of composites, which in turn would have reduced the
stiffness degradation. The measured surface temperature rise in non-prestressed
composites was greater than that in prestressed composites for low stress level fatigue. This
also correlates with the stiffness degradation results and also supports the concept that the
damage in prestressed composites is delayed when compared to non-prestressed
composites.
From the macroscopic failures it was found that, for low stress level fatigue the degree of
longitudinal splitting for prestressed composites was less than in non-prestressed
composites. This could be due to the reduction in residual tensile stresses in the matrix.
However, for high stress level fatigue, the degree of damage in non-prestressed and
prestressed composite was similar. The microscopic analysis of composites tested from
high stress level fatigue showed both compressive and tensile failures. However, the extent
of compressive failure in composites reduces with prestress. This could be attributed to the
improvement in fibre alignment. In low stress level fatigue region, the extent of fibre
damage in prestressed composites is less when compared to non-prestressed composites.
In prestressed composites the damage is predominantly in the matrix. The reason for less
fibre fracture in prestressed composites could be due to the improvement in fibre
alignment and an increase in tensile residual stress in the fibres.
220
CHAPTER 10
Conclusions and directions for

future research
OVERVIEW
Conclusions
Future work
T
HE PRESENT STUDY DEMONSTRATED a flat-bed prestressing methodology to
prepare prestressed composites in an autoclave and to monitor the applied load
and residual strain development online during the manufacturing process. The
residual stress development, quasi-static tensile and compression, and fatigue properties
(both tension-tension and tension-compression) of non-prestressed and prestressed
unidirectional E-glass/913 epoxy composites were investigated in this research. This
chapter presents the overall conclusions and contributions from the investigation.
10.1 Fibre prestress methodology
A new approach to fibre prestressing was presented using a flat-bed prestress methodology
for manufacturing prestressed composites. The FBPM was designed to apply a
unidirectional prestress to the fibres. From the design calculations, it was shown that the
rig is capable of applying 100 MPa prestress to the fibres. The FBPM was designed to
monitor and measure the applied load to the fibres via the load cell. The salient advantages
of flat-bed prestress methodology are 1) one parameter to control the applied load (load
CHAPTER 10 Conclusions and directions for future research
screw), 2) ability to monitor the applied load online, 3) enables conventional vacuum
bagging and 4) allows processing prepreg-based composites in an autoclave.
10.2 Evaluation of prestress methodology and sensors
Prestress rig evaluation:

From the static tensile prestress rig evaluation test, it can be concluded that the strain
distribution in the steel plate is uniform. A variation of 4% strain through the width of the
panel was observed. This could be due to (i) the Poisson’s effect, (ii) small misalignment in
the ERSG sensors and (iii) measurement errors.
A linear relationship was found between the strain measured from the steel plate and the
applied load using the load screw in the prestress rig. This is because the load applied is
within the elastic region of the steel plate.
Prepreg end-tab jig design and evaluation:

A flat-bed end-tab jig was designed to circumvent the fibre misalignment near the end-tab
region. Using this end-tab jig it was shown that fibre misalignment near the end-tab region
was eliminated.
Results from the surface profile of the composite shows that the prestressed composite
manufactured with the aid of an end-tab jig eliminated the bending of the composite panel.
Sensor evaluation:
A linear relationship was observed in the load cell with an increase in temperature. A
change in load of about 0.8 kN was measured for a temperature rise of 94°C. This small
change in load observed is negligible and shows that the load cell is temperature
compensated.
The measured thermal expansions in the unidirectional composite (in the fibre direction)
from embedded optical fibre sensors are in good agreement with the theoretically-predicted
thermal expansion from the micro-mechanical model.
The temperature measured from the embedded FBG temperature sensors was in good
agreement with the temperature measured from the thermocouple in the autoclave.
The temperature and pressure gave no significant change in the measured strain from the
reference optical fibre sensors. This shows that the recorded residual strain development
from optical fibre sensors in composites was not influenced by the autoclave parameters
(temperature and pressure).
222
10.3 Residual strain development
Preliminary studies show that the embedded optical fibre sensor in the prepreg recorded a
compressive strain of about 80 µε for -830 milibars (-0.084 MPa) vacuum. However, when
the vacuum is released not all the compressive strain in the optical fibre was removed. This
is because prior to the solidification of the resin it is difficult to control the relative
orientation of the FBG sensor.
10.3 Residual strain development
Results from the residual strain measurement using the EFPI sensors in non-prestressed
and prestressed unidirectional [0º16] composites show that fibre prestressing methodology
can be used to reduce and control the residual strain development in composites. Also it
was shown that by applying 108 MPa prestress, a composite with approximately zero final
residual strain can be prepared.
Experimental residual strain results measured using EFPI and FBG strain sensors are
greater than those predicted from Tsai and Hahn, and Wagner models. This is because the
models predict only the thermal stresses and on a micro-mechanical scale. On the other
hand the experimental results are measured in macroscopic laminates which would include
the cure-induced and thermally-induced residual stresses, and the interaction of
neighbouring fibres on the magnitude of residual stress in the host fibre.
Experimentally, it was also shown that the measured strain release from EFPI, FBG and
ERSG sensors were in good agreement with each other. The difference in strain measured
through the width of the composite is conjectured that to be due to Poisson’s ratio and
edge effects.
10.4 Theoretical analysis of prestressed composites
A theoretical model was developed to predict the strain released to the composite for a
given applied fibre prestress. The theoretical prediction is in good agreement with the
experimental results measured from EFPI, FBG and ERSG sensors.
This model was further developed to predict the Poisson’s effect on the strain variation
through the width of a unidirectional panel during the strain release. A small variation in
strain distribution of about 1% across the width of the composite was predicted.
10.5 Fibre alignment
Fibre prestressing improves the alignment of fibres in composites. This improvement in

fibre alignment resulted in an increase from 20% of fibres oriented in the 0° direction in
non-prestressed composites to 75% of fibres oriented in the 0° direction for a 108 MPa
223
prestressed composite. Above this prestress level no further improvement in alignment of

