You are on page 1of 249

The Pennsylvania State University

The Graduate School



Department of Engineering Science and Mechanics
CHARACTERIZATION OF INTERLAMINAR FRACTURE
TOUGHNESS OF A CARBON/EPOXY COMPOSITE MATERIAL

A Thesis in

Engineering Mechanics

by

Ye Zhu
2009 Ye Zhu
Submitted in Partial Fulfillment
of the Requirements
for the Degree of
Master of Science


May 2009



The thesis of Ye Zhu was reviewed and approved* by the following:

Charles E. Bakis
Distinguished Professor of Engineering Science and Mechanics
Thesis Advisor

George A. Lesieutre
Professor of Aerospace Engineering
Head of the Department of Aerospace Engineering

Kelvin L. Koudela
Senior Research Associate
Applied Research Laboratory

Judith A. Todd
Professor of Engineering Science and Mechanics
P. B. Breneman Department Head Chair
Head of the Department of Engineering Science and Mechanics

*Signatures are on file in the Graduate School

iii

ABSTRACT
The primary objective of this investigation is to characterize the Mode I, Mode II,
and mixed Mode I/II interlaminar fracture of a proprietary carbon/epoxy composite
material system. A state-of-the-art review of the literature on quasi-static and cyclic test
methods for interlaminar fracture testing is given. The Mode I, Mode II, and mixed
Mode I/II interlaminar fracture behavior of the carbon/epoxy laminated material in quasi-
static and fatigue loadings was investigated using the double-cantilever-beam (DCB)
specimen, the end-notched flexure (ENF) specimen, and the single leg bending (SLB)
specimen, respectively. It was found that the Mode I interlaminar fracture toughness at
crack onset (G
Ic
) was low for the investigated material system in comparison to results
reported in the literature for carbon/brittle epoxy material system. In addition, the Mode I
fracture toughness increased by about 40% after 50 mm crack extension. The Mode II
quasi-static tests were conducted with precracked and un-precracked specimens.
Compared to results reported in the literature, the Mode II fracture toughnesses (G
IIc
) of
the investigated material were in the common range for carbon fiber composites made
with brittle epoxies. The G
IIc
value of an un-precracked specimen was 44% -60% higher
than that of a precracked specimen. The mixed-mode fracture toughness (G
Tc
) was found
to be low in comparison to the results in the literature and it increased by 11 to 53% after
20 mm crack extension. For all fatigue tests, the modified Paris law was used to fit the
experimentally determined crack growth rate per cycle (da/dN) versus the applied
maximum strain energy release rate (SERR, G
max
). The delamination growth rate
decreased rapidly with decreasing applied SERR, which gave rise to high exponents of
the Modified Paris law for Mode I, Mode II, and mixed Mode I/II fatigue tests, with the
highest in mixed Mode I/II. To assess the capability of commercial finite element
software in solving delamination growth problems, a crack propagation analysis of the
DCB specimen was carried out using the virtual crack closure technique (VCCT) for
Abaqus and Abaqus/Standard V6.7. Preliminary results showed good agreement of load
versus displacement behavior between the finite element analysis (FEA) and
iv

experimental results. However, the crack front shape predicted by FEA did not agree well
with experimental results.



















NAVAIR Public Release 08-1184
Distribution: Statement A Approved for public release; distribution is unlimited
v

TABLE OF CONTENTS
LIST OF FIGURES ..................................................................................................... ix
LIST OF TABLES....................................................................................................... xvii
ACKNOWLEDGEMENTS......................................................................................... xix
Chapter 1 Introduction................................................................................................ 1
1.1 Background.............................................................................................. 1
1.2 Introduction to Fracture Mechanics of Composite Materials.................. 2
1.3 Problem Statement and Research Objectives .......................................... 6
Chapter 2 Literature Review....................................................................................... 9
2.1 Mode I Interlaminar Fracture Toughness (IFT) Testing ......................... 9
2.1.1 Geometry and analysis of the Double Cantilever Beam (DCB)
specimen.......................................................................................... 9
2.1.1.1 Modified Beam Theory (MBT) method............................... 11
2.1.1.2 Compliance Calibration (CC) method (Berrys Method)... 14
2.1.1.3 Modified Compliance Calibration (MCC) method............. 15
2.1.1.4 Elastic Foundation Model (EFM) method.......................... 16
2.1.2 Experimental aspects of the DCB test........................................... 18
2.1.2.1 Initial defect type................................................................. 18
2.1.2.2 Definition of critical point for crack onset.......................... 21
2.1.2.3 Method of loading/unloading.............................................. 23
2.1.2.4 Crack resistance curve and fiber bridging........................... 24
2.2 Mode II Interlaminar Fracture Toughness (IFT) Testing........................ 26
2.2.1 Geometry and analysis of the End Notched Flexure (ENF)
specimen........................................................................................ 26
2.2.1.1 Classical Plate Theory (CPT) (Davidson et al. 1996) ........ 28
2.2.1.2 Beam Theory (BT) with shear correction (Carlsson et al.
1986) ........................................................................................ 31
2.2.1.3 Compliance Calibration (CC) method (Davidson et al.
1996) ........................................................................................ 33
2.2.2 Experimental aspects of the ENF test............................................ 34
2.2.3 Other test configurations of Mode II IFT testing .......................... 36
2.3 The Mixed Mode I/II Interlaminar Fracture Toughness Testing............. 38
2.3.1 Geometry and analysis of Single Leg Bending (SLB) specimen.. 39
2.3.1.1 Classical Plate Theory (CPT) Method (Davidson and
Sundararaman 1996) ................................................................ 40
2.3.1.2 Beam Theory based analyses .............................................. 42
vi

2.3.1.3 Compliance Calibration (CC) method (Polaha et al.
1996) ........................................................................................ 43
2.3.2 Experimental aspects of the SLB test............................................ 44
2.3.3 Other configurations of mixed Mode I/II IFT test ........................ 44
2.3.4 Mixed mode delamination failure criterion................................... 46
2.4 Interlaminar Fracture Toughness Test Under Cyclic loading ................. 47
2.4.1 Fatigue delamination growth models ............................................ 51
2.4.1.1 Pure Mode I or II................................................................. 51
2.4.1.2 Mixed-mode........................................................................ 52
2.4.2 IFT fatigue test methods................................................................ 53
2.4.2.1 Fatigue threshold strain energy release rate
determination ........................................................................... 53
2.4.2.2 Fatigue delamination growth test ........................................ 54
2.5 Finite Element Modeling of Crack Propagation...................................... 55
2.5.1 The Virtual Crack Closure Technique (VCCT) ............................ 56
2.5.1.1 The virtual crack closure technique formulation ................ 56
2.5.1.2 Crack growth criterion for VCCT....................................... 60
2.5.2 Cohesive element (Abaqus 2007) ................................................. 62
2.5.2.1 Elastic behavior of the cohesive element ............................ 63
2.5.2.2 Damage initiation criteria of the cohesive element............. 65
2.5.2.3 Damage evolution criteria of the cohesive element ............ 65
2.6 Preview of the Following Chapters ......................................................... 66
Chapter 3 The Mode I Interlaminar Fracture Toughness Testing............................... 67
3.1 Material, Specimen and Test Configuration............................................ 67
3.2 The Mode I Quasi-static IFT Testing ...................................................... 69
3.2.1 Mode I quasi-static test method .................................................... 70
3.2.2 Mode I quasi-static test results ...................................................... 72
3.2.2.1 Load-displacement curves................................................... 73
3.2.2.2 Compliance calibration ....................................................... 78
3.2.2.3 G
Ic
onset values ................................................................... 81
3.2.2.4 Mode I resistance curve ...................................................... 84
3.2.2.5 Discussion of special issues ................................................ 85
3.3 The Mode I Fatigue IFT Testing ............................................................. 92
3.3.1 Mode I fatigue test method............................................................ 92
3.3.2 Mode I fatigue test results ............................................................. 95
3.3.2.1 Compliance calibration ....................................................... 95
3.3.2.2 Crack growth....................................................................... 98
3.3.2.3 Crack growth rate (da/dN) vs. maximum SERR (G
Imax
)
plots.......................................................................................... 99
Chapter 4 The Mode II Interlaminar Fracture Toughness Testing ............................. 102
vii

4.1 Material, Specimen and Test Configuration............................................ 102
4.2 The Mode II Quasi-static IFT Testing..................................................... 107
4.2.1 Mode II quasi-static test method ................................................... 107
4.2.2 Mode II quasi-static test results..................................................... 111
4.2.2.1 Load-displacement curves................................................... 111
4.2.2.2 Compliance calibration ....................................................... 114
4.2.2.3 G
IIc
onset values .................................................................. 120
4.2.2.4 A short Mode II fracture resistance curve........................... 122
4.2.2.5 Special issues on the ENF test ............................................ 126
4.3 The Mode II Fatigue IFT Testing............................................................ 129
4.3.1 Mode II fatigue test method .......................................................... 129
4.3.2 Mode II fatigue test results............................................................ 132
4.3.2.1 Crack growth....................................................................... 132
4.3.2.2 da/dN - G
IImax
plots.............................................................. 138
Chapter 5 Mixed-mode I/II Interlaminar Fracture Toughness Testing....................... 141
5.1 Material, Specimen and Test Configuration............................................ 141
5.2 The Mixed-mode I/II Quasi-static IFT Testing....................................... 144
5.2.1 Mixed-mode I/II quasi-static test method ..................................... 145
5.2.2 Mixed-mode I/II test results .......................................................... 147
5.2.2.1 Load-displacement curves................................................... 147
5.2.2.2 Compliance calibration ....................................................... 149
5.2.2.3 Mixed-mode I/II critical strain energy release rate (G
Tc
)
onset value ............................................................................... 152
5.2.2.4 Mixed-mode fracture resistance curve................................ 156
5.3 Mixed-mode I/II Fatigue IFT Testing ..................................................... 158
5.3.1 Mixed-mode fatigue test method................................................... 158
5.3.2 Mixed-mode fatigue test results .................................................... 160
5.3.2.1 Crack growth....................................................................... 160
5.3.2.2 Crack growth rate (da/dN) vs. maximum SERR (G
Tmax
)
plots.......................................................................................... 162
Chapter 6 Finite Element Modeling of Crack Propagation in DCB Specimens......... 165
6.1 Two-dimensional Modeling of the DCB Specimen ................................ 166
6.1.1 Geometry, loading and boundary conditions of 2D models.......... 166
6.1.2 Modeling techniques for 2D models ............................................. 166
6.1.3 Meshing of 2D models .................................................................. 170
6.1.4 Results of 2D models .................................................................... 171
6.1.4.1 Crack propagation............................................................... 171
6.1.4.2 Load-displacement curve .................................................... 173
6.1.4.3 Stress distribution................................................................ 174
6.2 Three-dimensional Modeling of the DCB Specimen .............................. 176
viii

6.2.1 Geometry, loading and boundary conditions of 3D models.......... 177
6.2.2 Modeling techniques for 3D models ............................................. 177
6.2.3 Meshing of 3D models .................................................................. 181
6.2.4 Results of 3D Models.................................................................... 183
6.2.4.1 Crack front shape observation............................................. 183
6.2.4.2 Load vs. displacement curve behavior ................................ 188
6.2.4.3 Stress distribution................................................................ 190
6.3 Conclusions ............................................................................................. 191
Chapter 7 Conclusions and Recommendations........................................................... 193
7.1 Mode I Interlaminar Fracture Toughness Characterization..................... 193
7.2 Mode II Interlaminar Fracture Toughness Characterization ................... 194
7.3 Mixed Mode I/II Interlaminar Fracture Toughness Characterization ..... 195
7.4 Preliminary Finite Element Modeling Results ........................................ 196
Bibliography ................................................................................................................ 198
Appendix A IFT Test Results From Literature........................................................... 207
Appendix B Additional Specimen Information and Test Results............................... 213
1. Dimensions of Specimens ......................................................................... 213
2. Mode I Quasi-static Fracture Toughness at Crack Onset .......................... 216
3. Mode I Fatigue Crack Growth Rate vs. Maximum SERR Plots ............... 217
4. Mode II Quasi-static Fracture Toughness at Crack Onset......................... 221
5. Mode II Fatigue Crack Growth Rate vs. Maximum SERR Plots.............. 222
6. Mixed Mode I/II Quasi-static Fracture Toughness at Crack Onset........... 225
7. Mode II Fatigue Crack Growth Rate vs. Maximum SERR Plots.............. 226
Appendix C Non-Technical Abstract.......................................................................... 229

ix

LIST OF FIGURES
Fig. 1-1: Load and displacement on a crack body ....................................................... 4
Fig. 1-2: The basic fracture modes. ............................................................................. 5
Fig. 2-1: Geometry of the ASTM D5528 Double Cantilever Beam (DCB)
specimen ............................................................................................................... 10
Fig. 2-2: A schematic of the DCB specimen (side view) ............................................ 12
Fig. 2-3: Determination of in the modified beam method (MBT) ........................... 13
Fig. 2-4: Determination of n in compliance calibration method.................................. 15
Fig. 2-5: Determination of
1
in the Modified Compliance Calibration (MCC)
method .................................................................................................................. 16
Fig. 2-6: An elastic foundation model of the DCB specimen, based on (Ozdil and
Carlsson 1999) ...................................................................................................... 17
Fig. 2-7: A schematic of microscopic view of longitudinal section of specimen
near the end of starter film.................................................................................... 20
Fig. 2-8: Crack initiation definitions............................................................................ 21
Fig. 2-9: Typical load-displacement curves for a DCB specimen with multiple
loading/unloading cycles ...................................................................................... 23
Fig. 2-10: Typical delamination resistance curve (R curve) from a DCB test
(ASTM D5528-01 2002) ...................................................................................... 24
Fig. 2-11: Picture showing fiber bridging (Mode I loading) ....................................... 25
Fig. 2-12: End-notched flexure test schematic ............................................................ 27
Fig. 2-13: A schematic of the ENF specimen (side view) ........................................... 29
Fig. 2-14: Schematic of an ENF specimen subject to three-point bending ................. 32
Fig. 2-15: Typical load-displacement curves of an ENF test....................................... 34
Fig. 2-16: ELS test configuration, based on (O'Brien 1998a) .................................... 37
Fig. 2-17: The SLB specimen geometry...................................................................... 39
x

Fig. 2-18: A schematic of the SLB specimen geometry and detailed notation............ 40
Fig. 2-19: A schematic of MMB test configuration (Kim and Mayer 2003)............... 45
Fig. 2-20: A schematic of the mixed mode end load split (MMELS) specimen,
based on (Szekrnyes and Jzsef 2006)................................................................ 45
Fig. 2-21: A schematic of the crack lap shear (CLS) specimen, based on (Tracy et
al. 2003) ................................................................................................................ 45
Fig. 2-22: A total fatigue life model of composite materials, based on
(Shivakumar et al. 2006)....................................................................................... 49
Fig. 2-23: The Mode I delamination onset SERR versus number of cycles, based
on (ASTM standard D6115-97 1997(R2004)) ..................................................... 54
Fig. 2-24: Crack closure for VCCT, based on (Krueger 2002) ................................... 57
Fig. 2-25: VCCT for four-node 2D element (plane strain or plane stress), based on
(Krueger 2002)...................................................................................................... 58
Fig. 2-26: VCCT for eight-node solid element (3D view) (Krueger 2002)................. 59
Fig. 2-27: VCCT for eight-node solid element (top view) (Krueger 2002)................. 60
Fig. 2-28: Traction-separation response of cohesive element, based on (Abaqus
2007) ..................................................................................................................... 63
Fig. 3-1: Diagram of the DCB panel............................................................................ 68
Fig. 3-2: DCB specimen geometry and notation ......................................................... 69
Fig. 3-3: Photograph of DCB test set up...................................................................... 70
Fig. 3-4: Constant fracture toughness after certain length of crack extension............. 74
Fig. 3-5: Load vs. displacement plot (Specimen 1-1).................................................. 75
Fig. 3-6: Load vs. displacement plot (Specimen 1-4).................................................. 76
Fig. 3-7: A C
1/3
versus a plot for Mode I tests............................................................. 79
Fig. 3-8: An a/2h versus (bC)
1/3
plot for Mode I DCB tests........................................ 80
Fig. 3-9: Load vs. displacement curve -- small load drop occurred at crack onset
(Specimen 1-1)...................................................................................................... 82
xi

Fig. 3-10: Load vs. displacement curve -- large load drop occurred at crack onset
(Specimen 1-3)...................................................................................................... 82
Fig. 3-11: G
Ic
onset values ........................................................................................... 83
Fig. 3-12: Mode I IFT resistance curves for Specimen 1-2......................................... 84
Fig. 3-13: Overall Mode I IFT resistance curve for five DCB specimens, by MCC
method .................................................................................................................. 85
Fig. 3-14: Crack surfaces of a quasi-static DCB test specimen................................... 87
Fig. 3-15: A schematic of crack propagation............................................................... 88
Fig. 3-16: Fracture resistance curve for Specimen 1-4, with crack length
calculated by compliance calibration by MCC method........................................ 90
Fig. 3-17: Fiber bridging observed through long distance microscope ....................... 91
Fig. 3-18: A schematic showing the loading and unloading procedures for the
Mode I precrack test ............................................................................................. 93
Fig. 3-19: Mode I maximum SERR reduction as crack grows for a displacement
controlled DCB fatigue test .................................................................................. 95
Fig. 3-20: Crack growth by different methods (Specimen 4-2)................................... 97
Fig. 3-21: Crack growth by various methods............................................................... 99
Fig. 3-22: da/dN vs. G
Imax
plots for four DCB fatigue specimens (with crack
growth by visual measurement)............................................................................ 100
Fig. 3-23: da/dN vs. G
Imax
plots for four DCB specimens (with crack growth
calculated by compliance calibration) .................................................................. 101
Fig. 4-1: The ENF and SLB panel diagram (Panel B)................................................. 103
Fig. 4-2: The ENF panel diagram (Panel AA)............................................................. 104
Fig. 4-3: A schematic of ENF specimen geometry and test configuration.................. 105
Fig. 4-4: Schematic of ENF un-precracked test configuration A................................ 106
Fig. 4-5: Schematic of ENF precracked test configuration B...................................... 106
xii

Fig. 4-6: A photograph of the ENF test setup with un-precracked test
configuration......................................................................................................... 108
Fig. 4-7: A photograph of the ENF test setup with precracked test configuration ...... 109
Fig. 4-8: Markings on ENF specimen edge for compliance calibration...................... 110
Fig. 4-9: A representative load vs. displacement plot for ENF crack onset test
(Specimen 2-2, un-precracked, a
0
/L 0.5)........................................................... 111
Fig. 4-10: A load vs. displacement plot for ENF crack onset test (Specimen 2-5,
precracked, test configuration B).......................................................................... 112
Fig. 4-11: A quasi-stable load vs. displacement plot for ENF crack onset test
(Specimen 2-6, precracked) .................................................................................. 113
Fig. 4-12: Construction method for the 5% compliance offset line............................. 114
Fig. 4-14: C(8bh
3
) vs. a
3
plots for all ENF specimens................................................. 116
Fig. 4-15: A C vs. a plot for compliance calibration, by Eq. (2.39) ............................ 118
Fig. 4-16: G
IIc
onset values .......................................................................................... 122
Fig. 4-17: Mode II resistance curve (Specimen 2-6, precracked)................................ 125
Fig. 4-18: A load vs. displacement plot for ENF test with constant G curves
shown.................................................................................................................... 126
Fig. 4-19: A schematic of load vs. displacement curves for different states of
crack growth ......................................................................................................... 127
Fig. 4-20: Plot of crack growth rate in a quasi-static ENF test versus normalized
crack length........................................................................................................... 129
Fig. 4-21: Maximum Mode II SERR (G
IImax
) vs. normalized crack length plot for
an ENF test with fixed displacement amplitude (F
1
is a factor related to initial
crack length and maximum opening displacement) ............................................. 131
Fig. 4-22: Crack growth for ENF fatigue specimens................................................... 133
Fig. 4-23: A microscopic view of a specimen edge while the specimen was in
different loadings (Specimen 5-3) ........................................................................ 134
Fig. 4-24: Photographs of fracture surfaces................................................................. 137
xiii

