Failure Analysis of Composite Top-hat-

stiffeners using Acoustic Emission and

Embedded Fibre Bragg Gratings

Raju

A dissertation presented for the degree of

Doctor of Philosophy

May 2011

School of Mechanical and Manufacturing Engineering

University of New South Wales, Sydney, NSW 2052, Australia
Originality Statement

I hereby declare that this submission is my own work and to the best of my knowledge it
contains no materials previously published or written by another person, or substantial
portions of material which have been accepted for the award of any other degree or diploma
at UNSW or any other educational institution, except where due acknowledgement is made in
the thesis. Any contribution made to the research by others, with whom I have worked at
UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual
content of this thesis is the product of my own work, except to the extent that assistance from
others in the project’s design and conception or in style, presentation and linguistic expression
is acknowledged.

Signed

Date

Page | i
Copyright Statement

I hereby grant the University of New South Wales or its agents the right to archive and to make
available my thesis or dissertation in whole or part in the University libraries in all forms of
media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain
all proprietary rights, such as patent rights. I also retain the right to use in future works (such
as articles or books) all or part of this thesis or dissertation.

I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation
Abstract International.

I have either used no substantial portions of copyright material in my thesis or I have obtained
permission to use copyright material; where permission has not been granted I have
applied/will apply for a partial restriction of the digital copy of my thesis or dissertation.

Signed

Date

Page | ii
Authenticity Statement

I certify that the Library deposit digital copy is a direct equivalent of the final officially
approved version of my thesis. No emendation of content has occurred and if there are any
minor variations in formatting, they are the result of the conversion to digital format.

Signed

Date

Page | iii
Dedicated to my parents for their eternal love and support

Smt. Ashwathalakshmi T.L. and late Sri Rathnaiah Setty V.

Page | iv
Abstract

Modern aerospace, automotive, marine, mechanical and civil structures rely on advanced
composites for their added benefits over conventional metallic structures. The complex
damage process in composites involves a variety of failure modes, such as matrix cracking,
fibre-matrix debonding, fibre fracture and delamination. In structural composites, the initiation
of a crack or fracture does not indicate a catastrophic failure. A stable crack propagation stage
associated with a steady increase in external load precedes catastrophic failure. Further, due
to the alignment of the fibres to the load path, some designs exhibit reserve strength after
crack formation, increasing until final collapse. The design and assessment of such structures
require adequate consideration of strength and fracture behaviour. An accurate prediction of
static collapse behaviour for marine structures is essential for assessing the reserve strength
and likelihood of damage. This thesis attempts to gain an insight into the complex stress
distribution and numerically model the secondary reserve strength of structures.

The issue of frequent keel failures of yachts made out of marine composites is a mystery to
many boat builders and designers and the past two decades have seen a number of casualties
and material loss. The main thrust of this study was to identify the key variables responsible
for these keel failures and to understand the out-of-plane load transfer in Top-hat-stiffeners
(THS). This research investigates the damage mechanisms of curved composite structures (L-
bends and THSs) and estimates the residual strength after the initial failure. This is vital in
estimating the time required to obtain assistance before the boat capsizes.

Experimental and numerical investigations are carried out to understand the out-of-plane load
transfer in the THS. Specimens were manufactured by hand-layup and vacuum-infused
process, and tested until final collapse. Acoustic-emission, embedded optical-fibre sensors and
surface bonded fibre Bragg grating optic sensor, to sense acoustic-emissions, are employed to
gain insight into failure initiation and progression. The key objective was to understand the
different failure modes in composite materials.

Page | v
Acoustic-emission is used to discriminate between the various failure modes using
piezoelectric sensors. Failure mode discrimination is performed by analysing the acoustic-
emission (signal) waveform and examining various parameters such as arrival time, peak
amplitude, rise-time, signal duration, counts or threshold crossings, elastic energy and RMS
from individual acoustic-emission signals. In addition, an attempt is made to investigate the
possibility of using a state of the art surfaced bonded optical fibre sensor as an acoustic sensor.
The performance of a strain-insensitive fibre Bragg grating sensor was compared with that of
piezoelectric sensors for composite structures. The feasibility of the system was demonstrated
in typical in-situ structural health monitoring applications, using acoustic-emission techniques.

Optical-fibre sensors provide benefits such as low specific weight, smaller size, and good
compatibility with polymeric composites. Fibre Bragg grating sensors are expected to provide a
replacement for traditional electric sensors as they are not susceptible to electromagnetic
signatures, provide multi-functionality and enable multiplexing. Optical-fibre sensors were
embedded in the composite specimen to measure the non-linear strain at the bend and to
analyse the failure progression with strain anomalies.

Non-linear finite element analysis (FEA) was conducted using MSC MARC to understand both
the normal and interlaminar stress distribution at the bends. Progressive delamination failure
was successfully modelled for all three types of layups using interface elements and good
agreement was achieved between the experiments and finite element analysis. The reserve
strength identified by experimental testing was successfully modelled. Factors affecting the
fibre-matrix interface strength were studied. The failure location and mode for fracture
initiation and propagation were determined.

Page | vi
Acknowledgements

I wish to express my greatest appreciation to my supervisor A/Prof. Gangadhara Prusty for his
guidance, encouragement, valuable time and support throughout the course of the thesis. I
would also like to thank Prof. Don Kelly, Mr. David Lyons and Prof. Gang Ding Peng, for their
guidance and scrutiny of the research program. This study could not be eventuated without
their supervision, direction and continued encouragement.

First and foremost I would like to acknowledge the financial support offered by the Faculty of
Engineering, UNSW through the Faculty of Engineering Award and Supplementary Engineering
Award.

Special thanks to Asrul Izam for his invaluable time and expertise with the photonics support
through the course of this PhD. My sincere appreciation goes to Jun Ikeda who conducted
preliminary study on Top-hat-stiffeners during his undergraduate thesis.

Thanks to the laboratory staff Jason Nhieu, Russell Overhall, Andrew Higley and many more
who have provided on time support with the manufacturing and testing of the specimens.
Many thanks to Lachlan Welch, James Diamond, Justine Bensley, Ian Linton and Anthony Brann
from EMP Composites, Sydney for their continuous technical support during the manufacture
of the test specimens.

Great appreciation goes to my colleagues who have helped along the way; Luke Djukic, Garth
Pearce, Jay Sul remembering all the fun, encouragement and support we had during the
course of this study.

I owe my thanks to Dr. Michael Bannister (CRC-ACS) and Prof. William R Walsh (Prince of Wales
Clinical School) for providing the Acoustic-emission (AE) Equipment for experimental testing of
the specimens.

I extend my thanks to the Faculty of Engineering for providing, state of the art Design Research
Lab for effective conduction of this thesis.

Page | vii
I wish to express my gratitude to A/Prof. NEA Ahmed and Mrs. Sharon Turnbull for their kind
assistance during the completion of MEngSc in Aerospace engineering. I can never forget their
invaluable support during the critical phase.

Last, and by no means least, special thanks go to my family and friends and in particular, my
wife Lakshmi dearly for the support and love she has given. Special thanks to my daughter
Spoorthi as well. Both of them have been supporting my unscheduled and untimely hours of
work, missing most of the weekends and supporting during the entire tenure of this research
while I had to spend a lot of mental energy away from my family.

Page | viii
Table of Contents

Originality Statement.............................................................................................................. i
Copyright Statement .............................................................................................................. ii
Authenticity Statement......................................................................................................... iii
Abstract ................................................................................................................................. v
Acknowledgements ............................................................................................................. vii
Table of Contents.................................................................................................................. ix
List of Tables ....................................................................................................................... xvi
List of Figures ....................................................................................................................... xx
Abbreviations and Acronyms ...........................................................................................xxviii
Symbols and Nomenclature ..............................................................................................xxxii

Chapter 1: Introduction.......................................................................................................... 1
1.1 Background Information .................................................................................................. 2
1.2 Aims and Objectives ......................................................................................................... 3
1.3 Thesis Outline ................................................................................................................... 5
1.4 Related Publications ......................................................................................................... 6
Edited Book Chapters ........................................................................................................ 6
Journal Papers ................................................................................................................... 6
International Conferences ................................................................................................. 7

Chapter 2: Literature Review ................................................................................................. 8

2.1 Introduction...................................................................................................................... 9
2.2 Curved Composites .......................................................................................................... 9
2.3 Out-of-plane Joints ......................................................................................................... 11
2.3.1 T- Joints ................................................................................................................. 12
2.4 Top-hat-stiffener ............................................................................................................ 12
2.4.1 Design Rules and Standards for Top-hat-stiffeners .............................................. 13
2.4.2. Past Studies on Top-hat-stiffeners ...................................................................... 16

Page | ix
2.5 Keels ............................................................................................................................... 18
2.5.1 Types of Keels ....................................................................................................... 19
2.5.2 Keel-hull Connection ............................................................................................ 20
2.5.3 Evolution of Keel Design ....................................................................................... 23
2.5.4. Major Keel-related Accidents .............................................................................. 25
2.5.5 Comments on Keel-hull Joints .............................................................................. 28
2.6 Composite Failure .......................................................................................................... 29
2.6.1 Matrix Failure ....................................................................................................... 30
2.6.2 Fibre Failure .......................................................................................................... 31
2.6.3 Interface Failure.................................................................................................... 33
2.6.4. Interlaminar Failure – Delamination ................................................................... 33
2.7 Failure Theories for Composite Structures .................................................................... 34
2.7.1 Limiting Value Theories ........................................................................................ 34
2.7.2 Interactive Theories .............................................................................................. 35
2.7.3 Failure Mode Theories .......................................................................................... 36
2.7.4 World-wide Failure Exercise (WWFE) Studies ...................................................... 39
2.8 Fracture-mechanics Approach to Composite Failures ................................................... 40
2.9 Delamination Prediction in Composites Using Numerical Methods .............................. 43
2.9.1 Stress-based Numerical Methods......................................................................... 43
2.9.2 Fracture-Mechanics-based Numerical Modelling ................................................ 45
2.9.3 Cohesive Zone or Interface Elements ................................................................... 46
2.9.4 Strain – Invariant Failure Theory (SIFT) ................................................................ 47
2.9.5 Summary of numerical methods for delamination failure ................................... 47
2.10 Non-destructive Composite Evaluation ....................................................................... 47
2.10.1 Types of NDE....................................................................................................... 48
2.11 Acoustic-Emission ......................................................................................................... 50
2.11.1 Acoustic-emission Analysis ................................................................................. 51
2.11.2 AE Sensors .......................................................................................................... 52
2.11.3 AE Data Analysis ................................................................................................. 54
2.11.4 Parametric or Feature Set Analysis .................................................................... 56
2.11.5 Transient or Modal Analysis or Waveform Digitisation...................................... 56
2.11.6 Summary of AE ................................................................................................... 57
2.12 Optical-fibre Sensing .................................................................................................... 57
Page | x
 2.12.1 Basic Principle of Optical fibre sensors ............................................................... 60
2.12.2 Optical-fibre Modulation Techniques................................................................. 61
2.12.3 Bragg Grating Sensors......................................................................................... 63
2.12.4 Multiplexing ........................................................................................................ 65
2.12.5 FBG Interrogation ............................................................................................... 66
2.13 Damage Characterisation of Advanced Composites Using Optical-fibre Sensors ....... 67
2.13.1 Matrix cracking ................................................................................................... 67
2.13.2 Debonding .......................................................................................................... 68
2.13.3 Crack Monitoring in a Bonded Repair System .................................................... 69
2.13.4 Detection of Transverse Cracks .......................................................................... 70
2.13.5 Delamination ...................................................................................................... 70
2.13.6 Effects of Embedded Optical Fibres on Structural Integrity ............................... 74
2.13.7 Summary of Optical-fibre Sensors ...................................................................... 75
2.14 Acoustic-Emission Measurement with Optical-fibre Sensors ...................................... 75
2.14.1 Interferometric Acoustic Sensors ....................................................................... 76
2.14.2 Intensity-type Acoustic Sensor ........................................................................... 79
2.15 Structural Changes ....................................................................................................... 81
2.16 Conclusions................................................................................................................... 82

Chapter 3: Material Characterisation of Marine Composites ................................................ 85

3.1 Introduction.................................................................................................................... 86
3.2 Marine Structures........................................................................................................... 86
3.2.1 Reinforcements .................................................................................................... 89
3.2.2 Matrix ................................................................................................................... 90
3.2.2.1 Polyester .............................................................................................................. 91
3.2.2.2 Vinylester............................................................................................................. 91
3.2.2.3 Epoxy ................................................................................................................... 91
3.3 Manufacturing Techniques in the Marine Industry ....................................................... 92
3.3.1 Wet Hand-Layup (HL) ........................................................................................... 92
3.3.2 Vacuum-bag Infusion Process (VIP) ...................................................................... 93
3.3.3 Use of Prepregs..................................................................................................... 94
3.3.4 Summary of Manufacturing Methods .................................................................. 94
3.4 Postcuring effect on composite structures .................................................................... 95
Page | xi
 3.5 Material Characterisation .............................................................................................. 95
3.5.1 Determination of Material Properties .................................................................. 96
3.5.2 Raw materials used............................................................................................... 97
3.5.3 Resin Burn-off Test ............................................................................................... 97
3.5.4 Coupon Tests ........................................................................................................ 97
3.5.5 Test Matrix ............................................................................................................ 98
3.5.6 Tensile Test Summary ........................................................................................... 99
3.5.7 Tensile test: Hand-Layup and VIP Manufacturing Process Comparison............. 101
3.5.8 Three-point Bend Test Summary ........................................................................ 104
3.6 Conclusions................................................................................................................... 107

Chapter 4: Experimental Investigation of curved composites ............................................. 108

4.1 Introduction.................................................................................................................. 109
4.2 Experimental Investigation of L-bend Laminates ......................................................... 109
4.2.1 Experimental Setup ............................................................................................ 109
4.2.2 L-Bend Experimental Results .............................................................................. 112
4.2.3 L-bend Layup 1 results ........................................................................................ 112
4.2.4 L-bend Layup 2 results ........................................................................................ 113
4.2.5 L-Bend Layup 3 Results ....................................................................................... 114
4.2.6 L-bend Test Summary ......................................................................................... 116
4.3 Experimental Investigation of Hand-laid Top-hat-stiffener ......................................... 117
4.3.1 Hand-laid Top-hat-stiffener Experimental test configuration ............................ 118
4.3.2 HL Top-hat-stiffener experimental results ......................................................... 120
4.3.3 HL Top-hat-stiffener Layup 1 experimental results ............................................ 120
4.3.4 HL Top-hat-stiffener Layup 2 experimental results ............................................ 122
4.3.5 HL Top-hat-stiffener Layup 3 experimental results ............................................ 124
4.3.6 HL Top-hat-stiffener – comparison of experimental strains .............................. 126
4.3.7 HL Top-hat-stiffener experimental Summary ..................................................... 128
4.4 VIP Top-hat-stiffener .................................................................................................... 129
4.4.1 VIP Top-hat-stiffener manufacturing process .................................................... 129
4.4.2 VIP Top-hat-stiffener Results .............................................................................. 131
4.4.3 VIP Top-hat-stiffener Layup 4 experimental results ........................................... 131
4.4.4 VIP Top-hat-stiffener Layup 5 experimental results ........................................... 133
Page | xii
 4.4.5 VIP Top-hat-stiffener Layup 6 experimental results ........................................... 135
4.4.6 VIP Top-hat-stiffener Test Summary .................................................................. 136
4.5 Comparison of Manufacturing Processes for THS ........................................................ 136
4.5.1 Resin Burn-off Tests ............................................................................................ 137
4.5.2 Strength and Stiffness Comparison .................................................................... 138
4.5.3 Strength and Stiffness Comparison at 10 kN ...................................................... 141
4.5.4 Discussion of the Two Manufacturing Processes ............................................... 141
4.6 Top-hat-stiffener experimental study conclusions ...................................................... 142

Chapter 5: Numerical Analysis ........................................................................................... 144

5.1. Introduction................................................................................................................. 145
5.1.1 Modelling Methodology ..................................................................................... 145
5.1.2 Material Properties ............................................................................................. 146
5.1.3 Element Type and Meshing ................................................................................ 147
5.1.4 Failure Modelling ................................................................................................ 148
5.2 L-bend Laminates ......................................................................................................... 150
5.2.1 L-Bend FE Model definition ................................................................................ 150
5.2.2 L-Bend FEA Results ............................................................................................. 152
5.2.3 L-Bend Layup 1 (All CSM) FEA results ................................................................. 152
5.2.4 L-Bend Layup 2 (CSM+DB) FEA results ............................................................... 153
5.2.5 L-Bend Layup 3 (CSM+DB+UD) FEA results ........................................................ 154
5.2.6 L-Bend Interlaminar tensile and shear stress distributions ................................ 155
5.2.7 L-bend Modelling summary ................................................................................ 156
5.3 Hand Laid Top-hat-stiffeners (HL-THS) ......................................................................... 158
5.3.1 Top-hat-stiffener geometry and material layup ................................................. 158
5.3.2 Top-hat-stiffener Structural Modelling .............................................................. 159
5.3.3. Top-hat-stiffener Loading and Boundary Conditions ........................................ 160
5.3.4 Top-hat-stiffener Contact Conditions and Load Case ......................................... 160
5.4 Top-hat-stiffener FEA Results ....................................................................................... 161
5.4.1 HL Top-hat-stiffener Layup 1 FEA results ........................................................... 161
5.4.2 HL Top-hat-stiffener Layup 2 FEA results ........................................................... 164
5.4.3 HL Top-hat-stiffener Layup 3 FEA results ........................................................... 166
5.4.4 HL Top-hat-stiffener Interlaminar Strain Distribution around bends ................. 170
Page | xiii
 5.4.5 HL Top-hat-stiffener: Comparison of Strains from FEA and Experiment............ 173
5.4.6 VIP Top-hat-stiffener Layup 4 FEA results .......................................................... 175
5.5 FEA of L-Bend and Top-hat-stiffener Conclusions ........................................................ 178

Chapter 6: Acoustic Emission ............................................................................................. 179

6.1 Acoustic-emission Introduction ................................................................................... 180
6.1.1 Parametric Analysis of AE signal ......................................................................... 180
6.1.2 Transient Analysis of AE signal ........................................................................... 182
6.2 AE Instrumentation and Setup ..................................................................................... 183
6.2.1 AE Data collection procedure ............................................................................. 184
6.2.2 AE Failure Characterisation ................................................................................ 185
6.3 L-bend Composite AE Failure Characterisation ............................................................ 185
6.3.1 L-bend Layup 1 (CSM) AE Correlation................................................................. 186
6.3.2 L-bend Layup 2 (CSM+DB) AE Correlation .......................................................... 188
6.3.3 L-bend Layup 3 (CSM+DB+UD) AE Correlation ................................................... 190
6.3.4 Summary of L-bend AE Testing ........................................................................... 192
6.4 Hand-laid Top-hat-stiffener AE Failure Characterisation ............................................. 192
6.4.1 Hand-laid Top-hat-stiffener Layup 3 AE Failure Characterisation ...................... 193
6.5 AE Detection with Optical Fibre Sensors ...................................................................... 197
6.5.1 Phase-shifted Fibre Bragg Grating AE Measurement System ............................ 198
6.5.2 Pencil Lead Break Test ........................................................................................ 200
6.5.3 Deployment of the PS-FBG Sensor in Continuous Monitoring of Composite
Failures ........................................................................................................................ 201
6.5.4 Fibre Bragg Grating (FBG) Characterisation ....................................................... 202
6.6 VIP THS AE Failure Characterisation ............................................................................. 202
6.6.1 VIP THS Layup 4 Top-hat-stiffener AE Testing - Setup........................................ 203
6.6.2 VIP THS Layup 4 (CSM+DB)—AE Parametric Analysis ........................................ 203
6.6.3 VIP THS Layup 4 Parametric Analysis—FBG-AE sensor ...................................... 206
6.6.4 VIP THS - Layup 4 Waveform Failure Characterisation for Piezoelectric and FBG-
AE Sensor ..................................................................................................................... 211
6.7 AE Conclusions ............................................................................................................. 220

Page | xiv
Chapter 7: Optical-fibre Sensors......................................................................................... 222
7.1 Introduction.................................................................................................................. 223
7.2 Fabrication of Fibre Bragg Grating (FBG) sensor .......................................................... 224
7.3 Layup and Embedding Optical Fibres ........................................................................... 227
7.4 FBG Strain Measurement Setup ................................................................................... 235
7.4.1 Test Procedure .................................................................................................... 238
7.5 VIP THS Layup 4 FBG Results ........................................................................................ 238
7.5.1 FBG Survival after Specimen Manufacture ......................................................... 238
7.5.2 VIP THS Layup 4 FBG-strain Results .................................................................... 239
7.6 FBG-strain sensing Conclusions .................................................................................... 246

Chapter 8: Conclusion and Future Work ............................................................................. 247

8.1 Overview ...................................................................................................................... 248
8.2 Summary of Key findings .............................................................................................. 248
8.3 Future work .................................................................................................................. 250

References ......................................................................................................................... 252

Appendix ........................................................................................................................... 280
Appendix A – Hand-layup - Tensile test Result .................................................................. 281
Appendix B – Hand-layup 3-point bend test results .......................................................... 285
Appendix C – AE Piezo-sensor Calibration ......................................................................... 289
Appendix D – HL THS - Layups 1 and 2 AE Failure Characterisation – Piezo-sensor .......... 291
Appendix E – VIP THS - Layups 5 and 6 AE failure characterisation – Piezo-sensor .......... 295
Appendix F – VIP THS AE Failure Characterisation – FBG sensor ....................................... 301
Appendix G – Optical fibre embedding procedure with VIP Technique ............................ 303
Appendix H – Layup 5 and Layup 6 FBG strain results ....................................................... 305

Page | xv
List of Tables

Table 2.1: Effects of geometry and material variations [Shenoi and Wellicome, 1993] ............ 11
Table 2.2: Recent keel-related failures ....................................................................................... 26
Table 2.3: Types of defects found in composite materials [Heslehurst and Scott, 1990] .......... 30
Table 2.4: Limiting value failure theories.................................................................................... 35
Table 2.5: Interactive failure theories ......................................................................................... 36
Table 2.6: Summary of failure mode theories ............................................................................ 38
Table 2.7: Summary of the grades scored by each failure theory .............................................. 40
Table 2.8: Fracture toughness characterisation ......................................................................... 42
Table 2.9: Summary of selected NDE techniques for damage monitoring and inspection for
composites [Staszweski et al., 2004; Li et al., 2004] ................................................................... 49
Table 2.10: Summary of FBG studies. ......................................................................................... 75
Table 3.1: Comparison of various manufacturing techniques [Mazumdar, 2001; Schwartz,
1996b] ......................................................................................................................................... 94
Table 3.2: Various tests on composite material: British Standards (BS), ISO, European Standards
(EN), ASTM, and Australian Standards (AS) ................................................................................ 96
Table 3.3: Various safety factors [Shenoi and Dodkins, 2000] ................................................... 96
Table 3.4: Comparison of the two hardeners [FGI, 2002] .......................................................... 97
Table 3.5: Tensile test for hand-layup(HL) process — test matrix.............................................. 98
Table 3.6: Tensile test for vacuum-infused process (VIP) — test matrix. ................................... 98
Table 3.7: Three-point bend test for hand-layup(HL) — test matrix. ......................................... 99
Table 3.8: Three-point bend test for vacuum-infused process (VIP) — test matrix. .................. 99
Table 3.9: Hand-layup tensile test summary .............................................................................. 99
Table 3.10: VIP tensile test summary. ...................................................................................... 100
Table 3.11: Hand-layup three-point bend test summary ......................................................... 104
Table 3.12: VIP three-point bend test summary....................................................................... 105
Table 4.1: Specimen layup ........................................................................................................ 110
Table 4.2: L-bend Layup 1 initial and final failure load, deflection and stiffness at first failure 112
Table 4.3: L-bend Layup 2 initial and final failure load, deflection and stiffness at first failure 113
Table 4.4: L-bend Layup 3 initial and final failure load, deflection and stiffness at first failure 115
Table 4.5: Basic materials for composite hand-layup manufacturing technique ..................... 118
Page | xvi
Table 4.6: Hand-laid Top-hat-stiffener hand-layup laminate configuration ............................. 119
Table 4.7: Hand laid Top-hat-stiffener – Layup 1 initial and final failure load, deflection and
stiffness ..................................................................................................................................... 121
Table 4.8: Hand laid Top-hat-stiffener – Layup 2 initial and final failure load, deflection and
stiffness ..................................................................................................................................... 122
Table 4.9: Hand laid Top-hat-stiffener – Layup 3 initial and final failure load, deflection and
stiffness ..................................................................................................................................... 124
Table 4.10: Basic materials for VIP ............................................................................................ 130
Table 4.11: VIP Top-hat-stiffener laminate configuration ........................................................ 130
Table 4.12: VIP-Top-hat-stiffener Layup 4 initial and final failure load, deflection and stiffness
.................................................................................................................................................. 132
Table 4.13: VIP-Top-hat-stiffener Layup 5 initial and final failure load, deflection and stiffness
.................................................................................................................................................. 133
Table 4.14: VIP-Top-hat-stiffener Layup 6 initial and final failure load, deflection and stiffness
.................................................................................................................................................. 135
Table 4.15: Top-hat-stiffener HL and VIP comparison .............................................................. 137
Table 4.16: Burn-off test results for Top-hat-stiffener – HL and VIP ........................................ 137
Table 4.17:THS strength and stiffness comparison based on the flexure modulus and Young’s
modulus at the first and ultimate failures ................................................................................ 140
Table 4.18: THS Flexure stress, shear stress and laminate stiffness at a 10 kN load................ 141
Table 5.1: Material properties–hand-layup process ................................................................. 146
Table 5.2: Layup configuration for L-bend laminates ............................................................... 151
Table 5.3: Hand laid Top-hat-stiffener laminate configuration ................................................ 159
Table 5.4: Elements and nodal numbers for three different layups of hand laid Top-hat-
stiffener ..................................................................................................................................... 160
Table 5.5: VIP Top-hat-stiffener layup 4 ................................................................................... 175
Table 5.6: Material properties of the VIP Top-hat-stiffener ..................................................... 175
Table 6.1: Summary of AE parametric studies .......................................................................... 181
Table 6.2: Summary of AE transient studies ............................................................................. 182
Table 6.4: AE parametric values for failure characterisation of composites ............................ 185
Table 6.5:HL Top-hat-stiffener Layup 3, four-channel amplitude distribution ranges (dB) ..... 194
Table 6.6: HL Top-hat-stiffener Layup 3 AE failure characterisation ........................................ 194
Table 6.7: Properties of PS-FBG and tuneable laser diode, and sensing performance of the
designed system ....................................................................................................................... 200
Table 6.8: VIP THS Layup 4, four-channel amplitude distribution ranges (dB) ......................... 205
Table 6.9: VIP THS Layup 4, AE parametric analysis for piezoelectric sensor ........................... 205

Page | xvii
Table 6.10: VIP THS Layup 4, AE parametric analysis for FBG-AE sensor ................................. 206
Table 6.11: Waveform analysis of VIP THS Layup 4 specimens–piezoelectric sensor .............. 212
Table 6.12: Waveform analysis of VIP THS, Layup 4 specimens–FBG-AE sensor ..................... 212
Table 7.1: FBG location in Top-hat-stiffener Layup 4 specimen ............................................... 228
Table 7.2: FBG location in Top-hat-stiffener Layup 5 specimen ............................................... 228
Table 7.3: FBG location in Top-hat-stiffener Layup 6 specimen ............................................... 229
Table 7.4: FBG wavelength survival details for VIP THS Layup 4 specimens ............................ 233
Table 7.5: FBG wavelength survival details for VIP THS Layup 5 specimens ............................ 234
Table 7.6: FBG wavelength survival details for VIP THS Layup 6 specimens ............................ 234
Table 7.7: VIP THS Layup 4 FBG survival summary ................................................................... 239
Table 7.8: VIP THS Layup 5 FBG survival summary ................................................................... 239
Table 7.9: VIP THS Layup 6 FBG survival summary ................................................................... 239
Table 7.10: Layup 4 FBG failure characterisation ..................................................................... 244
Table A1: CSM 925H hardener hand-layup tensile test ............................................................ 281
Table A2: DB 925H hardener hand-layup tensile test............................................................... 281
Table A3: UD 925H hardener hand-layup tensile test .............................................................. 282
Table A4: CFM 925H hardener VIP tensile test ......................................................................... 282
Table A5: CFM CHM50 hardener VIP tensile test ..................................................................... 282
Table A6: DB 925H hardener VIP tensile test ........................................................................... 283
Table A7: DB CHM50 hardener - post curing VIP tensile test ................................................... 283
Table A8: UD-900 CHM50 hardener VIP tensile test - L=175mm .............................................. 283
Table A9: UD-00 CHM50 hardener VIP tensile test ................................................................... 284
Table B1: CSM 925H hardener hand-layup 3-point bend test - L=110 mm .............................. 285
Table B2: DB 925H hardener hand-layup 3-point bend test - L=110 mm................................. 285
Table B3: UD-00 925H hardener hand-layup 3-point bend test - L=110 mm ............................ 286
Table B4: CSM 925H hardener VIP 3-point bend test - L=110 mm ........................................... 286
Table B5: CSM 925H hardener VIP 3-point bend test - L=110 mm ........................................... 287
Table B6: CFM 925H hardener VIP 3-point bend test - L=110 mm ........................................... 287
Table B7: CFM CHM50 hardener VIP 3-point bend test - L=110 mm ....................................... 287
Table B8: DB CHM50 hardener VIP 3-point bend test - L=110 mm .......................................... 288
Table B9: UD-900 CHM50 hardener VIP 3-point bend test - L=110 mm ................................... 288
Table B10: UD-900 CHM50 hardener VIP 3-point bend test - L=110 mm ................................. 288
Table D1: HL THS Layup 1 four channel amplitude distribution ranges ................................... 291
Table D2: HL THS Layup 1 AE failure characterisation .............................................................. 291
Page | xviii
Table D3: HL THS Layup 2 four channel amplitude distribution ranges ................................... 293
Table D4: HL THS Layup 2 AE failure characterisation .............................................................. 293
Table E1: VIP THS Layup 5 piezo-sensor amplitude distribution ranges................................... 295
Table E2: VIP THS Layup 5 Piezo-sensor acoustic-emission failure characterisation ............... 295
Table E3: VIP THS Layup 6 piezo-sensor amplitude distribution ranges................................... 298
Table E4: VIP THS Layup 6 Piezo-sensor acoustic-emission failure characterisation ............... 298
Table F1: VIP THS Layup 5 FBG sensor acoustic-emission failure characterisation .................. 301

Page | xix
List of Figures

Figure 2.1: Curved structures subjected to opening moments inducing interlaminar

tensile/compressive stress [Kedward et al., 1989] ....................................................................... 9
Figure 2.2: Typical composite T-joint [Junhou and Shenoi, 1996] .............................................. 12
Figure 2.3: Typical Top-hat-stiffener configuration .................................................................... 13
Figure 2.4: (a) Design of Top-hat-stiffeners [ABS, 2000] and (b) effective width of FRP plating
and FRP stiffener details [ABS, 1994].......................................................................................... 15
Figure 2.5: Loading configurations: three-point bend; reverse-bend; pull-off [Phillips et al.,
1999] ........................................................................................................................................... 17
Figure 2.6: Failure modes for Layups A and B [Eksik et al., 2004) .............................................. 18
Figure 2.7: Major types of fixed keels [Encyclopaedia Britannica, 2010] ................................... 19
Figure 2.8: (a) Typical keel with keel bolts; (b) keel-hull mating [Courtesy of EMP Composites,
Sydney]........................................................................................................................................ 20
Figure 2.9: Keel bolt - Washer plate arrangement [Courtesy of EMP Composites, Sydney] ...... 21
Figure 2.10: Typical keel floor arrangement [Courtesy of EMP Composites, Sydney] ............... 21
Figure 2.11: Typical keel showing the keel bolts and keel design [ABS, 2000] ........................... 22
Figure 2.12: Keel-hull assembly .................................................................................................. 23
Figure 2.13: Comparison of Keel-hull mating surface area over the years [Yachting World, 2008]
.................................................................................................................................................... 24
Figure 2.14: Changes in keel design relative to time [Yachting World, 2008] ............................ 24
Figure 2.15: Bavaria 42 Match, April 2005, Croatia [www.sailing.hu, 2005] .............................. 27
Figure 2.16: Hooligan V, February 2007, UK [MIAB, 2007] ......................................................... 27
Figure 2.17: (a) George Phydias, sister vessel to Cynthia Woods; (b) Cynthia Woods after the
accident on June 6, 2008 [Texas A&M University, 2009] ........................................................... 28
Figure 2.18: Cynthia Woods – Keel fracture surface [Texas A&M University, 2009] ................. 28
Figure 2.19: (a) Failure sequence in unidirectional composite with fibre-dominated strength
under longitudinal tensile loading; (b) microbuckling, leading to formation of kink zones with
excessive deformation or fracture planes for ductile or brittle fibres, respectively [Daniel and
Ishai, 1994]. ................................................................................................................................. 31
Figure 2.20: Failure mode of unidirectional composite under in-plane shear [Daniel and Ishai,
1994] ........................................................................................................................................... 31

Page | xx
Figure 2.21: (a) Progressive micro-cracking leading to ultimate failure in unidirectional
composite under transverse tension [Daniel and Ishai, 1994]; (b) shear failure mode in
unidirectional composite under transverse compression .......................................................... 32
Figure 2.22: Fibre and inter-fibre failure modes [Maligno, 2007] .............................................. 32
Figure 2.23: Delamination sources—material and geometric discontinuities [Sridharan, 2008]
.................................................................................................................................................... 33
Figure 2.24: Basic crack-propagation modes [Wang, 1996] ....................................................... 41
Figure 2.25: Critical strain energy release rates (SERR) for crack initiation (G1i); Crack
propagation (G1c) [Mouritz et al., 1999] ..................................................................................... 42
Figure 2.26: (a) Delamination modelled with bilinear three-dimensional solid elements; (b)
Delamination modelled with bilinear plate/shell type elements [Krueger, 2004] ..................... 45
Figure 2.27: (a) using AE as an alarm system; (b) using AE as a forecasting tool [Muravin, 2009]
.................................................................................................................................................... 50
Figure 2.28: Typical AE process chain [Hartmut, 2002] .............................................................. 52
Figure 2.29: Typical AE signal ...................................................................................................... 53
Figure 2.30: AE typical time-history AE plot – Duration and Load vs. time ................................ 55
Figure 2.31: AE Typical cumulative plot – counts v. time ........................................................... 55
Figure 2.32: Schematic of the basic structure of an optical fibre [Schwartz, 2009] ................... 60
Figure 2.33: Schematic diagram of a polarimetric sensor configuration [Schwartz, 2009] ........ 61
Figure 2.34: Schematic configurations of IFPI, EFPI, ILFE, FBG, LPG, TFBG and DFBG sensors
[Schwartz, 2009] ......................................................................................................................... 62
Figure 2.35: Schematic diagram of a FBG ................................................................................... 63
Figure 2.36: a) Schematic of a FBG b) graphical representation of light interaction when a
broadband optical field, I0 is launched into FBG resulting the reflected, IR and transmitted, IT
field ............................................................................................................................................. 64
Figure 2.37: Data acquisition system for FBG interrogation [Ling et al., 2005b] ........................ 66
Figure 2.38: (a) A matrix crack grown fully across the coupon; (b) reflected FBG spectra at
increasing applied strains for a coupon with a crack .................................................................. 67
Figure 2.39: (a) Optical output from the sensor when five cracks pass it; (b) Band-pass filtered
signal using FFT band pass range of 100–300Hz ......................................................................... 68
Figure 2.40: (a) Integral strain v. time plot from the interferometer; (b) a detailed view of the
jump ............................................................................................................................................ 68
Figure 2.41: (a) T-Joint before lamination; (b) comparison of bond-line longitudinal strain
distribution from FEA, FBG and experimental results ................................................................ 69
Figure 2.42: (a) Embedding an FBG sensor in a composite cross-ply laminate; (b) crack density
and stress as a function of strain measured from a strain gauge and FBG ................................ 70
Figure 2.43: (a) Experimental setup; (b) integral strain v. load position for an experimental FEA
comparison ................................................................................................................................. 71

Page | xxi
Figure 2.44: (a) Composite specimen with an embedded FBG sensor; (b) reflection spectra
from the FBG sensor with an increasing applied load ................................................................ 72
Figure 2.45: Response of FBG sensors to (a) uniform; (b) non-uniform strain distribution ....... 72
Figure 2.46: (a) Specimen with embedded FBG sensors; (b) transverse crack evolution; (c)
change in wavelength distribution of reflected light from the FBG sensor................................ 73
Figure 2.47: (a) T-joints embedded with artificial delaminations during manufacturing; (b)
strain distribution along the bond-line of a T-joint with embedded delaminations .................. 74
Figure 2.48: Optical-fibre interferometric sensing system ......................................................... 77
Figure 2.49: High-performance interferometric sensor arrangement with phase-modulation
scheme and fibre-laser sensor .................................................................................................... 78
Figure 2.50: Setup for an intensity acoustic sensor .................................................................... 80
Figure 2.51: Actual Top-hat-stiffener with hull and keel fin ....................................................... 81
Figure 2.52: Modified loading condition for the Top-hat-stiffener (Upside down) .................... 82
Figure 2.53: Research methodology ........................................................................................... 84
Figure 3.1: Marine structure lifecycle [Moan, 2003] .................................................................. 87
Figure 3.2: Forces on a yacht [Larsson and Eliasson, 1994] ........................................................ 87
Figure 3.3: Structural design based on direct analysis of load effects and strength [Moan, 2003]
.................................................................................................................................................... 88
Figure 3.4: Reinforcement types [Larson and Eliasson, 1994] .................................................... 90
Figure 3.5: Hand-layup technique for manufacturing composites [Carbon Fibre Guru, 2010] .. 93
Figure 3.6: Vacuum-bag infusion process for manufacturing composite [Tygavac Advance
Materials, Ltd.] ............................................................................................................................ 93
Figure 3.7: CFM and CSM composite tensile strength comparison .......................................... 101
Figure 3.8 CFM and CSM composite Young's modulus comparison ......................................... 101
Figure 3.9: Double bias composite tensile strength comparison ............................................. 102
Figure 3.10: Double bias composite Young's Modulus comparison ......................................... 102
Figure 3.11: Unidirectional composite tensile strength comparison ....................................... 103
Figure 3.12: Unidirectional composite Young's Modulus comparison ..................................... 103
Figure 3.13: CSM/CFM composite flexure strength comparison.............................................. 105
Figure 3.14: CSM/CFM composite flexure modulus comparison ............................................. 105
Figure 3.15: Double bias composite flexure strength comparison ........................................... 106
Figure 3.16: Double bias composite flexure strength modulus comparison ............................ 106
Figure 3.17 Unidirectional composite flexure strength comparisons ...................................... 106
Figure 3.18: Unidirectional composite flexure modulus comparison....................................... 107
Figure 4.1: (a) L-bend specimen geometry (b) L-bend experimental setup ............................. 110
Figure 4.2: (a) First hinge fixture (b) Three-hinge setup ........................................................... 111
Page | xxii
Figure 4.3: L-bend layup 1 load-deflection plot ........................................................................ 112
Figure 4.4: Progressive failure of L-bend Layup 1 specimens ................................................... 113
Figure 4.5: L-bend Layup 2 load-deflection plot ....................................................................... 114
Figure 4.6: Progressive failure of L-bend Layup 2 specimens ................................................... 114
Figure 4.7: L-bend Layup 3 load-deflection plot ....................................................................... 115
Figure 4.8: Progressive failure of L-bend Layup 3 specimens ................................................... 116
Figure 4.9: Top-hat-stiffener manufacturing process ............................................................... 118
Figure 4.10: Geometry of the Top-hat-stiffener ....................................................................... 119
Figure 4.11: Top-hat-stiffener experimental test configuration ............................................... 120
Figure 4.12: Hand laid Top-hat-stiffener load deflection for Layup 1 ...................................... 121
Figure 4.13: Hand laid Top-hat-stiffener Layup 1 damage process .......................................... 122
Figure 4.15: Hand laid Top-hat-stiffener Layup 2 damage process .......................................... 124
Figure 4.16: Hand laid Top-hat-stiffener load deflection for Layup 3 ...................................... 125
Figure 4.17: Hand laid Top-hat-stiffener Layup 3 damage process .......................................... 126
Figure 4.18: Hand laid Top-hat-stiffener tensile strain (Hyy) for SG1 and SG2 .......................... 127
Figure 4.19: Hand laid Top-hat-stiffener Transverse shear strain (Hyz) for SG1 and SG2 .......... 128
Figure 4.20: Top-hat-stiffener manufacturing with VIP ............................................................ 131
Figure 4.21: VIP Top-hat-stiffener load deflection plot for Layup 4 ......................................... 132
Figure 4.22: VIP Top-hat-stiffener Layup 4 damage process .................................................... 133
Figure 4.23: VIP Top-hat-stiffener load deflection plot for Layup 5 ......................................... 134
Figure 4.24: VIP Top-hat-stiffener Layup 5 damage process .................................................... 134
Figure 4.25: VIP Top-hat-stiffener load deflection plot for Layup 6 ......................................... 135
Figure 4.27: Top-hat-stiffener CSM+DB HL and VIP comparison .............................................. 138
Figure 4.28: Top-hat-stiffener Layups 3 and 5 comparisons .................................................... 139
Figure 5.1 Element orientations at the bend; local co-ordinate system .................................. 147
Figure 5.2: 3D linear interface element [MSC MARC Volume A, 2008] .................................... 148
Figure 5.3: Damage evolution curve for bilinear cohesive element [MSC MARC Volume A, 2008]
.................................................................................................................................................. 150
Figure 5.4: a) L-Bend Experimental setup; b) Corresponding FE model ................................... 151
Figure 5.5: L-Bend Layup 1 ILTS and ILSS distribution .............................................................. 153
Figure 5.6: L-Bend Layup 1 experiment and FEA load deflection plot ...................................... 153
Figure 5.7: L-Bend Layup 2 ILTS and ILSS distributions ............................................................. 154
Figure 5.8 L-Bend Layup 2 – Experiment and FEA load deflection plot .................................... 154
Figure 5.9: L-Bend Layup 3 ILTS and ILSS distributions ............................................................. 155
Page | xxiii
Figure 5.10: L-Bend Layup 3 experiments and FEA load-deflection plot .................................. 155
Figure 5.11: L-Bend ILTS distribution just before first failure for all three layups .................... 156
Figure 5.12: L-Bend ILSS distribution just before first failure for all three layups .................... 156
Figure 5.13: L-Bend FEA load-deflection plot for three different layups .................................. 157
Figure 5.14: Top-hat-stiffener – Geometry of the finite element model ................................. 158
Figure 5.15: A typical contact table defined in MSC.MARC Mentat for the Top-hat-stiffener
model ........................................................................................................................................ 161
Figure 5.16: HL Top-hat-stiffener Layup 1 ILTS and ILSS distribution just before first failure .. 162
Figure 5.17: HL Top-hat-stiffener Layup 1 load-deflection plots—experiment and FEA
comparison ............................................................................................................................... 162
Figure 5.18: HL Layup 1 correlation of experimental failure and FEA failure prediction ......... 163
Figure 5.19: Top-hat-stiffener Layup 1 ILTS and ILSS distribution just before first failure ....... 164
Figure 5.20: HL Top-hat-stiffener Layup 2 load-deflection plot—experiment and FEA
comparison ............................................................................................................................... 165
Figure 5.21: HL Top-hat-stiffener Layup 2 experimental failure and FEA failure-prediction
correlation................................................................................................................................. 166
Figure 5.22: HL Top-hat-stiffener Layup 3 ILTS and ILSS distribution just before first failure .. 166
Figure 5.23: HL Top-hat-stiffener Layup 3 load-deflection plot—experiment and FEA
comparison ............................................................................................................................... 167
Figure 5.24: HL Top-hat-stiffener Layup 3 experimental failure and FEA failure prediction
correlation................................................................................................................................. 168
Figure 5.25: HL Top-hat-stiffener Layup 3 final stage of failure; the structure takes a trapezium
shape—FEA and experimental correlation ............................................................................... 169
Figure 5.26: Delamination failure index at first failure for all three layups at left crown bend 170
Figure 5.27: Hand laid Top-hat-stiffener through-thickness ILTS distribution ......................... 171
Figure 5.28: Hand laid Top-hat-stiffener through-thickness ILSS distribution ......................... 171
Figure 5.29: Hand laid Top-hat-stiffener ILTS distribution across the width ............................ 171
Figure 5.30: Hand laid Top-hat-stiffener ILSS distribution across the width ............................ 172
Figure 5.31: Hand laid Top-hat-stiffener ILTS distribution along a bend (radius) .................... 172
Figure 5.32: Hand laid Top-hat-stiffener ILSS distribution along the bend (radius) ................. 172
Figure 5.33: Hand laid Top-hat-stiffener Comparison of FEA results and strain gauge 1 (SG1) Hyy
.................................................................................................................................................. 173
Figure 5.34: Hand laid Top-hat-stiffener Comparison of FEA results and strain gauge 1 (SG1) Hyz
.................................................................................................................................................. 174
Figure 5.35: Hand laid Top-hat-stiffener Comparison of FEA results and strain gauge 2 (SG2) Hyy
.................................................................................................................................................. 174

Page | xxiv
Figure 5.36: Hand laid Top-hat-stiffener Comparison of FEA results and strain gauge 2 (SG2) Hyz
.................................................................................................................................................. 174
Figure 5.37: VIP Top-hat-stiffener Layup 4 load-deflection plot experimental and FEA
comparison ............................................................................................................................... 176
Figure 5.38: VIP Top-hat-stiffener Layup 4 Failure - FEA and experimental comparison ......... 177
Figure 6.1: Sensors and preamplifiers used in this study ......................................................... 183
Figure 6.2 AE setup for L-bend composites .............................................................................. 186
Figure 6.3: HL L-Bend AE load correlation for Layup 1 specimens ........................................... 188
Figure 6.4: HL L-bend AE correlation for Layup 2 specimen ..................................................... 190
Figure 6.5: HL L-bend AE correlation for Layup 3 specimen ..................................................... 192
Figure 6.6: AE experimental setup and location of piezo-sensors on Top-hat-stiffener .......... 193
Figure 6.7: Hand-laid Top-hat-stiffener Layup 3 specimens—AE parameter correlation ........ 196
Figure 6.8: AE-FBG and AE piezo-sensors surface bonded to the Top-hat-stiffener ................ 197
Figure 6.9: Intensity type AE sensor designed for failure monitoring in the current study ..... 198
Figure 6.10: (a) Transmission spectrum of the fabricated PS-FBG; b) Power spectrum of the
tuneable laser diode ................................................................................................................. 198
Figure 6.11: Adjustment of laser wavelength for optimum operation ..................................... 199
Figure 6.12: Arrangement for the pencil lead break test ......................................................... 200
Figure 6.13: Comparison of waveforms recorded from the pencil lead break test between
electronic microphone and PS-FBG AE sensor.......................................................................... 201
Figure 6.14: Setup for FBG characterisation ............................................................................. 202
Figure 6.15: AE-FBG setup during the test process .................................................................. 203
Figure 6.16: VIP THS Layup 4—AE comparison of piezo and FBG sensor amplitude................ 207
Figure 6.17: VIP THS Layup 4—AE comparison of piezo and FBG sensor energy released ...... 207
Figure 6.18: VIP THS Layup 4—AE comparison of piezo and FBG sensor signal duration ........ 208
Figure 6.19: VIP THS Layup 4—AE piezo and FBG sensor signal rise time comparison ............ 208
Figure 6.20: VIP THS Layup 4—AE piezo and FBG sensor count comparison ........................... 209
Figure 6.21: VIP THS Layup 4—AE piezo-sensor signal strength .............................................. 209
Figure 6.22: VIP THS Layup 4—AE piezo and FBG sensor cumulative counts and load v. time
plots .......................................................................................................................................... 210
Figure 6.23: VIP Top-hat-stiffener Layup 4—piezo-sensor hits and load v. time plot .............. 211
Figure 6.24: VIP Top-hat-stiffener Layup 4—piezo-sensor cumulative hits and load v. time plot
.................................................................................................................................................. 211
Figure 6.25 (a): Piezo and FBG-AE sensor AE waveforms at 224.66 sec – failure 1.................. 213
Figure 6.25 (b): Piezo and FBG-AE sensor AE FFT at 224.66 sec–failures 1 .............................. 213
Figure 6.26 (a): Piezo and FBG-AE sensor AE waveform at 299.83 sec–failures 2 ................... 213
Page | xxv
Figure 6.26 (b): Piezo and FBG-AE sensor AE FFT at 299.83 sec–failures 2 .............................. 214
Figure 6.27 (a): Piezo and FBG-AE sensor AE waveform at 320.23 sec–failures 3 ................... 214
Figure 6.27 (b): Piezo and FBG-AE sensor AE FFT at 320.23 sec–failures 3 .............................. 214
Figure 6.28 (a): Piezo and FBG-AE sensor AE waveform at 345.35 sec–failures 4 ................... 215
Figure 6.28 (b): Piezo and FBG-AE sensor AE FFT at 345.35 sec— failures 4............................ 215
Figure 6.29 (a): Piezo and FBG-AE sensor AE waveform at 383.36 sec— failures 5 ................. 215
Figure 6.29 (b): Piezo and FBG-AE sensor AE FFT at 383.36 sec— failures 5............................ 216
Figure 6.30 (a): Piezo and FBG-AE sensor AE waveform at 422.50 sec— failures 6 ................. 216
Figure 6.30 (b): Piezo and FBG-AE sensor AE FFT at 422.50 sec—failure 6 .............................. 216
Figure 6.31 (a): FBG-AE sensor AE waveform at 486.50 sec— failures 7 ................................. 217
Figure 6.31 (b): FBG-AE sensor AE FFT at 486.50 sec— failures 7 ............................................ 217
Figure 6.32 (a): Piezo and FBG-AE sensor AE waveform at 530.77 sec— failures 8 ................. 217
Figure 6.32 (b): Piezo and FBG-AE sensor AE FFT at 530.77 sec— failures 8............................ 218
Figure 6.33 (a) Piezo and FBG-AE sensor AE waveform at 592.52 sec— failures 9 .................. 218
Figure 6.33 (b): Piezo and FBG-AE sensor AE FFT at 592.52 sec— failures 9............................ 218
Figure 6.34 (a): Piezo and FBG-AE sensor AE waveform at 655.54 sec—failure 10 ................. 219
Figure 6.34 (b): Piezo and FBG-AE sensor AE FFT at 655.54 sec—failure 10 ............................ 219
Figure 6.35 (a): Piezo and FBG-AE sensor AE waveform at 743.00 sec—failure 11 ................. 219
Figure 6.35 (b) Piezo and FBG-AE sensor AE FFT at 743.00 sec—failure 11 ............................. 220
Figure 7.1: Setup of FBG writing system in UNSW (Photonics lab) .......................................... 224
Figure 7.2: Phase-mask technique for writing FBG ................................................................... 225
Figure 7.3(a): Laser alignment being performed prior to writing process to optimise lasing
power and (b) the grating fabrication enclosure (left) and control unit (right) (Photonics lab)
.................................................................................................................................................. 226
Figure 7.4: a) Inside view of the grating writing enclosure before; b) during writing process . 226
Figure 7.5: Equipment for annealing FBG (Photonics lab) ........................................................ 226
Figure 7.6: Location of four FBGs on a Top-hat-stiffener specimen ......................................... 228
Figure 7.7: Optical-fibre preparation for embedding into the laminate .................................. 229
Figure 7.8: Positioning of optical fibres on the THS mould prior to resin infusion ................... 230
Figure 7.9: (a) Vertical positioning of bare FBG sensors on the marked area; (b) folding of upper
layer E-glass fibre in getting optical fibre through the fabric ................................................... 231
Figure 7.10: (a) Preparation for VIP; (b) Fully infused specimen – still under the bag ............. 232
Figure 7.11: VIP Top-hat-stiffeners with embedded FBGs, sliced and ready for testing .......... 233
Figure 7.12 System setup for strain measurement. All bare FBG sensors, except the reference,
were embedded in laminate. .................................................................................................... 235
Page | xxvi
Figure 7.13: FBG interrogation unit used in the test procedure............................................... 236
Figure 7.14: Measured reflectivity of FBG sensors from each wavelength for a typical specimen
.................................................................................................................................................. 237
Figure 7.15: Measured reflection spectrum of all sensors multiplexed together .................... 238
Figure 7.16: Top-hat-stiffener Layup 4 FBG strain-sensor measurement—left outer.............. 240
Figure 7.17: VIP THS Layup 4 FBG strain-sensor measurement—right outer .......................... 241
Figure 7.18: VIP THS Layup 4 FBG strain-sensor measurement—left inner ............................. 241
Figure 7.19: VIP THS Layup 4 FBG strain-sensor measurements—right inner ......................... 242
Figure 7.20: VIP Top-hat-stiffener Layup 4 Bragg wavelength shift during different phases of
the test procedure .................................................................................................................... 245
Figure C1: AE sensor calibration certificate 01 ......................................................................... 289
Figure C2: AE sensor calibration certificate 02 ......................................................................... 290
Figure D1: HL THS Layup 1 piezo acoustic-emission and Load-deflection curve correlation ... 292
Figure D2: HL THS Layup 2 piezo acoustic-emission and Load-deflection curve correlation ... 294
Figure E1: VIP THS Layup 5 piezo acoustic-emission and Load-deflection curve correlation... 297
Figure E2: VIP THS Layup 6 piezo acoustic-emission and Load-deflection curve correlation... 300
Figure F1: VIP THS Layup 5 FBG acoustic-emission and Load-deflection curve correlation ..... 302
Figure H1: VIP THS Layup 5 FBG strain-sensor measurement—right outer ............................. 305
Figure H2: VIP THS Layup 5 Bragg wavelength shift during different phases of the test
procedure .................................................................................................................................. 305
Figure H3: VIP THS Layup 6 FBG strain-sensor measurement—Left outer .............................. 306
Figure H4: VIP THS Layup 6 FBG strain-sensor measurement—Right outer ............................ 306
Figure H5: VIP THS Layup 6 FBG strain-sensor measurement—Right Inner ............................. 307
Figure H6: VIP THS Layup 5 Bragg wavelength shift during different phases of the test
procedure .................................................................................................................................. 307

Page | xxvii
Abbreviations and Acronyms

2D Two Dimensional
3D Three Dimensional
A/D Analog to Digital
ABS American Bureau of Shipping
AC Alternating Current
AE Acoustic-emission
AI Artificial Intelligence
ALS Accidental Collapse Limit state
ANN Artificial Neural Network
AS Australian Standards
ASE Amplified Spontaneous Emission
AST Automatic Sensor Test
ASTM American Society of Testing and Materials
BBO Beta-Barium Borate
BE Bi-Axial
BS British Standards
BVID Barely visible Impact damage
CDA Centroid Detection Algorithm
CFM Continuous Filament Mat
CFRP Carbon Fibre Reinforced Plastic
CG Centre of Gravity
CHM Cumyl hydroperoxide
CLPT Classic Laminate Plate Theory
CLS Cracked Lap Shear
CSM Chopped Strand Mat
CTS Compact Tension Shear
DB Double Bias
DC Direct Current
Page | xxviii
 DCB Double Cantilever Beam
DFBG D-shaped fibre Bragg gratings
DM Detectors Module
DNV Det Norske Veritas
DSP Digital signal processing
ECT Edge Crack Torsion
EDT Edge Delamination Test
EFPI Extrinsic Fabry-Perot Interferometric sensor
ELS End Loaded Split
EMI Electro Magnetic Induction
EMP Engineering Materials Processing
EN European Standards
ENF End Notch Flexure
EPFM Elastic Plastic Fracture-mechanics
FBD Free Body Diagram
FBG Fibre Bragg Grating
FE Finite Element
FEA Finite Element Analysis
FF Fibre Failure
FFT Fast Fourier Transform
FLS Fatigue Failure Limit state
FPF First Ply Failure
FPTF Fabry-Perot Tunable Filter
FRP Fibre Reinforced Plastic
GFRP Glass Fibre Reinforced Plastic
GRP Glass Reinforced Polymer
HDT Hit Definition Time
HL Hand Layup
HLT Hit Lockout Time
IFF Inter-Fibre Failure
IFPI Intrinsic Fabry-Perot Interferometric sensor
ILFE Inline Fibre Etalon
ILSS Interlaminar Shear Stress
Page | xxix
 ILTS Interlaminar Tensile Stress
ISO International Organisation for Standardisation
LM Laser Module
LEFM Linear Elastic Fracture-mechanics
LPG Long Period Gratings
LVDT Linear Variable Differential Transformer
MAIB Marine Accident Investigation Branch
MEKP Methyl ethyl ketone peroxide
MMB Mixed Mode Bending
MoD Ministry of Defence
MSPS Mega samples per second
NDE Non-Destructive Evaluation
OFS Optical Fibre Sensor
OHS Operation, Health and Safety
OSA Optical Spectrum Analyser
PAC Physical Acoustic Corporation
PDT Peak Definition Time
PGC Phase Generated Carrier
POF Polymer Optical Fibre
PS-FBG Phase Shifted Fibre Bragg Grating
RM Righting Moment
RINA Royal Institute of Naval Architects
RMS Root Mean Square value
RTM Resin Transfer Moulding
SDM Spatial Division Multiplexing
SERR Strain Energy Release Rate
SG Strain Gauge
SHM Structural Health Monitoring
SIFT Strain Invariant Failure Theory
SLB Single Leg Bending
SMF Single Mode Fibre
TDM Time Division Multiplexing
TFBG Tilted Fibre Bragg gratings
Page | xxx
 THS Top-hat-stiffener
TLS Tunable Laser Source
UD Unidirectional
UF Ultimate Failure
ULS Ultimate Failure Limit state
UV Ultra-Violet
VCCT Virtual Crack Closure Technique
VIP Vacuum Infused Process
WDM Wavelength Division Multiplexing
WR Woven Roving
WWFE World-wide Failure Exercise

Page | xxxi
Symbols and Nomenclature

This section summarises commonly used symbols and abbreviations throughout the thesis. The
list of symbols is not comprehensive; variations exist within the text but are defined explicitly
whenever they are used.

eu 1 atto = 10-18
Xc, Yc, Zc Allowable compressive stress in the X, Y and Z directions
Sij Allowable shear stress in the ij plane
γ12, γ23, γ31 Allowable shear strain in the XY, YZ and ZX directions
ε1, ε2, ε3 Allowable tensile/compressive strain in the X, Y and Z directions
Xt, Yt, Zt Allowable tensile stress in the X, Y and Z directions
Atm Atmosphere
M Bending moment
λB Bragg wavelength
Vc Critical effective opening displacement
δa Crack increment length
Gc Critical strain energy release rate
D Damage variable
dB Decibels
neff Effective refractive index
ρ11, ρ12 Elasto-optic coefficients
Wf Fibre weight fraction
FPF First Ply Failure
Ge Germanium
gsm Grams per square metre
l Grating length
Λ Grating period
h Height
Hz Hertz

Page | xxxii
 St Interlaminar shear Strength
τrθ Interlaminar shear Stress
σr Interlaminar tensile Stress
kN Kilo Newton
IL Lasing power
MPa Mega Pascal
μ Micro (10-6)
mm Millimetre
mW Milliwatts
σs Mode I crack initiation stress
σp Mode I crack propagation stress
GIc or G1c Mode I strain energy release rate
τs Mode II crack initiation stress
τp Mode II crack propagation stress
GIIc or G2c Mode II strain energy release rate
GIIIc or G3c Mode III strain energy release rate
I Moment of Inertia
N Newton
Nm Newton metre
ε Normal strain
σ Normal stress
pe Optical constant
IR, IS Optical Intensities for reference and sensing path

IR , IS Optical phase delay for reference and sensing path

Pa Pascal
νij Poisson's ratio in the ij plane
IR Reflected power
V Shear load
Gij Shear modulus in the ij plane
W12, W23, W31 Shear stress in XY, YZ and ZX directions
Sxy, Syz, Szx Shear stress in X, Y and Z directions
T Temperature

Page | xxxiii
 ΔT Temperature change
V1, V2, V3 Tensile/Compressive stress in the X, Y and Z directions
αΛ Thermal expansion coefficient
αn Thermo-optic coefficient
Zt Through-thickness tensile strength
Sn Tin
U Total energy of the system
tn, ts, tt Traction vector terms
IT Transmitted Power
UF Ultimate Failure
v Voltage
b Width
Vy Yield strength of the material
E Young's modulus

Page | xxxiv
Chapter 1: Introduction

http://modern-luxury-homes.blogspot.com/2009/10/luxury-boats.html

Page | 1
1.1 Background Information
The use of composites in structural applications dates back at least 60 years. Due to limited
research and scarce availability of composites, early structural applications were limited; only
military and defence projects could afford their high cost. More recently, reinforced polymer
composite applications have become popular for marine, aerospace and automobile
structures, sporting goods and biomedical components. Composite materials are preferred
over metal structures for added benefits such as a high strength-to-weight ratio, lower
maintenance, small electromagnetic signature, longer life and customisable properties. Glass
fibre reinforced plastic (GFRP) composites are popular in the marine industry due to their low
cost and ease of manufacture. Using composites in marine structures reduces the weight by
10–30 per cent compared to metal structures, increasing design efficiency and reducing fuel
consumption. The absence of oxidation and rust formation reduces the need for continual
maintenance. At the same time, disadvantages of GFRP include high initial cost, weaker matrix
low toughness, joining difficulty and environmental degradation of the matrix.