the fibre was observed.
Tensile properties:
Fibre prestressed composites show an improvement in strain at deviation from linearity in
stress-strain curve. This suggests that prestressing could have increased the resistance to
the onset of damage (matrix cracking, fibre-matrix debonding and weak fibre failure). This
improvement in SDL was observed up to 108 MPa prestress, above which it reduces. This
suggests that an optimum prestress limit for the improvement in strain at deviation from
linearity exists.
A theoretical prediction for the effect of prestress on failure strain of composites was
presented. The theory predicts that the fracture strain of a composite decreases with an
increase in prestress. This is because the residual tensile stress in the fibre increases with
prestress. This reduces the fracture strain of the composite. However, from the
experimental results no significant change in the tensile fracture strain was observed. This
could be due to greater scatter in the measurements. It was also shown that the tensile
strength and modulus of composites are independent of fibre prestressing.
From the macroscopic examination it can be concluded that there is no change in the
composite failure mechanism with fibre prestress. However, from the microscopic
examination it was found that at higher prestress levels more fibre fragmentation was
observed. This could be due to (i) breaking the weak fibres during prestressing results in
predominantly strong fibres in the composite and (ii) an increase in tensile residual stress in
the fibre with prestress. This results in increased strain energy stored in the fibres.
Compressive properties:
Fibre prestressing shows an improvement in the average compressive strength and
modulus by 9%. However, above an optimum prestress limit (108 MPa for compressive
strength and 80 MPa for compressive modulus) there is an indication of reductions in
compression properties. This could be because of the increase in residual stresses at the
interface with fibre prestress, above the optimum limit the fibre-matrix interface is prone to
debonding, which could lead to a reduction in compressive properties.
Fibre prestressing shows a change in the compression damage mechanism. The non-
prestressed (0 MPa) composites fractured by fibre kink-band formation, whereas
prestressed composites fractured by extensive longitudinal splitting and matrix
deformation. This change in damage mechanism suggests that the improvement in fibre
224
10.7 Fatigue behaviour
alignment as a result of prestressing has changed the fibre kink-band failure mode to a
longitudinal splitting and crushing mode.
An unknown matrix failure mode was observed in prestressed composites in the form of
curling. This type of fracture mode has not been reported in the literature previously. The
possible mechanisms for matrix curling could be due to the relaxation of the residual stress
in the interface and matrix during fibre-matrix debonding.
10.7 Fatigue behaviour
Tension-tension fatigue:
Normalised tension-tension fatigue data showed that the fatigue life of E-glass/913
composites at low stress levels was improved by prestressing. The fatigue life improvement
achieved at a 524 MPa peak stress level for the 51 MPa and 108 MPa prestressed
composites were 75% and 55% greater than that of the non-prestressed composites. In low
stress level fatigue the damage is concentrated in the matrix leading to matrix cracking and
interfacial debonding failure. The reduction in tensile residual stress in the matrix by
prestressing could delay the matrix damage, which would have increased the fatigue
endurance limit of the composite.
In the high stress level region no significant difference was observed between the non-
prestressed and prestressed composites. This is because in high stress level fatigue the
damage is dominated by the fibre properties of the composite. As the fibres are subjected
to tensile residual stresses by prestressing, the applied fatigue load would be higher and this
would reduce the fatigue life of the composites. The stiffness degradation and surface
temperature rise in non-prestressed and prestressed composites showed a three-stage
behaviour. It was found that the prestressed composites were more effective in delaying
the stiffness degradation for low stress level fatigue. The surface temperature rise in non-
prestressed composites was slightly higher than that of prestressed composites in stage I.
This also suggests that the micro-damage in prestressed composites could be delayed. The
enhanced resistance to stiffness degradation could be due to the reduction in residual stress
in the matrix, which in turn reduces the micro-damage in the matrix.
In high stress level fatigue the stiffness degradation in prestressed composites was rapid
and this could be due to the increase in tensile residual stresses in fibres and micro-residual
stresses at the interface. This in turn reduces the resistance to interfacial crack initiation and
crack growth of the composite. Also the surface temperature measurements recorded a
slightly higher surface temperature in prestressed composites compared to non-prestressed
composites.
From the macroscopic failure examination it is concluded that for low stress level fatigue
the extent of longitudinal splitting was significantly reduced in prestressed composites.
225
However, at high stress level fatigue both the non-prestressed and prestressed composite
showed a similar brush-like failure. From the microscopic failure analysis it was found that
at both low and high stress levels there is no difference in failure between non-prestressed
and prestressed composites.
Tension-compression fatigue:
A significant improvement in the fatigue life of prestressed composites was observed in the
low stress level region of normalised fatigue data. This fatigue life improvements achieved
at a 524 MPa peak stress for the 51 MPa and 108 MPa prestressed composites were over
200% greater than that of the non-prestressed (0 MPa) composite. The reduction in tensile
residual stress in the matrix would delay the damage initiation and crack growth in the
matrix, which would have increased the fatigue life of composites at low stress levels.
Also at high stress level fatigue, the 51 MPa and 108 MPa prestressed composites showed
an improvement in fatigue life of 94% and 50% respectively when compared with the non-
prestressed composite. This improvement in fatigue life could be due to increased tensile
residual stresses in the fibres, which would have reduced the compressive damage in
composites. However, at the intermediate stress levels there appeared to be no influence of
prestress on fatigue life. This could be due to the dominance of fibre, matrix and interfacial
properties acting together.
The stiffness degradation in prestressed composites was significantly less when compared
to non-prestressed composites for both low and high stress level fatigue cycles. The reason
for the delay in stiffness degradation for low stress level fatigue could be due to the
minimisation of tensile residual stresses in the matrix and improvement in the fibre
alignment. This will increase the resistance to matrix and interfacial fatigue damage.
For high stress level fatigue, the increase in tensile residual stress in the fibre will increase
the resistance to compressive damage of composites, which in turn would have reduced the
stiffness degradation. The measured surface temperature rise in non-prestressed
composites was greater than that in prestressed composites for low stress level fatigue. This
also correlates with the stiffness degradation results and also supports the concept that the
damage in prestressed composites is delayed when compared to non-prestressed
composites.
From the macroscopic failures it was found that, for low stress level fatigue the degree of
longitudinal splitting for prestressed composites was less than in non-prestressed
composites. This could be due to the reduction in residual tensile stresses in the matrix.
However, for high stress level fatigue, the degree of damage in non-prestressed and
prestressed composite was similar. The microscopic analysis of composites tested from
high stress level fatigue showed both compression and tensile failure. However, the extent
of compressive failure in composites reduces with prestress. This could be attributed to the
improvement in fibre alignment. In low stress level fatigue region, the extent of fibre
226
10.8 Directions for future research
damage in prestressed composites is less when compared to non-prestressed composites.

In prestressed composites the damage is predominantly in the matrix. The reason for less
fibre fracture in prestressed composites could be due to the improvement in fibre
alignment and an increase in tensile residual stress in fibres.
10.8 Directions for future research
This research investigated the effect of fibre prestress on residual stress, fibre waviness,
static mechanical and fatigue properties in unidirectional composites. From the
investigations it was found that fibre prestressing enhances the static and fatigue
performance by reducing the residual stress and improving the fibre alignment in the
composites. The following are the suggestions for continuing the research.
• One of the important criteria for aerospace structures is that the structure should
have high fatigue resistance. This research demonstrated that fibre prestressing
improves the tension-tension and tension-compression fatigue performance of
unidirectional composites. Further research could investigate the effect of prestress
on fatigue performance of cross-ply and quasi-isotropic laminates with glass/epoxy,
carbon/epoxy and aramid /epoxy and also hybrid fibre composites.
• Prestress methodology could be designed to enable prestressing all the ply
directions of cross-ply and quasi-isotropic laminates.
• Theoretical study of cure-induced residual stresses and the interaction of
neighbouring fibres on the host fibre residual stress could be made. In addition, the
theoretical analysis presented for the unidirectional composites in this research can
be extended to a cross-ply and quasi-isotropic laminates to predict the effect of
prestress on residual stresses and mechanical properties.
• The reduction in matrix residual stresses in prestressed composites may improve
the fracture toughness properties. Therefore studies on the Mode I, II and III,
fatigue crack propagation (delamination, transverse cracking) of prestressed multi-
directional laminates could be investigated.
227
CHAPTER 11
References
1 D Hull, TW Clyne, “An introduction to composite materials” Cambridge solid state

science series, Second edition, ISBN 0-521-38855-4, 1996.
2 ME Tuttle, “A mechanical/thermal analysis of prestressed composite laminates”, Journal

of Composite Materials, 22, 1998, 780-792.
3 K Shulte, R Marissen, “Influence of artificial prestressing during curing of CFRP

laminates on interfibre transverse cracking”, Comoposite Science and Technology, 44,
1992, 361-367.
4 LT Drzal, MJ Rich, PF Lloyd, “Adhesion of graphite fibers to epoxy matrices: 1. The role
of fiber surface-treatment”, Journal of Adhesion, 16 (1), 1982, 1-30.
5 J Kalantar, LT Drzal, “The bonding mechanism of aramid fibres to epoxy matrices: Part1
A review of the literature”, Journal of Materials Science, 25, 1990, 4186-4193.
6 FC Campbell, “Manufacturing processes for advanced composites”, Elsevier Ltd., ISBN

1-85617-415-8, 2004.
7 S-Y Lee, GS Springer, “Effects of cure on the mechanical properties of composites”,

Journal of Composite Materials, 22, 1988, 15-29.
CHAPTER 11 References
8 D Kugler, TJ Moon, “Identification of the most significant processing parameters on the

development of fibre waviness in thin laminates”, Journal of Composite Materials”, 36
(12), 2002, 1451-1479.
9 G Twigg, A Poursartip, G Fernlund, “Tool-part interaction in composites processing.