Fig. 4-25: Crack growth rate against Mode II maximum SERR plot, with crack
growth calculated by compliance calibration ....................................................... 139
Fig. 4-26: Crack growth rate against Mode II maximum SERR plot, with crack
growth measured visually..................................................................................... 140
Fig. 5-1: The SLB panel diagram................................................................................ 142
Fig. 5-2: A schematic of SLB specimen geometry and test configuration.................. 143
Fig. 5-3: SLB test configuration .................................................................................. 144
Fig. 5-4: A photograph of SLB test set up................................................................... 145
Fig. 5-5: Markings on SLB specimen edge ................................................................ 146
Fig. 5-6: Load vs. displacement curve for SLB quasi-static tests................................ 147
Fig. 5-7: Load-displacement plot near the critical onset point (Specimen 3-2)........... 148
Fig. 5-8: C vs. a plot for all SLB specimens............................................................... 150
Fig. 5-9: C(8bh
3
) vs. a
3
plot for all SLB specimens ................................................... 150
Fig. 5-10: G
Tc
values determined from SLB tests........................................................ 153
Fig. 5-11: The Reeders linear mixed mode failure locus and test data ...................... 155
Fig. 5-12: The B-K Law failure locus and test data..................................................... 155
Fig. 5-13: Load vs. displacement plot for Specimen 3-4 with constant G
TR
curves
shown.................................................................................................................... 157
Fig. 5-14: Fracture resistance curves for SLB specimens............................................ 158
Fig. 5-15: Crack growth of SLB fatigue specimens .................................................... 160
Fig. 5-16: A sketch of opening crack........................................................................... 161
Fig. 5-17: Fracture surfaces of a SLB specimen.......................................................... 162
Fig. 5-18: A da/dN vs. G
max
plot for SLB specimens (crack length by compliance
calibration method)............................................................................................... 163
Fig. 5-19: A da/dN vs. G
max
plot for SLB specimens (crack length by visual
measurement)........................................................................................................ 164
xiv

Fig. 6-1: Finite element analysis and the IFT test........................................................ 165
Fig. 6-2: Geometry and boundary conditions of 2-dimensional DCB models ............ 166
Fig. 6-3: Idealized Mode I fracture toughness with crack extension for Specimen
1-2......................................................................................................................... 168
Fig. 6-4: Mesh configuration for 2D Model #1 and #3................................................ 171
Fig. 6-5: Mesh configuration for 2D Model #2 ........................................................... 171
Fig. 6-6: Crack growth versus opening displacement from test data and 2D finite
element modeling.................................................................................................. 172
Fig. 6-7: Load vs. displacement curves from test data and 2D finite element
modeling ............................................................................................................... 173
Fig. 6-8: Contour plot of the stress in the longitudinal direction (
xx
), in MPa,
around the crack tip (2D Model #2) ..................................................................... 175
Fig. 6-9: Contour plot of the stress in the thickness direction (
22
), in MPa, around
the crack tip (2D Model #2).................................................................................. 176
Fig. 6-10: Geometry of 3D models of the DCB specimen .......................................... 177
Fig. 6-11: Mesh configuration for 3D Model #1 ......................................................... 181
Fig. 6-12: Mesh configuration for 3D Model #2 ......................................................... 182
Fig. 6-13: Mesh configuration for 3D Model #3 ......................................................... 182
Fig. 6-14: Crack surfaces of a quasi-static DCB test specimen................................... 183
Fig. 6-14: Crack fronts predicted by 3D Model #1...................................................... 184
Fig. 6-15: Crack fronts predicted by 3D Model #3...................................................... 185
Fig. 6-16: Crack fronts predicted by 3D Model #2...................................................... 187
Fig. 6-17: Load vs. displacement curves from 2D and 3D finite element analyses .... 189
Fig. 6-18: Comparison of damping energy to total strain energy for 3D models........ 190
Fig. 6-19: Distribution of longitudinal stress (
11
) in a 3D DCB specimen ................ 191
Fig. A-1: Lay-up of specimens, based on(Polaha et al. 1996)..................................... 209
xv

Fig. B-1: da/dN - G
Imax
plot for Specimen 4-2 (Crack growth measured visually.) .... 217
Fig. B-2: da/dN - G
Imax
plot for Specimen 4-2 (Crack growth calculated by
compliance calibration.) ....................................................................................... 217
Fig. B-3: da/dN - G
Imax
plot for Specimen 4-3 (Crack growth measured visually.) .... 218
Fig. B-4: da/dN - G
Imax
plot for Specimen 4-3 (Crack growth calculated by
compliance calibration.) ....................................................................................... 218
Fig. B-5: da/dN - G
Imax
plot for Specimen 4-4 (Crack growth measured visually.) .... 219
Fig. B-6: da/dN - G
Imax
plot for Specimen 4-4 (Crack growth calculated by
compliance calibration.) ....................................................................................... 219
Fig. B-7: da/dN - G
Imax
plot for Specimen 4-5 (Crack growth measured visually.) .... 220
Fig. B-8: da/dN - G
Imax
plot for Specimen 4-5 (Crack growth calculated by
compliance calibration.) ....................................................................................... 220
Fig. B-9: da/dN - G
IImax
plot for Specimen 5-3 (Crack growth rate was calculated
with crack growth by compliance calibration. The points excluded from
linear fit are near the crack growth arrest domain.).............................................. 222
Fig. B-10: da/dN - G
IImax
plot for Specimen 5-3 (Crack growth rate was calculated
with crack growth by visual measurement. The points excluded from linear
fit are taken near the beginning of the test.) ......................................................... 222
Fig. B-11: da/dN - G
IImax
plot for Specimen 5-4 (Crack growth rate was calculated
with crack growth by compliance calibration. The points excluded from
linear fit are taken near the beginning of the test.) ............................................... 223
Fig. B-12: da/dN - G
IImax
plot for Specimen 5-5 (Crack growth rate was calculated
with crack growth by compliance calibration. The points excluded from
linear fit are taken near the beginning of the test.) ............................................... 224
Fig. B-13: da/dN - G
IImax
plot for Specimen 5-5 (Crack growth rate was calculated
with crack growth by visual measurement. The points excluded from linear
fit are taken near the beginning of the test.) ......................................................... 224
Fig. B-14: da/dN - G
max
plot for Specimen 6-1 (Crack growth was calculated by
compliance calibration.) ....................................................................................... 226
Fig. B-15: da/dN-G
max
plot for Specimen 6-1 (Crack growth was measured
visually.) ............................................................................................................... 226
xvi

Fig. B-16: da/dN-G
max
plot for Specimen 6-2 (Crack growth was calculated by
compliance calibration.) ....................................................................................... 227
Fig. B-17: da/dN - G
max
plot for Specimen 6-2 (Crack growth was measured
visually.) ............................................................................................................... 227
Fig. B-18: da/dN-G
max
plot for Specimen 6-3 (Crack growth was calculated by
compliance calibration.) ....................................................................................... 228
Fig. B-19: da/dN-G
max
plot for Specimen 6-3 (Crack growth was measured
visually.) ............................................................................................................... 228


xvii

LIST OF TABLES
Table 2-1: DCB specimen dimensions required in ASTM D5528.............................. 10
Table 2-2: Some specific ENF specimen geometries from the literature .................... 27
Table 3-1: Parameters of compliance calibration by the MBT method....................... 80
Table 3-2: Parameters of compliance calibration by MCC method ............................ 80
Table 3-3: Mode I Fatigue test parameters .................................................................. 94
Table 3-4: Crack growth prediction approaches for DCB fatigue tests....................... 96
Table 4-1: Dimensions for ENF configuration A (un-precracked specimen).............. 106
Table 4-2: Dimensions for ENF test configuration B (precracked specimen)............ 107
Table 4-3: Parameters A and B, determined by CC 1) for ENF specimens ................. 117
Table 4-4: Parameters C
0
, C
1
, C
2
, and C
3
, determined by CC 2) for ENF
specimens.............................................................................................................. 119
Table 4-5: Testing parameters for ENF fatigue tests................................................... 130
Table 5-1: SLB test configuration dimensions ............................................................ 144
Table 5-2: Coefficients determined for compliance calibration by Eq. (5.1) ............. 151
Table 5-3: Coefficients determined for compliance calibration by Eq. (5.2) .............. 152
Table 5-4: A summary of Mode I, Mode II, and Mixed-mode I/II fracture
toughness at crack onset ....................................................................................... 154
Table 5-5: Testing parameters for SLB fatigue tests ................................................... 159
Table A-1: DCB tests results from literature ............................................................... 207
Table A-2: ENF test results from literature ................................................................ 210
Table A-3: Fatigue test results from literature............................................................ 211
Table B-1: Dimensions of DCB specimens used for quasi-static tests........................ 213
Table B-2: Dimensions of DCB specimens used for fatigue tests............................... 213
xviii

Table B-3: Dimensions of ENF specimen (un-precracked, tested in quasi-static
tests)...................................................................................................................... 214
Table B-4: Dimensions of ENF specimen (precracked, tested in quasi-static tests) ... 214
Table B-5: Dimensions of ENF specimen (precracked, tested in fatigue tests) .......... 214
Table B-6: Dimension of SLB quasi-static specimens ................................................ 215
Table B-7: Dimension of SLB fatigue specimens ....................................................... 215
Table B-8: A summary of Mode I quasi-static test results .......................................... 216
Table B-9: A summary of Mode II quasi-static test results......................................... 221
Table B-10: A summary of Mode I quasi-static test results ........................................ 225


xix

ACKNOWLEDGEMENTS
I would like to express my sincere gratitude to my advisor, Professor Charles E.
Bakis, for his invaluable guidance and encouragement along the course of my study at
Pennsylvania State University. Also, I would like to thank my lab colleagues for
assistance.
I am also grateful to Dr. Lesieutre and Dr. Koudela for serving on my committee
and taking the time to read this thesis and provide helpful advice.
Sincere thanks goes to my parents for their support and encouragement during my
study in Pennsylvania State University.
The financial support of this work is provided by Rhombus Consultants Group. I
would like to thank them for providing me the opportunity to work on this project.
1


Chapter 1

Introduction
1.1 Background
Composite materials are finding increased applications in many engineering fields.
In the aerospace industry, the use of composite materials in commercial and military
aircrafts has increased greatly over the last 20 years. For example, the usage of
composites has evolved from less than 5 percent of the structural weight in the Boeing
737 and 747 to about 50 percent in the Boeing 787 Dreamliner. By contrast, aluminum
will comprise only 12 percent of the Boeing 787 aircraft. According to Chambers (2003),
while the use of composites is less than 10% of the structural weight in the F14 fighter it
has increased to about 40% of the structural weight in the F22 fighter. In the ship-
building industry, thick-section glass and carbon fiber composites and sandwich
composites are more widely incorporated into ship structures than before to fulfill special
demands, such as light-weight, good insulation, low maintenance cost, and resistance to
corrosion (Daniel and Ishai 2006). In civil structures, such as bridges, the use of carbon
fiber-reinforced plastics (CFRP) has extended from only internal reinforcement in
structures to both internal and external reinforcement. In addition to structures, wide
applications of composite materials can be found in automobile parts and frames, trucks,
sports equipments, etc. Among these composite materials, the laminated fiber-reinforced
composite material is becoming commonplace in primary load bearing members of
structures and machines as a high performance material. Compared to metallic materials,
laminated fiber-reinforced materials can provide not only the primary advantage of high
strength to weight ratio, but also offer extra benefits of low coefficient of thermal
expansion (CTE), good resistance to corrosion, low maintenance cost, and low pollution.
These advantages together make the laminated fiber-reinforced material an attractive
candidate for modern structure and machine design.
2

Despite all their advantages, laminated fiber-reinforced composite materials have
certain disadvantages as well. The basic building block of a laminated fiber-reinforced
material is a lamina (ply). Within a lamina, high-strength fibers are combined with a
light-weight matrix. By selecting a bonded sequence of laminae with various orientations
of principal material directions and/or different materials, a wide range of mechanical
properties of laminated composites can be designed according to needs. With this special
material design methodology, the material anisotropy and heterogeneity are greatly
increased. As a result, many issues that do not exist for isotropic and homogeneous
materials arise when using laminated composites. One of the more prominent issues is the
large number of potentially interacting damage modes, such as fiber breakage,
intralaminar matrix cracking, fiber/matrix debonding, fiber pull-out, and delamination
(Daniel and Ishai 2006). Among these damage modes, delamination is one of the most
important and least understood. Delamination is especially important because it can cause
a catastrophic loss of compressive strength. Small delaminations cannot always be
detected by nondestructive inspections, and can potentially grow to unstable
configurations due to in-service loads. In some cases, the structure or machine can fail
catastrophically without any external warning signs. This failure scenario makes
delamination a major obstacle to wider utilization of advanced composite materials in
structures and machinery. Therefore, it is crucial to develop a better understanding of
delamination.
1.2 Introduction to Fracture Mechanics of Composite Materials
The linear elastic fracture mechanics (LEFM) approach has become a generally
accepted practice for the characterization of laminated composites behavior. The LEFM
approach was first developed for brittle homogeneous materials, such as certain types of
steel and ceramics, which exhibit no or small scale plastic deformation before fracture.
However, some of the theories within LEFM have been found to be applicable to predict
delamination in laminated composite materials.
3

From the LEFM approach, delamination growth kinetics are predicted by
comparing the crack driving force or energy release rate (ERR), G, to a critical value of
ERR, G
c
. The crack driving force (G) is related to the loadings, geometry of the crack
body, and constraints, while the critical value is a property of the material, which is called
fracture toughness.
The energy approach for fracture was proposed by Irwin in 1956 (Irwin 1956).
The energy release rate (ERR, G) is a measure of the energy available for an increment of
crack extension. Generally, in mathematical form it is defined by Eq. (1.1), (Anderson
2005).
The potential of an elastic body, , is defined by Eq. (1.2).
where U is the strain energy stored in the body, F is the work done by external forces,
and A is area of crack surface.
Another expression for ERR derived from the definition and commonly used for
fracture toughness test specimens is given by Eq. (1.3) (Anderson 2005).
where C=u/P is the compliance of the crack body; b is width of the body, and a is the
crack length, as shown in Figure 1-1.
d
G
dA

= (1.1)
U F = (1.2)
2
2
P dC
G
b da
= (1.3)
4

For a delamination growing under a constant displacement, which is usually the
case for a displacement controlled interlaminar fracture toughness (IFT) test, no work is
done by external forces and the energy released is only from the elastic strain energy in
the cracked body (U). Hence, the Strain Energy Release Rate (SERR), defined by
Eq. (1.4) , is commonly used as a measure of energy available for crack extension in an
IFT test specimen (ASTM D5528-01 2002).
Fracture Toughness is determined as the value of critical stress intensity factor (K
c
)
or critical strain energy release rate (G
c
) of a material. For an isotropic and homogeneous
material, the stress field in the vicinity of the crack tip can be characterized by a single
parameter, stress intensity factor (K). However, for composite materials, the stress field
near the crack tip is more complicated and sometimes shows oscillatory behavior. Thus
another parameter, ERR (G), based on energy released during the creation of new
surfaces, is more commonly used for composite materials. By comparing the energy
release rate, G, to the critical value, G
c
, of a material, one can predict this materials
capability to resist crack growth. In the JIS standard (JIS K 7086 1993), interlaminar
fracture toughness (IFT) is defined as the critical value of the energy required to create a
unit area of an interlaminar crack.

b
a
da
P
u
b
a
da
P
u

Fig. 1-1: Load and displacement on a crack body
1 dU
G
b da
= (1.4)
5

According to the relative displacement crack surfaces, three modes of fracture
existing for laminated composite materials (Figure 1-2) are defined as (ASTM standard
D5528-01 2002):
crack opening mode (Mode I) fracture mode in which the delamination faces
open away from each other and no relative crack face sliding occurs; the critical value of
G for delamination growth in this mode is named Mode I interlaminar fracture toughness
(IFT), denoted by, G
Ic
;
crack sliding mode (Mode II)fracture mode in which the delamination faces
slide over each other in the direction of delamination growth and no relative crack face
opening occurs (in the direction normal to the leading edge); Mode II interlaminar
fracture toughness is denoted by, G
IIc
;
crack tearing mode (Mode III)fracture mode in which the delamination faces
slide over each other in the direction parallel to the leading edge.
The mixed mode fracture toughness, G
c,
is defined as the critical value of strain
energy release rate, G, for delamination growth in mixed-mode.
Many international organizations and groups are actively involved in carrying out
research on interlaminar fracture toughness (IFT) testing. Some of them are: i) ASTM
Subcommittee D30.06; ii) the Polymers & Composites Task Group of the European
Structural Integrity Group (ESIG, formerly the European Group on Fracture); iii) the
Japan High Polymer Center (JHPC). In 1990, an international round robin exercise was


v
u
w
x
y
z

Figure 1-2: The basic fracture modes
6

carried out under collaboration between the ASTM group, the ESIG group and the
Japanese Industrial Standards (JIS) group. International cooperation at a government
level, the Versailles Project on Advanced Materials & Standards (VAMAS), has also
dealt with IFT since 1986.
In fatigue, the delamination process of a composite material is often characterized
in terms of the relationship between the crack growth rate per cycle and the applied range
of ERR on a log-log plot. Generally, there are three regimes of crack growth when
plotting crack growth rate against ERR on a log-log scale, with the first regime showing a
fast decelerating growth rate with increased ERR, the second regime a linear relationship
between crack growth rate and ERR, and the third regime markedly increasing crack
growth rate with increasing ERR. The behavior of laminated materials in the second
regime is of great interest for developing a damage tolerance design approach. Applying
such an approach, a designer needs to take into consideration how fast an existing crack
can grow while the structure is in service. The Paris law ( ( )
n
G B
dN
da
= ) and modified
Paris law ( ( )
n
G B
dN
da
max
= ) between crack growth rate per cycle (da/dN) and the applied
ERR (G) are commonly used to characterize crack growth in the second regime .
1.3 Problem Statement and Research Objectives
Laminated fiber-reinforced composites are well-known to be susceptible to
delamination. Advanced stress analysis tools and validated failure criteria are needed to
determine conditions for crack onset and growth under design loads, and also the
delamination growth rate in fatigue loading. To apply the fracture mechanics based
criterion with confidence, standard test methods are needed to characterize the fracture
resistance as a generic property of a material and a database of such properties for
commonly used composite materials needs to be established for design and material
application purposes.
7

Based on previous research by the many IFT research organizations and
individual researchers, several test configurations have been proposed for studying
delamination fracture behavior under various kinds of loading. For Mode I tests, the
double-cantilever-beam (DCB) specimen is the most common specimen. Several national
and an international standard (ASD-STAN prEN 6033 1995; ASTM D5528-01 2002;
ISO 15024 2002; JIS K 7086 1993) already exist for the quasi-static Mode I IFT testing.
For Mode II, the end-notched-flexure (ENF) specimen is one of the most popular ones
but the unstable crack growth issue exists for the common configurations, and hence
other test configurations were proposed. A Japanese and an European standard (ASD-
STAN prEN 6034 1995; JIS K 7086 1993) are available for the quasi-static Mode II
testing using the ENF specimen. For mixed Mode I/II, more test configurations exist, e.g.
mixed mode bending (MMB), mixed mode end load split (MMELS), cracked lap shear
(CLS), and single-leg-bending (SLB). An ASTM standard using the MMB specimen
exists for the quasi-static mixed mode I/II IFT testing. Even though experts have more or
less agreed on a few test specimens for quasi-static Modes I, II, and I/II IFT testing, there
are still many practical issues being debated, and some should be treated differently
according to the material system of interest. For fatigue testing, no standard exists for
characterizing the crack growth law in the stable crack growth region, where the
relationship between the delamination growth rate and strain energy release rate (SERR)
follows a power law.
In addition to experimental methods, numerical methods, such as finite element
modeling, are valuable tools for validating the LEFM approach in predicting
delamination. However, performing crack propagation analysis using finite element
method is numerically intensive. Further, the ERR based failure criteria are not yet
available in most commercial finite element analysis software. Although a numerical
technique, the virtual crack closure technique (VCCT), has been proposed, research on
applying this technique to advanced structures is preliminary. Extensive efforts are
needed to develop efficient modeling techniques for crack propagation analysis.
With the need for better understanding of delamination, and establishing the IFT
property database for structure design, this thesis research aims to:
8

1) assess the interlaminar fracture toughness testing methods found in the
literature for static and fatigue testing of laminated composite materials;
2) characterize the Mode I, Mode II, and mixed-mode I/II interlaminar fracture
and interlaminar fatigue properties of a proprietary carbon/epoxy composite material
system using test methods from the literature or, if necessary, test methods tailored to suit
the current material;
3) critically evaluate the test results and make recommendations for future
investigations;
4) preliminarily assess the capability of the current commercial finite element
software to perform crack propagation analysis.
9


Chapter 2

Literature Review
In this chapter, the widely-used interlaminar fracture toughness (IFT) testing
specimens are introduced based on a wide survey of the literature. Experimental aspects,
data reduction and analysis methods regarding IFT testing are reviewed.
2.1 Mode I Interlaminar Fracture Toughness (IFT) Testing
To-date, the Double Cantilever Beam (DCB) specimen is the dominant Mode I
testing specimen. The specimen is easy to manufacture and a pure Mode I stress state at
the tip of the crack is easy to create with commonly available mechanical testing
equipment. Several analytical approaches for interpreting the results of DCB testing are
summarized in this section. Different views on typical practical issues involving Mode I
testing, e.g. the initial defect type, critical point definition, fiber bridging, etc., are
presented.
2.1.1 Geometry and analysis of the Double Cantilever Beam (DCB) specimen
Some national and international standards are available for reference regarding
Mode I IFT testing. ASTM standard (ASTM D5528-01 2002) using a DCB specimen for
Mode I IFT testing was first published in 1994. A DCB specimen is also used in Japanese
standard (JIS K 7086 1993) and in European standard (ASD-STAN preEN 6033 1995).
An ISO (ISO 15024 2002) standard is also available.
The geometry of the DCB specimen described in (ASTM D5528-01 2002) is
shown in Figure 2-1.