The issue of frequent keel failures is a mystery to many boat builders and designers. EMP
Composites in Sydney initiated this research to focus on keel failures as the marine industry
has suffered a number of casualties and material loss during the past two decades. The main
thrust of this study was to identify the key variables responsible for those keel related failures.
This research aims at investigating the failure mechanisms of Top-hat-stiffeners for keel
structures, redesigning the keel structure and estimating the residual strength after an initial
failure. The residual strength is important for estimating the time available for obtaining
assistance before the boat with a keel failure, capsizes.

In structural composite context, the initiation of cracks (or fractures) does not indicate a
catastrophic failure. A stable crack propagation stage, associated with a constant external load
precedes a catastrophic failure. This is frequently seen in the structural application of
composites. Nevertheless, due to the alignment of the fibres to the load path, some designs
exhibit reserve strength after crack initiation has occurred. Understanding and assessing such
designs requires adequate consideration of strength and fracture behaviour of the composite
laminate. An accurate prediction of static collapse behaviour for marine structures is essential
for assessing the reserve strength and likelihood of damage. This thesis also attempts to gain
insight into stress distribution and model the secondary reserve strength of the Top-hat-
Page | 2
stiffener. Failure characterisation is performed using acoustic-emission and optical fibre
sensing techniques.

1.2 Aims and Objectives

The focus of modern structural design is its enhanced performance and reduced cost.
Developing designs that meet these goals requires tremendous amount of preparation and
data. It is vital to understand the strength and stiffness of the structure, its failure modes and
different techniques for monitoring its structural health condition. In this study, experimental
and numerical studies are conducted to analyse failures of the composite structure.

The primary objectives of this research are to:

x understand the out-of-plane load transfer mechanisms in a Top-hat-stiffener;

x get an insight into the key variables that govern the delamination strength in curved
composites;

x develop an experimental technique for real-time detection of delamination in the

composite Top-hat-stiffeners so that the full process of damage initiation and
progression can be analysed;

x comprehend the stress distribution around the bend of curved composites subjected
to flexure load;

x use Acoustic-emission and optical fibre sensor NDE (Non Destructive Evaluation)
methods to predict, measure and analyse structural failure;

x compare the performance of composite laminated Top-hat-stiffeners manufactured by

hand-layup and Vacuum Infused Process (VIP); and

x develop a numerical framework for modelling the out-of-plane load transfer

mechanism and failure of composite structures that explicitly accounts for the damage
in the form of inter-ply separation.

Page | 3
These objectives led to two goals:

x Experimental characterisation of the failures of curved composites using acoustic-

emission (AE) and embedded fibre Bragg grating (FBG) optic sensors.

x Development of validated numeric models to analyse the failures and lessen the need
for future experimental studies.

The following approach was taken to achieve these goals:

x Extended literature review and identification and development of a research

methodology to study the different composite failure mechanisms.

x Development of a validated finite element model to predict and analyse the load
transfer mechanisms, damage mechanisms and complex stress distributions (in-plane
and interlaminar).

x Preliminary Experimental and numerical analysis of L-bend curved laminates subjected

to bending loads.

x Experimental investigation of different layup configurations of Top-hat-stiffeners

subjected to bending loading and identification of the key variable governing failures.
This was achieved using specimens manufactured by two different techniques, hand-
layup and vacuum-bag infusion (VIP) techniques.

x Real-time progressive failure analysis was conducted using embedded optical-fibre

strain sensors, surface bonded piezoelectric acoustic-emission (AE) sensors and
surface bonded optical fibre AE sensors.

x Identifying the reserve strength in composite laminates with unidirectional fibres and
numerically modelling this secondary reserve strength.

Page | 4
1.3 Thesis Outline
Chapter 2 - Literature Review: This comprises a comprehensive review of related literature to
typical marine structures, curved composites, accidents related to keel failure, design and
application of Top-hat-stiffeners, composite failures, numerical methods to predict failures,
fracture-mechanics, acoustic-emission and optical fibre Bragg grating techniques to detect and
analyse the different modes of composite failures. The existing space/gaps in the area to which
this research aims to contribute are also identified.

Chapter 3 - Material characterisation of marine composites: An overview of the commonly

used manufacturing systems in the marine industry is presented in this chapter. Matrix burn
off tests to obtain the fibre weight fraction (Wf) of the composite material is discussed.
Material characterisation testing procedure for measuring tensile and flexure properties using
Uniaxial tensile test (ASTM D 3039/D 3039M) and three point bend test (ASTM D 7264/D
7264M) are discussed.

Chapter 4 - Experimental program: Experimental investigation of various composite curved

structures is carried out. This chapter details the different manufacturing process, testing and
failure analysis of L-Bend and Top-hat-stiffeners. Load deflection plots and photograph
evidence of the test show, in-detail failure initiation and progression.

Chapter 5 - FE analysis of Top-hat-stiffeners: Non-linear finite element analysis has been

carried on the curved composite structures (L-bend and Top-hat-stiffeners) to predict the
failure initiation progression. The main failure mode for curved composites structures is
delamination, which has been successfully been modelled using cohesive zone and interface
elements. Failure initiation is predicted by the stress-based failure criteria (interlaminar tensile
and shear stresses) and crack progression by fracture-mechanics criteria. Failure prediction
and progression correlated well with the experimental results.

Chapter 6 - Acoustic-emission: Failure characterisation using Acoustic-emission technique is

investigated and validated with experimental and photographic evidence. Two types of sensors
are used; conventional piezoelectric sensor and modern optical fibre sensor. Different
parameters such as amplitude, signal duration, signal rise time, signal strength, energy
released and number of counts are determined during the in-situ testing of the composite
Page | 5
structures. Based on these parameters, the type of composite failure mode is determined.
Waveform analysis with fast Fourier transformation (FFT) technique is also performed to
complement the parametric AE investigation. AE is carried for three kinds of structures: L-
bend, Top-hat-stiffener – hand-layup and Top-hat-stiffener – vacuum infusion process

Chapter 7 - Optical fibre Sensors: Optical fibre sensors are small, versatile, immune to
electromagnetic induction and are very appropriate to composite materials. Four optical fibre
sensors were embedded in Top-hat-stiffeners to measure the strain during structural loading.
Based on the local strain variation, composite structural failures were determined and
validated.

Chapter 8 - Conclusions and Recommendations: Key findings and outcomes of this research
have been summarised in this chapter and some important questions have been answered. As
research never ends, unresolved issues are discussed and recommendations are suggested for
the future work in the relevant field of study.

Appendices: Appendices contain detailed information on the coupon tests - Uni-axial tensile
test and Three-point bend test for hand-layup (HL) and vacuum infused specimens (VIP).
Acoustic-emission plots are given for Top-hat-stiffener Layups 5 and 6. Procedure to embed
optical fibre in the curved composite laminate is explained in detail. The calibration detail of
the AE piezo-sensor is presented. FBG strain and strain rate plots for Layups 5 and 6 are
presented in detail.

1.4 Related Publications

Edited Book Chapters

1. Raju, Prusty B.G. (2011). AE Failure Characterization of Advanced Composites using
Piezo-Electric Sensors and Fibre Optic Sensors, for the edited book titled Theory and
uses of Acoustic emission, Nova Publications, USA. ISBN: 978-1-61209-960-6.

Journal Papers
2. Raju, Prusty B.G., Kelly D.W., Lyons D., Peng G.D. (2010). Top hat stiffeners: A Study on
Keel Failures. Ocean Engineering 37: 1180-1192.

Page | 6
 3. Raju, Prusty B.G., Kelly D.W., Lyons D., Peng G.D. (2010). Failure characterisation of L-
bend curved composite laminates, Transactions of Royal Institute of Naval Architects
(RINA) Part B2, International Journal of Small Craft Technology 152: 93-105.
4. Raju, Azmi A.I., Prusty B.G. (2012). Acoustic Emission Techniques for Failure
Characterisation in Composite top-hat Stiffener. Journal of Reinforced Plastics and
Composites 31:495-516.
5. Raju, Prusty B.G., Kelly D.W (2012). Finite element analysis of the delamination Failure
of composite top hat stiffeners. Proceedings of the Institution of Mechanical
Engineers, Part M Journal of Engineering for the Maritime Environment - Accepted.
6. Raju, Prusty B.G., (2012). Failure Investigation of Top-Hat Composite Stiffened Panels.
Ships and Offshore Structures (SOS) – Special Issue – Under review.

International Conferences
7. Raju, Gangadhara Prusty, Don Kelly, Jun Ikeda, David Lyons (2008). Experimental and
Numerical Analysis of top hat stiffeners for Keel Structures. Proceedings of Pacific
2008, International Maritime Conference, Sydney.
8. Raju, Prusty B.G., Kelly D.W., Lyons D., Peng G.D. (2009). Progressive failure analysis of
top hat stiffeners for keel structures. Proceedings of 17th Annual International
Conference on Composites/Nano Engineering (ICCE-17), Honolulu, Hawaii, USA.
9. Raju, Prusty B.G., Kelly D.W., Lyons D., Peng G.D. (2010). Delamination
Characterisation of Curved Composites Using Acoustic Emission and Fibre Bragg
Gratings. Proceedings of Pacific 2010, International Maritime Conference, Sydney,
Australia.
10. Raju, Prusty B.G., Kelly D.W., Peng G.D., Lyons D. (2010). Failure Analysis of Laminated
Composites with embedded Fibre Bragg Gratings (FBG). Joint Workshop (Australia &
Japan) on Frontier Photonics and Electronics, Sydney, Australia.
11. Azmi A.I., Raju, Peng G.D. (2012). Progressive Failure Monitoring of EGlass/Vinylester
Curve Composites Using Embedded FBG Array. SPIE 8351, 835138, Proceedings of Asia
Pacific Optical Sensors (APOS) Conference 2012, Sydney, Australia.
12. Azmi A.I., Raju, Peng G.D. (2012). Application of fiber grating-based acoustic sensor in
progressive failure testing of e-glass/vinylester curve composites. SPIE 8351, 835110,
Proceedings of Asia Pacific Optical Sensors (APOS) Conference 2012, Sydney, Australia.

Page | 7
Chapter 2: Literature Review

http://www.boatdesign.net/gallery/data/500/2009-Lazzara-Yachts-LSX-Ninety-Two-Night-1600x1200.jpg

Page | 8
2.1 Introduction
This chapter presents a comprehensive literature review in areas related to the thesis. This
thesis is a multi-disciplinary work, covering aspects of curved composites, out-of-plane joints,
keels and their failures, composite failure theories, delamination modelling, acoustic-emission
and optical-fibre sensors for strain sensing.

2.2 Curved Composites

Layered composites allow the manufacture of structures with complex shapes and sizes. These
complex structures generally consist of curved segments, which may enhance efficiency but
generally involves complex failure mechanisms. Premature failures of curved composites
occurred in the past were predominantly due to a lack of appreciation of the interlaminar
tensile strength, which may be less than three per cent of the in-plane strength of a composite
laminate. The primary failure mode in curved composites occurs when bends are opened or
closed by external loading or pressure, as shown in Figure 2.1.

Figure 2.1: Curved structures subjected to opening moments inducing interlaminar tensile/compressive
stress [Kedward et al., 1989]
Page | 9
Interlaminar tensile and compressive stresses generally develop when curved composite
laminates are subjected to tensile or compressive loading in the plane of curvature [Kedward
et al., 1989]. When the curved composites are subjected to bending loads, the inner surface
tends to expand, generating circumferential tensile stress, whereas the outer surface tends to
compress, creating circumferential compressive stress. For design accuracy and failure
prediction, it is essential to understand the interlaminar stress (tensile/compressive/shear)
distribution during the design and analysis of a curved composite structure. The interlaminar
tensile stress is a function of the basic laminate constituents, such as the type of reinforcement
and matrix, stacking sequence, manufacturing process, fabrication quality, aging and type of
loading, environmental degradation [Heinz et al., 2000; Davies and Petton, 1999; Kim and
Sham, 2000]. The stress distribution is complex in curved composites; they more often
delaminate due to lack of through-thickness strength. A study of interlaminar stress
distribution at the bends provides a tool for understanding and predicting structural
delamination. In the laminate, the interlaminar tensile/compressive stress range is minimum at
the inner and outer surfaces and maximum at the mid-thickness of the composite laminate.

Kedward et al. [1989] studied the radial stress distribution for a curved beam using classic
elastic theory. They simplified the strength of materials approach and conducted finite
element analysis (FEA) using Nastran 2D and 3D elements. A finite element (FE) code was
developed [Graff, 1989] to obtain the stresses and strains in the curved composite strap
laminates (Graphite-Epoxy, Kevlar-Epoxy, Glass-Epoxy) numerically using the Tsai-Wu failure
criterion. Chang and Springer [1986] developed numerical codes to calculate the interlaminar
stress in a simple right angle bend subjected to pure moment. In-plane failure strength was
predicted using Tsai-Hill failure criterion and out-of-plane strength using the Chang-Springer
failure criterion. Later comparisons between fracture-mechanics and strength-based
delamination predictions were made [Martin and Jackson, 1993]. The difficulty of interlaminar
tensile/compressive or shear stresses was highlighted at the bonded joints and attachments of
the marine composite structures [Dodkins and Shenoi, 1994]. Wisnom [Wisnom, 1996;
Kaczmarek and Wisnom, 1998] was one of the primary investigators, studying delamination
failure through experimental, analytical and numerical approaches to detect interlaminar
failure and determine the flexural strength of composite laminates. Lekhnitskii’s equations are
applicable only for pure bending or edge loading; thus, they cannot sustain the circumferential
force without radial restraints. Shenoi and Wang [2001] developed equations based on
Page | 10
elasticity theory for delamination and the flexural strength of curved composites. The effects
of key variables, such as the stacking sequence and radius of curvature on stress distribution
within a curved layered beam and sandwich beam were also studied. Among the different
failure modes of composite materials, delamination mode predominates for curved laminates.
Also failure analysis becomes more critical when the load is transferred in an out-of-plane
mode.

2.3 Out-of-plane Joints

One of the key challenges in designing composite structures is the high-strength joints. Often,
the load needs to be transferred in an out-of-plane mode. The joints carrying such loads must
be carefully designed. Out-of-plane joints are formed by placing additional composite layers on
both sides of the structure and filling the gap with an appropriate filler material. Generally
used out-of-plane joints are the T-joints and flanges of Top-hat-stiffeners [THS].

In case of a hull-bulkhead joint (T-joint), the load is transmitted between the bulkhead and
shell through the laminate in its in-plane direction. When the load is applied to the joint, high
through-thickness stress is developed at the bends of the joint and the structure fails by
delamination. As the load increases, the structure suffers multiple delaminations until the
structure fails. Traditionally, one criterion considered in designing this joint is that the sum of
the in-plane properties of the two laminates is equivalent to that of the weakest member
being joined. In most applications, the material used in the laminate is the same as that used in
the structure. In that case, the thickness of each laminate is specified as a minimum of half the
thickness of the thinner structural element. Such joints are susceptible to bond-line porosity
and localised disbonding due to manufacturing defects and high/repetitive operational loads.

Generally, the designer is faced with the challenge of opting for either strength (requiring a
thicker laminate) or flexibility (requiring a thin laminate) [Shenoi and Wellicome, 1993]. These
opposing design requirements present different failure mechanisms based on in-plane and
through-thickness stresses, as shown in Table 2.1.

In-plane stress in
Response feature Through-thickness stress in laminate
laminate
Increasing thickness of laminate Decreases Increases
Increasing bend radius Decreases Decreases
Table 2.1: Effects of geometry and material variations [Shenoi and Wellicome, 1993]
Page | 11
2.3.1 T- Joints

A T-joint is constructed by connecting two orthogonal panels (web and flange) perpendicular
to each other. This joint is formed by connecting the two plates with laminate around the joint.
T-joints are used between the hull and the bulkhead, where tensile, shear or combined loads
are transferred. A typical T-joint is shown in Figure 2.2. The main failure mode with T-joints is
delamination between the overlap and the base panel when the vertical member is pulled
against the base panel. Matrix cracking is typically the initial failure mechanism, which
precedes delamination and fibre breakage.

Figure 2.2: Typical composite T-joint [Junhou and Shenoi, 1996]

2.4 Top-hat-stiffener
Top-hat-stiffeners are generally used in the fabrication of stiffened panels, such as decks,
bulkheads or the hull shell, joining the keel to the boat hull [Dodkins and Shenoi, 1994]. The
key benefits of using Top-hat-stiffeners are high bending stiffness and torsional resistance. As
the name suggests, the stiffener has a hat shape with a flange, web and a crown, as shown in
Figure 2.3. The top bend between the crown and the web is stiffened by adding more
composite layers (preferably unidirectional).

Due to the absence of through-thickness reinforcements and the lack of bond ductility, out-of-
plane joints suffer from relatively low strength and fatigue resistance against out-of-plane
tensile, bending and shear loads [Smith, 1990; Shenoi and Hawkins, 1995]. In addition, they are

Page | 12
susceptible to failure by peel or delamination well before the ultimate in-plane laminate
material stress is reached [Dodkins and Shenoi, 1994].

Figure 2.3: Typical Top-hat-stiffener configuration

2.4.1 Design Rules and Standards for Top-hat-stiffeners

x Gibbs and Cox [1960]: One of the earlier design recommendations for maximum
strength and rigidity suggested the use of woven roving reinforcement for a Top-hat-
stiffener with a thickness approximately equal to that of the shell laminate. Based on
industry experience, Gibbs and Cox published design graphs for varying stiffness and
strength, simply by changing the cross-sectional dimensions.

x Lloyd’s Register of Shipping [1983]: Lloyd’s Register recommends the necessary

weight, strength and modulus of the laminate where the materials were assumed to
be chopped strand mat (CSM), woven rowing (WR) and unidirectional fibres. The
required dimensions are obtained by direct calculation, based on the vessel speed and
displacement, when the length of the vessel exceeds 30 m. Section moduli based on
stiffener depth and laminate thicknesses were defined. The width of the flange
connection to the plate laminate was taken as 25 mm+12 mm per 600 gsm of
reinforcement in the stiffener web, or 50 mm, whichever is greater.

Page | 13
x Det Norske Veritas (DNV) [1991]: They proposed a similar approach, suggesting the
design of Top-hat-stiffeners based on first principles to calculate the dimensions and
section modulus. However, no specific information is given on the overall details of the
out-of-plane joint and design allowables.

x United Kingdom (UK) Ministry of Defence (MoD) [1982]: Based on the mine hunter
project, UK MoD came up with an efficient design procedure for obtaining the design
variables for Top-hat-stiffeners. They recommended that the thickness of the laminate
for the Top-hat-stiffener must be at least half (preferably two thirds) the thickness of
the weakest member at the joint. Design details about the stacking sequence, layup,
over laminate dimensions and fillet radius were provided. They also suggested
repeating the laying of CSM and WR layers until the desired thickness is achieved.

x American Bureau of Shipping (ABS) [2000]: ABS standards specify the stiffener width,
height, laminate thickness and the length of the overlap at the flange. The thickness of
the laminate was taken as t mm, as shown in Figure 2.4.

Page | 14
 Figure 2.4: (a) Design of Top-hat-stiffeners [ABS, 2000] and (b) effective width of FRP plating and FRP
stiffener details [ABS, 1994]

ABS recommends that the minimum overlap length at the flange should be 20 per cent
of the stiffener height, or 50 mm, whichever is greater. However, there is no
information given for the bend radius, laminate layup or manufacturing process.

x ISO 12215-5 [2008]: The International Organisation for Standardisation (ISO) uses the
weight of the stiffener laminate, rather than its thickness. Section 5, Annex G describes
the procedure to derive the dimensions of structural panels and stiffeners—the
different types (squat, square and tall) of FRP Top-hat-stiffeners, in particular. This
document provides a number of design recommendations to determine the thickness
of the laminate based on the reinforcement weight and fibre weight fraction.

x Australian Standards [AS 4132.31993]: This suggests that the layup configuration
should contain a unidirectional layer on the outer surface of the Top-hat-stiffener if it
is a hollow former (no core). It suggests the relationship between the width of the
crown, the overall depth of the stiffener and laminate thickness.

The available standards recommend a number of design parameters for the Top-hat-stiffener.
The dimensional parameters are dependent on the laminate thickness. None of the design
standards suggests detailed design procedures for the bend radius between flange-web and
crown-web for a Top-hat-stiffener. In addition, the published standards do not provide any
information on the manufacturing procedure, layup configuration, fatigue properties, failure
mechanism or residual strength of the structure. None of the standards addresses the out-of-
plane mechanical properties.

Page | 15
2.4.2. Past Studies on Top-hat-stiffeners

Green and Bowyer [1981] studied the failure mechanism of Top-hat-stiffeners. They used plain
weave and woven roving glass cloth with polyester resin. The laminate was manufactured by
hand-layup method, where a foam core was glued to the base panel, over which layers of fibre
mat were laid. Straight pull-off tests were performed and delaminations were observed
between the flange and the base panel. Dodkins et al. [1994] developed numerical models to
assess the attachment of the Top-hat-stiffeners to the base plate. They studied various fillet
radii (25–125 mm) and fillet backfill angles inside the stiffener (0q – 45q). A parametric study
was performed on laminate thickness, gap size, backfill angle and the fillet radius of the
composite Top-hat-stiffener. Delaminations between the laminate and the base panel were
examined. Shenoi and Hawkins [1995] numerically investigated the design of Top-hat-
stiffeners for shell plating joints for ships and boats using the commercial FEA software, ANSYS.
Load-transfer paths and failure mechanisms were determined.

Junhou and Shenoi [1996] conducted experimental studies and proposed three tests on Top-
hat-stiffeners: three-point bend, reverse three-point bend and pull off, as shown in Figure 2.5.
The foam core was glued to the base panel and the laminates were laid over it. For the three-
point bend, initial failures were observed at the fillet-laminate interface followed by
progressive delaminations in the laminate, due to excessive interlaminar stresses. Failure of
the panel skin occurred during the reverse three-point bend test. Complete disbonding of the
laminate from the base panel was observed during the pull-off test. Phillips et al. [1999]
conducted detailed experimental and numerical analysis on these three testing mechanisms to
determine the internal load-transfer characteristics and failure mechanisms of Top-hat-
stiffeners. They suggested that the delamination in the curved region close to the fillet was
due to high through-thickness stress.

Page | 16
 Figure 2.5: Loading configurations: three-point bend; reverse-bend; pull-off [Phillips et al., 1999]

Eksik et al. [2004] conducted experimental and numerical studies on top hat-shaped glass-
reinforced plastic (GRP) beams, testing three specimens for each of two different layups, using
four-point bend test procedure until failure. The failure sequence observed was matrix
cracking, root whitening caused by delamination at the web-flange corner and final failure with
fibre matrix breakage or catastrophic shear failure, depending on the layup. The locations of
Page | 17
the failures for the two layups are shown in Figure 2.6. Experimental and numerical study of
top-hat stiffened panels by Eksik et al. [2007] provided a valuable insight into identifying the
critical locations of structural damage and the failure mechanisms associated with in-service
loading conditions.

Figure 2.6: Failure modes for Layups A and B [Eksik et al., 2004)

2.5 Keels
The keel is considered as the basic structural member of the ship which runs from bow to
stern. It serves as the backbone of the structure; it is the major source of structural strength of
the hull and provides greater directional control and stability. Traditionally, laying the keel was
the first part of shipbuilding; however, modern ships are built with a pre-fabrication process
where the keels are attached to the hull at a later stage using keel bolts. In non-sailing vessels,
the keel helps the hull to move forward, rather than slipping to the side. In the case of sailing
ships, keels provide adequate lift to counteract the lateral force generated by the mast.
Generally, a fair amount of weight is added to the keel to assist with the righting effect of the
hull's hydrodynamic forces. A keel serves as a dual-purpose hydrofoil. Firstly, it counteracts
any heeling movement; secondly, it minimizes sideways drift. The key factors in keel design are
the root chord, span, taper, twist, longitudinal position at tip, cant angle, junction angle,
junction fairing and section characteristics [Larsson and Eliasson, 1994]. Earlier designs
concentrated on lowering the centre of gravity by flattening the wings, so that more weight
could be added. However, in modern keels, the weight is placed on the lower part of the keel
as a bulb.

Page | 18
2.5.1 Types of Keels

The two types of commonly used keels are fixed and non-fixed. Fixed keels include full keels,
long keels, fin keels, winged keels, bulb keels, bilge keels, skeg keels and centreboard keels,
shown in Figure 2.7. Non-fixed keels are swing and canting keels.

Figure 2.7: Major types of fixed keels [Encyclopaedia Britannica, 2010]

x Full keel: This keel runs at least 50 per cent of the length of the hull. Its main
advantages are providing for safe grounding and improved directional stability,
reducing the chance of fall-off as well as less swinging off course due to wind gusts and
wave action. However, this type of keel makes it difficult for the vessel to make sharp
turns. The large keel produces significant drag and the speed is lower than other keel
types. There is a large contact area between the keel and the ship. This keel suits
stability more than speed.

x Fin keel: The shape is similar to that of a shark fin, deep, flat and sharp edged; its
length is typically less than 50 per cent of the hull length. There is less wetted area
with reduced drag, thus speed is enhanced. The turning radius is small and therefore,
this is the keel generally used in racing boats. The reduced length provides less
resistance to forces such as wind gusts and waves. Bulbs are a type of fin keel that
have a ballast weight at the lower end of the fin, allowing for a shallower design. A
Page | 19
 wing keel at the trailing edge of the keel provides additional hydrodynamic stability
with reduced tip-vortex turbulence.

x Swing keels: These are weighted, narrow, fin-type keels providing both ballast and
lateral stability. Centreboard keels are similar to swing keels but not weighted. Swing
keels can be inserted in the water or retracted, depending on the depth of the water.
A major drawback is the required maintenance. Swing keels are generally lighter than
fixed-fin keels.

x Canting keels: A canting keel mounts on a hinge and is hydraulically rotated to

windward when the boat heels. Rotating the keel away from the heeling direction
generates force that both rights the boat and propels it forward. The drawback is the
required maintenance and the complexity of the mechanism. Racers prefer canting
keels for increased speed and stability.

2.5.2 Keel-hull Connection

The keel is normally attached using a series of bolts that penetrate the hull (Figure 2.8). These
bolts load the hull through washer plates (Figure 2.9), which bear down upon the flanges of
Top-hat-stiffeners called keel floors (Figure 2.10).

Figure 2.8: (a) Typical keel with keel bolts; (b) keel-hull mating [Courtesy of EMP Composites, Sydney]

Page | 20
Figure 2.9: Keel bolt - Washer plate arrangement [Courtesy of EMP Composites, Sydney]

Figure 2.10: Typical keel floor arrangement [Courtesy of EMP Composites, Sydney]

Page | 21
With such high loadings (possibly a combination of tensile/compression, flexure and torsional
loads), the design of the supporting structure in this region is critical. However, because of the
complex loading, this region is often over designed to prevent failure. Wind force on the sails
causes a sailboat to heel. Resistance to heeling, called righting moment (RM), results from the
lateral movement of the boat's centre of buoyancy away from the centre of gravity (CG). The
RM is the heart of loads throughout the boat and its structure. Placing the weight as low as
possible (often in a large bulb at the bottom of the keel) extracts the maximum righting
moment from a given mass. However, increasing the RM can result in a rapid increase of
power and, in turn, the structural stresses, especially where the keel attaches to the hull. The
parameters required to calculate the minimum diameter of the keel bolts are shown in Figure
2.11.

Figure 2.11: Typical keel showing the keel bolts and keel design [ABS, 2000]

The keel bolts usually connect the keel direct to the hull laminate in pockets between the
transverse floors and longitudinals, relying on the high compressive strength of the laminate.
They do not, as in timber construction, go up through the floors because the core of the floors
is not sufficiently strong in compression to take the imposed loads. Compression tubes are not
used through floors either, because they would interfere with the pre-tension loads in the
bolts. The laminate typically fails by inter-laminar shear at the junction between the floor or
longitudinal and the bottom laminate, often at or close to the load-bearing steel plate. Figure
2.12 shows the structural configuration of a keel attached to the boat hull and the Top-hat-

Page | 22
stiffener (THS). Steel/titanium alloy bolts are used to attach the keel to the hull. The top hat
stiffener is tilted upside down in the current study to facilitate the experimental program (keel
load is managed by the actuator in INSTRON).

Figure 2.12: Keel-hull assembly

2.5.3 Evolution of Keel Design

Keel-related accidents were rare in the twentieth century; however, the last two decades have
seen numerous accidents in which keels were damaged or fallen off, and the vessel capsized,
leading to human and capital loss. A comprehensive study of the keel failures in this thesis
concluded that design and manufacturing inaccuracies were the cause of majority of the

Page | 23
accidents. Maintenance was also an issue, but had far less an effect than the combined effect
of design and manufacturing factors. Although conventional keels provide ample space for keel
bolt attachment, modern keels provide lesser contact area and may require a different design
approach.

The keel is mounted to the hull with keel bolts. The area of contact between the keel surface
and the yacht hull has varied over time. Earlier keels were light and had fat sections and huge
fin areas. Those designs had a high centre of gravity and a large contact area, which reduced
the load and fatigue on keel bolts. An ample number of keel bolts would distribute the load
adequately, as shown in Figure 2.13. The need for higher speed changed the design of modern
keels, making them stiffer, with high-lift and low-drag fins, resulting in less contact surface
area, which produces a high load on the keel bolts. The evolution of keel design is shown in
Figure 2.14. In some cases, the keel acts only as a strut supporting the heavy bulb to minimise
drag, and the keel bolts are closer to the centreline, resulting in higher stress in the bolts.

Figure 2.13: Comparison of Keel-hull mating surface area over the years [Yachting World, 2008]

Figure 2.14: Changes in keel design relative to time [Yachting World, 2008]

Page | 24
2.5.4. Major Keel-related Accidents

Keel failure may be attributed to inadequate design, sub-standard manufacturing and lack of
maintenance of the structure (for ex: polyester resin can shrink/compress over time, resulting
in abnormal variation in the torque setting of the keel bolts). A summary of keel-related
accidents from the past two decades is presented in Table 2.2. Keel failures relevant to the
current study are shown in Figures 2.15 to 2.18. The common factor behind most of the
failures is attributed to the change in design needed for an ultra-light hull and high-speed
racing qualities. The demand for high speed and a low centre of gravity has influenced
designers to use a smaller contact mating area to the keel structure. This generates enormous
stress concentrations on the keel bolt. To sustain these high stress concentrations, the
structural components (Top-hat-stiffener, backing plate and hull) had to be made stiffer and
stronger. As the hull and backing plate design modifications are complicated, design
enhancement of the THS is beneficial in overcoming this hurdle as the major failures were due
to inferior design and flawed manufacturing processes, as opposed to operating and
environmental conditions.

Date Vessel Location Type of failure Casualties

14000 kg keel fell off and boat

1 10/08/1985 Drum Ireland 16 Injured
capsized

Due to Inadequate hull

Rising NSW, 2 dead, 4
2 01/04/2001 thickness, fin keel separated
Farrster Australia injured
from the hull

NSW, Keel broke off due to poor 4 dead, 2

3 16/09/2002 Excalibur
Australia welding of the keel sections injured

Crack generated and

propagated through repeated
4 02/02/2002 SchockaZulu San Diego Nil
grounding and finally keel broke
off

Crack generated and

Everest propagated through repeated
5 20/06/2003 Bermuda Nil
Horizontal grounding and keel finally broke
off

Tasmania, Canting keel piston failure, keel

6 28/12/2004 Skandia Nil
Australia detached and yacht capsized

Page | 25
 Keel snapped off due to fatigue,
7 03/02/2005 Ecover France Nil
35000 miles service

Lack of thickness of backing

Bavaria 42
8 26/04/2005 Croatia plates, keel ripped off and boat 1 dead
Match
capsized

Substandard lamination of keel-

hull joint, poor workmanship,
South
9 16/09/2005 Moquini insufficient periodic inspection. 6 dead
Africa
Keel detached and yacht
capsized

Sweden 42 Crack near keel-hull joint, water

10 29/11/2005 UK Nil
Caliso leakage, yacht abandoned

Keel dropped into ocean, but

San
11 04/04/2006 Cheetah 30 boat was kept upright due to Nil
Francisco
favourable sea conditions

Inadequate design and

1 dead, 5
12 02/02/2007 Hooligan V UK miscalculation of stresses, keel
injured
snapped off and boat capsized

Inadequate design of keel-hull

13 30/06/2007 Bounder UK joint, keel sheared off the hull Nil
and boat capsized

Indian Keel snapped off, yacht adrift

14 24/09/2007 Gi Go 2 2 Injured
Ocean for 9 days and finally rescued

Crack propagated under hull

Maxi - New
15 17/12/2007 near leading edge at the top of Nil
Maximus Zealand
fin. Keel finally snapped off

Inadequate number of keel

TP 52
16 21/03/2008 Hong Kong bolts, keel snaps off, boat kept Nil
Strewth
upright by the crew

Inadequate hull thickness and

Cynthia
17 06/06/2008 Mexico keel backing plate thickness, 1 dead
Woods
boat capsized

Table 2.2: Recent keel-related failures

Some of the major keel-related accidents relevant to the current study are Bavaria 42 Match
(April 2005), Hooligan V (February 2007) and Cynthia Woods (June 2008). These are discussed
in the following section.
Page | 26
x Bavaria 42 Match: Bavaria 42 Match capsized due to keel failure. The backing plates were
too thin and the keel structure could not hold the keel for long. Poor fatigue resistance
caused the keel to detach, as shown in Figure 2.15.

Figure 2.15: Bavaria 42 Match, April 2005, Croatia [www.sailing.hu, 2005]

x Hooligan-V: Another flawed design caused Hooligan–V to capsize (Figure 2.16) with one
fatality. Inadequate stress calculation caused fatigue failure of the keel in the fillet weld
area, which had been subjected to high bending stresses. Defects were also found in the
keel taper box welds and two of the three keel bolts had failed [Marine Accident
Investigation Branch, 2007]. The primary reason for failure was the fatigue of the keel fin
plating. A recommended factor of safety of 2.5 – 3.5 is commonly used for keel fin platings,
whereas the calculated safety factor for Hooligan V was just 1.38.

Figure 2.16: Hooligan V, February 2007, UK [MIAB, 2007]

x Cynthia Woods: Cynthia Woods failed in 2008 due to poor calculation of the stresses at
the backing plate and keel bolts, as shown in Figure 2.17. The hull itself was very thin and
had insufficient shear load strength when heeling due to a deficient safety factor [Texas

Page | 27
 A&M University, 2009]. Strangely, the hull was only one third of the minimum fibreglass
laminate thickness recommended by the American Bureau of Shipping [ABS, 2000]
standards. The backing plates were too thin and narrow and generated a huge stress
concentration at the connection point (Figure 2.18). The minimal hull thickness and shear
loading-heeling were the main causes of the keel separating from the hull.

Figure 2.17: (a) George Phydias, sister vessel to Cynthia Woods; (b) Cynthia Woods after the accident on
June 6, 2008 [Texas A&M University, 2009]

Figure 2.18: Cynthia Woods – Keel fracture surface [Texas A&M University, 2009]

2.5.5 Comments on Keel-hull Joints

From the keel-related failures section discussed above, it is clear that these failures resulted
from either poor design or manufacturing. During the design phase, more focus is needed for
the keel-hull joint. Moreover, the various design standards available for Top-hat-stiffeners and
keel structures do not specifically concentrate on the Top-hat-stiffeners at keel-hull joints.
Thus, it is necessary to focus on these issues to minimise the human and material loss, which
can be achieved by understanding the out-of-plane load-transfer mechanism and its effect on
composite failures. This determines the load-carrying capacity and the damage mechanism of
Top-hat-stiffeners.

Page | 28
2.6 Composite Failure
Despite their numerous advantages, composite structural applications may pose a serious risk
if the load-transfer and damage mechanisms are not properly analysed. Structural components
are subjected to multiple loadings such as tension, compression, shear and flexure. The
increase in the number of composite structure applications also demand in-depth study of the
process of load transfer between the fibre and matrix at the microscopic level to large
delamination failures at the macroscopic level. Stress distribution and failure modes in GFRP
composites are complex in nature. Geometric and material non-linearity complicates
composite failure prediction; however, a fundamental feature of composite structures is that
the failure is not usually a unique event, but an accumulation of a gradual sequence of micro
cracking and delaminations leading to structural collapse [Hinton et al., 2004].

This could be overcome by better understanding of the material properties, which may lead to
reduced factor of safety, lower costs, lighter structures, superior design and lower potential
hazards. Unlike metals, composites are brittle, and the possible failure parameters are matrix
failure, fibre failure, interface failure and interlaminar failure. The different types of defects
found in composite materials are presented in Table 2.3.

Page | 29
 Matrix damage: Cracking Creep
Matrix damage: Crazing Crushing
Matrix damage: Incorrect cure Cuts and scratches
Matrix damage: Moisture pick-up Dents
Matrix damage: Porosity Incorrect fibre volume ratio
Fibre damage: Broken Incorrect materials
Fibre damage: Fibre/matrix debonding Marcelled fibres
Fibre damage: Misalignment Mismatched parts
Fibre damage: Miscollination Missing plies
Fibre damage: Uneven distribution Overaged prepreg
Fibre damage: Wrinkles or kinks Over-/undercured
Delamination: Bearing surface damage Ply underlap or gap
Delamination: Blistering Prepreg variability
Delamination: Contamination Reworked areas
Delamination: Corner radius delamination Surface damage
Delamination: Corner/edge splitting Surface oxidation
Delamination: Edge damage Thermal stresses
Delamination: Fastener holes Translaminar cracks
Delamination: Fibre/matrix debond Variation in density
Delamination: Holes and penetration Variation in thickness
Delamination: Fills or fuzz balls Variation in resin fraction
Delamination: Surface Swelling Missing Plies
Barely visible impact damage (BVID) Erosion
Contamination of bonded surfaces Voids
Corner cracking Warping
Table 2.3: Types of defects found in composite materials [Heslehurst and Scott, 1990]

2.6.1 Matrix Failure

The principal role of the matrix in FRP composites is to secure the fibres in place, to transfer
the load between fibres and also to protect the brittle fibres from damage. Matrix
imperfections, such as micro-cracks, voids and porosity lead to other failure modes. Voids
cannot be eliminated but can be minimised with optimised manufacturing processes. Borg
[2004] showed that the concentration of voids at weak points might reduce the compressive
strength of the structure by as much as 20 per cent. The factors affecting the matrix strength

Page | 30
are the manufacturing process, temperature during manufacture, cure and post-cure
conditions, void content and environmental degradation.

2.6.2 Fibre Failure

The fibre is the primary load-carrying member in the composite structure. Three modes of
fibre failure are tensile, compression and shear failure. Fracture behaviour of E-glass (electrical
grade glass) is brittle in nature; if the tensile load/strain exceeds its critical limit, the glass fibre
fracture may occur locally or globally. Fibre failure does not initiate catastrophic failure, as the
load is generally distributed to neighbouring fibres through the medium of the matrix. Typical
longitudinal tensile and compressive failures are shown in Figures 2.19 (a) and (b). In-plane
shear failure process of unidirectional composites is shown in Figure 2.20.

Figure 2.19: (a) Failure sequence in unidirectional composite with fibre-dominated strength under
longitudinal tensile loading; (b) microbuckling, leading to formation of kink zones with excessive
deformation or fracture planes for ductile or brittle fibres, respectively [Daniel and Ishai, 1994].

Figure 2.20: Failure mode of unidirectional composite under in-plane shear [Daniel and Ishai, 1994]

Typical transverse tensile and transverse compression failures are shown in Figures 2.21 (a)
and (b). Impact failure may cause fibre failure at the impacted area, depending on the size of
the impact zone and the impact velocity. During compression loading, the glass fibres suffer

Page | 31
micro-buckling and kink-band formation [Purslow, 1988]. Berbinau and Soutis [1999] and
Soutis [1994] showed that micro buckling initiates compressive failure in unidirectional
composites. Compressive shear failure may lead to kink-band formation, where few fibre
failures with large matrix shear deformation are observed [Vogler and Kyriakides, 2001].
Typical fibre failure (tensile and compressive) and inter-fibre failures are shown in Figure 2.22.

Figure 2.21: (a) Progressive micro-cracking leading to ultimate failure in unidirectional composite under
transverse tension [Daniel and Ishai, 1994]; (b) shear failure mode in unidirectional composite under
transverse compression

Figure 2.22: Fibre and inter-fibre failure modes [Maligno, 2007]

Page | 32
2.6.3 Interface Failure

Fibre-matrix debonding may initiate due to microscopic cracks arising in the matrix material
during loading. When fibre breakage occurs, the fibre tips act as microscopic cracks and cause
debonding. Good fibre-matrix interface strength controls the fibre pull-out.

2.6.4. Interlaminar Failure – Delamination

Delamination is the primary failure mode in composites and reduces the strength and integrity
of the structure, particularly through matrix cracking [Sridharan, 2008; Garg, 1988].
Delamination is induced by interlaminar tension and shear due to inherent factors such as
stiffness (elastic modulus) mismatch between the layers, structural discontinuities, free edge
effects around holes, impact by a foreign object and through-thickness stress in curved
composites. External factors such as fatigue, moisture and temperature variation and voids
may also cause delamination. Other failure modes, such as matrix cracking, can also induce
delamination. Most common sources of delamination in structural composites are material
and structural discontinuities, as shown in Figure 2.23.

Figure 2.23: Delamination sources—material and geometric discontinuities [Sridharan, 2008]

Page | 33
Initial failure through delamination can be undetected, as delamination is frequently
embedded between layers within the composite structure. In some cases, delamination may
reduce the residual strength of the structure as much as 60 per cent [Antonio and Miguel,
2002]. The problem of interlaminar tensile/compressive or shear stress was highlighted at the
bonded joints and attachments of marine composite structures [Dodkins et al., 1994]. Wisnom
[1996 and 1998] was one of the primary investigators to study delamination failures, using
experimental, analytical and numerical approaches to detect interlaminar failure and measure
the flexural strength of a composite laminate.

The interlaminar tensile and shear stresses are the key parameters in defining the
delamination strength of the composite. However, these are matrix-related properties; it is
very difficult to increase the delamination strength of the specimen without external
fastenings. Many methods have been suggested for arresting interlaminar failure, such as
using tougher matrix polymers, interleaf layers, through-thickness reinforcement, improved
fibre/matrix strength and optimisation of fabrication [Mouritz et al., 1997a; Mouritz et al.,
1997b].

2.7 Failure Theories for Composite Structures

Many theories have been proposed concerning composite structure failures and the three
major ones are the limiting value, interactive and failure mode theories. A comparison of the
widely used failure criteria is presented.

2.7.1 Limiting Value Theories

With these theories, a failure occurs when any stress component exceeds its corresponding
strength values. This technique may be applied either for stress- or strain-based methods, as
shown in Table 2.4.

Page | 34
Stress
Tensile
Compressive ߪଵ ߪଶ ߪଷ ߪଵ ߪଶ ߪଷ ߬ଵଶ ߬ଶଷ ߬ଷଵ
݉ܽ‫ ݔ‬൤ ǡ ǡ ǡ ฬ ฬ ǡ ฬ ฬ ǡ ฬ ฬ ǡ ฬ ฬ ǡ ฬ ฬ ǡ ฬ ฬ൨ ൒ ͳ
Transverse ܺ௧ ܻ௧ ܼ௧ ܺ௖ ܻ௖ ܼ௖ ܵଵଶ ܵଶଷ ܵଷଵ
Shear
Strain
Tensile ߝଵ ߝଶ ߝଷ ߝଵ ߝଶ ߝଷ ߛଵଶ ߛଶଷ ߛଷଵ
Compressive ݉ܽ‫ ݔ‬ቈ ǡ ǡ ǡ ฬ ฬ ǡ ฬ ฬ ǡ ฬ ฬ ǡ ቤ ቤ ǡ ቤ ቤ ǡ ฬ ฬ቉ ൒ ͳ
ܺ௧ ܻ௧ ܼ௧ ܺ௖ ܻ௖ ܼ௖ ܵ௫௬ ܵ௬௭ ܵ௭௫
Transverse
Shear
Polynomial type ሺߪଵ െ ்ܺ ሻሺߪଵ െ ܺ஼ ሻሺߪଶ െ ்ܻ ሻሺߪଶ െ ܻ஼ ሻሺߪଷ െ ்ܼ ሻሺߪଷ െ ܼ஼ ሻሺߪଵଶ
maximum stress െ ߬ଶଷ ሻሺߪଵଶ ൅ ߬ଶଷ ሻሺߪଶଷ െ ߬ଵଶ ሻሺߪଶଷ ൅ ߬ଵଶ ሻሺߪଷଵ െ ߬ଵଶ ሻሺߪଷଵ
criterion ൅ ߬ଵଶ ሻ ൌ Ͳ

Polynomial type ሺߝଵ െ ܺఌ் ሻሺߝଵ െ ܺఌ஼ ሻሺߝଶ െ ܻఌ் ሻሺߝଶ െ ܻఌ஼ ሻሺߝଷ െ ܼఌ் ሻሺߝଷ െ ܼఌ஼ ሻሺߝଵଶ
maximum strain െ ߛଶଷ ሻሺߝଵଶ ൅ ߛଶଷ ሻሺߝଶଷ െ ߛଵଶ ሻሺߝଶଷ ൅ ߛଵଶ ሻሺߝଷଵ െ ߛଵଶ ሻሺߝଷଵ
criterion ൅ ߛଵଶ ሻ ൌ Ͳ

Table 2.4: Limiting value failure theories

2.7.2 Interactive Theories

Interactive failure theories contain both linear and quadratic stress terms. A simple interactive
theory for isotropic materials was proposed by von Mises [1913], which is also known as the
maximum distortion energy criterion. This criterion states that failure occurs when the energy
of distortion reaches the same energy as that of yield/failure in uniaxial tension. Hill [1965]
defined a new failure criterion based on distortion energy. Tsai-Wu [1971] proposed a criterion
which consists of seven non-interactive (F1, F22...) and interactive (F12, W12...) coefficients. The
non-interactive coefficient terms are generally determined by uniaxial and shear strengths.
The interactive coefficient terms are determined by biaxial strengths. Later, to account for
different strengths in tension and compression for orthotropic materials, Hoffman [1967]
extended Hill's equation (the basis for the Tsai-Hill criterion). Biaxial terms are not required
with this criterion. Six coefficients are needed that are determined by uniaxial stress and pure
shear tests. The key advantage of Hoffman’s theory is that the interactive terms are based on
biaxial data, so it is easier to fit them into the Tsai-Wu criterion. However, the major
disadvantage is that, although the Hoffman criterion successfully predicts the onset of failure,
it fails in identifying the corresponding failure modes. Tsai-Azzi [1965] simplified the Hill
criterion to account for transverse isotropy and plane stress conditions. The Yeh-Stratton
criterion [1994], initially proposed for isotropic materials, was later modified for composite
materials for failure prediction and progressive failure analysis.
Page | 35
 ͳ
ሾሺߪ െ ߪଶ ሻଶ ൅ ሺߪଶ െ ߪଷ ሻଶ ൅ ሺߪଷ െ ߪଵ ሻଶ ሿ ൑ ߪ௬ଶ
von Mises, ʹ ଵ
1913
ଶ ଶ
‫ݏݏ݁ݎݐݏ݈݁݊ܽ݌ݎ݋ܨ‬ǡ ߪଷ ൌ Ͳǡߪ௬ଶ ൌ ߪଵଵ ൅ ߪଶଶ ൅ ߪଵଵ ߪଶଶ

Tsai-Hill, ߪଵ ଶ ߪଶ ଶ ߪଵ ߪଶ ߪଵଶ ଶ
ቀ ቁ ൅ ቀ ቁ െ ቀ ቁቀ ቁ ൅ ቀ ቁ ൒ ͳ
1965 ܺ ܻ ܺ ܻ ܵ

‫ܨ‬ଵ ߪଵ ൅ ‫ܨ‬ଶ ߪଶ ൅ ‫ܨ‬ଵ ߪଵଶ ൅ ‫ܨ‬ଶ ߪଶଶ ൅ ʹ‫ܨ‬ଵଶ ߪଵ ߪଶ ൅ ‫߬ ଺଺ܨ‬ଵଶ

ଶ
൒ͳ
Tsai-Wu,
1971 ͳ ͳ ͳ ͳ ͳ ͳ ͳ
‫ܨ‬ଵ ൌ ൅ ǡ ‫ܨ‬ଶ ൌ ൅ ǡ ‫ܨ‬ଵ ൌ െ ǡ ‫ܨ‬ଶ ൌ െ ǡ ‫ ଺଺ܨ‬ൌ ଶ
ܺ௧ ܺ௖ ܻ௧ ܻ௖ ܺ௧ ܺ௖ ܻ௧ ܻ௖ ܵ

ଶ ଶ ଶ
‫ܣ‬ሺߪଵ െ ߪଶ ሻଶ ൅ ‫ܤ‬ሺߪଶ െ ߪଷ ሻଶ ൅ ‫ܥ‬ሺߪଷ െ ߪଵ ሻଶ ൅ ‫ߪܦ‬ଵଶ ൅ ‫ߪܧ‬ଶଷ ൅ ‫ߪܨ‬ଷଵ

Hoffman, ͳ ͳ ͳ ͳ ͳ ͳ
‫ܣ‬ൌ ൅ െ ǡ‫ ܤ‬ൌ െ ൅ ൅
1967 ʹܺ௧ ܺ௖ ʹܻ௧ ܻ௖ ʹܼ௧ ܼ௖ ʹܺ௧ ܺ௖ ʹܻ௧ ܻ௖ ʹܼ௧ ܼ௖
ͳ ͳ ͳ ͳ ͳ ͳ
‫ܥ‬ൌ െ ൅ ǡ‫ ܦ‬ൌ ǡ ‫ܧ‬ ൌ ǡ ‫ܨ‬ ൌ
ʹܺ௧ ܺ௖ ʹܻ௧ ܻ௖ ʹܼ௧ ܼ௖ ܵ‫ ݕݔ‬ଶ ܵ‫ ݖݔ‬ଶ ܵ‫ ݖݕ‬ଶ

Tsai-Azzi, ߪଵଶ ߪଵ ߪଶ ߪଶଶ ߪଵଶ

ଶ

1965 െ ଶ ൅ ଶ൅ ଶ ൒ͳ
ܺଶ ܺ ܻ ܵ

ଶ
ߪଵ ߪଶ ߬ଵଶ
Yeh- ൅ ൅ ‫ܤ‬ଵଶ ߪଵ ߪଶ ൅ ଶ ൒ ͳ
ܺ ܻ ܵ
Stratton, where B is the interaction constant to be determined by experimental data. (for
1994 example, B = -1.0X10-9 for cross ply laminates and -1.5X10-8 for angle ply
laminates

Table 2.5: Interactive failure theories

2.7.3 Failure Mode Theories

Even though limiting value and interactive theories dominated the early stages of composite
application, failure mode theories are now widely adopted, as they are more appropriate for
the design of composite structures. These theories allocate individual equations for each
failure mode. The basic aim is to identify the stress/strain values leading to a particular failure
mode.

Page | 36
Hashin and Rotem [1973] developed stress-based failure criteria to determine critical failure
modes. Sun [2000] proposed that the apparent shear strength increases when combined with
a normal compressive stress by modifying the relevant Hashin-Rotem criterion [1973] for the
1–3 plane. Sun and Tao [1997] presented a failure criterion based on classic laminate plate
theory (CLPT) to obtain the failure envelopes of Hashin-Rotem criterion for unidirectional and
multidirectional composites. Table 2.6 summarises the interactive failure criteria.

ߪଵ ȁߪଵ ȁ
Hashin- ‫݁݃ܽ݇ܽ݁ݎܤ݁ݎܾ݅ܨ‬ǣ ൌ ͳǡ ൌ ͳǡ
ܺ௧ ܺ௖
Rotem,
ߪଶ ଶ ߬ଵଶ ଶ ߪଶ ଶ ߬ଵଶ ଶ
1973 ‫݁ݎݑ݈݅ܽܨݔ݅ݎݐܽܯ‬ǣ ൬ ൰ ൅ ቀ ቁ ൌ ͳ ൬ ൰ ൅ ቀ ቁ ൌ ͳ
ܻ௧ ܵ ܻ௖ ܵ

ߪଵ ଶ ߬ଵଶ ଶ
ܶ݁݊‫݁ݎݑ݈݂݅ܽ݁ݎܾ݂݈݅݁݅ݏ‬ǣ ൬ ൰ ൅ ቀ ቁ ൌ ͳ
ܺ௧ ܵ
‫݁ݎݑ݈݂݅ܽ݁ݎܾ݂݅݁ݒ݅ݏݏ݁ݎ݌݉݋ܥ‬ǣȁߪଵଵ ȁ ൌ ܺ௖
Hashin, ߪଶ ଶ ߬ଵଶ ଶ
1980 ܶ݁݊‫݁ݎݑ݈݂݅ܽݔ݅ݎݐ݈ܽ݉݁݅ݏ‬ǣ ൬ ൰ ൅ ቀ ቁ ൌ ͳ
ܻ௧ ܵ
ߪଶ ଶ ܻ௖ ߪଶ ߬ଵଶ ଶ
‫݁ݎݑ݈݂݅ܽݔ݅ݎݐܽ݉݁ݒ݅ݏݏ݁ݎ݌݉݋ܥ‬ǣ ൬ ൰ ൅ ൤ െ ͳ൨ ൅ ቀ ቁ ൌ ͳ
ʹܵ௧ ʹܵ௧ ܻ௖ ܵ

ͳ ͳ
‫ݏ݈ܽ݅ݎ݁ݐ݈݈ܽ݉ܽݎ݋ܨ‬ሺߪ௧ ൑ ߪ௖ ሻǣ൬ െ ൰ ሺߪଵ ൅ ߪଶ ൅ ߪଷ ሻ
ܺ௧ ܺ௖
Christensen, ͳ ͳ
൅ ൜ ሾሺߪ െ ߪଶ ሻଶ ൅ ሺߪଶ െ ߪଷ ሻଶ ൅ ሺߪଷ െ ߪଵ ሻଶ ൅ሿ
1988 ܺ௧ ܺ௖ ʹ ଵ
ଶ ଶ ଶ
൅ ͵൫ߪ௫௬ ൅ ߪ௬௭ ൅ ߪ௭௫ ൯ൠ

Initial Failure
ܶ‫݊݋݅ݏ݊݁ݐ݁ݏݎ݁ݒݏ݊ܽݎ‬ǣߪଶ ൒ ‫ܨ‬ଶ௧
ߪଶ ଶ ߬ଵଶ ଶ
‫݊݋݅ݏ݊݁ݐ݁ݏݎ݁ݒݏ݊ܽݎݐ݀݊ܽݎ݄ܽ݁ݏܾ݀݁݊݅݉݋ܥ‬ǣ ൬ ൰ ൅ ൬ ൰ ൒ ͳ
‫ܨ‬ଶ௧ ‫ܨ‬ଵଶ
ܶ‫݊݋݅ݏݏ݁ݎ݌݉݋ܿ݁ݏݎ݁ݒݏ݊ܽݎ‬ǣߪଶ ൑ ‫ܨ‬ଶ௖ ሺܾ‫ߪ݄ݐ݋‬ଶ ܽ݊݀‫ܨ‬ଶ௖ ݊݁݃ܽ‫݁ݒ݅ݐ‬ሻ
Grant-
Sanders, Final Failure
1996; ‫݊݋݅ݏ݊݁ݐ݈ܽ݊݅݀ݑݐ݅݃݊݋ܮ‬ǣߪଵ ൒ ‫ܨ‬ଵ௧
Edge, 1996 ‫݊݋݅ݏݏ݁ݎ݌݉݋݈ܿܽ݊݅݀ݑݐ݅݃݋ܮ‬ǣߪ௧ ൑ ‫ܨ‬ଵ௖ ሺܾ‫ߪ݄ݐ݋‬ଶ ܽ݊݀‫ܨ‬ଵ௖ ݊݁݃ܽ‫)݁ݒ݅ݐ‬
‫ ݊ܫ‬െ ‫ݎ݄ܽ݁ݏ݈݁݊ܽ݌‬ǣȁ߬ଵଶ ȁ ൒ ‫ܨ‬ଵଶ
ߪଵ ȁ߬ଵଶ ȁ
‫ݎ݄ܽ݁ݏ݀݊ܽ݊݋݅ݏݏ݁ݎ݌݉݋݈ܿܽ݊݅݀ݑݐ݅݃݊݋݈ܾ݀݁݊݅݉݋ܥ‬ǣ ൅
‫ܪ‬ଵ௖ ‫ܪ‬ଵଶ
൒ ͳሺߪଵ ܽ݊݀‫ܪ‬ଵ௖ ܾ‫݁ݒ݅ݐ݄ܽ݃݁݊ݐ݋‬ሻ
‫݊݋݅ݐ݈ܽ݊݅݉ܽ݁ܦ‬ǣ߬ଵଶ ‫ݐ‬௖ ൒ ߱௖௥

Page | 37
Puck and Fibre failure
Schurmann, ்ܺ ܺ௖
ܺ௙் ൌ ‫ܧ‬௙ଵ ൌ‫א‬ଵ் ‫ܧ‬௙ଵ ܽ݊݀ܺ௙஼ ൌ ‫ܧ‬௙ଵ ൌ‫א‬௙ଵ ‫ܧ‬௙ଵ
1998 ‫ܧ‬ଵ ‫ܧ‬ଵ

ߪଵ ଶ ͳ ଶ ଶ ሻ
‫݁ݎݑ݈݂݈݅ܽ݁݅ݏ݊݁ݐݎܾ݁݅ܨ‬ǣሺߪଵ ൒ Ͳሻǣ൬ ൰ ൅ ଶ ሺߪଵଶ ൅ ߪଵଷ ൌͳ
ܺ௧ ܵ
ȁߪଵ ȁ
‫݁ݎݑ݈݂݅ܽ݁ݒ݅ݏݏ݁ݎ݌݉݋ܿݎܾ݁݅ܨ‬ሺߪଵ ൏ Ͳሻǣ ൌͳ
ܺ஼
‫݁ݎݑ݈݂݈݅ܽ݁݅ݏ݊݁ݐݔ݅ݎݐܽܯ‬ሺߪଶ ൅ ߪଷ ൒ Ͳሻ
Hashin, ͳ ͳ ͳ ଶ
1980 ଶ
ሺߪଶ ൅ ߪଷ ሻଶ ൅ ଶ ሺߪଷଶ ൅ ߪଶ ߪଷ ሻ ൅ ଶ ሺߪଵଶ ൅ ߪଵଷ ଶ ሻ
ൌͳ
ܻ௧ ܵ௧ ܵ
‫݁ݎݑ݈݂݅ܽ݁ݒ݅ݏݏ݁ݎ݌݉݋ܿݔ݅ݎݐܽܯ‬ሺߪଶ ൅ ߪଷ ൒ Ͳሻ
ଶ
ͳ ܻ௖ ͳ ͳ ͳ ଶ
ቈ൬ ൰ െ ͳ቉ ሺߪଶ ൅ ߪଷ ሻ ൅ ଶ ሺߪଶ ൅ ߪଷ ሻଶ ൅ ଶ ሺߪଶଶ െ ߪଶ ߪଷ ሻ ൅ ଶ ሺߪଵଶ ଶ ሻ
൅ ߪଵଷ
ܻ௖ ʹܵ௧ Ͷܵ௧ ܵ௧ ܵ
ൌͳ

Yamada and ߪଵ ଶ ߪଵଶ ଶ

ቀ ቁ ൅൬ ൰ ൌͳ
Sun, 1978 ܺ ܵ௜௦

ଶ
‫ܧ‬௠ ሺߝ஺ ሻߝ஺ ߪ் ଶ ߪ஺் ଶ
Rotem, ቆ ቇ ൅൬ ൰ ൅൬ ൰ ൒ͳ
േܵ௠ േ்ܵ ܵ஺்
1996
Where m = matrix material
T = direction transversed to the fibres
AT = shear in a plane axial and transverse to the fibre direction.