Part I: experimental investigation and analytical model”, Composites Part A, 35, 2004,
121-133.
10 HT Hahn, “Effects of residual stresses in polymer matrix composites”, Journal of

Astronautical Science 32 (3), 1984, 253-267.
11 MR Piggott, “The effect of fibre waviness on the mechanical properties of

unidirectional fibre composites: A Review”, Composite Science and Technology, 53,
1995, 201-205.
12 MS Madhukar, MS Genidy, JD Russell, “A new method to reduce cure-induced stresses

in thermoset polymer composites, Part I: Test Method”, Journal of Composite
materials, 34 (22), 2000, 1882-1904.
13 JD Russell, “Cure shrinkage of thermoset composites”, SAMPE quarterly, January 1993,

28-33.
14 SR White, HT Hahn, “Mechanical property and residual stress development during cure
of a Gr/BMI composite”, Polymer Engineering Science, 30 (22), 1990, 1465-73.
15 FR Jones, “Laminates – residual thermal and related strains”, in Handbook of Polymer-

Fibre Composites, Editor: F R Jones, ISBN 0-582-06554-2, 1994, 254 - 260.
16 NJ Rendler, I Vigness, “Hole-drilling strain-gauge method of measuring residual

stresses”, Experimental Mechanics, 1966, 577-586.
17 RY Kim, HT Hahn, “Effect of curing stress on the first ply failure in composite
laminates” Journal of Composite Materials, 13, 1979, 2-16.
18 PG Ifju, X Niu, BC Kilday, SC Liu, SM Ettinger, “Residual stress measurement in

composites using the cure-referencing method”, Experimental Mechanics, 40, 2000,
22-30.
19 B Benedikt, M Kumosa, PK Predecki, L Kumosa, MG Castelli and JK Sutter, “An

analysis of residual thermal stresses in a unidirectional graphite/PMR-15 composite
based on X-ray diffraction measurements”, Composites Science and Technology,
61(14), 2001, 1977-1994.
230
20 X Yan, T Ohsawa, “Measurement of the internal local stress-distribution of composite-
materials by means of laser imaging methods”, Composites, 25 (6), 1994, 443-450.
21 Q Ma and DR Clarke, “Measurement of residual stresses in sapphire fiber composites

using optical fluorescence”, Acta Metallurgica et Materialia, 41(6), 1993, 1817-1823.
22 RJ Young, RJ Day, M Zakikhani, IM Robinson, “Fibre deformation and residual

thermal stresses in carbon fibre reinforced PEEK”, Composites Science and
Technology, 34, 1989, 243-258.
23 IM Daniel, T Liber, “Effect of laminate construction on residual stresses in

graphite/polyimide composites”, Experimental Mechanics, 17, 1977, 21-25.
24 Liu T, Wu M, Rao Y, Jackson DA, Fernando GF, “A multiplexed optical fibre-based

extrinsic Fabry-Perot sensor system for in-situ strain monitoring in composites”,
Smart Materials and Structures, 7 (4), 1998, 550-556.
25 CM Lawrence, DV Nelson, JR Spingarn, TE Bennett, “Measurement of process-

induced strains in composite materials using embedded fibre optic sensors”, SPIE,
2718, 1997, 60-68.
26 CM Lawrence, DV Nelson, TE Bennett, JR Spingarn, “An embedded fibre optic-sensor

method for determining residual stresses in fibre-reinforced composite materials”,
Journal of Intelligent Material Systems and Structures, 9, 1998, 788-799.
27 RM Measures, “Fibre optic sensing for composite smart structures”, Composites

Engineering,; 3 (7-8), 1993, 715-750.
28 V Bhatia, KA Murphy, RO Claus, TA Tran, JA Greene, “Recent developments in

optical-fibre-based extrinsic Fabry-Perot interferometric strain sensing technology”,
Smart Materials and Structures, 4, 1995, 246-251.
29 GF Fernando, DJ Webb, P Ferdinand, Guest Editors, “Optical-fibre sensors”, MRS

Bulletin, May 2002, 359-364.
30 B Degamber, GF Fernando, “Process monitoring of fibre-reinforced polymer

composites”, MRS Bulletin, 27 (5), 2002, 370-380.
31 C Doyle, A Martin, T Liu, M Wu, S Hayes, PA Crosby, GR Powell, D Brooks, GF

Fernando, “In-situ process and condition monitoring of advanced fibre-reinforced
composite materials using optical fibre sensors”, Smart Structures and Materials, 7,
1998, 145-158.
231
32 R Badcock, S Krishnamurthy, G.F Fernando, T Butler, R Chen and J Tetlow, “Health

monitoring of composite structures using a novel fibre optic acoustic emission
sensors”, The 11th European Conference on Composite Materials, 31 May - 3 June
2004 Rhodes , Greece.
33 JS Sirkis, “Optical fibre strain sensing in engineering mechanics”, Topics in applied

Physics, 77, 2000, 233-272.
34 A Othonos, K Kalli, “Fibre Bragg Gratings”, Artech House, Boston, London, ISBN:
0-89006-344-3, 1999.
35 FM Haran, JK Rew, PD Foote, “A strain-isolated fibre Bragg grating sensor for

temperature compensation of fibre Bragg grating strain sensors “, Measurement
Science and Technology, 9 (8), 1998, 1163-1166.
36 M Jinno, S Sakai, K Osaka, T Fukuda, “Smart autoclave processing of thermoset resin

matrix composites based on temperature and internal strain monitoring”, Advanced
Composite Materials, 12 (1), 2003, 57-72.
37 H Kang, D Kang, C Hong, C Kim, “Simultaneous monitoring of strain and temperature

during and after cure of unsymmetric composite laminate using fibre-optic sensors”,
Smart Materials and Structures, 12, 2003, 29-35.
38 SW.Tsai, H.T.Hahn, “Introduction to Composite Materials”, Technomic Publications

Co., Inc. Westport, CT, ISBN 0-87762-288-4, 1980, 402-407.
39 HD Wagner, “Residual stresses in microcomposites and macrocomposite”, Journal of