10

Specimen dimensions required in ASTM standard D5528 (ASTM D5528-01 2002)
are listed in Table 2-1.
The ASTM DCB specimen shown in Figure 2-1 consists of a rectangular cross
section and uniform thickness and width. Opening forces are applied to the specimen by
piano hinges or loading blocks bonded to one end of the specimen. The ends of the
specimen are opened by controlling either the opening displacement or the crosshead
movement. The load, crosshead (or crack opening) displacement and delamination length
are recorded continuously during the test. Load versus displacement plots are generated
during or after the test. The delamination length is determined as the distance from the
loading line to the front of delamination. The initiation and propagation value of G
Ic
can
be calculated based on these recorded data using beam theory and so-called compliance
calibration methods.
Several analytical models can be used to reduce the data for a DCB test. The three
data reduction methods recommended in (ASTM D5528-01 2002) are: (1) Modified

a0
l
2h
b
a0
l
2h
b
Piano
hinge Aluminum
block
Insert
Insert
a0
l
2h
b
a0
l
2h
b
Piano
hinge Aluminum
block
Insert
Insert
Fig. 2-1: Geometry of the ASTM D5528 Double Cantilever Beam (DCB) specimen
Table 2-1: DCB specimen dimensions required in ASTM D5528
Length Width Thickness Initial crack length
L, mm b, mm 2h, mm a
0
, mm
125 (5 in.) 20-25 (0.8-1.0in.) 3-5(0.12-0.2 in.) ~50 (2 in.)


11

Beam Theory (MBT) method, (2) Compliance Calibration (CC) method, and (3)
Modified Compliance Calibration (MCC) method. The main difference in these methods
lies in the different models used for the compliance vs. crack length relation. Both the
MBT and MCC methods assume that compliance (C) is related to crack length (a) by a
third order polynomial, but in different forms. According to the simplest of beam theories,
the delaminated portion of the DCB specimen can be treated as two Euler-Bernoulli
cantilever beams each built in at the crack tip and point-loaded near the opposite end.
Some inaccuracy can result from the perfect-built-in-end assumption, and the simple
beam model can be improved by considering two effects: i) end displacement and
rotation of the cantilever beams, and ii) transverse shear deformation of the delaminated
beams. The MBT method assumes that the end rotation effect can be incorporated simply
by adding a correction factor to the crack length and this correction factor is constant for
a certain material. Another beam analysis model, developed by Ozdil and Carlsson (Ozdil
and Carlsson 1999) and used for angle ply ([]
n
) laminates, assumes that the un-
delaminated portions of the beams are bonded by an elastic medium of infinitesimal
thickness which allows beam end displacement and rotation. The effect of shear
deformation in a DCB specimen is typically small (<1%) and can be neglected, according
to one analytical study (Ozdil and Carlsson 1999). The most commonly used analytical
models for improving elementary beam theory in the analysis of DCB test results are
presented in more detail next.
2.1.1.1 Modified Beam Theory (MBT) method
This method assumes that the cracked part of the specimen consisting of an upper
arm and lower arm can be represented as two cantilever beams built-in at a distance in
front of the crack tip (i.e., embedded in the un-delaminated part of the specimen), as
shown in Figure 2-2. (ASTM D5528-01 2002)
12

From Euler-Bernoulli beam theory, the predicted compliance can be written as a
third order polynomial, by Eq. (2.1).
Applying the SERR-compliance relation given by Eq. (2.2),
the Mode I SERR is then given by Eq. (2.3).
where P and are the load and total crack opening displacement as shown in Figure 2-2,
respectively. b is the specimen width, and a is the length of the delamination from
loading points to crack tip. C is the compliance and defined by /P. The parameter is
determined experimentally by the abscissa intercept of a straight line fit by least squares
to C
1/3
versus a data as shown in Figure 2-3.

P
P

2h
a

P, /2
P, /2
a +

Fig. 2-2: A schematic of the DCB specimen (side view)
( )
3 3
+ = a m C (2.1)
da
dC
b
P
G
2
2
= (2.2)
( ) +
=
a b
P
G
I
2
3

(2.3)
13

Eq. (2.3) can be modified to include the correction for the loading block
reinforcement effect and the large displacement effect. In this case, the expression for
SERR is shown in Eq. (2.4) (de Morais and Pereira 2007).
where,
I
is a correction factor for crack tip displacement and rotation, F a correction
factor for large displacements, and N a correction factor for loading block effects.
Expressions for F and N are given in (ASTM D5528-01 2002).
Assume the experimentally obtained compliance C vs. a data points are
approximated by the linear polynomial of Eq. (2.5),
where, A
0
and A
1
are parameters obtained from a linear least-square fit to C/N vs. a data.
Then the end correction factor,
I
, is given by Eq. (2.6),
and an estimate of the flexural modulus is given by Eq. (2.7),

a
C
1/3

0

Fig. 2-3: Determination of in the modified beam method (MBT)
( )
Ic
3
2
I
P F
G
b a N

=
+

(2.4)
( ) a A A N C
1 0
3 / 1
/ + = (2.5)
1
0
A
A
I
=
(2.6)
14

On the other hand, if the material properties are known, the correction factor
I

can be predicted from the theoretical solution for a beam on an elastic foundation by
Williams (1989), Eq. (2.8),
where,
and E
3
and G
13
are the through-thickness Youngs modulus and transverse shear
modulus, respectively.
The end correction factor
I
calculated by Eq. (2.6) and the estimated flexural
modulus E
1
calculated by Eq. (2.7) are commonly found to be greater than the beam
theory prediction by Eq. (2.8) and the experimentally measured flexural modulus,
respectively (de Morais and Pereira 2007).
2.1.1.2 Compliance Calibration (CC) method (Berrys Method)
From the simple beam theory model, the deflection, v, of the tip of a cantilever
beam with length a may be written as Eq. (2.10),
For a double cantilever beam, the end deflection is assumed to be related to the load by
Eq. (2.11). (ASTM D5528-01 2002; Davies et al. 1990)
( )
3
1
1
8
hA b
E =
(2.7)

=
2
13
1
1
2 3
11G
E
h
I
(2.8)
13
3 1
18 . 1
G
E E
= (2.9)
EI
Pa
v
3
3
= (2.10)
n
RPa = 2 / (2.11)
15

Hence, the predicted compliance is in the form of Eq. (2.12).
The strain energy release rate G
I
may thus be written as Eq. (2.13).
where n is experimentally determined by the slope of straight line that has been fit to
lnC lna data as shown in Figure 2-4. The logarithm of R is the intersection of the best-fit
straight line with the ordinate.
2.1.1.3 Modified Compliance Calibration (MCC) method
In the MCC method, the normalized compliance (bC) is related to the normalized
delamination length, a/2h, by Eq. (2.14) (JIS K 7086 1993).
where
1
can be determined experimentally by the slope of a straight line that has been fit
to a plot of a/2h versus (bC)
1/3
data as shown in Figure 2-5.
n
Ra C = (2.12)
2 1
I
2 2
n
nP nP Ra
G
ba b


= = (2.13)


Fig. 2-4: Determination of n in compliance calibration method
0
3 / 1
1
) ( 2 / + = bC h a (2.14)
16

The expression for SERR by MCC method is given by Eq. (2.15).
An estimate of the coefficient
1
and
0
is given by Eq. (2.16) and Eq. (2.17),
respectively,
where, is given by Eq. (2.9). E
1
and G
13
are the longitudinal Youngs modulus and
transverse shear modulus, respectively.
2.1.1.4 Elastic Foundation Model (EFM) method
Unlike the three methods illustrated above, where compliance versus crack length
relations are determined experimentally, the compliance of a DCB specimen as a function
of crack length is theoretically predicted based on a Euler-Bernoulli beam and Winkler

Fitted line

1
(bC)
1/3
a/2h

Fig. 2-5: Determination of
1
in the Modified Compliance Calibration (MCC) method
) 2 ( 2
3
2
1
3 / 2 2
h b
C P
G
I

=
(2.15)
4
3
1
1
E
=
(2.16)
2 / 1
2
13
1
0
1
2 3
11 2
1

=
G
E
(2.17)
17

elastic foundation model analysis. This model was derived by Ozdil and Carlsson (Ozdil
and Carlsson 1999) for angle ply laminates. Compliance and SERR of a DCB specimen
can be calculated based on material properties. Hence, by conducting this analysis, the
compliance calibration procedures can be avoided if the material properties are known.
The mathematical model of half of the DCB specimen is shown in Figure 2-6.
The governing differential equation for the DCB is given by Eq. (2.18),
where H(x) and is given by Eq. (2.19) and Eq. (2.20), respectively.
and
where k
e
is the foundation modulus and E
x
is the effective bending modulus of the
laminate. Applying boundary conditions and solving the governing equation, the
expression for the deflection of the beam was obtained. Thus, the compliance of the beam
is given by Eq. (2.21).

a
/2
h
x
z
P
k
e
Figure 2-6: An elastic foundation model of the DCB specimen, based on (Ozdil and
Carlsson 1999)
( )
( ) ( ) 0 4
4
4
4
= + x w x H
dx
x w d
(2.18)
( )

<
>
=
0 , 0
0 , 1
x
x
x H
(2.19)
3
4
3
bh E
k
x
e
=
(2.20)
18

and the expression for the Mode I SERR is given by Eq. (2.22),
where E
z
is the Youngs modulus in the through-thickness direction. The term outside the
parentheses of Eq. (2.22) is identical to the SERR expression for a DCB specimen given
by the beam theory, which is strictly valid only for infinitely long crack lengths. The term
inside the parentheses is the elastic foundation correction factor for an angle-ply laminate
that accounts for the finite crack length by incorporating the through-thickness elasticity
of the uncracked region of the specimen.
2.1.2 Experimental aspects of the DCB test
Although the standard method for the Mode I IFT testing has been established
since the early 1990s, some experimental aspects are still not completely resolved. For
some issues, different recommendations could be made for different materials. Among
them, the issues of most concern for researchers are the effects of initial defect types,
crack initiation definition, effect of fiber bridging, etc.
2.1.2.1 Initial defect type
Many studies have shown the defect type may have a great influence on
experimental results (Brunner 2000). Generally, three approaches have been used for the
initial defect:
a. A thin film insert is put between two lamina during the layup process of a
laminated material and no precracking is performed prior to testing;

=
4 / 3 2 / 1
2
4 / 1
3
39 . 0 22 . 1 92 . 1 1
8
z
x
z
x
z
x
x
E
E
a
h
E
E
a
h
E
E
a
h
h
a
b E
C
(2.21)
1/ 4 1/ 2
2
2 2
I 2 3
12
1 1.28 0.41
x x
x z z
E E P a h h
G
E b h a E a E



= + +





(2.22)
19

b. A thin film insert is put between two lamina during layup, and a wedge is
driven into the initial crack to grow the delamination beyond the starter film tip,
or a brief Mode I test is performed before the test;
c. A thin film insert is put between two lamina during layup, and the test
specimen is precracked by performing a brief Mode II test to grow the
delamination beyond the starter film tip.
Method (a) is obviously the easiest method since no precracking is involved. The insert
film allows the formation of a macroscopically well defined shape of the starter crack tip
as well (usually straight across the specimen width, at least on a macroscopic scale).
However, one problem associated with this type of starter crack is that a resin pocket can
form beyond the tip of the insert film, as shown schematically in Figure 2-7. This
disturbance of local material distribution may become negligible as the thickness of the
insert film decreases. However, the limiting value of the thickness of the insert films
depends on the material type ((Hojo et al. 1995), (Murri and Martin 1993)). Early round-
robin tests have indicated that insert thickness has to be less than or equal to 13 m to
yield results not affected by insert thickness in Mode I DCB testing, at least for
unidirectional carbon fiber reinforced polymer laminates. For glass fiber laminates, due
to the larger diameter of some types of glass fibers compared with carbon fibers, thicker
film inserts might still be suitable without significantly affecting the fracture toughness
measurement. Another disadvantage of using insert film as starter crack noticed by Hojo
et al. (1995) is that crack propagation from the starter film was unstable in brittle epoxy
unidirectional laminates without mechanically-induced precracks.
20

Precracking by mechanical loading (Methods b and c) has several advantages and
disadvantages. Precracking in Mode I has been shown to yield conservative values of
fracture toughness compared with insert films for a number of carbon and glass fiber
composites, in particular for Mode II testing (Hojo et al. 1995; Murri and Martin 1993).
Mode I or wedge precracking has been shown to yield starter tip shapes similar to those
observed during Mode I testing. However, precracking may form a damage zone ahead of
the crack tip and irregular crack front shape as well. An additional issue is that fibers tend
to bridge between interfaces with crack extension, which results in a higher initiation
value than an insert film defect type. Even so, it is suggested by the Japanese Standards
Association (JIS K 7086 1993) that a precrack should be done for a DCB specimen
before testing.
The ASTM DCB test (ASTM D5528-01 2002) suggests to use an insert film with
a thickness less than 13 m for the initial defect of a DCB specimen so that an initiation
G
Ic
value based on the insert film could be obtained. Then after the crack had initiated
and grown for 3-5 mm, the specimen shall be unloaded and reloaded so that a G
Ic
value
based on Mode I precrack can be obtained as well.

End of insert film
Resin pocket

Figure 2-7: A schematic of microscopic view of longitudinal section of specimen near the
end of starter film
21

2.1.2.2 Definition of critical point for crack onset
Four approaches have been developed to define the critical crack onset point as
shown in Figure 2-8.


(a)

(b)
Figure 2-8: Crack initiation definitions
22

a. Non-linearity of the load-displacement plot (NL)
The nonlinear critical point is the first point that deviates from linearity on the
load vs. displacement curve. Work done at EMPA (Swiss Federal Laboratory for
Material Testing and Research) in Switzerland using in-situ micro focus radiography
has proved that the NL point is close to the real onset of delamination (Kalbermatten
et al. 1992). However, this definition is somewhat ambiguous, for the closer one looks
at the plots, the earlier nonlinear behavior can be detected. Additionally, nonlinear
behavior may be due to other reasons, for example, yielding of material at the crack
tip or local instead of global crack growth.
b. Maximum load point
The maximum load point is the critical point where load reaches its maximum
value during for the whole loading process.
c. 5% compliance increase (5% offset)
The 5% compliance offset point, is obtained by constructing a straight line whose
slope has decreased by 5% of the initial slope of load-displacement curve. Then the
critical point is intersection of this straight line and load-displacement curve. The
advantage of this definition is that it is less scattered than the NL method, but it
produces higher values.
d. Visual observation (VIS)
The visual observation point is the point where crack onset is observed visually.
The difficulty with this method is that it is operator dependent.
e. Acoustic emission (AE), and strain gages
These methods are relatively complicated compared to other methods and
require additional specialized equipment. They will not be discussed further in this
investigation.

Two typically observed types of load vs. displacement curves are shown in
Figure 2-8. For the first type, Figure 2-8 (a), it has a slight nonlinear behavior before the
load reaches the maximum, and the maximum load was reached before the compliance of
specimen increased by 5%. In this case, as suggested by the Japanese Standards
23

Association (JIS K 7086 1993), the maximum load point is usually taken as the critical
point. For the second type of curve, Figure 2-8 (b), a relatively greater degree of
nonlinearity is observed before the maximum load and the 5% compliance offset critical
point occurs before the maximum load. In the second case, the percentage compliance
increase criterion is commonly used for critical point definition for crack onset.
2.1.2.3 Method of loading/unloading
For a typical DCB test, the total amount of crack growth from the starter crack tip
is about 50-60 mm. Two kinds of loading/unloading methods have been widely used,
according to the literature. The first method includes only one or two loading/unloading
cycles, with the short loading/unloading cycle creating a crack extension of about 3-5 mm
and a second long loading/unloading creating a crack extension of about 50 mm. In the
second method, several loading/unloading cycles were conducted. A typical load against
opening displacement curve of loading and unloading procedures is shown in Figure 2-9.
The purpose of the loading/reloading process is to obtain a more accurate compliance
value for the crack lengths at which the specimen is unloaded.


Fig. 2-9: Typical load-displacement curves for a DCB specimen with multiple
loading/unloading cycles
24

2.1.2.4 Crack resistance curve and fiber bridging
A typical critical strain energy release rate, G
Ic
, against crack length, a, curve is
shown in Figure 2-10.
As shown in Figure 2-10, a resistance type of fracture behavior is commonly
found for laminated composite material (Brunner et al. 2006; Russell 1988). This
behavior commonly features a monotonically increasing G
Ic
in the first few millimeters
(i.e., 3-5 mm) of Mode I crack extension, and then stabilizes with further crack growth.
According to many studies (Brunner et al. 2006; de Morais and Pereira 2007; Russell
1988), fiber bridging is the primary reason for this history dependent behavior. As the
crack starts to extend, fibers begin to pull out of the delaminated surfaces immediately
ahead of the crack tip, and gradually a zone of fibers bridging the gap is developed
between the delamination faces directly behind the crack tip, as shown in Figure 2-11. As
the delamination extends and the crack opening displacement increases, these bridged
fibers continue to pull out and in some cases break due to applied tensile stress.
Consequently, the bridging fibers divert some of the available strain energy away from
the crack tip.


Deviation from linearity
Visual onset
5% offset
Propagation
Crack length, a, mm
G
Ic
,
J/m
2

Fig. 2-10: Typical delamination resistance curve (R curve) from a DCB test (ASTM
D5528-01 2002)
25

As a consequence of the fiber bridging mechanism, the measured compliance of
the DCB specimen is less than that predicted by modified beam theory for the visually
observed crack length (Brunner et al. 2006). Additionally, if dense fiber bridging occurs
on the edge of the specimen, it can be difficult to visually locate the tip of the crack,
which results in unreliable crack length measurements.
Other effects that may be involved in the shape of a resistance curve include:
matrix cracking, tow cracking, multiple delamination, tow bridging and tow breaking in
the case of woven fiber composites, etc. The implication of R-curve behavior for
structural designers is that the initiation fracture toughness must be measured as well as
the increase of fracture toughness as a delamination grows if a damage tolerance criterion
is part of the design process.



Fig. 2-11: Picture showing fiber bridging (Mode I loading)
26

2.2 Mode II Interlaminar Fracture Toughness (IFT) Testing
For Mode II interlaminar fracture toughness testing, at least six types of test
specimens are available. They are the end notched flexure (ENF) specimen (Carlsson et
al. 1986; Ozdil et al. 1998), the stabilized end notched flexure (SENF) specimen (Tanaka
et al. 1995), the four point bend end notched flexure (4ENF) specimen (Davies et al. 2005;
Schuecker and Davidson 2000), the end load split (ELS) specimen (Hashemi et al. 1990a;
Hashemi et al. 1990b; Wang and Vu-Khanh 1996), the over notched flexure (ONF)
specimen (Szekrnyes and Uj 2005; Tanaka et al. 1998; Wang and Qiao 2003), and the
tapered end-notched flexure (TENF) specimen (Qiao et al. 2004; Wang and Qiao 2003).
Among them, the ENF specimen has been most widely used for Mode II IFT testing of
fiber composites because of its simple test configuration. Many analytical models have
been developed for ENF specimens, and many discussions on experimental testing can be
found in the literature as well. No international standard exists for a Mode II Interlaminar
Fracture Toughness (IFT) test.
2.2.1 Geometry and analysis of the End Notched Flexure (ENF) specimen
In an ENF test, the specimen is placed in a three point bending fixture which
consists of two supporting points and one loading point. The generic ENF specimen
geometry is shown in Figure 2-12. Specific ENF specimen geometries found in the
literature are shown in Table 2-2. Two national standards were established in Japan (JIS
K 7086 1993) and Europe (ASD-STAN prEN 6034 1995) for Mode II testing using ENF
specimens.