ߪଷ ଶ ߬௦ ଶ
Sun, 2000 ൬ ൰ ൅൬ ൰ ൌͳ
‫ܨ‬ଷ௖ ‫ܨ‬௦ െ ߟߪଷ

Table 2.6: Summary of failure mode theories

Puck and Schurmann [1998] developed a theory that separates composite failure into fibre
failure and inter-fibre failure using Mohr's and Hashin's reflections on brittle fractures. The
stress-based Grant-Sanders [1982] method was developed to produce envelopes of initial and
final failure mechanisms. Christensen [1997] proposed a stress-based criterion for transversely
isotropic material, which included only the effect of the deformation in the direction of the
fibre. Sun and Yamada [1978] proposed a failure criterion that assumed that a laminate would
fail when all the laminas have failed with cracks along the fibre direction; the failure strength
of the lamina is obtained from a cross-ply laminate. Chang and Chang's failure criteria [1987]
deals with the stiffness degradation model, including matrix cracking, fibre-matrix shearing and
fibre breakage based on CLPT. For fibre failure, the stiffness of the failed layer was assumed to
be depended upon the extent of damage, which was determined using a micromechanics

Page | 38
approach. Rotem [1996] developed the Hashin-Rotem criteria [1973] with a quadratic function
to obtain a better approximation of the failures involved. Hashin [1980] developed a
progressive damage model that proposed separate failure criteria for fibre and matrix failures
in both tension and compression loadings. Its major limitation is that the model was assumed
as a 2D structure. Hashin [1980] extended the Hashin-Rotem failure criterion to include
interactions between different stress terms. This criterion could be easily modified to
incorporate 3D failure of composites. Various investigators, such as Hart-smith [1998], Zinoviev
[2002], Sun and Tao [1997], have developed their own failure theories.

It is generally difficult to use the failure mode theories, as they require accurate material
properties for successful application. This requires extensive coupon study and the whole
exercise may become very expensive.

2.7.4 World-wide Failure Exercise (WWFE) Studies

No single failure criterion can satisfy all the needs of an engineer in the course of structural
design; thus, there is a lack of faith in the currently available failure criteria. In addition, the
definition of 'failure' does not have a universal explanation. Hinton et al. [2004] conducted a
worldwide failure exercise (WWFE) intended to establish the current level of failure theories,
bridge the gap between the theoreticians and designers and convince designers by presenting
a robust failure criteria. Failure theories tested in the exercise were maximum stress,
maximum strain, Tsai, Puck, Cuntze, McCartney, Eckold, Bogetti, Chamis, Edge, Hart-Smith,
Huang, Mayes and Hansen, Rotem, Sun and Tao, Wolfe and Butalia and Zonoviev. Fourteen
test cases were considered and the extent of their accuracy and reliability for different failure
modes were evaluated.

The exercise was carried out in three distinctive stages:

x Part A: Details the description of the fourteen leading theories and predictions by each
of a set of test cases.
x Part B: Compares the theoretical predictions and experimental results for all the
tested failure theories.
x Part C: Includes four more failure theories that were omitted during the initial stages
of WWFE.

Page | 39
Finally, all the failure theories were graded according to their performance, as shown in Table
2.7. The following grades were awarded for each prediction.

x Grade A if the prediction was within ±10 per cent of the measured value
x Grade B if the prediction was between ±10 and ±50 per cent of the measured value
x Grade C if the prediction was below 50 per cent or above 150 per cent of the
measured value

Grade type A B C A+B

Cuntze-B 59 38 28 97
Zinoviev 53 43 29 96
Bogetti 51 44 30 95
Puck 64 30 31 94
Cuntze 62 32 31 94
Tsai-B 57 36 32 93
Table 2.7: Summary of the grades scored by each failure theory

2.8 Fracture-mechanics Approach to Composite Failures

The fracture-mechanics based approach was initially suggested by Griffith [1921], who
suggested that a crack would propagate if the available energy exceeds the energy required to
create the surfaces of the crack. Irwin [1948 and 1958] modified this approach and suggested
that a critical elastic strain energy release rate (SERR), or crack extension force G C, describes
the fracture toughness. He calculated the work required to close a crack of length a+δa to a
length a. Fracture toughness might be defined as the ability of the material to resist fracture,
which is an indication of the amount of stress required to propagate a pre-existing crack. SERR
is the quantitative measure of the fracture toughness of the material [Gdoutos et al., 2000].
Figure 2.24 shows the three crack propagation modes: (a) in-plane, (b) shear and (c) out-of-
plane shear.

Page | 40
 Figure 2.24: Basic crack-propagation modes [Wang, 1996]

The total SERR GC is the summation of three components of fracture-mechanics:

‫ܩ‬஼ ൌ  ‫ܩ‬ூ஼ ൅ ‫ܩ‬ூூ஼ ൅ ‫ܩ‬ூூூ஼ (Equation 2.1)

Where GIC, GIIC and GIIC are the respective SERR for the three modes of crack propagation
[Irwin, 1957].

For a mode I failure, SERR may be defined as

ఋ௎ ఋ௎
‫ܩ‬ூ஼ ൌ െ ൌെ (Equation 2.2)
஻Ǥఋ௔ ఋ஺

Where
U is the total energy of the system
B is the uniform thickness of the structure
δa is the minute crack increment length, which results in the change in energy δU and
δA is the change in crack area for the minute crack progression (δA=B. δa)

Various tests have been developed over the years for the fracture toughness characterisation
for different failure modes and are presented in Table 2.8.

Failure Mode Testing Standards

Mode I Double cantilever beam (DCB) test, ASTM D5528 - 01(2007) e3.
End notch flexure (ENF) test [Prel et al., 1989]
Four-point end notch flexure (4ENF) test [Martin and Davidson, 1997]
Mode II End-loaded split (ELS) test [Wang, 1996; Corletto and Bradley, 1989]
Cracked-lap shear (CLS) test [Brussat et al., 1977]
Edge delamination test (EDT) [O'Brien, 1982]
Page | 41
 Arcan test specimen [Arcan et al., 1978]
Modified rail shear test [Lakshminarayana, 1984]
Mode III Edge-crack torsion (ECT) test [Lee, 1993].
Mixed-mode bending, (MMB) test, ASTM D6671/D6671M-06 [Reeder and
Crews, 1990]
Compact tension shear (CTS) test [Rikards et al., 1998]
Mixed Mode
Iosipescu specimen [Bansal and Kumosa, 1995]
Single leg bending (SLB) specimen [Davidson et al., 1995; Sundararaman et
al., 1998]

Table 2.8: Fracture toughness characterisation

Structural delamination and debonds involve all three failure modes. The mixed-mode failure
criteria beneficial for accurate failure prediction. Typical GIC values for crack initiation and
propagation are shown in Figure 2.25.

Figure 2.25: Critical strain energy release rates (SERR) for crack initiation (G1i); Crack propagation (G1c)
[Mouritz et al., 1999]

Mixed-mode fracture occurs in composites during delamination failures. For isotropic

materials, Mode I fractures prevail, whereas for composites, Modes I and II are the dominate

Page | 42
fracture modes. The interlaminar tensile/compressive stress give rise to Mode I fractures and
the interlaminar shear stresses result in Mode II and III fracture modes. The interaction
between shear and tensile fractures further complicates the failure mechanism in composites
[O'Brien, 1984; Lee, 1990]. Delamination is generally dominated by the resin property. As the
resin gets tougher, the delamination becomes less brittle, which may limit the linear analysis of
toughness [Broek, 1982]. Delamination strength is calculated by the strain energy release rate
(SERR) or fracture toughness Gc, based on linear elastic fracture-mechanics (LEFM). For brittle
materials, LEFM is applicable, but for tougher matrices, J-integral, which is based on elastic
plastic fracture-mechanics (EPFM), is often used.

2.9 Delamination Prediction in Composites Using Numerical

Methods
The different methods of predicting delamination failure are summarised in this section.
Techniques based on stress, fracture-mechanics, a combination of stress and fracture-
mechanics, cohesive zone and strain invariants are discussed.

2.9.1 Stress-based Numerical Methods

Stress-based methods are efficient in predicting the initiation or onset of delamination in

composite materials. Widely used failure stress-based criteria are Puck [1998 and 2002],
Grant-Sanders [1982], Choi [1992], Kim and Soni [1984] and Ye [1988]. With these methods,
stresses are compared to the allowable stress of the structure to determine when
delamination has occurred in each load increment.

The basic form of delamination failure theory is

ఙయయ
൒ͳ (Equation 2.3)
௓೟

Where Zt is the through-thickness tensile strength.

Hou et al. [2001] studied the effect of shear stress distribution on delamination of the
structure when the structure was subjected to compressive load. The out-of-plane tension
accelerated the development of delamination, while out-of-plane compression limited
delamination initiation. Hou’s theory suggests that the critical shear stress required to initiate

Page | 43
delamination is much higher when the laminate experiences through-thickness compression.
Shear delamination is eliminated altogether if the compressive load exceeds a critical value.

ଶ ଶ
ߪଷଷ ଶ ߪଶଷ ൅ ߪଵଷ
‫݊݋݅ݐ݈݈ܽ݊݅݉ܽ݁݀݁݅ݏ݊݁ݐݎ݋ܨ‬ሺߪଷଷ ൒ Ͳሻǡ ൬ ൰ ൅ቆ ቇ൒ͳ
ܼ௧ ܵ௧

ଶ ଶ ଶ ଶ ଶ
ߪଵଷ ൅ ߪଶଷ ߪଵଷ ൅ ߪଶଷ െ ͺߪଷଷ
‫ ݊݋݅ݐ݈ܽ݊݅݉ܽ݁݀݁ݒ݅ݏݏ݁ݎ݌݉݋ܿݎ݋ܨ‬ቌቐඨ ቑ ൑ ߪଷଷ ൏ Ͳቍ ǡ ൒ͳ
ͺ ܵ௧

ଶ ଶ
ߪଵଷ ൅ ߪଶଷ
ܽ݊݀ܰ‫ߪ݄݊݁ݓ݊݋݅ݐ݈ܽ݊݅݉ܽ݁݀݋‬ଷଷ ൏ െ ቌඨ ቍ
ͺ

Ye [1988] proposed a failure criterion for delamination based on the interaction between the
normal and shear through-thickness stresses.

ߪଷଷ ଶ ߪଵଷ ଶ ߪଶଷ ଶ

ܶ݁݊‫݊݋݅ݐ݈݈ܽ݊݅݉ܽ݁݀݁݅ݏ‬ሺߪଷଷ ൒ Ͳሻǣ൬ ൰ ൅ ቀ ቁ ൅ ൬ ൰ ൒ ͳ
ܼ௧ ܵ ܵ௧

ߪଵଷ ଶ ߪଶଷ ଶ
‫݊݋݅ݐ݈ܽ݊݅݉ܽ݁݀݁ݒ݅ݏݏ݁ݎ݌݉݋ܥ‬ሺߪଷଷ ൏ Ͳሻǣ ቀ ቁ ൅൬ ൰ ൒ͳ
ܵ ܵ௧

Zhang [1998] modified Ye's criteria into tensile- and shear-failure modes.

ߪଷଷ
ܶ݁݊‫݁݀݋݈݉݁݅ݏ‬ǣ ൬ ൰൒ͳ
ܼ௧

ଶ ଶ
ߪଵଷ ൅ ߪଶଷ
݄ܵ݁ܽ‫݁݀݋݉ݎ‬ǣ ቆ ቇ ൒ ͳ
ܵ௧

Where Zt = through-thickness tensile strength

S = in-plane shear strength
St = interlaminar shear strength.

Although stress-based methods determine crack initiation, they require a finer mesh in the
through-thickness direction to account for a complex stress field. Moreover, it is difficult to
model the crack progression using stress-based methods.

Page | 44
2.9.2 Fracture-Mechanics-based Numerical Modelling

Numerical methods based on LEFM have proven effective in predicting delamination growth in
composites; however, these methods require the pre-existence of an initial crack front. Once a
crack occurs, delamination growth is achieved when the combination of the components of
energy release rate is equal to or greater than a critical value. Various methods have been
used to obtain the delamination growth, such as the virtual crack-closure technique (VCCT)
and J-integral.

The mostly widely employed delamination criteria is the VCCT, proposed by Rybicky and
Kanninen [1977], which is an extension to finite element analysis (FEA) of Irwin's crack-closure
integral. The technique is based on the assumption that the energy released during crack
propagation is the same as the energy required closing the crack, as shown in Figure 2.26.

Figure 2.26: (a) Delamination modelled with bilinear three-dimensional solid elements; (b) Delamination
modelled with bilinear plate/shell type elements [Krueger, 2004]

Page | 45
VCCT is applicable for both shell and solid models based on the SERR. The main drawback of
VCCT is the requirement of the pre-existence of a crack front.

The major difficulty that arises in FEA of a failure using VCCT is the mesh dependency in the
stress-based methods, particularly when the delamination is present. The fracture-mechanics
approaches rely on the definition of an initial flaw or crack; however, in a practical situation,
the location of damage initiation is not obvious. Therefore, stress-based methods can be used
to predict the onset of delamination and fracture-mechanics provides accurate delamination
propagation modelling.

2.9.3 Cohesive Zone or Interface Elements

Cohesive damage models have been developed based on damage mechanics to simulate the
onset and growth of fractures. They do not depend on a predefined crack/defect. The cohesive
elements combine the aspects of strength-based analysis to predict the onset of damage at
the interface and fracture-mechanics to predict the propagation of delamination. Interface
elements are separate FE entities modelled between the substructures of a composite
material as a means of inserting a damageable layer for delamination modelling. Interface
elements can be modelled in various ways, from nodal 2D spring connections [Cui et al., 1993;
Lammerant and Verpoest, 1996] to full 3D solid element formulations [Goyal et al., 2004;
Petrossian and Wisnom, 1998; de Moura et al., 2000]. These elements are designed to
represent the separation at the zero-thickness interface between the layers of 3D elements
during delamination. In addition, these elements are sufficiently stiff in compression to
prevent interpenetration of the delaminated surfaces. The main advantage of the use of
cohesive elements is the ability to predict both onset and propagation of delamination without
previous knowledge of the crack location and propagation direction. However, it is limited to a
very fine mesh and can produce unacceptably inaccurate predictions when large elements are
employed. According to Turon et al. [2007], the cohesive element is an efficient approach to
model a fracture when the crack propagation is known a priori.

Page | 46
2.9.4 Strain – Invariant Failure Theory (SIFT)

Gosse and Christensen [2001] proposed a micromechanical 3D strain-based approach to

predict failures in composites, called the strain-invariant failure theory (SIFT). Strain tensors
extracted from experiments are substituted into strain invariant parameters of J 1 (first
invariant of strain) and εvm (von Mises strain). The critical properties required for this theory
are the critical first invariant of strain and the critical equivalent strain for both the matrix and
fibre reinforcement phases.

‫ܬ‬ଵ  ൌ  ߝଵ ൅ ߝଶ ൅ ߝଷ (Equation 2.4)

ଵ ଶଷ ଶ ଶ ሿ
ߝ௩௠ ൌ ට ሾሺߝଵ െ ߝଶ ሻଶ ൅ ሺߝଶ െ ߝଷ ሻଶ ൅ ሺߝଷ െ ߝଵ ሻଶ ሿ ൅ ሾߛଵଶ ൅ ߛଶଷ ൅ ߛଷଵ (Equation 2.5)
ଶ ସ

Failure occurs at either the fibre or the matrix phase when any of these invariants first exceeds
the critical values.

2.9.5 Summary of numerical methods for delamination failure

A brief discussion about the various numerical methods available for the simulation of
delamination failure is presented. Only VCCT and cohesive zone modelling are relevant to the
current study. In order to predict the failure location, cohesive zone modelling is preferred
over VCCT.

2.10 Non-destructive Composite Evaluation

Non-destructive evaluation (NDE) refers to the family of test techniques that evaluate the
material integrity of a laminate (as opposed to its performance or functionality) by detecting
subsurface flaws without actually damaging the actual test specimen. Composite structures
are non-homogeneous and anisotropic, which makes damage detection in those structures
more difficult. Laminated composite materials can have a widely varying set of material
properties based on the chosen fibres, matrix and manufacturing process. This makes
modelling composites complex and are often non-linear. Reliable damage detection
techniques are needed to maintain the strength and integrity of the structure. Measurements
from the different techniques can be interpreted to provide four possible level of diagnosis:

Page | 47
 x Detection and progression of cracks
x Location of cracks/damage
x Damage force magnitude
x Residual strength/remaining life of the structure

Different commercially available NDE techniques have been developed. The reliability of the
NDE technique is the essential issue; however, a comparison of techniques is only significant if
it refers to a single task. Each NDE technique has its own set of advantages and disadvantages
and some are better suited for a particular application. The following section provides an
overview of the available techniques.

2.10.1 Types of NDE

The types of commercially available NDE methods are:

Visual inspection Tap testing

Ultrasound testing Liquid penetrant testing
Infrared and thermal testing Eddy current testing
Microwave testing Magnetic measurements
Laser interferometry Magnetic particle testing
Acoustic-emission testing X-Ray radiographic testing
Thermographic testing Shearographic testing
Optical fibre sensing Acousto-ultrasonics
The NDE techniques commonly used with the composites are described in Table 2.9.

Type of NDE Damage Type Advantages Limitations

Fatigue Does not require sophisticated - Time consuming
Visual Inspection Cracks equipment Limited accuracy
Delamination Relatively expensive
Well established and understood Point Monitoring
Fatigue Requires coupling
Sensitive to small damage
Cracks Often does not detect -
Possible damage location
Ultrasonics BVID closed crack
Good depth ranging
Corrosion Sensitive to geometry
Relatively inexpensive
Delamination
Possible in-situ monitoring
Detection of small cracks Used mostly for crack -
Possible non-contact testing detection
Fatigue
Eddy Current Does not require coupling Point monitoring
Cracks
Possible data storage Requires specific-
Relatively inexpensive monitoring skills
Page | 48
 Requires calibration
Poor penetration
Fatigue
Cracks Large structures are monitored Not sensitive to small -
Acoustic-
BVID Non-sensitive to geometry damage
emission
Corrosion Possible in-situ monitoring
Delamination
Fatigue Not sensitive to small -
Cracks Fast monitoring damage
X-Ray
BVID Good penetration Not possible for large -
Radiography
Corrosion Relatively inexpensive structures
Delamination
Fatigue Expensive
Cracks Fast monitoring Not sensitive to small -
Thermography damage
BVID Large structures are monitored
Delamination Poor penetration
Fatigue
Cracks Very expensive
Fast monitoring Not well developed
Shearography BVID
Large structures are monitored Passive technique
Corrosion
Delamination
Damage detection in physically - Expensive
Fatigue Passive technique
accessible area
Optical fibre - Cracks Limited sensitivity in -
Possible in-situ monitoring
sensing BVID dynamic mode
High signal-to-noise ratio
Delamination Localised detection
Distributed sensing possible
Susceptible surface -
Fatigue contamination
Acousto- Can yield false alarms
Cracks Non-passive technique
Ultrasonics Signal includes reflected -
Delamination
and scattered modes.

Table 2.9: Summary of selected NDE techniques for damage monitoring and inspection for composites
[Staszweski et al., 2004; Li et al., 2004]

Marine composite structures are generally large and not all NDE techniques can cover such
large areas. Delaminations are generally embedded in the structure, thus penetration is
required. The cost can also affect the application of NDE methods. Ultrasonic devices are
sensitive to geometry and the current thesis is based on curved composites and out-of-plane
load-transfer mechanisms; thus, ultrasonics has limited applicability in the current study.
Among the various forms of NDE technologies presented in Table 2.9, acoustic-emission (AE)
and optical fibre sensing are well suited for composite damage analysis, as they are very well
capable of in-situ failure monitoring/structural health monitoring (SHM).

Page | 49
2.11 Acoustic-Emission
People have observed acoustic-emissions (AE) from the earliest days. Primitive people heard,
and reacted to, cracking stones, fracturing bones, and the crackling of wood in the fire, as
shown in Figure 2.27.

Figure 2.27: (a) using AE as an alarm system; (b) using AE as a forecasting tool [Muravin, 2009]

People used to listen and pay attention to AE in different situations to detect danger [Muravin,
2009], as shown in Figure 2.27. Research on AE began in the twentieth century. Initially a tin
cry (the characteristic sound of tin being bent) was analysed during the twinning of tin and zinc
crystals [Czochralski, 1916]. Later, Chatelier [1923] reported small, high-frequency audible
sounds that could be heard during plastic deformation of an alloy of aluminium, manganese
and copper. A year later, Ehrenfest and Yoffe [1924] observed that the shear deformation of
salt and zinc was accompanied by clicking sounds. The first AE instrumentation was used in
1933 when seismologist Kishinouye [1934] studied the AE from wood fracture. In 1936, Forster
and Scheil created and applied instrumentation for the registration of AE generated during
martensitic transformations. Kaiser [1950] investigated different engineering materials and
reported on the absence of AE in materials under stress levels below those previously applied
to that material. This became the 'Kaiser effect', which is currently used in acoustic-emission
testing. From 1960 onwards, many industries have used AE technology for non-destructive
testing of structures. Today, almost all industries have tested a variety of structures using AE
technique.

Page | 50
2.11.1 Acoustic-emission Analysis

For a loaded structure, elastic energy is produced and stored in the material. When the elastic
energy exceeds the critical value, a crack or defect is generated and a rapid release of elastic
energy is observed. This rapid release of energy is generally termed acoustic-emission [Liptai
and Harris, 1971]. The examination of AE is a useful tool for the sensitive detection and
location of active damage in polymer blends, particle-filled and FRP composites [CARP, 1993;
Bohse and Kroh, 1992]. AE measurement is based on the detection of microscopic surface
movements from stress waves in a material during the fracture process [Bohse, 2000]. The
interpretation of the signals and the evaluation of the damage progression are crucial in
assessing the health of the structure. Key applications of AE technology in the composite
structural context are damage activity monitoring, identification of damage locations using
multiple AE sensors, damage mechanism identification (for example, matrix/fibre failure or
delamination) and strength predictions to assess the residual structural strength.

Successful application of AE technology considers three aspects: acoustic-emission generation,

its propagation (within the material, from the source to sensors) and an active detection
system. AE generation is dependent on the inherent properties of the material and the applied
load. The amplitude of the generated acoustic wave signal decreases over time. This is termed
attenuation, and is caused by absorption, geometric spreading, dissipation of acoustic energy
into adjacent media and dispersion. Decreasing the signal amplitudes complicates the accurate
determination of the arrival time at sensors, which are more distant from the source. A
logarithmic amplitude scale is often used to overcome this difficulty. Once an acoustic wave is
generated and propagated, this elastic signal must be detected, conditioned and recorded for
analysis.

Page | 51
A simple AE acquisition hardware design includes a transducer, preamplifier, band-pass filter,
amplifier and several digital signal processors. Two approaches used in AE analysis are
transient and parametric analysis. A typical AE process chain is shown in Figure 2.28.

External Load
Mechanical Stress
(force, pressure...)

Material
Properties Stimulation of AE Source (such as crack
Material
(strength, Behaviour formation) by release of elastic energy
tenacity...)

Environmental Propagation of Wave

Conditions
(temperature,
environmental Detection of AE Signal
materials...) Sensors
AE Channel (1..N) Personal
Computer
Parametric Channel
(1..M) AE System

Evaluation, Report

Figure 2.28: Typical AE process chain [Hartmut, 2002]

2.11.2 AE Sensors

Piezoelectric transducers are industrial-grade conventional sensors used to sense acoustic-

emissions. A piezoelectric crystal is housed in a suitable enclosure with a wear plate and a
connector. This is mounted on the surface of the structure using an appropriate clamping
mechanism, or may be embedded in the structure itself. The operating principle is that, when
a stress wave is transmitted in the structure, the piezoelectric crystal resonates, converting the
mechanical energy of the elastic wave into electric energy. This energy is measured by the
appropriate A/D (analogue-to-digital) converter in a data-acquisition system. Commercially
available data-acquisition systems can filter, amplify and post-process signals. As the voltages
generated by the piezoelectric sensors are very small, a preamplifier is often used to boost the
signal strength. Some sensors house an internal preamplifier and this usually adds 40 dB of
gain to the signal. A wide range of sensors is commercially available in various sizes and shapes

Page | 52
to accommodate the specimens and testing geometries. AE sensors can be mounted on the
material surface using non-magnetic means, such as elastic ties, tape, tacks, clamps or glue.

Depending on the application, sensors can be either broadband or resonant at a particular

frequency. Broadband sensors capture the entire AE signal with low sensitivity. The drawback
is that they also capture unwanted background noise due to their wide frequency range and
their sensitivity is generally less than that of properly selected resonant sensors. Resonant
sensors are sensitive to a small range of frequencies, which helps eliminate background noise.
Appropriate resonant sensors for particular frequencies need to be properly selected.

A typical AE signal is shown in Figure 2.29.

Figure 2.29: Typical AE signal

Page | 53
Some of the terms used in AE analysis are defined as per ASTM E1316

x Event (AE event): A local material change giving rise to acoustic-emission.

x AE signal: An electrical signal obtained by detection of one or more AE events.

x Channel: The assembly of a sensor, preamplifier, filters, secondary amplifier,

connecting cables and a processor.

x Voltage threshold: A set voltage level such that signals are recognised when they
exceed the specified limit, usually defined in decibels (dB).

x Arrival time: Time of first threshold crossing, usually defined in microseconds.

x Rise time: Time interval between first threshold crossing and the peak amplitude,
usually defined in microseconds.

x Signal amplitude: Absolute value of the peak voltage of the waveform of one or more
AE events, usually defined in dB, using the following equation [ASTM E1316]:
௏
‫ ܣ‬ൌ ʹͲ݈‫ ݃݋‬൬ ൰ (Equation 2.6)
௏ೝ೐೑

Where V is the measured voltage and Vref is an arbitrarily chosen reference voltage.
Generally, a voltage of 1 μV (Voltage generated by 1 mbar pressure of the face of the
sensor) at the sensor output is used in AE as the standard reference voltage.

x Count (AE count): Number of times the AE signal crosses the predetermined threshold.

x Duration (Hit duration): Time difference between the first and last threshold crossing,
shown in milliseconds.

x Energy (AE signal energy): Integral of the squared amplitude over the time of signal
duration, usually defined in euJ (1eu=10-18).

x Hit: The detection and measurement of an AE signal on a channel.

x Signal strength: The area under the envelope of the linear voltage time signal from the
sensor.

2.11.3 AE Data Analysis

For each set of structural tests, a tremendous amount of AE data will be collected—this can be
very difficult to analyse. Thousands of hits are generated for each event and numerous AE
parameters are associated with each hit. Before the analysis, unwanted background noise and

Page | 54
signals from non-structural sources must be carefully filtered out. Various correlation plots are
used to analyse the data and discriminate among the failure modes for composite structures.
Distribution plots and cumulative plots are generally used to analyse and characterise the AE
signal.

x Time-history plots: These can be plotted with any available AE parameter, such as
amplitude, duration, energy released, rise time, or counts v. time. These plots are used to
find trends in the AE parameters when the specimen approaches failure. They can provide
the starting and ending point for characterisation of AE data. An increase or decrease in
the value of AE parameters shows the condition and extent of structural damage. With
composites, many signals can arise under a constant load, showing the crack propagation
or damage progression. A typical energy released v. time plot is shown in Figure 2.30.

Figure 2.30: AE typical time-history AE plot – Duration and Load vs. time

x Cumulative plots: These involve plotting cumulative signal strength v. time for the AE
parameters discussed above. Generally, an increase in the slope of the curve indicates that
damage has begun. A typical cumulative plot is shown in Figure 2.31. Once the damage is
initiated, a steep increase in the plot can be observed, which is generally termed as ‘knee’
in the curve.

Figure 2.31: AE Typical cumulative plot – counts v. time

Page | 55
x Channel Plots: They show the distribution of the detected emissions by channel. This
facilitates in identifying the location and extent of damage when multi channels are used.

2.11.4 Parametric or Feature Set Analysis

Parametric AE analysis is based on the extraction of a number of parameters such as; arrival
time, peak amplitude, rise-time, signal duration, counts or threshold crossings, elastic energy
released during fracture and crack propagation (measured in atto Joules, 1 atto=10-18, referred
as eu) and RMS from individual AE signals. Parametric analysis provides high event rates, which
may be needed for AE testing. In addition, the circuits can measure signals over a wide
dynamic range, which is often quoted to be as high as 80-100 dB. However, the real dynamic
range may be less due to the smaller dynamic range of signal conditioning elements including
preamplifiers and amplifiers [Shull, 2002].

2.11.5 Transient or Modal Analysis or Waveform Digitisation

As the name suggests, waveform digitisation analyses the modes of wave propagation
produced by the AE sources. This analysis can lead to enhanced noise discrimination and
superior source characterisation. In addition, once the waveforms are recorded, they can be
analysed multiple times with different analysis techniques, which are explained later in the
following section. The amplitude of the signal depends on the structural geometry, sensor
type, location and the type of source material. The value of the amplitude is generally more
with flexural loading for out-of-plane source motion compared to the in-plane source motion.
The primary drawback in using the transient analysis is the lack of automated methods to
analyse the modes contained in the signal. In many cases, different modes involve different
frequency components. Hence, some sort of analysis and filtering is needed to minimise these
drawbacks. In order to overcome this hindrance, either time-frequency methods {short time
Fourier transform, Fast Fourier transform (FFT), wavelets, instantaneous frequency, and
Wigner-Ville distribution} or artificial intelligence (AI) {pattern recognition, artificial neural
networks (ANN)} are used. AI methods require extensive data collection and expensive, time
consuming and laborious experiments to develop adequate training data; hence time-
frequency methods may be practical to use.

Page | 56
2.11.6 Summary of AE

AE is generally advantageous as it can provide continuous monitoring of structures in situ,

while the structure is being tested or in a practical application. AE signals can be utilised to
determine the damage source location, extent of damage, type of damage, residual strength (if
the material property and strength are known) and provide structural health monitoring.
Although AE provides a valid analysis of structural failure mechanisms, precautions must be
taken to minimise errors and misleading results; the level of noise must be analysed and its
effect needs to be minimised in the analysis. AE signals may arise from non-damaged sites;
these locations or sources need to be eliminated or minimised during testing. If this is not
possible, then the signals must be eliminated during the data analysis. Human error needs to
be minimised. Propagation of the wave signal is dependent on attenuation, velocity dispersion,
multiple modes of propagation, reflections and more conversions. AE measurement is affected
by sensor location, type, its response, filtering mechanism and the hardware used. If the AE is
measured on small structures, attenuation does not pose a threat to analyse the damage
mechanism of the structure. However, effective post-processing of the data may be required
to discriminate among the various failure modes in composite structures.

2.12 Optical-fibre Sensing

The different types of loads on a marine structure are quasi-static loads (wave loading, vertical
and lateral combined), dynamic loads (slamming load, whipping load and springing load),
thermal effects and the weight of the structure. These loads create a complex state of stresses
in the structure. Stresses in solids cannot be measured directly; however, the strain can be
directly measured and the corresponding stress values are calculated. Acoustic-emission and
optical-fibre sensing are passive forms of in-situ structural health monitoring, as no additional
excitation is applied to the structure other than the operational load. Various researchers
[Davis et al., 2002; Li et al., 2004] have shown that the occurrence of damage in a structure
alters the local strain distribution under structural load due to the changing load path. Thus,
the occurrence and extent of damage could be detected when the measured strain at a given
location significantly diverges from the value expected for a healthy structure.

Page | 57
Strain-based damage detection is not limited by the physical properties of the target structure.
This makes it suitable for a wide range of applications, from traditional metallic structures to
modern fibre-reinforced composites. The major limitation of this technique is that it measures
strain variations at a local scale or near locations of the installed sensors. Thus, the design of
the sensor network requires knowledge of damage locations, which could be obtained from
numerical simulation, experimental testing and previous experience. For accurate strain
measurement, various strain-measuring systems exist, such as the semiconductor strain gauge,
foil gauges, mercury-in-rubber strain gauge, capacitive strain gauge and optical-fibre sensing.
This study focuses on the application of embedded optical-fibre sensors to measure the
localised damage in glass-fibre composites subjected to out-of-plane loading mechanisms.

Another NDE method suitable for damage monitoring of composite structures is optical-fibre
sensing. The primary application of optical-fibre sensors (OFS) has been for data transmission
in the telecommunications industry. Optical fibres can transmit light over a long distance with
minimal loss and the properties of the light inside the fibre are not affected by physical
parameters outside the fibre. This implies that the fibre can be used as both the sensing
element and the communication path for the signal between the sensor and the optical
interrogator. These sensors are lightweight and deliver superior speed, sensitivity and
bandwidth. The key attributes of OFS are immunity to electro-magnetic interference (EMI),
intrinsic or extrinsic placement, water and corrosion resistance, low maintenance, ruggedness,
smaller size, low cost and they can be multiplexed in parallel or in series. Modern OFS are
suitable for the measurement of temperature, pressure, strain, angular rotation speed,
acceleration, curvature, flow, refractive index, and many other parameters [Staszweski, 2004].
The combination of high precision, high measurement speed and a broad wavelength
measurement range is still quite expensive; however, in large measuring systems with many
sensors per interrogator, the cost per channel can drop to a low level. Optical fibres are
steadily becoming more cost effective due to recent advances in telecommunications,
optoelectronic, and micro and nano industries.

Optical-fibre sensors (OFS) offer new capabilities, provide an alternative to electronic sensors
and have a good record in terms of performance, reliability and cost. Development of OFS
focuses on niche areas; the sensors are made to match the performance of electronic sensors
while retaining their inherent advantages. Those advantages of OFS are [Lee and Jeong, 2002]:

Page | 58
 x Optical fibre is immune to electromagnetic fields, high voltage, high temperature,
(~1000ºC for fused silica) and chemical reactions (corrosive, ionisation). These
attributes allow safe deployment of OFS in almost any environment that normally will
harm electronic sensors.

x Optical fibre has small diameter (125 μm), is lightweight and compatible with fibreglass
composites. Therefore, a large amount of OFS can be embedded arbitrarily in
composites with negligible effects on the host structure.

x Performance of bare fibre sensors can be optimized by packaging techniques to meet

specific requirements of a targeted application. Sensor packaging is essential, as
demonstrated in the previous work on improving the performance of pressure
sensitivity [Lin et al., 2005; Cole et al., 2002; Goodman et al., 2008], linearity,
frequency response (of either broadband [Morris et al., 2005] or resonance [Lee and
Jeong, 2010], mechanical strength and temperature stability [Horowitz and Amey,
2001].

x With a variety of multiplexing schemes available, the number of sensors per optical
cable can be multiplied to reduce the cost per sensor. Among the available
multiplexing schemes are time, frequency, wavelength, code, coherence and hybrid
multiplexing [Cranch et al., 2003; Childs et al., 2010].

Some drawbacks also need to be addressed:

x Optical fibres (glass) have very poor elastic properties. They are very brittle and can
easily break. Risk mitigation procedures are required during sensor fabrication,
measurement and installation to avoid sensor failure.

x In-situ measurement is cumbersome and time consuming since measurement setup of

optical-fibre sensor systems involves several different modules, including the light
source, coupler and receiver. In contrast, the well-established electronic sensor has all
its system components on a single board, which is very convenient for on-site
measurement.

Page | 59
2.12.1 Basic Principle of Optical fibre sensors

Optical fibres consist of a high refractive index core, where broadband light travels. A low
refractive index cladding surrounds the core (mostly glass) to reflect light back into the core, as
shown in Figure 2.32. The total internal reflection guides the light is guided inside the core. A
coating (typically acrylate or polyimide) is applied to fibres to protect against environmental
damage. Single-mode fibre (for distances >550 m) has a small core of 9 μm and can transmit
broadband light from 1300–1550 nm. Multimode fibres (for shorter distances) have larger
cores of 62.5 μm, capable of transmitting broadband light from 850–1300 nm.

Sensor Single mode Multimode

Material
Parameter fibre (μm) fibre (μm)
Core 2-10 50-150 Silica-based
Cladding 80-120 100-250 Doped Silica
Coating 250 250 Polymer
Buffer 900 900 Plastic

Figure 2.32: Schematic of the basic structure of an optical fibre [Schwartz, 2009]

Polymer optical fibre (POF) is gaining importance due to its increased strain limit of 10% (the
strain limit of glass fibre is 1%, which makes it very brittle). The application of POF to advanced
composite structures is limited, as the minimum diameter of POF is 1 mm, which is too thick
for composites and may cause stress concentrations. Fundamental optical-fibre sensors used
with composite structures are fibre Bragg gratings (FBG), intensity-based optical fibres,
polarimetric sensors and a range of optical fibre interferometric sensors based on the Fabry-
Perot, Mach-Zehnder and Michelson configurations [Kuang, 2003].

Page | 60
2.12.2 Optical-fibre Modulation Techniques

Commonly used modulation techniques are found in polarimetric and interferometric sensors.

1. Polarimetric sensors: These sensors rely on the principle of coherent interference

between two light beams from a common source travelling along two different
polarisation axes of a common optical fibre, as shown in Figure 2.33.

Figure 2.33: Schematic diagram of a polarimetric sensor configuration [Schwartz, 2009]

2. Interferometric sensors: These sensors rely on the phase change between two
interfering paths of light in response to external stimuli, such as strain, pressure,
temperature or chemical reaction. The different forms of interferometric sensors
shown in Figure 2.34 are

x IFPI: Intrinsic Fabry-Perot interferometric sensor

x EFPI: Extrinsic Fabry-Perot interferometric sensor
x ILFE: In-line fibre etalon
x FBG: Fibre Bragg grating
x LPG: Long period gratings
x TFBG: Tilted fibre Bragg gratings
x DFBG: D-shaped fibre Bragg gratings

Page | 61
Figure 2.34: Schematic configurations of IFPI, EFPI, ILFE, FBG, LPG, TFBG and DFBG sensors [Schwartz,
2009]

Page | 62
2.12.3 Bragg Grating Sensors

Among the various types of sensors, FBG sensors are widely used in the strain monitoring of
advanced composites. The basic principle of FBG technique is shown in Figure 2.35. A periodic
variation etched directly into the core of an optical fibre creates a partially reflective mirror
that returns selective wavelengths in the form of a narrow spectral peak determined by the
grating’s period.

Figure 2.35: Schematic diagram of a FBG

FBG has been demonstrated as a fundamental component in the detection of different types
of measurands including strain, temperature and acoustic-emission. The FBG is formed by
introducing a periodic change of the refractive index in a continuous length of optical fibre,
which can be achieved by exposure of the fibre core to an intense optical interference. Each
index perturbation reflects the incoming field and constructively forms a back-reflected peak
with the centre wavelength determined by the grating parameters. The centre or Bragg
wavelength, λB of an FBG is given by:

O B 2neff / (Equation 2.7)

Where neff is the effective refractive index of the optical fibre and /is the grating period.
Figure 2.36 shows a schematic of a FBG and light interaction inside the FBG.

Page | 63
 Figure 2.36: a) Schematic of a FBG b) graphical representation of light interaction when a broadband
optical field, I0 is launched into FBG resulting the reflected, IR and transmitted, IT field

The wavelength of a FBG is affected by changes of strain and temperature. Both factors change
the effective refractive index of the core as well as the index change period. The shift of Bragg
wavelength due to strain and temperature change is given by Othonos and Kalli [1999]:

§ wn w/ · § wn w/ ·
'OB 2¨¨ / eff  neff ¸¸'l  2¨¨ / eff  neff ¸¸'T
© wl wl ¹ © wT wT ¹ (Equation 2.8)

Where T is the temperature and l is the grating length. The first term of equation (2.8)
represents the strain effect, which can also be written as [Meltz and Morey, 1992]:

'O B O B 1  pe H z (Equation 2.9)

Where εz is the longitudinal strain and pe is the strain optic constant given by:

neff2
pe p12 Q ( p11  p12 )
2 (Equation 2.10)

where ν is Poisson’s ratio and ρ11 and ρ12 are the elasto-optic coefficients. For a
germanosilicate optical fibre, these parameters have the following values: ν= 0.16, ρ11 = 0.113,
ρ12 = 0.252 and neff = 1.482.

The second term of equation (2.8) gives the wavelength shift due to the temperature change
ΔT given by [Meltz and Morey, 1992]:

'OB OB (D /  D n )'T (Equation 2.11)

where αΛ and αn are the thermal expansion coefficient and thermo-optic coefficient,
respectively. For silica, αΛ= 0.55 × 10-6 and αn = 8.6 × 10-6.

Page | 64
With proper calibration and compensation for thermal variations, an accurate localised
measurement of strain at the sensor can be obtained. The length of the sensor (grating) varies
from 1–20 mm to achieve versatile sensing requirements and the grating reflectivity may be
approximately 100 per cent in practice, depending on the length of the fibre [Rao, 1997].

The gratings may be written using the following techniques:

x Split-beam interferometer
x Diffractive optical element phase mask technique and
x Point-by-point method

To achieve a high Bragg reflectivity, the photosensitivity of the pristine fibre is enhanced,
either by doping the core with elements such as Ge or Sn, or by hydrogen loading the optical
fibre via immersion in high-pressure hydrogen. It was observed [Wei, 2002] that FBGs written
in hydrogen loaded fibres have less than 50 per cent of the mechanical strength of FBGs that
have not been hydrogen loaded. To achieve high reflectivity, this loading is unavoidable and
the strength issue must be addressed.

The manufacturing costs of FBGs are still relatively high; current manufacturing technology is
not suitable for mass production. Although FBG manufacturing methods have improved
considerably, current technology only manages to produce one FBG at one time and most of
processes still required manual handling. In addition, running FBG facilities is expensive due to
high electric consumption, expensive ultrahigh purity nitrogen, and the requirement for highly
skilled operators.

2.12.4 Multiplexing

Optical fibres are well suited to multiplexed networks, as each FBG can be written at a unique
wavelength, as done in the telecommunications industry. This gives FBG sensor systems the
important property of being able to simultaneously read large numbers of sensors on very few
fibres, reducing the cabling requirements and simplifying installation. Three different
techniques for multiplexing are:

x Wavelength-division multiplexing (WDM): With WDM, the sensors in the network are
initialised with their own wavelength such that they can only respond to a particular
range of wavelengths.
Page | 65
 x Time-division multiplexing (TDM): TDM demodulates a series of discrete pulses from
individual sensors, which are separated by fibres that are longer than the optical pulse.

x Spatial-division multiplexing (SDM): Here a single light source is required and the
outputs from each sensor are directed to individual detectors; therefore, there is no
multiplexing advantage at the detection end of the system.

2.12.5 FBG Interrogation

A typical FBG sensing system, shown in Figure 2.37, consists of a broadband light source,
coupler (to guide the emissive light to the specimen and the reflected light back to the optical
spectrum analyser), an array of FBG sensors and an optical spectrum analyser (OSA).

Figure 2.37: Data acquisition system for FBG interrogation [Ling et al., 2005b]

The reflected signals from the specimen (sensor) are then recorded by the OSA, in which the
reflection spectrum is observed with a sharp single peak. The reflected wavelength (λB) is
obtained using the equation in Figure 2.31, where λB is the wavelength at maximum
reflectivity, neff is the effective mode index of refraction and /is the grating period [Erdogan,
1997]. Both the grating period and refractive index are dependent on the temperature and
mechanical strain that the FBGs experienced.

Page | 66
2.13 Damage Characterisation of Advanced Composites Using
Optical-fibre Sensors
The sensor signal is compared to the response of a healthy structure (pre-determined data)
and any statistically significant variation is assigned to the presence of damage or damage
progression. Optical-fibre sensor applications have focussed on concrete structures, composite
structures and metal alloys. Composites suffer various failure modes, not seen in metallic
alloys, such as delamination, fibre breakage and matrix cracking. A significant number of
research papers have been published concerning the use of optical-fibre sensors with
advanced composite materials.

2.13.1 Matrix cracking

x [Ussorio et al., 2006]: They embedded a single-mode optical fibre with a 10 mm FBG in a
GFRP cross-ply laminate to monitor crack initiation and development due to matrix
cracking. The spectra reflected in the case of undamaged material exhibit a single peak and
a symmetrical shape. After the formation of a crack at 0.3% strain, the spectra became
skewed, with a broadening of the spectra at higher wavelengths, as shown in Figure 2.38.
Secondary peaks in the spectrum at higher wavelengths became more prominent with
increasing applied strain.

Figure 2.38: (a) A matrix crack grown fully across the coupon; (b) reflected FBG spectra at increasing
applied strains for a coupon with a crack
Page | 67
x [Wang et al., 2005]: A polarimetric optical-fibre sensor was embedded in cross-ply GFRP
laminate to monitor transverse ply crack initiation and propagation in real time. The
combined observation of an extensometer, load recordings and a video recording of the
crack development enable a direct correlation between matrix crack growth past the
sensor and a step-change in the sensor response, as shown in Figure 2.39.

Figure 2.39: (a) Optical output from the sensor when five cracks pass it; (b) Band-pass filtered signal
using FFT band pass range of 100–300Hz

2.13.2 Debonding

x [Xu et al., 2005]: Experimental and numerical studies were carried out on debonding
detection based on optical-fibre interferometry using a surface-mounted optical fibre
along the length of the repair material in a carbon-fibre reinforced polymer (CFRP)
composite specimen. The location and extent of damage was identified using induced
perturbations on the plot of the fibre integral strain v. load position, as shown in Figure
2.40. The load was applied at different positions and the integral strain (or total
elongation) was measured with optical-fibre interferometry. The technique was simple and
suitable for in-situ damage monitoring.

Figure 2.40: (a) Integral strain v. time plot from the interferometer; (b) a detailed view of the jump

Page | 68
2.13.3 Crack Monitoring in a Bonded Repair System

x [McKenzie et al., 2000]: They embedded an optical fibre with an array of FBGs for in-situ
monitoring of crack propagation beneath the bonded repair systems on cracked
aluminium skins. The FBGs were multiplexed using a combination of wavelength and
spatial techniques using a tuneable Fabry-Perot filter to track individual gratings. The
sensors were located appropriately to monitor the changing stress fields associated with
the propagation of cracks beneath the bonded patch. 3D FEA was supplemented with the
experimental study. During crack propagation, the strain gradient was monitored to assess
the extent of damage. In addition to this test, they conducted tests on graphite epoxy
doubler, bonded to both the upper and lower surfaces of the specimen to evaluate
delamination damage in both the adhesive and the composite doubler. A range of disbond
and delamination lengths were considered: 0, 26, 31, 36, 41 and 46 mm from the edge of
the patch.

x [Li et al., 2006]: They conducted experimental and numerical investigations on marine
composite structural joints with various lengths of artificial disbonds. A methodology was
developed for health monitoring of composite marine T-joints based on strain
measurements using embedded FBG under operational loads, as shown in Figure 2.41.

Figure 2.41: (a) T-Joint before lamination; (b) comparison of bond-line longitudinal strain distribution
from FEA, FBG and experimental results

They argued that the presence of disbonding significantly altered the local strain
distribution, and that the damaged-induced strain anomalies could be successfully
detected using the embedded sensors. Using the localised strain gradient variations, the
location and intensity of damage was monitored.

Page | 69
2.13.4 Detection of Transverse Cracks

x [Okabe et al., 2000]: FBG sensors were used to detect transverse cracks, which cause non-
uniform strain distribution within the gauge length in CFRP cross-ply laminate, as shown in
Figure 2.42. With an increase in the crack density, the shape of the reflection spectrum
was distorted. The intensity of the highest peak became small, some peaks appeared
around it, and the spectrum broadened. As the crack density approached close to
saturation, the spectrum became narrow again and the highest peak recovered its height.
The changes in the spectrum were thought to be due to the strain distribution caused by
the transverse cracks in 90° ply. The occurrence of transverse cracks was detected from
the changes in the reflection spectrum. One of the practical issues they encountered was
that during the early stages of damage initiation, the spectrum split into two peaks due to
the transverse thermal residual stresses.

Figure 2.42: (a) Embedding an FBG sensor in a composite cross-ply laminate; (b) crack density and stress
as a function of strain measured from a strain gauge and FBG

x [Takeda et al., 1999]: Takeda is one of the key researchers in the usage of optical fibres for
crack detection. In this study, they tested flat coupons embedded with plastic optical fibre.
It was concluded that the optical power of the fibre decreases nonlinearly after the
initiation of transverse cracks when the specimen is loaded by tension or bending.

2.13.5 Delamination

x [Grouve et al., 2007]: The influence of delamination on the resonant frequencies of

carbon-polyetherimide flat composite specimens was studied using artificial
delaminations (16 μm foil polyimide). A clear shift in the resonance frequency was
observed due to the presence of delaminations. The FBGs could measure the strains up to
Page | 70
 950 Hz. At higher frequencies, the strain variation was too small for the FBG sensors to
measure. It was also observed that the standard deviation of the FBG sensor before and
after manufacturing the specimen were one and 3.7 με, respectively, demonstrating the
negative influence on the signal resolution during specimen manufacturing.

x [Leung et al., 2005]: An interferometric sensor was used to detect the delamination in
CFRP composites with artificial Teflon inserts, as shown in Figure 2.43. A quasi-impulse
load was applied at different points along the specimen length. When delamination
occurred, a sudden jump in the magnitude of the integral strain was observed. It was
shown that the location and extent of damage could be deduced from the points where
the slope of the integral strain v. load position curve shows a sudden change.

Figure 2.43: (a) Experimental setup; (b) integral strain v. load position for an experimental FEA
comparison

x [Ling et al., 2005b]: They conducted FEA and experimental three-point bending tests on
GFRP composites with different edge delaminations in the through-thickness direction for
eight different stacking sequences, as shown in Figure 2.44. An optical fibre with FBG
sensors was embedded to characterise the delaminations in the specimen using a static-
strain method. An abrupt change in the strain gradient due to the delamination caused
deformation of the reflection spectra.

Page | 71
Figure 2.44: (a) Composite specimen with an embedded FBG sensor; (b) reflection spectra from the FBG
sensor with an increasing applied load

x [Sorensen et al., 2007]: Non-uniform strains resulting from the propagation of mode I
delamination in DCB specimens were measured. They embedded FBG sensors with a
length of 22/35 mm in carbon fibre–polyphenylene sulphide composite specimens.
The residual strains were used as the reference data. Whenever the delamination
crack grew to a desired length, the FBG readings were compared to the reference
data. The measured distribution of strains in the FBG sensor and numerical modelling
were used in an inverse procedure to determine the actual distribution of bridging
stresses during delamination.

x [Takeda et al., 2005]: They developed a technique for the quantitative evaluation of
delamination length in CFRP laminates using Lamb wave sensing with embedded FBG
sensors.

Figure 2.45: Response of FBG sensors to (a) uniform; (b) non-uniform strain distribution

Page | 72
 Figure 2.46: (a) Specimen with embedded FBG sensors; (b) transverse crack evolution; (c)
change in wavelength distribution of reflected light from the FBG sensor

A piezo-ceramic actuator generated Lamb waves in a CFRP laminate. During wave

propagation in the laminate, FBG sensors detect the transmitted waves, which
correspond to the interlaminar delamination. When the Lamb waves passed through
the delamination, a reduction in the peak amplitude (reflectivity) was observed and a
new wave mode emerged, as shown in Figures 2.45 and 2.46. This was confirmed with
FEA simulation. Analysing the changes in the amplitude ratios and the arrival time of
the new mode, they were able to evaluate the delamination length quantitatively.

x [Kesavan et al., 2008]: Fourteen optical-fibre sensors were embedded in GFRP

composite T-joints with artificial disbonds (Teflon inserts) to detect the damage
mechanism, as shown in Figure 2.47. The sensors were able to detect the presence of
multiple delaminations and determine the location and extent of all delaminations
present in the T-joint structure, regardless of the load (angle and magnitude) acting on
the structure. Two dimensional FEA, using MSC.NASTRAN, was conducted to study the
strain behaviour of T-joints with multiple delaminations.

Page | 73
 Figure 2.47: (a) T-joints embedded with artificial delaminations during manufacturing; (b) strain
distribution along the bond-line of a T-joint with embedded delaminations

2.13.6 Effects of Embedded Optical Fibres on Structural Integrity

It is a general assumption that the strain measured by embedded FBG sensors is the same as
the strain experienced by the host material. Nevertheless, in practical applications, not all the
strain is transferred to the sensor. Some energy may be damped by the protective coating,
spray adhesives used and other impurities. Successful and accurate strain measurement
depends on factors such as sensor network design, laminate manufacturing process and sensor
location. The strain needs to be successfully transferred from the fibre/resin host to the
core/cladding of the optical fibre through the primary resin-rich region, which acts as an
interface. Past studies [Ling et al., 2005a] concluded that the strength and fracture behaviour
of the structure could be significantly affected by improper alignment and placement of optical
fibres in the laminate. They concluded that utilisation of embedded optical fibres for damage
detection is accurate and reliable if the interaction between the optical fibre and the
delamination is known. Despite their small size, compared to the typical diameter of the
reinforcement fibres, the diameter of optical fibre is large enough to induce stress
concentrations and geometric discontinuities in the structure. Previous studies on the effect of
embedding optical fibre in the mechanical properties [Jensen et al., 1992; Lee et al., 1995; Mall
et al., 1996; Choi et al., 1997; Shivakumar and Emmanwori, 2004; Silva et al., 2003] concluded
that structures where the optical fibres are embedded in the mid-plane of the composite
laminate have the greatest effect. The severity depends on the stacking sequence,
manufacturing process and type of reinforcement (cross-plied, weaved, short fibres or
unidirectional) used.
Page | 74
2.13.7 Summary of Optical-fibre Sensors

Studies have been conducted using optical-fibre sensors to evaluate the damage mechanism in
composites. The damage could be qualitatively analysed by using the damage detection
schemes, as presented in Table 2.10.

Type of Damage Researchers

Matrix cracking [Ussorio et al., 2006; Wang et al.2005]
Debonding [Xu et al., 2005]
Crack monitoring in
[McKenzie et al., 2000; Li et al., 2006]
bonded repair system
Detection of
[Okabe et al., 2000; Takeda,1999]
transverse cracks
[Le Blanc, 1992; Measures et al., 1989; Martin et al., 1997; Chang and
Impact damage
Sirkis, 1998; Pearson et al., 2007; Kitade et al., 1995]
Fatigue induced
[Badcock and Fernando, 1995; Lee et al., 2001; Rodrigo et al., 2009]
damage
[Rowe et al., 1986; Everall et al., 2000; Davies et al., 1994; Froggatt
Strain measurement
and Moore, 1998; Tao et al., 2000]
[Leung et al., 2005; Elvin and Leung, 1997; Kesavan et al., 2008;
Delamination Grouve et al., 2007; Ling et al., 2005; Sorensen et al., 2007; Takeda et
al., 2005]

Table 2.10: Summary of FBG studies.

It is evident that the failure mode of composites could be determined by analysing the change
in wavelength distribution, change in strain rate and wavelength shift. This technique of
damage detection requires a low level of signal processing and yields highly accurate results. In
addition to strain variation, the FBGs could also provide information on temperature change.
Although research has been carried out to determine the damage mechanism, some failure
modes, such as matrix cracking, delamination, transverse cracks and crack propagation, have
not been investigated to characterise the damage based on the sensor response.

2.14 Acoustic-Emission Measurement with Optical-fibre Sensors

Conventional AE detection technology is based on piezoelectric transducers that are highly
sensitive in the frequency range of interest, are robust and well established. Piezoelectric
transducer sensitivity can be maximized by making transducers with a resonant frequency
Page | 75
within the frequency range of interest. However, piezoelectric transducers have certain
drawbacks. Firstly, highly sensitive piezoelectric transducers are made possible by using a
resonant frequency, although at a cost of losing information from the AE frequency content.
Secondly, they are susceptible to interference from strong electric fields and cannot be used in
some harsh or hazardous environments. Finally, their relatively large dimensions make them
unsuitable for embedding in composite structures [Chen et al., 2008].

The potential of optical fibres in acoustic sensing has been recognised since the mid-1970s.
Subsequently, manufacturers and researchers have actively engaged in improving and
discovering new technologies for optical systems for sensing. Optical-fibre sensors can sense
acoustic-emissions from stress waves generated in a structure during the damage mechanism.
Perez et al. [2001] performed a series of experiments to study the feasibility and limitations of
FBG sensors for detecting acoustic-emissions during damage processes. Read et al. [2002]
conducted experimental investigations on carbon-fibre composites by embedding and surface
mounting optical-fibre sensors. Other researchers [Pappu et al., 2009; Kim et al., 2004;
Vandenplas, 2005; Chen et al., 2008] confirmed that FBG sensors are promising for damage
detection based on the acoustic-emission approach.

In general, optical-fibre acoustic sensors can be classified into two major types: interferometric
type and intensity type. The techniques for acoustic sensing using optical fibre sensors are
briefly reviewed and discussed.

2.14.1 Interferometric Acoustic Sensors

The operating principle of the interferometric or phase-modulated sensor is based on

detection of a relative optical-phase shift induced between two or more optical fields.
Utilisation of interferometer in acoustic sensing has been considered since experiments with
the first optical-fibre acoustic sensor in 1977 by the Naval Research Laboratory Group [Bucaro
et al., 1977]. This is considered the most precise technique and has found use in many
applications, including military sonar [Cranch et al., 2003; Kullander and Vahlberg, 2005],
geophones [Nash et al., 2000], intrusion detection system [Bush et al., 1999; Juan and Henry,
2005], ground traffic monitoring [Furstenau et al., 1998; Scott et al., 1997] and structural
health monitoring [Liang et al., 2009].

Page | 76
 Sensing arm
Photodetectors
Light source Coupler
IS I1
IR I2
Demodulator
Reference arm

Mach-Zehnder Interferometer
Figure 2.48: Optical-fibre interferometric sensing system

The basic setup of an interferometric sensor comprises of interferometer, light source and
sensing head, as shown in Figure 2.48. One arm of the interferometer is assigned as the
sensing head, with another arm as the reference. The reference arm may need to be shielded
from environmental noise so that the detected intensity function can discriminate phase
changes of the optical field due to acoustic pressure. The output intensities detected by the
photodetector can be written as the following functions:

I1 I R  I S  2 I R I S cos IR  IS
(Equation 2.12)

I2 I R  I S  2 I R I S cos IR  IS
(Equation 2.13)

Where IR and IS are the optical intensities and, IR and IS are the optical phase delays of the
reference and sensing path. The optical phase delay can be written by:

I kneff L
(Equation 2.14)

where k is the free space propagation constant given by k=2π/λ, neff is the effective refractive
index and L is the path length. The acoustic wave information can be retrieved by
implementing a certain algorithm on the photodetector signal. This setup can be further
classified into balanced and unbalanced schemes. In the balanced scheme, the optical path
lengths of both arms are closely matched. This allows utilization of a low-coherence length
optical source and the system is not prone to phase noise of the light source. In the
unbalanced scheme, a long path difference (30–200 m) with a high-coherence source is used.
The unbalanced system yields higher sensitivity than the balanced system due to phase

Page | 77
amplification by the long path difference. Other types of interferometers can be used in
realizing a sensor system, including Michelson [Pang et al., 2007], Fabry-Perot [Furstenau et
al., 1998] and Sagnac [Vakoc et al., 1999]. Each interferometer has its own characteristics and
advantages.