Adhesion, 52, 1995, 131-148.
40 JA Nairn, “Thermoelastic analsyis of residual stresses in unidirectional high-

performance composites”, Polymer Composites, 6 (2), 1985, 123-130.
41 Y Chen, Z Xia, F Ellyin, “Evolution of residual stresses induced during curing

processing using a viscoelastic micromechanical model”, Journal of Composite
Materials, 35 (6), 2001, 522-542.
42 Y Zhang, Z Xia, F Ellyin, “Evolution and influence of residual stresses/strains of fibre

reinforced laminates”, Composites Science and Technology, 64, 2004, 1613-1621.
43 PJ Joyce, D Kugler, TJ Moon, “A technique for characterising Process-Induced fibre

waviness in unidirectional composite laminates-using optical microscopy”, Journal of
Composite Materials, 31 (17), 1997, 1694-1727.
232
44 RH Knibbs, JB Morris, “The effect of fibre orientation on the physical properties of
composites”, Composites, 5, 1974, 209-218.
45 AM Mrse, MR Piggott, “Compressive properties of unidirectional carbon fibre

laminates: II. The effects of unintentional and intentional fibre misalignment”,
Composites Science and Technology, 46, 1993, 219-227.
46 HM Hsiao, IM Daniel, “Effect of fibre waviness on stiffness and strength reduction of

unidirectional composites under compressive loading” Composites Science and
Technology, 56, 1996, 581-593.
47 DO Adams, MW Hyer, “Effect of layer waviness on the compression fatigue

performance of thermoplastic composite laminates”, Fatigue, 16, 1994, 385-391.
48 HJ Chun, JY Shin, IM Daniel, “Effect of material and geometric nonlinearities on the

tensile and compressive behaviour of composite materials with fibre waviness”,
49 SW Yurgartis, “Techniques for the quantification of composite mesostructure”,

50 AR Clarke, G Archenhold, NC Davidson, “A novel technique for determining the 3D

spatial distribution of glass fibres in polymer composites”, Composite Science and
Technology, 55, 1995, 75-91.
51 SR White, HT Hahn, “Cure cycle optimisation for the reduction of processing-induced

residual stresses in composite materials”, Journal of Composite Materials, 27 (14),
1993, 1352-1378.
52 MS Geindy, MS Madhukar, JD Russell, “A new method to reduce cure-induced stresses

in thermoset polymer composites, Part II: Closed loop feedback control system”,
Journal of Composite Materials, 34, (22), 2000, 1905-1925.
53 JD Russell, MS Madhukar, MS Genidy, AY Lee, “A new method to reduce cure-induced

stresses in thermoset polymer composites, Part III: Correlating stress history to
viscosity, degree of cure and cure shrinkage” Journal of Composite Materials, 34 (22)
2000, 1926-1947.
54 AK Gopal, S Adali, VE Verijenko, “Optimal temperature profiles for minimum residual

stress in the cure process of polymer composites”, Composite Structures, 48, 2000, 99
– 106.
55 SP systems Ltd, SE 84LV product data sheet, 2005.
233
56 A Singh, “Radiation processing of carbon fibre-reinforced advanced composites”,

Nuclear Instruments and Methods in Physics Research B, 185, 2001, 50-54.
57 S Korenev, “Electron beam curing of composites” Vacuum 62, 2001, 233-236.
58 DL Goodman, GR Palmese, “Curing and bonding of composites using electron beam

processing”, Handbook of Polymer Blends and Composites, Volume 1, A
Kulshreshtha and C Vasile, eds., Published by Rapra Technology Ltd, ISBN 1-85957-
249-9, 2003, UK.
59 FR Jones, “Interphase formation and control in fibre composite materials”, Key

Engineering Materials, 116-117, 1996, 41-60.
60 LE Asp, LA Berglund, R Talreja, “Effects of fibre and interphase on matrix-initiated

transverse failure in polymer composites”, Composites Science and Technology, 56,
1996, 657-665.
61 DL Hiemstra, NR Sottos, “Thermally induced interfacial microcracking in polymer

matrix composites”, Journal of Composite Materials, 27(10), 1993, 1030-1051.
62 NR Sottos, DL Hiemstra, WR Scott, “Correlating interphase glass transition and

interfacial microcracking in polymer composites”, Fracture Mechanics, 25, ASTM STP
1220, F Erdogan and RJ Hartranft, Eds., 1994.
63 AL Roytburd, J Slutsker, M Wuttig, “Smart composites with shape memory alloys”,

Comprehensive Composite Materials, Volume 5, Published by Elsevier Science Ltd.,
ISBN 0-08-042993-9, 2000, 507-522.
64 GC Psarras, J Parthenios, C Galiotis, “Adaptive composites incorporating shape

memory alloy wires, Part I: Probing the internal stress and temperature distributions
with laser Raman sensor”, Journal of Materials Science, 36, 2001, 535-546.
65 J Parthenious, GC Psarras, C Galiotis, “Adaptive composites incorporating shape

memory alloy wires, Part II: development of internal recovery stresses as a function of
activation temperature”, Composites Part A, 32, 2001, 1735-1747.
66 LS Penn, RCT Chou, ASD Wang, WK Binienda, “The effect of matrix shrinkage on
damage accumulation in composites”, Journal of Composite Materials, 23, 1989, 570-
586.
67 M Shimbo, MOchi, T Inamura, M Inoue, “Internal stress of epoxide resin modified

with spiro ortho-ester type resin”, Journal of Materials Science, 20, 1985, 2965-2972.
234
68 J Orso, AJ Vizzini, “The effects of an expanding monomer on the tensile properties of
graphite / epoxy”, Journal of Composites Technology and Research, 16 (3), 1994, 270-
274.
69 GJ Dvorak, AP Suvorov, “The effect of fibre pre-stress on residual stresses and the
onset of damage in symmetric laminates”, Composites Science and Technology, 60,
2000, 1129-1139.
70 The Oxford English dictionary, Second Edition, prepared by JA Simpson and ESC
Weiner, Volume XII, Clarendon Press, Oxford, 1989.
71 LDA Jorge, AT Marques, PMST De Castro, “The influence of prestressing on the

mechanical behaviour of uni-directional composites”, Developments in the science
and Technology of the composite materials, Elsevier Applied Science, London and
New York, 1990, 897 – 902.
72 DH Rose, JM Whitney, “Effect of prestressed fibres upon the response of composite

materials”, Proceedings of American Society of Composites, 8th Technical conference
on composites, Lancaster, PA, 1993, 489-498.
73 GX Sui, G Yao, BL Zhao, “Influence of artificial pre-stressing during the curing of

VIRALL on its mechanical properties”, Composite Science and Technology, 53, 1995,
361-364.
74 ME Tuttle, RT Koehler and D Keren, “Controlled thermal stresses in composites by

means of fibre prestress”, Journal of Composite Materials, 30(4), 1996, 486-502.
75 S Motahhari, J Cameron, “Measurement of micro-residual stresses in fibre pre-stressed

composites”, Journal of Reinforced Plastics and Composites, 16 (12), 1997, 1129-1137.
76 S Motahhari, J Cameron, “Impact strength of fibre pre-stressed composites”, Journal of

Reinforced Plastics and Composites, 17 (2), 1998, 123-130.
77 J Zhao, J Cameron, “Polypropylene matrix composites reinforced with pre-stressed

glass fibres”, Polymer Composites, 19 (3), 1998, 218-224.
78 AS Hadi and JN Ashton, “On the influence of pre-stress on the mechanical properties
of a unidirectional GRE composite”, Composite Structures, 40(3-4), 1998, 305-311.
79 S Motahhari, J Cameron, “Fibre prestressed composites: improvement of flexural

properties through fibre prestressing”, Journal of Reinforced Plastics and Composites,
18 (3), 1999, 279-288.
235
80 MP Jevons, GF Fernando, and GS Kalsi, “Effect of pre-tensioning on the low velocity

impact performance of glass fibre composites”, Tenth European conference on
composite materials (ECCM-10), June 3-7, 2002, Brugge, Belgium.
81 DAL Lee, “The effects of residual stresses in unidirectional Kevlar 49/epoxy composite
upon their mechanical properties”, MSc Dissertation, University of Manchester and
Institute of Science and Technology, 1992.
82 N Ali, “The effects of residual stresses in unidirectional Twaron/epoxy composite upon

their mechanical properties”, MSc Dissertation, University of Manchester and Institute
of Science and Technology, 1993.
83 YK Huang, PH Frings, E Hennes, “Mechanical properties of Zylon/epoxy composite”,