27


In the ENF test, load is applied to the top of specimen at the mid-span in a
displacement controlled mode. The load, center point displacement and crack length are
measured and recorded during the test. Load versus displacement plots are generated


Fig. 2-12: End-notched flexure test schematic
Table 2-2: Some specific ENF specimen geometries from the literature
Material
Half span
length
Specimen
thickness
Specimen
width
Initial
crack
length
Reference
L, mm 2h, mm b, mm a
0
, mm
E-glass/epoxy 65 5 20 35
(Ducept et al.
1997)
Carbon/epoxy 40 4.2 20 N/A
(Davies et al.
1990)
E-glass/polyester 50 4.4 20 25
(Ozdil et al.
1998)
Carbon/epoxy 50 3 20 20
(Tanaka et al.
1995)

28

during or after the test. The initiation and propagation G
IIc
can be evaluated based on
these recorded data using classical plate theory, beam theory and compliance calibration
methods.
Several analysis methods can be applied to an ENF test, including classical plate
theory, beam theories with shear correction, and the compliance calibration method.
These analysis methods are described next.
2.2.1.1 Classical Plate Theory (CPT) (Davidson et al. 1996)
The assumptions made in applying classical plate theory to the ENF test are:
i) the crack is at the midplane of the laminate;
ii) the laminate and two sublaminates created by the crack are specially
orthotropic and symmetric (A
16
= A
26
= B
ij
= D
16
= D
26
= 0), but not necessarily
homogeneous or unidirectional;
iii) as a result of the material symmetry, residual thermal stresses do not affect
the strain energy release rate;
iv) a multidirectionally reinforced ENF specimen may exhibit non-classical
bending behavior; however, with bending rigidities of the cracked and uncracked regions
chosen appropriately, 2D plate theory is assumed to be able to accurately predict
deflections and SERRs.
The ENF specimen is treated as a bending plate consisting of cracked and
uncracked regions as shown in Figure 2-13. Two extreme cases of constraint conditions
are considered by Davidson et al., with the first one referred to generalized plane stress
condition and the second one generalized plane strain condition (Davidson et al. 1996).
For generalized plane stress condition, it is valid for specimens where the cracked and
uncracked regions are long and narrow. In this case, the moments are zero on the edges of
the laminate, and the effective bending rigidity per unit width, D, of either region can be
expressed as Eq. (2.23),
22
2
12 11
/ D D D D = (2.23)
29

where D
11
, D
12
, and D
22
come from the D matrix of classical lamination theory (Daniel
and Ishai 2006).
The generalized plane strain condition is valid for specimens where the cracked and
uncracked regions are short and wide. In this case, the bending rigidity per unit width of
either region, D, can be expressed as Eq. (2.24).
where D
11
comes from the D matrix of classical laminated plate theory (Daniel and Ishai
2006) for the region of interest in the specimen. The difference between the plane stress
and plane strain flexural rigidities may be characterized by the ratio, D
c
, defined by
Eq. (2.25).
Small values of D
c
indicate that the effect of finite width is relatively unimportant, while
the larger value of D
c
indicate a greater influence of the three-dimensional effect as
explained in the following Section 2.2.2.
Using the moment-curvature relationship, the equation for compliance may be
derived using classical plate theory as Eq. (2.26),


L
P,
a
Uncracked
region
Cracked
region
L

Fig. 2-13: A schematic of the ENF specimen (side view)
11
D D = (2.24)
22 11
2
12
D D
D
D
c
= (2.25)
( )
u
CPT
bD
R a L
C
24
2 4
3 3
+
=
(2.26)
30

where, b is the specimen width, C is the compliance, defined as center point deflection
divided by load, R is the ratio of flexural rigidities of the uncracked and cracked regions,
such that
c u
D D R / = , and D
c
is the bending rigidity of either of the identical uncracked
arms.
Substituting the expression for compliance into Eq. (2.2), expression for the
SERR of the ENF specimen may be obtained as Eq. (2.27)
or
For a homogeneous specimen, which could be a unidirectional or an angle-ply specimen
with a uniform through-the-thickness fiber distribution, for example, R = 8.
For a homogeneous orthotropic specimen configured as a narrow beam, the bending
rigidity may be written as,
where, E
11
is the Youngs modulus in the x direction, I is entire specimens area moment
of inertia about its neutral axis, and h is the specimens half thickness. As a result, the
compliance of such a specimen is,
Substituting R = 8 and Eq. (2.30) into Eq. (2.27), yields the expression of SERR by
Eq. (2.31).
Normally, the accuracy of Eq. (2.28) is higher than Eq. (2.31) since material
properties are not required by Eq. (2.28). The accuracy of Eq. (2.28) in reducing data from
( )
u
CPT
II
D b
R a P
G
2
2 2
16
2
=
(2.27)
( )
( ) [ ] 2 4
2
2
3
3 3
2
+

=
R a L
R
b
a P
G
CPT
II


(2.28)
3 / 2 /
3
11 11
h E b I E D
u
= = (2.29)
3
11
3 3
8
3 2
bh E
a L
C
HP
ENF
+
=


(2.30)
3 2
11
2 2
16
9
h b E
a P
G
CPT
II
=
(2.31)
31

an ENF test depends upon whether the experimental compliance, C, and its derivative,
C/a, obey the functional form of Eq. (2.26) (and its derivative).
For ENF tests with test fixtures that use loading pins, R will generally be less than
theory predicts due to the finite diameter of the pins ((Davidson et al. 1996) (O'Brien et
al. 1989)). For ENF fixtures that use knife edge supports, R may be either greater or
less than theory, and depends upon the fixture itself as well as the manufacturing process
for the specimen.
2.2.1.2 Beam Theory (BT) with shear correction (Carlsson et al. 1986)
Russell and Street presented a beam theory solution for the SERR of the ENF
specimen in 1985 (Russell and Street 1985). Based on their work, Carlsson et al.
(Carlsson et al. 1986) extended the beam theory solution to include the effect of shear
deformation on SERR.
This beam theory solution with shear correction can be applied to the ENF
specimen with the common geometry, for which delamination occurs at mid-thickness, as
shown schematically in Figure 2-14. The center point deflection can be treated
decomposed into three parts as by Eq. (2.32),
where,
AB
,
BC
, and
CD
are defined in Figure 2-14. The beams BC and CD are treated
as cantilever beams and it is assumed that the cross-section at C does not warp because of
the line of load introduction is an approximate line of symmetry. Expressions for
deflection
BC
, by Eq. (2.33), and
CD
, by Eq. (2.34), may be obtained from Timoshenko
beam theory.

2
CD BC AB
+ +
= (2.32)
[ ] ( )


+
+ +
=
bh G
a L P
bh E
a aL L P
BC
13
3
11
3 2 3
3 . 0
8
3 2

(2.33)
32

where, E
11
and G
13
are the longitudinal Youngs modulus and the transverse shear
modulus of the composite laminate, respectively.
For the delaminated region, AB, the displacement,
AB
, has two components, one
due to the bending and shearing deformations of the two delaminated beams and the other
due to the rotation of the cross-section at point B. The ends of the parallel beams are here
assumed to be allowed to deform freely under the action of shearing stress. Assuming
that each beam of the delaminated region carries the same load, P/4, the displacement
component due to bending and shearing,
AB,1
, is obtained as Eq. (2.35) .
The rotation of the cross section at B is expressed by the slope of the cross section
with respect to the horizontal axis. The displacement component due to the slope,
AB,2
, is

+ =
bh G
PL
bh E
PL
CD
13
3
11
3
3 . 0
4

(2.34)

L L
a
2h
P
A
P/2
P/2
B
C
D
L L


CD

AB

BC
A B C D
a
(a) Loading of the ENF specimen
(b) Deflection of the ENF specimen

Fig. 2-14: Schematic of an ENF specimen subject to three-point bending

+ =
13
2
1
2
3
1
3
1 ,
8
3
1
G a
E h
bh E
Pa
AB

(2.35)
33

simply the slope at B multiplied by the length of the delaminated region, which yields the
following expression of Eq. (2.36).
The final expression for compliance and SERR of an ENF specimen incorporating
shear deformation is Eq. (2.37) and Eq. (2.38), respectively.
For geometries and material properties commonly in use (unidirectional lay-ups), the
error in the toughness value calculation induced by neglecting shear deformation is less
than 10% according to the analysis conducted by Carlsson et al. (Carlsson et al. 1986).
Ozdil et al. (Ozdil et al. 1998) pointed out that compliance and SERR determined for
relatively thin unidirectional and angle-ply laminate ENF specimens were in good
agreement with a previous classical plate theory formulation. For thicker laminates,
however, effects of shear deformation on the compliance of ENF specimen become
significant.
2.2.1.3 Compliance Calibration (CC) method (Davidson et al. 1996)
At any given crack length, the compliance can be obtained from the slope of a
linear least-squares curve fit of the deflection versus load data. A cubic polynomial,
shown in Eq. (2.39), can be used to fit the compliance vs. crack length data.
Other expressions typically used to fit compliance data may exclude the first and second
order terms, such as, Eq. (2.40).

+
13
1
2
3 2
3
1
2 ,
8
3
G
E ah
a aL
bh E
P
AB

(2.36)
( )
( )

+
+
+
+
=
13
3 3
1
2
3
1
3 3
,
3 2
9 . 0 2 . 1 2
1
8
3 2
G a L
E h a L
bh E
a L
C
SH ENF

(2.37)

+ =
2
13
1
3 2
1
2 2
,
2 . 0 1
16
9
a
h
G
E
h b E
P a
G
SH ENF

(2.38)
2 3
0 1 2 3
C C C a C a C a = + + + (2.39)
34

Substituting expression for C given by Eq. (2.2) yields,
2.2.2 Experimental aspects of the ENF test
Similar to Mode I testing, the issues of initial defect type and critical point
definition exist as well for Mode II testing with ENF specimens. However, in many cases
the definition of crack onset critical point is quite clear for an ENF test because the crack
growth immediately after crack onset is often unstable. This instability is manifested as a
large load drop as shown in a typical stroke-controlled test result as shown in Figure 2-15.
Other issues concerning ENF testing include the effect of initial defect type, the
effect of friction between the crack surfaces, and three dimensional effects. They are
described below.
(a) Effect of initial defect type for Mode II testing
3
3 0
a C C C + = (2.40)
2
2
1 2 3
( 2 3 )
2
IIc
P
G C C a C a
b
= + + (2.41)


Fig. 2-15: Typical load-displacement curves of an ENF test
35

The apparent inconsistency between G
IIc
values measured by growing the
crack front from a thin midplane insert and G
IIc
values measured by growing the
crack from an initial shear precrack was noticed by many researchers (Murri and
Martin 1993; O'Brien 1998a). For Mode I testing, a characteristic toughness value
may be obtained as the insert thickness is decreased because the G
Ic
value reaches an
asymptotic limit as the insert film thickness decreases. However, according to some
researchers (Murri and Martin 1993; O'Brien 1998a), G
IIc
values decrease
continuously with decreased insert thickness, never reaching an asymptotic value that
may be considered a characteristic of the composite material. Furthermore, G
IIc

values measured from the insert are sometimes greater than and sometimes less than
G
IIc
measured from a shear precrack. In most cases, the precrack value is lower than
the insert value. However, for two materials (S2/SP250 glass/epoxy and IM7/F3900
graphite/epoxy) the reverse has been shown to be true (O'Brien 1998a).

(b) Effect of friction between fracture surfaces
Friction between the crack surfaces is inevitable for Mode II testing.
According to finite element analyses by Davies (Davies et al. 1996), this effect is not
significant for typical specimen geometries. However, it is recommended that a
PTFE film be placed between crack faces to reduce the influence of friction (JIS K
7086 1993). According to Carlsson et al. (1986), an analysis shows that, for
reasonable values of coefficients of friction (0.25-0.5), the error in G
II
induced by
neglecting friction is only 2-4%.

(c) Three-dimensional effects
Davidson et al. (1995a) noticed that significant 3-D effects, in the form of
concentrations in the Mode II SERR and occurrence of a Mode III SERR component
at specimens free edges, may occur during an ENF test with multidirectional
specimen. To a lesser extent, this 3-D effect has also been shown to occur in
unidirectional specimens.
36

According to Davidson et al. (Davidson et al. 1995a), 3-D effects may occur
in ENF geometries due to two causes: 1) the bending-stretching or bending-twisting
behavior of sublaminates; 2) the finite width of an ENF specimen. As a result of it,
the SERR and/or the mode ratio varies across the delamination front. The bending-
stretching or bending-twisting behavior may cause local gradients in the SERR and
mode ratio across the delamination front, and may be avoided by choosing specimen
geometries in which the cracked and uncracked sections are individually specially
orthotropic. A finite-width specimen will exhibit a local increase in the Mode II
SERR at each of its free edges. Also, due primarily to the anticlastic curvature of the
individual sublaminates, a local Mode III SERR component will occur at the
specimens edges. This effect is small in unidirectional laminates, but significant for
some other layups with a large value of D
c
.
2.2.3 Other test configurations of Mode II IFT testing
One disadvantage of using the ENF specimen is that the crack growth is unstable
until the crack grows to a certain point where the crack length to half span ratio, a/L, is
about 0.7 (Davies et al. 2001). Hence a fracture resistance versus crack length curve (R
curve) may be not generated from an ENF test. Many other test configurations are able to
develop stable crack growth, such as: SENF, 4ENF, ELS, but these have their own
advantages and disadvantages which are reviewed next.
Stabilized End Notched Flexure (SENF) test uses the same test configuration as the
ENF test. But rather than using a fixed cross-head displacement rate as in the ENF test,
the SENF test is controlled by a feedback signal from the test so that the crack growth
can be stabilized. The feedback can be the crack shear displacement (CSD) measured
between the top and bottom halves of the specimen at the insert end, or a coordinate
conversion control (Davies et al. 1998; Tanaka et al. 1995) signal which is a function of
the load and displacement output. This configuration allows generation of a R curve, but
requires a relatively more sophisticated test setup and procedure.
37

The 4ENF test uses a four-point-bending fixture rather than the more typical 3 point
fixture. It was shown by Paris et al. (2003) and Zile and Tamusz (2005) that crack growth
is stable in this test configuration and a resistance curve similar to Mode I can be
obtained. Schuecker and Davidson (2000) show that if the crack length and the
compliance are measured accurately, the ENF and 4ENF give the same fracture
toughness values. The primary difficulty with the 4ENF test is the nonlinear effect, such
as, effect of friction. As a result, the values of G
IIc
measured by 4ENF tests are often
significantly larger than those obtained from 3ENF tests. Values of G
IIc
obtained by the
4ENF test has been found to be 9-60% higher than those obtained by the 3ENF test
(Martin and Davidson 1999; Martin et al. 1998; Schuecker and Davidson 2000).
The End Load Split (ELS) specimen is shown in Figure 2-16. The condition for
stable crack growth for the ELS test specimen is a/L > 0.55, whereas for the ENF
specimen it is a/L > 0.7 (Davies et al. 2001). Hence, the ELS test specimen offers an
advantage over the ENF specimen in promoting stable propagation. This enables several
values of compliance and critical SERR to be obtained from one specimen at different
crack lengths and allows an experimental compliance calibration to be performed, as in a
DCB test. The disadvantage of the ELS test is that it requires a complicated fixture to
ensure a clamped end condition.


P
2h
a
L

Fig. 2-16: ELS test configuration, based on (O'Brien 1998a)
38

2.3 The Mixed Mode I/II Interlaminar Fracture Toughness Testing
For mixed mode I/II testing, many specimens have been developed, such as the
mixed mode bending (MMB) specimen, the mixed mode ELS specimen, the single leg
bending (SLB) specimen, the mixed-mode flexure (MMF) specimen, the cracked-lap
shear (CLS) specimen, the asymmetric double cantilever beam (ADCB) specimen. These
test methods are compared in (Szekrnyes and Uj 2007). ASTM has a standard for mixed
mode testing by MMB unidirectional specimen (ASTM standard D6671 2006). Among
these specimens, the MMB specimen is likely to be the most widely used specimen for
mixed mode testing, for the wide range of mode mixity it can create. The single leg
bending specimen is also an attractive candidate since it requires a relatively simple
loading apparatus and test configuration. Davidson et al. performed extensive theoretical
and experimental investigations using SLB coupons (Davidson et al. 1997; Davidson et
al. 2000; Davidson and Koudela 1999; Davidson and Sundararaman 1996; Davidson et
al. 1995c). Later, based on this SLB specimen, Tracy et al. proposed a single-leg four
point bend (SLFPB) geometry (Tracy et al. 2003), and Szekrnyes et al. introduced the
over leg bending (OLB) specimen (Szekrnyes and Uj 2007). One advantage of the SLB
specimen over the MMB specimen is that the compliance calibration method can be used
for the SLB specimen to obtain more accurate critical SERR values. In a MMB test,
SERR calculation involves some material properties. Large discrepancies commonly
exist between theoretically predicted and as-manufactured values of material properties,
such as bending rigidities, shear modulus and through thickness modulus (especially for
multidirectional laminates). As a result, the MMB test is not suitable for multidirectional
lay-ups. Considering all these factors, the single leg bending (SLB) specimen was
selected as the mixed mode IFT testing specimen in the current investigation. Hence,
further details on the SLB specimen and associated experimental aspects are described
next.
39

2.3.1 Geometry and analysis of Single Leg Bending (SLB) specimen
The SLB specimen has a beam type of geometry, as shown in Figure 2-17. A
portion of the lower arm is removed from the cracked end in an SLB specimen, so that
the reaction force act directly to the upper arm. The removed part can be used as a
spacer under the cut end, so that the upper surface of the specimen is horizontal before
loading. A standard three point bending fixture is used in this test. The load, center point
displacement and crack length are measured and recorded during the test. Compliance
and critical strain energy release rate, G
Tc
, can be calculated by compliance calibration
and classical plate theory methods, and mode ratios can be obtained by finite element
analysis incorporating a crack tip element.