A more advanced setup, using a fibre laser and phase modulation scheme, is illustrated in
Figure 2.49. The nonlinear interferometer output shown by equations (2.12) and (2.13) (in
cosine term) can be linear with high dynamics by phase modulating one of the interferometer
arms. Utilisation of a D-shaped fibre Bragg laser with very narrow line width (~20 kHz),
allowing a large path difference, further enhances the pressure sensitivity of the system. An
inline polarisation controller ensures high fringe visibility at the output of the interferometer.
The most successful technique for passive demodulation is the phase-generated carrier (PGC)
technique. In pioneering work, Dandridge et al. [1982] accomplished demodulation by
computation on a quadrature pair from Bessel’s expansion of the phase-modulated signal.
Other techniques include time domain orthogonal sampling [Bush et al., 1997] and a
symmetrical 3×3 coupler (i.e. 120° phase difference of each port) [Brown et al., 1991].

Pump laser WDM coupler DFB fibre laser sensor

diode

Fibre coil
Isolator

Coupler Photodetector
Demodulator

Phase modulator Polarisation controller

P

Figure 2.49: High-performance interferometric sensor arrangement with phase-modulation scheme and
fibre-laser sensor

The overall performance of the sensor depends significantly on the extrinsic packaging and the
demodulation technique. The typical minimum detectable pressure is less than 100 μPa
[Kirkendall and Dandridge, 2004; Nash, 1996; Cranch, 2003] (-80 dB re Pa Hz-0.5), which is much
lower than sea state zero. It is also has a large dynamic range (>200 dB) with good linearity at a
lower pressure range. Since the interferometric sensor was primarily developed for the
hydrophone, it is normally configured with flat frequency response at low frequencies (<30

Page | 78
kHz) [Kirkendall and Dandridge, 2004; Nash, 1996; Cranch, 2003; Kilic et al., 2008; Azmi et al.,
2010]. There are also sensors configured for ultrasonic applications [Fisher et al., 1998].

For continuous monitoring of high-speed signals in composite failure characterisation (up to a

few MHz), the digital signal processor should able to handle large amounts of data at high
speed. This poses a stringent requirement for the demodulator as the typical demodulation
algorithm involves intensive computation. Further, other tasks, such as signal conditioning and
recording, add to the processing load. Another issue related to this technique is polarisation-
induced fading, where the signal level drops significantly due to destructive interference at the
output of the interferometer. Several polarisation-mitigation techniques have been proposed,
such as active-feedback current loop [Matsuo et al., 2006] and the Faraday rotating mirror
[Faxiang et al., 2009], but the addition of these devices makes them cumbersome and
expensive. Therefore, the intensity sensor is preferred, as it avoids complexity and achieves
reasonably sensitivity.

2.14.2 Intensity-type Acoustic Sensor

For intensity sensors, there are a few mechanisms available to modulate light in optical fibres,
including attenuation, reflection, scattering, modal changes, spectral changes and sub-carrier
modulation. The early research works in this area were initiated by Webb et al. [1996] and
Takahashi et al. [2001]. The latter work demonstrated good linearity, attainable within 150 dB
re 1 μPa pressure for a FBG with bandwidth of 2 nm. They also demonstrated a minimum
detectable pressure of 65.3 dB re 1 μPa. Lee and Jeong [2010] reported that resonant response
can be accomplished by two fibre-end bonding methods: a) partially bonded and b) both ends
bonded to the structure. The resonant frequencies can be configured by the unbounded
sensor length. Changing the length produces different standing wave formations on the
sensing head. High acoustic sensitivity has been achieved to about a few Pico strains, and this
could prove beneficial in narrowband applications.

Tsuda et al. [2009] reported an effective desensitisation method against quasi-static strain by
applying a thin polyimide coating. This method allows maintaining the operating point set by
the tuneable laser by making the FBG non-responsive to quasi-static strain while retaining
sensitivity to dynamic strain. The coating technique can reduce the applied pressure by 99 per
cent, tested up to few MPa. For temperature compensation, one technique is to use a FBG pair

Page | 79
that is tightly packaged with a broadband source. The FBG assigned as the reference reflects
constant light intensity from a broadband source to the sensing FBG [Tanaka et al., 2005]. A
relative spectral shift, due to temperature variation, is nullified since both FBGs experience a
similar change. In other research work, Perez et al. [2001] demonstrated the highest
detectable frequency of this type of sensor was 2.5 MHz, which covers almost the entire
frequency range required for failure monitoring. Other reported works based on similar
sensing principles, in different applications, include investigation of spatial response in robotic
structural health monitoring [Wild and Hinckley, 2010], multifunctional FBGs for structural
damage identification [Betz et al., 2006a; Betz et al., 2006b ] and vibroacoustics by
incorporating adaptive filtering for disturbance mitigation [Tosi et al., 2008].

Figure 2.50 shows the setup and working principle of a typical intensity sensor. The lasing
wavelength of the laser diode is tuned prior to the sensing operation at the optimum point.

(a)
Circulator
I2 FBG
Tuneable Laser 1 2
Diode
3

I3 Photodetector
Computer

(b) Initial condition During acoustic sensing

I2 I2 Δλ
Intensity going out
FBG reflection
from port 2
Light spectrum

λB0 λ λB0 λBB1 λ

I3 I3
Intensity going out
from port 3

λ λ

Figure 2.50: Setup for an intensity acoustic sensor

Page | 80
The lasing wavelength is typically set at 50 per cent of the FBG reflectivity, where the slope
transition has a linear response and the distance to the peak lasing provides a sufficient
dynamic range for measurement. At this initial point, the FBG reflects a small amount of
constant optical power, which the photodetector detects as a DC voltage. As the FBG spectrum
is shifted by the acoustic pressure, a higher intensity level is reflected towards the
photodetector. This dynamic optical power encodes the acoustic signal into amplitude
modulation and the photodetector detects it as an AC voltage.

2.15 Structural Changes

The work done in this thesis is of interest in regards to light displacement keel yachts. Figure
2.51 represents the keel Top-hat-stiffener arrangement for a typical yacht.

Keel load

Figure 2.51: Actual Top-hat-stiffener with hull and keel fin

To simplify the joint between the Top-hat-stiffeners and the keel, the hull shell laminate was
removed. As the hull shell laminate bending stiffness would be negligible in comparison to that
of the keel floors, this simplification was deemed valid. The Top-hat-stiffener was turned
upside down to facilitate the experimental testing, as shown in Figure 2.52. The experimental
investigation and numerical study for the rest of the thesis is based on this ‘upside down’
arrangement of the Top-hat-stiffener for keel structures.

The floor boards are generally made of balsa wood. This balsa supports the web of the Top-
hat-stiffener, in order to minimise buckling of the web. The bends of the Top-hat-stiffener fail
(delamination) prior to the web failure. Preliminary FEA simulations showed that the balsa has

Page | 81
negligible effect on the failure behaviour of the Top-hat-stiffener. Hence balsa wood is not
considered in the current study.

The supporting top hat section for keels is continuous. To represent the continuity, the flange
is secured by exerting external force on the flange plates, which is explained in detail in
Chapter 4.

Keel load

Base
Fixture

Figure 2.52: Modified loading condition for the Top-hat-stiffener (Upside down)

2.16 Conclusions
A comprehensive literature survey has been conducted on various aspects of curved
composites, keel failures, failure modes of composites and failure criteria. Failures in curved
composites are mainly due to the interlaminar stresses creating delaminations within the
structure. The out-of-plane load transfer in T-stiffeners and Top-hat-stiffeners is discussed.
Various failure modes were explored and the failures occurring in the keel structures
identified. The need to analyse the Top-hat-stiffeners for keel structures was determined as
critical in minimising future casualties. Although a number of failure criteria are available for
numerical modelling, delamination prediction is limited to techniques such as VCCT and
interface elements.

The study helped in deciding the type of test specimens used in this thesis. Initially, a simple
bend (three different layups) subjecting to opening force is tested and finally Top-hat-stiffener
(two manufacturing process and 6 layups) is tested until failure.

Page | 82
Various types of sensors and sensing techniques were discussed and their applicability to
composite marine structures assessed. Due to reduced laminate thickness and various modes
of failure involved, acoustic-emission and optical-fibre sensors have been identified as the
most appropriate techniques for composite structures. Acoustic-emission is based on stress-
wave detection and optical-fibre sensors are based on both strain variation and stress-wave
detection. Acoustic-emission measurement has not been exploited to its full potential in
damage evaluation of composite structures. The work done on acoustic-emission and optical-
fibre sensors for this thesis addressed a problem that has not received much attention in the
literature - characterising the various failure modes through analysis of stress waves and strain
variation generated by the structure.

The research methodology is given in Figure 2.53.

Page | 83
 Aim
To understand the damage mechanism of Top-hat-stiffeners

Coupon Test
Material
Properties
Finite
Experimental L-Bend
Element specimen
Analysis Type Investigation
Element type Analysis
Boundary -
Top-hat
Conditions
Stiffener
Contact
conditions

Load Drop
Findings Findings
Strength Stiffness Piezo electric
Load Drop Stiffness Strength sensor
Failure location Failure Mode – Acoustic
Normal Stress - Failure Location Emission
distribution
Interlaminar -
Optical fibre
Delamination stress -
sensor – Strain
distribution
anomalies
Failure mode
Optical fibre
sensor
– Acoustic
Emission

Output
Understanding structural behaviour of top-hat-stiffeners under static loads
Understanding the progressive failure of the structure
Determining the reserve strength of the structure
Providing validated numerical model
Prediction of failure mode and location

Figure 2.53: Research methodology

Page | 84
Chapter 3: Material
Characterisation of Marine
Composites

http://www.boatdesign.net/forums/attachments/materials/7601d1150612245-vacuum-
infusion-questions-picture-2035.jpg

Page | 85
3.1 Introduction
By definition, composites contain two or more dissimilar materials that provide the mechanical
properties, which the individual constituents (materials) fail to deliver. The reinforcement
(fibres) and matrix for each material retains its own physical and chemical properties, yet the
combination enhances the structural properties. The overall performance of composites is
influenced by the constituent materials, their distribution and their interaction. The properties
of the composite structure are determined by the properties of the fibre and resin, the ratio of
fibre to resin in the composite (fibre volume fraction) and the geometry and orientation of the
fibres in the composite. Reinforcements are primarily the load carrying members in the
composites and the matrix essentially transfers the load between the fibres. Fibres give
composites their stiffness, strength, thermal stability and other structural properties. The
matrix keeps the fibres in place, and distributes and transfers the load among the fibres.
Metals such as steel and aluminium are generally isotropic, i.e., the mechanical properties
remain same in all the directions; however, the mechanical properties in composites are
direction specific. For example, the longitudinal tensile strength of unidirectional glass
composites is at least 9–10 times higher than the transverse tensile strength (742 v. 72 MPa)
[Appendix A]. Discontinuous fibre composites (chopped strand mat) have comparatively lower
strength than continuous fibre composites, but provide equal mechanical properties in all
directions in the plane of the structure. Similar to metals, composites are sensitive to stress
concentrations; thus, any cut outs, notches, joints, voids and discontinuities need to be
carefully analysed before manufacturing [Tan, 1994].

3.2 Marine Structures

Although naval architecture is one of the oldest technologies in the world, the need for speed
and weight carrying capacity requires the invention of new materials and design. The life cycle
of general marine structures consists of three phases: design, fabrication and operation, as
shown in Figure 3.1. Structural design is based on providing serviceability, manufacturability
and safety.

Page | 86
 Figure 3.1: Marine structure lifecycle [Moan, 2003]

A structure must sustain various types of loads during its life. For example, the forces due to
loading on a typical yacht are shown in Figure 3.2. The area of interest relevant to the current
study is the region where the keel attaches to the hull. The laminate must be thick enough to
withstand high local pressures from the keel bolts. Transversely, the area must be stiff enough
not to let the keel act like a pendulum when beating in a seaway [Moan, 2003].

Figure 3.2: Forces on a yacht [Larsson and Eliasson, 1994]

Page | 87
The loading mechanisms may vary, due to static load, dynamic load and extreme impact loads,
as shown in Figure 3.3. To sustain complex loads in an efficient design, it is vital to understand
the load effects on the structure and the required strength corresponding to the relevant
failure modes. The failure may be an Ultimate failure Limit State (ULS), Fatigue failure Limit
State (FLS) or failure due to global loss of structural integrity due to the Accidental collapse
Limit State (ALS). ALS is often caused by initial damage, typically by an accidental load or
deficient strength [Moan, 2003].

Figure 3.3: Structural design based on direct analysis of load effects and strength [Moan, 2003]

The traditional materials used for the construction of marine structures are wood, steel, and
aluminium [Chalmers, 1994]. Beginning in the 1960s, GRP material gained prominence as a
basic material for boat hulls and other key structures. Since then, the need for the
understanding of its properties, directional strength and methods for manufacturing
composites has become widespread. Traditionally, glass is the basic reinforcement for marine
structures, but exotic laminates include Kevlar and carbon fibres along with glass for high
speed vessels. The core materials include balsa and foam.

Unfortunately, the various failure modes of laminated composites make failure prediction
difficult and unpredictable. The failure mode of a composite is strongly affected by the type of
Page | 88
matrix material used, as well as its compatibility with the fibres. In some cases, the matrix also
serves to protect the fibres from damage such as corrosion and oxidation.

Various factors to be considered for the design of composite structural components are:

x Reinforcement: Carbon, glass, aramid, ceramic, boron, organic fibres and hybrids.
x Reinforcement type: Short-fibre composites (chopped strand mat, for example),
stitched fabrics (bi-axial, double bias, tri-axial, quadri-axial, combination fabrics) uni-
directional, filament wound and woven rovings.
x Resin type: Polyester, vinylester, epoxy and phenolic.
x Core type (sandwich structures): Balsa, polymeric foam, honeycomb {aramid (Nomex),
Kraft, paper} and timber (Western red cedar).
x Mould type: Male (small-scale production) and female (large-scale production) mould.
x Manufacturing process: Hand-layup (HL), vacuum-bagged infusion (VIP), resin-transfer
moulding (RTM), prepreg layup, injection moulding, filament winding, autoclave
moulding and spray-up moulding.
x Curing: Ambient temperature, elevated temperature and pressurised or non-
pressurised curing.
x Finish: Painted, fairing or gel coated.
x Machinability: For example, Kevlar/epoxy has high impact resistance, but it is
relatively very difficult to machine.
x Occupational Health and Safety (OHS) issues: Styrene smell and itching.

Selection of the appropriate factor in each category depends on cost, areal weight, mechanical
strength (including fatigue, impact, tensile, compressive and shear strength), aesthetic aspects,
location, load transfer mechanism, exposed environment, strength deflection, core shear and
skin buckling requirements.

3.2.1 Reinforcements

The reinforcement is the dimensionally controlled constituent (fixed geometry). It can be

particles, fibres or flakes. Fibres may be continuous (unidirectional) or discontinuous (chopped
strand mat). The important dimension of the reinforcement fibre is its length or the length-to-
diameter ratio. Fibres in composites carry the majority of the load and contribute most to the

Page | 89
stiffness in polymer composites. The stiffness of the matrix is typically 1/10 to 1/100 that of
the reinforcement. Common types of fibres used with composites are shown in Figure 3.4.

In marine composites, glass fibres are predominantly used because of their low weight and
cost. Glass fibres are generally produced by pulling molten glass through orifices at a
temperature where the glass has the right viscosity. Unlike carbon or Kevlar fibres, glass fibres
are isotropic, thus do not lose their properties when loaded in a transverse direction.

Figure 3.4: Reinforcement types [Larson and Eliasson, 1994]

For exotic and high-speed vessels, where cost is not a limiting factor, carbon and aramid fibres
are used.

3.2.2 Matrix

The matrix holds the structure in place, distributes the load, protects the fibres and carries
interlaminar shear. A good resin system should possess:

x good mechanical, adhesive and toughness properties

x resistance to chemical spills and environmental degradation
x low shrinkage and co-efficient of thermal expansion
x the ability to be processed easily

Page | 90
Although different resin systems are used in the marine industry, commonly used resin
systems are polyester, vinylester and epoxy.

3.2.2.1 Polyester
Polyesters are unsaturated thermoset resin systems, and are most widely used in the marine
industry for the manufacture of small vessels such as dinghies, yachts and workboats. Up to 50
per cent of the volume of polyesters is styrene, which helps reduce the viscosity and ease
handling during processing. The properties of polyester resins depend on the cross-link
density. The key advantages of these resins are low cost, low viscosity and fast cure time; the
main drawback is high volumetric shrinkage, which causes uneven depressions on moulded
surfaces, which are undesirable for exterior surfaces requiring high surface finish. Polyester is
also prone to hydrolysis.

3.2.2.2 Vinylester
To reduce the intensity of degradation of composites in a marine environment, vinylester is
preferred over polyester. Vinylester resins have excellent chemical resistance and tensile
strength, low viscosity and fast curing. They have a volumetric shrinkage of 5–10 per cent.
They are less prone to damage by hydrolysis than polyester resins. For exotic yachts and racing
boats, carbon fibres with epoxy or vinylester may be used. Vinylester resin systems are
combinations of methacrylated epoxy compounds and styrene, wherein styrene is used as a
reactive diluent and the vinylester serves as the cross-linking agent. Cure is achieved by free
radical bulk polymerisation with rates and degrees of cure being highly dependent on the
actual system formulation and cure regime used. Typical vinylester systems contain 20–60 per
cent styrene monomer by weight, which increases overall hydrophobicity and decreases
homogeneity. The network formation process can be described as a combination of three
ongoing processes, homopolymerization of vinylester and styrene separately, and the co-
polymerisation of both processes with progression taking place at different rates in the same
period [Karbhari, 2004].

3.2.2.3 Epoxy
Epoxies are high-performance resins that offer added advantages, such as low shrinkage,
resistance to creep, fatigue, environmental degradation, chemicals and solvents, excellent
adhesion and good electrical properties. However, they are highly viscous, comparatively

Page | 91
expensive, difficult to combine toughness, high-temperature resistant and the material
produces a large amount of smoke in fire. Epoxies have a higher cure time and poor
performance in hot-wet conditions [Talreja and Manson, 2001]. Epoxies are also used as
adhesives, caulking compounds, casting compounds, sealants, varnishes and paints [SP
Systems]. The main drawback of epoxies is that they are expensive, compared to vinylester
and polyester.

3.3 Manufacturing Techniques in the Marine Industry

The quality of an end product depends on the manufacturing process. Composite
manufacturing is a complex process requiring an understanding of the various polymers, their
chemical reactions, gelling time and emissions that may be produced. The common
manufacturing processes in the marine industry are hand-layup, vacuum-bag infusion, resin-
transfer moulding and the use of prepregs.

3.3.1 Wet Hand-Layup (HL)

This is perhaps the oldest and simplest process. It involves low volume, labour intensive
procedures suited for large structures, such as boat hulls. In a typical open mould process, the
mould may be male or female, depending on the production quantity. Female moulds are
suited for moderate production volumes. The fibre mat is impregnated with resin by hand. The
advantages include easy tool change, to accommodate engineering design, no requirement for
curing ovens and semi-skilled workers can be readily trained to do the work. The
disadvantages are more voids, high matrix volume, poor uniformity and comparatively low
mechanical properties [Strong, 1993].

The mould is cleaned and a layer of gel coat (resin coat) is applied. The first layer of fabric is
laid on the mould and resin is applied using a paintbrush or roller. Working it with a
consolidation roller squeezes out air pockets and ensures that the layer is properly wetted, as
shown in Figure 3.5. The sequence is repeated until the desired thickness is obtained. Curing is
generally carried out at room temperature. This process requires extensive labour and
produces low fibre content specimens. For sandwich construction, the manufacturing involves
a two-step process; bonding each skin separately to the core. External heating with infrared
lamps or hot air blowers is sometimes used to accelerate the curing process.

Page | 92
 Figure 3.5: Hand-layup technique for manufacturing composites [Carbon Fibre Guru, 2010]

3.3.2 Vacuum-bag Infusion Process (VIP)

VIP is a closed-mould manufacturing technique that produces higher quality specimens than
hand-layup. Reinforcement are cut to size, placed on the open mould and then enclosed by an
impermeable film commonly referred to as a 'vacuum bag'. The resin is infused under the
vacuum while the volume inside the bag is evacuated, as shown in Figure. 3.6.

Figure 3.6: Vacuum-bag infusion process for manufacturing composite [Tygavac Advance Materials, Ltd.]

The sealing of the bag need to be administered carefully. Any small leaks in the bag could draw
air in, during the infusion process; residual air can cause dry spots and micro voids that
degrade the mechanical properties of the laminate.

Page | 93
3.3.3 Use of Prepregs

To obtain stiffer, stronger and lighter structures, prepregs can be used with a mixture of glass,
carbon or Kevlar reinforcements. The prepregs need to be stored, wrapped and sealed in a
polythene bag at -18°C for maximum shelf life. During manufacturing, they are cut to size and
used for fabrication. The moulded area is generally covered with vacuum bag to remove the air
and to allow the laminate to solidify. To obtain a more thoroughly cured product, the whole
structure needs to be placed in an oven or covered with heat blankets. The temperature and
the duration of curing affect the strength and rigidity of the structure. Generally, to obtain
optimum curing, the structure needs to be cured at 20°C [Larsson and Eliasson, 1994].

3.3.4 Summary of Manufacturing Methods

A comparison of various manufacturing techniques commonly used in the marine industry is
presented in Table 3.1. Cost is the main criteria for the selection of an appropriate
manufacturing technique.

Tool Labour
Method Advantages Disadvantages
Cost Cost
Inconsistent part quality
Low tooling costs
Require heavy labour
Hand- Control over fibre orientation
High levels of emissions Low High
layup Flexibility in resin/matrix choice
Fibre volume fraction
Low raw material costs
cannot be maintained
High fibre volume fraction Wrinkles may be formed in
Vacuum Air voids are virtually eliminated curved areas
Low Medium
Bagging Low styrene emissions Higher skill is needed
Low weight
High strength components Require heavy labour
Prepreg Consistent & high fibre fraction Very expensive Very
High
Layup Control over fibre orientation Low production volumes High
High raw material costs High equipment setup cost
Low cost setup High resin concentration
Small-medium production volume Heavier components
Spray Very
Low technology Low dimensional accuracy Low
Layup Low
Small to large components High emissions
Low raw material costs Inconsistent quality
Complex tooling design
High dimensional tolerance
Higher tooling costs
Intermediate production volume
Preform needs extra
RTM Can produce complex parts High Low
manufacturing process
Higher fibre volume fractions
Complex parts may need
Zero emissions from closed mould
resin flow simulation
Table 3.1: Comparison of various manufacturing techniques [Mazumdar, 2001; Schwartz, 1996b]
Page | 94
3.4 Postcuring effect on composite structures
Low/high level temperature curing promotes completion of the cross-linking process of the
resin and achieves a higher mechanical stability. The increase of cross-linking density improves
the mechanical stability of the material, and the relaxation of the molecular network can
increase its ductility and, thus, the energy absorption during fracture [Pearson and Yee, 1989].
Post-curing also facilitates large-scale deformation involving a substantial portion of the
polymer chains, which enhances the fracture toughness. Different methods of post-curing
used are thermal, microwave, ultraviolet radiation, electron beam and radio-frequency-energy
curing. The results of coupon tests with and without thermal curing are discussed later in this
Chapter.

3.5 Material Characterisation

To achieve better mechanical properties, composite manufacturing must be carried out in a
temperature-controlled environment (16–26o C). Carrying out production trials is advised, as it
helps in overcoming practical issues, such as resin gel time, uniform infusion, location of resin
infusion initiation and optimising the manufacturing process. Standard tests (Table 3.2) are
conducted to obtain mechanical properties such as:

x Physical properties: Density, volume fraction and porosity

x Mechanical: E11, E22, ν12 and G12
x Interlaminar properties: Interlaminar tensile and shear stresses
x Fracture-mechanics: GIC, GIIC and GIIIC
x Specific loading types: Static, fatigue, creep and impact

BS (Obsolete –
ASTM ISO now merged EN AS
with EN)

ASTM 2584
Constituent ISO 1172/7822
ASTM 2374
properties ISO 14127 --- --- ---
ASTM D3171
Tensile ASTM D2585 ISO 3268 AS 1145.4
BS 2782 EN 2597
properties ASTM D3039 ISO 527 AS 3572.13

Compressive ASTM D3410 ISO 8515/604

EN 2850
properties ASTM D6641 ISO 14126 --- ---
Page | 95
 ASTM D790
Flexural ASTM D747 ISO 178
BS 2872 EN 2561
properties ASTM D6272 ISO 14125 ---
ASTM D7264

ASTM D3479
ASTM D3518
In-plane
ASTM D4255 ISO 14129
shear AS 3572.19
ASTM D3846 ISO 15310 --- ---
properties
ASTM D5379
ASTM D7078

ASTM D2344
Interlaminar ISO 4585
ASTM E1922 BS 2782 EN 2563 AS 3572.20
properties ISO 14130
ASTM D6415

Fracture ASTM D5528

ISO 15024
toughness ASTM D6671 --- --- ---

Table 3.2: Various tests on composite material: British Standards (BS), ISO, European Standards (EN),
ASTM, and Australian Standards (AS)

Once the directional strength is obtained, an appropriate safety factor is introduced. The
typical factor of safety for various loading mechanisms is presented in Table 3.3.

Load Action Factor of Safety

Static Short term loads (Tension) 3.0
Static Short term loads (Compression) 2.0
Static long term loads (dry) 4.0
Static long-term loads (immersed) 6.0
Load reversal 5.0
Local buckling (stiffeners parallel to load) 1.5
Column buckling of plate/stiffener combinations 2.0
Buckling (Stiffeners perpendicular to load) 3.5
Table 3.3: Various safety factors [Shenoi and Dodkins, 2000]

3.5.1 Determination of Material Properties

The mechanical properties of the composite material depend on the manufacturing process,
type of catalyst used and post-curing procedure. The basic mechanical properties of the
materials used for this research are determined using the appropriate ASTM standards.
Typically, material characterisation is based on the following aspects: Type of material and
fabrication process.

Page | 96
3.5.2 Raw materials used

The materials studied in this research were selected to be representative of those used in the
construction of composite marine vessels. The specimens were manufactured at EMP
Composites, Sydney, one of the premier boat builders in Australia. All specimens meet industry
standards. The materials used were from Fibre Glass International [www.fgi.com.au]:

x Chopped strand mat (CSM; short-fibre reinforcement) 450 gsm

x Continuous filament mat (CFM) 450 gsm
x Cross ply double bias (DB) (oriented ±45° to the loading axis) 611 gsm
x Unidirectional (UD) (oriented 0° to the loading axis) 450 gsm
x Resin system: Vinylester SPV-6036
x Hardener: Norox Methyl ethyl ketone peroxide (MEKP-925H) 1.5%;
Norox Cumylhydroperoxide (CHM50) 1.5%

A comparison between the two hardeners (accelerators) is presented in Table 3.4

Peak Barcol Hardness @

Initiator Gel time Exotherm temp. (°F) 3 hrs 4 hrs 5 hrs 24 hrs
CHM50 16 min 198 41 50 53 14
925H 13 min 289 28 35 44 53
Table 3.4: Comparison of the two hardeners [FGI, 2002]

3.5.3 Resin Burn-off Test

A resin burn-off test was conducted for DB specimens to determine the fibre weight for hand-
layup and VIP manufacturing processes. The samples were weighed into crucibles, charred
using a gas flame, then heated to 750o C for six hrs, cooled and re-weighed. The fibre weights
for hand-layup and VIP were 50.2 and 68.4%, respectively.

3.5.4 Coupon Tests

x Tensile Test: Tensile specimens were cut for different types of reinforcements. Tabs
were manufactured using a three-layer double bias (611gsm) and vinylester (SPV 6036)
resin system. The tabs were glued to the specimens using FGI R045 epoxy resin base
and FGI H045 epoxy hardener. The test was conducted as per the ASTM D3039-
D3039M-00 standard. The specimens were tested on an INSTRON 50 kN testing
Page | 97
 machine. Longitudinal deflections were measured using 50 mm Instron extensometer.
All specimens were tested under a displacement-controlled setup at the rate of 2
mm/min and the test was terminated when the load dropped to 40 per cent of the
previous load increment. Tensile stress was calculated by the ratio of load and cross-
sectional area and the Young's modulus was determined by the slope of the tensile
stress v. tensile strain plot. The effects of post-curing the laminates at 20°C for 24 hrs,
50°C for 24 hrs and 82°C for 2 hrs were determined.

x Three-point bend test: The flexure properties of the materials were determined by
conducting a three-point bend test, carried out as per ASTM D7264-D7264M [ASTM
D7264 2007] standards. Displacement-controlled loading was applied at the rate of 1
mm/min. Flexure stress was calculated using: ߪ௙௟௘௫௨௥௘ ൌ ሺ͵ܲ‫ܮ‬ሻȀሺͶܾ݄ଶ ሻ. The flexure
modulus was determined by the slope of the flexure stress v. flexure strain plot. The
effects of post-curing the laminates at 20°C for 24 hrs, 50° C for 24 hrs and 82°C for 2
hrs were determined.

3.5.5 Test Matrix

A number of specimens were tested to determine the tensile and flexure properties of the
composite laminate. Tables 3.5 and 3.6 present the test matrix for tensile testing and Tables
3.7 and 3.8 present the test matrix for the three-point bend test. The effects of three post-
curing temperatures (20°C for 24 hrs, 50°C for 24 hrs and 82°C for 2 hrs) on the tensile and
flexure strength were determined.

Post-curing CSM925H DB925H UD925H

20°C 24 hrs 3 3 3
50°C 24 hrs 3 3 3
82°C 2 hrs 4 4 4
Table 3.5: Tensile test for hand-layup(HL) process — test matrix.

CFM CFM DB DB UD-90deg UD-0deg

Post-curing
925H CHM50 925H CHM50 CHM50 CHM50
20°C 24 hrs 6 5 4 4 5 5
50°C 24 hrs 0 0 0 4 0 0
82°C 2 hrs 0 0 0 4 0 0

Table 3.6: Tensile test for vacuum-infused process (VIP) — test matrix.

Page | 98
 Post-curing CSM925H DB925H UD925H
20°C 24 hrs 4 4 4
50°C 24 hrs 4 4 4
82°C 2 hrs 4 4 4
Table 3.7: Three-point bend test for hand-layup(HL) — test matrix.

CSM DB CFM CFM DB UD-90deg UD-0deg

Post-curing
925H 925H 925H CHM50 CHM50 CHM50 CHM50
20°C 24 hrs 4 4 5 5 3 6 5
50°C 24 hrs 3 3 0 0 0 0 0
82°C 2 hrs 3 3 0 0 0 0 0
Table 3.8: Three-point bend test for vacuum-infused process (VIP) — test matrix.

3.5.6 Tensile Test Summary

For hand-layup CSM, DB and UD were tested for three different post cure temperatures: 20°C
for 24 hrs, 50°C for 24 hrs and 82°C for 2 hrs. Only 925H hardener was used for the hand-layup
specimens. Though CSM offers adequate permeability, the binder in CSM may hinder the resin
flow. Hence, for the VIP process, the continuous filament mat (CFM) is used in place of CSM.
The tensile test results are presented in Tables 3.9 and 3.10 for the specimens manufactured
by hand-layup and VIP process. The effects of the two hardeners, 925H and CHM50, were
compared to optimise the VIP process.

Tensile Young's % diff. % diff.

Fibre Hardener Post-curing Stress Modulus Stress Modulus
(MPa) (MPa) (MPa) (MPa)
CSM 925H 20°C 24hrs 176.30 9607 0.00 0.00
CSM 925H 50°C 24hrs 144.41 9961 -18.09 3.68
CSM 925H 82°C 2hrs 143.97 10161 -18.34 5.77
DB 925H 20°C 24hrs 75.59 7872 0.00 0.00
DB 925H 50°C 24hrs 115.75 9508 53.13 20.78
DB 925H 82°C 2hrs 97.88 9001 29.49 14.34
UD-0deg 925H 20°C 24hrs 443.81 23576 0.00 0.00
UD-0deg 925H 50°C 24hrs 461.40 22512 3.96 -4.51
UD-0deg 925H 82°C 2hrs 434.86 22849 -2.02 -3.08

Table 3.9: Hand-layup tensile test summary

Page | 99
 Tensile Young's % diff. % diff.
Fibre Hardener Post-curing Stress Modulus Stress Modulus
(MPa) (MPa) (MPa) (MPa)
CFM 925H 20°C 24hrs 137.82 8344 N/A N/A
CFM CHM50 20°C 24hrs 111.43 7074 N/A N/A
DB CHM50 20°C 24hrs 103.18 6703 0.00 0.00
DB CHM50 50°C 24hrs 123.68 7737 19.87 15.43
DB CHM50 82°C 2 hrs 119.45 7928 15.77 18.28
UD-90deg CHM50 20°C 24hrs 72.21 15128 N/A N/A
UD-0deg CHM50 20°C 24hrs 747.66 37232 N/A N/A

Table 3.10: VIP tensile test summary.

For CSM laminates manufactured by hand-layup, the tensile strength was reduced by 18 per
cent for post-curing temperatures of 50 and 82°C; however, the stiffness (Young's Modulus)
increased marginally—by 3.68 and 5.77% with post curing at 50 and 82°C, respectively. For DB,
post curing at 50°C increased the stiffness by a massive 53.13% and the strength by 20.78%.
Surprisingly, for UD laminates, post-curing decreased the stiffness by 4.51 and 3.08% for 50
and 82°C, respectively. The strength increased by a marginal 3.96% for 50°C and decreased
slightly (2.02%) at 82°C. These mixed results suggest that post-curing may increase the
stiffness for CSM and DB but actually reduce the tensile strength for CSM.

For CFM tensile test specimens manufactured by the VIP process, the 925H hardener yielded a
higher strength (137.82 MPa) compared to the CHM50 hardener (111.43 MPa). The stiffness
was also slightly higher for 925H specimens. The effects of post curing were determined for DB
specimens for three different post-cure temperatures. The strength increased by 19.87 and
15.77% while the stiffness increased by 15.43 and 18.28% for specimens post cured at 50 and
82°C, respectively. CHM50 hardener (103.18 MPa) yielded higher strength than 925H hardener
(95.01 MPa) for DB specimens. Longitudinal (0 deg) and transverse (90 deg) tensile strength of
72.21 and 747.66 MPa were obtained for UD specimens. The detailed results of the tensile test
are given in Appendix A. Graphical representation of the tensile strength and Young’s modulus
for CSM, CFM, DB and UD are presented in Figures 3.7 to 3.12.

Page | 100
 Figure 3.7: CFM and CSM composite tensile strength comparison

Figure 3.8 CFM and CSM composite Young's modulus comparison

3.5.7 Tensile test: Hand-Layup and VIP Manufacturing Process

Comparison

CSM specimens (176.30 and 9607 MPa) yielded higher strength and stiffness compared to CFM
specimens (137.82 and 8344 MPa) when cured at room temperature with 925H hardener.
However, for CFM with VIP, CHM50 provided lower tensile strength and stiffness of 111.43 and
7074 MPa, respectively.

Page | 101
DB specimens manufactured by VIP with CHM50 showed a significant increase in strength
(103.18 MPa) but lower stiffness (7074 MPa) compared to those made with the hand-layup
process.

Figure 3.9: Double bias composite tensile strength comparison

Figure 3.10: Double bias composite Young's Modulus comparison

For UD specimens, a huge increase in the strength (68 per cent) and stiffness (58 per cent) was
observed by using the VIP process with CHM50.

Page | 102
 Figure 3.11: Unidirectional composite tensile strength comparison

Figure 3.12: Unidirectional composite Young's Modulus comparison

The short fibres in CSM layers are dense; they restrict smooth resin flow. Hence a continuous
filament mat (CFM) is generally used to ease the resin flow. The major risk in using CFM is that
it can wrinkle at bends and corners, resulting in decreased strength and stiffness. The difficulty
in resin flow for CSM layers could be reduced by decreasing the amount of catalyst (hardener),
thus, increasing the gel time with VIP process. A significant increase in the tensile strength and
stiffness of the laminates with the VIP process and CHM50 hardener was observed for DB and
UD specimens. CSM was preferred for the VIP process.

Page | 103
3.5.8 Three-point Bend Test Summary

The flexural strength and stiffness of laminates were determined using three-point bend test,
in accordance with ASTM D7264. Test samples were cut from the same panel as for the tensile
test samples. The span and width of the specimen were 110 and 15 mm, respectively.

Flexure Flexure % diff. % diff.

Fibre Hardener Post-curing Stress Modulus Stress Modulus
(MPa) (MPa) (MPa) (MPa)
CSM 925H 200C 24hrs 222.80 1822 0.00 0.00
0
CSM 925H 50 C 24hrs 250.74 2356 12.39 29.31
CSM 925H 820C 2hrs 243.24 2150 9.17 18.00
DB 925H 200C 24hrs 74.67 1169 0.00 0.00
DB 925H 500C 24hrs 127.82 1561 71.18 33.53
DB 925H 820C 2hrs 138.78 1631 85.86 39.52
UD-0deg 925H 200C 24hrs 436.81 4235 0.00 0.00
0
UD-0deg 925H 50 C 24hrs 535.31 4407 22.55 4.06
UD-0deg 925H 820C 2hrs 536.61 4462 22.85 5.36

Table 3.11: Hand-layup three-point bend test summary

The flexural stress and modulus for CSM and CFM made with hand-layup and VIP are
presented in Figures 3.13 and 3.14. The detailed results of the tensile test are given in
Appendix B. For 20°C post-cured CSM/CFM specimens, the performance of VIP with 925H was
around 20 per cent greater than that of the hand-layup specimens. Similarly, for post-cured
specimens, VIP with 925H had a 10 per cent increase in strength. Surprisingly, for VIP with
CHM50, the strength was reduced by 30 per cent. On the other hand, CFM with VIP and 925H
had 46, 26 and 42 per cent increases in stiffness for the three post-cure temperatures,
respectively. With CHM50, the stiffness increased by 29 per cent. VIP with CHM50 provided
higher strength and stiffness for DB and UD specimens as shown in Figures 3.15, 3.16, 3.17 and
3.18.

Page | 104
 Flexure Flexure % diff. % diff.
Fibre Hardener Post-curing Stress Modulus Stress Modulus
(MPa) (MPa) (MPa) (MPa)
CFM 925H 200C 24hrs 265.99 2668 0.00 0.00
0
CFM 925H 50 C 24hrs 273.34 2966 2.76 11.17
CFM 925H 820C 2hrs 277.88 3052 4.47 14.39
DB 925H 200C 24hrs 152.12 1881 0.00 0.00
DB 925H 500C 24hrs 153.85 1814 1.14 -3.56
DB 925H 820C 2hrs 171.07 1945 12.46 3.40
CFM CHM50 200C 24hrs 152.71 2341 N/A N/A
0
DB CHM50 20 C 24hrs 176.90 2869 N/A N/A
UD-90deg CHM50 200C 24hrs 108.44 2467 N/A N/A
UD-0deg CHM50 200C 24hrs 579.01 6748 N/A N/A

Table 3.12: VIP three-point bend test summary

Figure 3.13: CSM/CFM composite flexure strength comparison

Figure 3.14: CSM/CFM composite flexure modulus comparison

Page | 105
 Figure 3.15: Double bias composite flexure strength comparison

Figure 3.16: Double bias composite flexure strength modulus comparison

Figure 3.17 Unidirectional composite flexure strength comparisons

Page | 106
 Figure 3.18: Unidirectional composite flexure modulus comparison

3.6 Conclusions
The two manufacturing processes currently used in the marine industry, hand-layup and
vacuum-bag infusion, were discussed in this chapter. The specimens were post-cured at three
different temperatures—20°C for 24 hrs, 50°C for 24 hrs and 82°C for 2 hrs. Two hardeners
were tested: 925H and CHM50. As expected, the VIP process provided higher strength and
stiffness due to its higher fibre volume fraction; however, large ship structures still must be
manufactured by hand-layup, while small and critical structures are manufactured by the VIP
process. Hence, for the experimental test program in this thesis, all specimens were
manufactured by both hand-layup and vacuum infusion to assess and compare their strengths
and failure mechanisms.

Page | 107
Chapter 4: Experimental
Investigation of curved composites

Page | 108
4.1 Introduction
Most composite structural components used in aerospace and marine applications contain
curved beam sections or form curved panels themselves. Similar to an automobile tyre, Top-
hat-stiffeners undergo various loading mechanisms, including flexure and shear stresses in the
laminate at the steel plate ends, tensile/compressive stress in the web and interlaminar tensile
and shear stresses in the curved region. These can lead to failures such as matrix cracking,
delamination and fibre failure. Inadequate design of the structural members can lead to
catastrophic structural failures. Thus, an experimental program was carried out to assess the
strength performance of the structure and to analyse the failure mechanisms. All specimens
were manufactured in a boat builders’ workplace, hence all specimens manufactured and
tested for this thesis meet industry standards.

The test specimens investigated in this thesis are selected to represent the curved top-hat
section typical of the stiffeners/keel plate arrangements identified in Section 2.15. The
experimental program involves flexural testing of composite L-bend specimens and Top-hat-
stiffeners. Three different types of L-bend specimens were used, labelled L-bend Layups 1-3.
Top-hat-stiffeners were manufactured and tested with hand-layup (HL) and vacuum-bag
infusion processes (VIP), using six different layups, referred to as THS HL Layups 1–6.

4.2 Experimental Investigation of L-bend Laminates

Well-established methods exist to analyse the load-carrying capabilities of flat composite
panels and structures; however, the behaviour of curved composites under external loads
involving out-of-plane load transfer is less well understood. One objective of this investigation
is to estimate the maximum load-carrying capacity of curved composite laminated structures.
Furthermore, an attempt is made to identify the failure mode during damage initiation and
progression.

4.2.1 Experimental Setup

For experimental investigation, composite laminates were manufactured using E-glass fibre
and vinylester resin (FGI – SPV 6036) using hand-layup method and cured at room temperature
for 24 hrs. The specimens were manufactured with three different layups (Table 4.1) with a
uniform thickness of 4.62 mm—each lamina 0.66 mm thick. Layup 1 had seven chopped strand
Page | 109
mat (CSM) layers, Layup 2 had four CSM and three double bias (DB) layers and Layup 3
consisted of two CSM, three DB and two unidirectional (UD) layers. The ply layup sequence for
layers one to seven from the inner surface to outer surface is presented in Table 4.1. The
specimens (Figure 4.1a) each had an arm of 115 mm and a leg of 100 mm, and were cut to a
width of 25 mm. The radius of the bend was 19 mm.

L-Bend Layup 1 - CSM Layup 2 - CSM+DB Layup 3 - CSM+DB+UD

Layer 1 CSM 450gsm CSM 450gsm DB 611gsm
Layer 2 CSM 450gsm DB 611gsm CSM 450gsm
Layer 3 CSM 450gsm CSM 450gsm UD 451gsm
Layer 4 CSM 450gsm DB 611gsm DB 611gsm
Layer 5 CSM 450gsm CSM 450gsm CSM 450gsm
Layer 6 CSM 450gsm DB 611gsm UD451gsm
Layer 7 CSM 450gsm CSM 450gsm DB 611gsm
Laminate Thickness 4.62mm 4.62mm 4.62mm
Normalised fibre weight 1.00 1.15 1.10

Table 4.1: Specimen layup

Figure 4.1: (a) L-bend specimen geometry (b) L-bend experimental setup

The experimental setup of the L-bend specimens is shown in Figure 4.1b. The base of the
specimen leg was secured using the clamping arrangement shown in Figure 4.2a. A three-hinge
test arrangement using an Instron test machine is established to allow bending load (Figure

Page | 110
4.2b). As no load was applied in the transverse direction (z-axis), the effect of second hinge is
negligible. A hole was drilled in the specimen arm and connected to first hinge by an 8 mm
bolt. The loading mechanism was designed such that flexure load is applied and the structure
fails through delamination, due to interlaminar tensile stresses. Appropriate care was taken
while installing the specimens within the grips of the clamp to ensure proper alignment.

Figure 4.2: (a) First hinge fixture (b) Three-hinge setup

The curved laminates were loaded such that the radius of curvature increased and
interlaminar tensile stresses were induced in the inner surface of the curved laminate. As this
mechanism was described by Li and Kelly [2003], the interlaminar tensile stresses are generally
maximum within 20–50 per cent of the thickness, while they are zero at the surfaces. The load
was applied to pull the arm of the specimen, i.e., to straighten it. Peak interlaminar tensile
stress occurred at the mid-width of the specimen as the loading point was at mid-width of the
specimen. Displacement-controlled loading was applied at a rate of 15 mm/min for first four
minutes and thereafter 5 mm/min for next eight minutes to achieve a total deflection of 100
mm. During the static tests, the first audible sign of damage corresponds to a loss of bending
stiffness of the curved portion of the specimen. A small drop in the load-carrying capacity can
be seen in the load-deflection plot. To obtain statistical reliability of the test data, five
specimens of each of the three different layups were tested.

Page | 111
4.2.2 L-Bend Experimental Results

The specimens were numbered L-bend A01 to A05 for Layup 1, L-bend B01 to B05 for Layup 2
and L-bend C01 to C05 for Layup 3, with the sub-notation AE used to note the specimen with
which the acoustic-emission results are compared in Chapter 6. The load-displacement plots
for all five specimens of each layup are shown in Figures 4.3, 4.5 and 4.7. Figures 4.4, 4.6 and
4.8 show the crack initiation and progressive failure at various stages of testing.

4.2.3 L-bend Layup 1 results

The load and corresponding deflection at first failure initiation, together with the ultimate
failure load for all five specimens manufactured using Layup 1 configuration are presented in
Table 4.2.

First failure First failure Ultimate failure Stiffness @ first

Specimen
Load (N) Deflection (mm) Load (N) failure N/mm
L-Bend A01 189.92 65.13 790 2.916
L-Bend A02 59.41 44.40 1266 1.338
L-Bend A03 174.8 76.83 3214 2.275
L-Bend A04 138.91 64.98 2840 2.138
L-Bend A05 162.92 66.27 616 2.458
Table 4.2: L-bend Layup 1 initial and final failure load, deflection and stiffness at first failure

The load-deflection plots for all the five specimens are presented in Figure 4.3. The initial
delamination was observed at 162.92 N at a displacement of 66.27 mm (specimen # L-bend
A05). A crack was developed between the second and third layers, as shown in Figure 4.4. This
crack is represented in the load-displacement plot as a small kink.
1400

L-bend A01
1050
L-bend A02
Initial Failure
Load (N)

L-bend A03
700
L-bend A04
L-bend A05-AE
350

0
0 10 20 30 40 50 60 70 80 90 100
Deflection (mm)

Figure 4.3: L-bend layup 1 load-deflection plot

Page | 112
On further loading the specimen began to straighten, generating excessive interlaminar
stresses at the bend. Outer layers which were subjected to interlaminar compressive stresses,
buckled and the fibres sheared off in bending. The photographic representation of crack
development is shown in Figure 4.4. At this stage, no delaminations were observed between
other plies. During the final stages of loading, a small delamination occurred between the sixth
and seventh layers. This caused further delamination in the specimen and ultimate collapse
occurred at 616 N while the corresponding deflection was 85.58 mm.

Figure 4.4: Progressive failure of L-bend Layup 1 specimens

4.2.4 L-bend Layup 2 results

The second layup, consisting of chopped strand mat (CSM) + double bias (DB) followed similar
failure trend as Layup 1. The failure load and stiffness are presented in Table 4.3.

First failure First failure Final failure Stiffness @ first

Specimen
Load (N) Deflection (mm) Load (N) Failure N/mm
L-Bend B01 376.80 81.67 651 4.612
L-Bend B02 128.61 68.38 552 1.881
L-Bend B03 177.02 72.04 1038 2.457
L-Bend B04 57.02 61.04 574 0.934
L-Bend B05 85.14 59.75 1050 1.425
Table 4.3: L-bend Layup 2 initial and final failure load, deflection and stiffness at first failure

The experimental load deflection plot for Layup 2 is shown in Figure 4.5. An initial crack was
observed between the second and third layers at a displacement of 61.04 mm at a load of 57 N
for specimen L-bend B04.

Page | 113
 800

L-bend B01
600
L-bend B02

L-bend B03
Load (N)

Initial Failure
400
L-bend B04 - AE

L-bend B05
200

0
0 20 40 60 80 100
Deflection (mm)

Figure 4.5: L-bend Layup 2 load-deflection plot

The crack initiation and progression of the specimens is shown in Figure 4.6. Further loading
caused the crack to propagate in the longitudinal direction. Upon further loading, another
major crack occurred between the third and fourth layers. During the final stages of loading,
similar to Layup 1 specimens, the outer layers buckled and sheared at the bend. The double-
bias layer is very flexible, hence large deflection with minimal increase in load is observed. The
specimen failed at 573.6 N with a displacement of 93.03 mm.

Figure 4.6: Progressive failure of L-bend Layup 2 specimens

4.2.5 L-Bend Layup 3 Results

Interestingly, the third layup, consisting of CSM+DB+UD layers had high ultimate strength but
low stiffness until the first ply failure. The failure values for Layup 3 are presented in Table 4.4.

Page | 114
 First failure First failure Final failure Stiffness @ first
Specimen
Load (N) Deflection (mm) Load (N) Failure N/mm
L-Bend C01 182.48 72.62 1806 2.516
L-Bend C02 71.36 59.15 673 1.206
L-Bend C03 212.91 78.37 1074 2.717
L-Bend C04 221.3 73.16 3513 3.025
L-Bend C05 71.63 55.43 3274 1.292
Table 4.4: L-bend Layup 3 initial and final failure load, deflection and stiffness at first failure

The load-deflection plots for the five specimens tested with Layup 3 are presented in Figure
4.7. The post-failure behaviour exhibited a secondary stiffness with reserve strength up to
1800 N. For specimen L-bend C01, the first crack occurred at 182.48 N at a deflection 72.62
mm between the second and third layers, as shown in Figure 4.7. Further loading caused the
crack to grow; another crack appeared later, between the fifth and sixth layer at 427 N and
91.65 mm. Final collapse occurred at 99.96 mm with a load of 1806 N.

1800

L-bend C05
1350
L-bend C01-AE
Initial failure
Load (N)

L-bend C02
900
L-bend C03
L-bend C04
450

0
0 20 40 60 80 100
Deflection (mm)
Figure 4.7: L-bend Layup 3 load-deflection plot

The process of crack initiation and progression is shown in Figure 4.8. Further application of
load beyond the first failure point caused the CSM fibres to shear away in tension, shown as
kinks in the load-displacement plot. The DB layers were more flexible, and remained intact.
Upon further loading (after 91.65 mm deflection), the bend started to vanish and the specimen
began to straighten, thus behaving like a tensile specimen, as shown in Figure 4.7. Only the
unidirectional fibres carried the entire load until final collapse. The addition of unidirectional
fibres in composite curved specimen appeared to be beneficial.

Page | 115
 Figure 4.8: Progressive failure of L-bend Layup 3 specimens

4.2.6 L-bend Test Summary

Three layups of L-bend composite specimens (five each) were tested under displacement
control up to a displacement of 100 mm. The load-carrying capacity, failure initiation and
progression were studied for each layup. The bending load in this study generated more
interlaminar tensile stress than interlaminar shear stress.

The stiffness of the first layup (all CSM) was higher than the other two layups. The second
layup (CSM + DB) experienced reduced stiffness compared to the other two layups. The
addition of the double bias (±45°) layers added more flexibility to the structure. Odd numbers
of layers were chosen to minimise the location of neutral axis exactly between the layers, to
obtain a better failure study. Interestingly, the initial delamination was observed between the
second and third layer, irrespective of the layup. Therefore, this region was assumed to be the
region of highest interlaminar stress.

A considerable amount of scatter is seen for all three layups. This scatter can be attributed to:
(a) inherent issues of the hand-layup manufacturing process, (b) minor defects at the bend, (c)
inconsistent base clamping location and (d) small variations in the width of the specimen
during cutting. The failure modes for each layup were similar to each other. Structural
performance could be improved by manufacturing the specimens using the vacuum-infusion
process.

The performance of Layup 3 (CSM+DB+UD) was greater than that of the other two layups. The
secondary reserve strength of Layup 3, after the initial failure (due to the presence of
unidirectional fibres), helped carrying the load up to 1800 N. Even though the load-carrying
Page | 116
capacity of Layup 1 was greater than Layup 2, the CSM fibres suffered catastrophic brittle
failure. Layups 2 and 3 had DB layers, which managed to hold the structure in place and help
avoid the catastrophic failure seen with Layup 1.

In this section, L-bend structures with one bend subjected to flexure load are analysed.
However, structures such as Top-hat-stiffeners have two bends and the loading and failure
mechanism vary significantly, compared to the L-bend structures. As there is no free end (in L-
bend specimens), it has less deflection and is capable of carrying higher loads than L-bend
structures. The next section focuses on the experimental investigation of Top-hat-stiffeners
with different layups and manufacturing techniques.

4.3 Experimental Investigation of Hand-laid Top-hat-stiffener

Marine structures are subjected to in-plane and out-of-plane loads, and susceptible to both
through-thickness and bending failure modes. Top-hat-stiffeners are the primary load carrying
members, transferring the load from the keel to the hull. With Top-hat-stiffeners, the load is
transferred from the crown, down to the flange by shear load transfer; thus, shear stiffness is a
vital parameter in the structural design. In the structural context, Top-hat-stiffeners are stiffer
than the L-bend laminates due their geometry. The boat manufacturers typically lay ±45° fibre
mats in the web of the Top-hat-stiffener for efficient shear load transfer. The initial failure in
curved composites is generally due to interlaminar failure rather than tensile failure. There is a
general tendency by manufacturers to include a couple of unidirectional layers to improve the
joint tensile strength. The current study focuses on the inclusion of unidirectional fibres along
the loading axis in an attempt to optimise the strength of the stiffener. The study discussed in
this section has been published in two refereed international conference proceedings 1, 2. The
current study involves experimental testing of Top-hat-stiffener manufactured by hand layup
and vacuum infusion process.

1
Raju, Gangadhara Prusty, Don Kelly, Jun Ikeda, David Lyons (2008). Experimental and Numerical Analysis of top hat
stiffeners for Keel Structures. Proceedings of Pacific 2008 International Maritime Conference, Sydney.
2
Raju, Prusty B.G., Kelly D.W., Lyons D., Peng G.D. (2009). Progressive failure analysis of top hat stiffeners for keel
2
Raju, Prusty B.G., Kelly D.W., Lyons D., Peng G.D. (2009). Progressive failure analysis of top hat stiffeners for keel
structures. Proceedings of 17th Annual International Conference on Composites/Nano Engineering (ICCE-17),
Honolulu, Hawaii, USA.
Page | 117
4.3.1 Hand-laid Top-hat-stiffener Experimental test configuration

A series of composite Top-hat-stiffeners are manufactured using the basic materials presented
in Table 4.5.

E-glass - Chopped Strand Mat (CSM) - 451gsm and 225gsm

Fibre E-glass - Double Bias (DB) - 611gsm
E-glass - Unidirectional Fibres (UD) - 461gsm
Resin Vinylester – SPV 6036 – FGI
Hardener Norox CHM-50 – 1.5%

Table 4.5: Basic materials for composite hand-layup manufacturing technique

Typical composite laminated Top-hat-stiffener specimens were manufactured using E-glass

reinforcement and vinylester resin, using HL, and cured under room temperature for 24 hrs, as
shown in Figure 4.9, and sliced to the required dimensions.

Figure 4.9: Top-hat-stiffener manufacturing process

No post-curing was done to replicate the actual manufacturing process of the marine
composite structures. The specimen testing was conducted in the solids laboratory on an
Instron 8805 universal testing machine. The experiments were conducted in a 250 kN load cell
and the load was applied at the rate of two mm/min. The load-deflection plots were
monitored using Bluehill software compatible with the Instron.
Page | 118
The three layups used in this study are presented in Table 4.6.

Layup 1 Layup 2 Layup 3

Layer 1 CSM 450gsm CSM 450gsm CSM 225gsm
Layer 2 CSM 450gsm CSM 450gsm UD 461gsm
Layer 3 CSM 450gsm DB 611gsm CSM 225gsm
Layer 4 CSM 450gsm CSM 450gsm UD 461gsm
Layer 5 CSM 450gsm DB 611gsm CSM 225gsm
Layer 6 CSM 450gsm CSM 450gsm UD 461gsm
Layer 7 CSM 450gsm DB 611gsm CSM 225gsm
Layer 8 --- CSM 450gsm DB 611gsm
Layer 9 --- --- CSM 225gsm
Layer 10 --- --- DB 611gsm
Layer 11 --- --- CSM 225gsm
Layer 12 --- --- DB 611gsm
Laminate Thickness 4.41mm 5.22mm 5.79mm
Relative Fibre weight 1.0 1.29 1.45
Table 4.6: Hand-laid Top-hat-stiffener hand-layup laminate configuration

At least four specimens in each layup were tested in tension until failure. The specimen
dimensions are shown in Figure 4.10.

Figure 4.10: Geometry of the Top-hat-stiffener

Page | 119
Each specimen was 100 mm wide and 200 mm high. The crown was 200 mm long with radii of
19 and 23 mm at the top and bottom bends, respectively. Two 10 mm steel backing plates
below and one 10 mm steel backing plate above the crown were placed and fastened to
replicate the actual keel fixture (Figure 4.11). The flange was secured using clamps and the
displacement load was applied at a rate of 2 mm/min using the centre bolt. The loading
mechanism is similar to that of the load applied by a keel bolt to the hull transverse floor (See
Section 2.15). Special care was exercised when installing specimens within the grips to ensure
proper alignment.

Figure 4.11: Top-hat-stiffener experimental test configuration

4.3.2 HL Top-hat-stiffener experimental results

A series of Top-hat-stiffeners were manufactured using hand-layup for the three different
layup schemes outlined in Table 4.6. To maintain the statistical reliability of the results, at least
four specimens were manufactured and tested in each category. The load-deflection behaviour
and the failure mechanism for each sample are described in this section.

4.3.3 HL Top-hat-stiffener Layup 1 experimental results

Layup 1 consists of all CSM layers and is homogenous throughout the laminate. This layup
consists of randomly oriented short-length fibres and does not have continuous fibres running
along the structure. This makes the structure brittle and prone to catastrophic failure
compared with the other two layups. Four specimens (‘THS HL A02’ to ‘THS HL A05’) were
tested until final failure and Table 4.7 presents the first and final failure values for Layup 1.
Page | 120
 First failure First failure Final failure Final failure Initial Stiffness
Specimen
Load (kN) Deflection (mm) Load (kN) Deflection (mm) kN/mm
A02 17.96 8.57 22.07 15.92 2.096
A03 13.20 6.20 17.11 14.75 2.129
A04 18.20 8.70 19.33 12.78 2.092
A05 11.29 9.11 12.86 13.93 1.239
Table 4.7: Hand laid Top-hat-stiffener – Layup 1 initial and final failure load, deflection and stiffness

The load-deflection plots for Layup 1 specimens are shown in Figure 4.12 for comparison. The
initial delamination failure occurred at a displacement range of 7–9 mm at the corresponding
loading range of 15–18 kN.