Composite: Part B, 33, 2002, 109-115.
84 KS Fancey, “Prestressed polymeric composites produced by viscoelastically strained

nylon 6,6 fibre reinforcement”, Journal of Reinforced Plastics and Composites, 19
(15), 2000, 1251-1266.
85 GF Fernando, FAA Al-Khodairi, “The fatigue of hybrid composites”, Chapter 7 in

Fatigue in Composites, Edited by Bryan Harris, Published by Woodhead Publishing
Limited, 2003, 189-241, ISBN 1 85573608 X.
86 JF Mandell, “Fatigue behaviour of fibre-resin composites”, Developments in reinforced

plastics-2, edited by G Pritchard, Applied Science Publishers, 1982, 67-107.
87 IM Daniel and A Charewicz, “Fatigue damage mechanisms and residual properties of

graphite/epoxy laminates”, Engineering Fracture Mechanics, 25 (5-6), 1986, 793-808.
88 KL Reifsnider, “ Damage and damage mechanics”, in Fatigue of composites materials,

Composite Materials Series, Vol. 4, Edited by KL Reifsnider, Elsevier Science
publications, 1991.
89 PT Curtis, “A review of the fatigue of composite materials”, Technical Report No.

87031, May 1987.
90 JW Dally, LJ Broutman, “ Frequency effects on the fatigue of glass reinforced plastics”,

Journal of Composite Materials, 1, 1967, 424-442.
91 FR Jones, “ Moisture absorption – anomalous effects”, in Handbook of Polymer-Fibre

Composites, Editor: FR Jones, ISBN 0-582-06554-2, 1994, 254 - 260.
236
92 B Harris, “Fatigue – glass fibre reinforced plastics”, in Handbook of polymer-Fibre
Composites, Editor: FR Jones, ISBN 0-582-06554-2, 1994, 309 - 316.
93 B Harris, “Fatigue and accumulation of damage in reinforced plastics”, Composites, 8

(4), 1977, 214-220.
94 GX Sui, G Yao, BL Zhou “Effects of prestrain on the fatigue properties of VIRALL

laminate”, Composite science and technology, 56 (8), 1996, 929-932.
95 CT Lin, PW Kao, FS Yang, “Fatigue behaviour of carbon fibre-reinforced aluminium

laminates”, Composites, 22 (2), 1991, 135 – 141.
96 BW Rosen, “Tensile failure of fibrous composites”, AIAA Journal, 2 (11), 1964, 1985-
1991.
97 M Fuwa, AR Bunsell, B Harris, “Tensile failure mechanisms in carbon fibre reinforced

plastics”, Journal of Materials Science, 10, 1975, 2062-2070.
98 HT Hahn, JG Williams, “ Compression failure mechanisms in unidirectional

composites” Composite Materials: Testing and Design, ASTM STP 893, edited by JM
Whitney, American Society for Testing and Materials, Philadelphia, 1986, 115-139.
99 B Budiansky, NA Fleck, “Compressive failure of fibre composites”, Journal of

Mechanics and Physics of Solids, 41 (1), 1993, 183-211.
100 PM Jelf, NA Fleck, “Compression failure mechanisms in unidirectional composites”,

Journal of Composite Materials, 26 (18), 1992, 2706-2726.
101 MR Piggot, B Harris, “Compression strength of carbon, glass and Kevlar-49 fibre
reinforced polyester resins” Journal of Material Science, 15, 1980, 2523-2538.
102 SH Lee, AM Wass, “Compressive response and failure of fibre reinforced

unidirectional composites”, International Journal of Fracture, 100, 1999, 275-306.
103 P Berbinau, C Soutis, IA Guz, “Compressive failure of 0° unidirectional carbon-fibre-

reinforced plastic (CFRP) laminates by fibre microbuckling”, Composite Science and
Technology, 59, 1999, 1451-1455.
104 CR Schultheisz, AM Waas, “Compressive failure of composites, Part I: Testing and

micromechanical theories”, Progress in Aerospace Science, 32, 1996, 1-42.
105 HM Jensen, “Residual stress effects on the compressive strength of unidirectional fibre
composites”, Acta Materialia, 50, 2002, 2895-2904.
237
106 MS Madhukar, LT Drzal, “Fibre-matrix adhesion and its effect on composite

mechanical properties. III Longitudinal compressive properties of graphite/epoxy
composites”, Journal of Composite Materials, 26, 1992, 310-333.
107 R Talreja, “Fatigue of composite materials”, Technomic Publishing Company, Inc,

ISBN 87762-516-6, 1987.
108 FP Beer, ER Johnston, “Mechanics of Materials” McGraw-Hill, ISBN 0-07-004284-5,

1981.
109 JE Shigley, “Mechanical Engineering Design” McGraw-Hill, 1986.
110 PC Powell, “Engineering with fibre-polymer laminates”, Chapman & Hall, ISBN 0-
412-49620-8, 1994.
111 Hexcel prepreg selector guide, Publication number FTU064d, Published by Hexcel
Ltd, Cambridge, UK, 2004.
112 Hexcel composites product data sheet, Publication FTA054b, Oct 2002.
113 MH Geier, “Quality handbook for composite materials”, Published by Chapman and
Hall, London, UK, First edition, ISBN 0412431203, 1994, 169-171.
114 PT Curtis, “CRAG test methods for the measurement of the engineering properties of
fibre reinforced plastics”, Technical Report No.88012, Published by Royal Aerospace
Establishment, 1988.
115 JG Häberle, EW Godwin, “The Imperial college method for testing composite
materials in compression” The centre for composite materials, Technical Report No.
99/03, Imperial College, London, 1999.
116 ISO 13003, Fibre-reinforced plastics – Determination of fatigue properties under cyclic
loading conditions, 2003.
117 SW Yurgartis, “Measurement of small angle fibre misalignments in continuous fibre

composites”, Composite Science and Technology, 30, 1987, 279-293.
118 J Lange, S Toll, J.-AE Manson and A Hult, “Residual stress build-up in thermoset
films cured below their ultimate glass transition temperature”, Polymer, 38 (4), 1997,
809-815.
119 S Goutianos, T Peijs, C Galiotis, “Comparative assessment of stress transfer efficiency

in tension and compression”, Composites Part A, 33 (10), 2002, 1303-1309.
238
120 I Partridge, P Virlouvet, J Chubb,and P Curtis, “Effect of fibre volume on tensile
fatigue behaviour of unidirectional glass/epoxy composite” Third European
Conference on Composite Materials (ECCM 3), AR Bunsell ed., France, 1989, 451.
121 RF Dickson, G Fernando, T Adam, H Reiter, B Harris, “Fatigue behaviour of hybrid

composites Part 2: Carbon-glass hybrids”, Journal of Materials Science 24, 1989, 227-
233.
122 WR Broughton, MJ Lodeiro, “Cyclic fatigue of continuous unidirectional polymer