Fig. 2-17: The SLB specimen geometry
40

2.3.1.1 Classical Plate Theory (CPT) Method (Davidson and Sundararaman 1996)
A schematic of side view of the SLB specimen geometry and detailed notation is
shown in Figure 2-18. By classical plate theory, the expressions of compliance, C,
derived by Davidson (Davidson and Sundararaman 1996) for the SLB specimen is given
by Eq. (2.42),
and total critical strain energy release rate G
TC
is given by Eq. (2.43).
Alternatively, the expression for total SERR in the case where bending rigidity per unit
width of the uncracked region (D
u
) is unknown, is given by Eq. (2.44).
where,
C = the compliance, defined by center-point deflection divided by the load;
a = crack length;
b = specimen width;
L = half span length as indicated in Figure 2-18;
( )
3 3
2 1
12
CPT
SLB
u
L a R
C
bD
+
= (2.42)
( )
2 2
, 2
1
8
CPT
Tc SLB
u
P a R
G
b D

= (2.43)



t
1

L
a
L
t
2
P,
Uncracked
region
Cracked
region
Top plate

Fig. 2-18: A schematic of the SLB specimen geometry and detailed notation
( )

+

=
1 2
1
2
3
3 3
2
,
R a L
R
b
a P
G
CPT
SLB Tc


(2.44)
41

R = the ratio of the flexural rigidity of the uncracked region, D
u
, to the bending
rigidity of the top plate of the cracked region, D
c
, as given by Eq. (2.45).
The bending rigidity of the uncracked region can be estimated by considering the
appropriate constraint condition as described in Section 2.2.1.1.
For a unidirectional specimen configured so that delamination occurs at the
midplane, the ratio R is 8. In this case, the expression for the total SERR can be
simplified to Eq. (2.46).
Mode decomposition can be achieved by a finite element approach using a crack
tip element (Davidson et al. 1995a; Davidson et al. 1995b). A crack tip element analysis
predicts the SERR components as Eq. (2.47) and Eq. (2.48).
where N
c
and M
c
are the concentrated crack-tip force and moment per unit width,
respectively. These parameters are given by Eq. (2.49) and Eq. (2.50), respectively.
where M
p
is the moment resultant at the crack tip, i.e.,
The constants c
1
, c
2
, a
12
, and a
22
are related to the plate rigidities as indicated by Eq. (2.52)
and Eq. (2.53), Eq. (2.55), Eq. (2.54):
/
u c
R D D =
(2.45)
3 3
,
7 2
7
2
3
a L b
P
G
CPT
SLB Tc
+
=

(2.46)
2
2
2 c
I
M c
G = (2.47)
2
2
1 c
II
N c
G = (2.48)
p c
M a N
12
=
(2.49)
P c
M a
t a
M

=
22
1 12
2

(2.50)
b
Pa
M
P
2
= (2.51)
42

where, t
1
and t
2
are the thickness of top plate and bottom plate respectively. A
1
and D
1
are
terms from the stiffness matrix of classical lamination theory (Daniel and Ishai 2006). A
1
,
D and D
1
are terms from the compliance matrix of classical lamination theory.
Specifically, A
1
is the in-plane rigidity of the top plate; A
1
is its in-plane compliance; D
1

is the flexural rigidity of the top plate; D
1
is the flexural compliance of the top plate, and
D is the flexural compliance of the uncracked region.
2.3.1.2 Beam Theory based analyses
Szekrnyes and Uj have applied Euler-Bernoulli and Timoshenko beam theories
in conjunction with a Winkler-Pasternak Foundation analysis, a Saint-Venant effect
analysis at the crack tip, and a crack tip shear deformation analysis to determine the
compliance and SERR of an SLB specimen with an initial crack at the midplane
(Szekrnyes and Uj 2007). The result for compliance is shown in Eq. (2.56),
while the Mode I and II components of SERR are given by Eq. (2.57) and Eq. (2.58),
respectively,
2
'
' 2
2
1 1
1 1
t D
A c + = (2.52)
1 2
' 2D c = (2.53)
2
'
2 1
12
t D A
a

= (2.54)
'
1 22
D D a = (2.55)
1/ 2
2
3 3 3
11 11
3 3
11 13 11 33 33
1/ 2 1/ 4 1/ 2
2
2 3
11 11 11
2 3
11 33 11 33 33
7 2 2
0.98 0.43
8 8 8
1 3
5.07 8.58
4 8
SLB
E E a L a L a h h
C
bh E bhkG bh E a E a E
E E E a a h h
bh E E bh E a E a E

+ +
= + + +





+ + +


3/ 4
3
11
33
2.08
E h
a E



+




(2.56)
43

where,
k = 5/6, shear correction factor;
E
11
= flexural modulus;
E
33
= through thickness modulus;
G
13
= longitudinal shear modulus.
2.3.1.3 Compliance Calibration (CC) method (Polaha et al. 1996)
A compliance versus crack length curve can be obtained by fitting a cubic
polynomial in the form of Eq. (2.59), to the compliance versus crack length data obtained
experimentally.
Other expressions typically used to fit compliance data may exclude the first and second
order term, as indicated by Eq. (2.60).
Therefore, the critical SERR can be obtained using either Eq. (2.61) or Eq. (2.62).
1/ 4 1/ 2
2
2 2
11 11
2 3
11 33 33
1/ 2
2
11 11
33 33
12
1 0.85 0.71
16
0.32 0.1
SLB
I
E E P a h h
G
b h E a E a E
E E h h
a E a E



= + +




+ +



(2.57)
1/ 2
2
2 2
11 11
2 3
11 33 33
9
1 0.22 0.048
16
SLB
II
E E P a h h
G
b h E a E a E



= + +





(2.58)
3
3
2
2 1 0
a C a C a C C C
SLB
+ + + = (2.59)
3
3 0
a C C C
SLB
+ = (2.60)
( )
2
3 2 1
2
,
3 2
2
a C a C C
b
P
G
c
SLB Tc
+ + = (2.61)
( )
2
3
2
,
3
2
a C
b
P
G
c
SLB Tc
= (2.62)
44

2.3.2 Experimental aspects of the SLB test
Since in a mixed-mode I/II test the specimen is subject to a combination of Mode
I and II loading, the experimental issues identified previously as potentially important in
pure mode tests, such as initial defect type, fiber bridging, and three-dimensionality, may
also be present during an SLB test to some extent depending on the specific mode ratio
and material type. Regarding the definition of critical point on the load-deflection curve,
the method used in Mode I and Mode II tests can as well applied as well for the SLB test.
2.3.3 Other configurations of mixed Mode I/II IFT test
For mixed Mode I/II IFT testing, another widely used test configuration is mixed
mode bending (MMB). The mixed-mode bending (MMB) test apparatus was designed by
Reeder and Crews (Reeder and Crews 1990), and redesigned later to reduce nonlinear
effects (Reeder and Crews 1992). Although the MMB test was created for thin,
unidirectional, symmetric laminates, it has been used for thick, asymmetric, off-axis ply
laminates as well (Kim and Mayer 2003). A schematic of the MMB test configuration is
shown in Figure 2-19. A great advantage of using the MMB test is that it provides the
ability to characterize delamination onset and growth for a wide range of mode ratios.
The loading in this test is a simple combination of the double cantilever beam Mode I and
the end notch flexure Mode II tests. This test method allows the generation of a wide
range of mode ratio by changing the lever length c as shown in Figure 2-19.
45

Many other mixed mode test configurations can be found in (Szekrnyes and
Jzsef 2006; Tracy et al. 2003), including, mixed mode end load split (MMELS) shown
in Figure 2-20, and crack lap shear (CLS) in Figure 2-21.



P

2h
a
L L
c
Base
Level arm
Specimen

Fig. 2-19: A schematic of MMB test configuration (Kim and Mayer 2003)

P
h
a
L
h

Fig. 2-20: A schematic of the mixed mode end load split (MMELS) specimen, based on
(Szekrnyes and Jzsef 2006)

P
P


Fig. 2-21: A schematic of the crack lap shear (CLS) specimen, based on (Tracy et al.
2003)
46

2.3.4 Mixed mode delamination failure criterion
In most realistic situations, both tensile and shear stresses can be present at the
delamination front, which result in a mixture of Mode I and Mode II loading. It is
commonly observed that Mode I fracture dominates the fracture failure process of
isotropic materials, however, both tensile and shear fracture can be significant in
laminated composites. Many delamination failure criteria proposed by researchers are
based on the decomposition of the total SERR, G
T
, into separate modes, G
I
, G
II,
and G
III
,
and an empirical relation between the applied SERRs and the critical value of each mode
(G
Ic
, G
IIc
, and G
IIIc
). However, there is some debate over whether Mode II fracture
toughness is a true material property (O'Brien 1998b; Tay 2003). Based on microscopic
examination, it has been suggested that the so called pure Mode II tests may be in fact
mixed-mode tests with variable local Mode I contributions. This complicates the
establishment of a general delamination failure criterion which can be suitable for all
materials at various mode ratios. Hence, there are many different forms of delamination
onset criterion found for different materials. Among them, the most commonly used ones
are listed below:
The linear criterion (Reeder and Crews 1991), by Eq. (2.63):
The power law criterion (Reeder 1993), by Eq. (2.64):
The B-K law (Benzeggagh and Kenane 1996), by Eq. (2.65) and Eq. (2.66):
The bilinear criterion (Reeder 1993), by Eq. (2.67) and Eq. (2.68):
1 =

IIc
II
Ic
I
G
G
G
G

(2.63)
1 =


IIc
II
Ic
I
G
G
G
G
(2.64)
II I T
G G G + = (2.65)
( )
m
T
II
Ic IIc Ic Tc
G
G
G G G G

+ = (2.66)
47

where, and are the slopes of the two line segments used in the bilinear criterion.
Kim and Mayer (Kim and Mayer 2003) used a relation by Eq. (2.69) :
Some other criteria addressing Mode I and Mode II interaction were developed by
Hashemi et al. (Hashemi et al. 1990a; Hashemi et al. 1991). Reeder and Crews (Reeder
and Crews 1991) found that a linear form is adequate for AS4/PEEK composites. Also,
Rhee (Rhee 1994) obtained experimental results indicated a linear form G
I
-G
II
behavior.
Ducept et al. (Ducept et al. 1997) developed a nonlinear failure locus for glass/epoxy.
Nonlinear failure loci was also developed by Kinloch et al. (Kinloch et al. 1993; Reeder
1993), and Singh et al.(Singh and Partridge 1995) for carbon/epoxy. Mathews and
Swanson (Mathews and Swanson 2007) found that both the linear and power law criteria
considered give a reasonable representation of the experimental delamination fracture
data for AS4/3501-6 carbon/epoxy. The fact that so many different mixed-mode criteria
have been suggested and used indicates that there is still much to be understood about the
mixed mode failure mechanism.
2.4 Interlaminar Fracture Toughness Test Under Cyclic loading
For metallic materials, it is common practice to use Paris law (Paris 1964) to
predict crack growth rate under a certain stress intensity factor range, K. It is also
assumed that there is a threshold level in the stress intensity amplitude below which no
crack growth will take place. Above this threshold value, fatigue crack propagation is
governed by Eq. (2.70),
Ic
m
IIc
m
Ic
G G G + = (2.67)
IIc
m
IIc
m
Ic
G G G = (2.68)
( )
k
II
Ic IIc Ic c
G
G
G G G G

+ = 2 (2.69)
( )
p
K A
dN
da
= (2.70)
48

where da/dN is the fatigue crack growth rate, A and p are material property parameters.
For fiber reinforced composites, the stress field near the crack tip is oscillatory,
and hence the SERR, or crack driving force, G, is more commonly used for composite
materials. A relationship similar to Eq. (2.70), based on SERR, was proposed by Wilkins
et al. (Wilkins et al. 1982) to predict the crack growth rate for composite materials, as in
Eq. (2.71).
For tests at small R, G
min
is small compared to G
max
, and hence the expression for G can
be written as, Eq. (2.72),
By this approach, the modified Paris law becomes Eq. (2.73),
Based on the modified Paris law, a total fatigue life model of composite materials was
suggested by Shivakumar et al. (Shivakumar et al. 2006). A brief description of this
model is given next.
Similar to metallic materials, the fatigue delamination process of a composite
material is often characterized in terms of the relationship between the crack growth rate
per cycle and the applied SERR as visualized on a log-log plot (Figure 2-22). The first
region of the crack growth process occurs at low SERR values and consists of either no
crack growth below a certain threshold of SERR or a decelerating growth rate with
increased SERR on the log-log scale. The second region, occurring at increased SERR
values, shows a linear relationship between crack growth rate and SERR on the log-log
scale. The third region, occurring at still higher SERR values, corresponds to markedly
increasing crack growth rate with increasing SERR, leading eventually to unstable crack
growth in a single loading cycle as SERR approaches its critical value. According to
Shivakumar et al., the delamination growth rate depends on microscopic details of
interactions between the fibers and resin in Region 1, crack driving force (strain energy
( )
n
G B
dN
da
= (2.71)
max min max
G G G G =
(2.72)
( )
n
G B
dN
da
max
= (2.73)
49

release rate G or G) in Region 2, and on interlaminar fracture characteristics of the
laminate in Region 3. In Figure 2-22, the delamination growth rate curve is bounded on
the left by G
max
= G
th
and on the right by G
max
= G
R
, where G
max
is the maximum cyclic
strain energy release rate, G
th
is the threshold strain energy release rate, and G
R
is the
interlaminar fracture toughness resistance as the delamination grows.
The total life model for Mode I fatigue growth proposed by Shivakumar et al.
(Shivakumar et al. 2006) is given by Eq. (2.74).


Regime
1
Regime
2
Regime
3
m
Log A
Log(G
max
/G
R
)
Log(da/dN)
G
max
= G
th
G
max
= G
R

Fig. 2-22: A total fatigue life model of composite materials, based on (Shivakumar et al.
2006)

=
2
1
Im
Im
Im
1
1
D
IR
ax
D
ax
Ith
m
IR
ax
G
G
G
G
G
G
A
dN
da

(2.74)
50

According to the model, the parameters needed to characterize interlaminar crack growth
under cyclic loading include: 1) fatigue interlaminar fracture property parameters: A, m,
D
1
, D
2
and threshold SERR, G
th
, at or below which the delamination growth rate, da/dN,
is nearly zero; 2) static interlaminar fracture parameters, G
R
(as a function of Mode I
interlaminar fracture toughness and delamination length). In Region 1, the delamination
growth rate, da/dN, is dominated by D
1
if the maximum strain energy release rate G
max
is
greater than G
th
. In Region 2 (stable crack growth domain), the crack growth rate, da/dN,
follows the power law dependency on either the maximum strain energy release rate,
G
max
, or range of strain energy release rate, G, under cyclic loading. The fatigue
interlaminar properties of material can be characterized by parameters in this power law
relation (A and m in Eq. (2.74)). In Region 3, the dominant parameter is D
2
as long as the
maximum strain energy release rate G
max
is less than the interlaminar fracture toughness
G
R
.
Substantial research work (Asp et al. 2001; Blanco et al. 2004; Gregory and
Spearing 2005; Gustafson and Hojo 1987; Hojo et al. 1994; Hojo et al. 2006; Matsubara
et al. 2006; Nakai and Hiwa 2002; Ramkumar and Whitcomb 1985; Tanaka 1997;
Vinciquerra et al. 2002) has contributed to the characterization of delamination laws in
Region 2. For carbon fiber composites, the threshold toughness is often found to be
important since the slope of the stable crack growth region is very high. Hence, research
work has also focused on determining the threshold SERR (Asp et al. 2001; Gregory and
Spearing 2005; Hojo et al. 1987; Tanaka 1997). The fatigue fracture resistance of
laminated fiber composite materials has been found to be closely related to matrix
properties, matrix toughness and the strength of the fiber/matrix interface. The effect of
matrix material on delamination growth was studied by Hojo et al. (Hojo et al. 1994).
The effect of fiber surface treatment was studied by Hojo et al. (Hojo et al. 2001). For the
same laminate material, the delamination growth law has been found to depend on the
stress ratio (Matsubara et al. 2006), the loading frequency and environment (Gustafson
and Hojo 1987; Nakai and Hiwa 2002), and mode ratio (Gustafson and Hojo 1987).

51

2.4.1 Fatigue delamination growth models
Since the delamination mechanism is considered to be different for different
modes of crack tip stress state, the analysis of the crack growth rate under cyclic loading
should be considered in terms of the individual operative modes.
2.4.1.1 Pure Mode I or II
In the stable crack growth region, fatigue crack growth rate in pure Mode I or
Mode II loading is typically related to range of strain energy release rate by the modified
Paris law, Eq. (2.75):
where, i = I or II for Mode I and Mode II, respectively, B
i
and n
i
are material constants, a
is the crack length, N is loading cycle, and
max min
G G G = is the range of strain energy
release rate.
In the case of homogeneous materials, such as metals and plastics, Eq. (2.75) is
often found to be in the form of Eq. (2.70), where, the range of stress intensity factor, K,
instead of strain energy release rate is often used (Gustafson and Hojo 1987).
Assuming that the material is homogeneous and isotropic, the stress intensity factor,
K
i
, can be related to the strain energy release rate, G
i
, (i = I, II) by Eq. (2.76).
where H
i
is a function of the elastic constants of the material (E
ij
,
ij
) (Hojo et al. 1987).
The stress ratio, R, is defined by Eq. (2.77) (Hojo et al. 1994):
When R is positive, substituting Eq. (2.76) into Eq. (2.77) yields, Eq. (2.78) :
( )
i
n
i
G B
dN
da
= (2.75)
2
i i i
K H G = (2.76)
max
min
K
K
R =
(2.77)
52

When R is small, G
min
is small compared to G
max
, and thus the power law relation can be
rewritten as Eq. (2.73) (Asp et al. 2001).
Many studies have found that the exponent in the power law relationship for
composite materials is much higher than that for metals. This indicates that small
uncertainties in applied loads will lead to large uncertainties in predicted delamination
growth rate.
2.4.1.2 Mixed-mode
Several models exist for relating loading and delamination growth rate. Some
typical models used are given from Eq. (2.79) to Eq. (2.82).
Ramkumar and Whitcomb, by Eq. (2.79) (Ramkumar and Whitcomb 1985):
Gustafson and Hojo, by Eq. (2.80) (Gustafson and Hojo 1987):
Russell and Street, by Eq. (2.81) and Eq. (2.82) (Russell and Street 1989):
where,
In these equations, B and n are material parameters.

max
min
G
G
R =
(2.78)
II I
n
IIc
II
II
n
Ic
I
I
G
G
B
G
G
B
dN
da

= (2.79)
( ) ( )
II I
n
II II
n
I I
G B G B
dN
da
+ = (2.80)
m
n
c
m
G
G
B
dN
da


= (2.81)
IIc
II
Ic
I
c
G
G
G
G
G
G
+


(2.82)
53

2.4.2 IFT fatigue test methods
2.4.2.1 Fatigue threshold strain energy release rate determination
The fatigue crack growth onset test can be used to determine threshold strain
energy release rate values. The threshold strain energy release rate value is defined as the
maximum SERR value below which no delamination onset occurs after a certain number
of cycles (e.g. 10
6
cycles in (Shivakumar et al. 2006), 1,800,000 cycles in (Asp et al.
2001)). This SERR value can be expected to depend on the fidelity of the crack length
measurement technique. By conducting fatigue onset tests, a plot of the maximum SERR
values, G
max
, versus number of cycles at delamination onset, N, can be obtained. A curve
is then fitted to the data. Based on this curve, the cyclic strain energy release rate to cause
delamination onset after a specified number of cycles can be predicted. A typical G-N
plot from Mode I delamination onset tests is given in Figure 2-23. A threshold SERR,
G
Ith
, can thus be determined for a selected number of loading cycles for no crack growth.
ASTM D6115 (ASTM D6115-97 1997(R2004)) was established for Mode I fatigue
delamination onset testing. For many kinds of composite materials, the exponent of the
Paris law relation by Eq. (2.71) or Eq. (2.73) is very high. This implies that a small error
in SERR calculation will result in a large error in crack growth and service life prediction,
if such a material is used in a structure. Hence, for such cases, the no-growth design
criterion is suggested and the threshold SERR is proposed to be used as the fatigue
interlaminar fracture toughness.
54

2.4.2.2 Fatigue delamination growth test
Fatigue delamination growth tests are used to establish a relation for crack growth
rate prediction in the form of Eq. (2.71) or Eq. (2.73). Before conducting a fatigue test, a
quasi-static test of the corresponding deformation mode is usually needed to obtain an
estimate of static fracture properties, e.g. critical load and displacement at crack onset,
and critical strain energy release rate for the type of material investigated. The same
specimen type and test configuration are then used for fatigue tests.
Fatigue tests can be either displacement controlled (Blanco et al. 2004; Gregory
and Spearing 2005) or load controlled (Bureau et al. 2002; Nakai and Hiwa 2002;
Vinciquerra et al. 2002). However, the displacement control method is more commonly
used because it is considered to be more likely to create stable crack growth. To obtain a
single power law relation for crack growth prediction as by Eq. (2.71) or Eq. (2.73),
stress ratio,
min max
/ R G G = , (or
max min
/ for displacement controlled tests; for linear
elasticity and small deflections (/a < 0.4),
max min
/ is identical to the R-ratio), are kept

G
Imax
J/m
2
G
I
N
a
N, cycles to delamination growth onset
G
Imax
J/m
2
G
I
N
a
N, cycles to delamination growth onset

Fig. 2-23: The Mode I delamination onset SERR versus number of cycles, based on
(ASTM standard D6115-97 1997(R2004))
55

constant during a test. Typical load ratios used are within the range of 0.1-1, and typical
cyclic loading frequencies are within 1-10 Hz, for Mode I, Mode II, and mixed-mode I/II
tests. Typical test configurations for each type of fatigue test are the same as those
presented for quasi-static testing in Section 2.1-2.3.
For data acquisition, the crack growth rate and SERR values need to be obtained
on a regular basis. The crack growth rate is obtained by recording cycles and crack
growth length during fatigue test. Two methods are generally used to measure the
delamination length and data are usually taken either at certain intervals of delamination
length or certain intervals of number of cycles. One way to measure delamination length
is to obtain all the data optically using a microscope. The other way is to use the
compliance calibration method. In the second method, a compliance versus crack length
relationship is obtained for this material either based on quasi-static tests or fatigue tests.
Then the crack length at certain number of cycles is calculated based on the measured
compliance. Hence load-displacement curves are recorded to obtain the compliance at
selected cycle intervals. Measurements can be taken periodically until the crack
propagation rate reaches a certain level, e.g. below 10
-8
m/cycle, or the number of load
cycles reaches 10
5
(this criterion was used in (Gregory and Spearing 2005)). The SERR is
then calculated based on the maximum load, load amplitude, crack length, displacement,
etc., at certain numbers of cycles.
2.5 Finite Element Modeling of Crack Propagation
The widely used techniques for modeling crack propagation interface in
composite materials include: modeling the crack interface with the virtual crack closure
technique (VCCT), with a tiebreak contact, and with a cohesive layer (Meo and Thieulot
2005). Among them, only the virtual crack closure technique is a fracture mechanics
based method. This technique is discussed in more detail in the following paragraph. In
the second technique, coincident nodes are tied together with a constraint relation and
remain joined until the maximum interlaminar stresses are reached. Once this maximum
value is exceeded, the nodes associated with the constraint are released to simulate the
56

initiation of delamination. Meo and Thieulot found that poor results were obtained when
using the tiebreak contact technique to model the delamination behavior of the double
cantilever beam specimen (Meo and Thieulot 2005). The cohesive element technique,
which is capable of simulating the initial elastic behavior, damage initiation, and damage
evolution of the material, is used to model the cohesive layer. However, in order to
simulate the delamination behavior correctly, criteria and parameters defining each
loading stage of the cohesive element need to be selected appropriately. In the following
sections, fundamentals and basic principles for using VCCT and cohesive elements are
briefly reviewed.
2.5.1 The Virtual Crack Closure Technique (VCCT)
The virtual crack closure technique (VCCT) implemented in finite element
analyses is gaining popularity. This is especially true in aerospace engineering, because
of the practical computation time and suitability of the technique for modeling the
behaviors of anisotropic materials. VCCT predicts crack propagation by comparing local
SERR to the critical SERR (fracture toughness), and gives results of SERR from node to
node in a finite element model. VCCT is available as an add-on capability of Abaqus
V6.7-1 to efficiently simulate delamination behaviors for composite materials. Using
VCCT for Abaqus, an engineer can simulate the crack propagation process and estimate
the residual strength of a composite structure after damage. VCCT for Abaqus is based
on patent-pending technology by Abaqus, Inc. (VCCT for Abaqus Manual 2007).
2.5.1.1 The virtual crack closure technique formulation
Using VCCT, the crack propagation is modeled with a contact interaction defined
between a pair of crack surfaces. Therefore, the crack path is predefined using VCCT.
Crack onset or growth prediction is achieved by comparing the calculated SERR (G) with
the interlaminar fracture toughness property (G
c
) of the material. VCCT is based on the
57

assumption that the energy released when the crack is extended from a to a+a is
identical to the energy required to close the crack (Figure 2-24, (Krueger 2002)).