25

THS HL A05
20
THS HL A02

15 THS HL A03
Load (kN)

THS HL A04
10

0
0 4 8 12 16
Displacement (mm)
Figure 4.12: Hand laid Top-hat-stiffener load deflection for Layup 1

For specimen ‘THS HL A04’, at least six significant load drops are visible, showing the
delamination progression. The failure initiated at the left crown bend with a deflection and
load of 8.7 mm 18.2 kN, respectively. Later, the undamaged right crown bend carried most of
the load until it suffered failure (delamination) at 9.96 mm deflection and the corresponding
load of 18.1 kN.

As both bends suffered extensive delaminations, further loading caused multiple

delaminations at both bends; the structure finally collapsed catastrophically at the base, at a
load of 19.33 kN with the corresponding deflection of 12.78 mm. The average stiffness of the
specimens (except ‘THS HL A05’) was 2.11 kN/mm. The failure initiation and progression for a
typical specimen of Layup 1 is shown in Figure 4.13.
Page | 121
 Figure 4.13: Hand laid Top-hat-stiffener Layup 1 damage process

4.3.4 HL Top-hat-stiffener Layup 2 experimental results

Layup 2 consists of five CSM layers and three DB layers. The fibres in the DB layers are oriented
to ±45°, but the loading is applied at 0°. Because of this alignment, the laminate is generally
able to sustain large displacement and prevent snapping of the structure into two or more
pieces.

First failure First failure Final failure Final failure Initial Stiffness
Specimen
Load (kN) Deflection (mm) Load (kN) Deflection (mm) kN/mm
B01 23.47 11.91 26.76 16.26 1.971
B02 21.15 9.51 29.83 15.95 2.22
B03 20.69 10.41 28.86 21.22 1.988
B04 22.74 10.03 27.30 19.64 2.267
B05 21.26 9.76 23.52 16.39 2.178
Table 4.8: Hand laid Top-hat-stiffener – Layup 2 initial and final failure load, deflection and stiffness

Page | 122
Five specimens (‘THS HL B01’ to ‘THS HL B05’) were tested until failure; the experimental load-
deflection plots are shown in Figure 4.14 and the failure values are presented in Table 4.8. For
specimen ‘THS HL B04’ the initial crack occurred at displacement of 10.03 mm at 22.74 kN at
the left bend between the second and third layers. Six significant load drops can be seen.

30
THS HL B01
25
THS HL B02

20 THS HL B03
Load (kN)

THS HL B04
15
THS HL B05

10

0
0 5 10 15 20
Displacement (mm)
Figure 4.14: Hand laid Top-hat-stiffener load deflection for Layup 2

The presence of the DB layers produced a large deflection with a minimum increase in load
after the first failure. Multiple delaminations occurred at both crown bends and the structure
failed at a load of 27.30 kN for a deflection of 19.64 mm. All specimens had similar stiffness
around 2.2 kN/mm. The progressive damage of Layup 2 can be seen in Figure 4.15.

Page | 123
 Figure 4.15: Hand laid Top-hat-stiffener Layup 2 damage process

4.3.5 HL Top-hat-stiffener Layup 3 experimental results

Layup 3 consists of three DB layers (611 gsm) and three UD layers (461 gsm), each sandwiched
between six CSM layers (225 gsm). The UD layers were placed at the lower part of the laminate
and DB layers on the upper end. Four specimens (‘THS HL C01’ to ‘THS HL C05’, except ‘THS HL
C04’) were tested until failure. The base of specimen THS HL C05 slipped after 23.56 mm
deflection. An international journal paper has been published based on the failure analysis of
Top-hat-stiffener with this layup3.

First failure First failure Final failure Final failure Initial Stiffness
Specimen
Load (kN) Deflection (mm) Load (kN) Deflection (mm) kN/mm
C01* 15.20 5.59 33.68 32.59 2.719
C02* 11.93 4.59 38.06 34.07 2.599
C03 15.23 5.95 45.29 28.53 2.559
C05* 9.46 4.86 N/A N/A 1.947
* Base slipped
Table 4.9: Hand laid Top-hat-stiffener – Layup 3 initial and final failure load, deflection and stiffness

The experimental load-deflection result for Layup 3 is shown in Figure 4.15 and the
failure/stiffness values are presented in Table 4.9. Four specimens were tested until failure.
For specimen THS HL C03, initial cracking occurred at 5.95 mm for a load of 15.23 kN at a
similar location as for the other two layups.

3
Raju, Prusty B.G., Kelly D.W., Lyons D., Peng G.D. (2010). Top Hat Stiffeners: A Study on Keel Failures.
Ocean Engineering 37: 1180-1192.

Page | 124
A mismatch between the elastic properties of the adjacent CSM and UD layers might have
created excessive interlaminar tensile stress, causing the structure to suffer premature failure
between the second and third layers. Similar to earlier layups, the crown bend failed prior to
the flange bend. One significant behaviour observed was the damage at the bottom two bends
(web-flange interface), which is absent in the first two layups. Continuing application of the
load after the initial failure caused other bends to fail without a significant increase in the load
up to a deflection of 22 mm. Thirteen visible load drops can be identified to assess the
structural damage.

50
THS HL C01
Secondary Stiffness
THS HL C02
40
THS HL C03

THS HL C05 - slipped

30
Load (kN)

20

10

0
0 5 10 15 20 25 30 35
Displacement (mm)

Figure 4.16: Hand laid Top-hat-stiffener load deflection for Layup 3

The progressive damage can be seen in Figure 4.16. At 20.20 mm deflection and 21.24 kN load,
all the four bends began to straighten and all the layers suffered delamination. The structure
resembled a trapezium with relatively sharp corners. The secondary strength mechanism is
triggered at this stage due to the presence of UD fibres. The secondary slope (4.399 kN/mm) in
Figure 4.14 shows the reserve strength—it continued to handle the load beyond 45 kN. During
this stage, slippage at the grips was observed and the test was terminated. The average initial
specimen stiffness (except ‘THS HL C05’) was 2.625 kN/mm.

Page | 125
 .
Figure 4.17: Hand laid Top-hat-stiffener Layup 3 damage process

Two behavioural changes were observed compared to the first two layups. At 20.2 mm
deflection, all the bends completely disappeared and the structure was shaped like a
trapezium. The secondary reserve strength (secondary stiffness shown in Figure 4.12) carried
the load up to 45 kN, beyond which the base slipped due to the high loads.

4.3.6 HL Top-hat-stiffener – comparison of experimental strains

Because of the material and geometric nonlinearity, stresses and strains are different at
different parts of the Top-hat-stiffener. Until now, the focus of the failure process was on the
four curved bends of the Top-hat-stiffener. The flat crown section is sandwiched between the
steel plates, and is not being damaged as the laminate does not carry much of the load (steel
plates take most of the load). The only part of the Top-hat-stiffener that has not been
investigated is the web, between the crown bend and flange bend. This region of the Top-hat-
stiffener is subjected to tensile stress, but due to the difference in the radius of the crown and
flange bends, this tensile stress is not the same throughout the length of the web.

Page | 126
The simplest way to assess the behaviour of the web is to measure the strain at various
locations throughout the length of the web and compare it to the corresponding tensile and
transverse shear strains until first failure. This was achieved by attaching two strain gauge
rosettes (three-gauge 45°) to the specimen (SG1 and SG2) on the inner side of the web (Figure
4.11). A data acquisition system (Hottinger Baldwin Messtechnik, GmbH) was used to record
the strains at the marked location. In this study, the strain gauges were attached to the left
web of specimen ‘THS VIP B04’ with Layup 2, as shown in Figure 4.9. The longitudinal tensile
strain (y-direction) and transverse shear strain (yz-direction) for the two strain gauges were
measured. The normal tensile and transverse shear strains are shown in Figures 4.18 and 4.19.
The strain gauges were fixed at 60 and 130 mm from the base of the Top-hat-stiffener.

7000

6000

5000
SG1 - eyy
Microstrain

4000
SG2 - eyy

3000

2000

1000

0
0 5 10 Load (kN) 15 20 25

Figure 4.18: Hand laid Top-hat-stiffener tensile strain (Hyy) for SG1 and SG2

The normal tensile strain Hyy for SG1 and SG2 were 5873 and 2660 μH, respectively, at the
point of first failure (22.74 kN). Similarly, the transverse shear strain Hyz for strain gauges SG1
and SG2 were 470 and 194 μH. It is evident that the upper half of the web is subjected to
higher tensile stress than the lower half. The shape of the strain curve indicates the material
and geometric non-linear behaviour of the structure. Thus, more care must be taken when
designing the Top-hat-stiffeners for keel floors, as most of the damage occurs in the top half of
the Top-hat-stiffener.

Page | 127
 450

SG1 - eyz
350

SG2 - eyz
Microstrain

250

150

50

-50
0 5 10 Load (kN) 15 20 25

Figure 4.19: Hand laid Top-hat-stiffener Transverse shear strain (Hyz) for SG1 and SG2

The measured strains present an excellent opportunity to validate the numerical model by
comparing the strains at the same location in the numerical mode. This is discussed in detail in
Chapter 5, where the finite element analysis of Top-hat-stiffeners for damage prediction is
presented.

4.3.7 HL Top-hat-stiffener experimental Summary

This section presented the experimental investigation of hand laid Top-hat-stiffeners that fail
primarily due to delamination. An interesting feature of the load deflection response has been
the secondary stiffening of the laminate when unidirectional fibres are included in the
laminates (Layup 3). These fibres do not lose strength due to the delamination when the load
is applied in the axial direction of the 0° fibres.

A considerable amount of observed scatter might be caused due to variations in the surface
finish, geometric symmetry, excess resin on the flat sections of the beam, and may include
variations in the void content. Proper care was taken to minimise these effects during the
manufacturing process.

Page | 128
The specimens were manufactured by hand-layup, these inconsistencies are unavoidable and a
small variation in the specimen geometry can significantly alter the structural behaviour. This
may be the reason why nominally identical specimens exhibit different failure mechanisms.
Other factors affecting the inconsistent behaviour of the structure are slight variations in the
boundary conditions, the location of clamps being slightly misplaced, inconsistent clamping
(tightening) of the specimen, and specimen slippage at higher loads.

The material selection appears to have an effect the load bearing and damage mechanism of
the structure. Structures made of only CSM are prone to catastrophic failure as the material is
brittle in comparison with the other laminates tested. Hence, such a layup might be avoided in
the design of Top-hat-stiffeners. DB layers are very flexible; addition of DB layers is advisable
as they hold the structure in place and maintain structural integrity even though their
contribution to carrying additional load is limited.

Inclusion of unidirectional fibres in Layup 3 showed a secondary stiffness in the load-deflection

plot. In the event of a keel-structural failure in the ocean, the Top-hat-stiffener might still be
able to support the load until further assistance arrives by utilising this reserve strength.

The next section will focus on the experimental investigation of Top-hat-stiffeners

manufactured by the vacuum infusion process (VIP), which has higher fibre volume ratio.

4.4 VIP Top-hat-stiffener

The closed-mould manufacturing process has become increasingly popular as it provides
superior strength, stiffness, high surface finish and virtually no emissions. Hence, to compare
and validate the manufacturing process, the next set of Top-hat-stiffeners were manufactured
using the vacuum-infusion process (VIP).

4.4.1 VIP Top-hat-stiffener manufacturing process

CSM is denser and preferable for hand-layup but continuous filament mat (CFM) is generally
preferred for VIP. However, CFM often wrinkles around bends and makes the structure
unsuitable, thus a small amount of CSM is used with VIP in the current study. The gel time is
increased to assist the resin flow throughout the mould. The basic materials used are given in
Table 4.10.

Page | 129
 E-glass - Chopped Strand Mat (CSM) - 451 gsm and 225 gsm
E-glass - Double Bias (DB) - 611 gsm
Fibre
E-glass - Unidirectional Fibres (UD) - 461 gsm
E-glass – B-axial (BE) - 461 gsm
Resin Vinylester – SPV 6036 – FGI
Hardener Norox CHM-50 – 1.5%

Table 4.10: Basic materials for VIP

Three different layups are tested and the layup configurations presented in Table 4.11.

Layup 4 Layup 5 Layup 6

CSM+DB Thickness CSM+DB+UD Thickness CSM+DB+UD+BE Thickness
Mould Surface (mm) Mould Surface (mm) Mould Surface (mm)
DB-450 0.26 CSM-225 0.3 CSM-225 0.3
CSM-450 0.6 CSM-225 0.3 DB-450 0.26
DB-450 0.26 DB-450 0.26 CSM-225 0.3
CSM-450 0.6 CSM-225 0.3 DB-450 0.26
DB-450 0.26 DB-450 0.26 CSM-225 0.3
CSM-450 0.6 CSM-225 0.3 BE-450 0.465
DB-450 0.26 UD-461 0.26 UD-461 0.26
DB-450 0.26 CSM-225 0.3 CSM-225 0.3
UD-461 0.26 BE-450 0.465
CSM-225 0.3 UD-461 0.26
Total thickness 3.1 Total thickness 2.84 Total thickness 3.17
Table 4.11: VIP Top-hat-stiffener laminate configuration

Although VIP is expensive and requires more complex tools and equipment than hand-layup
technique, the end product has a higher fibre fraction, consistent thickness, has fewer voids
and is a superior quality specimen. A snapshot of the Top-hat-stiffener specimens
manufactured by VIP is shown in Figure 4.20. The specimens are cured under the vacuum bag
for 24 hours at a temperature of 20°C. Later they are sliced to the required dimension, the
edges sanded and cleaned thoroughly with acetone.

Page | 130
 Figure 4.20: Top-hat-stiffener manufacturing with VIP

4.4.2 VIP Top-hat-stiffener Results

Four different specimens were tested for each layup scheme under displacement control of 2
mm/min. An identical test arrangement for top-hat layup using the INSTRON machine is
ensured in this series of experiments. A total of twelve specimens with three layups were
tested to failure to determine the failure behaviour and progression. The specimens were
colour coded with a permanent marker at the thickness of the laminate. During the
delamination/splitting process, it was possible to identify the failure location by measuring the
thicknesses on both sides of the crack, either during testing or during the analysis of various
photographs taken during the test procedure. The results for each layup scheme are presented
in the following subsections.

4.4.3 VIP Top-hat-stiffener Layup 4 experimental results

Layup 4 consisted of five DB-450 gsm and three CSM-450 gsm layers. The load-deflection
behaviour of this layup is shown in Figure 4.21. An initial stiffness of 1 kN/mm is seen until the
first failure (Table 4.12).

Page | 131
 First failure First failure Final failure Final failure Initial Stiffness
Specimen
Load (kN) Deflection (mm) Load (kN) Deflection (mm) kN/mm
D01 9.05 11.42 22.01 25.48 0.792
D02 12.48 11.85 29.57 27.04 1.053
D03 9.54 10.37 22.76 23.20 0.920
D05 10.71 10.99 30.83 24.60 0.975
Table 4.12: VIP-Top-hat-stiffener Layup 4 initial and final failure load, deflection and stiffness

Initial failure of the four specimens tested occurred around 11 mm deflection. For specimen
THS VIP D02, twelve distinguishable load drops are seen on the load-deflection plot in Figure
4.21. Matrix cracking occurred at 12.48 kN and 11.85 mm deflection.

35
THS VIP D01
30
THS VIP D02
25
THS VIP D03
Load (kN)

20
THS VIP D05
15

10

0
0 5 10 15 20 25
Deflection (mm)

Figure 4.21: VIP Top-hat-stiffener load deflection plot for Layup 4

A significant load drop of 2.76 kN occurred at 13.87 mm deflection. This was a crack between
the second and third layers at the radius of the left crown bend. With further application of
load, the structure continued to take up the load until the right crown bend failed due to
delamination and eccentric loading began on one side of the specimen at 15.66 mm (14.48 kN
load). Upon further loading, both the crown bends started to disappear and, at 20 mm (18.91
kN load), the right flange bend suffered a major crack between the fourth and fifth layers. A
secondary stiffness started to appear and, at 23.33 mm deflection (24.22 kN load), the left
flange bend lost its stiffness and delamination occurred between the fourth and fifth layers.
Further loading caused delamination propagation between other layers and the structure

Page | 132
finally collapsed at 27.04 mm d and 29.57 kN The laminate sheared away at the left flange
base below the edge of the steel plate. The damage process is shown in Figure 4.22.

Figure 4.22: VIP Top-hat-stiffener Layup 4 damage process

4.4.4 VIP Top-hat-stiffener Layup 5 experimental results

Layup 5 consisted of two DB-450 gsm, four CSM-225 gsm and two UD-461 gsm layers. The first
and final failure load/deflection of all the four specimens tested is shown in Table 4.13. The
first failure load and deflection were comparable for all the specimens, even though a small
amount of scatter was observed.

First failure First failure Final failure Final failure Initial Stiffness
Specimen
Load (kN) Deflection (mm) Load (kN) Deflection (mm) kN/mm
E01 8.02 10.01 24.26 26.30 1.248
E02 7.61 8.39 22.05 23.42 0.907
E03 8.52 12.54 19.18 30.02 0.679
E04 6.49 7.66 22.08 24.52 0.847
Table 4.13: VIP-Top-hat-stiffener Layup 5 initial and final failure load, deflection and stiffness

The load-deflection plot for Layup 5 is shown in Figure 4.23. As with the hand-laid specimens,
the addition of unidirectional layers created secondary stiffness in the structure. Considerable
scatter is seen at the initial failure point but the failure mechanism and secondary stiffness
process are consistent in all specimens. For specimen THS VIP E01, ten distinguishable load
drops can be seen in Figure 4.23.

Page | 133
 25
THS VIP E01 Secondary Stiffness
20
THS VIP E02

15 THS VIP E03

Load (kN)

THS VIP E05

10

0
0 5 10 15 20 25 30
Deflection (mm)
Figure 4.23: VIP Top-hat-stiffener load deflection plot for Layup 5

The first failure was matrix cracking with a small crack between the second and third layers at
10.01 mm deflection and 8.02 kN load, as shown in Figure 4.24. The failure process was similar
to that of Layup 4—the first major crack occurred at the left crown bend between the second
and third layers in the radial direction.

Figure 4.24: VIP Top-hat-stiffener Layup 5 damage process

At 20.65 mm deflection, all four bends were damaged due to delamination. Until the first
failure, the stiffness was 1.248 kN/mm and further loading caused the bends to straighten. A
secondary stiffness was noticed, with a stiffness of 3.666 kN/mm. After observing a noticeable
secondary stiffness, the final failure occurred at 26.30 mm deflection at a load of 24.26 kN at
the right flange bend, underneath the edge of the steel plate.

Page | 134
4.4.5 VIP Top-hat-stiffener Layup 6 experimental results

Layup 6 consisted of four CSM-225 gsm, two DB-450 gsm, two UD-461 gsm and two BE-450
gsm layers. Similar to the other two layups, four specimens were tested until failure. The
failure load, deflection and initial stiffness for all four specimens are given in Table 4.14.

First failure First failure Final failure Final failure Initial Stiffness
Specimen
Load (kN) Deflection (mm) Load (kN) Deflection (mm) kN/mm
F02 9.19 9.26 23.34 24.36 0.992
F03 10.17 13.82 18.37 29.40 0.736
F04 9.31 7.80 20.77 23.21 1.194
F05 10.32 8.04 21.52 22.82 1.284
Table 4.14: VIP-Top-hat-stiffener Layup 6 initial and final failure load, deflection and stiffness

The initial failure load was around 10 kN and the failure behaviour was similar. Similar to Layup
5, secondary stiffness was observed due to the presence of unidirectional and BE
reinforcements. For specimen THS VIP F02, the initial and secondary stiffness were 0.992 and
2.48 kN/mm. Eleven load drops can be seen from the load-deflection plot (Figure 4.25).

25
Secondary Stiffness
THS VIP F02
20
THS VIP F03
THS VIP F04
15
Load (kN)

THS VIP F05

10

0
0 5 10 15 20 25 30
Deflection (mm)
Figure 4.25: VIP Top-hat-stiffener load deflection plot for Layup 6

The initial failure was matrix cracking at 7.56 mm and the major delamination occurred
between the second and third layers in the right crown bend, as shown in Figure 4.26.
Successive load increments caused all bends to fail due to delamination. Secondary stiffness
began when the bends disappeared and the structure started to take the form of a trapezium.

Page | 135
 Figure 4.26: VIP Top-hat-stiffener Layup 6 damage process

4.4.6 VIP Top-hat-stiffener Test Summary

Based on the experimental observations, most specimens failed at the base (at the flange end)
due to flexure, providing a warning to increase the stiffness and strength at the flange. This
may be achieved by increasing the number of fibre reinforcement layers at the flange region.
The final failure was caused by stress concentration at the edge of the steel washer plates, due
to bending. In structural context, addition of DB layers in the laminate restricts transverse
crack growth in the through-thickness direction of the laminate. The inclusion of unidirectional
layers (for Layups 5 and 6) proved to be beneficial (similar to Layup 3); the secondary strength
helps maintain the structural integrity even after the initial structural failure.

4.5 Comparison of Manufacturing Processes for THS

The performance of two representative layups is compared to examine the potential of VIP
composites as replacements for laminates that are traditionally hand laid. Specimens of both
manufacturing techniques were tested using similar equipment and test conditions. The two
different manufacturing processes with three layups each are summarised in Table 4.15.
Layups 1, 2 and 3 were tested to assess the behaviour of the structure by increasing one type
of reinforcement for each layup—CSM, DB and UD respectively. Layup 4 had CSM+DB and
Layups 5 and 6 had additional UD and BE layers. This assisted in optimising the Top-hat-
stiffener layup with experimental testing; however, the relative fibre weight is not the same
for all layups. Although there is a mismatch in the fibre weights, the VIP specimens exhibited
an increased ultimate load carrying capacity.
Page | 136
 1 2 3 4 5 6
Hand- Hand-
Process Hand-layup VIP VIP VIP
layup layup
CSM+DB+
Layer CSM CSM+DB CSM+DB+UD CSM+DB CSM+DB+UD
UD+BE
1 CSM-450 CSM-450 CSM-225 DB-450 CSM-225 CSM-225
2 CSM-450 CSM-450 UD-461 CSM-450 CSM-225 DB-450
3 CSM-450 DB-611 CSM-225 DB-450 DB-450 CSM-225
4 CSM-450 CSM-450 UD-461 CSM-450 CSM-225 DB-450
5 CSM-450 DB-611 CSM-225 DB-450 DB-450 CSM-225
6 CSM-450 CSM-450 UD-461 CSM-450 CSM-225 BE-450
7 CSM-450 DB-611 CSM-225 DB-450 UD-461 UD-461
8 CSM-450 DB-611 DB-450 CSM-225 CSM-225
9 CSM-225 UD-461 BE-450
10 DB-611 CSM-225 UD-461
11 CSM-225
12 DB-611
Thickness 4.41 5.22 5.79 3.10 2.84 3.17
Relative
1.00 1.30 1.45 1.14 1.01 1.15
Fibre wt.
Table 4.15: Top-hat-stiffener HL and VIP comparison

4.5.1 Resin Burn-off Tests

Burn-off tests were conducted to determine the fibre weight fraction in the laminate for both
hand laid and vacuum bag infused Top-hat-stiffeners. The samples were weighed into
crucibles, charred using a gas flame, then heated to 750°C for six hours, cooled and re-
weighed. The results of the test are presented in Table 4.16. For hand-laid specimens, the fibre
weight is 51 per cent, but for VIP specimens, fibre weight is around 68 per cent.

Manufacturing Layup Layers Relative Fibre Relative Polymer

Process No. Weight Weight
1 Hand-layup Layup 2 CSM+DB 0.516 0.484
2 Hand-layup Layup 3 CSM+DB+UD 0.512 0.488
3 VIP Layup 4 CSM+DB 0.688 0.312
4 VIP Layup 5 CSM+DB+UD 0.644 0.356
Table 4.16: Burn-off test results for Top-hat-stiffener – HL and VIP

Page | 137
4.5.2 Strength and Stiffness Comparison

To determine the influence of the manufacturing technique on the mechanical properties, two
approaches were considered: strength and stiffness based design. In a broader sense, Layups 2
and 4 are compared in Figure 4.27 and Layups 3 and 5 are compared in Figure 4.28.

32
THS HL Layup 2 B01 Hand laid specimens
THS HL Layup 2 B02
THS HL Layup 2 B03
THS HL Layup 2 B04
24
THS HL Layup 2 B05
THS VIP Layup 4 D01
THS VIP Layup 4 D02
THS VIP Layup 4 D03
Load (kN)

16 THS VIP Layup 4 D05

VIP specimens
8

0
0 5 10 15 20 25 30
Deflection (mm)

Figure 4.27: Top-hat-stiffener CSM+DB HL and VIP comparison

Page | 138
 45
THS HL Layup 3 C01
THS HL Layup 3 C02
THS HL Layup 3 C03
36
THS HL Layup 3 C05
THS VIP Layup 5 E01
Hand laid specimens
THS VIP Layup 5 E02
27
THS VIP Layup 5 E03
Load (kN)

THS VIP Layup 5 E05

18

VIP specimens

0
0 5 10 15 20 25 30 35
Deflection (mm)
Figure 4.28: Top-hat-stiffener Layups 3 and 5 comparisons

The stiffness of the VIP specimens is around 60 per cent that of hand-laid specimens. This
difference is mainly due to the reduced cross-section of the laminates from the higher fibre
volume fraction in VIP. The large unemployed structural reserve capacity from the inclusion of
unidirectional fibres must be exploited; however, the stiffness/strength must not decrease
below the designed limit load.

Table 4.17 presents the strength and stiffness values at the first ply failure and final failure
loads for the six layups tested. For strength-based analysis, the flexure stress and shear stress
are determined, while for stiffness-based analysis, the flexure modulus and Young’s modulus
are calculated, where the values of the flexure modulus and Young’s modulus are determined
by coupon tests.

VIP specimens show increased specific strength and specific stiffness, making them suitable for
load-bearing structures, such as keels. The higher fibre fraction reduces the thickness
considerably, and this has huge bearing on the stiffness of the structure.
Page | 139
 Manufacturing First Failure Final Failure
Technique Flex Young’s Strength Strength Stiffness Stiffness Strength Strength Stiffness Stiffness
and mm Mod Mod Flex stress Shear stress Flex Young’s Flex stress Shear stress Flexure Young’s
Layup MPa MPa MPa MPa Mod Mod MPa MPa Mod Mod
1 HL - CSM 4.41 8035 42367 482.1 27.97 0.0272 0.0052 573.70 33.29 0.0324 0.0061
2 HL - CSM+DB 5.22 8159 46557 458.47 31.49 0.0215 0.0038 574.34 39.45 0.0270 0.0047
3 HL - CSM+DB+UD 5.79 13614 77596 215.37 16.41 0.0055 0.001 769.95 58.66 0.0195 0.0034
4 VIP - CSM+DB 3.10 7248 26714 619.56 25.27 0.0551 0.015 1469.60 63.61 0.1388 0.0377
5 VIP - CSM+DB+UD 2.84 9289 40846 541.11 20.22 0.0410 0.0093 1547.24 57.82 0.1173 0.0267
6 VIP - CSM+DB+UD+BE 3.17 11176 51400 552.83 23.06 0.0312 0.0068 1191.19 49.68 0.0672 0.0146

FPF - First ply failure

UF - Ultimate failure
σ=MY/I - Flexure Stress
τmax= VQ/It - Maximum bending shear stress at the crown - steel plate edge (VQ/IT)
Flex Mod - The stiffness calculations is based on the flexure modulus based on three point bend test results (ASTM D3039)
Young’s Mod - The stiffness calculations is based on the Young’s modulus based on tensile bend test results (ASTM D7264)

Table 4.17:THS strength and stiffness comparison based on the flexure modulus and Young’s modulus at the first and ultimate failures

Page | 140
4.5.3 Strength and Stiffness Comparison at 10 kN

In the previous section, the strength and stiffness values were calculated at the respective first
and final failure loads, which are significantly different for each layup. However, for accurate
comparison of the strength and stiffness of the structure, the values are calculated for an
assumed load of 10 kN and given in Table 4.18. It can be seen that the specific stiffness of VIP
specimens is much higher (at least eight times) than that of hand-laid specimens.

Strength Strength Stiffness Stiffness

Layup Specimen Layup My/I τmax M/EI M/EI
Flexure Stress Shear Stress Flex Mod Young’s Mod
1 HL - CSM 132.92 17.01 0.0165 0.0031
2 HL - CSM+DB 80.15 14.37 0.0098 0.0017
3 HL - CSM+DB+UD 58.73 12.95 0.0043 0.0008
4 VIP - CSM+DB 382.67 24.19 0.0528 0.0143
5 VIP - CSM+DB+UD 497.68 26.41 0.0536 0.0122
6 VIP - CSM+DB+UD+BE 357.87 23.66 0.0320 0.0070
Table 4.18: THS Flexure stress, shear stress and laminate stiffness at a 10 kN load

4.5.4 Discussion of the Two Manufacturing Processes

Compared with hand-layup, the VIP samples show expected increase in first-ply failure (FPF)
and ultimate failure (UF) stresses. This arises from the corresponding increase in fibre volume
fraction that VIP produces through greater laminate consolidation and exclusion of resin under
vacuum during infusion and curing process. UF is particularly increased for all VIP samples
compared with HL.

However, there is no consistent increase in Young’s modulus for the VIP samples, compared
with the HL samples. In the case of the CSM+DB there is a reduction in the calculated Young’s
Modulus.

Combined with the marked reduction in cured laminate thickness due to VIP there is a
dramatic reduction in I-value. This has the effect of producing markedly reduced EI values. The
geometry of the test pieces and the loading they are placed under, puts a premium on
effective stiffness if the objective is to preserve the part geometry as the load is applied. The
reduced EI in the infused samples means that they deflect out of shape much more than the
hand laid samples and this generates non-linear behaviour, as laminate displacement is a
Page | 141
multiple of laminate thickness. Thus, the loading on the radiused corners of the samples is
mixed mode.

Therefore, geometries such as these are felt to be highly dependent on EI values that, for the
operating load of the structure, limit displacement and preserve part geometry. They are, in
effect, stiffness limited. The load-displacement plots show the slope of the VIP curves is
approximately half that of the HL samples.

Design choices that encourage EI are therefore desirable for parts such as keel floors. The use
of cores to increase EI with little weight gain would be an obvious solution enhancing EI values.

4.6 Top-hat-stiffener experimental study conclusions

Singly curved L-bend specimens and Top-hat-stiffeners for keel structures were tested. Three
layups of L-bend composite specimens were tested under displacement control up to 100 mm
displacement.

The manufacturing process had a significant effect on the stiffness performance of the Top-
hat-stiffeners. Experimental investigation on six different layups with two manufacturing
process was conducted. As expected, vacuum-infused specimens showed higher strength and
stiffness compared to the hand-laid specimens. At the same time, due to lower resin content,
laminate thickness was reduced by at least 40 per cent. Hence, low structural stiffness was
observed with VIP specimens as the structure failed at a lower load than hand-laid specimens.

The initial failure mode and failure location of the Top-hat-stiffeners were similar to L-bend
specimens, between the second and third layers at the lower part of the crown bend, due to
delamination. It is evident that the interlaminar stresses (ILTS in particular) play a vital role
during the damage process. Final collapse of most of the hand-laid Top-hat-stiffener specimens
occurred at the crown bend, whereas ultimate failure of the VIP specimens was at the base
(flange). Hence, adding additional layers at the base is recommended to make it stiffer and
delay final collapse.

The residual strength of the structure was identified to assist the boat builders in designing the
curved structures. Addition of unidirectional fibres proved to be beneficial in gaining secondary

Page | 142
strength, even after extensive delaminations in the laminate. This knowledge of the low ratio
of delamination strength to ultimate tensile strength in laminates is significant to designers.

Due to their thinner cross sections, VIP specimens are unsuitable for this use. A different
approach is required to validate the VIP process for marine structural applications. The
thickness could be increased by adding more reinforcement layers, which may increase the
weight slightly but would increase the strength and stiffness of the structure multifold.

The next chapter deals with the numerical analysis of the L-bend specimen and Top-hat-
stiffener. The numerical results are compared with the experimental results in order to obtain
a validated numerical FEA model.

Page | 143
Chapter 5: Numerical Analysis

Page | 144
5.1. Introduction
Experimental verification of large structures is both time consuming and expensive. It is more
beneficial to use a validated numerical model to obtain basic structural information such as
stiffness, strength, failure location, failure load and failure mode. A validated numerical model
allows the examination of structural behaviour under different scenarios in virtual
computational world. It can be used to identify manufacturing and performance complexities,
provide a safe, and relatively cheap test bed for evaluating side effects and optimising the
performance of a structure before transferring it to the real world. FEA is used to predict out-
of plane displacement and failure mechanisms for comparison with experimental results.

In the current study, preliminary FE modelling is carried out for L-bend composites to
understand the interlaminar tensile stress (ILTS) and interlaminar shear stress (ILSS)
distribution and failure. Failure location and mode were identified and successfully modelled
for the L-bend composites. Once the L-bend model is validated, contact conditions are
introduced for the Top-hat-stiffener model and the damage mechanism is investigated. Two
papers covering L-bend failure prediction presented in this chapter were published in the
proceedings of a refereed international conference4 and an international journal paper5.
Another paper on the failure prediction of Top-hat-stiffeners has been submitted to an
international journal paper6.

5.1.1 Modelling Methodology

Delamination is the primary failure mode in curved composite structures. Among the available
modelling techniques, the virtual crack-closure technique (VCCT) and cohesive zone modelling
provide promising results. One aim of this research is to identify the location of failure

4
Raju, Prusty B.G., Kelly D.W., Lyons D., Peng G.D. (2010). Delamination characterisation of curved
composites using acoustic emission and fibre Bragg gratings, Proceedings of Pacific 2010, International
Maritime Conference, Sydney, Australia, January 2010.
5
Raju, Prusty B.G., Kelly D.W., Lyons D., Peng G.D. (2010). Failure characterisation of L-bend curved
composite laminates, Transactions Royal Institute of Naval Architects (RINA) Part B2, International
Journal of Small Craft Technology 152: 93-105.
6
Raju, Prusty B.G., Kelly D.W (2012). Finite element analysis of the delamination Failure of composite
top hat stiffeners. Proceedings of the Institution of Mechanical Engineers, Part M Journal of Engineering
for the Maritime Environment – Accepted.
Page | 145
initiation. VCCT requires the existence of a predefined crack for crack progression; thus, it
cannot be used in the current numeric model. However, cohesive zone models can predict
crack initiation and progression, as well as the corresponding reduction in strength and
stiffness of the structure.

Numerical analysis of a curved composite panel is carried out using commercially available FE
package—MSC MARC MENTAT 2008r1. This implicit FE solver can perform linear and non-
linear stress analysis in a static or dynamic framework. To employ the implicit FE simulation
procedure, the solver is equipped with additional isotropic/orthotropic material models and
delamination failure criteria for description of composite materials and their failure to perform
the non-linear static analysis of the structural model.

5.1.2 Material Properties

Four different fibre mats were used in the testing configuration. Two different weights of
chopped strand mat (CSM) (450 and 225 gsm), double bias (DB ±45q) 461 gsm, unidirectional
(UD) 450 gsm and biaxial (BE) 450 gsm were used. The material properties of chopped strand
mat, double bias and unidirectional fibres manufactured by hand-layup are presented in Table
5.1.
Hand-layup Material Properties
Property CSM DB UD
b
E11 (MPa) 9607 7872 23576
b
E22 (MPa) 9607 7872 6560
E33 (MPa)c 6062 6673 6560
c
υ12 (MPa) 0.347 0.62 0.369
υ23 (MPa) c 0.139 0.146 0.249
c
υ31 (MPa) 0.108 0.133 0.087
c
G12 (MPa) 2602 7157 2265
G23 (MPa) c 1847 2050 1847
c
G31 (MPa) 1847 2050 2265
ILTS (MPa)a 9.5 10.5 10.5
a
ILSS (MPa) 28 30 28
b
Flexure Modulus (MPa) 1822 1169 4235
Flexure Strength (MPa)b 222 75 437
2 a
Fracture Toughness G1C (kJ/m ) 0.68 1.04 0.84
Critical opening displacement (mm)a 0.05 0.05 0.05
a
Literature [ Johnson, 1986; Dirand et al., 1996; Perrot et al., 2007; Park and Jang, 2004;
Kitching et al., 1984]
b
Experimental testing
c
Using ‘Component Design and Analysis (CoDA)' software with resin and fibre properties
Table 5.1: Material properties–hand-layup process
Page | 146
CoDA is ‘Component Design Analysis’ a standard commercial code for laminate design used in
composite maritime industry. It is based on the classical laminate theory where the input may
be individual fibre and resin types and properties. The output is the full 3D mechanical
properties for the laminate based on the manufacturing technique, suggesting amount of resin
to be used prior to the manufacture.

5.1.3 Element Type and Meshing

Eight-node isoparametric hexahedral elements (element type 7 of MSC MARC) were used with
geometric and material non-linearity. These eight-node composite brick elements allow the
definition of layer-by-layer material parameters, layer thickness, and orientation angles for a
laminated composite material. Whenever contact bodies are used in the model, element type
7 is generally preferred over higher order elements. A finer mesh with a global edge length of 1
mm was used around the bend (crown-web interface) to increase the number of elements in
the complex geometry. The assumed strain formulation was also triggered to improve the
bending accuracy for the 3D element. All the elements were modelled with local co-ordinate
system. A snapshot of the bend of composite laminate bend is shown in Figure 5.1.

Figure 5.1 Element orientations at the bend; local co-ordinate system

Page | 147
5.1.4 Failure Modelling

The FE software has a built-in progressive failure capability, with or without element
elimination and stiffness reduction, using various failure criteria such as Hoffman, Tsai-Wu,
Tsai-Hill, Hashin, Puck and the maximum stress criterion. The primary mode of failure in curved
composites is delamination, which is caused by the interaction of interlaminar tensile (σr) and
shear (τrθ) stresses. One aim of this study is to better understand the variation of radial (ILTS)
and tangential (ILSS) stresses resulting from laminate curvature, stacking sequence and fibre
orientation in curved composite structures. Thus, in the current model, interface elements are
used to simulate the onset and progress of delamination. The constitutive behaviour of these
interface elements is expressed in terms of traction v. displacement between the top and
bottom edge/surface of the elements (Figure 5.2).

Figure 5.2: 3D linear interface element [MSC MARC Volume A, 2008]

Element 188 in MSC MARC MENTAT 2008r1 is designed for a cohesive zone element, which is a
linear, eight-node, 3D element typically used to model the interface between different
materials, where nodes 1, 2, 3 and 4 correspond to the bottom of the interface and nodes 5, 6,
7 and 8 to the top. The stress components of the element are one normal traction and two
shear tractions, which are expressed with respect to the local coordinate system; the
corresponding deformations are the relative displacements between the top and the bottom
face of the element. The element is allowed to be infinitely thin, in which case faces 1-2-3-4
and 5-6-7-8 coincide.

For a 3D interface element, the relative displacement components are given by one normal
and two shear components, expressed with respect to the local element co-ordinate system
shown in Equation 5.1.

Page | 148
 ௧௢௣
‫ݒ‬௡ ൌ ‫ݑ‬ଵ െ ‫ݑ‬ଵ௕௢௧௧௢௠
௧௢௣
‫ݒ‬௦ ൌ ‫ݑ‬ଶ െ ‫ݑ‬ଶ௕௢௧௧௢௠ (Equation 5.1)
௧௢௣
‫ݒ‬௧ ൌ ‫ݑ‬ଷ െ ‫ݑ‬ଷ௕௢௧௧௢௠

Based on the relative displacement components, the effective opening displacement is defined
in Equation 5.2:

‫ ݒ‬ൌ ඥ‫ݒ‬௡ଶ ൅ ‫ݒ‬௦ଶ ൅ ‫ݒ‬௧ଶ (Equation 5.2)

Damage onset is predicted using a quadratic stress criterion, allowing the mesh to split
between the materials using Equation 5.3.

஢ ଶ ஢ ଶ
ቀ ౤ቁ + ቀ ౪ቁ = 1 (Equation 5.3)
ୗ౤ ୗ౪

where σn and σt are the normal and tangential stresses and Sn and St, are the critical values of
normal and tangential stresses. Values of Sn and St are the ILTS and ILSS shown in Table 5.1.

The traction-separation model assumes linear elastic behaviour, written in terms of an elastic
constitutive tensor K relating the nominal stresses (traction vector) t=(tn, ts, tt) to the nominal
strains (opening displacement vector) v(vn, vs, vt) across the interface. The effective traction is
introduced as a function of the effective opening displacement and characterised by an initial
reversible response, followed by an irreversible response as soon as a critical effective opening
displacement has been reached, as shown in Figure 5.3. The irreversible part is characterised
by increasing damage, ranging from zero (onset of delamination) to one (full delamination).
For maximum effective traction, tc corresponding to the critical effective opening displacement
vc is expressed in Equation 5.4.
ଶீ೎
‫ݐ‬௖ ൌ (Equation 5.4)
௩೘

Page | 149
 Figure 5.3: Damage evolution curve for bilinear cohesive element [MSC MARC Volume A, 2008]

Once the corresponding initiation criterion is reached, the specified damage evolution law
describes the rate the material stiffness degrades. A scalar damage variable ‘D’ ሺͲ ൑ ‫ ܦ‬൑ ͳሻ
represents the overall damage in the material and captures the combined effects of all active
degradation mechanisms. In the current model, a bilinear model is used to obtain the traction
t. When the overall damage variable reaches its limit Dmax at all material points, the cohesive
element corresponding to complete fracture of the interface between layers can be removed
and is considered as delamination propagation.

5.2 L-bend Laminates

The comparative study to validate the FE model with the experimental results for L-bend
curved laminates discussed in Chapter 4 is presented in this section. Non-linear static analysis
was conducted on L-bend curved composite specimens subjected to bending load. The failure
modelling is based on the ILTS and ILSS distributions. The curved beam is subjected to pure
bending such that the angle of the bend is increased (i.e., the bend is opened). The maximum
ILTS occurs at the mid-section of the beam in the thickness direction and minimum at the
outer surfaces, which creates a crack within the laminate thickness.

5.2.1 L-Bend FE Model definition

The geometry and the experimental setup of the L-bend specimen are shown in Figure 5.4.
Two boundary conditions are used; all nodes at the base are fixed and the loading node is
subjected to a displacement of 100 mm.

Page | 150
 Figure 5.4: a) L-Bend Experimental setup; b) Corresponding FE model

A 90 mm 2D beam is used to connect the laminate and the loading point to simulate the
behaviour at hinges during experimental testing. Each element thickness represents the
thickness of one layer in the composite laminate, except the first layer, which has three
elements in the through-thickness direction to increase the ILTS accuracy at the inner surface.
A static load case was defined with 100 fixed load steps. Tests were carried out on the three
different layups described in Table 5.2.

Layers Layup 1 - CSM Layup 2 - CSM+DB Layup 3 - CSM+DB+UD

Layer 1 CSM 450gsm CSM 450gsm DB 611gsm
Layer 2 CSM 450gsm DB 611gsm CSM 450gsm
Layer 3 CSM 450gsm CSM 450gsm UD 451gsm
Layer 4 CSM 450gsm DB 611gsm DB 611gsm
Layer 5 CSM 450gsm CSM 450gsm CSM 450gsm
Layer 6 CSM 450gsm DB 611gsm UD451gsm
Layer 7 CSM 450gsm CSM 450gsm DB 611gsm
Laminate Thickness 4.62mm 4.62mm 4.62mm

Table 5.2: Layup configuration for L-bend laminates

Page | 151
5.2.2 L-Bend FEA Results

The curved laminates were loaded such that the radius of curvature expanded and
interlaminar tensile stresses were induced in the inner surface. This mechanism was described
by Li and Kelly [2003]—the ILTS generally is at maximum within 20–50 per cent of the thickness
and zero at the surfaces. The load was applied to pull the arm of the specimen, straightening
it. The peak ILTS occurs in the middle of the width of the specimen as the loading point is
there. The tests were terminated when the displacement reached 100 mm.

Through-thickness ILTS was maximum at the lower end of the radius of the bend and this
location was predicted as the probable location of initial failure. For all three layups,
interlaminar matrix cracking occurred due to excessive ILTS and delamination initiated at the
lower end of the bend. The critical interlaminar tensile and shear stresses of the three layups
are presented in Figures 5.5, 5.7 and 5.9. The theoretical ILTS on the extreme surface is zero;
however, due to positioning of the Gauss points, which do not sit on the surface of the
elements, the interlaminar stress calculated at this point is a non-zero value.

5.2.3 L-Bend Layup 1 (All CSM) FEA results

All layers of the laminate in Layup 1 are CSM. The ILTS and ILSS stress distributions prior to the
first failure are shown in Figure 5.5. The ILTS curve is smooth and linear until mid-thickness and
drops to near zero at the surface, as shown in Figure 5.11. With this layup, the initial failure
occurred at a displacement of 65 mm, predominantly due to interlaminar tensile stress of 9.22
MPa, between the second and third layers, mid-width and at an angle of 18.26q. The
corresponding ILSS at the point was 4.23 MPa. Further loading caused the laminate to
straighten, creating ILTS in the inner surface and interlaminar compressive stress on the outer
surface. Experiments for five specimens of similar configuration and FEA load-deflection plots
are shown in Figure 5.6. As the FE model is free of manufacturing and testing inaccuracies,
relatively higher strength and stiffness is observed compared to the experimental results.

Page | 152
 Figure 5.5: L-Bend Layup 1 ILTS and ILSS distribution

Figure 5.6: L-Bend Layup 1 experiment and FEA load deflection plot

5.2.4 L-Bend Layup 2 (CSM+DB) FEA results

Layup 2 consisted of four CSM layers and three DB layers. The first failure occurred between
the second (DB) and third (CSM) layer. The ILTS and ILSS stress distributions in the bend at
increment number 66 for Layup 2 are presented in Figure 5.7. Delamination occurred at an
angle of 21.89q due to the ILTS peak of 10.52 MPa and ILSS peak of 6.89 MPa. Delamination
progressed until all the layers failed and the model collapsed. The addition of DB layers slightly
increased the load-carrying capacity, as well as the deflection of the first failure—66 mm
compared to 65 mm for all CSM. A comparison of the FEA load deformation with the
experiments for five specimens is shown in Figure 5.8. It can be seen that the stiffness of the
FE model is comparatively higher than that of experimental results.

Page | 153
 Figure 5.7: L-Bend Layup 2 ILTS and ILSS distributions

Figure 5.8 L-Bend Layup 2 – Experiment and FEA load deflection plot

5.2.5 L-Bend Layup 3 (CSM+DB+UD) FEA results

Layup 3 consists of three CSM layers, three DB layers and two UD layers. Figure 5.9 shows the
ILTS and ILSS stress distributions in the bend at increment number 79 for Layup 3 just before
the initial failure. The addition of unidirectional fibres increased the stiffness, and the first
crack occurred at 79 mm. The maximum ILTS of 10.61 MPa was at an angle of 14.60q and the
maximum shear stress of 10.22 MPa occurred at the lower end of the laminate curvature. With
the third layup, the first crack occurred between the third (UD) and fourth (DB) layers.
Experiments on five specimens are compared with the FEA load-deflection comparative plot in
Figure 5.10. Due the presence of UD layers, the structure was able to take the load beyond
1600 N for most of the specimens.

Page | 154
 Figure 5.9: L-Bend Layup 3 ILTS and ILSS distributions

Figure 5.10: L-Bend Layup 3 experiments and FEA load-deflection plot

5.2.6 L-Bend Interlaminar tensile and shear stress distributions

Implicit FEA attempts to force convergence of loads or displacements, and this convergence
becomes more complex and computationally expensive when the material becomes highly
distorted and damaged. Yet, the MSC MARC MENTAT 2008r1 was able to predict the initial
failure, as well the failure progression. The numerical model discussed in this section
attempted to determine the regions of maximum interlaminar tensile and shear stress and the
predictions showed good agreement with the experimental results for all layups. The ILTS and
ILSS plots for all three layups are presented in Figures 5.11 and 5.12, respectively. The ILTS
distribution is comparable for all the three layups; however, Layups 2 and 3 have higher ILSS

Page | 155
values, which may be attributed to a mismatch in the elastic properties between adjacent
layers.

12

9
ILTS (MPa)

6
Layup 1
Layup 2
3
Layup 3

0
0 0.66 1.32 1.98 2.64 3.3 3.96
Thickness (mm)

Figure 5.11: L-Bend ILTS distribution just before first failure for all three layups

-2
ILSS (MPa)

-6

Layup 1 ILSS
-10 Layup 2 ILSS

Layup 3 ILSS
-14
0 0.66 1.32 1.98 2.64 3.3 3.96
Thickness (mm)

Figure 5.12: L-Bend ILSS distribution just before first failure for all three layups

5.2.7 L-bend Modelling summary

Specimens of three different layups were tested up to 100 mm deflection. The specimens were
not loaded by pure moment; hence, the failures do not occur exactly at 45q. The FEA results
for Layups 1 and 2 were comparable to the experimental data. The maximum ILSS for all three
layups occurred at the lower end of the laminate bend. The failure criterion for FEA is a
function of both the interlaminar tensile and shear stresses. Interlaminar tensile stress
dominated the initial failures for all three layups.

Page | 156
The FEA load-deflection plot of all layups, up to the first failure, is shown in Figure 5.13. CSM
demonstrates higher stiffness than the other two layups due to its homogeneity and
brittleness. Most CSM laminate specimens were snapped into two pieces at the final failure.
DB is very flexible and the inclusion of DB layers slightly reduced the stiffness of the structure
for Layups 2 and 3. Layup 3 had two unidirectional layers, which increased the ultimate failure
load; however, the stiffness is lower than that of Layup 1. DB and UD layers held the structure
of Layup 3 specimens together, even after the ultimate failure. However, at this stage, the CSM
layers were completely damaged. The stiffness seems to have been dominated by the DB
layers and the ultimate strength is boosted by the presence of UD layers.

Figure 5.13: L-Bend FEA load-deflection plot for three different layups

Reasonable agreement was obtained between the experiment and FEA predictions. Notable
discrepancies are attributed to minute differences in the geometry and the boundary
conditions between the test specimens and numerical models. Accurate damage location and
loading can be determined through the interaction of the interlaminar stresses. Shapes of
composite structures are generally more complex than simple L-bends; the next section will
discuss the failure prediction of Top-hat-stiffeners with more complex boundary conditions,
loadings and contact issues.

Page | 157
5.3 Hand Laid Top-hat-stiffeners (HL-THS)
The studies with the simple L-bend specimens outlined in the previous section showed that the
proposed numeric model could predict failure initiation and progression under flexure load.
This section examines failure prediction in a realistic keel joint structure with a more complex
geometry. The crown of the Top-hat-stiffener is embedded between two 10 mm steel plates
and the load is applied at the centre bolt representing the keel bolt of a typical yacht.

5.3.1 Top-hat-stiffener geometry and material layup

The FE model for the Top-hat-stiffener is created using the geometry shown in Figure 5.14. The
crown and flange had radii of 19 and 23 mm, respectively. The specimens were 100 mm wide.
The web has a 3q offset from the y-axis. The width of crown and the height of specimen were
of both 200 mm each. The crown is sandwiched between two 10 mm steel plates.

Figure 5.14: Top-hat-stiffener – Geometry of the finite element model

Three fibre reinforcement types were used: chopped strand mat (CSM), double bias (DB), and
unidirectional (UD) fibres with the fibre aligned with the local x-axis, as shown in Figure 5.1.
For each layup, the flange-web bends had an additional two layers of CSM to stiffen the Top-
hat-stiffener at the base. To confirm the strains from FEA with experimental data, two strain

Page | 158
gauges were glued to the specimens (shown in Figure 5.14). The lamina thicknesses for CSM
(450 gsm), DB (611 gsm) and UD (461 gsm) were 0.63, 0.69 and 0.66 mm respectively. The
lamina thickness for CSM (225 gsm) was taken as 0.315 mm. The crown of the Top-hat-
stiffener was sandwiched between two ten-mm mild steel backing plates. The three layups for
the Top-hat-stiffeners are presented in Table 5.3, where the bottom layer is labelled layer 1.

Layup 1 Layup 2 Layup 3

Layer 1 CSM 450gsm CSM 450gsm CSM 225gsm
Layer 2 CSM 450gsm CSM 450gsm UD 461gsm
Layer 3 CSM 450gsm DB 611gsm CSM 225gsm
Layer 4 CSM 450gsm CSM 450gsm UD 461gsm
Layer 5 CSM 450gsm DB 611gsm CSM 225gsm
Layer 6 CSM 450gsm CSM 450gsm UD 461gsm
Layer 7 CSM 450gsm DB 611gsm CSM 225gsm
Layer 8 --- CSM 450gsm DB 611gsm
Layer 9 --- --- CSM 225gsm
Layer 10 --- --- DB 611gsm
Layer 11 --- --- CSM 225gsm
Layer 12 --- --- DB 611gsm
Laminate Thickness 4.41mm 5.22mm 5.79mm
Table 5.3: Hand laid Top-hat-stiffener laminate configuration

5.3.2 Top-hat-stiffener Structural Modelling

The basic skeleton of the FE model (pre-processing) was performed using MSC PATRAN MD
R2.1. During the experimental investigation, a load was applied at the centre of the crown,
using a 20 mm diameter bolt at the centre of the specimen, at a rate of 2 mm/min.
Considerable through-thickness shear deformation was observed. To simulate experimental
behaviour, FEA was performed on a 3D model. Local co-ordinate systems were defined for
each element to match the local fibre orientation. Each element thickness represents the
thickness of one layer in the composite laminate to obtain an accurate distribution of the
interlaminar stresses. As expected, the damage initiated at the mid-width of the specimen due
to the load being applied at the centre of the crown. As observed in the experiments, damage
did not occur in both halves of the specimen simultaneously, hence, the Top-hat-stiffener is
modelled as a single body and a symmetry option is not considered. The numbers of elements
and nodes used for three different models are presented in Table 5.4.

Page | 159
 Layup 1 Layup 2 Layup 3
No. of elements 18582 20506 23978
No. of Nodes 14152 15908 19612
Table 5.4: Elements and nodal numbers for three different layups of hand laid Top-hat-stiffener

Hou et al. [2001] studied the effects of shear stress distribution due to compressive loading on
structural delamination. Out-of-plane tension accelerated the development of delamination
failure, while out-of-plane compression limited delamination initiation. He suggested that even
small compressive stresses have a large restraining influence on the delamination growth.
Hou’s theory also suggests that the critical shear stress required to initiate delamination is
much higher when the laminate experiences through-thickness compression. Shear
delamination is eliminated altogether if the compressive load exceeds a critical value. Thus, in
the current simulation, no delamination in the laminate was allowed to occur between the
steel backing plates at the crown of the Top-hat-stiffener.

5.3.3. Top-hat-stiffener Loading and Boundary Conditions

For curved laminates, the boundary conditions play a key role in predicting the accuracy of the
results. Three boundary conditions were defined to represent the arrangement existing in the
Top-hat-stiffener experimental analysis. The nodes at the flange were secured in all directions
and a displacement load was applied on the bolt nodes in the y-direction while the steel
backing plate nodes were fixed in x and z directions. The load was applied linearly with respect
to time.

5.3.4 Top-hat-stiffener Contact Conditions and Load Case

Contact condition was modelled using the direct constraint method in MSC. MARC. This
method requires the definition of contact bodies—i.e., bodies that potentially come into
contact. The efficiency of the contact bodies was further improved by using a contact table
available in MSC MARC. These contact tables define the contact bodies which are likely to be in
contact with each other (may be either glued or touching contact) during an analysis step. Four
contact bodies—laminate, upper backing plate (crown top), lower backing plate (crown
bottom) and the central bolt were defined in the FE model. The nodes of the upper and lower
backing plates were glued (by defining them in the contact table) to the bolt nodes. The

Page | 160
laminate nodes were modelled as touching the steel plates. The contact table used in this
section is shown in Figure 5.15.

Figure 5.15: A typical contact table defined in MSC.MARC Mentat for the Top-hat-stiffener model

A single load case was defined in this analysis. Generally, a small number of increments
provide an inaccurate damage mechanism while a large number of increments increase the
computer runtime; hence an optimum number of increments need to be provided. After
certain number of iterations, it was decided that the displacement load needs to be applied in
200 fixed increments (load steps) to identify the damage initiation and progression with high
accuracy and optimal processing time.

5.4 Top-hat-stiffener FEA Results

Three different models for three layups were generated from MSC. PATRAN R2.1 and imported
to MSC MARC. A similar failure criterion was used for all three models. For Layup 3, large strain
analysis was conducted to simulate the straightening of the bend at all four corners. Reserve
strength modelling is discussed in Section 5.4.3.

5.4.1 HL Top-hat-stiffener Layup 1 FEA results

This layup consists of randomly oriented short-length fibres (CSM) and without continuous
fibres running along the structure. This makes the structure very brittle and prone to
catastrophic failure compared with the other layups. The numerical analysis was conducted
using small strain analysis and backing plates with elastic-plastic behaviour. Interlaminar
stresses just before the first failure are shown in Figure 5.16.

Page | 161
 Figure 5.16: HL Top-hat-stiffener Layup 1 ILTS and ILSS distribution just before first failure

To validate the numerical model, the FEA result was compared with the experimental results
for specimen ‘THS HL A04’, shown in Figure 5.17. When the interlaminar stresses exceed the
critical values, interlaminar cracks are introduced. Crack growth is obtained by the insertion of
interface elements.

25

Simulation
20
Experiment CSM04

15
Load (kN)

10

0
0 3 6 9 12 15 18
Deflection (mm)

Figure 5.17: HL Top-hat-stiffener Layup 1 load-deflection plots—experiment and FEA comparison

The first crack occurred at 18.64 kN with a deflection of 8.42 mm in the left bend. At this point,
the peak ILTS and ILSS were predicted to be 9.54 and 1.77 MPa respectively, at the mid-width,
between the second and third layers at an angle of 18q from the lower end of the left bend.
Successive load increments produces multiple cracks in the left bend, but the right bend was
still intact and continued to take more load until a deflection of 14.66 mm and 21.86 kN, when
Page | 162
the structure suffered a major delamination. At this stage, both bends at the crown were
damaged and the structure catastrophically collapsed at 17.37 mm deflection with an ultimate
failure load of 19.38 kN. Comparison of damage initiation and progression between the
experimental and numerical results are shown in Figure 5.18.

Failure
Initiation

Failure
Progression

Ultimate
Failure

Figure 5.18: HL Layup 1 correlation of experimental failure and FEA failure prediction

Page | 163
5.4.2 HL Top-hat-stiffener Layup 2 FEA results

Layup 2 consists of CSM layers and three DB layers. The fibres in the DB layers are oriented to
±45q, but the loading is applied at 0q. This alignment allows the laminate to sustain a large
displacement without the structure snapping into two or more pieces. Due to the mismatch in
the elastic modulus, the interlaminar tensile stress distribution is not uniform, as shown in
Figure 5.19.Similar to Layup 1, small strain analysis was performed. At first failure, the peak
interlaminar tensile and shear stresses were 11.96 and 3.56 MPa, respectively, as shown in
Figure 5.19.

Figure 5.19: Top-hat-stiffener Layup 1 ILTS and ILSS distribution just before first failure

FEA results of this layup were compared to specimen ‘THS HL B04’ to validate the numerical
model, as shown in Figure 5.20. Similar to with Layup 1 failure behaviour, the initial crack
occurred between the second and third layers at mid-width, 18q from lower end of the left
bend. The numerical simulation predicted the delamination fracture at a deflection of 9.84 mm
for a load of 23.65 kN. At this increment, a significant load drop of 2.3 kN was observed and
interface elements were inserted to simulate crack progression. The next fracture occurred at
the right bend, as it sustained large stresses during successive load increments at 10.80 mm
and 22.80 kN. Once both bends were damaged, there was a large deflection with a minor
increase in load. Later, more delaminations occurred between the outer layers of the laminate
due to interlaminar compressive stress and shifting of the neutral axis in the through-thickness
direction. Multiple delaminations occurred between layers six and seven, and seven and eight,
at 14.16 mm with a load of 24.28 kN. Finally, at a displacement of 20.1 mm, the structure
suffered multiple delaminations at both bends and collapsed at a load of 21.79 kN.
Page | 164
 30

Simulation
25
Experiment CSMDB 04

20
Load (kN)

15

10

0
0 5 10 15 20
Deflection (mm)
Figure 5.20: HL Top-hat-stiffener Layup 2 load-deflection plot—experiment and FEA comparison

A comparison of the damage analysis between the experimental and numerical results is
shown in Figure 5.21.