matrix composites”, NPL Report CMMT (A) 249, January 2000, published by
National Physical Laboratory, UK.
123 J Scherf, HD Wagner, “Interpretation of fibre fragmentation in carbon/epoxy single

fibre composites: Possible fibre pre-tension effects”, Polymer Engineering and
Science, 32 (4), 1992, 298-304.
124 FL Matthews, “Chapter 5 – Compression” in the book Mechanical testing of advanced

fibre composites, edited by Hodgkinson JM, Woodhead Publishing Ltd and CRC
Press, ISBN 1-85573-312-9, 2000.
125 J Lankford, “Compressive failure of fibre-reinforced composites: buckling, kinking and

the role of the interphase”, Journal of Materials Science, 30, 1995, 4343-4348.
126 B Harris, H Reiter, T Adam, RF Dickson and G Fernando, “Fatigue behaviour of

carbon-fibre reinforced plastics”, Composites, 21 (3), 1990, 232-242.
127 J Lee, B Harris, DP Almond and F Hammett, “Fibre composite fatigue-life

determination”, Composites: Part A, 28A, 1997, 5-15.
128 NH Tai, MC Yip, CM Tseng, “Influences of thermal cycling and low-energy impact on
the fatigue behaviour of carbon/PEEK laminates”, Composites: Part B, 30, 1999, 849-
865.
129 T Hobbiebrunken, B Fiedler, M Hojo, S Ochiai, K Schulte, “Microscopic yielding of

CF/epoxy composites and the effect on the formation of thermal residual stresses”,
130 RA Badcock, GF Fernando, “An intensity-based optical fibre sensor for fatigue
damage detection in advanced fibre-reinforced composites”, Smart Materials and
Structures, 4, 1995, 223–230.
131 HT Hahn and RY Kim, “Fatigue behaviour of composite laminate”, Journal of

Composite Materials, 10, 1976, 156-180.
239
132 PT Curtis, “The fatigue behaviour of fibrous composite materials”, Journal of Strain
Analysis, 24, (4), 1989, 235-244.
133 CKH Dharan, “Fatigue failure in graphite fibre and glass fibre-polymer composites”,
Journal of Materials Science, 10, 1975, 1665–1670.
134 PS Steif, “Stiffness reduction due to fibre breakage”, Journal of Composite Materials,
17, 1984, 153-172.
135 PT Curtis, “Tensile fatigue mechanisms in unidirectional polymer matrix composite

materials”, International Journal of Fatigue, 13, (5), 1991, 377–382.
136 MR Bhat and CRL Murthy, “Fatigue damage stages in unidirectional glass fibre/epoxy
composites: identification through acoustic emission technique”, International Journal
of Fatigue, 15 (5), 1993, 401-405.
137 R Richards-Frandsen, Y Naerheim, “Fracture morphology of graphite/epoxy

composites”, Journal of Composite Materials, 17, 1983, 105.
138 D Purslow, “Matrix fractography of fibre-reinforced epoxy composites”, Composites,

17 (4) 1986, 289-303.
139 WD Bascom, DJ Boll, DL Hunston, B Fuller, PJ Phillips, “Fractographic analysis of

interlaminar fracture”, ASTM STP 937, 1987, 131-149.
140 GD Sims, DG Gladman, “Effect of test conditions on the fatigue strength of a glass-
fabric laminate: Part A-frequency, Plastics and Rubber, May 1978, 41-48.
141 NH Tai, CCM Ma, SH Wu, “Fatigue behaviour of carbon fibre/PEEK laminate
composites”, Composites, 26, 1995, 551-559.
240
241
APPENDIX A
Bi-axial loading frame method
OVERVIEW
Jevons et al.80 method for manufacturing prestressed composites contains a
frame and four clamps as shown in Figure 2.16. The pre-load is applied to the
prepreg by using a tensile testing machine and locked by the bolts to the frame.
When the rig is removed from the tensile testing machine, pre-load applied to
the prepreg may change because of the frame bending.
Calculation of the change in the pre-load applied to the prepreg using classical
mechanics is presented.
A.1 Pre-load calculations
Figure A-1 shows dimensions of the prepreg and its cross-section (A-A). A ply of glass
fibre/epoxy prepreg has a thickness of 125 µm. An eight-ply laminate has a thickness of 1
× 10-3 m. The volume fraction of fibres is assumed to be Vf = 60%.
Cross-sectional areas (CSA) of the fibres are expressed as follows:
Prepreg CSA: APrepreg = (250 × 10-3) × (1 × 10-3) = 250 × 10-6 m2
Fibre CSA: Af = (250 × 10-6) × (0.60) = 150 × 10-6 m2

To obtain 100 MPa stress in the fibres, the required force to apply is calculated as shown
below:
Appendix A Bi-axial loading frame method
F
σ= (A-1)
Af
F = (150 × 10-6) × (100 × 106) = 15 kN

This force is used in the following calculations.
A-A
A A
280 mm 1 mm
250 mm
Figure A-1: A schematic diagram of prepreg and its cross-section with

dimensions.
A.2 Jevons et al. C-Channel frame
A schematic diagram of the frame is shown in Figure A-2. The frame has holes to
accommodate the bolts and clamps. However, in the following calculation, the holes are
neglected for simplification.
A.2.1. Beam deflection

A uniaxial loading of the fibre prepreg clamped in the frame is considered here. In this
case, the frame can be broken down into four elements: two beams, where the forces are
applied, and two columns as shown in the Figure A-3. The beam can be considered as a
simply supported beam. When the pre-load is locked by bolts, the force acting on the beam
is a two point load. This is represented in Figure A-4.
242
A.2 Jevons et al. C-Channel frame
A-A
37 mm
464 mm
75 mm
63 mm
A A
6 mm
390 mm
Figure A-2: A schematic representation of the frame.
Beam Locking bolts
Column
Clamp Prepreg
Figure A-3: Uniaxial loading in Jevons et al. prestress method.
243
Appendix A Bi-axial loading frame method
F F
R1 R2
390 mm
55 mm
Figure A-4: Force diagrams of the beam.
Deflection of the beam can be determined by the following expression:
cFL3
δ= (A-2)
EI
where c is the numerical coefficient (for a two point load c = 11/768), F is the force, L is
the span length, E is the modulus of elasticity (for steel E = 200 GPa) and I is the moment
of inertia of the beam.
The force F is chosen to be 15 kN because this is the load required to prestress the prepreg
to 100 MPa as calculated in the previous section.
For a uniaxial loading, beam and column deflections were calculated to find the total
displacement of the frame per unit force. The results are summarised in Table A-1.
Table A-1: Total displacement of the frame subjected to uniaxial loading.
C-Channel box section Two point load

[ × 10-9 m/N]
Beam Deflection δB 38.30
Column Displacement ∆δC 1.08
Total Displacement ∆δ
39.38
[∆δ = δB + ∆δC]
A.2.2. Bi-axial loading of frame