The VCCT method assumes that a crack extension of a from a+a (node i) to
a+2a (node k) does not significantly alter the state at the crack tip. Therefore, the
displacements behind the crack tip at node i are approximately equal to the displacements
behind the original crack tip at node l. The basic formulae for SERR calculation by using
VCCT are presented next.



X
1l
Z
1l
l i
a Da Da
Crack closed
x, u, X
z, w, Z
l
i
a Da Da
Crack extended
x, u, X
z, w, Z
Du
2l
Dw
2l

Fig. 2-24: Crack closure for VCCT, based on (Krueger 2002)
58

(1) 2D 4-node element
In a two-dimensional plane stress or plane strain model, the Mode I and II
components of SERR, G
I
and G
II
, for a four node element are calculated by Eq. (2.6) and
Eq. (2.7), (Figure 2-25, (Krueger 2002)):
where, a is the length of the elements immediately in front of the crack along the crack
growth direction. X
i
and Z
i
are the forces at the crack tip (node point i). u
l
and w
l
are the
nodal displacements behind the crack tip of the upper crack face, and u
l*
and w
l*
are the
nodal displacements behind the crack tip of the lower crack face. (Figure 2-25). G
III
= 0
for two-dimensional cases. The crack is represented as a one-dimensional discontinuity
by a line of nodes on the top and bottom fracture surfaces. Nodes on bottom and top
fracture surfaces initially have identical coordinates before loading. As loading proceeds,
the bonded nodes are released along the crack sequentially.

( )
*
2
1
l
l i I
w w Z
a
G = (2.83)
( )
*
2
1
l
l i II
u u X
a
G = (2.84)


Fig. 2-25: VCCT for four-node 2D element (plane strain or plane stress), based on
(Krueger 2002)
59

(2) 3D solid 8-noded element
The Mode I, Mode II and Mode III components of SERR, G
I
, G
II
and G
III
, for an
eight node element are calculated by Eq. (2.85), Eq. (2.86) and Eq. (2.87), respectively.
where, b a A = as shown in Figure 2-26. Other notations are defined in Figure 2-26
and Figure 2-27.

( )
*
2
1
Ll Ll Li I
w w Z
A
G

= (2.85)
( )
*
2
1
Ll Ll Li II
u u X
A
G

= (2.86)
( )
*
2
1
Ll Ll Li III
v v Y
A
G

= (2.87)

w
Ll
b
v
Ll
u
Ll
w
Ll*
v
Ll*
u
Ll*
Z
Li
X
Li
Y
Li
L
l
l*
i
a
a
a
x, u,
X
z, w, Z
z, w,
Z
x, u,
X
y, v,
Y
Global
system
Local crack
tip system

Fig. 2-26: VCCT for eight-node solid element (3D view) (Krueger 2002)
60

2.5.1.2 Crack growth criterion for VCCT
Crack onset or growth is predicted by comparing the calculated SERR, G, with
interlaminar fracture toughness properties of material, G
c
. The SERR along the
delamination front is calculated at the end of a converged increment. Once the SERR
exceeds the critical SERR, the node at the crack tip is released at the following
increment, which allows the crack to propagate. To avoid sudden loss of stability, the
force at the crack tip before advance is released gradually during succeeding increments
in such a way that the force is brought to zero no later than the time at which the next
node along the crack path begins to open.
For pure Mode I case, debonding of nodes occurs when: (VCCT for Abaqus
Manual 2007)

b
/
2
b
/
2
b
b
l i
l
a a a
i
v
Ll
u
Ll
X
Li
Y
Li
y, v, Y
x, u, X
L
Delamination front
Delaminated
area
x, u,
X
y, v,
Y
Global
system

Fig. 2-27: VCCT for eight-node solid element (top view) (Krueger 2002)
61

For mixed mode, debonding of nodes occurs when:
VCCT for Abaqus releases a bonded node at the crack tip when Eq. (2.89) is satisfied,
however, when the ratio of the difference between the computed and critical SERR to the
critical SERR exceeds the specified release tolerance, as described by Eq. (2.90), a
cutback operation (rerun the current iteration with smaller time increment) is requested
by VCCT for Abaqus. (VCCT for Abaqus Manual 2007)
In the above two equations, G
equiv
is the equivalent SERR calculated at a node in the
general mixed mode case and G
equivC
is the critical SERR calculated based on the user-
specified mode-mix criterion and the critical SERR of the interface.
For mixed mode fracture, VCCT for Abaqus provides three mixed-mode
criterions to model failure. The users can choose from three of the most commonly used
mixed mode fracture criterion, the B-K law by Eq. (2.91), the Power law by Eq. (2.92),
and the Reeder law by Eq. (2.93).
G
I
> G
Ic
(2.88)
G
equ
> G
equC
(2.89)
tolerance release >

equivC
equivC equiv
G
G G

(2.90)

+ +
+
+ =
III II I
III II
IC IIC IC equivC
G G G
G G
G G G G ) ( (2.91)
ao
IIIC
III
an
IIC
II
am
IC
I
equivC
equiv
G
G
G
G
G
G
G
G

= (2.92)
( )
( )

+ +
+

+
+

+ +
+
+ =
III II I
III II
III II
III
IIC IIIC
III II I
III II
IC IIC Ic equivC
G G G
G G
G G
G
G G
G G G
G G
G G G G


(2.93)
62

For the B-K law and the Reeder law, G
IC
, G
IIC
, G
IIIc
and are material properties to be
specified by users to define the model. For the Power law, G
IC
, G
IIC
, G
IIIc
, am, an, and
ao are material properties need to be specified.
2.5.2 Cohesive element (Abaqus 2007)
Besides VCCT, cohesive elements are sometimes used to model fracture behavior
of laminated composites. The advantage of cohesive elements over VCCT is that the
former does not require the existence of an initial crack in the material. However, extra
material property parameters for defining elasticity, damage initiation and evolution
criterion, are needed for cohesive elements. This makes the usage of cohesive element
more versatile, but more complicated as well.
Cohesive elements are capable of modeling the initial elastic loading, the
initiation and the propagation of damage leading to eventual failure at a bonded interface.
Unlike the usual elements, the thickness of a cohesive element can be as small as zero
before loading is applied. The model need not have any crack to begin with. However,
the cracks are restricted to propagate along the layer of cohesive elements.
Unlike usual elements, the constitutive response of cohesive elements can be
defined using a traction-separation law. A typical traction-separation response of a
cohesive element, as shown in Figure 2-28, involves three stages: initial elastic loading,
damage initiation and damage evolution. The constitutive thickness used for traction-
separation response is generally different from the geometric thickness, which is typically
close or equal to zero. By default, if the traction-separation law is specified, Abaqus sets
the constitutive thickness of the element to be 1.0, so that, the nominal strain is equal to
the relative normal displacements (separation) between the top and bottom faces. The
nominal stresses are the force components divided by the original area at each integration
point, while the nominal strains are the separations divided by the original thickness at
each integration point.
63

2.5.2.1 Elastic behavior of the cohesive element
The nominal traction stress vector, t, consists of three components in three-
dimensional problems: t
n
, t
s
, and t
t
, which represent the normal and the two shear
tractions, respectively. The corresponding separations are denoted by
n
,
s
, and
t
.
Denoting by T
o
the original thickness of the cohesive element, the nominal strains are
defined as Eq. (2.94),
The elastic behavior can then be written as,

( )
0 0 0
,
t s n
t t t
( )
0 0 0
,
t s n

( )
f
t
f
s
f
n
,
Traction
Separation
Initial
loading
Damage
initiation
Damage
evolution

Fig. 2-28: Traction-separation response of cohesive element, based on (Abaqus 2007)
o
t
t
o
s
s
o
n
n
T T T

= = = , ,
(2.94)
K =

=
t
s
n
tt st nt
st ss ns
nt ns nn
t
s
n
K K K
K K K
K K K
t
t
t
t


(2.95)
64

The terms in the stiffness matrix, K, are material properties. The off-diagonal terms in the
stiffness matrix are zero if uncoupled behavior between the normal and shear components
is desired.
The traction-separation response can be better understood by comparing it to the
load-displacement response of a rod, as described by Eq. (2.96).
or Eq. (2.97)
where the end displacement of the truss is . The length is denoted by L, the elastic
stiffness by E, the original area by A, and the axial load by P. A P S / = is the nominal
stress and L E K / = is the stiffness that relates the nominal stress to the displacement. If
the bulk adhesive material of the cohesive elements has stiffness E
c
, the stiffness of the
interface is given by
c c c
T E K / = . When the constitutive thickness of the cohesive layer
is artificially set to 1.0, ideally K
c
should be specified as the material stiffness as
calculated with the true thickness of the cohesive layer. As the thickness of the interface
layer is often very small, the equation
c c c
T E K / = implies that the stiffness, K
c
, tends to
infinity. This stiffness is sometimes chosen as a penalty parameter. Very large penalty
stiffnesses may result in ill-conditioning of the element operator in Abaqus/Standard. To
ensure the cohesive elements will have no adverse effect on the stable time increments,
the Abaqus manual suggests to choose,
e c
K K 1 . 0 = , where K
c
is the cohesive element
stiffness, and K
e
is the stiffness of surrounding material.
AE
PL
= (2.96)
K
S
= (2.97)
65

2.5.2.2 Damage initiation criteria of the cohesive element
Four criteria can be used for defining damage initiation: maximum nominal stress,
maximum nominal strain, quadratic nominal stress, and quadratic nominal strain criterion.
The mathematical forms of these criteria are given by Eq. (2.98) to Eq. (2.101).
In the above equations, t
n
0
, t
s
0
, and t
t
0
are ultimate stresses in the normal, first and second
shear directions, respectively, as shown in Figure 2-28.
n
0
,
s
0
, and
t
0
are ultimate strains
in the normal, first and second shear directions, respectively, as shown in Figure 2-28.
2.5.2.3 Damage evolution criteria of the cohesive element
For damage evolution many criteria are available for Abaqus. Generally, they can
be divided into linear and nonlinear categories. For nonlinear damage evolution criteria,
the Power Law form, given by Eq. (2.102), and the B-K form, given by Eq. (2.103), are
commonly used.
Maximum nominal stress:
0 0 0
max , , 1
n s t
n s t
t
t t
t t t

=



(2.98)
Maximum nominal strain:
0 0 0
max , , 1
n s t
n s t




=



(2.99)
Quadratic nominal stress:
2 2 2
0 0 0
1
n s t
n s t
t t t
t t t

+ + =


(2.100)
Quadratic nominal strain:
2 2 2
0 0 0
1
n s t
n s t



+ + =


(2.101)
1
n s t
C C C
n s t
G G G
G G G


+ + =


(2.102)
( )
C C C C S
n s n
T
G
G G G G
G


+ =


(2.103)
66

In the Power law form, G
n
, G
s
, and G
t
are the work done by the traction and its conjugate
relative displacement in the normal, the first and the second shear directions, respectively.
The quantities G
n
C
, G
s
C
, and G
t
C
are the critical fracture energies required to cause failure
in the normal, the first, and the second shear directions, respectively. In the B-K form,
S s t
G G G = + ,
T n s
G G G = + , and is a material parameter. The mixed mode fracture
energy
C
n s t
G G G G = + + . The user needs to specify G
n
C
, G
s
C
, and .
2.6 Preview of the Following Chapters
Similar to using the traditional failure criteria of materials, crack onset and growth
is predicted based on comparing the local crack driving force (G) with the material
fracture resistance (G
c
). The local drive force, or named strain energy release rate
(SERR), G, represents the local loading conditions, while the fracture resistance, G
c
,
represents the material strength. To predict delamination behavior accurately, firstly, a
repeatable and conservative value of the fracture toughness needs to be characterized as a
material property. Secondly, a reliable and effective method for evaluating local SERR is
required. The interlaminar fracture toughness (IFT) test is the type of test to characterize
the capability of a laminated composite material to resist fracture. The finite element
method (FEM) is potential to accurately predict local SERR distribution and
delamination behavior based on a fracture mechanics approach. For most of the currently
popular commercial FE software, the SERR based failure and crack propagation analysis
methods have not yet been incorporated, except for Abaqus. During this investigation, the
fracture behavior in Mode I, Mode II, and mixed Mode I/II of a proprietary carbon/epoxy
composite material system were characterized under static and fatigue loadings. The
methods and results for Mode I, Mode II, and Mixed Mode I/II testing are presented in
Chapter 3, 4 and 5, respectively. Additionally, in an attempt to preliminary assess the
capability of Abaqus in solving crack propagation problems, several finite element
models of the DCB specimen were analyzed using VCCT for Abaqus. The modeling
technique and results obtained for finite element modeling are presented in Chapter 6.
67


Chapter 3

The Mode I Interlaminar Fracture Toughness Testing
Mode I interlaminar fracture toughness (IFT) tests were conducted under quasi-
static and cyclic loading conditions with the Double Cantilever Beam (DCB) specimen.
In this chapter, DCB test configurations and methods used in this investigation are
introduced. The Mode I interlaminar fracture toughness at crack onset and fracture
resistance with crack extension were characterized with DCB quasi-static specimens.
Mode I fatigue tests were conducted in displacement control with displacement ratio R =
0.2. Crack growth rate (da/dN) was related to maximum Mode I strain energy release rate
(G
Imax
) by the power law relation ( )
I
n
ax I
G B dN da
Im
/ = .
3.1 Material, Specimen and Test Configuration
The specimens used in this investigation are machined from a flat carbon/epoxy
panel of [0]
12
lay-up as shown in Figure 3-1. A Teflon film of 12.7 m thickness was
inserted at the mid-plane during the panel lay-up process to define the initial starter crack.
The DCB specimens were cut from the panel at the Penn State Composites Lab using a
water-cooled diamond abrasive cut-off wheel. The distribution of tested DCB specimens
in the panel is shown in the panel diagram, Figure 3-1.
68

The DCB specimen geometry and notations are shown in Figure 3-2. Five
specimens (Specimen 1-1, 1-2, 1-3, 1-4 and 1-5 in Figure 3-1) were tested in quasi-static
loading and four specimens (Specimen 4-2, 4-3, 4-4, and 4-5 in Figure 3-1) in fatigue.
The specimen dimension of length width thickness is approximately 150 25.4 3.9
mm, with an initial artificial crack created by the embedded thin film of about 76.2 mm
length. Specific dimensions of each specimen are listed in Appendix A Table A-1 and
Table A-2, for static and fatigue case, respectively.



00.0005 Teflon film
0-degree fiber direction
Specimen 1-1
Specimen 1-2
Specimen 1-3
Specimen 1-4
Specimen 1-5
Specimen 4-2
Specimen 4-3
Specimen 4-4
Specimen 4-5

Fig. 3-1: Diagram of the DCB panel
69

3.2 The Mode I Quasi-static IFT Testing
Although an ASTM standard (ASTM D5528-01 2002) has been established for
the Mode I quasi-static IFT testing, certain issues are still debated in the literature, such
as whether to use an initiation G
Ic
value from a Mode I precrack or from an insert film,
and whether it is advantageous to use several reloading/unloading cycles or to use just
one reloading/unloading cycle. The test methods used in this investigation follow the
general suggestions by the ASTM standard; however, certain modifications were made
for the special properties of the tested material and also to compare the results of different
methods.

a0
l
2h
b
Piano
hinge
Insert
film

Fig. 3-2: DCB specimen geometry and notation
70

3.2.1 Mode I quasi-static test method
The Mode I interlaminar fracture toughness tests using DCB specimens were
performed according to ASTM standard D5528 (2002). The specimens were loaded by a
13.5 kN (3 kip) servo-hydraulic MTS machine. The load was measured using a 110 N (25
lb) load cell and the displacement was measured by the MTS LVDT built into the
actuator. A photograph of the test setup is shown in Figure 3-3.
Loading was applied to the specimen through piano hinges. Before the test, the
piano hinges were bonded to specimen surfaces using M-bond adhesive. A thin layer of
white correction fluid (slightly transparent) was painted on one specimen edge for
convenience of crack growth observation. The position of the end of the insert film was
marked on the painted edge before testing, and pencil markings were made by hand at

Top Grip (stationary)
110 N Load Cell
Universal Joint
Specimen
Bottom Grip (actuated)
Optical Fiber Light Source
Not shown:
- Servo-hydraulic actuator
- Long distance instrumented
stage microscope
Top Grip (stationary)
110 N Load Cell
Universal Joint
Specimen
Bottom Grip (actuated)
Optical Fiber Light Source
Not shown:
- Servo-hydraulic actuator
- Long distance instrumented
stage microscope

Fig. 3-3: Photograph of DCB test set up
71

every one millimeter in the first 5 mm beyond the insert film, and every 2 or 3 mm in the
next 45 mm. The exact distances from the first marking (insert film tip marking) to the
following markings were measured through a Questar QM-1 long distance telescope
before the test.
One modification to the loading/unloading method was made to the ASTM
standard D 5528 test method. In the current investigation, a DCB specimen was subjected
to five successive loading and unloading cycles, instead of one initial loading/unloading
cycle and one reloading/unloading cycle as specified in the ASTM standard. For the first
loading cycle, loading was stopped and unloading started after the crack growth had
reached 5 mm or unstable crack growth had arrested. For each of the next four loading
cycles, the loading was stopped and unloading started after 10-15 mm crack growth had
been created in a loading cycle.
During the test, the specimen was loaded at a constant displacement rate of 1
mm/min and unloaded at 2 mm/min. Crack tip position was pinpointed with a Questar
QM-1 long distance telescope mounted on an instrumented stage with 0.001 mm
resolution. Load and displacement data were recorded continuously for each loading
cycle, together with the elapsed time from the beginning of the test, by a digital data
acquisition system. The distance from each subsequent marking to the first marking,
where the initial crack is, was measured by the Questar telescope before the test. The
crack length at each subsequent marking was calculated by the sum of the initial crack
length and the distance to the first marking (initial crack front marking). The elapsed time
from the beginning of test to the moment at which the crack propagates to each marking
was recorded during the test, so that after the test the crack length could be matched with
the concurrent load and displacement based on the data recorded by the data acquisition
system.

72

3.2.2 Mode I quasi-static test results
The characterization of the fracture resistance is based on the load-displacement
plot generated after the tests. From this plot, the compliance values of the specimen at
several crack lengths were obtained and related to crack lengths by several models. This
compliance vs. crack length relation was then used for critical strain energy release rate
(SERR) calculation, together with the load and/or displacement recorded during the test.
Some general observations of crack growth during the tests are: the crack growth
was typically unstable immediately after initiation in the first loading cycle, while the
crack growth in other reloading/unloading cycles was typically quasi-stable and very fast,
especially at shorter crack lengths. As the crack extended, the growth rate tended to slow
down. This trend coincides with the prediction by the crack length vs. displacement
relationship from the modified beam theory, Eq. (3.1) and Eq. (3.2) (assuming constant
fracture resistance with crack extension).
where, a
eff
is the effective crack length taking into consideration rotation of the cantilever
beam at the built-in end. Also, G
IR
is the fracture resistance as the crack grows; m is the
coefficient from the compliance-crack length relation in Eq. (2.1); b is the specimen
width, and is the opening displacement. For a displacement controlled test, d(a
eff
)/d is
proportional to the speed of crack growth (d(a
eff
)/dt, where t is time). As implied by
Eq. (3.2), d(a
eff
)/d and thus the speed of crack growth decreases as the loading
displacement increases.