Failure
Initiation

Failure
Progression

Page | 165
 Ultimate
Failure

Figure 5.21: HL Top-hat-stiffener Layup 2 experimental failure and FEA failure-prediction correlation

5.4.3 HL Top-hat-stiffener Layup 3 FEA results

Layup 3 consists of three DB layers (611 gsm) and three UD layers (461 gsm), each sandwiched
between six CSM layers (225 gsm). The UD layers were placed on the lower part of the
laminate and DB layers on the upper end.

Interlaminar stress distributions around the top bends of the Top-hat-stiffener just before the
first failure are shown in Figure 5.22.

Figure 5.22: HL Top-hat-stiffener Layup 3 ILTS and ILSS distribution just before first failure

To validate the FEA model, specimen ‘THS HL C03’ was compared with the FEA results shown
in Figure 5.23. In the FEA model, for a load of 17.10 kN, the initial crack occurred at 6.71 mm,
between the second and third layers at an angle of 18q from the lower end of the left bend.
The contribution of ILTS to crack initiation was greater than that of ILSS. A comparison of the
Page | 166
damage process between the experimental and FEA results is shown in Figure 5.24. The
number of failures (significant load drop in the FEA load-deflection plot) is less than that of the
experimental results due to factors such as the absence of material impurities, manufacturing
inconsistencies and testing inaccuracies. Successive load increments caused the right bend to
fail at 13.63 mm and 33.46 kN. The next two failures were at the lower bends at 17.62 and
19.51 mm with loads of 35.67 and 27.83 kN, respectively.

50

Simulation Large Strain - Elastic backing plate

40
Experiment CSM+DB+UD 03

30
Load (kN)

20

10

0
0 5 10 15 20 25 30
Deflection (mm)

Figure 5.23: HL Top-hat-stiffener Layup 3 load-deflection plot—experiment and FEA comparison

At this stage, all bends completely disappeared, as shown in Figure 5.25. Secondary reserve
strength was observed at 20.98 mm and 29.44 kN. The secondary reserve strength of the
structure in the FE model continued to take up the load up to 54 kN.

Page | 167
 Failure
Initiation

Failure
Progression

Ultimate
Failure

Figure 5.24: HL Top-hat-stiffener Layup 3 experimental failure and FEA failure prediction correlation

During the experimental testing, two behavioural changes were observed, in comparison with
the first two layups. At 20.2 mm deflection, all four bends started straightening up and the
bends began to disappear. By the end of the test, all the bends were straightened and the Top-
hat-stiffener looked like a trapezium, as shown in Figure 5.25. Large strain analysis was used to
model this trapezoidal behaviour of the Top-hat-stiffener. A reserve strength that carried the

Page | 168
structure from 20 to 45 kN loading was identified. This secondary reserve strength occurred
due to the presence of unidirectional fibres. During FE modelling with cohesive zone elements,
it was not possible to model this reserve strength. However, modelling the material of the
steel backing plates as elastic instead of elastic-plastic improved the numerical stability of the
algorithm. This helped in modelling the reserve strength from 20 kN load.

Figure 5.25: HL Top-hat-stiffener Layup 3 final stage of failure; the structure takes a trapezium shape—
FEA and experimental correlation

Page | 169
5.4.4 HL Top-hat-stiffener Interlaminar Strain Distribution around bends

The ILTS reached maximum at the lower end of the left bend and this location was predicted as
the probable location of the initial failure. The delamination index just before the first failure is
shown in Figure 5.26. This index is the value of the RHS of the Equation 5.3 with values equal
to 1, indicating failure.

Figure 5.26: Delamination failure index at first failure for all three layups at left crown bend

The critical interlaminar tensile and shear stresses of all three layups are shown in Figures 5.27
to 5.32 with the marked failure locations. Figures 5.27 and 5.28 show the through-thickness
interlaminar tensile and shear stress distributions, whereas Figures 5.29 and 5.30 detail the
ILTS and ILSS distributions across the width of the specimen between the second and third
layers and at an angle of 18q from the bottom of the left bend. Figures 5.31 and 5.32 depict the
ILTS and ILSS distributions between the second and third layers around the radius of the bend.
As the specimens are symmetric about the vertical axis, the stress distribution is same at both
bends until the first failure occurs. An interlaminar matrix crack occurred due to excessive ILTS
and delamination initiated at the lower end of the bend for all three layups.

Page | 170
 12
Layup 1
Layup 2
9
Layup 3
ILTS (MPa)

Failure
3 Location

0
0 0.66 1.32 1.98 2.64 3.3 3.96 4.62 5.28
Through-thickness (mm)
Figure 5.27: Hand laid Top-hat-stiffener through-thickness ILTS distribution

4
Layup 1
Layup 2
3
Layup 3
ILSS (MPa)

Failure
1 Location

0
0 0.66 1.32 1.98 2.64 3.3 3.96 4.62 5.28 5.94
Through-thickness (mm)
Figure 5.28: Hand laid Top-hat-stiffener through-thickness ILSS distribution

12

11
Failure
Location
10
ILTS (MPa)

9
Layup 1
Layup 2
8
Layup 3

7
0 10 20 30 40 50 60 70 80 90 100
Width (mm)
Figure 5.29: Hand laid Top-hat-stiffener ILTS distribution across the width

Page | 171
 20
Layup 1
Layup 2
16
Layup 3
Failure
12
ILSS (MPa)

Location

0
0 10 20 30 40 50 60 70 80 90 100
Width (mm)
Figure 5.30: Hand laid Top-hat-stiffener ILSS distribution across the width
13

-13
ILTS (MPa)

-26
Failure
Layup 1
Location
-39
Layup 2
Layup 3
-52

-65
0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90
Angle (Degrees)
Figure 5.31: Hand laid Top-hat-stiffener ILTS distribution along a bend (radius)
35

28
Layup 1 Failure
Layup 2 Location
21
ILSS (MPa)

Layup 3
14

-7
0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90
Angle (Degrees)
Figure 5.32: Hand laid Top-hat-stiffener ILSS distribution along the bend (radius)

Page | 172
5.4.5 HL Top-hat-stiffener: Comparison of Strains from FEA and
Experiment

Two rectangular foil strain gauges were attached to the inner surface of the web of specimen
B04 with Layup 2, as shown in Figure 5.14. The web of the specimen was fitted with two
rectangular three-gauge 45q strain gauge rosettes. Longitudinal tensile strain (y-direction) and
transverse shear strain (yz-direction) for the two strain gauges were compared with the
numerical model results. To validate the FEA model, the strain at this location was compared
to the experimental results (Figures 5.33 to 5.36) until first failure.

The experimental tensile strain in yy-direction at SG1 just before the first failure was 5873 μH
at 22.74 kN and the FEA strain was 6709 μH (14.2% higher) at the same load. The experimental
SG1 transverse shear strain in yz-direction was 470 μH and the FEA yz strain was 451 μH (4.2%
lower). For SG2, the experimental tensile strain was 2660 μH compared to an FEA strain of
2364 μH (12.5% lower). The SG2 experimental transverse shear strain was 194 μH and the FEA
transverse shear strain was 240 μH (23.7% higher). The shape of the strain curve indicates the
non-linear behaviour of the structure. An acceptable correlation between the experimental
and FEA results was achieved.

7500

SG1- eyy
6000
FEA
Strain (Microstrain)

4500

3000

1500

0
0 5 10 Load (kN) 15 20 25

Figure 5.33: Hand laid Top-hat-stiffener Comparison of FEA results and strain gauge 1 (SG1) Hyy

Page | 173
 500

SG1- eyz
400
FEA YZ
Strain (Microstrain)

300

200

100

0
0 5 10 Load (kN) 15 20 25

Figure 5.34: Hand laid Top-hat-stiffener Comparison of FEA results and strain gauge 1 (SG1) Hyz

3000
SG2 - eyy
2500
FEA
Strain (Microstrain)

2000

1500

1000

500

0
0 5 10 Load (kN) 15 20 25

Figure 5.35: Hand laid Top-hat-stiffener Comparison of FEA results and strain gauge 2 (SG2) Hyy

225
FEA YZ
175
SG2 - eyz
Strain (Microstrain)

125

75

25

0 5 10 15 20 25
-25
Load (kN)

Figure 5.36: Hand laid Top-hat-stiffener Comparison of FEA results and strain gauge 2 (SG2) Hyz
Page | 174
5.4.6 VIP Top-hat-stiffener Layup 4 FEA results

Due to the manufacturing process, VIP specimens had a higher fibre volume fraction, as
discussed in Chapter 4. This resulted in reduced thickness of the structure; the specimens were
very flexible. An FEA model was created for layup 4 specimens in order to validate VIP
specimens. The material layup for the Layup 4 specimens is given in Table 5.5 and the material
properties used to model Layup 4 VIP specimens are given in Table 5.6.

Layup 4 (CSM+DB) Thickness

DB-450 0.26
CSM-450 0.6
DB-450 0.26
CSM-450 0.6
DB-450 0.26
CSM-450 0.6
DB-450 0.26
DB-450 0.26
Total thickness 3.1
Table 5.5: VIP Top-hat-stiffener layup 4

Vacuum Infusion Process Material Properties

Property CSM DB UD
b
E11 (MPa) 10000 6703 37232
b
E22 (MPa) 10000 6703 15128
E33 (MPa)c 6500 6173 15128
c
υ12 (MPa) 0.347 0.62 0.369
c
υ23 (MPa) 0.139 0.146 0.249
c
υ31 (MPa) 0.108 0.133 0.087
c
G12 (MPa) 2602 7157 2265
G23 (MPa) c 1847 2050 1847
c
G31 (MPa) 1847 2050 2265
a
ILTS (MPa) 9.5 10.5 10.5
a
ILSS (MPa) 28 30 28
b
Flexure Modulus (MPa) 2668 2869 6748
Flexure Strength (MPa)b 266 177 579
2 a
Fracture Toughness G1C (kJ/m ) 0.68 1.04 0.84
a
Critical opening displacement (mm) 0.05 0.05 0.05
a
Literature [ Johnson, 1986; Dirand et al., 1996; Perrot et al., 2007; Park and Jang, 2004;
Kitching et al., 1984]
b
Experimental testing
c
Obtained through 'Component Design and Analysis (CoDA)' software using resin and fibre
properties
Table 5.6: Material properties of the VIP Top-hat-stiffener
Page | 175
The same failure criterion (interface element) is used, and the load-deflection plots are shown
in Figure 5.37. Four major failure points are seen at 12.54, 13.75, 15.44 and 18.97 kN. The
stiffness and failure behaviour are in good agreement with the experimental results.

35
CSM+DB 01
30
CSM+DB 02
25 CSM+DB 03
CSM+DB 05
Load (kN)

20
FEA
15

10

0
0 5 10 15 20 25
Deflection (mm)
Figure 5.37: VIP Top-hat-stiffener Layup 4 load-deflection plot experimental and FEA comparison

Page | 176
 Figure 5.38: VIP Top-hat-stiffener Layup 4 Failure - FEA and experimental comparison

A comparative study validating experimental and FE model results is presented in this section.
FEA predicted reasonably higher stiffness for Layup 1 (7.5%) and Layup 2 (15.35%) than the
experimental stiffness. For Layup 3, FEA predicted 19.80% lower stiffness. These deviations are
attributed to geometric variances, boundary condition variations and the slippage/slackness at
the base clamps during experimental testing.

Page | 177
5.5 FEA of L-Bend and Top-hat-stiffener Conclusions
Failure progression is performed by automatic insertion of cohesive elements between
damaged layers. Successive failures are triggered by delaminations between other layers.
Applying a pure bending load on the bend causes failure at the neutral axis at the mid-width of
the bend due to large ILTS and small ILSS; however, in the current study, an offset in the
loading is observed because of the presence of steel backing plates and a slightly inclined web
(3q). Finite element modelling was used to predict the failure load and damage prediction was
based on cohesive zone modelling. Good agreement with the experimental results was
obtained for the L-bend composites and hand-laid and VIP Top-hat-stiffeners. The secondary
stiffening was successfully modelled.

As the damage mechanisms in composites involve multiple failure modes, different techniques
are needed to determine and discriminate the failure modes. Among the various NDE
techniques available, acoustic-emission (Chapter 6) and optical-fibre strain sensing (Chapter 7)
have proven useful for composites. The next chapter deals with failure characterisation of the
experiments conducted in this chapter using acoustic-emission techniques, with conventional
piezoelectric sensors and state-of-the art fibre Bragg grating sensors.

Page | 178
Chapter 6: Acoustic Emission

Prosser W.H.(2002). Acoustic-emission, Non-destructive evaluation, Theory, Techniques and Applications.

Marcel Dekker, Inc., New York. ISBN: 0-8247-8872-9

Page | 179
6.1 Acoustic-emission Introduction
The experimental failure analysis of curved composite structures based on the corresponding
load drop and photographic evidence was discussed in Chapter 4. Various NDE techniques are
discussed in Chapter 2. Among these, acoustic-emission (AE) is one of the better techniques
compatible with composite structures. This chapter presents the usage of AE technique for
failure characterisation of L-bend and Top-hat-stiffener specimens using conventional
piezoelectric AE sensors and modern fibre Bragg Grating (FBG) sensors. The history of AE
events represents the damage initiation and progression in a material. When a laminated
composite structure begins to generate more AE events, it indicates damage occurrence within
the material. Damage source location and failure mode discrimination is achieved by
systematic analysis of the various parameters (parametric analysis) or the corresponding
waveforms (transient analysis). Digital recording of the AE data stores the signal event
information for further review and analysis. Typical AE applications in the aerospace and
marine industries determine fatigue cracking and corrosion and monitor failures in composite
structures. A book chapter in an edited book7 and a journal paper has been published on the
results of Acoustic emission analysis8.

6.1.1 Parametric Analysis of AE signal

The current study carries out a systematic approach in identifying the failure mode, failure
intensity and extent of failure using parametric analysis of AE signals. The waveform is
analysed and the parameters are extracted. Failure characterisation of composite materials is
performed based on these extracted values. Parametric AE analysis relies on a number of
parameters, such as: arrival time, peak amplitude, rise-time, signal duration, counts or
threshold crossings and elastic energy released during fracture/crack propagation.

Many researchers have worked with parametric analysis and developed their own parametric
values for failure modes such as matrix cracking, interface debonding, fibre breakage,

7
Raju, Prusty B.G. (2011). AE Failure Characterization of Advanced Composites using Piezo-Electric
Sensors and Fibre Optic Sensors, for the edited book titled Theory and uses of Acoustic emission, Nova
Publications, USA. ISBN: 978-1-61209-960-6.
8
Raju, Azmi A.I., Prusty B.G. (2012). Acoustic Emission Techniques for Failure Characterisation in
Composite top-hat Stiffener. Journal of Reinforced Plastics and Composites 31:495-516.
Page | 180
delamination and fibre pull-out. One of the initial studies [Nahas, 1985] explained only two
levels of failure—matrix cracking and fibre fracture. Through improvements in data-acquisition
technology, the detection of other failure modes is now possible. A number of investigators
have attempted to use the AE technique to detect and evaluate the damage mechanisms in
composites [Bohse, 2000; Bhat et al., 1994; Leszek and Kanji, 1995; Benmedakhene et al.,
1999; Yuris and Jie, 2001]. Huguet et al. [2002] studied the damage mode and the
corresponding amplitude output of GFRP composites. They proposed that the amplitude of
matrix cracking was 40–55 dB and that of debonding was 60–65dB; fibre pull-out was 65–85
dB and fibre fracture was 85–95 dB. A study by Ageorges et al. [1999] showed that interface
debonding had amplitude of 40–60 dB and fibre breakage of 75–90 dB. In classifying the failure
mechanism based on event duration, debonding was about ten milliseconds, fibre breakage
approximately fifteen and delamination approximately twenty-five milliseconds. A summary of
parametric values proposed by researchers for the different composite failure modes is given
in Table 6.1.

Failure Mode Amplitude (dB) Rise time (μs) Duration (ms)

<40 [Ageorges et al., 1999]
<1 [Ageorges et al.,
40-55 [Barre and Benzeggagh, 1994]
1999]
Matrix 40-55 [Benmedakhene et al., 1999]
N/A
Cracking <40 [Barnes and Ramirez, 1998]
>0.1 [Ono, 1988]
21-24 [Berthelot and Rhazi, 1990]
<40 [Li and Zhao, 1986]
40-60 [Ageorges et al., 1999]
60-65 [Barre and Benzeggagh,1994]
Interface 30-60 [Park et al., 2002] 10-15 [Ageorges et
1-5 [Ujjin, 2007]
Debonding 40-50 [Ujjin, 2007] al., 1999]
55-65 [Benmedakhene et al., 1999]
40-55 [Kotsikos et al., 2000]
75-90 [Ageorges et al., 1999]
85-95 [Barre and Benzeggagh, 1994] 15-25 [Ageorges et
60-80 [Park et al., 2002] al., 1999]
Fibre
50-80 [Ujjin, 2007] 5-20 [Ujjin, 2007]
breakage
60-62 [Berthelot and Rhazi, 1990] <0.1 [Ono, 1988]
>76 [Li and Zhao, 1986]
<50 [Ono, 1988]
>100 [Park et al., 2002]
2-10 [Barnes and
80-100 [Ujjin, 2007]
Ramirez, 1998]
55-70 [Kotsikos et al., 2000]
Delamination >20 [Ujjin, 2007]
40-60 [Barnes and Ramirez, 1998]
~0.2 [Ono, 1988]
52-78 [Li and Zhao, 1986]
50-70 [Ono, 1988]
Fibre pull-out 65-85 [Barre and Benzeggagh, 1994] N/A N/A
Table 6.1: Summary of AE parametric studies
Page | 181
6.1.2 Transient Analysis of AE signal

The waveform of the AE signal has been studied by a number of researchers. Along with the
parametric correlation, waveform analysis has provided validated failure discrimination among
the various failure modes in composite materials. The peaks in the spectrum can be associated
with specific types of damage events, as well as ultimate failure [Eckold, 1994]. A summary of
the frequencies for different failure modes for composites is given in Table 6.2.

Failure Mode Frequency Range (kHz)

250 and 700 [Ageorges et al., 1999]
150-350 [Park et al., 2002]
90-180 [Groot et al., 1995]
Matrix Cracking 50-150 [Russel and Henneke, 1977]
30-150 [Suzuki et al., 1988]
160-200 [Oskouei and Ahmadi, 2010]
<100 [Iwamoto et al., 1999]
100 and 250 [Ageorges et al., 1999]
200-250 [Park et al., 2002]
Interface Debonding 240-310 [Groot et al., 1995]
230-450 [Suzuki et al., 1988]
250-310 [Oskouei and Ahmadi, 2010]
100 and 300 [Ageorges et al., 1999]
200-300 [Park et al., 2002]
300 [Groot et al., 1995]
Fibre breakage 140-180 [Russel and Henneke, 1977]
250-400 [Suzuki et al., 1988]
360-450 [Oskouei and Ahmadi, 2010]
400-450 [Iwamoto et al., 1999]
700-800 [Park et al., 2002]
230-450 [Suzuki et al., 1988]
Delamination
>450 [Oskouei and Ahmadi, 2010]
150-350 [Iwamoto et al., 1999]
Fibre pull-out 0-250 [Suzuki et al., 1988]
Table 6.2: Summary of AE transient studies

Kinjo et al. [1997] conducted pattern-recognition analysis of AE waves monitored by

conventional AE equipment for glass-fibre composites. Autoregressive coefficients determined
by maximum entropy method were used to classify individual AE signals. Based on the analysis,
the AE signals were classified into four types: Type 1 (Mode I), tensile fibre/matrix breakage;
Page | 182
Types 2 and 3 (Mode I), disbonding due to tensile inter-fibre fracture with different crack
normal; and Type 4 (Mode II), disbonding due to shear inter-fibre fracture. Extensive work
[Ageorges et al., 1999; Park et al., 2002; Groot et al., 1995; Russel and Henneke, 1977; Suzuki
et al., 1988; Oskouei and Ahmadi, 2010; Iwamoto et al., 1999] has been carried out in this
area.

6.2 AE Instrumentation and Setup

In the current study, four-channel AE equipment—DISP, from Physical Acoustic Corp was used
to acquire the stress waves generated by the loaded specimen. This AE equipment has four 16-
bit high-speed A/D converters, high-performance multiple digital signal processor (DSP)
technology on a single four-channel PCI card, which can assess progressive real-time AE
features in both the time and frequency domains. Both Parametric and transient analyses are
used in the current study for failure discrimination. Figure 6.1 shows the typical sensors and
preamplifiers used in this study.

Figure 6.1: Sensors and preamplifiers used in this study

Page | 183
In the current study, a ‘Nano 30’ piezoelectric sensor from PAC, with a resonant frequency of
140 kHz is used. Resonant sensors are sensitive to a small range of frequency while the wide
range of materials may produce stress waves with several frequencies. The higher frequency
components will attenuate more quickly as the wave travels and can only be detected close to
the source. These sensors were attached on the outer surface of the laminate at the bend in
order to ensure data was collected from the region of the specimen where most damage
would occur.

Initially the piezo-sensors were clamped to the laminate using G-clamps to secure them in
position. But during the failure process, when the layers separated due to interlaminar
delamination, the clamps and sensors flew away with a huge noise and was an OHS issue.
Hence, in the current study Blu-tack (versatile, reusable putty-like pressure-sensitive adhesive)
was used to apply a constant hold-down force on the sensor, on one side of the laminate for
effective signal sensing. ‘2211-silicone compound’ (Geophysical grade/high vacuum grease)
was used as a couplant to provide a good contact between the material surface and the
piezoelectric sensor.

A 40 dB system gain and a 45 dB threshold were used for the AE acquisition. The AE data
acquisition was synchronized with the mechanical loading and information collected was
stored in a parametric AE file. This allowed correlation of the AE parameters with the load-
deflection at the time the AE signal was generated.

6.2.1 AE Data collection procedure

AE was recorded on a personal computer laptop using ‘AEwin’ software for DISP, which is
commercially available software from PAC. All the sensors were tested using the available
Automatic Sensor Test (AST) function to ensure comparable data recorded by sensors of the
same type. Data collection parameters for the four channels used are the peak definition time
(PDT), hit definition time (HDT) and hit lockout time (HLT) which were set to be 50 μs, 200 μs
and 300 μs respectively for each of the four channels.

Page | 184
In each AE event, the following parameters are analysed; arrival time, rise time, decay time,
amplitude, duration of the signal, counts, hits, signal strength and energy released. The analog
signals from the specimen were pre-amplified and converted into digital signals using the A/D
converter. The Standard channel setup and the signal-processing filter for the DISP equipment
were set as (recommended by the manufacturer, exclusively for composite materials):
Threshold type : Fixed
Threshold limit : 45 dB
Pre amplification : 40 dB
Signal processing filter : 100-300 kHz
Sample rate : 1 MSPS (Mega samples per second)
Pre-Trig ger time : 256 μs

6.2.2 AE Failure Characterisation

Based on the extensive literature survey, the parametric values and the frequency range for
the different failure modes in composite materials are presented in Table 6.4.

Amplitude Energy Duration Rise time Frequency

Failure Mode
(dB) (eu) (ms) (μs) (kHz)
Matrix Cracking 40-55 <5000 8-10 N/A 50-150
Interface Debonding 55-65 5000-8000 10-15 1-5 150-350
Fibre breakage 65-85 9000-12000 >15 5-20 350-400
Delamination 90-100 >12000 >25 >20 >450
Fibre pull-out 65-85 N/A N/A N/A 0-250
Table 6.4: AE parametric values for failure characterisation of composites

6.3 L-bend Composite AE Failure Characterisation

In the current experimental program, the failure behaviour of the specimens with different
composite layups were analysed by comparing four AE parametric variables along with the
load v. time plot. The results of the parametric correlations are presented for each layup. The
results contain six plots: (a) amplitude v. time; (b) energy release v. time; (c) event duration v.
time; (d) event-rise time v. time; (e) counts v. time; and (f) cumulative counts v. time. In the
case of a fracture, elastic energy is released which can be detected and measured by AE
technique. The AE equipment setup for L-bend composite specimens is shown in Figure 6.2.

Page | 185
 Figure 6.2 AE setup for L-bend composites

6.3.1 L-bend Layup 1 (CSM) AE Correlation

These are L-bend specimens with an all CSM layup. It is assumed that no difference exists in
the elastic modulus between the individual layers. Four distinguishable damages were
analysed by AE, as shown in Figure 6.3. The initial scatter, up to 375 sec, accounts for matrix
cracking. The first major damage was observed at 375 sec, between the second and third
layers at 30 per cent of ultimate load. The primary sign of initial delamination is shown as the
energy release (measured in euJ, 1eu=10-18J) of 563 euJ, amplitude of 78 dB, event rise time of
510 μs and event duration of 7139 μs. The small load drop in the load-time plot complements
the failure analysis of AE. Further damage up to 438 sec was attributed to matrix cracking and
delamination growth. The next major delamination occurred at 438 sec, between the third and
fourth layers, with an energy release of 521 eu, amplitude of 92 dB, event rise time of 546 μs
and event duration of 4327 μs. The third major delamination occurred between the fourth and

Page | 186
fifth layers, at 499 sec with an energy release of 490 eu, amplitude of 99 dB, event rise time of
726 μs and event duration of 1819 μs. No significant reduction in load was observed between
the initial delamination at 375 sec and final collapse at 546 sec. The final collapse was a
combination of matrix cracking, fibre failure and delamination, with an energy release of
15609 eu, amplitude of 100 dB, event rise time of 464 μs and duration of 162365 μs. Until the
first failure, few counts were observed (Figures 6.3 (e) and 6.3(f)). A steep increase in the
number of cumulative counts shows increase in AE activity, indicating extensive and rapid
failure progression during the ending stage of the test after 500 sec.

Page | 187
 Figure 6.3: HL L-Bend AE load correlation for Layup 1 specimens

6.3.2 L-bend Layup 2 (CSM+DB) AE Correlation

The parametric distribution of AE events for a CSM+DB layup is shown in Figure 6.4. The first
damage occurred at 10 per cent of the ultimate load at 247 sec, as a crack between the second
and third layers. The AE parametric values at 247 sec were: amplitude, 100 dB; energy, 2680
eu; rise time, 708 μs and duration, 53027 μs. Further damage was characterized by matrix
cracking and delamination growth. The amplitude of the AE signal was steady from the initial
failure until final collapse. Progressive damage was analysed using other parameters—mainly
the elastic energy release rate. Delamination between the third and fourth layers occurred at
416 sec with a duration of 5521 μs energy of 1091 eu and an amplitude of 91 dB. The third
highest peak in the duration plot occurred at 485 sec with a duration of 8602 μs, energy of
1475 eu and an amplitude of 95 dB. Further AE data was widely scattered, indicating
continuous matrix cracking and delamination growth. The rise in number of counts during the
failure progression is almost stagnant; a gradual increase in the number of counts is seen in
Figure 6.4(f). Final collapse of the structure occurred at 636 sec with an amplitude of 92 dB and
duration of 2331 μs at 573.6 N.

Page | 188
Page | 189
 Figure 6.4: HL L-bend AE correlation for Layup 2 specimen

6.3.3 L-bend Layup 3 (CSM+DB+UD) AE Correlation

Layup 3 had two CSM, two UD and three DB layers. The initial damage was observed at 370 sec
between the second and third layers, as shown in Figure 6.5. Although no reduction of load
was observed (due to the presence of UD fibres, which carried most of the load), other AE
parameters showed extensive matrix cracking and fibre breakage. At this stage, energy of 2103
eu, amplitude of 95 dB and event duration of 4076 μs were observed. The six peaks in the
energy and event-duration plots demonstrate progressive delamination until final collapse.
The first major damage was observed at 438 sec at 14 per cent of ultimate load, as
delamination between the second and third layers. The energy release was 6496 eu, amplitude
was 100 dB, rise time was 469 μs and duration was 20912 μs during the second stage of
damage. Further delaminations were observed at 550 sec (14 per cent of ultimate load), 587
sec (17 per cent), 639 sec (26 per cent) and final collapse at 720 sec (1800 N). Up to 26 per
cent of the ultimate load, the load-deflection plot was linear (until this point, the stiffness of
the structure remained constant). At this point, all the CSM and DB layers were completely
damaged and incapable of higher load increments.

The time of first failure can be determined by analysing the cumulative counts v. time plot
(Figure 6.5(f)). Until the failure initiation, the number of counts does not increase, and a
sudden increase in the slope of the curve indicates the occurrence of extensive damage in the
structure.

Page | 190
Page | 191
 Figure 6.5: HL L-bend AE correlation for Layup 3 specimen

6.3.4 Summary of L-bend AE Testing

Layup 2 showed signs of saturation with the accumulation rate being stagnant from initial
failure until final failure. Layups 1 and 3 showed an increase in the number of AE events—
demonstrating the damage progression. The accumulation rate of AE increased, even though a
decrease in load was observed, indicating crack progression in the laminate. All three layups
displayed extensive damage due to non-homogenous stress fields between the layers
composed of strongly anisotropic plies with different fibre orientations.

Failure initiation is triggered by matrix cracking, delamination due to interlaminar stresses and,
finally, fibre failure due to in-plane tension. A structure with a smooth curve in the cumulative
counts v. time plot is generally preferred, such as for Layup 2. Layup 1 had a steep increase in
the slope (cumulative counts v. time plot), which indicates an undesirable rapid degradation of
the structure within a short time span. Layup 3 has two different slopes and the second slope
is much steeper than the first.

The next section focuses on an AE investigation into Top-hat-stiffeners for keel structures.

6.4 Hand-laid Top-hat-stiffener AE Failure Characterisation

AE interrogation on L-bend composite specimens was carried out. The damage mechanism of
curved specimens was investigated and failure mode was determined, based on the AE
parameters. Similar to the earlier test, this AE study on Top-hat-stiffeners analyses and
discriminates among the failure modes. The parametric AE plots for Layups 1 and 2 are similar
to that of Layup 3, which is discussed in this section. As their failure patterns are similar to that
of Layup 3, the AE plots for Layups 1 and 2 are presented in Appendices to avoid repetitive
discussion. The sensor placement location on the Top-hat-stiffener is shown in Figure 6.6.
Page | 192
 Figure 6.6: AE experimental setup and location of piezo-sensors on Top-hat-stiffener

6.4.1 Hand-laid Top-hat-stiffener Layup 3 AE Failure Characterisation

Analysis of AE parameters provides a good insight into the failure characterisation of laminate
composite structures. The amplitude distribution and parametric analysis for Layup 3 are
presented in Tables 6.5 and 6.6. The load v. time, and corresponding AE parameters v. time
plots are presented in Figure 6.7.

Similar to the results discussed in the L-bend test, six different AE parameters were used to
analyse the failure of Top-hat-stiffeners. Until the first failure, very little AE activity was
observed, as the structural strength was intact. The first failure occurred at 185.1 sec with a
load drop of 1029 N. The duration was 83.535 ms, which is a long duration, with an energy
release of 12645 eu. Although there were 5498 counts, the rise time was 789 μs,
corresponding to delamination failure, and the experimental and FEA analysis results
confirmed the same. The next two failures were delamination propagation with a low drop in
load but a significant energy release of 16000 eu and event duration of 246 ms.

Page | 193
Channel Total (40-45) (45-50) (50-55) (55-60) (60-65) (65-70) (70-75) (75-80) (80-85) (85-90) (90-95) (95-100)
1 234 23 105 55 29 8 6 2 2 0 0 3 1
2 19023 1657 8892 5024 2215 718 257 150 58 29 13 6 4
3 42510 1472 13555 11609 8296 4515 1768 675 330 148 99 28 15
4 11882 878 5737 2922 1405 544 227 101 40 10 9 4 5
Table 6.5:HL Top-hat-stiffener Layup 3, four-channel amplitude distribution ranges (dB)

Failure Time Load Load drop Amplitude Energy (eu) Duration Rise Time Counts (Threshold Failure
No. (Sec) (N) (N) (dB) (1eu=10-18 J) (ms) (μs) Crossings) Mode
1 185.1 15228 1029 100 12645 83.535 789 5498 Delamination
2 238.7 19077 637 100 16080 246.115 584 23390 Delamination
3 251.9 18768 937 100 16372 249.488 544 23734 Delamination
4 294.6 19960 540 100 4622 21.322 800 375 Crack Progression
5 310.7 20252 291 100 4374 19.717 493 314 Crack Progression
6 398.9 26055 118 75 2234 72.016 125 2812 Crack Progression
7 428.2 29348 702 100 10990 70.193 383 3145 Crack Progression
8 431.6 27908 6232 100 16355 195.330 769 9626 Delamination
9 501.9 26953 4374 100 8113 20.855 516 753 Crack Progression
10 537.8 22233 452 100 9827 248.687 1015 10153 Delamination
11 587.6 26551 7362 100 6831 13.512 840 536 Crack Progression
12 652.5 24692 2585 78 171 10.385 1096 199 Crack Progression
13 909.0 46703 1424 100 32211 63.061 488 3701 Crack Progression
Table 6.6: HL Top-hat-stiffener Layup 3 AE failure characterisation

Page | 194
The next two load drops were comparatively low and the energy release was 375 and 315 eu
at 294.6 and 310.7 sec, respectively. These correspond to fibre/matrix debonding, as the
durations were 21.32 and 19.71 ms. The debonding was due to crack propagation. The next
failure was thought to be delamination propagation, as the amplitude was 75 dB with rise time
of 125 μs, energy release of 2234 eu and a load drop of 118 N at 399 sec. An energy release of
10990 eu and duration of 70.19 ms corresponds to major delamination propagation at 428.2
sec. At 431.6 sec, another delamination occurred at both top bends —CSM and DB fibres began
to shear with an energy release of 16355 eu, duration of 195.33 ms and a huge load drop of
6232 N. CSM and DB fibre breakage continued until 652.5 sec with progressive fibre fracture.
Although DB showed a higher strain-to-failure ratio than CSM and UD, it was unable to carry
the load. Once all the CSM fibres were sheared, most of the UD fibres sustained the load until
909 sec. Between 678 and 909 sec, very little AE activity was observed, as the UD fibres were
in tension and most of the CSM fibres were completely damaged at the edge of the steel
backing plate. The test was terminated at 909 sec as the specimen began slipping from the test
equipment at 45 kN.

The cumulative counts plot (Figure 6.7(f)) shows the extent of damage in the material. The
failures could be divided into three distinctive zones. Zone 1 is up to 185 sec, where very little
AE activity was observed; zone 2 is from 185 to 587.6 sec, where a steep increase in the slope,
representing extensive damage, is observed. The number of counts increased from 4123 to
1317424 within a span of 402.6 sec. Zone 3 is from 587.6 sec to final failure at 906.8 sec. The
slope was lower and the number of counts was 213576, indicating that most damage due to
delamination occurred during Zone 2. AE plots for Layups 2 and 3 are presented in Appendix D.

Page | 195
Figure 6.7: Hand-laid Top-hat-stiffener Layup 3 specimens—AE parameter correlation
Page | 196
6.5 AE Detection with Optical Fibre Sensors
Modern NDE approaches provide an excellent opportunity to use a variety of AE techniques
with composites. The conventional AE sensing mechanism is based on highly sensitive
resonant piezoelectric sensors that are typically operated in a limited frequency range. This
resonant frequency can produce some loss of information from the AE signal. Their relatively
large size makes it difficult to embed them in composite structures. They are generally
susceptible to strong electric fields and are unsuitable for corrosive, high temperature and
hazardous environments. Each AE sensor is used in a channel, which includes a preamplifier
and cables. This makes them heavy and bulky—awkward to use with laminated composite
structures.

An alternative technique uses fibre Bragg grating (FBG) sensors. A FBG can be written on an
optical fibre, which can be embedded in a composite laminate without significantly altering the
strength and stiffness of the structure. Multiplexing FBGs for AE detection increases the
number of sensors without actually increasing the weight of the test equipment.

An optical fibre with a single FBG sensor was surface bonded to VIP specimens and the AE was
recorded during the testing. To validate the FBG sensor, four piezoelectric sensors were also
surface bonded on the outer surface of the Top-hat-stiffener, as shown in Figure 6.8. FBG
manufacturing and equipment setup is presented in the following section.

Figure 6.8: AE-FBG and AE piezo-sensors surface bonded to the Top-hat-stiffener

Page | 197
6.5.1 Phase-shifted Fibre Bragg Grating AE Measurement System

As shown in Figure 6.9, the designed acoustic measurement system consists of a tuneable laser
system, as the interrogation source, and a phase-shifted FBG (PS-FBG) as the sensor head.

Figure 6.9: Intensity type AE sensor designed for failure monitoring in the current study

As distinct from the previous work, the FBG in this study incorporates a centrally located π-
phase shift. This design creates a narrow transmission channel at a Bragg wavelength (1556.13
nm), as shown in Figure 6.10(a). The laser output spectrum is indicated in Figure 6.10(b).

0 10

0
-10
Transmission (dB)

-10
Power (dBm)

-20 -20

-30
-30
-40

-40 -50
1555 1555.5 1556 1556.5 1557 1554 1555 1556 1557 1558
Wavelength (nm) Wavelength (nm)

Figure 6.10: (a) Transmission spectrum of the fabricated PS-FBG; b) Power spectrum of the tuneable laser
diode

The maximum power output from the laser is 5 mW. The laser has a linear power transition,
which means that it can achieve a linear sensing response. Prior to measurement, the laser
wavelength was lightly offset from the centre wavelength, as shown in Figure 6.11. TLS
corresponds to Tunable laser source. The offset point that offers the best compromise
between sensitivity and dynamic range is 0.026 nm. Table 6.7 summarised the properties of
the fabricated PS-FBG, tuneable laser and the sensing performance.

Page | 198
 0 20
Offset

-10 0
Transmission (dB)

Power (dBm)
PS-FBG TLS
-20 -20

-30 -40

-40 -60
1555 1555.5 1556 1556.5 1557
Wavelength (nm)

Figure 6.11: Adjustment of laser wavelength for optimum operation

The slope of the narrow transmission band of the PS-FBG is much higher than that of a normal
FBG; therefore, higher sensitivity and resolution are expected. This scheme uses direct
transmission intensity, which eliminates the switching of optical intensity from port 2 to port 3.
This reduces power loss at the circulator, which means further enhancement of sensitivity. A
circulator is the interface between the light source and sensor that blocks the unused reflected
light from the destabilised laser diode.

Photo sensitive Fibre Bragg Grating

Fibre type Hydrogen loaded single mode fibre (SMF-28)
Non-apodised, centrally located π phase
Fabrication profile
shift
Length 8 mm
3 dB / 20 dB passband bandwidth 4 pm / 20 pm
Maximum stopband loss 27 dB
Tuneable laser diode
Maximum configured power 10 mW
3 dB bandwidth 50 pm
Usable offset 60 pm
Optimum offset 26 pm

Page | 199
 Sensing performance
Dynamic range ~7.5 MPa (using 26 pm offset)
Frequency range 512.5 kHz (half of maximum sampling rate)
Rate of power change per wavelength
35 μW/pm
(maximum),
Rate of power change per wavelength (average) 0.12 pW/pm
Rate of power change per pressure 3.47 × 10-6 pm/Pa
Receiver gain 2V/mW
DAQ card type National Instrument PCI-6251
DAQ card input sensitivity 6 μV
DAQ card input maximum limit 5V
Table 6.7: Properties of PS-FBG and tuneable laser diode, and sensing performance of the designed
system

6.5.2 Pencil Lead Break Test

The arrangement for the pencil lead break test of the PS-FBG sensor is shown in Figure 6.12.
An electric microphone compared the resulting acoustic wave from the pencil lead breakage.
The distance between the breaking point and acoustic sensors was set at eight centimetres.
Lab view 8.0 software provided data acquisition, DSP and recording during the real-time
measurements.

Composite specimen surface

Electronic
microphone

8 cm

Breaking position
PS-FBG

Figure 6.12: Arrangement for the pencil lead break test

The sampling rate was set to 100 kHz and the examples of the recorded waveform shown in
Figure 6.13 suggest that the designed sensor is sensitive enough for acoustic detection and is
comparable to the electronic microphone.

Page | 200
 Figure 6.13: Comparison of waveforms recorded from the pencil lead break test between electronic
microphone and PS-FBG AE sensor

6.5.3 Deployment of the PS-FBG Sensor in Continuous Monitoring of

Composite Failures

An acoustic coupling gel was used to ensure good acoustic coupling between the composite
surface and the PS-FBG. The gel also stabilised the temperature around the sensor. The data
acquisition sampling rate was set to its maximum value (1 MS s-1) resulting in a maximum
detectable frequency of 500 kHz. Each data array has 32768 data points (fifteen bits);
therefore, the number of data recorded per second was limited to four, due to the processing

Page | 201
power required to handle the large amount of data. The acoustic waveform was automatically
saved by a trigger level when its amplitude exceeded a certain level. The laser operating point
was offset by 40 pm with maximum lasing power. The PS-FBG spectrum was checked
frequently and the offset adjusted to maintain constant sensing conditions for all
measurements.

6.5.4 Fibre Bragg Grating (FBG) Characterisation

FBG measurement is necessary at each stage of production to assess the grating growth using
certain fabrication parameters. The required adjustment for the next writing process is known
once the current sample measurements are taken. The setup for simultaneously measuring the
transmission and reflection spectra of a FBG consists of a tuneable laser diode and coupler, as
shown in Figure 6.14. Simultaneous measurement is provided by dual detectors packaged in
the system. During measurement, the FBG was held in a straight orientation and fixed loosely
on a thermal conducting plate, so that it is not subjected to strain and temperature variations.
The significant parameters are the Bragg wavelength, three-dB bandwidth and transmission
loss at the Bragg wavelength.

Figure 6.14: Setup for FBG characterisation

6.6 VIP THS AE Failure Characterisation

Similar to Section 6.5, AE recording was performed on three different layups with five
specimens of each layup. Along with the piezoelectric sensing technique, a surfaced bonded
FBG sensor is used to assess the validity of the FBG system and to compare it to the
piezoelectric sensing system. The results for Layup 4 are discussed in this chapter; the results
for Layups 5 and 6 are presented in Appendix E to avoid repetitive explanation.
Page | 202
6.6.1 VIP THS Layup 4 Top-hat-stiffener AE Testing - Setup

Four piezoelectric sensors were surface bonded at the four corners of the Top-hat-stiffeners;
the other end was connected to the DISP processor and a laptop computer. A FBG AE sensor
was surface bonded at the top of the web. The setup for the AE-FBG measurement is shown in
Figure 6.15 where a personal computer is connected to a optical-fibre interrogator. Labview
8.0 software was used for recording and analysing the data.

Figure 6.15: AE-FBG setup during the test process

6.6.2 VIP THS Layup 4 (CSM+DB)—AE Parametric Analysis

Parametric analysis of the piezoelectric and FBG sensor data are discussed in this section.
Table 6.8 presents the amplitude distribution range for the Layup 4 specimens. Only a couple
of signals reached amplitude of 100 dB. Most were situated between 40–60 dB, indicating
crack propagation in the laminate. Table 6.9 presents the failure characterisation of the
piezoelectric sensor. Four signals which were delamination failures were observed and the rest
of the signals were crack-propagation signals. As the peak amplitude for most of the failures is
99 dB, the amplitude parameter is not considered for failure-mode discrimination.

Page | 203
The initial failure was thought to be matrix cracking with a small drop in load of 7 N at 224.66
sec, and a high energy release and duration of 5914 eu and 25.80 ms. The major crack
occurred at 299.83 sec, delamination failure with a significant drop of 484 N. The signal lasted
18.88 ms and energy released was 6549 eu. The signal at 320.33 sec corresponded to crack
propagation with a 437 N drop in the load, but the rise time and counts were just 0.56 ms and
467. A large energy release, of 10358 eu, was observed at 422.50 sec with a load drop of 1575
N corresponding to a delamination failure. The duration and counts were 68.51 ms and 3039,
respectively. Further loading caused the cracks to propagate in the laminate. Three peak
values were determined at 486.50, 530.77 and 592.52 sec. The signal duration was small
compared to the delamination signal. Another major crack occurred at 655.54 sec with a 1122
N drop in load. The parametric values of energy, duration, rise time and counts were 5804 eu,
17.70 ms, 0.86 ms and 347 respectively. Final collapse, an observed drop in load of 12457 N
was assumed a fibre failure.

The plots of amplitude, energy, duration, rise time, counts and signal strength are presented in
Figures 6.16 to 6.21.

Page | 204
 Channel Total (40-45) (45-50) (50-55) (55-60) (60-65) (65-70) (70-75) (75-80) (80-85) (85-90) (90-95) (95-100)
1 4086 652 1625 838 477 272 144 48 17 3 6 0 4
2 7375 1040 3225 1484 762 455 245 107 34 9 5 1 8
3 7165 873 2944 1578 829 439 240 137 70 34 11 2 8
4 9151 1148 3864 2110 1025 529 265 127 54 18 5 1 5
Table 6.8: VIP THS Layup 4, four-channel amplitude distribution ranges (dB)

Failure Time Load Load drop Amplitude Energy (eu) Duration Rise Time Counts Signal Failure
No. (Sec) (N) (N) (dB) (1eu=10-18J) (ms) (ms) (Threshold Crossings) Strength Mode
1 224.66 7571 7 99 5914 25.80 1.61 406 36945024 Matrix Cracking
2 299.83 10818 484 99 6549 18.88 2.74 315 40910904 Delamination
3 320.23 11404 223 99 3927 18.33 8.45 927 24532492 Crack progression
4 345.35 11254 437 99 3654 22.12 0.56 467 22830236 Delamination
5 383.36 12337 72 99 4890 39.38 26.37 2290 41089040 Crack progression
6 422.50 14200 1575 99 10358 68.51 1.43 3039 64706032 Delamination
7 486.50 16150 93 66 53 1.55 0.22 85 536223 Crack progression
8 530.77 17456 631 99 7641 17.23 9.93 410 47732564 Crack progression
9 592.52 19690 257 78 258 6.08 0.22 129 1617729 Crack progression
10 655.54 24699 1122 99 5804 17.70 0.86 347 36260192 Delamination
11 743.00 30345 12457 99 8659 21.64 15.60 1433 54093544 Fibre failure
Table 6.9: VIP THS Layup 4, AE parametric analysis for piezoelectric sensor

Page | 205
6.6.3 VIP THS Layup 4 Parametric Analysis—FBG-AE sensor

The failure values of the FBG–AE sensor are presented in Table 6.10. The values are
comparatively lower than those for a piezoelectric sensor but the trend is similar to piezo-
sensors; failure modes were successfully identified using these values. Only one FBG AE sensor
was used, whereas, four piezo-sensors were used. Parameters such as amplitude, energy,
duration, rise time and counts are compared with the piezo-sensor values in Figures 6.16 to
6.20.

Energy
Rise Counts
Failure Time Load Amplitude (eu) Duration Failure
Time (Threshold
No. (Sec) (N) (dB) (1eu=10- (ms) Mode
18 (μs) crossings)
J)
1 224.66 7571 74.27 13074 15.39 6.87 3231 Matrix Cracking
2 299.83 10818 73.45 1048 11.19 3.80 1063 Delamination
Crack
3 320.23 11404 63.92 386 8.68 0.52 936
progression
4 345.35 11254 66.52 1337 3.24 3.66 458 Delamination
Crack
5 383.36 12337 61.35 276 18.36 1.50 1974
progression
6 422.50 14200 52.10 1308 5.97 2.83 230 Delamination
Crack
7 486.50 16150 62.08 760 9.78 4.51 840
progression
Crack
8 530.77 17456 49.31 2453 10.32 1.56 804
progression
Crack
9 592.52 19690 64.02 1612 9.59 1.90 1112
progression
10 655.54 24699 71.22 227 6.26 2.58 521 Delamination
11 743.00 30345 72.94 1707 10.77 7.33 412 Fibre failure
Table 6.10: VIP THS Layup 4, AE parametric analysis for FBG-AE sensor

Page | 206
 Figure 6.16: VIP THS Layup 4—AE comparison of piezo and FBG sensor amplitude

Figure 6.17: VIP THS Layup 4—AE comparison of piezo and FBG sensor energy released

Page | 207
Figure 6.18: VIP THS Layup 4—AE comparison of piezo and FBG sensor signal duration

Figure 6.19: VIP THS Layup 4—AE piezo and FBG sensor signal rise time comparison

Page | 208
Figure 6.20: VIP THS Layup 4—AE piezo and FBG sensor count comparison

Figure 6.21: VIP THS Layup 4—AE piezo-sensor signal strength

Page | 209
The overall damage process could be characterised into three distinctive zones. Figure 6.22
shows the cumulative counts v. time and load v. time plots.

Figure 6.22: VIP THS Layup 4—AE piezo and FBG sensor cumulative counts and load v. time plots

Zone 1 covers up to 210 sec, where limited AE activity is observed, because of environmental
noise and machine vibration. The number of counts recorded in this zone was 2395. Zone 2
corresponds to the time between 210 and 413 sec, where a steep slope that indicates
extensive damage in the structure is observed. There were 211,051 counts within a span of
203 sec and most critical failures, which proved to be delaminations, occurred during this
stage. Zone 3 consisted mainly of crack propagation and an ultimate failure between 413 and
Page | 210
743 sec. The slope of the curve was lower with 115,905 counts within a span of 330 sec. The
histogram of hits v. time and the plot of cumulative hits v. time for the piezo-sensor are shown
in Figures 6.23 and 6.24, respectively. Comparing the three stages, it is evident that stage two
had higher rate of hits, indicating large damage to the structure.

Figure 6.23: VIP Top-hat-stiffener Layup 4—piezo-sensor hits and load v. time plot

Figure 6.24: VIP Top-hat-stiffener Layup 4—piezo-sensor cumulative hits and load v. time plot

6.6.4 VIP THS - Layup 4 Waveform Failure Characterisation for

Piezoelectric and FBG-AE Sensor

Waveform analysis of the recorded transient data of the piezoelectric sensor is presented in
this section. The peak AE signal voltage was recorded and a frequency analysis was performed
by Fast Fourier transform (FFT) with the commercial software WinPost from the Physical
Acoustics Corporation (PAC). The FFT spectra of the recorded waveforms were compared and
changes in the FFT spectra were noted as a function of damage evolution. Table 6.11 presents
the failure characterisation of the Layup 4 specimen using the signal peak voltage and FFT peak
magnitude. Waveform analysis of the transient data from the piezoelectric sensor is presented
in this section.

Page | 211
 Failure Time Load Load drop Voltage Peak FFT Failure
No. (Sec) (N) (N) (mV) Magnitude (kHz) Mode
1 224.66 7571 7 0.490 55.894 Matrix Cracking
2 299.83 10818 484 9.608 647.614 Delamination
3 320.23 11404 223 0.235 4.613 Crack progression
4 345.35 11254 437 9.913 874.377 Delamination
5 383.36 12337 72 0.246 8.795 Crack progression
6 422.50 14200 1575 0.201 4.356 Crack progression
7 486.50 16150 93 N/A N/A N/A
8 530.77 17456 631 1.228 87.692 Crack progression
9 592.52 19690 257 0.760 26.505 Crack progression
10 655.54 24699 1122 9.902 617.553 Delamination
11 743.00 30345 12457 1.214 61.479 Fibre failure
Table 6.11: Waveform analysis of VIP THS Layup 4 specimens–piezoelectric sensor

The designed sensor has a non-uniform frequency response, which is typically encountered in
an unpackaged optical-fibre sensor. The sensor has a peak response between 2 and 8 kHz and
weak response above 20 kHz. In the following analysis, the response of the optical-fibre sensor
is divided into a region below 20 kHz and one above 20 kHz, so that the peak FFT can be
recognised for acoustic signals above 20 kHz. Note that failures 1, 8, 9 and 11 (between 20 and
500 kHz) can be recognised by separating the signal FFT into regions below and above 20 kHz.
The FBG-AE failure waveform values are tabulated in Table 6.12. The AE waveforms and FFT
spectra for the piezo and FBG-AE sensors at the point of each distinctive failure are presented
in Figures 6.25 to 6.35.

Amplitude Peak FFT Peak FFT

Failure No. Comment
(mV) (< 20 kHz) (> 20 kHz)
1 150.7 5.951 55.050 Weak signal detected at 55 kHz
2 33.11 4.059 23.830 Signal beyond frequency range
3 12.34 4.883 N/A Signal beyond frequency range
4 13.33 14.95 24.41 Signal beyond frequency range
5 9.247 7.172 N/A Signal beyond frequency range
6 40.98 2.136 73.67 Half
7 35.16 7.507 24.6 NA
8 57.42 3.265 76.69 Higher freq signal detected at 76.69 kHz
9 54.27 5.157 24.66 Higher freq signal detected at 24.66 kHz
10 12.45 7.599 20.87 Signal beyond freq range
11 60.26 6.104 57.10 Higher freq signal detected at 57.10 kHz
Table 6.12: Waveform analysis of VIP THS, Layup 4 specimens–FBG-AE sensor
Page | 212
 0.2

0.15

0.1

Voltage (V)
0.05

-0.05

-0.1

-0.15
0 5 10 15 20 25 30
Time (ms)

Figure 6.25 (a): Piezo and FBG-AE sensor AE waveforms at 224.66 sec – failure 1

-5
2 x 10
6

5
1.5
Amplitude (V)

4
Amplitude (V)

1 3

2
0.5
1

0 0
0 10 20 50 60 70 80
Frequency (kHz) Frequency (kHz)

Figure 6.25 (b): Piezo and FBG-AE sensor AE FFT at 224.66 sec–failures 1

0.2

0.15

0.1
Voltage (V)

0.05

-0.05

-0.1

-0.15
0 5 10 15 20 25 30
Time (ms)
Figure 6.26 (a): Piezo and FBG-AE sensor AE waveform at 299.83 sec–failures 2

Page | 213
 -3
x 10
1.4 3

1.2
2.5

1
2
Amplitude (V)

Amplitude (V)
0.8
1.5
0.6
1
0.4

0.2 0.5

0 0
0 5 10 15 20 20 25 30 35 40
Frequency (kHz) Frequency (kHz)

Figure 6.26 (b): Piezo and FBG-AE sensor AE FFT at 299.83 sec–failures 2

0.015

0.01

0.005
Voltage (V)

-0.005

-0.01

-0.015
0 5 10 15 20 25 30
Time (ms)
Figure 6.27 (a): Piezo and FBG-AE sensor AE waveform at 320.23 sec–failures 3

-3
x 10
4

3.5

3
Amplitude (V)

2.5

1.5

0.5

0
0 5 10 15 20
Frequency (kHz)

Figure 6.27 (b): Piezo and FBG-AE sensor AE FFT at 320.23 sec–failures 3

Page | 214
 0.015

0.01

0.005

Voltage (V)
0

-0.005

-0.01

-0.015
0 5 10 15 20 25 30
Time (ms)

Figure 6.28 (a): Piezo and FBG-AE sensor AE waveform at 345.35 sec–failures 4

-4 -5
x 10 x 10
2

1.5
2
Amplitude (V)

Amplitude (V)

1
0.5

0 0
5 10 15 20 20 25 30 35 40
Frequency (kHz) Frequency (kHz)

Figure 6.28 (b): Piezo and FBG-AE sensor AE FFT at 345.35 sec— failures 4

0.01

0.005
Voltage (V)

-0.005

-0.01
0 5 10 15 20 25 30
Time (ms)

Figure 6.29 (a): Piezo and FBG-AE sensor AE waveform at 383.36 sec— failures 5
Page | 215
 -4
x 10
3.5

2.5

Amplitude (V)
2

1.5

0.5

0
5 10 15 20
Frequency (kHz)

Figure 6.29 (b): Piezo and FBG-AE sensor AE FFT at 383.36 sec— failures 5

0.05

Voltage (V)
0

-0.05
0 5 10 15 20 25 30
Time (ms)

Figure 6.30 (a): Piezo and FBG-AE sensor AE waveform at 422.50 sec— failures 6

-3 -5
x 10 x 10
1.5 1.4

1.2

1
1
Amplitude (V)

Amplitude (V)

0.8

0.6
0.5
0.4

0.2

0 0
0 5 10 15 20 60 65 70 75 80
Frequency (kHz) Frequency (kHz)

Figure 6.30 (b): Piezo and FBG-AE sensor AE FFT at 422.50 sec—failure 6

Page | 216
 0.04

0.03

0.02

Voltage (V)
0.01

-0.01

-0.02

-0.03
0 5 10 15 20 25 30
Time (ms)

Figure 6.31 (a): FBG-AE sensor AE waveform at 486.50 sec— failures 7

-5
0.02 x 10
4

0.015 3
Amplitude (V)

Amplitude (V)
0.01 2

0.005 1

0 0
0 5 10 15 20 20 25 30 35 40
Frequency (kHz) Frequency (kHz)

Figure 6.31 (b): FBG-AE sensor AE FFT at 486.50 sec— failures 7

0.06

0.04

0.02
Voltage (V)

-0.02

-0.04

-0.06
0 5 10 15 20 25 30
Time (ms)

Figure 6.32 (a): Piezo and FBG-AE sensor AE waveform at 530.77 sec— failures 8

Page | 217
 -6
x 10
0.07 6

0.06 5

0.05
4
Amplitude (V)

Amplitude (V)
0.04
3
0.03
2
0.02

0.01 1

0 0
0 5 10 15 20 60 65 70 75 80 85 90
Frequency (kHz) Frequency (kHz)

Figure 6.32 (b): Piezo and FBG-AE sensor AE FFT at 530.77 sec— failures 8

0.06

0.04

0.02

Voltage (V)
0

-0.02

-0.04

-0.06
0 5 10 15 20 25 30
Time (ms)

Figure 6.33 (a) Piezo and FBG-AE sensor AE waveform at 592.52 sec— failures 9

-5
-3 x 10
x 10 8
4

3.5 7

3 6
Amplitude (V)
Amplitude (V)

2.5 5

2 4

1.5 3

1 2

0.5 1

0 0
0 10 20 30 40 50 20 25 30 35 40
Frequency (kHz) Frequency (kHz)

Figure 6.33 (b): Piezo and FBG-AE sensor AE FFT at 592.52 sec— failures 9

Page | 218
 0.015

0.01

0.005

Voltage (V)
0

-0.005

-0.01

-0.015
0 5 10 15 20 25 30
Time (ms)

Figure 6.34 (a): Piezo and FBG-AE sensor AE waveform at 655.54 sec—failure 10

-3
x 10
4

3.5

Amplitude (V)
2.5

1.5

0.5

0
0 10 20 30 40 50
Frequency (kHz)
-5
x 10
8

6
Amplitude (V)

0
20 25 30 35 40
Frequency (kHz)
Figure 6.34 (b): Piezo and FBG-AE sensor AE FFT at 655.54 sec—failure 10

2
Voltage (V)

-2

-4

-6

-8
0 5 10 15 20 25 30
Time (ms)

Figure 6.35 (a): Piezo and FBG-AE sensor AE waveform at 743.00 sec—failure 11
Page | 219
 140 1.4

120 1.2

100 1

Amplitude (V)
Amplitude (V)

80 0.8

60 0.6

40 0.4

20 0.2

0 0
0 10 20 30 40 50 50 60 70 80 90 100
Frequency (kHz) Frequency (kHz)

Figure 6.35 (b) Piezo and FBG-AE sensor AE FFT at 743.00 sec—failure 11

A visual comparison yielded distinctive differences between the FFT spectra and signal voltage
with damage growth. There seemed to be a continuous trend with the increasing damage in
the laminate. The waveform for failure seven at 486.5 sec is not shown, as the peak voltage
and FFT magnitude was much lower than that of the other failures, which was deemed small
crack propagation. For delamination initiation, the FFT spectra were always above 600 kHz and
the voltage more than 9 mV. Delamination progression with the FFT ranged between 4 and 90
kHz, while the voltage was less than 1.3V. It was difficult to distinguish matrix cracking from a
fibre failure based on the FFT magnitude as they were at 55.89 and 61.48 kHz, respectively.
The peak voltages for matrix cracking and fibre failure were 0.49 and 1.214 mV.