In the above calculations, only one-directional (uni-axial) loading was considered. For
prestressing cross-ply laminates, the load is applied in two directions, which deforms the
frame in the respective loading directions. The deflection of the frame will be twice the
deflection calculated for uni-axial loading.
244
APPENDIX B
Flat-bed prestress
methodology
OVERVIEW
The engineering drawings of flat-bed prestress methodology are presented.
The photograph shows the manufactured prestress rig.
The conventionally vacuum-bagged prepreg inside the autoclave is depicted.
Appendix B Flat-bed prestress methodology
Figure B-1: Two-dimensional drawing of flat-bed prestress methodology.
246
Figure B-2: Two dimensional drawing of load screw.
247
Appendix B Flat-bed prestress methodology
Figure B-3: Three-dimensional solid model of flat-bed prestress

methodology.
248
Figure B-4: Photograph of manufactured flat-bed prestress rig instrumented
with load cell.
Figure B-5: Photograph of conventionally vacuum-bagged prepreg

laminate clamped to the FBPM inside the autoclave.
249
APPENDIX C
Measurements of static
mechanical properties
OVERVIEW
The static tensile and compressive test results of individual samples from all
the prestress levels studied in this research are presened.
C.1 Static tensile properties
Table C-1: 0 MPa Prestressed composite
Specimen Tensile strength Failure strain Secant modulus Strain deviation

ID σ (MPa) ε (%) E (GPa) from linearity (%)
Ref1b_1 1175.788 3.292 39.2 1.441
Ref1b_4 1373.597 3.092 44.6 1.531
Ref1b_10 1252.387 3.252 45.4 -
Ref2_1 1372.174 3.281 40.2 1.556
Ref3_1 1347.807 3.025 45.4 1.445
Ref3_4 1349.983 3.182 44.8 1.502
Average 1311.956 3.187 43.3 1.495
Standard
80.308 0.108 2.6 0.051
deviation
Appendix C Measurements of static mechanical properties
Table C-2: 51 MPa prestressed composite.
Specimen Tensile strength σ Failure strain Secant modulus Strain deviation

ID (MPa) ε (%) E (GPa) from linearity %
UPT2_1 1461.293 3.047 46.9 1.851
UPT1_2 1201.067 3.101 37.5 1.744
UPT2_4 1162.757 2.933 39.7 1.760
UPT2_7 1245.276 2.818 44.6 1.689
UPT3_1 1220.184 3.196 40.7 1.734
UPT3_5 1177.680 3.027 40.5 1.609
Average 1244.710 3.050 41.6 1.731
Standard
110.112 0.157 3.4 0.067
deviation

UP2_7kN_1 1381.640 3.164 44.6 1.804
UP2_7kN_2 1305.861 3.245 42.2 1.802
UP2_7kN_3 1248.431 3.159 41.3 1.780
UP2_7kN_5 1215.297 3.018 42.6 1.827
UP2_7kN_6 1327.834 3.003 47.5 1.716
Average 1295.812 3.117 43.6 1.785
Standard
65.613 0.103 2.4 0.042
deviation
Table C-4:108 MPa prestressed composite

UP1_14kN_2 1117.044 2.933 39.0 2.241
UP1_14kN_4 1187.524 3.014 41.4 1.897
UP2_14kN_4 1131.685 2.854 42.0 1.803
UP2_14kN_7 1434.793 3.305 46.6 1.887
UP4_14kN_1 1354.885 3.235 43.6 1.802
UP4_14kN_4 1337.816 3.256 43.6 2.017
UP4_14kN_6 1436.600 3.135 46.0 1.816
252
Average 1285.763 3.104 43.2 1.923
Standard
138.025 0.173 2.6 0.159
deviation

UP1_24kN_2 1243.290 3.188 41.4 1.750
UP1_24kN_3 1228.880 3.103 40.6 1.833
UP2_24kN_1 1228.843 2.929 42.6 1.780
UP2_24kN_2 1373.165 2.902 43.2 1.785
UP2_24kN_3 1249.423 3.096 42.1 1.855
UP2_24kN_4 1223.303 3.052 42.6 1.729
UP2_24kN_5 1365.666 3.298 45.4 1.882
Average 1273.224 3.081 42.6 1.802
Standard
66.363 0.138 1.5 0.056
deviation
C.2 Static compressive properties
Table C-6: 0 MPa prestressed composite
Specimen Compression strength Modulus

ID σ (MPa)
E (MPa)
UR9_2 1171.907 41.0
UR9_3 1187.087 40.4
UR9_4 1163.510 41.3
UR9_5 1138.495
UR9_6 1265.391 40.3
UR9_7 1135.659
Average 1177.008 40.8
STDEV 47.569 0.5
253
Appendix C Measurements of static mechanical properties

ID σ (MPa)
E (MPa)
UPT7_1 1282.286 45.2
UPT7_2 1242.497 -
UPT7_3 1300.230 41.7
UPT7_4 1192.514 43.2
UPT7_5 1310.592 43.4
UPT7_6 1320.467 -
UPT7_7 1411.099 44.2
Average 1294.241 43.5
STDEV 68.087 1.3

ID σ (MPa)
E (MPa)
UP1_7kN_2 1313.541 44.8
UP1_7kN_4 1202.050 44.1
UP1_7kN_5 1226.011
UP1_7kN_6 1338.159 45.7
UP1_7kN_7 1386.543 44.3
Average 1293.261 44.7
STDEV 77.412 0.7
Specimen Compression Modulus

ID strength σ (MPa)
E (MPa)
UP8_14kN_2 1231.993 44.3
UP8_14kN_3 1377.195 43.7
UP8_14kN_5 1343.674 43.8
UP8_14kN_6 1222.486 44.2
UP8_14kN_7 1283.325 45.5
254
UP8_14kN_8 1219.676 43.3
Average 1279.725 44.1
STDEV 67.473 0.7
Specimen Compression Strength σ Modulus

ID (MPa)
E (MPa)
UP1_24kN_1 1163.199 42.6
UP1_24kN_2 1274.530 -
UP1_24kN_3 1130.041 43.6
UP1_24kN_4 1085.053 42.6
UP1_24kN_5 1149.451 44.1
Average 1160.455 43.2
STDEV 70.271 0.7
255
APPENDIX D
Fatigue test results
OVERVIEW
Individual sample results of stiffness degradation and surface temperature rise
measurements during tension-tension and tension-compression fatigue test are
presented.
Table D-1: Summary of the stiffness degradation for the non-prestressed and the prestressed composites at a peak stress of 655
MPa in T-T fatigue.

Normalised
0 MPa Prestress 51 MPa Prestress 108 MPa Prestress
life
UR4_5 UR3_6 UR2_5 UPT2_3 UPT8_3 UPT3_2 UP2_14kN_6 UP1_14kN_5 UP1_14kN_6
0.1 110 100 874 80 90 30 80 70 30
0.3 170 150 1128 100 80 40 90 90 50
0.5 200 200 1149 110 80 - 90 90 60
0.7 680 310 1245 110 90 - 90 80 200
0.9 710 3990 4464 120 130 - 100 100 210
Table D-2: Summary of the surface temperature changes for the non-prestressed and the prestressed composites at a peak
stress of 655 MPa in T-T fatigue.
Temperature rise (°C)

Normalised
0 MPa Prestress 51 MPa Prestress 108 MPa Prestress
life
0.1 3.7 4.3 4.2 2.1 3.81 2.66 2.8 3.64 3.63
0.3 3.6 4.49 4.4 3.3 4.59 3.85 4.5 4.09 -
0.5 4.4 5.13 4.7 5.0 4.81 - 5.3 3.53 -
0.7 4.7 6.63 5.1 5.7 4.93 - 6.2 3.55 -
0.9 5.6 7.56 6.3 6.2 6.81 - 6.0 3.04 -
Table D-3: Summary of the stiffness degradation for the non-prestressed and the prestressed composites at a peak stress of
1050 MPa in T-T fatigue.