2 / 1
4
3
2
3

IR
eff
G bm
a a = + =
(3.1)
( )
2 / 1
4
3
2
1
2
3

= =

IR
eff
G bm d
da
d
a d

(3.2)
73

3.2.2.1 Load-displacement curves
The load-displacement curve consists of two parts. The first part is the linear part,
where load is linearly related to displacement. During this portion of test, there is neither
crack extension nor nonlinear material behavior in the specimen as loading proceeds. For
this part, simple or advanced beam theory can be used to predict the loading curve. The
second part is the nonlinear part, where there is either nonlinear material behavior or
local/global crack growth in the specimen. If nonlinear material behavior or local crack
growth occurs in the specimen, the loading curve deviates slightly from the straight line
that can be fitted into the initial linear part of loading curve. However, if global crack
growth occurs, the loading curve exhibits a marked deviation from the straight line fitted
into initial loading curve. For the global crack growth regime, fracture mechanics
approach can be used to predict loading curve.
For the linear regime, the slope of the load-displacement curve is mainly related
to the modulus in the longitudinal direction of specimen, E
1
, and the crack length, as by
Eq. (3.3), which is derived from MCC expression of compliance (Section 2.1.1.2).
where
0
is given by Eq. (2.17), and involves the material properties E
1
, E
3
, and G
13
.
For the nonlinear regime, if the nonlinearity is caused by global crack growth, the
trend of the loading curve is dependent on fracture resistance of the specimen with crack
extension. Derived from the modified beam theory (Section 2.1.1.1), the load of the DCB
specimen is related to the loading displacement by Eq. (3.4) if the materials fracture
toughness with crack extension is G
IR
.
A special case when the material demonstrates constant fracture toughness after a
certain length of crack extension (as shown in Figure 3-4) is considered next. Assume a
specimen demonstrates a constant fracture toughness
IR Ic
G kG = , after the crack extends
( )
3
1
3
0
1
8
2
E bh P
a h


=



(3.3)
3/ 4
1/ 2 IR
2
3
bG
P
m


=


(3.4)
74

the distance of a
1
, where G
Ic
denotes the fracture toughness at crack initiation. At crack
initiation, Eq. (3.5) is predicted by MBT.
where P
cr
and
cr
are load and displacement at crack onset, respectively. After the crack
length reaches
1 0 1
a a a = + , we have,
Comparing Eq. (3.5) to Eq. (3.6), one obtain, Eq. (3.7),
and thus the loading curve is predicted by, Eq. (3.8).
when,

3/ 4
1/ 2 Ic
2
3
cr cr
bG
P
m


=


(3.5)
3/ 4
1/ 2 IR
2
3
bG
P
m


=


(3.6)
3/ 4
1/ 2
IR
1/ 2
cr cr Ic
G P
P G


=


(3.7)
( ) ( )
3/ 4
1/ 2 1/ 2 3/ 4 1/ 2 1/ 2 IR
cr cr cr cr
Ic
G
P P k P
G



= =


i i i i (3.8)
2
1/ 2 1
1
0
cr
a
k
a

+
=

+

(3.9)

Crack length, a, mm
G,
J/m
2
G
Ic
a
0 a
1
IR Ic
G kG =

Fig. 3-4: Constant fracture toughness after certain length of crack extension
75

In Figure 3-5 and Figure 3-6, load vs. displacement data from two tests were
plotted along with constant G
IR
curves, as predicted by Eq. (3.8). The initial linear portion
of the load-displacement curve from experimental data coincides with the theoretical
prediction. For the nonlinear portion (fracture portion) of the P- curve, the data
intersects with curves of increasing G as the crack extends. The higher order constant G
IR

curves plotted coincide with the experimental data fairly well for the second, third and
fourth loading cycles of Specimen 1-1, and the third, fourth and fifth cycles of Specimen
1-4. This indicates that for the tested material, the fracture resistance tends to stay
constant after certain distance of crack extension, and a somewhat flat fracture resistance
vs. crack length curve may be obtained for this material.


0
10
20
30
40
0 2 4 6 8 10 12 14 16
Displacement, mm
L
o
a
d
,

N
G
IR
=G
Ic
G
IR
=1.6G
Ic

Fig. 3-5: Load vs. displacement plot (Specimen 1-1)
76

Two types of behavior not predicted by the theories were also observed in the
load-displacement plots. The first one is the sudden load drop immediately after crack
initiation in the first loading cycle for most tested specimens. The second one is the
stick-slip behavior of the fracture portion of load-displacement curve. These two types
of behavior are described in more details as follows.
a) Sudden load drop after crack initiation
The sudden load drop at the crack initiation point of the load vs. displacement
curve often corresponds to an unstable crack growth at crack initiation. However, this
unstable crack growth is not predicted by basic theories in the fracture mechanics
approach. Based on the amount of load drop and unstable crack growth, two types of
load-displacement behavior exist at crack onset among the tests conducted were observed.
In the first type, as loading proceeds, a relatively small amount of load drop occurs
immediately after the crack onset of the first loading cycle, as shown in Figure 3-5, and
stick-slip behavior occur after crack starts to propagate. In this case, the unstable crack
growth at crack onset is less than 5 mm. In the second type, a relatively large amount of

0
5
10
15
20
25
30
35
40
45
0 2 4 6 8 10 12 14
Displacement, mm
L
o
a
d
,

N
G
IR
=G
Ic
G
IR
=1.1G
Ic

Fig. 3-6: Load vs. displacement plot (Specimen 1-4)
77

load drop occurs immediately after crack onset, with a noticeably unstable crack growth
observed through the telescope and sometimes an audible noise. For this case, the
unstable crack growth arrests after 5-10 mm crack extension. As loading proceeds, the
load-displacement curve becomes linear and then stick-slip behavior occurs with
further crack extension. The second type of load-displacement curve is shown in
Figure 3-6.
For the first loading cycle, four out of five specimens showed a load-displacement
behavior of the second type at initial crack onset, with dramatically unstable crack growth
from the insert film. However, for the reloading cycles, the crack growth after onset was
stable but a few cases. This unstable crack growth from the insert film might imply the
formation of a resin pocket near the front of insert film, as was shown schematically in
Figure 2-7 (Chapter 2). Further, the less than 13 m thickness of insert film criterion
for initial defect type, proposed by ASTM to eliminate the resin rich zone beyond the
insert film, might not be appropriate for the material system of the current investigation.
This thickness limit might still not be small enough to eliminate the effect of resin rich
zone just in front of the insert film for some materials. If a resin rich zone exists in front
of the insert film, the perceived onset G
Ic
value may be not indicative of the fracture
toughness of the laminated material. If there exists a big difference of the fracture
resistance of the local material and the delamination resistance of the laminate, a rather
large amount of load drop might occur at the instance of initial crack onset.
Additionally, unlike load-displacement curves observed for some other materials,
e.g. some GFRP material, in which significant nonlinear behavior occurs before the load
starts to decrease, for the investigated material the loading curve is particularly linear
before the initial and subsequent crack onset.
b) Stick-slip behavior
The stick-slip behavior of the load vs. displacement curve can be explained as
follows. The load vs. displacement curve is continuous and smooth as predicted
theoretically if the crack growth proceeds slowly and continuously across the width in
infinite small increments. However, in a real test, the crack growth proceeds step by step
in uneven increments. If the crack growth jumps ahead by a small amount, a small load
78

drop will be observed in the load vs. displacement curve, which is called slip behavior
of the loading curve. Since the crack length has increased, the crack driving force (G) is
less than the material fracture toughness at the new crack length in a displacement
controlled test. As loading proceeds, the load vs. displacement curve is linear again, until
the crack driving force (G) exceeds the fracture toughness of the material and the crack
starts to propagate again. If the crack advances in finite small steps progressively, the
corresponding loading-displacement curve exhibits small zigzag shape or even is
possibly close to a smooth shape. On the other hand, if the crack grows in a large sudden
increment, an unstable crack growth behavior may be observed.
3.2.2.2 Compliance calibration
Compliance values of each specimen at certain measured crack lengths were
obtained by two methods. During the test, each DCB specimen was unloaded and
reloaded at four crack lengths in addition to an initial loading at the initial crack length.
At these crack lengths, compliance values were obtained by inverting the slope of the
linear portion of the loading curves. Meanwhile, compliance values at other crack lengths
measured without unloading/reloading were obtained by dividing the instantaneous
displacement by the instantaneous load at the selected crack lengths.
Compliance calibration was conducted individually for every test specimen. Two
expressions were used to relate compliance (C) to crack length (a). The first one by the
Modified Beam Theory (MBT) (ASTM standard D5528-01 2002), Eq. (2.1),
and the other is based on the Modified Compliance Calibration method, (JIS standard
(JIS K 7086 1993), Eq. (2.14)).
( )
3 3
+ = a m C (2.1)
0
3 / 1
1
) ( 2 / + = bC h a (2.14)
79

To determine the coefficients for compliance calibration, two plots were made as shown
in Figure 3-7 and Figure 3-8, respectively. In Figure 3-7, the cubic root of compliance
(C
1/3
) was plotted against crack length (a) along with a straight line fitted by least squares
for each specimen, so that the coefficient m of Eq. (2.1) can be determined by the slope of
the fitted line, and can be determined by the abscissa intersection. Coefficients of
Eq. (2.1) determined for each specimen are shown in Table 3-1. In the second plot,
Figure 3-8, normalized crack length, a/2h, was plotted against compliance normalized by
width, (bC)
1/3
, along with a straight line fitted by least squares for each specimen. For
each specimen, coefficient
1
is determined by the slope of the corresponding fitted line,
and
0
determined by the ordinate intersection. In the second type of plot, the coordinates
are normalized by specimen dimensions, and thus the fitted lines are not subject to
variations of specimen dimensions. Coefficients determined by line-fitting in the second
plot are listed in Table 3-2.


0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
-20 0 20 40 60 80 100 120
Crack length, a , mm
C
1
/
3
,

(
m
m
/
N
)
1
/
3
Specimen 1-1
Specimen 1-2
Specimen 1-3
Specimen 1-4
Specimen 1-5
( )
3 3
+ = a m C

Fig. 3-7: A C
1/3
versus a plot for Mode I tests
80



Table 3-1: Parameters of compliance calibration by the MBT method
Specimen
,
mm
m,
1/(mm
2/3
N
1/3
)
1-1 9.594 7.580E-03
1-2 6.626 7.315E-03
1-3 6.238 7.196E-03
1-4 8.378 7.368E-03
1-5 7.386 7.479E-03



0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
0.5 1.0 1.5 2.0 2.5 3.0
(bC)
1/3
, (mm
2
/N)
1/3
a
/
2
h
Specimen 1-1
Specimen 1-2
Specimen 1-3
Specimen 1-4
Specimen 1-5
0
3 / 1
1
) ( 2 / + = bC h a

Fig. 3-8: An a/2h versus (bC)
1/3
plot for Mode I DCB tests
Table 3-2: Parameters of compliance calibration by MCC method
Specimen
0

1

1-1 -2.5720 12.240
1-2 -1.6155 11.872
1-3 -1.5513 12.094
1-4 -2.0847 11.681
1-5 -1.8749 11.730


81

Based on the two compliance calibration plots, an interesting observation was
made. If the initial compliance vs. initial crack length data were also plotted, in the first
plot by the MBT method, this data point will fall below the fitted straight line for the
corresponding specimen, while it will fall above the fitted straight in the second plot by
the MCC method. Noticing that the compliance is involved in the y-coordinate in the
first plot and x-coordinate in the second plot, both plots seem to consistently indicate that
the specimen at the initial crack length is stiffer than that predicted by the fitted
compliance calibration relation. However, in fact, this disagreement is caused by the
dissimilarity of crack front shapes at the initial crack length and at the subsequent
propagated crack lengths, which will be explained in detail in Section 3.2.2.5.
3.2.2.3 G
Ic
onset values
In the first loading cycle, the specimen was loaded until the crack had propagated
for about 5 mm or unstable crack growth had arrested. Then, the DCB specimen was
unloaded and subject to another four loading/unloading cycles. The onset value of SERR,
G
Ic
, was determined for the crack onset from the insert film, at the critical (maximum
load) point of the first loading cycle. The crack onset from a Mode I precrack was
determined from the critical point of the first reloading cycle (second loading cycle). The
critical points were found from load vs. displacement curves. The loading curve is
typically linear up to the maximum load point, which is typically the visual crack onset
point as well. As discussed before, two kinds of load-displacement behavior were
observed at crack onset of first loading cycle, with the first one having a small amount of
load drop at the critical onset point, and the second one a large amount of load drop at the
instance of unstable crack growth of about 3-7 mm in one step. These two kinds of
behaviors correspond to the load vs. displacement curves for first loading/unloading cycle
in Figure 3-9 and Figure 3-10, respectively.
82



y = 13.427x - 0.1079
R
2
= 0.9994
y = 9.1699x - 1.9322
R
2
= 0.9998
0
10
20
30
40
50
0 1 2 3 4 5
Displacement, mm
L
o
a
d
,

N
Loading
Unloading
1st Pmax
Linear (Loading)
Linear (Unloading)
P
cr

Fig. 3-9: Load vs. displacement curve -- small load drop occurred at crack onset.
(Specimen 1-1)

y = 11.03x + 0.4456
R
2
= 0.9996
y = 16.549x + 1.6032
R
2
= 0.9997
0
10
20
30
40
50
0 0.5 1 1.5 2 2.5 3 3.5
Displacement, mm
L
o
a
d
,

N
Loading
Unloading
1st Pmax
Linear fit (Unloading)
Linear fit (Loading)

Fig. 3-10: Load vs. displacement curve -- large load drop occurred at crack onset.
(Specimen 1-3)
83

The onset G
Ic
values from starter film and precrack (first reload) were calculated
with MBT (Eq. (2.3)) and MCC (Eq. (2.15)) methods, respectively. The results from
different specimens are compared in Figure 3-11 and numerical values are available in
Appendix Table B-8. It was obtained that the onset value from the insert film calculated
by MCC was always higher than that by MBT. For Specimens 1-1 and 1-2, for which a
small load drop occurred at the crack onset, the G
Ic
value based on the precrack is higher
than that based on the insert film. The opposite holds for the other cases, for which
significant unstable crack growth occurred in the initial loading cycle. From the DCB test
results found from the literature (Appendix A, Table A-1), the G
Ic
values for crack onset
from an insert film range from 100 to 290 J/m
2
. For the investigated material, the G
Ic

onset values based on an insert film determined by different calculation methods range
from 110 to 150 J/m
2
. Compared to the results from the literature, the Mode I quasi-static
fracture toughness of the investigated material is near the low end of the common range
for carbon/brittle epoxy materials.

0
20
40
60
80
100
120
140
160
180
200
1-1 1-2 1-3 1-4 1-5
Specimen No.
G
I
c
,

J
/
m
2
MCC, insert film MBT, insert film
MCC, precrack Average, by MCC
Average, by MBT Average, precrack, by MCC

Fig. 3-11: G
Ic
onset values
84

3.2.2.4 Mode I resistance curve
During the test, the crack lengths were measured visually at certain points using
an instrumented-stage telescope with 0.001 mm resolution. After the test, strain energy
release rates were calculated at these points with the measured crack length and
compliance calibration relations. The modified beam theory (Eq. (2.3)) and modified
compliance calibration method (Eq. (2.6)) were used for SERR calculations. However,
the SERR values obtained by the two methods are very similar except for the first point
(crack onset), as shown in a typical case in Figure 3-12. An overall Mode I IFT resistance
curve fit to the data from all five specimens is shown in Figure 3-13.


0
20
40
60
80
100
120
140
160
0 10 20 30 40 50 60
a, mm
G
I
c
,

J
/
m
2
MCC method
MBT method
Onset value by MBT method
Onset value by MCC method
Power (MCC method)
Power (MBT method)

Fig. 3-12: Mode I IFT resistance curves for Specimen 1-2
85

A relatively flat resistance curve was obtained for the investigated material,
compared to other laminated material system. For example, for a glass/vinyl ester
material, Shivakumar et al. (2006) obtained that G
IR
value increased by about 140%
comparing to onset value G
Ic
after 50 mm crack extension. For all the DCB specimens in
this investigation, the toughness increased by less than 40% after about 50 mm crack
growth. The flat resistance curve is typical of brittle matrix laminates as was found in
some round robin tests (Hojo et al. 1995).
3.2.2.5 Discussion of special issues
Some special issues concerning Mode I testing are discussed in this section based
on experimental observations and results obtained for the currently investigated material.
These issues, either unresolved in the literature or not reported for carbon/brittle epoxy
material, include: i) whether to measure critical SERR at crack onset (G
Ic
) based on a


Fig. 3-13: Overall Mode I IFT resistance curve for five DCB specimens, by MCC method
86

starter film or a precrack, ii) accuracy of SERR calculation at crack onset using the MBT
or MCC method, iii) effect of reloading/unloading, and iv) fiber bridging. They are
discussed as follows.
i) Whether to measure G
Ic
based on a starter film or precrack?
The advantages of measuring G
Ic
based on the starter film is that it forms an
identical straight shape of crack front, and additionally, it is not dependent on the loading
history. The latter advantage is important for materials with an increasing fracture
resistance curve with crack extension because G
Ic
measured from a precrack is higher
than that measurement from the starter film, and the difference between the two depends
on the precrack length and precracking methods. However, for the material system in this
investigation, the fracture resistance curve is very flat after ~ 5 mm crack extension. Even
though a very thin (13 m-thickness) insert film was used as the starter film, it is still
likely to create a resin rich pocket and unstable crack growth occurs at crack onset. For
this reason, as shown in the G
Ic
results by different methods in Figure 3-11, the average
G
Ic
based on the precrack and MCC is even more conservative than G
Ic
based on the
insert film and MCC. Furthermore, the variation of G
Ic
values based on the precrack is
less than that based on insert film. Therefore, the G
Ic
value based on a Mode I precrack is
recommended to be a more conservative and repeatable material property for the material
system in this investigation.

ii) Accuracy of SERR calculation of the MBT and MCC methods at crack onset
and propagation
Comparing MBT and MCC methods (Figure 3-12), the results of critical SERR
calculations by these two methods are very similar except for the onset value (G
Ic
). The
onset value (G
Ic
) obtained using the MBT method is always less than that obtained using
the MCC method. This disagreement is a result of the dissimilarity of crack front shapes
at the initial crack length and at the subsequent propagated crack lengths, as mentioned in
the earlier section. The theoretical methods assume plane strain or plane stress condition,
and the crack front shape across the specimen width with crack extension is always
straight. However, in reality, the crack front shape in a DCB specimen develops from an
87

initially straight shape into a curved shape after certain length of propagation, as shown
in Figure 3-14. This implies that, shortly after crack initiation, the crack growth at the
center of specimen is more than that on the edge. The differentiation of compliance to
crack length, dC/da, predicted by theories is somewhat inaccurate. Hence, the SERR at
crack onset predicted by either MBT or MCC is somewhat inaccurate because the crack
front changes from a straight to a convex shape.
A comparison of how the crack propagates theoretically and experimentally is
shown in Figure 3-15. Theoretically, the crack starts from the straight front formed by
insert film (shown in solid blue line), and advances with a same straight front shape as
shown in dashed blue lines. However, during a test, the crack propagates in a curved
shape rather than straight, as shown in solid black lines.