6.7 AE Conclusions
Real-time damage detection and identification of damage events for curved composite
structures using AE signals is presented in this chapter. Two different types of AE sensing
mechanisms, conventional piezoelectric sensors and optical-fibre AE sensors were used to
obtain AE signals during the damage process. The different analyses of standard piezo-sensor
AE signal parameters, frequency spectra of AE waveforms and FBG-AE signal parameters all

Page | 220
yielded indications of changes correlating with the growth of damage in the composite
laminate. Four distinctive failure modes were determined using these analyses; matrix
cracking, delamination initiation, delamination propagation and fibre failure. Significant
differences in the peak amplitude and frequency content were seen among the different
failure modes. Although the waveform analysis provided sufficient information to distinguish
between delamination and crack propagation, it was not possible to identify the failure modes
clearly as matrix cracking and fibre failure. Delamination and crack propagation are the
primary failure modes of curved composite structures; therefore, the proposed damage
mechanism is acceptable. Parametric analysis provides more promising results in
discrimination among failures.

A novel non-destructive evaluation (NDE) approach to use optical fibre sensor as acoustic-
emission sensor has been proposed. In the current study, one FBG-AE sensor was compared
with four piezo AE sensors. As a result, fewer hits were recorded with the FBG-AE sensor. The
FBG-AE sensor can be embedded or surface mounted (which may be re-used) and multiplexed
to minimise the hardware needed. Parameters such as amplitude, signal duration, energy
released, counts and signal rise time were determined from the waveform, using commercially
available data-analysis software. These parameters from the FBG sensor were compared with
the piezoelectric sensors and the results are in good agreement.

Page | 221
Chapter 7: Optical-fibre Sensors

Li H.C.H.(2005). Damage monitoring in large composite ship joints using smart sensors, PhD thesis,
RMIT University, Melbourne, Australia

Page | 222
7.1 Introduction
Fibre Bragg gratings (FBGs) are excellent strain sensors for marine/aerospace structure
applications. In Chapter 6, FBG-AE sensors were used to acquire stress waves from Top-hat-
stiffener subjected to static loading. The application of FBG for the damage monitoring of
laminated composite structures is dealt in this chapter. The manufacturing procedure for FBG
as a strain sensor is explained and the test setup process is detailed. In the AE study in Chapter
6, a FBG-AE sensor was surface bonded to the specimen; however, in this chapter, four optical-
fibre sensors (FBG) were embedded in the curved section of the Top-hat-stiffener where the
maximum strain variation is expected. A minor change in FBG wavelength after the specimen
manufacture process is observed due to temperature variation and residual stresses during
curing. A procedure for successfully embedding the optical fibres in the curved section of the
Top-hat-stiffener is presented in Appendix G. This procedure provides satisfactory sensor
performance and offers adequate protection of the optical fibres at the ingress and egress. An
international conference has been published on the results of embedding FBG strain sensor 9.

Top-hat-stiffeners embedded with optical fibres at appropriate locations were manufactured

with VIP. Five specimens of each of three layups (Layups 4, 5 and 6) were tested until failure.
The layup details were discussed in earlier chapters. Local strain variation during the damage
process was observed and the abnormal strain variations were captured successfully. Many
researchers have applied the FBG strain sensing technique to flat specimen monitoring
[Ussorio et al., 2006; Wang et al., 2005; Xu et al., 2005; McKenzie et al., 2000; Li et al., 2006;
Okabe et al., 2000; Takeda, 1999; Rowe et al., 1986; Everall et al., 2000; Davies et al., 1994;
Froggatt and Moore, 1998; Tao et al., 2000; Leung et al., 2005; Elvin and Leung, 1997; Kesavan
et al., 2008; Grouve et al., 2007; Ling et al., 2005a and 2005b; Sorensen et al., 2007;Takeda et
al., 2005]. In this study, strain sensing is performed on the curved section of the structure
where huge strain variation could be observed.

9
Azmi A.I., Raju, Peng G.D. (2012). Application of fiber grating-based acoustic sensor in progressive
failure testing of e-glass/vinylester curve composites. SPIE 8351, 835110, Proceedings of Asia Pacific
Optical Sensors (APOS) Conference 2012, Sydney, Australia.
Page | 223
7.2 Fabrication of Fibre Bragg Grating (FBG) sensor
The grating for the sensor was written using the phase-mask technique at the University of
New South Wales [Photonics and Optical Communications, School of Electrical Engineering and
Telecommunications] writing facilities. As shown in Figure 7.1, the writing system consists of
three main components: an ion laser, grating fabrication enclosure and control system. The ion
laser produces output power at 488 nm, which is then frequency doubled at 244 nm by a beta-
barium borate (BBO) crystal. The standard operating power for the laser is between 200 and
300 mW at a tube current of 50 A. Standard safety procedures were implemented for working
with the Class 4 laser, including wearing ultra-violet (UV) goggles and a protective coat. UV
radiation exposure was minimised even when working with low power levels. The UV beam is
aligned from the laser front towards the phase mask through multiple optics components, as
shown in Figure 7.1.

Figure 7.1: Setup of FBG writing system in UNSW (Photonics lab)

At the turning mirror, a small portion of the beam (~1 per cent) is passed through to a
photodetector to observe beam vibration and provide feedback to the second tip/tilt mirror to
compensate for the vibration. The UV beam was highly focused by a cylindrical lens so that the
maximum UV power enters the fibre core. When the UV beam incident is on the phase mask,
the diffracted plus and minus one orders are maximized (each contains approximately 35 per
cent of the total incident power), while the zero order is suppressed. This scenario is shown in
Figure 7.2.

Page | 224
 Figure 7.2: Phase-mask technique for writing FBG

For writing a grating, a section of stripped optical fibre is placed close to the phase mask and
clamped to secure its position. The UV beam is scanned along the intended FBG length by
moving the translation stage at a controlled speed. A near field fringe pattern (grating) is
formed on the fibre by the plus and minus first-order diffracted beam. The period of the fringe
pattern is half of the phase mask period (ΛB=Λpm/2). The strength of the fringe pattern is
controlled by the speed of the translation stage and the laser output power. For fabrication,
several phase mask periods were used (1056.5, 1064.7, 1076.2 and 1081.9 nm) to obtain Bragg
wavelengths of 1529, 1540, 1555 and 1565 nm, respectively. Different Bragg wavelengths were
attained by applying different tensions during writing. A tension of approximately 100 g
produces about one-nm wavelength spacing. The control unit handles all process of writing,
gives feedback to system and monitors the equipment status.

Once the writing process is completed, the grating position is labelled and annealed at 300qC
for five minutes. Annealing is required to age the grating and to stabilise the wavelength and
spectral drifts. If the grating needs to exhibit stable long-term behaviour, either a
hydrogenated or non-hydrogenated fibre is required. The final process is to bake the entire
fibre in oven at 80qC for twenty-four hours to remove remaining hydrogen from other parts of
the fibre. Then the FBG samples are ready for experimental testing. The writing facilities and
fabrication procedures yield highly reliable and repeatable processes while producing excellent
quality of FBG. Figures 7.3 to 7.5 show the main equipment and some of the processes
involved in manufacturing FBGs. Figure 7.3(a) shows the procedure of laser alignment prior to
FBG fabrication, while Figure 7.3(b) shows the in-process FBG manufacturing control unit.

Page | 225
 a b

Figure 7.3(a): Laser alignment being performed prior to writing process to optimise lasing power and (b)
the grating fabrication enclosure (left) and control unit (right) (Photonics lab)

Figure 7.4 shows a close-up view of the FBG writing equipment, while Figure 7.5 shows the
annealing equipment used in this study.

a b
Cylindrical lens

Translation stage Fibre clamp

Phase mask holder

Unused
phase mask
Granite block

Figure 7.4: a) Inside view of the grating writing enclosure before; b) during writing process

Figure 7.5: Equipment for annealing FBG (Photonics lab)

Page | 226
The optical fibre is hydrogen loaded to increase reflectivity of the sensor. As hydrogen loading
increases the photosensitivity of the boron and germanium co-doped silica fibre [Lemaire et
al., 1993], hydrogen loading into single mode fibre was performed prior to grating writing. This
process is a relatively simple, cost effective and highly effective method of achieving high UV
photosensitivity. In this process, the required length of optical fibre was mixed with hydrogen
in a hydrogen cell using the standard recipe—a duration of twenty-four hours, pressure of 200
atm and temperature at -100°C. The unused hydrogen loaded fibre is kept in refrigerator at
low temperature (-20°C) to slow diffusion of hydrogen from the fibre core. This retains the
fibre photosensitivity for up to three days without the need of re-hydrogenation.

7.3 Layup and Embedding Optical Fibres

The interlaminar stresses (tensile, compressive or shear) are maximum around the mid-
thickness of the laminate (depending on the loading mechanism) and have a near zero value at
the outer surface during structural loading. Cracks initiate between the plies, where strain
anomalies within the thickness of the laminate are much greater than at the surface. Hence,
embedding an optical-fibre sensor in the composite laminate is preferable in comparison to
surface bonding for capturing strain variations due to interlaminar cracks. The FBG is well
protected inside the laminate rather being exposed on the outer surface, and embedding
procedure achieves good bonding between the host material and the sensor.

Each optical fibre consists of one FBG strain sensor. Two optical fibres were embedded at each
of the two crown bends, with sensor location as shown in Figure 7.6. Tables 7.1 to 7.3 show
the location of the FBG sensor at the crown bend of the Top-hat-stiffener in the laminate for
all three different layups.

Page | 227
 Figure 7.6: Location of four FBGs on a Top-hat-stiffener specimen

Layer No. Layup 4

8 ------------ DB450
7 ------------ DB450
------------------------ FBG A (Left)/ FBG B (Right) ─ Outer
6 ------------ CSM450 (White) (Yellow)
5 ------------ DB450
4 ------------ CSM450
3 ------------ DB450
------------------------ FBG C (Left)/ FBG D (Right) ─ Inner
2 ------------ CSM450 (Blue) (Green)
1 ------------ DB450
Mould Surface
Table 7.1: FBG location in Top-hat-stiffener Layup 4 specimen

Layer No. Layup 5

10 ------------ CSM225
9 ------------ UD450
8 ------------ CSM225
------------------------ FBG A (Left)/ FBG B (Right) ─ Outer
7 ------------ UD450 (White) (Yellow)
6 ------------ CSM225
5 ------------ DB450
4 ------------ CSM225
------------------------ FBG C (Left)/ FBG D (Right) ─ Inner
3 ------------ DB450 (Blue) (Green)
2 ------------ CSM225
1 ------------ CSM225
Mould Surface
Table 7.2: FBG location in Top-hat-stiffener Layup 5 specimen

Page | 228
 Layer No. Layup 6
10 ------------ UD450
9 ------------ BE450
8 ------------ CSM225
------------------------ FBG A (Left)/ FBG B (Right) ─ Outer
7 ------------ UD450 (White) (Yellow)
6 ------------ BE450
5 ------------ CSM225
4 ------------ DB450
------------------------ FBG C (Left)/ FBG D (Right) ─ Inner
3 ------------ CSM225 (Blue) (Green)
2 ------------ DB450
1 ------------ CSM225
Mould Surface
Table 7.3: FBG location in Top-hat-stiffener Layup 6 specimen

The four different wavelengths of FBGs for the four different locations are 1529, 1540, 1555
and 1565 nm. The optical fibres were colour coded, as shown in Figure 7.7.

Figure 7.7: Optical-fibre preparation for embedding into the laminate

Polyolefin heat-shrink tubes were found to be able to provide sufficient protection and
flexibility to optical fibre during VIP. When bent, this tube produces rounded bending rather
than V-shape bending, which ensures survival of optical fibre when small radius bending is
required. Each wavelength was designated with distinguished tube colours for easy tracking of
sensor location. The sensor head was left exposed, as this gives direct strain measurement,
while other section was protected by the tube. The diameter of primary tubing used was 3
mm, which further shrunk when blown with industrial heat gun. All tubes were applied to
optical fibres prior to embedding as this saves time compared to its application during
embedding process. The applied tubes were colour-coded for easy tracking of the FBGs. A
small amount of super glue was then applied at the tube edge to fix the optical fibre position,
Page | 229
and also to seal the tube against the entry of resin during vacuum infusion. A secondary
coating was applied at the ingress/egress points using heat shrink tubes with 3 mm diameter.
This would provide additional protection to optical fibre against sharp resin points during
residual removing.

The heat shrink tubes containing the optical fibre were penetrated into the fibre mat (in the
web section of the Top-hat-stiffener). An FEA analysis showed no reduction in strength of the
structure due to this penetration as the bends fail prior to the web during actual loading. A
detailed discussion of the procedure for embedding the optical-fibre sensors in the curved
bend of the Top-hat-stiffener is discussed in Appendix G. Figure 7.8 shows the optical fibres on
the mould prior to resin infusion.

Figure 7.8: Positioning of optical fibres on the THS mould prior to resin infusion

All specimens were manufactured at EMP Composites fabrication facilities (EMP Composites,
Warriewood, Sydney). Different E-glass reinforcement fabrics were measured to the required
size of manufacturing 5 specimens at a time. The first fabric layer was laid onto the mould
surface and fixed with of small quantity of 3M spray adhesive. The fabric layer was gently
rubbed in from top surface towards bottom flange-web bend to smoothen the fabric layer at
the bends. This procedure was applied to ensure that the produced shape and thickness will be
uniform. Unused space at all sides of the mould was created to make room for resin infusion
process. Figure 7.9 (a) shows the process of positioning of FBG sensors on the desired
composite layers. The two lines designate horizontal axis for sensors and fibre ingress/egress
point. For each side, a common ingress/egress point was used to retrieve inner and outer layer
Page | 230
FBGs. Figure 7.9 (b) shows that sensors with white and blue coatings (FBGs A and C) shared a
similar ingress/egress point. Optical fibres were temporarily held up vertically by applying
adhesive tape to the fibre. Top and bottom ends of in-positioned FBG sensors were applied
with small amount of super glue to ensure a constant position of FBG sensors during resin
infusion.

Figure 7.9: (a) Vertical positioning of bare FBG sensors on the marked area; (b) folding of upper layer E-
glass fibre in getting optical fibre through the fabric

The next fabric layer was placed on top of the fabric layer that was attached with the FBG
sensors. Initially only the flat top part was fixed with spray adhesive, while the left and right
fabrics at flange were folded upward as shown in Figure 7.9 (b). The fibre ends were stitched
through instantly made small holes in the top fabric layer. These holes would have quickly
shrunk as the tension of the weave brought the reinforcement fabric threads to its original
position. Experiment has shown that the structural strength will be not affected by this
penetration, as the web did not fail during the experiment. A small amount of spray adhesive
was then applied on other parts of the fabric. As all optical fibres were completely stitched
through, the upper fabric layer was then gradually unfolded. At the same time the optical
fibres were pulled to straighten the underneath alignment. The FBG sensors located at the
crown bend gradually obtained firmer hold from the fabric adhesion. At this time, the 3M
adhesive tapes used previously were removed. The alignment of underneath optical fibres was
checked to ensure no optical fibres were bent.

Page | 231
The following steps were related to standard VIP. Figure 7.10 (a) shows the arrangement of the
materials used for VIP. Peel ply, perforated film, and flow mesh were aligned and secured
around the mould perimeter. Additional polyolefin heat-shrunk tubes (black colour) with 5 mm
diameter were inserted for additional protection to the first coating. The cured resin may
cause optical fibre to become inflexible and fragile; therefore both ends of the black tube were
sealed using industrial wax. The ingress/egress pipes on LHS and RHS of the specimen suggest
the resin was infused from left to right, due to superiority of this procedure over top-to-
bottom infusion. The specimen was cured under a vacuum bag for 24 hours at 20°C
temperature. Figure 7.10 (b) shows a fully cured specimen. The process of residual removal
which included scissoring, cutting, and tearing, became complicated as the cured resin was
very stiff and strongly bonded to the structure. The weakest point of ingress/egress optical
fibre was at high risk of being cut by the sharp end of broken resin. At this stage, it was found
that stronger material for secondary reinforcement, e.g., Teflon or steel tubes, at the
ingress/egress point would be more advantageous in terms of protection. However, the usage
of these tubes was not preferred due to their inflexibility that could raise handling intricacy.

Figure 7.10: (a) Preparation for VIP; (b) Fully infused specimen – still under the bag

Coloured, protected optical fibres were again inserted into a 5 mm heat shrink tube and the
edges sealed to protect the optical fibre from resin during infusion. After curing, the Top-hat-
stiffener was sliced to the required dimensions using a bandsaw and the edges were sanded
and cleaned with acetone, as shown in Figure 7.11.

Page | 232
 Figure 7.11: VIP Top-hat-stiffeners with embedded FBGs, sliced and ready for testing

During the curing process, the optical-fibre FBG sensor may be subjected to high temperature
variations and residual stresses may be induced in the laminate. This can vary the wavelength
of the fibre Bragg grating significantly. Some optical fibres may be misaligned or broken during
the installation, resin infusion process and specimen trimming process. Post-cure wavelength
measurement was taken to examine survival of the embedded FBGs and also to record the
final Bragg wavelength which was required for the actual failure monitoring process. Tables 7.4
to 7.6 show the magnitude of the signal before and after embedding. FBGs marked ‘w’ had a
weak signal, which were unable to provide adequate results. Those marked ‘X’ were either
broken or misplaced from the original location, and were not tested during the experimental
study.

Left Outer Left Inner Right Outer Right Inner

FBG λ Before A1 1528.8510 C1 1556.6862 B1 1540.5393 D1 1565.1653
FBG λ After 1529.46 (w) 1557.2025 1541.1825 1565.4175 (w)
FBG λ Before A2 1528.7612 C2 1556.5934 B2 1540.4914 D2 1565.2115
FBG λ After 1529.0800 1556.7538 1541.0400 1565.4900
FBG λ Before A3 1528.8580 C3 1556.6044 B3 1540.5408 D3 1565.0314
FBG λ After 1529.6500 1557.48 (w) 1540.8800 X
FBG λ Before A4 1528.8746 C4 1556.7714 B4 1540.5485 D4 1565.1033
FBG λ After 1529.0400 1556.9970 1540.724 (w) 1565.5800
FBG λ Before A5 1528.8553 C5 1556.4556 B5 1540.5310 D5 1565.2342
FBG λ After 1529.5375 1556.8975 1541.1625 1566.2346

Table 7.4: FBG wavelength survival details for VIP THS Layup 4 specimens

Page | 233
 Left Outer Left Inner Right Outer Right Inner
FBG λ Before A1 1528.7818 C1 1556.8218 B1 1540.2971 D1 1564.9772
FBG λ After X X 1540.9800 X
FBG λ Before A2 1528.8474 C2 1556.7265 B2 1540.3359 D2 1564.6499
FBG λ After 1529.5672 X X X
FBG λ Before A3 1528.8322 C3 1556.8223 B3 1540.3054 D3 1565.1606
FBG λ After X X X X
FBG λ Before A4 1528.8223 C4 1556.6548 B4 1540.2222 D4 1565.1221
FBG λ After X X X X
FBG λ Before A5 1528.7583 C5 1556.5211 B5 1540.2006 D5 1565.1452
FBG λ After X 1557.3453 1540.6996 X
Table 7.5: FBG wavelength survival details for VIP THS Layup 5 specimens

Left Outer Left Inner Right Outer Right Inner

FBG λ Before A1 1528.8394 C1 1556.7849 B1 1540.2433 D1 1565.1597
FBG λ After X X X X
FBG λ Before A2 1528.8473 C2 1556.8592 B2 1540.5884 D2 1565.1291
FBG λ After 1529.6375 X 1541.0650 1565.3984
FBG λ Before A3 1528.7745 C3 1556.6637 B3 1540.1857 D3 1565.2083
FBG λ After 1529.6217 1557.8308 X X
FBG λ Before A4 1528.8197 C4 1556.5003 B4 1540.5235 D4 1565.1896
FBG λ After 1529.6000 1557.6800 X X
FBG λ Before A5 1528.8385 C5 1556.6013 B5 1540.5528 D5 1565.0603
FBG λ After 1529.4375 X 1540.9325 X

w – FBG with weak reflection which can be measured using tunable laser system
X – FBG either broken or drifted away from the original location
Table 7.6: FBG wavelength survival details for VIP THS Layup 6 specimens

Page | 234
7.4 FBG Strain Measurement Setup
The setup for measuring strain is similar that described in Section 6.6.1. The setup for the
strain measurement uses the typical parts shown in Figure 7.12. An optical spectrum analyser
(OSA) is required to interrogate the FBG sensor during the experimental test procedure.

Figure 7.12 System setup for strain measurement. All bare FBG sensors, except the reference, were
embedded in laminate.

With all four FBG strain sensors deployed in a parallel configuration, three 50:50 couplers were
required to combine the reflected light from the four sensors towards the Fabry-Perot
tuneable filter (FPTF). The reading from the reference FBG ensured compensation for
temperature-induced wavelength variation. Since the light from the FBG sensors experienced
high loss after propagating through two coupler stages, two C-band Amplified Spontaneous
Emission (ASE) sources were utilized to ensure that the FBG signal at the photodetector was
distinguishable from the noise floor. The computer is equipped with a National Instrument
PC1-6251 data acquisition card to digitally sample the analogue voltage at the photodetector
and continuously provide the FPTF with a configured ramp signal for linear-wavelength
scanning. The sampled data is processed and stored in the computer by the Labview 8.0
program. The spectral position detection is optimised by the Centroid Detection Algorithm
(CDA) [Binfeng et al., 2006].

Most of the FBG gratings designed for practical applications are not uniform—the needs of
different applications vary. The need for high performance and tailored spectral and dispersive
characteristics of FBGs imposes restrictions on the fabrication of special structures. These
structures can be achieved by varying numerous physical parameters, such as an induced index
change or periodicity along the grating, length, and fringe tilt. In this way, specialised gratings

Page | 235
(apodised, chirped, phase shifted and sampled) can be fabricated rather than the standard
uniform Bragg gratings [de Melo, 2006].

The FBG sensors were fabricated of hydrogenated GF1 fibre with a length of 4 mm. A short
cavity length was preferred to minimize effect of strain variation distributed vertically that can
produce a chirped grating. Chirped gratings can be produced accurately, simply by increasing
the amount of fibre translation each time the fibre is irradiated. Cosine apodisation was
applied to suppress noise during measurement due to side lobes. With the cosine apodisation
profile, the effective grating length is actually half of the total length. This requirement
increases the writing time and the FBG bandwidth. The experimental setup for FBG strain
measurement is shown in Figure 7.13.

Figure 7.13: FBG interrogation unit used in the test procedure

Figures 7.14 and 7.15 show the reflectivity spectra for each FBG sensor and multiplexed FBG
sensor, respectively.

Page | 236
 FBG 1 FBG 2
-10 -10

Reflectivity (dB) -20 -20

Reflectivity (dB)
-30 -30

-40 -40

-50 -50

-60 -60
1525 1527 1529 1531 1533 1535 1535 1537 1539 1541 1543 1545
Wavelength (nm) Wavelength (nm)

FBG 3 FBG 4
-10 0

-20 -10
Reflectivity (dB)

Reflectivity (dB)
-30 -20

-30
-40
-40
-50
-50
-60 -60
1554 1556 1558 1560 1562 1560 1562 1564 1566 1568 1570
Wavelength (nm) Wavelength (nm)

Figure 7.14: Measured reflectivity of FBG sensors from each wavelength for a typical specimen

From these measurements, the average FBG parameters are: 3 dB bandwidth ~ 0.4 nm; side-
mode suppression ratio >30 dB; peak reflectivity ~ 99 per cent; and wavelength spacing
between sensors >10 nm. However, the inclusion of the reference FBG reduced the dynamic
range of FBG2 and FBG3. Evidently, the positioning of the FBG was crucial for avoiding
wavelength crosstalk that can cause spectral detection errors. To avoid such problem, the two
FBGs with highest wavelengths (i.e., 1556 and 1565 nm) were embedded in an inner layer of
the Top-hat-stiffener. The tension in the inner layer increases the wavelength. The two FBGs
with the lowest wavelengths were embedded in the outer layer, which is subjected to
compressive strains during loading. This decreases the wavelength, avoiding collision with the
higher wavelengths.

Page | 237
 FBG 1 FBG Ref FBG 2 FBG 3 FBG 4

Figure 7.15: Measured reflection spectrum of all sensors multiplexed together

7.4.1 Test Procedure

Five Top-hat-stiffener specimens for each of three different layups were manufactured by
vacuum-infusion methods. Four sensors were embedded in each specimen. The experimental
procedure was discussed in-depth in Chapter 4. Analysis of FBG strain sensor measurement for
failure initiation and progression is discussed in the next section.

7.5 VIP THS Layup 4 FBG Results

When the Top-hat-stiffener is subjected to the designed loading condition (keel bolt pulled
up), the inner and outer layers of the laminate at the bend are subjected to in-plane
compression and tension, respectively. Thus, the inner FBGs (groups C and D) are in
compression and the outer FBGs (groups A and B) are in tension.

7.5.1 FBG Survival after Specimen Manufacture

After specimen manufacturing, some FBGs were lost due to weak reflection, optical fibre
breakage or misalignment. Summaries of the test specimens are presented in Tables 7.7, 7.8
and 7.9.

Page | 238
 Specimen No AE-piezo Strain-FBG
A1 Tested - OK 3 FBG Tested – Good
A2 Tested - OK 3 FBG Tested – Not Good / Lost spectral ref
A3 Tested - OK 2 FBG Tested – OK / Spectral split
A4 Tested - OK 2 FBG Tested – Not Good /
A5 Tested - OK 4 FBG Tested – Good
Table 7.7: VIP THS Layup 4 FBG survival summary

Specimen No AE-piezo Strain-FBG

B1 Tested - OK 1 FBG Tested – Good
B2 Tested - OK 1 FBG Tested – Good
B3 No FBG survival
B4 No FBG survival
B5 Tested - OK 1 FBG Tested – Good
Table 7.8: VIP THS Layup 5 FBG survival summary

Specimen No AE-piezo Strain-FBG

C1 No FBG survival
C2 Tested - OK 3 FBG Tested –Good
C3 No FBG survival
C4 Tested - OK 2 FBG Tested – Partially Good
Tested -
C5 Tested – Not OK / Lost spectral reference
Problem
Table 7.9: VIP THS Layup 6 FBG survival summary

The order of specimen manufacturing was Layup 5, 6 and then Layup 4. Most of the
manufacturing and handling mishaps were minimised by the last layup. As a result, as shown in
the tables, FBG survival rate was higher for Layup 4 compared with the other two. To avoid
repetition, the discussion of Layup 4 specimens is presented and the failure plots and
tabulation of individual failures for Layups 5 and 6 are presented in Appendix H.

7.5.2 VIP THS Layup 4 FBG-strain Results

Specimen ‘THS VIP D05’ from Layup 4 had all four FBG sensors in good working condition and
the discussion of results is focussed on this specimen. The 4 mm long FBG sensor yielded
satisfactory strain measurements at the top two bends of the Top-hat-stiffeners. This was the
expected failure location in the actual yacht due to excessive loading on the keel bolt.

Page | 239
Two different methods can be used for failure recognition based on strain measurements.
Firstly, the absolute changes in strain can be compared with those of the reference location.
The overall strain distribution shows the reduction or increase in strain values at the failure
points corresponding to the load drops. However, failure discrimination and determining the
magnitude of the failure are not possible by analysing the strain distribution alone. The second
method is based on the detection of abnormal strain gradients. The second plot in each figure
is the strain rate (με/sec) and load v. time, showing the strain gradients with respect to time
where the damage-induced strain variation can be clearly seen. This plot shows sensor output
where the sensor is in tension or compression, depending on the failure.

The cumulative strain (με) and load v. time is presented in Figures 7.16, 7.17, 7.18 and 7.19 for
the left outer, right outer, left inner and right inner FBG sensors, respectively.

Figure 7.16: Top-hat-stiffener Layup 4 FBG strain-sensor measurement—left outer

Page | 240
Figure 7.17: VIP THS Layup 4 FBG strain-sensor measurement—right outer

Figure 7.18: VIP THS Layup 4 FBG strain-sensor measurement—left inner

Page | 241
 Figure 7.19: VIP THS Layup 4 FBG strain-sensor measurements—right inner

A crude approach of demonstrating failure detection is made by observing the abnormal strain
gradients for the four FBG sensors. No statistical analysis of the data is undertaken in this
study. These abnormal strain gradients (strain rate) of the FBG sensors are presented in Table
7.10. The four FBG strain sensors are referred as per their location in the Top-hat-stiffener: LO
(left outer), RO (right outer), LI (left inner) and RI (right inner).

The damage characterisation of Top-hat-stiffeners is performed by monitoring the abnormal

strain rate gradients. These strain gradients have specific values (and can be positive or
negative) at any given point of time, irrespective of the initial or reference values. The positive
and negative values show whether the laminate is displaced up or down at that particular
instant in time.

First failure occurred as matrix cracking between the second and third layers on the left crown
bend at 224.66 sec, as discussed in Chapters 4 and 5 in the experimental and FEA analysis. At
this point, LO FBG sensor reported a 373.1 με/s jump in the strain rate. This failure was
assumed matrix cracking, as a load drop of only 7 N was observed. The inner FBG sensors are
placed between the second and third layers. Matrix cracking appeared to have initiated
between these layers. The LI sensor might have been glued to the second layer and, as a result,

Page | 242
a strain gradient of only -5.8 με/s is observed. At this point, compared to the left crown bend,
the right crown bend had a limited strain gradient (-81.7 and 0.6 με/s for RI and RO
respectively). A major delamination occurred at the same location at 299.83 sec, with strain
gradients of -167.9 and -108.7 με/s at LO and LI, respectively. For the minor crack propagation
at 320.23 sec, LO and RI recorded values of -32.7 and -55.51 με/s, respectively. The next major
delamination occurred at RI with a value of -228.9 με/s. A large delamination triggered the
strain rates of -121.9, 353.6, -297.3 and -306.1 με/s for LO, RO, LI and RI, respectively. The next
three failure points were crack propagations, where the failure at 592.52 sec triggered a large
deformation in the structure with the FBG strain rate reading of 789.6 με/s at LI. In comparison
with the experimental results and photographic evidence of the structural damage, this
particular failure was caused by the disappearance of the crown bend (straightening) at the
end of the steel plates. Final collapse occurred at 743 sec with readings of -343.0, -2291.9, -
2400.6 and -1181.5 με/s.

Page | 243
 Load LO RO LI RI
Failure Time Load Failure
drop FBG A FBG B FBG C FBG D FBG Response/ Interpretation
No. (Sec) (N) Mode
(N) (με/s) (με/s) (με/s) (με/s)
Matrix Wavelength gradually moved against initial direction, indicating the
1 224.66 7571 7 373.1 0.6 -5.8 -81.7
Cracking beginning of eccentric interlaminar stress response.
Transverse compressive stress caused spectral splitting of FBG A & FBG
C.
2 248.23 230 6.4 -48 3
Large strain rate in FBG A due to peak splitting and not from major
damage.
3 299.83 10818 484 -167.9 -18.2 -108.7 -78.55 Delamination Insignificant change of average strain, spectral and small load drop.
Crack Compressive stress anomalies caused significant spectral splitting of
4 320.23 11404 223 -32.7 4.8 -11.2 -55.51
progression FBG A and FBG C.
Significant change of average strain in FBG C almost to relaxation
5 345.35 11254 437 60.7 -67.4 -498.7 -228.9 Delamination
position, potentially due to delamination of surrounding layer.
Crack Failure occurred but cause insignificant change of average strain and
6 383.36 12337 72 6.3 -9.6 5.2 -32.2
progression spectral, suggesting crack progression.
Significant change of average strain in FBG B and D potentially caused
7 422.50 14200 1575 -121.9 353.6 -297.3 -306.1 Delamination
by delamination of surrounding layer.
Crack Failure occurred but caused insignificant change of average strain and
8 486.50 16150 93 7.2 -57.9 -44.1 -83.0
progression spectral, suggesting crack progression
Crack Failure occurred but caused insignificant change of average strain and
9 530.77 17456 631 -67 -122.1 -279.2 -165.4
progression spectral, suggesting crack progression
Crack Failure occurred but caused insignificant change of average strain and
10 592.52 19690 257 -51.3 -53.31 789.6 -55.3
progression spectral, suggesting crack progression
Significant change of average strain and no anomalies observed in FBG
11 655.54 24699 1122 -99.1 -124.8 -310.5 -103.2 Delamination C, while increased of reflected intensity in FBG A. Possible delamination
of surrounding layer
Ultimate failure occurs with large residual stress remained indicating
12 743.00 30345 12457 -343.0 -2291.9 -2400.6 -1181.5 Fibre failure
the layers were still intact; hence the failure was fibre breakage
Table 7.10: Layup 4 FBG failure characterisation

Page | 244
Changes in the level of intensity and bandwidth of the spectrum and a shift of Bragg
wavelength are observed during the study. The shift in Bragg wavelength may be due to the
alteration of the average strain distribution along the width of the grating. The abrupt changes
in the strain gradient from delamination and crack propagation caused the deformation of the
reflection spectra at various stages of testing. Snapshots of the FBG spectrum footprints at
221, 348 and 625 sec are shown in Figure 7.20.

Figure 7.20: VIP Top-hat-stiffener Layup 4 Bragg wavelength shift during different phases of the test
procedure

Page | 245
7.6 FBG-strain sensing Conclusions
A methodology for damage monitoring of composite structures based on strain
measurements, using embedded fibre Bragg grating sensors, is presented. FBG sensors were
embedded at various locations in the Top-hat-stiffener. The study shows that the proposed
technique is able to identify the occurrence and progression of damage, crack size, damage
location and intensity of damage in a complex composite structure. It has been successfully
demonstrated that the occurrence of interlaminar cracks within the laminate significantly
alters the local strain distribution, and that these strain anomalies can be detected and
measured using embedded FBGs. The sensors successfully detected failures from the change in
the local strain distribution. Failure detection is possible by monitoring either irregular strain
variations or abnormal large strain gradients—the latter method was adopted for the current
study. Understanding the failure behaviour of Top-hat-stiffeners is complex as they are
subjected to out-of-plane load transfers. The four bends of the Top-hat-stiffener add more
complexity to the load transfer mechanism and the involved damage processes. Embedded
FBGs successfully detected the occurrence and progression of delamination, demonstrating
that FBGs can be successfully employed for damage monitoring in curved composite
structures.

The major limitation of the proposed technique is the localised nature of damage-induced
strain variations. This makes damage detection in the structure only possible on a local scale,
or near the locations of the installed sensors, making knowledge of failure locations required a
priori. Damage location prediction could be obtained from experience, prior knowledge of the
structure, history of in-service structures, experimental tests and by conducting numerical
simulations (such as FEA). The potential of failure detection in real structures would be to
implement the technique where visual inspection and other evaluation techniques may not be
practical due to visibility, space, weight or cost constraints.

Page | 246
Chapter 8: Conclusion and Future
Work

http://files.geglobalresearch.com/wp-content/img/technologies/mst_composites_full.jpg

Page | 247
8.1 Overview
In curved composites such as Top-hat-stiffeners, interlaminar delamination is a common and
potentially dangerous mode of failure. Delaminations often go undetected at the subsurface
level, leading to a sudden loss of stiffness and strength which may lead to safety, reliability and
integrity issues. Therefore it is important to understand the failure mechanisms involved
during the damage process that will enable the designers to develop more effective damage
tolerant and damage resistant structures.

Failure analysis of various marine accidents in the past reveals the scarcity of available
knowledge in this area. The investigation was extended above and beyond what was currently
available in the published scientific literature. Experimental investigation on the L-bend
structures and Top-hat-stiffeners has been conducted. Damage initiation and progression was
monitored by acoustic-emission technique and embedded optical fibre strain gauges.
Numerical modelling of the Top-hat-stiffeners was also studied in this thesis.

8.2 Summary of Key findings

The objectives defined in Chapter 1, have been accomplished and the key achievements are
summarised in the following sections.

x Literature Review – A comprehensive literature review has been summarised in Chapter

2 where various aspects of keel design, keel related accidents, available failure
theories for different failure modes for composites, progressive failure modelling,
acoustic-emission and optical fibre sensing techniques for determining the damage
mechanisms within the composite laminate have been presented.

x Material properties – Mechanical properties of composite structures depend on the

basic materials, manufacturing process and post-processing of the laminates. An
attempt is made to assess the variation in material properties for CSM, DB and UD
with:
o Two manufacturing process: Hand-layup and Vacuum bag infusion Process
o Two types of hardeners: 925H and CHM 50
o Three post curing temperatures : 20°C for 24 hrs, 50°C for 24 hrs and 82°C for
2 hrs
o Two tests: Tensile test and three-point bend test.
Page | 248
x Preliminary test outcomes – L bend testing – Preliminary tests on the load carrying
capacity and the failure behaviour of L-bend curved composite laminates were
performed. It is observed that the delamination is the critical failure mode of failure in
composites. Also the fibre fracture tends to occur at a later stage of loading. It was
concluded that irrespective of the layup, delamination always initiated between the
second and third layers.

x Top-hat-stiffener testing – The Top-hat-stiffeners for keel structures are subjected to

interlaminar failure due to various in-service loading conditions. In an attempt to
optimise the laminate layup and manufacturing procedure, six different layups and
two manufacturing procedures were considered. Unidirectional fibres must to be
included in the layup but need to be placed far away from the mould in order to
reduce premature failures. It was seen that the vacuum bag infusion technique is
beneficial and increases the strength and stiffness of the curved composite structure.
However the specimen layup need to be optimised to achieve appropriate laminate
thickness.

x Validated numerical model for delamination failure – Implicit finite element modelling
of the delamination failure of Top-hat-stiffeners has been undertaken. Progressive
failure analysis with interface elements has been performed and the numerical models
have been validated with experimental testing for L-bend and Top-hat-stiffener
specimens.

x Analysis for the stress distribution – Interlaminar stresses are responsible for the
delamination failures of curved composite structures. Validated numerical models
provided an accurate interlaminar stress distribution around the bend and also
predicted the initial failure location.

x Acoustic-emission – The future of failure and health monitoring of composite structures

lies with the optical fibres. Conventional Piezo-sensors and modern FBG-AE sensors
were used to assess and analyse the different failure modes in composite structures.
The results of FBG-AE sensors were compared with the Piezo-electric sensors and the
failure prediction was satisfactory.

x Embedded optical fibre strain sensor – Fibre Bragg gratings were written in an optical
fibre and embedded in the composite structure to monitor the failures during the
Page | 249
 service. Two optical fibres were embedded for L-bend specimens and four optical
fibres were embedded for Top-hat-stiffeners at the curved section of the structure.
The continuous strain monitoring successfully captured the interlaminar failure
through strain gradients during crack initiation and propagation, proving an excellent
alternative to conventional foil strain gauges.

x Optical fibre Embedding procedure for curved structures: The procedure to embed the
optical fibre and to be held in position during the process of laminate manufacture is
both difficult and risky as the optical fibres are very brittle (1% strain limit). An optical
fibre embedding procedure was proposed and implemented successfully for the strain
monitoring of L-bend composites and Top-hat-stiffeners.

x Reserve strength – For curved composites, delamination is the primary failure. Reserve
strength was recognised and it was concluded that the secondary stiffening will be
more identifiable when the failure is delamination – neither through-thickness shear
nor fibre failure. This was observed only in those specimens where unidirectional
fibres are included in the layup.

8.3 Future work

Similar to any research program, this thesis has answered many questions but also raised
many more. Possible avenues of further research on the basis of this thesis are presented
below.

x Compressive and bending load on Top-hat-stiffeners – This thesis dealt with the
scenario where there is excessive tensile load on the keel bolt (may be due to self-
weight). In a real life situation the vessels are subjected various loading such as
collision, grounding and broaching. To fully understand the effect and interaction of
various loadings on the yacht, it might be necessary to carry out more experimental
tests.

x Fatigue analysis of Top-hat-stiffeners – Strength, stiffness and failure behaviour of Top-

hat-stiffeners need to be systematically assessed through fatigue testing to ensure
adequate long-term performance. Research into the effects of fatigue loading is a rich
area of possible research.

Page | 250
x Failure analysis of top hat stiffened panel – In this thesis, study was conducted on the
strength and failure analysis of Top-hat-stiffener slice (100mm wide). In real life, the
yacht floor board takes up the keel load. Hence it would be beneficial if the
experimental/numerical studies are conducted for that section of the floor board
where the whole keel is attached.

x Rate of loading – In this study a displacement loading rate of 2mm/min has been
attempted. It might be interesting to investigate the interdependence between
loading rate and the failure behaviour of the Top-hat-stiffener for lower or higher
loading rates in a situation such as, where couple of keel-bolts have failed and as a
result, remaining keel-bolts are subjected to excessive loadings.

x Effect of post curing – different temperatures and duration – During the material
characterisation investigation in this study it was proven that the strength and stiffness
would be increased to a huge extent (up 85% for DB in flexure test). More
experimental/numerical study on the post curing effect for the strength/stiffness
enhancement in the curved composite structures might be beneficial.

x Durability of FBG sensors – The durability of the FBG sensors during long term operation
needs to be assessed. Various studies are conducted for low and high cyclic stress in
the past but the information fatigue behaviour of FBG sensors in curved composite
structures is lacking. This might be an interesting avenue of further research.

x FEA model with involving other failure modes – Primary mode of failure for curved
composites is delamination which has been successfully addressed in this thesis. A
further investigation might be interesting to model delamination along with other
failure modes such as fibre breakage.

x Use of hybrid composites – Using two or more different kinds of fibres in a single
matrix. This can be achieved either by mixing the fibres in a single layer of having
layers, each of consisting of a single fibre type, alternating one with another. This also
helps in preventing hygro-thermal effects on the structure as well.

Page | 251
References

Page | 252
Ageorges C., Friedrich K., Schuller T., Lauke B. (1999). Single-fibre Broutman test: fibre - matrix
interface transverse debonding. Composites: Part A 30: 1423-1434.

AS 4132.3 – Boat and ship design and construction, Part 3: Fibre-reinforced plastics
construction (1993). Standards Australia, Australia.

Arcan M., Hashin Z., Voloshin A. (1978). A method to produce plane-stress states with
applications to fiber-reinforced materials. Experimental Mechanics 18: 141–146.

ASTM D5528-01 (2007). Standard test method for Mode I interlaminar fracture toughness of
unidirectional fibre-reinforced polymer matrix composites, ASTM International, West
Conshohocken, PA, USA.

ASTM D3039/D3039M (2000). Standard test method for tensile properties of polymer matrix
composite materials. ASTM International, West Conshohocken, PA, USA.

ASTM D7264/D7264M (2007). Standard test method for flexural properties of polymer matrix
composite materials. ASTM International, West Conshohocken, PA, USA.

ASTM Standard D6671/D6671M (2006). Standard test method for mixed Mode I-Mode II
interlaminar fracture toughness of unidirectional fiber reinforced polymer matrix composites.
ASTM International, West Conshohocken, PA, USA.

ASTM E1316 - Standard terminology for non-destructive examinations. ASTM International,

West Conshohocken, PA, USA.

Azmi A. I., Sen D., Peng G. D. (2010). Sensitivity enhancement in composite cavity fiber laser
hydrophone. IEEE Journal of Lightwave Technology 28(12): 1844-1850.

Azzi V.D., Tsai S.W. (1965). Anisotropic strength of composites. Experimental Mechanics 5(9):
283-288.

Badcock R.A., Fernando G.F. (1995). An intensity-based optical fiber sensor for fatigue damage
detection in advanced fiber-reinforced composites. Smart Materials and Structures 4: 223–230.
Page | 253
Bansal A., Kumosa M. (1995). Application of biaxial Iosipescu method to mixed-mode fracture
of unidirectional composites. International Journal of Fracture 71: 131–150.

Barre S. and Benzeggagh M.L. (1994). On the use of acoustic emission to investigate damage
mechanisms in glass-fibre-reinforced polypropylene. Composites Science and Technology 52:
369-376.

Barnes C.A., Ramirez G. (1998). Acoustic Emission Testing of Carbon Fiber Composite Offshore
Drilling Risers. The Sixth International Symposium on Acoustic Emission from Composite
Materials: 13-22. San Antonio: American Society for Nondestructive Testing, Inc.

Benmedakhene S., Kenane M., Benzeggagh L., (1999). Initiation and growth of delamination in
glass/epoxy composites subjected to static and dynamic loading by AE monitoring. Composites
Science and Technology 59: 201-208.

Berbinau, P., Soutis C., Guz I.A. (1999). Compressive failure of 0° unidirectional carbon-fibre-
reinforced plastic (CFRP) laminates by fibre microbuckling. Composites Science and Technology
59(9): 1451-1455.

Berthelot, J.M., Rhazi J. (1990). Acoustic Emission in Carbon Fibre Composites. Composites
Science and Technology 37: 411-428.

Betz D. C., Thursby G., Culshaw B., Staszewski W.J. (2006a). Identification of structural damage
using multifunctional Bragg grating sensors: I. Theory and implementation. Smart Materials
and Structures 15(5): 1305-1312.

Betz D. C., Staszewski W.J., Thursby G., Culshaw B. (2006b). Structural damage identification
using multifunctional Bragg grating sensors: II. Damage detection results and analysis. Smart
Materials and Structures 15(5): 1313-1322.

Bhat M.R., Majeet M.A., Murthy C.R.L. (1994). Characterisation of fatigue damage in
unidirectional GFRP composites through acoustic emission signal analysis. NDT & E
International 21(1): 27-32.

Page | 254
Bohse J., Kroh G. (1992). Micromechanics and acoustic emission analysis of the failure process
of thermoplastic composites. Journal of Material Science 27: 298-306.

Bohse J. (2000). Acoustic emission characteristics of micro-failure process in polymer blends

and composites. Composites Science and Technology 60: 1213-1226.

Borg, U. (2004). Compressive strength of fiber composite with porosity. Proceedings of 21st
International Congress of Theoretical and Applied Mechanics, Warsaw, Poland.

Broek D. (1986) Elementary engineering fracture mechanics. 4th edition, Kluwer Academic
Publishers, The Netherlands, ISBN: 90-247-2580-1.

Brown D. A., Cameron C. B., Keolian R. M., Gardner D. L., Garrett S. L. (1991). A symmetric 3×3
coupler based demodulator for fiber optic interferometric sensors. Proceedings of SPIE 1584:
328-335.

Brussat, T.R., Chiu, S. T. Mostovoy, S. (1977). Fracture Mechanics for Structural Adhesive Bonds,
AFNLTR-77-163, Air Force Materials Laboratory, Wright-Patterson AFB, Ohio, USA.

Bucaro J. A., Dardy H. D., Carome E. F. (1977). Fiber-optic hydrophone. Journal of the Acoustical
Society of America 62: 1136-1138.

Burlayenko V.N., Sadowski T. (2008). FE modelling of delamination growth in interlaminar

fracture specimens. Budownictwo i Architektura 2: 95-109.

Bush J., Cekorich A., Kirkendall C. K. (1997). Multichannel interferometric demodulator.

Proceedings of SPIE 3180: 19-29.

Bush J., Davis C., Davis P., Cekorich A., McNair F. (1999). Buried fiber intrusion detection sensor
with minimal false alarm rates. Proceedings of SPIE 3860: 285–295, Boston, USA.

Carbon Fiber Guru (2010). Carbon fiber processing Part 1 – Wet/Hand Lay-up from
http://www.carbonfiberguru.com/carbon-fiber-processing-part-1-wet-hand-lay-up

Page | 255
CARP Aerospace/Advanced composites subcommittee (1993). Guidance for development of AE
applications on composites. Journal of Acoustic Emission 11(3): C1-C24.

Chalmers D.W. (1994). The potential for the use of composite materials in marine structures.
Marine Structures 7:(2-5) 441-456.

Chang C.C., Sirkis J.S. (1998). Design of fiber optic sensor systems for low velocity impact
detection. Smart Materials and Structures 7: 166–177.

Chang F.K., Springer G.S. (1986). The strengths of fibre reinforced composite bends. Journal of
Composite Materials 20: 30-45.

Chang F.K., Chang K.Y. (1987). A Progressive damage model for laminated composites
containing stress concentrations. Journal of Composite Materials 21(9): 834-855.

Chen R., Theobald P., Gower M., Malik S., Burns J., Fernandes E., Bryce G., Fernando G. F.
(2008). A novel fibre optic acoustic emission sensor. Proceedings of SPIE 6932, 693237
DOI:10.1117/12.776170.

Childs P., Wong A.C.L., Yan B., Li M. Peng G.D. (2010). A review of spectrally coded multiplexing
techniques for fibre grating sensor systems. Measurement Science and Technology 21(9): 1-7.

Choi C., Seo D.C., Lee J.J., Kim J.K. (1997). Effect of embedded optical fibers on the interlaminar
fracture toughness of composite laminates, Composite Interfaces 5: 225–240.

Choi H.Y., Chang F.K. (1992). A model for predicting damage in graphite/epoxy laminated
composites resulting from low-velocity point impact. Journal of Composite Materials 26(14):
2134-2169.

Christensen, R.M. (1997). Stress based yield/failure criteria for fiber composites. International
Journal of Solids and Structures 34(5): 529-543.

Page | 256
Cole J. H., Sunderaman C., Tveten A. B., Kirkendall C., Dandridge A. (2002). Preliminary
investigation of air-included polymer coatings for enhanced sensitivity of fiber - optic acoustic
sensors. 15th Optical Fiber Sensors Technical Digest 1: 317-320.

Corletto, C.R., Bradley, W.L. (1989). Mode II delamination fracture toughness of unidirectional
graphite/epoxy composites. Composite Materials: Fatigue and Fracture, 2nd volume. P. A.
Lagace, ASTM STP 1012, American Society for Testing and Materials, Philadelphia: 201-221.

Cranch G. A, Nash P. J., Kirkendall C. K. (2003). Large-scale remotely interrogated arrays of

fiber-optic interferometric sensors for underwater acoustic applications. IEEE Sensors Journal
3(1): 19-30.

Cui W.C., Wisnom M.R. (1993). A combined stress-based and fracture-mechanics-based model
for predicting delamination in composites. Composites 24(6): 467-474.

Czochralski J. (1918). Ein neues Verfahren zur Messung des Kristallisationsgeschwindigkeit der
Metalle (A new method for the measurement of crystallization rate of metals), Zeit- schrift fur
Physikalische Chemie 92: 219-221. (The paper was received in the editorial office on 19 August
1916).

Dandridge A., Tveten A., Giallorenzi T. (1982). Homodyne demodulation scheme for fiber optic
sensors using phase generated carrier. IEEE Journal of Quantum Electronics 18(10): 1647-1653.

Daniel I.M., Ishai O., (1994). Engineering Mechanics of Composite Materials. Oxford University
Press, ISBN: 0-19-507506-4.

Davidson B.D., Kruger R., Konig M. (1995), Three-dimensional analysis of centre-delaminated

unidirectional and multidirectional single-leg bending specimens. Composites Science and
Technology 54: 385–394.

Davis C., Baker W., Moss S.D., Galea S.C., Jones R. (2002). In situ health monitoring of bonded
composite repairs using a novel fibre Bragg grating sensing arrangement. Smart Materials II,
Proceedings of a SPIE 4934: 140-149.

Page | 257
Davies M.A., Bellemore D.G., Kersey A.D. (1994). Structural strain mapping using a
wavelength/time division addressed fiber Bragg grating array. SPIE 2361: 342–345.

Davies P., Petton D. (1999). An experimental study of scale effects in Marine composites.
Composites Part A 30: 267-275.

De Melo M.A.R. (2006). Implementation and development of a system for the fabrication of
Bragg gratings with special characteristics. ME thesis. University of Porto.

de Moura M.F.S.F., Goncalves J.P.M., Marques A.T., de Castro P.M.S.T. (2000). Prediction of
compressive strength of carbon-epoxy laminates containing delamination by using a mixed-
mode damage model. Composite Structures 50: 151-157.

Dirand X., Hilaire B., Soulier J.P. Nardin M. (1996). Interficial shear strength in glass-
fiber/vinylester-resin composites. Composites Science and Technology 56: 533-539.

Dodkins A.R., Shenoi R.A., Hawkins G.L. (1994). Design of joints and attachments in FRP ship
structures. Marine Structures 7: 364-398.

Dzenis Y.A., Qian J. (2001). Analysis of microdamage evolution histories in composites.

International Journal of Solids and Structures 38: 1831-1854.

Eckold G. (1994). Design and Manufacture of composite structures. Woodhead Publishing,

England, ISBN 978-1-85573-051-9.

Edge E.C. (1996). Stress-based Grant-Sanders method for predicting failure of composite
laminates. Composites Science and Technology 58(7): 1033-1041.

Eksik O., Shenoi R.A., Blake J.I.R., Jeong H.K. (2004). Damage mechanisms of hat-shaped GRP
beams. Journal of Reinforced Plastics and Composites 23(14): 1497-1514.

Eksik O., Shenoi R.A., Moy S.J., Jeong H.K. (2007). Finite element analysis of top-hat-stiffened
panels of Fiber-Reinforced-Plastic Boat Structures. Marine Technology 44(1): 16-26.

Page | 258
Elvin N., Leung C. (1997). Feasibility study of delamination detection with embedded optical
fibers. Journal of Intelligent Material Systems and Structures 8: 824–828.

Erdogan T. (1997). Fiber grating spectra. Journal of Lightwave Technology 15: 1277–94.

Everall L., Gallon A., Roberts D. (2000). Optical fiber strain sensing for practical structural load
monitoring. Sensor Review 20: 113–119.

FGI Norox CHM 50 data sheet (2002). FGI from

http://www.fgi.com.au/files/images/stories/pdfs/products/catalysts/Norox_CHM_50_DS.pdf

Final investigation report on the S/V Cynthia Woods (2009). The office of General Counsel and
the Internal Audit Department of The Texas A&M University System, TX, USA.

Fisher N. E., Webb D. J., Pannell C. N., Jackson D. A., Gavrilov L. R., Hand J. W., Zhang L.,
Bennion I. (1998). Ultrasonic Hydrophone Based on Short In-Fiber Bragg Gratings. Applied
Optics 37(34): 8120-8128.

Forster F., Scheil E. (1936). Akustischie Untersuchung der Bildung von Marensitnadeln.
Zeitschrift fur Metallkunde 28(9): 245-247 (in German).

Froggatt M., Moore J. (1998). Distributed measurement of static strain in an optical fiber with
multiple Bragg gratings at nominally equal wavelengths. Applied Optics 37(10): 1741–1746.

Furstenau N., Horack H., Schmidt W. (1998). Extrinsic Fabry-Perot interferometer fiber-optic
microphone. IEEE Transactions on Instrumentation and Measurement 47(1): 138-142.

Garg A.C. (1988). Delamination - A damage mode in composite structures. Engineering

Fracture Mechanics 29(5): 5557-584.

Gdoutos E.E., Pilakoutas K., Rodopoulos C.A., (2000). Failure analysis of industrial composite
materials, McGraw-Hill, ISBN: 0-07-134517-5.

Page | 259
Ghezzo F., Nemat-Nasser S. (2007). Effects of embedded SHM sensors on the structural
integrity of glass fiber/epoxy laminates under in-plane loads. Proceedings of the SPIE - The
International Society for Optical Engineering 6530: 65300V1-V6.

Gibbs and Cox (1960). Marine design manual for fiber glass reinforced plastics. McGraw Hill
Book Company, New York.

Goodman S., Tikhomirov A., Foster S. (2008). Pressure compensated distributed feedback fibre
laser hydrophone. Proceedings of SPIE 7004: 700426.1-700426.4, Perth, Australia.

Gosse J.H., Christensen S. (2001). Strain invariant failure criteria for polymers in composite
materials. AIAA-2001-1184, American Institute of Aeronautics and Astronautics, USA.

Goyal V.K., Jaunky N.R., Johnson E.R., Ambur D.R. (2004). Intralaminar and interlaminar
progressive failure analyses of composite panels with circular cut-outs. Composite Structures
64(1): 91-105.

Graff E. (1989). Mechanical behaviour of thick, curved laminated composites. PhD thesis,
Stanford University.

Green A.K., Bowyer W.H. (1981). The development of improved methods for stiffening frames
on large GRP panels. Composites 12(1): 49–55.

Griffith A.A. (1921). The phenomena of rupture and flow in solids. Philosophical Transactions of
the Royal Society of London, Series A 221: 163-198.

Groot P.J.D., Wijnen P.A.M., Janssen R.B.F. (1995). Real-time frequency determination of
acoustic emission for different fracture mechanisms in carbon/epoxy composites. Composite
Science and Technology 55: 405-412.

Grouve W.J.B., Warnet L., de Boer A., Akkerman R., Vlekken J. (2008). Delamination detection
with fibre Bragg gratings based on dynamic behaviour. Composites Science and Technology 68:
2418–2424.

Page | 260
Guggarigoudar P.V. (2005). NDE studies of advanced composites.MS thesis, Wayne State
University.

Guide for building and classing offshore racing yachts (1994). American Bureau of Shipping
(ABS), New York, USA.

Guide for building and classing motor pleasure yachts (2000). American Bureau of Shipping
(ABS), Houston, TX, USA.

Hartmut V. (2002). AE testing fundamentals, equipment and applications. www.NDT.net 7(9).

Hart-Smith L.J. (1993). Should fibrous composite failure modes be interacted or superimposed?
Composites 24(1): 53-55.

Hart-Smith L.J. (1998). Predictions of the original and truncated maximum-strain failure models
for certain fibrous composite laminates. Composites Science and Technology 58: 1151-1178.

Hart-Smith L.J. (2000). What the textbooks won’t teach you about interactive composite failure
criteria. Composite Structures: Theory and Practice, ASTM STP 1383: 413-436. P. Grant and
C.Q. Rousseau, ASTM, West Conshohocken, PA, USA.

Hashin Z., Rotem A. (1973). A Fatigue Criterion for Fiber Reinforced Materials. Journal of
Composite Materials. 7: 448-464.

Hashin Z. (1980). Failure Criteria for Unidirectional Fiber Composites. Journal of Applied
Mechanics 47: 329-334.

Heinz. D., Ritcher B., Weber S. (2000). Application of advanced materials for ship construction –
Experience and problems. Materials and Corrosion 51: 407-412.

Heslehurst R.B., Scott M. (1990). Review of defects and damage pertaining to composite
aircraft components. Composite Polymers 3(2): 103-133.

Page | 261
Hill R. (1948). A theory of the yielding and plastic flow of anisotropic metals. Proceedings of the
Royal Society, Series A 193:1033 281-297.

Hinton M.J., Kaddour A.S., Soden P.D. (2004). Failure Criteria in Fibre Reinforced Polymer
Composites. The World-Wide Failure Exercise. Elsevier Science Publishers.

Hoffman O. (1967). The brittle strength of orthotropic materials. Journal of Composite

Materials 1: 200–206.

Horowitz S.J., Amey D.I. (2001). Ceramics meet next-generation fiber optic packaging
requirements. Photonics Spectra 35(11): 123-126.

Hou J.P., Petrinic N., Ruiz C. (2001). A delamination criterion for laminated composites under
low-velocity impact. Composites Science and Technology 61(14): 2069-2074.

Huguet S., Godin N., Gaertner R., Salmon L., Villard D., (2002). Use of acoustic emission to
identify damage modes in glass fibre reinforced polyester. Composite Science and Technology
62: 1433-1444.

Irwin G.R., (1948). Fracture Dynamics. Fracturing of Metals, 147-166. American Society for
Metals symposium (Trans. ASM 40A), Cleveland, USA.

Irwin G.R. (1957). Analysis of stresses and strains near the end of a crack transversing a plate.
Journal of Applied Mechanics 24: 361–364.

Irwin G.R. (1958). Fracture. Handbuch der Physik, Flügge, ed., 558-590.