Normalised 0 MPa Prestress 51 MPa Prestress 108 MPa Prestress
life
0.1 30 10 20 30 10 10 110 100 190
0.3 30 160 40 40 10 40 130 110 250
0.5 40 200 50 70 20 70 150 120 260
0.7 120 3400 60 70 30 1950 190 970 1080
0.9 3110 8300 2420 1680 2890 - 18150 2860 3890
stress of 1050 MPa in T-T fatigue.

life
0.1 0.3 0.5 1.0 0.1 0.54 0.49 1.3 1.0 -
0.3 2.0 2.0 2.0 2.4 1.2 2.41 2.6 2.0 -
0.5 3.3 4.5 2.5 3.9 2.46 3.96 3.5 3.4 -
0.7 4.4 5.5 3.5 5.5 3.06 5.46 4.9 4.6 -
0.9 4.7 6.0 4.0 5.3 4.07 6.36 6.3 5.9 -
259
Appendix D Fatigue test results
MPa in T-C fatigue.

life
0.1 2630 3700 8050 2190 190 220 1530 40 390
0.3 6980 20550 8560 2200 100 580 1740 30 890
0.5 11060 24250 8780 2200 130 780 1790 30 4420
0.7 12270 24580 9180 2220 230 800 1960 50 7020
0.9 15130 25090 9340 2810 110 570 2180 280 10730
stress of 524 MPa in T-C fatigue.

life
0.1 6.5 4.48 - 5.5 5.52 4.84 4.9 2.01 4.01
0.3 9.9 4.87 - 3.4 6.43 6.68 2.6 3.54 5.47
0.5 10.7 3.83 - 3.5 1.98 5.7 4.2 5.3 5.97
0.7 10.7 3.79 - 2.2 1.0 6.77 4.7 4.29 3.21
0.9 9.6 3.45 - 2.8 2.4 5.98 4.1 1.82 2.99
260
MPa in T-C fatigue.

life
0.1 1410 520 2210 780 150 280 390 210 -
0.3 1540 832 3520 1090 200 420 780 310 -
0.5 1950 1345 3590 1210 250 660 1260 460 -
0.7 2190 1671 3600 1270 280 1210 1700 1760 -
0.9 2220 3785 - 1330 280 1490 2330 - -
stress of 840 MPa in T-C fatigue.

life
0.1 3.1 - 4.43 3.0 0.72 1.83 2.4 0.5 2.2
0.3 4.2 - 6.01 3.3 1.16 4.33 4.1 1.99 4.61
0.5 5.3 - 7.09 4.3 2.82 5.82 4.6 2.87 5.41
0.7 6.3 - 8.27 3.6 2.77 5.68 5.5 3.48 6.13
0.9 7.3 - 7.53 3.6 3.66 6.03 4.8 4.38 6.55
261

2006 Krishnamurthy

Uploaded by

Copyright:

Available Formats

2006 Krishnamurthy

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2006 Krishnamurthy

Uploaded by

Copyright:

Available Formats

CRANFIELD UNIVERSITY

DEFENCE COLLEGE OF MANAGEMENT AND TECHNOLOGY

ENGINEERING SYSTEMS DEPARTMENT

PRESTRESSED ADVANCED FIBRE REINFORCED

Supervisors: Dr RA Badcock and Professor JG Hetherington

© Cranfield University 2006. All rights reserved. No part of this publication

CHAPTER 1 Introduction .......................................................... 1

1.1 RESEARCH MOTIVATION AND RATIONALE ......................................................................................... 1

1.2 AIMS AND OBJECTIVES ................................................................................................................... 2

1.4 WORK DISSEMINATIONS .................................................................................................................. 4

1.5 ORGANISATION OF THESIS ............................................................................................................... 5

CHAPTER 2 Literature review ................................................... 7

2.1 INTRODUCTION .............................................................................................................................. 7

2.2 PROCESSING OF COMPOSITES ......................................................................................................... 8

2.3 PROCESS-RELATED ISSUES .............................................................................................................. 10

2.4 RESIDUAL STRESSES ....................................................................................................................... 11

2.5 FIBRE WAVINESS ........................................................................................................................... 18

2.6.7. Fibre prestressing ..................................................................................................... 25

2.7 FIBRE PRESTRESSED COMPOSITES .................................................................................................... 26

2.8 THEORETICAL ANALYSIS OF FIBRE PRESTRESSING ............................................................................... 27

2.9 EVALUATION OF EXISTING FIBRE PRESTRESS METHODOLOGIES ............................................................ 31

2.10 STATIC MECHANICAL PROPERTIES ................................................................................................ 37

2.11 FATIGUE OF COMPOSITES............................................................................................................ 44

2.12 DAMAGE MECHANISMS ............................................................................................................. 49

2.13 SUMMARY ................................................................................................................................. 56

CHAPTER 3 Design and development of the prestress

3.1 DESIGN CRITERIA ......................................................................................................................... 57

3.2 FLAT-BED PRESTRESS METHODOLOGY ............................................................................................. 59

3.3 DESIGN CALCULATIONS ................................................................................................................ 62

3.4 CONCLUDING REMARKS............................................................................................................... 70

CHAPTER 4 Experimental procedure ...................................... 71

4.1 MATERIALS .................................................................................................................................. 71

4.2 SENSORS ..................................................................................................................................... 72

4.3 PREPARATION OF COMPOSITE ....................................................................................................... 72

4.4 EVALUATION OF PRESTRESS METHODOLOGY ................................................................................... 79

4.5 EVALUATION OF SENSORS ............................................................................................................. 81

4.6 RESIDUAL STRAIN MONITORING ...................................................................................................... 83

4.7 DIFFERENTIAL SCANNING CALORIMETRY ......................................................................................... 85

4.8 QUALITY CONTROL AND SPECIMEN PREPARATION ........................................................................... 85

4.9 MECHANICAL TESTING.................................................................................................................. 89

CHAPTER 5 Evaluation of prestress methodology and sensors 93

5.1 RESULTS AND DISCUSSIONS ON PRESTRESS RIG EVALUATION .............................................................. 93

5.1.1. Load cell calibration............................................................................................... 93

5.2 PREPREG END-TAB JIG DESIGN AND EVALUATION ............................................................................ 98

5.3 RESULTS AND DISCUSSIONS ON SENSOR EVALUATION ..................................................................... 102

5.4 CONCLUDING REMARKS............................................................................................................. 108

CHAPTER 6 Residual strain and fibre alignment in the

6.1 RESIDUAL STRAIN DEVELOPMENT IN NON-PRESTRESSED COMPOSITE ................................................. 112

6.2 RESIDUAL STRAIN DEVELOPMENT IN PRESTRESSED COMPOSITE .......................................................... 117

6.3 RESULTS OF FIBRE ALIGNMENT ...................................................................................................... 122

6.4 CONCLUDING REMARKS............................................................................................................. 126

CHAPTER 7 Theoretical analysis of prestressed composites .. 127

7.1 INTRODUCTION .......................................................................................................................... 127

7.2 COMPOSITE STRAIN .................................................................................................................... 131

7.3 RESULTS AND DISCUSSIONS .......................................................................................................... 134

7.4 CONCLUDING REMARKS............................................................................................................. 141

CHAPTER 8 Static mechanical properties ............................ 143

8.1 STATIC TENSILE RESULTS AND DISCUSSION ...................................................................................... 143

8.2 POST-TENSILE FAILURE ANALYSIS ................................................................................................... 149

8.4 COMPRESSIVE STRENGTH COMPARISON ....................................................................................... 158