Iwamoto M., Ni Q.Q., Fujiwara T., Kurashiki K. (1999). Intralaminar fracture mechanism in
unidirectional CFRP composites: Part I: Intralaminar toughness and AE characteristics.
Engineering Fracture Mechanics 64(6): 721-745.

ISO 12215-5 (2009). Small craft – Hull construction and scantlings, Part 5: Design pressures for
monohulls, design stresses, scantlings determination. The International Organisation for
Standardisation (ISO), Geneva, Switzerland.
Page | 262
Jensen D., Pascual J., August A. (1992). Performance of graphite/bismaleimid laminates with
embedded optical fibre optics, Part I: Uniaxial tension. Smart Materials and Structures 1: 24-30.

Johnson A.F. (1986). Comparison of the mechanical properties of SMC with laminated GRP
materials. Composites 3: 233-239.

Jones R., Galea S. (2002). Health monitoring of composite repairs and joints using optical fibres.
Composite Structures 58(3): 397–403.

Juan C. J., Henry F. T. (2005). Distributed fiber optic intrusion sensor system for monitoring long
perimeters. Proceedings of SPIE 5778: 692–703, Orlando, USA.

Junhou P., Shenoi R.A. (1996). Examination of key aspects defining the performance
characteristics of out-of-plane joints in FRP marine structures. Composites: Part A 27A: 89-103.

Kaczmarek K., Wisnom M.R., Jones M.I. (1998). Edge delamination in curved (04/±456)s glass-
fibre/epoxy beams loaded in bending. Composite Science and Technology 58: 155–161.

Kaiser J. (1950). Untersuchungen uber das Auftreten Geraushen beim Zugversuch (An
Investigation into the Occurrence of Noises in Tensile Tests). PhD dissertation, Technische
Hochschule, Munich Germany.

Karbhari V.M., (2004). E-Glass/Vinylester composites in aqueous environments: Effects on

short-beam shear strength. Journal of Composites for Construction 8(2): 148-156.

Kedward K.T., Wilson R.S., Mclean S.K. (1989). Flexure of simply curved composite shapes.
Composites 20(6): 527-536.

Keel (2010). In Encyclopædia Britannica. Retrieved September 14, 2010, from Encyclopaedia
Britannica Online: http://www.britannica.com/EBchecked/topic/314095/keel.

Kesavan A., John S., Herszberg I. (2008). Strain-based Structural Health Monitoring of Complex
Composite Structures. Structural Health Monitoring 7(3): 203-213.

Page | 263
Kilic O., Digonnet M., Kino G., Solgaard O. (2008). Photonic-crystal-diaphragm-based fiber-tip
hydrophone optimized for ocean acoustics. Proceedings of SPIE 7004: 700405.1-700405.4.

Kim R.Y., Soni S.R. (1984). Experimental and analytical studies on the onset of delamination in
laminated composites. Journal of Composite Materials 18(1): 70-80.

Kim D.H., Koo B.Y., Kim C.G., Hong C.S. (2004). Damage detection of composite structures using
a stabilized extrinsic Fabry-Perot interferometric sensor system. Smart Materials and Structures
13: 593-598.

Kim J.K., Sham M.L. (2000). Impact and delamination failure of woven-fabric composites.
Composite science and technology 60: 745-761.

Kinjo T., Hiroaki S., Takemoto M., Ono K., (1997). Fracture mode classification in glass-fibre
composites by AE source wave characterisation and autoregressive coefficients. (in Japanese)
Journal of Applied Physics 36: 3281-3286.

Kirkendall C.K., Dandridge A. (2004). Overview of high performance fibre-optic sensing. Journal
of Physics D: Applied Physics 37(18): 197-216.

Kishinouye F. (1934). An experiment on the progression of Fracture (A preliminary report).

Jishin 6:25-31 (in Japanese).

Kitade S., Fukuda T., Osaka K. (1995). Fiber optic method for detection of impact induced
damage in composite laminates. Journal of the Society of Materials Science 44: 1196–1200.

Kitching R., Tau A.L., Abu-Mansour T.M.N. (1984). The influence of through thickness properties
on glass reinforced plastic laminated structures. Composite Structures. 2: 105-151.

Kotsikos G. Evans J. T., Gibson A. G., Hale J. M. (2000). Environmentally enhanced fatigue
damage in glass fibre reinforced composites characterised by acoustic emission. Composites
Part A 31(9): 969-977.

Page | 264
Krueger R. (2004). Virtual crack closure technique: History, approach and applications. Applied
Mechanics review 57(2): 109-143

Kuang K.S.C. Cantwell W.J. (2003). Use of conventional optical fibers and fiber Bragg gratings
for damage detection in advanced composite structures: A review. Applied Mechanical Review
56(5): 493-513.

Kullander F., Vahlberg C. (2005). Towards a thin and lightweight fibre optic towed array sonar.
Archives of Acoustics 30: 91-94.

Lakshminarayana H.V. (1984). A symmetric rail shear test for Mode II fracture toughness (GllC)
of composite materials - finite element analysis. Journal of Composite Materials 18: 227-238.

Lammerant L., Verpoest I. (1996). Modelling of the interaction between matrix cracks and
delaminations during impact of composite plates. Composites Science and Technology 56:
1171-1178.

Larsson L., Eliasson R.E. (1994) Principles of Yacht Design, Adlard Coles Nautical, London,
ISBN:0-7136-3855-9, UK.

LeBlanc M., Measures R.M. (1992). Impact damage assessment in composite materials with
embedded fiber-optic sensors. Composites Engineering 2(5-7): 573–596.

Lee B., Jeong Y. (2002). Interrogation techniques for fiber grating sensors and the theory of
fiber gratings. Fiber Optic Sensors, F. T. S. Yu, S. Yin, eds., Marcel Dekker, New York, USA 295—
382.

Lee D.C., Lee J.J., Yun S.J. (1995). The mechanical characteristics of smart composite structures
with embedded optical fibre sensors. Composite Structures 32: 39-50.

Lee D.C., Lee J.J., Kwon I.B., Seo D.C. (2001). Monitoring of fatigue damage of composite
structures by using embedded intensity-based optical fiber sensors. Smart Materials and
Structures 10: 285–292.

Page | 265
Lee J. R., Jeong H. M. (2010). Design of resonant acoustic sensors using fiber Bragg gratings.
Measurement Science and Technology 21(5): 1–6.

Lee S.M. (1990). Failure mechanisms of edge delaminations of composites. Journal of

Composite Materials 24: 1200-1212.

Lee S.M. (1993). An edge crack torsion method for Mode III delamination fracture testing.
Journal of Composites Technology and Research 15(3) 193–201.

Lemaire P.J., Atkins R.M., Mizrahi V., Reed W.A. (1993). High pressure HZ loading as a
technique for achieving ultrahigh photosensitivity and thermal sensitivity in Ge02 doped optical
fibres. Electronics Letters 29:13 1191-1 193.

Leszek G, Kanji O. (1995). Acoustic emission analysis of laminate failure mechanisms with
reference to failure criteria, AECM-5: Fifth International Symposium on Acoustic emission from
composite 5: 170-178.

Leung C.K.Y., Yang Z., Ying X., Tong P., Lee S.K.L., (2005). Delamination Detection in Laminate
Composites with an Embedded Fiber Optical Interferometric Sensor. Sensors and Actuators A-
Physical, 119: 336-344.

Li H.C.H., Herszberg I., Mouritz A.P., Davis C.E., Galea S.C. (2004). Sensitivity of embedded fibre
optic Bragg grating sensors to disbonds in bonded composite ship joints. Composite Structures
66: 239-248.

Li H.C.H., Herszberg I., Davis C.E., Mouritz A.P, Galea S.C. (2006). Health monitoring of marine
composite structural joints using fibre optic sensors. Composite Structures, 75(1–4): 321–7.

Li L., Zhao J.H. (1986). The Monitoring Damage Growth Processes in Glass Fiber Reinforced
Composite by Amplitude Analysis. Second International Symposium on Acoustic Emission from
Reinforced Composites Montreal: Reinforced Plastics/Composites Institute: 90-95.

Li R., Kelly D., Ness R. (2003). Application of first invariant strain criterion for matrix failure in
composite materials. Journal of Composite Materials 37(22): 1977-2000.
Page | 266
Liang S., Zhang C., Lin W., Li L., Li C., Feng X., Lin B. (2009). Fiber-optic intrinsic distributed
acoustic emission sensor for large structure health monitoring. Optics Letters 34: 1858-1860.

Lin Y. B., Chang K. C., Chern J. C., Wang L. A.(2005). Packaging methods of fiber-Bragg grating
sensors in civil structure applications. IEEE Sensors Journal 5(3): 419-423.

Ling H.Y., Lau K.T., Lam C. (2005a). Effects of embedded optical fibre on mode II fracture
behaviours of woven composite laminates. Composites Part B: 36(6-7): 534-543.

Ling H.Y., Lau K.T., Cheng L., Jin W. (2005b). Utilization of embedded optical fibre sensors for
delamination characterization in composite laminates using a static strain method. Smart
Materials and Structures 14: 1377–1386.

Ling H.Y. (2006). Vibration and damage monitoring in advanced composites using multiplexed
fibre Bragg grating (FBG) sensors. PhD thesis, The Hong Kong Polytechnic University.

Liptai R.G., Harris D.O. (1971). Acoustic emission-An introductory review. Materials Research
and Standards 11(3): 8-10.

O'Brien TK. (1982a). Characterisation of delamination onset and growth in a composite

laminate, Damage in Composite Materials. ASTM STP 775, Reifsnider KL(ed), American Society
for Testing and Materials, Philadelphia, 140-167.

O'Brien T.K., Johnston N.J., Morris D.H. Simonds R.A. (1982b). A simple test for the interlaminar
fracture toughness of composites. SAMPE Journal 8-15.

Maligno A.R. (2007). Finite element investigations on the microstructures of composite

materials. PhD thesis, The University of Nottingham, UK.

Mall S. Dosedel S.B., Holl M.W. (1996). The performance of graphite-epoxy composite with
embedded optical fibres under compression. Smart Materials and Structures 5: 209-215.

Manual of GRP ship structural design, NES 140, (1982), Ministry of Defence (PE), Bath, UK.

Page | 267
Martin A.R., Fernando G.F., and Hale K.F. (1997). Impact damage detection in filament wound
tubes using embedded optical fiber sensors. Smart Materials and Structures 6: 470–476.

Martin R.H., Jackson W.C. (1993). Damage Prediction in Cross-Plied curved composite
laminates. Composite Materials. Fatigue and Fracture 4: 105-126.

Martin R.H., Davidson B.D. (1997). Mode II fracture toughness evaluation using a four point
bend end notch flexure test. Proceedings 4th International Conference on Deformation and
Fracture of Composites, Institute of Materials 243-252.

Martin R.H. Davidson B.D. (1999). Mode II fracture toughness evaluation using a four point
bend end notched flexure test. Plastics, Rubber Composites 28(8): 401–406.

Matsuo T., Cho H., Takemoto M. (2006). Optical fiber acoustic emission system for monitoring
molten salt attack. Science and Technology of Advanced Materials 7(1): 104-110.

Mazumdar S.K. (2001). Composites Manufacturing: Materials, Product, and Process

Engineering. CRC Press, 1st edition, ISBN: 978-0849305856, Florida, USA.

McKenzie I., Jones R., Marshall I.H., Galea S. (2000). Optical fiber sensors for health-monitoring
of bonded repair systems. Composite Structures 50: 405–416.

Measures R.M., Glossop N.D.W., Lymer J., LeBlanc M., Dubois S., Tsaw W., Tennyson R.C.
(1989). Structurally integrated fiber optic damage assessment system for composite materials.
Applied Optics 28: 2626 – 2633.

Meltz G. Morey W.W. (1992). Bragg grating formation and germanosilicate fiber
photosensitivity. Proceedings of SPIE 1516: 185-99.

Miguel J., Higuera L. (2002). Introduction to fibre optic sensing technology. Handbook of
Optical Fibre Sensing Technology, J. Miguel and L. Higuera, eds., John Wiley & Sons, West
Sussex, UK, 1-30.

Page | 268
Moan T. (2003). Marine structures for the future. Core report No. 2003-01, Centre for Offshore
Research and Engineering, National University of Singapore.

Morris P., Beard P., Hurrell A. (2005). Development of a 50MHz optical fibre hydrophone for the
characterisation of medical ultrasound fields, IEEE Ultrasonics Symposium, Rotterdam,
Netherlands: 1747-1750.

Mouritz A.P., Leong K.M., Herszberg I. (1997a). A review of the effect of stitching on the in-
plane mechanical properties of fibre-reinforced polymer composites. Composites Part A 28:
979-991.

Mouritz A.P., Gallagher J., Goodwin A.A. (1997b). Flexural strength and interlaminar shear
strength of stitched GRP laminates following repeated impact. Composites Science and
Technology 87: 509-522.

Mouritz A.P., Baini C., Herszberg I. (1999). Mode I interlaminar fracture toughness properties of
advanced textile fibreglass composites. Composites: Part A 30: 859-870.

Muravin B. (2009). Acoustic emission science and technology, Journal of Building and
Infrastructure Engineering of the Israeli Association of Engineers and Architects.

Muravin B., Acoustic emission method - History, fundamentals and applications.

www.muravin.com.

MSC MARC® 2008r1 -Volume A (2008). Theory and User Information. MSC. Software
Corporation, CA, USA.

Nahas M.N. (1985). Analysis of composite failure mechanisms using acoustic emission and
ultrasonic scanning techniques, Composites 16(2): 148-152.

Nash P. (1996). Review of interferometric optical fibre hydrophone technology. IEE

Proceedings-Radar, Sonar and Navigation 143(3): 204-209, 1996.

Page | 269
Nash P. J., Cranch G. A., Hill D. J. (2000). Large scale multiplexed fibre-optic arrays for
geophysical applications. Proceedings of SPIE 4202: 55-65, Boston, USA.

Okabe Y., Yashiro S., Kosaka T., Takeda N. (2000). Detection of transverse cracks in CFRP
composites using embedded fiber Bragg grating sensors. Smart Materials and Structures 9:
832–838.

Ono K. (1998). Acoustic emission behaviour of flawed unidirectional carbon fiber-epoxy

composites. Journal of Reinforced Plastics and Composites 7(1): 90-105.

Optical fiber sensors guide: Fundamentals and applications, Micron Optics Inc. GA, USA.

Oskouei A.R., Ahmadi M. (2010). Fracture strength distribution in E-glass fibre using acoustic
emission. Journal of Composite Materials 44: 693-705.

Othonos A., Kalli K. (1999). Fiber Bragg Gratings: Fundamentals and Applications in
Telecommunications and Sensing. Artech House Optoelectronics Library, Norwood, USA.

Pappu R.P., Zhang W., Bennion I., Sugden K., (2009). Acoustic emission detection using optical
fiber base perot sensor and quadrature recombination technique. Proceedings of European
conference on Lasers and Electro-optics and the European Quantum electronics conference
(CLEO/EQEC 2009).

Park J.M., Kim J.W., Yoon D.J. (2002). Interficial evolution and microfailure mechanisms of
single carbon fiber/Bismaleimide (BMI) composites by tensile and compressive fragmentation
tests and Acoustic Emission. Composite Science and Technology 62: 743-756.

Pang M., Zhang M., Wang L. W., Zou Q. L., Kuang W., Wang D. N., Liao, Y. B. (2007). Phase
Mode-Matching Demodulation Scheme for Interferometric Fiber-Optic Sensors. Photonics
Technology Letters, IEEE 19(1):39-41.

Park R. Jang J. (2004). Effect of surface treatment on the mechanical properties of glass
fiber/vinylester composites. Journal of Applied Polymer Science. 91: 3730-3736.

Page | 270
Pearce G.P. (2010). High Strain rate behaviour of bolted joints in carbon fibre composite
structures. PhD Thesis, University of New South Wales, Sydney.

Pearson J.D., Zikry M.A., Prabhugoud M., Peters K., (2007). Global-local assessment of low-
velocity impact damage in woven composites. Journal of Composite materials 41(23): 2759-
2783.

Pearson R.A., Yee A.F. (1989). Toughening mechanisms in elastomer-modified epoxies. Journal
of Materials Science 24: 2571–80.

Perez, I., Cui, H.L. Udd, E. (2001). Acoustic emission detection using fiber Bragg gratings. SPIE,
4328: 209-215.

Perrot Y., Baley C., Grohens Y. and Davies P. (2007). Damage resistance of composites based on
glass fibre reinforced low styrene emission resins for marine applications. Applied Composite
materials 14: 67-87.

Petrossian Z., Wisnom M.R. (1998). Prediction of delamination initiation and growth from
discontinuous plies using interface elements. Composites Part A: 29(5-6): 503-515.

Phillips H.J., Shenoi R.A., Moss C.E. (1999). Damage mechanics of Top-hat-stiffeners used in FRP
ship construction. Marine Structures 12: 1-19.

Portevin A., Le Chatelier F., (1923). Sur un phénomène observé lors de l'essai de traction
d'alliages en cours de transformation. Comptes Rendus de 'Académie des Sciences Paris 176:
507-510.

Prel Y.J., Davies P., Benzeggagh M.L., de Charentenay F.X. (1989). Mode I and Mode II
delamination of thermosetting and thermoplastic composites. Composite Materials: Fatigue
and Fracture, 2nd volume, ASTM STP 1012: 251-269. P.A. Lagace, American Society for Testing
and Materials, Philadelphia.

Puck, A., Schurmann, H. (1998). Failure analysis of FRP laminates by means of physically based
phenomenological models. Composite Science and Technology 58: 1045-1067.
Page | 271
Puck A., Schurmann H. (2002). Failure analysis of FRP laminates by means of physically based
phenomenological models. Composites Science and Technology 62(12-13): 1633-1662.

Purslow D., (1988). Fractography of fibre-reinforced thermoplastics, Part 3. Tensile,

compressive and flexural failures. Composites 19(5): 358-366.

Rao Y.J. (1997). In-fibre Bragg grating sensors. Measurement Science and Technology 8: 355-
375.

Read I., Foote P., Murray S. (2002). Optical fibre acoustic emission sensor for damage detection
in carbon fibre composite structures. Measurement Science and Technology 13: N5-N9.

Reeder J.R., Crews Jr. J.H. (1990). The mixed-mode bending method for delamination testing.
AIAA Journal 28(7): 1270-1276.

Reeder J.R., Crews Jr J.H. (1990). Nonlinear analysis and design of the mixed mode bending
delamination test. NASA Technical Memorandum 102777, 1991.

Report on the investigation of the keel failure, capsize and loss of one crew member from the
Max Fun 35 yacht Hooligan V (2007). Marine Accident Investigation Branch (MAIB),
Southampton, UK.

Rikards R., Buchholz F.G., Wang H., Bledzki A.K., Korjakin A., Richard H.A. (1998). Investigation
of mixed mode I/II interlaminar fracture toughness of laminated composites by using a CTS
type specimen. Engineering Fracture Mechanics 61: 325–342.

Roberts S.S.J., Davidson R. (1991). Mechanical properties of composite materials containing

embedded fiber optic sensors. SPIE 1588: 326–341.

Rules for Classification of high speed and light craft (1991) Det Norske Veritas (DNV), Hovik.

Silva-Muñoz R.A., Lopez-Anido R.A. (2009). Structural health monitoring of marine composite
structural joints using embedded fiber Bragg grating strain sensors. Composite Structures 89:
224–234.
Page | 272
Rotem A. (1996). Prediction of laminated failure with the Rotem failure criterion. Composite
Science and Technology 58: 1083-1094.

Rowe W.J., Rausch E.O., Dean P.D. (1986). Embedded optical fiber strain sensor for composite
structure applications. SPIE 718: 266–273.

Rules for Yachts and small craft (1983) Lloyd’s Register for shipping, London.

Russel S.S., Henneke E.G. (1977). Signature analysis of acoustic emission from graphite/epoxy
composites. Interim report number VPI-E-77-22, NASA Grant NSG 1238.

Rybicky E.F., Kanninen M.F. (1977). A finite element calculation of stress intensity factors by a
modified crack closure integral. Engineering Fracture Mechanics 9(4): 931-938.

Sailing.hu (2005). Tragedy on the Adriatic. Internet sailing Magazin, Hungary. http://www.
Sailing.hu.

Sanders R. C. and Grant P. (1982). The strength of laminated plates under in-plane loading. BAe
Report SOR(P)130.

Schwartz M.M. (1996a). Composite Materials - Properties, Non-destructive testing and Repair,
Volume 1. Prentice Hall Inc., NJ, USA, 1996, ISBN 0-13-300047-8.

Schwartz M.M. (1996b). Composite Materials - Processing, Fabrication and Applications,

Volume 2. Prentice Hall Inc., NJ, USA, 1996, ISBN 0-13-300039-7.

Schwartz M. (2009). Smart Materials. CRC Press, ISBN: 13:978-1-4200-4372-3, Florida, USA.

Scott A. M., de Vries M. J., Vivek A., Richard O. C., Noel Z. (1997). Advances in optical fiber
sensors for vehicle detection. Proceedings of SPIE 3207: 318-322, Pittsburgh, USA.

Shenoi R.A., Wellicome JF. (1993). Composite materials in maritime structures: Volume 2 -
Practical considerations, 1st edition, Cambridge University Press, New York, ISBN 0-521-45154-
X.
Page | 273
Shenoi R.A., Hawkins G.L., (1995). An investigation into the performance characteristics of Top-
hat-stiffener to shell plating joints. Composite Structures 30: 109-121.

Shenoi R.A., Dodkins A.R., (2000). Design of ships and marine structures made from FRP
composite materials. Comprehensive Composite Materials - Volume 6 - Design and
applications, Pergamon, ISBN: 0-08042-993-9

Shenoi R.A., Wang W. (2001). Through-thickness stresses in curved composite laminates and
sandwich beams. Composites Science and Technology 61: 1501-1512.

Shivakumar K., Emmanwori L. (2004). Mechanics of failure of composite laminates with an

embedded fiber optic sensor. Journal of Composite Materials 38(8): 669-680.

Shull P.J. (2002). Nondestructive evaluation: Theory, Techniques and Applications. Marcel
Dekker, Volume 6, 3rd edition, New York, ISBN:0-8247-8872-9.

Silva, J.M.A., Ferreira J.A.M., Devezas T.C. (2003). Fatigue damage of carbon-epoxy laminates
with embedded optical fibres. Materials Science and Technology 19: 809-814.

Smith C.S. (1990). Design of marine structures in composite materials. Elsevier Applied Science,
London, ISBN-10: 1851664165.

Sorensen L., Botsis J., Gmür T., Cugnoni J., (2007). Delamination detection and characterisation
of bridging tractions using long FBG optical sensors. Composites Part A: 38(10): 2087–2096.

SP Systems. Guide to composites. Gurit Inc., RI, USA. GTC-1-1098.

Soutis C. (1991). Measurement of the static compressive strength of carbonfibre/epoxy

laminates. Composites Science and Technology 42(4): 373-392.

Sridharan S. (2008). Delamination behaviour of composites. Boca Raton, FL: CRC Press, ISBN -
9-781420-079678.

Page | 274
Staszewski W., Boller C., Tomlinson G. (2004). Health monitoring of aerospace structures:
Smart sensor technologies and signal processing. John Wiley & Sons, Ltd, England, 2004, ISBN
0-470-84340-3.

Strong A. B., (1993). Fundamentals of composites manufacturing, Society of Manufacturing

Engineers. Michigan, ISBN: 0-87263-358-6.

Sun C.T., Tao J. (1997). Prediction of failure envelopes and stress/strain behaviour of composite
laminates. Composite Science and Technology 58(7): 1125-1136.

Sun C.T. (2000). Strength analysis of unidirectional composites and laminates. Comprehensive
Composite Materials. A. Kelly, C. Zweben, Elsevier Science Ltd., Oxford, UK Ch. 1.20: 2-25.

Sundararaman V. and Davidson B.D. (1998). An unsymmetric end notched flexure test for
interfacial fracture toughness determination. Engineering Fracture Mechanics 60: 361–377.

Suzuki M., Nakamishi H., Iwamoto M., Jinen E. (1988). Application of static fracture
mechanisms to fatigue fracture behaviour of class A-SMC composite. Proceedings of 4th Japan-
US Conference on Composite materials 297-306.

Takahashi N., Yoshimura K., Takahashi S. (2001). Vibration sensing with fiber Bragg grating.
Proceedings of SPIE 4513:1-6, Vladivostok, Russia.

Tanaka S., Yokosuka H., Takahashi N. (2005). Temperature–stabilized fiber Bragg grating
underwater acoustic sensor array using incoherent light. Proceedings of SPIE 5855: 699-702,
Bruges, Belgium.

Takeda N., Kosaka T., Ichiyama T. (1999). Detection of transverse cracks by embedded plastic
optical fiber in FRP laminates. SPIE 3670: 248–255.

Takeda N., Okabe Y., Kuwahara J., Kojima S., Ogisu T. (2005). Development of smart composite
structures with small-diameter fiber Bragg grating sensors for damage detection: Quantitative
evaluation of delamination length in CFRP laminates using Lamb wave sensing. Composites
Science and Technology 65(15– 16): 2575–87.
Page | 275
Talreja R., Manson J.A.E. (2001). Comprehensive Composite Materials - Volume 2 - Polymer
Matrix Composites, Pergamon, ISBN: 0-080437-257.

Tan S.C. (1994). Stress Concentrations in Laminated Composites. Technomic Publishing

Company Inc., USA, ISBN 1-56676-077-1.

Tao X., Tang L., Du W., Choy C., (2000). Internal strain measurement by fiber Bragg grating
sensors in textile composites. Composites Science and Technology 60: 657-669.

Tosi D., Olivero M., Perrone G. (2008). Low–cost fiber Bragg grating vibroacoustic sensor for
voice and heartbeat detection. Applied Optics 47(28): 5123–5129.

Tsai S.W., Wu E.M. (1971).A General Theory of Strength for Anisotropic Materials. Journal of
Composite Materials 5(1): 58-80.

Tsuda H., Sato E., Nakajima T., Nakamura H., Arakawa T., Shiono H., Minato M., Kurabayashi
H., Sato A. (2009) Acoustic emission measurement using a strain-insensitive fiber Bragg grating
sensor under varying load conditions., Optics Letters 34(19): 2942-2944.

Turon A., Da´vila C.G., Camanho P.P., Costa J. (2007). An engineering solution for mesh size
effects in the simulation of delamination using cohesive zone models. Engineering Fracture
Mechanics 74: 1665–1682.

Tygavac Advance Materials Ltd., Resin infusion from www.tygavac.co.uk/process/resin-

infusion.html.

Ujjin R. (2007). Prediction of bearing failure in pion-loaded laminates. PhD thesis, University of
New South Wales, Sydney, Australia.

Ussorio M., Wang H., Ogin S.L., Thorne A.M., Reed G.T., Tjin S.C., Suresh R. (2006).
Modifications to FBG sensor spectra due to matrix cracking in a GFRP composite. Construction
and Building Materials, 20:(1-2) 111-118.

Page | 276
Vakoc B. J, Digonnet M. J. F., Kino G. S. (1999). A novel fiber-optic sensor array based on the
Sagnac interferometer. IEEE Journal of Lightwave Technology 17(11): 2316-2326.

Vandenplas S., Papy J.M., Wevers M., Huffel S.V., Inaudi D. (2005). Acoustic emission
monitoring using a polarimetric single mode optical fiber sensor, Proceedings of 17th
International conference on optical fibre sensors SPIE 5855: 1064-1067.

Vogler T.J., Kyriakides S. (2001). On the initiation and growth of kink bands in fiber composites:
Part I. experiments. International Journal of Solids and Structures 38(15): 2639-2651.

von Mises R. (1913). Mechanik der festen Korper im plastisch deformablen Zustand.
Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, Mathematisch-
Physikalische Klasse 1913: 582-592.

Wang C.H. (1996a). Introduction to Fracture Mechanics, DSTO, Melbourne, Australia.

Wang H., Vu-Khanh T. (1996b). Use of end-loaded-split (ELS) test to study stable fracture
behaviour of composites under mode II loading. Composite Structures 36: 71–9.

Wang H., Ogin S.L., Thorne A.M., Reed G.T. (2005). Interaction between optical fibre sensors
and matrix cracks in cross-ply GFRP laminates. Part 2: Crack detection. Composites Science and
Technology 66(13): 2367-2378.

Webb D. J., Surowiec J. Sweeney M. Jackson D.A. Gavrilov L.R. Hand J.W., Zhang L. Bennion I.
(1996). Miniature fibre optic ultrasonic probe. Proceedings of SPIE 2839: 76-80, Denver, USA.

Wang J.T., Raju I.S. (1996). Strain energy release rates formulae for skin-stiffener debond
modeled with plate elements. Engineering Fracture Mechanics 54(2): 211-228.

Wisnom M.R. (1996). 3-D Finite element analysis of curved beams in bending. Journal of
Composite Material 30(11): 1178–1190.

Page | 277
Wei C. Y., Ye C. C., James S. W., Tatam R. P., Irving P. E. (2002). The influence of hydrogen
loading and the fabrication process on the mechanical strength of optical fibre Bragg gratings.
Optical Materials 20(4): 241-251.

Wild G., Hinckley S. (2010). Spatial performance of acousto-ultrasonic fiber Bragg grating
sensor. IEEE Sensors Journal 10(4): 805-806.

Xu Y., Leung C.K.Y., Tong P., Yi J., Lee S.K.L. (2005). Interfacial debonding detection in bonded
repair with a fiber optical interferometric sensor. Composites Science and Technology 65(9):
1428-1435.

Yachting World, Jan 2008

Yamada S.E. Sun C.T. (1978). Analysis of Laminate Strength and Its Distribution. Journal of
Composite Materials 12(3): 275-284.

Ye L. (1988). Role of matrix resin in delamination onset and growth in composite laminates.
Composites Science and Technology 33(4): 257-277.

Yeh H.Y., Chern C. (1998). The Yeh–Stratton Criterion for stress concentration in fibre-
reinforced composite materials. Journal of Composite Materials 32(2): 141–157.

Yun B., Wang Y., Li A., Cui Y. (2006). Tunable fiber laser based fiber Bragg grating strain sensor
demodulation system with enhanced resolution by digital signal processing. Microwave and
Optical Technology Letters 48(7): 1391-1393.

Zhang F., Zhang W., Li F., Liu Y. (2009). Four-element fiber laser hydrophone array. Proceedings
of SPIE 7514: 75140L.1-75140L.7.

Zhang X. (1998). Impact damage in composite aircraft structures-experimental testing and

numerical simulation. Proceedings of the Institution of Mechanical Engineers, Part G: Journal
of Aerospace Engineering 212(4): 245-259.

Page | 278
Zinoviev, P.A., Lebedeva, O.V., Tairova, L.P. (2002). A coupled analysis of experimental and
theoretical results on the deformation and failure of composite laminates under a state of
plane stress. Composites Science and Technology 62(12-13): 1711-1723.

Page | 279
Appendix

Page | 280
Appendix A – Hand-layup - Tensile test Result
The results for hand-layup manufactured tensile specimens are presented in Tables A1, A2 and
A3 whereas the results from vacuum infused process (VIP) are presented in Tables A4, A5, A6,
A7, A8 and A9. VIP specimens include CSM and CFM to compare the difference in strength for
both types of reinforcements (short fibre and continuous fibre). Specimen length of 175 mm
and gauge of 50 mm were used.

Hand Avg Avg Tensile Young's Avg Avg

Fibre No. Post-curing Layup- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
0
CSM AA1 20 C 24hrs 925H 24.82 4.30 186.97 9604
CSM AA2 200C 24hrs 925H 22.99 4.66 167.42 9578
CSM AA3 200C 24hrs 925H 21.63 4.78 174.51 9639 176.30 9607
0
CSM AB1 50 C 24hrs 925H 24.27 4.98 153.19 9502
CSM AB2 500C 24hrs 925H 22.43 5.11 125.28 10370
0
CSM AB3 50 C 24hrs 925H 22.58 4.95 154.76 10010 144.41 9961
CSM AC1 820C 2hrs 925H 23.39 5.23 144.64 9490
0
CSM AC2 82 C 2hrs 925H 22.49 5.05 127.35 11274
CSM AC3 820C 2hrs 925H 22.17 5.08 156.32 9527
0
CSM AC4 82 C 2hrs 925H 23.27 5.17 147.58 10352 143.97 10161
Table A1: CSM 925H hardener hand-layup tensile test

Hand Avg Avg Tensile Young's Avg Avg

Fibre No. Post-curing Layup- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
0
DB BA1 20 C 24hrs 925H 21.38 3.83 76.07 7859
DB BA2 200C 24hrs 925H 22.09 3.84 66.89 7614
DB BA3 200C 24hrs 925H 22.52 3.89 83.82 8144 75.59 7872
0
DB BB1 50 C 24hrs 925H 20.24 3.86 92.98 9515
DB BB2 500C 24hrs 925H 23.16 3.92 137.91 9723
0
DB BB3 50 C 24hrs 925H 24.39 3.82 116.38 9286 115.75 9508
DB BC1 820C 2hrs 925H 23.78 3.87 111.52 9582
0
DB BC2 82 C 2hrs 925H 22.83 3.87 85.41 8526
DB BC3 820C 2hrs 925H 21.96 3.83 106.97 9235
0
DB BC4 82 C 2hrs 925H 24.13 3.89 87.64 8662 97.88 9001
Table A2: DB 925H hardener hand-layup tensile test

Page | 281
 Hand Avg Avg Tensile Young's Avg Avg
Fibre No. Post-curing Layup- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
UD-0 CA1 200C 24hrs 925H 13.85 3.74 430.56 23377
UD-0 CA2 200C 24hrs 925H 13.74 3.96 452.39 23562
0
UD-0 CA3 20 C 24hrs 925H 13.42 3.66 448.49 23790 443.81 23576
UD-0 CB1 500C 24hrs 925H 15.61 3.81 456.85 22332
0
UD-0 CB2 50 C 24hrs 925H 14.14 3.72 485.54 22397
UD-0 CB3 500C 24hrs 925H 16.36 3.91 441.81 22809 461.40 22512
0
UD-0 CC1 82 C 2hrs 925H 14.60 3.85 445.55 23484
UD-0 CC2 820C 2hrs 925H 14.30 3.80 449.33 22774
0
UD-0 CC3 82 C 2hrs 925H 14.93 3.90 412.92 22397
UD-0 CC4 820C 2hrs 925H 15.95 3.77 431.65 22742 434.86 22849
Table A3: UD 925H hardener hand-layup tensile test

VIP - Tensile test (925H-Set 01, CHM50-Set 02)

Vacuum Avg Avg Tensile Young's Avg Avg

Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
0
CFM A01 20 C 24hrs 925H 23.87 5.56 144.64 7690
CFM A02 200C 24hrs 925H 24.36 5.39 139.46 9613
CFM A03 200C 24hrs 925H 23.61 5.53 156.32 8529
0
CFM A04 20 C 24hrs 925H 23.47 5.64 140.88 7601
CFM A05 200C 24hrs 925H 23.46 5.75 110.15 8242
0
CFM A06 20 C 24hrs 925H 22.88 5.73 135.45 8391 137.82 8344
Table A4: CFM 925H hardener VIP tensile test

Vacuum Avg Avg Tensile Young's Avg Avg

Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
0
CFM B01 20 C 24hrs CHM50 22.51 5.83 117.50 7464
CFM B02 200C 24hrs CHM50 22.59 5.86 108.03 6075
0
CFM B03 20 C 24hrs CHM50 21.19 5.76 117.90 7674
0
CFM B04 20 C 24hrs CHM50 21.83 5.91 104.36 7187
0
CFM B05 20 C 24hrs CHM50 22.25 5.85 109.36 6969 111.43 7074
Table A5: CFM CHM50 hardener VIP tensile test
Page | 282
 Vacuum Avg Avg Tensile Young's Avg Avg
Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
DB C01 200C 24hrs 925H 24.96 3.09 99.34 7544
DB C02 200C 24hrs 925H 24.99 3.10 97.60 8157
DB C03 200C 24hrs 925H 24.70 3.11 88.43 7277
DB C04 200C 24hrs 925H 25.34 3.10 94.65 6934 95.01 7478

Table A6: DB 925H hardener VIP tensile test

Vacuum Avg Avg Tensile Young's Avg Avg

Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
DB D01 200C 24hrs CHM50 26.16 3.50 97.31 6745
DB D02 200C 24hrs CHM50 25.82 3.37 103.18 7096
DB D03 200C 24hrs CHM50 24.79 3.48 101.44 6596
DB D04 200C 24hrs CHM50 24.62 3.51 102.56 6377 103.18 6703
DB D05 500C 24hrs CHM50 23.64 3.42 123.51 6237
DB D06 500C 24hrs CHM50 24.59 3.40 123.22 8063
DB D07 500C 24hrs CHM50 25.28 3.46 123.76 7191
DB D08 500C 24hrs CHM50 25.78 3.39 124.20 9456 123.68 7737
DB D09 820C 2hrs CHM50 24.70 3.46 124.48 8093
DB D10 820C 2hrs CHM50 24.95 3.50 115.31 7941
DB D11 820C 2hrs CHM50 25.79 3.44 124.52 8155
DB D12 820C 2hrs CHM50 24.54 3.54 113.49 7522 119.45 7928

Table A7: DB CHM50 hardener - post curing VIP tensile test

Vacuum Avg Avg Tensile Young's Avg Avg

Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
UD-90 E01 200C 24hrs CHM50 24.75 3.54 76.52 13448
UD-90 E02 200C 24hrs CHM50 24.64 3.53 74.57 16567
UD-90 E03 200C 24hrs CHM50 25.27 3.56 66.23 15166
UD-90 E04 200C 24hrs CHM50 24.79 3.54 70.86 16683
UD-90 E05 200C 24hrs CHM50 24.75 3.55 72.89 13774 72.21 15128

Table A8: UD-900 CHM50 hardener VIP tensile test - L=175mm

Page | 283
 Vacuum Avg Avg Tensile Young's Avg Avg
Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
UD-0 F01 200C 24hrs CHM50 15.22 3.63 736.76 35943
UD-0 F02 200C 24hrs CHM50 15.12 3.66 720.36 36692
UD-0 F03 200C 24hrs CHM50 14.64 3.55 747.66 36932
UD-0 F04 200C 24hrs CHM50 15.98 3.63 676.09 37232
UD-0 F05 200C 24hrs CHM50 17.33 3.55 728.69 36070 747.66 37232

Table A9: UD-00 CHM50 hardener VIP tensile test

Page | 284
Appendix B – Hand-layup 3-point bend test results
Hand Avg Avg Flexure Flexure Avg Avg
Fibre No. Post-curing Layup- w t Stress Modulus Stress Mod
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
CSM DA1 200C 24hrs 925H 11.74 4.28 233.85 1757
CSM DA2 200C 24hrs 925H 11.99 3.85 202.12 1563
CSM DA3 200C 24hrs 925H 13.42 4.46 206.74 1757
CSM DA4 200C 24hrs 925H 12.91 4.57 248.48 2209 222.8 1822
CSM DB1 500C 24hrs 925H 13.00 4.86 243.70 2350
0
CSM DB2 50 C 24hrs 925H 11.94 4.40 256.68 2418
0
CSM DB3 50 C 24hrs 925H 13.15 4.07 248.67 2298
0
CSM DB4 82 C 2hrs 925H 12.08 4.51 253.92 2361 250.74 2356
CSM DC1 820C 2hrs 925H 12.47 4.80 220.20 2154
CSM DC2 820C 2hrs 925H 12.58 4.37 231.62 1953
CSM DC3 820C 2hrs 925H 10.73 4.57 260.16 2024
CSM DC4 820C 2hrs 925H 12.05 4.10 260.98 2466 243.24 2150

Table B1: CSM 925H hardener hand-layup 3-point bend test - L=110 mm

Hand Avg Avg Flexure Flexure Avg Avg

Fibre No. Post-curing Layup- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
DB EA1 200C 24hrs 925H 13.37 4.04 71.38 1041
0
DB EA2 20 C 24hrs 925H 13.65 4.12 74.67 1201
0
DB EA3 20 C 24hrs 925H 13.23 4.24 81.86 1232
0
DB EA4 20 C 24hrs 925H 11.82 4.15 70.75 1201 74.67 1169
DB EB1 500C 24hrs 925H 12.48 3.95 127.31 1552
DB EB2 500C 24hrs 925H 14.05 3.94 139.18 1635
DB EB3 500C 24hrs 925H 13.57 3.94 132.67 1549
DB EB4 820C 2hrs 925H 12.40 4.28 112.12 1505 127.82 1561
0
DB EC1 82 C 2hrs 925H 13.15 4.11 142.86 1749
0
DB EC2 82 C 2hrs 925H 12.31 4.07 139.94 1597
0
DB EC3 82 C 2hrs 925H 12.22 4.06 138.56 1616
DB EC4 820C 2hrs 925H 13.35 4.24 133.75 1562 138.78 1631

Table B2: DB 925H hardener hand-layup 3-point bend test - L=110 mm

Page | 285
 Hand Avg Avg Flexure Flexure Avg Avg
Fibre No. Post-curing Layup- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
UD-0 FA1 200C 24hrs 925H 11.46 3.88 466.99 4331
0
UD-0 FA2 20 C 24hrs 925H 13.28 3.87 429.48 4179
UD-0 FA3 200C 24hrs 925H 11.34 3.87 402.85 4079
UD-0 FA4 200C 24hrs 925H 13.61 3.77 447.89 4350 436.81 4235
UD-0 FB1 500C 24hrs 925H 11.78 4.08 535.80 4457
UD-0 FB2 500C 24hrs 925H 12.88 3.76 475.40 4122
UD-0 FB3 500C 24hrs 925H 13.68 4.12 563.21 4522
0
UD-0 FB4 82 C 2hrs 925H 15.10 3.88 566.83 4525 535.31 4407
0
UD-0 FC1 82 C 2hrs 925H 14.26 4.02 538.20 4466
UD-0 FC2 820C 2hrs 925H 13.27 3.82 513.48 4728
UD-0 FC3 820C 2hrs 925H 11.83 4.08 579.97 4446
UD-0 FC4 820C 2hrs 925H 12.81 4.12 514.79 4210 536.61 4462

Table B3: UD-00 925H hardener hand-layup 3-point bend test - L=110 mm

VIP 3-point bend test

Vacuum Avg Avg Flexure Flexure Avg Avg

Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
CSM AA1 200C 24hrs 925H 16.06 5.13 285.96 2461
CSM AA2 200C 24hrs 925H 15.44 5.58 263.32 2748
CSM AA3 200C 24hrs 925H 13.99 6.27 260.67 2764
CSM AA4 200C 24hrs 925H 13.42 6.04 254.02 2670 265.99 2668
CSM AB1 50deg 24 hrs 925H 15.25 5.85 284.98 3194
CSM AB2 50deg 24 hrs 925H 15.62 5.46 279.95 2862
CSM AB3 50deg 24 hrs 925H 14.56 6.34 255.08 2844 273.34 2966
0
CSM AC1 82 C 2hrs 925H 15.15 5.23 282.56 2998
CSM AC2 820C 2hrs 925H 13.63 5.44 283.33 2887
CSM AC3 820C 2hrs 925H 14.87 5.48 267.74 3271 277.88 3052

Table B4: CSM 925H hardener VIP 3-point bend test - L=110 mm

Page | 286
 Vacuum Avg Avg Flexure Flexure Avg Avg
Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
DB BA1 200C 24hrs 925H 13.29 3.09 145.79 1825
DB BA2 200C 24hrs 925H 14.11 3.11 154.76 1934
DB BA3 200C 24hrs 925H 13.11 3.13 156.65 1876
DB BA4 200C 24hrs 925H 12.70 3.14 151.26 1889 152.12 1881
DB BB1 500C 24hrs 925H 13.11 3.09 156.54 1927
DB BB2 500C 24hrs 925H 12.95 3.10 152.38 1706
DB BB3 500C 24hrs 925H 12.58 3.16 153.29 1810 153.85 1814
DB BC1 820C 2hrs 925H 14.13 3.12 165.33 1984
DB BC2 820C 2hrs 925H 13.13 3.23 175.17 1899
DB BC3 820C 2hrs 925H 12.92 3.14 172.49 1953 171.07 1945

Table B5: CSM 925H hardener VIP 3-point bend test - L=110 mm

Vacuum Avg Avg Flexure Flexure Avg Avg

Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
CFM A01 200C 24hrs 925H 13.50 5.76 208.75 2504
CFM A02 200C 24hrs 925H 13.71 5.70 186.08 2358
CFM A03 200C 24hrs 925H 13.23 5.89 198.41 2711
CFM A04 200C 24hrs 925H 13.26 5.50 198.81 2753
CFM A05 200C 24hrs 925H 13.23 5.92 183.93 2526 195.20 2570

Table B6: CFM 925H hardener VIP 3-point bend test - L=110 mm

Vacuum Avg Avg Flexure Flexure Avg Avg

Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
CFM B01 200C 24hrs CHM50 13.51 5.77 153.97 2366
CFM B02 200C 24hrs CHM50 14.08 5.75 163.75 2376
CFM B03 200C 24hrs CHM50 14.50 5.88 151.37 2370
CFM B04 200C 24hrs CHM50 13.19 5.63 153.26 2399
CFM B05 200C 24hrs CHM50 14.17 5.76 141.92 2197 152.71 2341

Table B7: CFM CHM50 hardener VIP 3-point bend test - L=110 mm

Page | 287
 Vacuum Avg Avg Flexure Flexure Avg Avg
Fibre No. Post-curing
Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
DB D01 200C 24hrs CHM50 14.41 3.470 162.01 2776
DB D02 200C 24hrs CHM50 13.85 3.390 190.13 2991
DB D03 200C 24hrs CHM50 13.93 3.43 178.56 2540 176.90 2869

Table B8: DB CHM50 hardener VIP 3-point bend test - L=110 mm

Vacuum Avg Avg Flexure Flexure Avg Avg

Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
UD-90 E01 200C 24hrs CHM50 13.25 3.70 110.96 2394
UD-90 E02 200C 24hrs CHM50 13.44 3.69 119.17 2657
UD-90 E03 200C 24hrs CHM50 13.39 3.67 104.61 2605
UD-90 E04 200C 24hrs CHM50 13.07 3.66 105.32 2143
UD-90 E05 200C 24hrs CHM50 13.27 3.66 107.90 2520
UD-90 E06 200C 24hrs CHM50 13.27 3.77 102.67 2486 108.44 2467

Table B9: UD-900 CHM50 hardener VIP 3-point bend test - L=110 mm

Vacuum Avg Avg Flexure Flexure Avg Avg

Fibre No. Post-curing Infused- w t Stress Modulus Stress E
Catalyst (mm) (mm) (MPa) (MPa) (MPa) (MPa)
UD-0 F01 200C 24hrs CHM50 12.37 3.65 588.54 6246
UD-0 F02 200C 24hrs CHM50 12.15 3.64 574.58 7062
UD-0 F03 200C 24hrs CHM50 12.95 3.69 571.85 7058
UD-0 F04 200C 24hrs CHM50 13.18 3.73 573.41 6607
UD-0 F05 200C 24hrs CHM50 13.68 3.63 586.67 6767 579.01 6748

Table B10: UD-900 CHM50 hardener VIP 3-point bend test - L=110 mm

Page | 288
Appendix C – AE Piezo-sensor Calibration

Figure C1: AE sensor calibration certificate 01

Page | 289
Figure C2: AE sensor calibration certificate 02

Page | 290
Appendix D – HL THS - Layups 1 and 2 AE Failure Characterisation – Piezo-sensor

Channel Total (40-45) (45-50) (50-55) (55-60) (60-65) (65-70) (70-75) (75-80) (80-85) (85-90) (90-95) (95-100)
1 234 23 105 55 29 8 6 2 2 0 0 3 1
2 19023 1657 8892 5024 2215 718 257 150 58 29 13 6 4
3 42510 1472 13555 11609 8296 4515 1768 675 330 148 99 28 15
4 11882 878 5737 2922 1405 544 227 101 40 10 9 4 5
Table D1: HL THS Layup 1 four channel amplitude distribution ranges

Failure Time Load Load drop Amplitude Energy (eu) Duration Rise Time Counts (Threshold
Failure Mode
No. (Sec) (N) (N) (dB) (1eu=10-18J) (ms) (μs) Crossings)
1 185.4 13512 607 100 11643 35931 869 3190 Delamination
2 211.7 15005 1743 100 9223 31330 1087 1258 Delamination
3 258.4 16342 947 100 13444 68118 774 6896 Crack progression
4 287.7 16058 246 100 5306 36169 1006 16487 Crack progression
5 300.0 16105 657 100 6644 34362 551 1835 Crack progression
6 320.5 16581 1560 100 9959 115183 1518 7735 Delamination
7 350.0 16238 96 81 3593 147072 410 9266 Delamination
8 353.1 15816 133 92 38213 999999 854 28468 Crack progression
9 357.7 12534 344 76 65535 25061 1127 54040 Crack progression
10 442.5 9207 3117 100 13659 38309 1188 19154 Fibre failure
Table D2: HL THS Layup 1 AE failure characterisation

Page | 291
Figure D1: HL THS Layup 1 piezo acoustic-emission and Load-deflection curve correlation
Page | 292
Channel Total (40-45) (45-50) (50-55) (55-60) (60-65) (65-70) (70-75) (75-80) (80-85) (85-90) (90-95) (95-100)
1 1083 93 594 216 115 40 11 7 1 0 2 0 4
2 47378 17429 26828 2049 657 239 93 48 25 3 1 1 5
3 112119 27884 64391 7500 6113 4024 1558 456 132 34 13 4 10
4 39027 2156 14974 9605 6662 3666 1277 417 180 87 49 14 10
Table D3: HL THS Layup 2 four channel amplitude distribution ranges

Failure Time Load Load drop Amplitude Energy (eu) Duration Rise Time Counts (Threshold
Failure Mode
No. (Sec) (N) (N) (dB) (1eu=10-18J) (ms) (μs) Crossings)
1 317.7 20851 2371 100 17623 24418 18504 715 Matrix cracking
2 374.3 21922 1726 100 19984 100063 19984 4689 Delamination
3 386.2 21047 182 100 4563 26499 679 506 Crack progression
4 413.9 23403 2821 100 20601 75691 841 1210 Crack progression
5 424.2 21232 1669 68 19389 531045 1760 40445 Delamination
6 477.6 24331 41 84 5373 52428 1111 2894 Crack progression
7 519.9 27791 5621 100 5499 20144 843 516 Crack progression
8 554.3 23170 1241 90 1175 15174 604 270 Crack progression
9 636.5 28852 2654 100 10644 10203 894 941 Fibre failure
Table D4: HL THS Layup 2 AE failure characterisation

Page | 293
Figure D2: HL THS Layup 2 piezo acoustic-emission and Load-deflection curve correlation

Page | 294
Appendix E – VIP THS - Layups 5 and 6 AE failure characterisation – Piezo-sensor
Channel Total (40-45) (45-50) (50-55) (55-60) (60-65) (65-70) (70-75) (75-80) (80-85) (85-90) (90-95) (95-100)
1 10385 3116 4715 1467 603 261 98 61 35 19 3 4 3
2 10583 1845 5295 2224 698 279 121 61 27 13 3 2 5
3 36057 3950 13902 8497 5140 2705 1118 456 164 73 27 14 11
4 6245 1268 3084 1035 429 241 123 34 13 7 3 1 7
Table E1: VIP THS Layup 5 piezo-sensor amplitude distribution ranges

Failure Time Load Diff. Amplitude Energy (eu) Duration Rise Time Counts Signal
Load (N) -18 Failure Mode
No. (Sec) (N) (dB) (1eu=10 J) (ms) (μs) Strength
1 240.56 6528 135 92 7030 24.865 1899 474 43917916 Matrix Cracking
2 273.86 6599 258 98 7242 30.301 507 1197 45241540 Delamination
3 375.63 8479 21 91 2870 13.648 507 200 17928736 Crack Progression
4 388.33 8385 300 99 9366 23.082 1082 477 58509396 Crack Progression
5 411.62 8657 25 95 7113 99.999 50171 8886 44432472 Crack Progression
6 435.61 9433 217 99 12118 99.946 12118 5982 75700072 Delamination
7 455.98 9017 75 99 4099 25.137 248 777 25608334 Crack Progression
8 505.92 8458 1513 99 8505 21.298 17408 715 53127464 Delamination
9 563.40 9978 217 98 10777 99.987 8873 2344 67321504 Crack Progression
10 575.89 10156 1888 99 6330 15.983 10304 266 39544772 Delamination
11 610.97 10683 31 99 2425 10.804 97 164 15153498 Crack Progression
12 711.70 19744 9 82 9549 100.00 58774 28180 59652828 Crack Progression
13 738.56 16642 5888 99 24664 99.484 30144 6408 154062450 Fibre failure
Table E2: VIP THS Layup 5 Piezo-sensor acoustic-emission failure characterisation

Page | 295
Page | 296
Figure E1: VIP THS Layup 5 piezo acoustic-emission and Load-deflection curve correlation

Page | 297
Channel Total (40-45) (45-50) (50-55) (55-60) (60-65) (65-70) (70-75) (75-80) (80-85) (85-90) (90-95) (95-100)
1 48408 9160 27764 6367 2750 1292 541 284 131 57 26 20 16
2 9498 1140 3965 2265 1195 528 220 80 47 26 14 6 12
3 7530 1138 3298 1492 790 409 184 99 58 27 18 7 10
4 36730 5112 17252 8478 3317 1199 592 341 208 100 64 31 36
Table E3: VIP THS Layup 6 piezo-sensor amplitude distribution ranges

Failure Time Load Diff. Amplitude Energy (eu) Duration Rise Time Counts
Load (N) Failure Mode
No. (Sec) (N) (dB) (1eu=10-18J) (ms) (μs) (Threshold Crossings)
1 238.75 9210 359 97 10684 99999 25357 2118 Delamination
2 279.43 10680 23 99 7744 35405 1629 981 Crack progression
3 286.61 10937 267 99 23110 99939 8411 3280 Delamination
4 323.26 9709 13 99 13795 81095 3459 2006 Crack progression
5 352.20 10354 78 99 17925 99973 42381 3286 Delamination
6 423.28 12567 20 99 10824 25272 185 492 Crack progression
7 472.09 10127 559 99 10636 77316 59573 3395 Crack progression
8 552.04 13046 2382 99 13130 64497 46188 2156 Delamination
9 659.76 18005 171 99 6459 24603 545 538 Crack progression
10 686.53 20364 11 99 14577 36407 85 896 Crack progression
11 730.64 21661 8712 99 12295 75596 67 5693 Fibre failure
Table E4: VIP THS Layup 6 Piezo-sensor acoustic-emission failure characterisation

Page | 298
Page | 299
Figure E2: VIP THS Layup 6 piezo acoustic-emission and Load-deflection curve correlation

Page | 300
Appendix F – VIP THS AE Failure Characterisation – FBG sensor
Failure Time Amplitude Energy (eu) Duration Rise Time Counts
Load (N)
No. (Sec) (dB) (1eu=10-18J) (ms) (μs)
1 240.56 6528 41 22711 1200 515 50
2 273.86 6599 73 61 19605 6346 832
4 388.33 8385 40 41 19325 9280 266
6 435.61 9433 69 10717 3888 2618 1896
7 455.98 9017 66 2398 23719 2214 204
8 505.92 8458 64 2050 7620 3084 266
9 563.40 9978 41 28 3205 258 1572
10 575.89 10156 50 177 31696 18172 182
11 610.97 10683 54 61 8140 721 114
12 711.70 19744 50 14 15563 10328 223
13 738.56 16642 80 8882 32767 14387 2338
Table F1: VIP THS Layup 5 FBG sensor acoustic-emission failure characterisation

301
Figure F1: VIP THS Layup 5 FBG acoustic-emission and Load-deflection curve correlation

302
Appendix G – Optical fibre embedding procedure with VIP
Technique
1. Apply Mould Release coat 1 on the mould surface.
2. Plan the Layup. On the mould, at least 100 mm should be left free on all sides.
3. Plan for the resin ingress and exit end space allowance, which requires at least 50 mm
from the each end of the mould.
4. Calculate the length and width of specimen for cutting reinforcement fabric. Allow 25
mm on both ends and 2.5 mm per cut (including sanding/polishing).
5. Calculate the dimensions of the fabric for extra CSM layers at bottom bend - Ply drop.
6. Apply Mould Release coat 2 on the mould.
7. Cut Fabric to required size - for Laminate and for ply drop layers.
8. Organise the fibre mats as per the layup including ply-drop layers (Do not use Tack
Spray at this stage)
9. Once the layup is checked, lay the fibre mats on the mould and remember to use Ply-
drop layers as well. Use light tack-spray between the layers to prevent fabric slippage.
10. If two layups are manufactured at a time, then add a bridging material between the
layups over the mould surface to facilitate easy resin flow.
11. Cut Peel ply to size and lay it over the entire laminate, 10 mm extra on all sides - Use
tack-spray. The resin ingress and egress ends should cover the resin inlet pipe.
12. Cut perforated film and place it on top of peel ply - use tack-spray.
13. Cut flow mesh and align over the perforated film. Fix it using adhesive tape.
14. Cut and place flow medium (50 mm wide, white honeycomb structure) at the ingress
side. It should sit on top of peel ply and a bit on the laminate and flow mesh.
15. Fix tacky-tape all around the perimeter of the mould.
16. Cut the resin inlet hose to size and fix it at the ingress end using tacky-tape pieces
over the flow medium.
17. Take a T-joint; connect a hose to the central inlet and the sides to the resin inlet pipe.
18. Cut the resin inlet hose to size and fix it at the egress end using tacky-tape pieces.
19. The resin flow needs to be restricted at the egress end. Take 600 gsm CSM, cut into
pieces, apply tack-spray and fix it over the resin outlet pipe.
20. Connect the egress pipe to the outlet pipe using T-joint.
21. Connect the ingress and egress pipe to the mould using tacky-tape.

303
22. Cut the Nylon sheet (for vacuum bagging) to size; fix it on the mould with the glued
tacky-tape slowly. Make pleats at corners for efficient circulation of resin.
23. Secure the nylon sheet applying thumb pressure all along the length of the tacky-
tape.
24. Close the inlet pipe using a clamp. Egress pipe will be in catch pot.
25. Switch on the vacuum slowly and align the nylon cover, fibre mat, and everything in
the nylon bag to be secured on the mould properly. Increase the suction up to 100
kPa.
26. Conduct drop test, check to see if the vacuum pressure drops from 100 kPa, which
indicates an air leak. Use acoustic-emission headphones and fix all the leaks.
27. Estimate the amount of resin (Vinylester SPV 6036) and hardener (CHM 50/925H -
1.5% of resin).
28. Mix resin and hardener thoroughly in an appropriate container. The estimated gel
time is around 30 min.
29. Place the container near the mould and immerse ingress pipe end in the container.
30. Re-check the vacuum strength. Slowly release the clamp and allow resin to infuse into
the laminate slowly.
31. Observe the generation of air bubbles in the ingress pipe as well as in the laminate
(resin front).
32. Keep an eye on the resin container; the ingress pipe should always be immersed in the
container. If there is a shortage of resin, stop the flow using the clamp, prepare new
resin-hardener mix and place it carefully under the inlet pipe.
33. Allow the whole laminate to infuse, frequently checking for any leaks and fix the leaks
by patching with tacky-tape.
34. Disconnect the resin supply after successful infusion and seal the ingress and egress
hose.
35. Keep the mould under vacuum overnight.
36. Disconnect all the pipes, remove nylon bag, flow medium, perforated film along with
peel ply and slowly release the laminate from the mould.

304
Appendix H – Layup 5 and Layup 6 FBG strain results

Figure H1: VIP THS Layup 5 FBG strain-sensor measurement—right outer

Figure H2: VIP THS Layup 5 Bragg wavelength shift during different phases of the test procedure

305
Figure H3: VIP THS Layup 6 FBG strain-sensor measurement—Left outer

Figure H4: VIP THS Layup 6 FBG strain-sensor measurement—Right outer

306
 Figure H5: VIP THS Layup 6 FBG strain-sensor measurement—Right Inner

Figure H6: VIP THS Layup 5 Bragg wavelength shift during different phases of the test procedure

307