Approximated crack front
Fig. 3-14: Crack surfaces of a quasi-static DCB test specimen
88

Assume that the beam theory can be still used to predict the compliance versus
crack length relationship if the crack length measured on the edge, a, is replaced by an
average crack length, a
avg
, which is given by, Eq. (3.10)
Parameter k
1
is positive since the crack length is longer in the center than on the edge of
specimen. If the crack front shape during propagation is self-similar, then k
2
= 0. If the
crack front shape becomes flatter as crack propagates, then k
2
> 0; if it is the opposite,
then k
2
< 0.
Based on this assumption, the compliance versus crack length relation can be
written as,
From Eq. (3.12), C
1/3
is still in a linear relationship with respect to crack length measured
on the specimen edge, as assumed by the MBT and MCC methods. Comparing Eq. (3.12)
to the compliance vs. crack length relationship by the MCC method, by Eq. (2.5)
(neglecting the effect of loading block, N=1),

a
0
a
i
a
j
Insert
film
Loading line
(a
avg
)
i
(a
avg
)
j
End of insert
film
Straight crack
front assumed by
theories
Propagated curved
crack front
Fig. 3-15: A schematic of crack propagation
avg 1 2
a a k k a = +
(3.10)
1/ 3
0 1 avg 0 1 1 2
' '( ) ' '( ) C A A a A A a k k a = + = + +
(3.11)
( )
1/ 3
0 1 1 1 2
' ' '(1 ) C A A k A k a = + +
(3.12)
89

if a curved front shape exists in a DCB specimen, parameter A
0
will be larger than that
predicted by theoretical solution with straight front shape assumption . In the MBT
compliance calibration method, it will result in a higher value of end correction factor
than the theoretical prediction (Eq. (2.8)). However, since C
1/3
is still linearly related with
respect to crack length, as assumed by the MBT and MCC method, the SERR value for
crack propagation calculated by both methods are still accurate.
Based on the proceeding discussion, since the calculated G
Ic
values onset from an
insert film is not accurate by both MBT and MCC methods, it is again recommended to
use the initiation value based on a short Mode I precrack.

iii) Effect of reloading/unloading
From Figure 3-5 and Figure 3-6, one can see that after a loading/unloading cycle,
when the load returns to zero, there is a small positive offset in displacement relative to
the original displacement before loading. This might be caused by the plastic deformation
occurring near the crack tip, and/or fibers pulling out of the material after the formation
of new fracture surfaces. In addition, when another loading cycle is applied, the critical
load and displacement for crack onset is more than the load and displacement for crack
growth at the end of the last loading cycle, respectively. As a result, the unloading and
reloading procedure may have an effect of increasing the perceived toughness value (G
IR
).
In Figure 3-16, the critical SERR during crack extension calculated by MCC method, is
plotted against the crack length calculated by the MCC compliance calibration
relationship for Specimen 1-4. As shown in the circled part of the resistance curve, at the
same crack length the critical SERR for crack growth is increased after an unloading-
reloading procedure.
1/ 3
0 1
C A Aa = + (3.13)
90


iv) Fiber bridging
Small scale fiber bridging was observed during tests, as shown in Figure 3-17.
Comparing to laminates with tougher matrix, the effect of fiber bridging is small for the
material system in this investigation, as was indicated by the flat resistance curve as well.

0
20
40
60
80
100
120
140
160
40 50 60 70 80 90 100 110
Crack length, a, mm
G
I
R
,

J
/
m
2
Crack onset
where reloading-
unloading occurs

Fig. 3-16: Fracture resistance curve for Specimen 1-4, with crack length calculated by
compliance calibration by MCC method
91


(a)
(b)
Fig. 3-17: Fiber bridging observed through long distance microscope
Fiber bridging
Fiber bridging
92

3.3 The Mode I Fatigue IFT Testing
The DCB fatigue tests use the same test configuration as the quasi-static tests. The
test method involves a brief precrack process, fast cyclic loading and some slow
compliance calibration cycles. Test results are presented as crack growth with increasing
cycles and the relation between SERR and crack growth rate. No test standard exists for
characterizing crack growth behavior using the modified Paris law.
3.3.1 Mode I fatigue test method
The Mode I fatigue test uses the same setup as the Mode I quasi static test as
shown in Figure 3-3. Before the fatigue test, the specimen was precracked by a short
Mode I quasi-static test according to the following procedures:
1) The specimen was loaded by increasing crack opening displacement (COD)
until crack onset was visually observed. As seen in some of the previous quasi-static
tests, the crack growth was unstable immediately after crack onset during the test.
The actuator movement was stopped after the visual onset of crack growth was
observed.
2) After the unstable crack growth was arrested, the specimen was unloaded
slightly and then loaded again until stable crack propagation was observed. The total
precrack growth (a
pr
) was measured by a long distance microscope and recorded at
the maximum crack opening displacement. The maximum crosshead opening
displacement was recorded and used as the critical displacement (
Ic
) at the crack
length of a
0
+ a
pr
for the fatigue test.
3) The specimen was unloaded to zero opening displacement.
The loading procedures for the precrack test are shown schematically in Figure 3-18.
93

The fatigue test started right after the Mode I precrack test. The test was carried
out in displacement control. The opening displacement was cycled between the maximum
displacement (
Imax
) and minimum displacement (
Imin
) using a sine wave. The maximum
opening displacement was set to
Ic 0
0.8( ( ))
pr
a a + , corresponding to 80% of the Mode
I critical SERR at this crack length. A loading frequency of 10 Hz was used for cyclic
loading and a slower loading frequency of 0.0028 Hz was used for compliance
measurement at certain cycle numbers (N =1, 100, 200, 500, 1000, 2000 50,000).
Maximum and minimum displacements were recorded for every cycle. During slow
cycles, crack growth was measured by a QM-1 long distance telescope. Additionally,
load-displacement data were recorded throughout every slow cycle to obtain the
specimens compliance. Testing parameters for the four fatigue tests are summarized in
Table 3-3. The displacement ratio R was selected such that the minimum SERR is much
smaller than maximum SERR, but the minimum displacement is still large enough to
avoid crack closure.

Ic
(a
0
+ a
pr
)
P


Fig. 3-18: A schematic showing the loading and unloading procedures for the Mode I
precrack test
94

The maximum SERR reduces very quickly as the Mode I crack grows in a
displacement controlled DCB test, and it is predicted by beam theory to be inversely
proportional to the fourth order of the effective crack length, (a + )
4
, as expressed in
Eq. (3.14).
A plot of normalized maximum SERR vs. crack length predicted by simple beam theory
for the current nominal specimen geometry (a
0
= 50 mm) is shown in Figure 3-19.
Table 3-3: Mode I Fatigue test parameters
Specimen No.
4-2 4-3 4-4 4-5
Displacement ratio: R 0.2 0.2 0.2 0.2
Loading frequency (Hz): f 10 10 10 10
Maximum displacement (mm):
Imax
3.6 3.85 4.12 3.687
Minimum displacement (mm):
Imin
0.72 0.77 0.82 0.737
Critical displacement before
fatigue test (mm):

Ic
(a
0
+a
pr
) 4.02 4.30 4.60 4.12
Precrack length (mm): a
pr
9.716 12.89 12.866 12.852
* Except for periodic crack and compliance measurements, where f = 0.0028 Hz.

( )
( )
2 4
0
Imax Imax
4
Ic Ic
a
G
G
a

+
=

+

i
(3.14)
95

3.3.2 Mode I fatigue test results
3.3.2.1 Compliance calibration
The purpose of compliance calibration for fatigue tests is to 1) theoretically
predict crack growth based on the compliance obtained from the load-displacement curve,
and 2) obtain the correction factor for SERR calculation by the MBT method, or the
multiplier A
1
for SERR calculations by the MCC method. Similar to quasi-static tests,
compliance calibration could be done based on two analytical methods, the MBT method
(Figure 2-3) and the MCC method (Figure 2-5). However, based on these two
compliance calibration methods, crack growth could be predicted by at least four
alternative approaches:

0%
20%
40%
60%
80%
100%
50 55 60 65 70 75 80
Crack length, a , mm
G
I
m
a
x
/
G
I
c
,

%

Imax
= 1.1
Ic
(a
0
)

Imax
= 1.2
Ic
(a
0
)

Imax
= 1.3
Ic
(a
0
)

Imax
= 1.0
Ic
(a
0
)

Fig. 3-19: Mode I maximum SERR reduction as crack grows for a displacement
controlled DCB fatigue test
96

i) obtain C-a relationship by MBT from quasi-static test results,
1/ 3
( )
s s
C m a = + ; predict crack growth by
1/ 3
1
( ) /
s
a C m = ;
ii) obtain C-a relationship by MBT method from current fatigue test result,
1/ 3
( )
f f
C m a = + ; predict crack growth by
1/ 3
2
( ) /
f
a C m = ;
iii) obtain C-a relationship by MCC method from quasi-static test results,
s s
B C A h a + =
3 / 1
1
2 / ; predict crack growth by
1/ 3
3 1
2 ( )
s
a hA C = ;
iv) obtain C-a relationship by MCC method from current fatigue test result:
s f
B C A h a + =
3 / 1
1
2 / ; predict crack growth by ) ( 2
3 / 1
1 4
C hA a
f
= ;
where the subscript s indicates a parameter from static tests, and f indicates one from
a fatigue test. Details of calculation methods are listed in Table 3-4.

Table 3-4: Crack growth prediction approaches for DCB fatigue tests

Approach i) Approach ii) Approach iii) Approach iv)
C-a data for
calibration
from
five quasi-static
tests
one fatigue test
with the
specimen of
interest
five quasi-static
tests
one fatigue test
with the
specimen of
interest
Theory,
plotting
method
MBT,
Figure 2-3
MBT,
Figure 2-3
MCC,
Figure 2-5
MCC,
Figure 2-5
Calibration
parameter
obtained
m
s
,
s
m
f
,
f

1 0
,
s s

1 0
,
f f

C-a relation
obtained
1/3
( )
s s
C m a = +
1/3
( )
f f
C m a = +
1/3
1 0
( )
2
s s
a
bC
h
= +
1/3
1 0
( )
2
f f
a
bC
h
= +
Crack growth
predicted by
1/ 3
1
( ) /
s
a C m =
1/ 3
2
( ) /
f
a C m =
1/3
3 1
2 ( )
s
a h bC =
1/3
4 1
2 ( )
f
a h bC =
SERR
calculated by ( )
3
2
I
s
P
G
b a

=
+

( )
3
2
I
f
P
G
b a

=
+

2 2 / 3
2
1
3
2 (2 )
I
s
P C
G
b h
=

2 2 / 3
2
1
3
2 (2 )
I
f
P C
G
b h
=


97

Crack growth calculations based on the compliance measured at certain cycles
were done using these four approaches for one specimen (Specimen 4-2), as an example.
The results of crack growth predicted by these four approaches are compared to crack
growth measured visually (a) in Figure 3-20. The crack growth predicted by the MBT
method is very similar to that predicted by the MCC method. However, the crack growth
calculated using the compliance calibration relation obtained from quasi-static tests
(approach i) and iii)) under-predict the crack growth by 0-1 mm relative to the crack
growth calculated from the compliance calibration relation from the fatigue test
(approach ii) and iv)). One possible explanation for this could be: at a same crack length,
the compliance of a fatigue specimen is less than the compliance of a quasi-static
specimen because more severe fiber bridging occurs in a fatigue specimen and thus make
the fatigue specimen stiffer. The difference between the MBT and MCC method is small
as shown in Figure 3-20.


0
1
2
3
4
5
6
7
8
0 20,000 40,000 60,000 80,000
N, cycle

a
,

c
r
a
c
k

l
e
n
g
t
h
,

m
m
a measured visually
a calculated by Eqn. (1)
a calculated by Eqn. (2)
a calculated by Eqn. (3)
a calculated by Eqn. (4)
1/ 3
1
( ) /
s
a C m =
1/ 3
2
( ) /
f
a C m =
1/ 3
3 1
2 ( )
s
a hA C =
) ( 2
3 / 1
1 4
C hA a
f
=
, measured visually a
i)
ii)
iii)
iv)
MBT
MCC

Fig. 3-20: Crack growth by different methods (Specimen 4-2)
98

Since the compliance calibration by approaches i) and iii) usually under-predict
the crack growth and the results by approaches ii) and iv) are similar, crack growth
calculation for other specimens was done using approach iv) only.
3.3.2.2 Crack growth
Crack growth measured visually and that calculated by the MCC method
(approach iv)) based on the measured compliance were plotted against number of cycles
in Figure 3-21 for all Mode I fatigue specimens. From Figure 3-21, the crack growth by
visual measurement is relatively noisy compared to the crack growth calculated by the
MCC compliance calibration method. This is caused by the fact that the visual
measurement focuses on the crack growth on one specimen edge, while the compliance
calibration method is reflective of crack growth across the entire specimen width. Hence,
the error in visual measurements could be easily introduced by a small scale fiber
bridging in the edge area, or a thicker white painting in some local areas.

99

3.3.2.3 Crack growth rate (da/dN) vs. maximum SERR (G
Imax
) plots
The crack growth rate, da/dN, with crack growth obtained by visual measurement,
was plotted against Mode I maximum strain energy release rate, G
Imax
, for all Mode I
fatigue specimens, as shown in Figure 3-22. Also, the crack growth rate, da/dN, with
crack growth calculated by compliance calibration of MCC method, was plotted against
Mode I maximum strain energy release rate, G
Imax
, for all Mode I fatigue specimens, as
shown in Figure 3-23. The power law relationship by Eq. (2.73) was used to fit the data
from all test specimens. The exponent of the power law relation found by using visually
measured crack growth and by using calculated crack growth are 11.12 and 12.29,
respectively. Additional crack growth rate (da/dN) vs. maximum SERR (G
Imax
) plots for
individual specimens are given in Appendix B (Figure B-1 to Figure B-8).


60
62
64
66
68
70
72
74
0 10,000 20,000 30,000 40,000 50,000 60,000 70,000 80,000
Cycle No., N
C
r
a
c
k

l
e
n
g
t
h
,

a
,

m
m
4-2, by CC 4-3, by CC
4-4, by CC 4-5, by CC
4-2, by visual 4-3, by visual
4-4, by visual 4-5, by visual

Fig. 3-21: Crack growth by various methods
100




Fig. 3-22: da/dN vs. G
Imax
plots for four DCB fatigue specimens (with crack growth by
visual measurement)
101

Comparing Figure 3-22 to Figure 3-23, there is more scatter in the plot for which
crack growth is obtained by visual means, which is expected because of the inherent
uncertainty in crack length measurement. The exponents of the Modified Paris law
( ( )
I
n
ax I
G B dN da
Im
/ = ) are similar in the two plots. Comparing to the exponents found in
literature (Appendix Table A-3), which are in the range of 3.6 to 15, the exponents found
in this investigation (11.12 and 12.29) are near the high end of the common range. This
indicates that the crack growth rate decreases very quickly as the crack driving force
decreases. Also, the high exponent indicates that a small error in the crack driving force
prediction will result in a large error in the crack growth prediction, and thus large error
in life prediction of a structure made of this material.



Fig. 3-23: da/dN vs. G
Imax
plots for four DCB specimens (with crack growth calculated
by compliance calibration).
102


Chapter 4

The Mode II Interlaminar Fracture Toughness Testing
Mode II interlaminar fracture toughness tests under quasi-static and cyclic
loadings were conducted with the End Notched Flexure (ENF) specimen. ENF test
configurations, methods, and results are presented in this chapter.
4.1 Material, Specimen and Test Configuration
The specimens used in this investigation are machined from two flat
carbon/epoxy panels of [0]
12
lay-up as shown in Figure 4-1 and Figure 4-2. A thin Teflon
film of 12.7 m thickness was inserted at the mid-plane during the lay-up process of each
panel to define the initial starter crack. The ENF specimens were cut from the panels at
the Penn State Composites Lab using a water-cooled diamond abrasive cut-off wheel.
Distributions of tested ENF specimens in the panels are shown in Figure 4-1 and
Figure 4-2.
103





00.0005 Teflon film
0-degree fiber direction
Specimen 2-1
Specimen 2-2
Specimen 2-3
Specimen 2-4 Specimen 3-1
Specimen 3-3

Fig. 4-1: The ENF and SLB panel diagram (Panel B)
104

The ENF specimen geometry and notations are shown in Figure 4-3. For the
three-point bending fixture used in this investigation, the three loading and supporting
pins are attached to the bending fixture by three springs, so that the loading pin can rotate
about its centerline and the two supporting pins can roll along the longitudinal direction
of the bending fixture. The quasi-static tests were conducted with seven ENF specimens.
In four of these (Specimen 2-1, 2-2, 2-3, and 2-4 in Figure 4-1) the crack initiated from
an insert film (un-precracked specimens) and in three (Specimen 2-5, 2-6, and 2-7 in
Figure 4-2) the crack initiated from a short Mode I precrack that extended beyond the
insert film (precracked specimens). Fatigue tests were conducted with three ENF
specimens, namely Specimen 5-3, 5-4, and 5-5 in Figure 4-2. The dimensions of ENF
specimens in terms of length width thickness are approximately 150 25.4 3.9 mm,
with an initial artificial crack created by the embedded thin film of about 50.8 mm length



00.0005 Teflon film
0-degree fiber direction
Specimen 2-5
Specimen 2-6
Specimen 5-3
Specimen 5-4
Specimen 5-5
Specimen 2-7

Fig. 4-2: The ENF panel diagram (Panel AA)
105

for the un-precracked specimens. For the precracked specimens, the length of the
embedded thin film is about 76.2 mm length and this artificial crack was extended further
by about 3-5 mm in Mode I loading before testing. Dimensions of un-precracked
specimens tested quasi-statically are listed in Appendix B Table B-3 , while dimensions
of precracked specimens tested quasi-statically and in fatigue are listed in Appendix B
Table B-4 and Table B-5, respectively.
The ENF tests were conducted with two kinds of configurations, since specimens
with two different initial crack lengths were used. Because the precracked specimens had
longer initial delamination lengths, in order to create enough room for crack growth, the
three loading pins were shifted a certain distance away from the initially cracked end of
specimen. The un-precracked test configuration, denoted Configuration A, is described in
Figure 4-4 and Table 4-1. The precracked test configuration, denoted Configuration B, is
described in Figure 4-5 and Table 4-2.

l

b
2
h

a
0
Support
Loading
direction
Support

Fig. 4-3: A schematic of ENF specimen geometry and test configuration
106




L L
c
r
c
l
l
Supporting line
Supporting line
r
1
r
2
a
0
Test
specimen
Loading line
Loading
direction
Fig. 4-4: Schematic of ENF un-precracked test configuration A
Table 4-1: Dimensions for ENF configuration A (un-precracked specimen).
Notation Parameter measured Dimension, mm (in.)
l Specimen length 152.4 (6.0)
L Distance from support to loading point 50 (2.0)
r
1
Radius of loading nose 3.2 (0.125)
r
2
Radius of support 6.4 (0.25)
c
l
Left overhang 25.4 (1.0)
c
r
Right overhang 25.4 (1.0)


L L
c
r
c
l
l
r
1
r
2
a
p
Loading
direction
Loading
line
Supporting line
Supporting line
Test
specimen
L L
c
r
c
l
l
r
1
r
2
a
p
Loading
direction
Loading
line
Supporting line
Supporting line
Test
specimen

Fig. 4-5: Schematic of ENF precracked test configuration B
107


4.2 The Mode II Quasi-static IFT Testing
The Mode II quasi-static IFT tests were conducted with ENF precracked and un-
precracked specimens. Since it was found that a large amount of unstable crack growth
occurred at initiation in both quasi-static and fatigue tests with ENF un-precracked test
configuration, an ENF precracked test configuration was used in an attempt to solve the
instability issue. The result did show stable crack growth with ENF test configuration B
(precracked specimen) in the fatigue tests. From quasi-static tests, the initiation value of
G
IIc
obtained for precracked specimens is much lower than that obtained for un-
precracked specimens.
4.2.1 Mode II quasi-static test method
Photographs of the ENF test setup are shown in Figure 4-6 for the un-precracked
test configuration and in Figure 4-7 for the precracked test configuration. A three-point
bending fixture with a maximum span length of 203.2 mm was used as the loading fixture.
The specimen was loaded by a servo-hydraulic MTS 810 machine. Load was measured
by a 13.5 kN MTS load cell using the 2.2 kN (500 lb) load range, and a 1.1 KN (250 lb)
Table 4-2: Dimensions for ENF test configuration B (precracked specimen)
Notation Section of test specimen measured
Dimension of specimen,
mm (in.)
l Specimen length 152.4 (6.0)
L Half span length of bending fixture 50 (2.0)
r
1
Radius of loading roller
6.4 (0.25)
(same as configuration A)
r
2
Radius of supporting rollers
3.2 (0.125)
(same as configuration A)
c
l
Left overhang 43.4 (1.709)
c
r
Right overhang 9.2 (0.362)

108

load cell was also connected into the load frame in order to check the accuracy of 500 lb
load cell measurement. The loading point displacement was measured by the MTS LVDT
built into the actuator.


MTS built-in load
cell (stationary)
1.1 kN (250 lb)
load cell
Bending fixture
Specimen
MTS built-in load
cell (stationary)
1.1 kN (250 lb)
load cell
Bending fixture
Specimen

Fig. 4-6: A photograph of the ENF test setup with un-precracked test configuration
109

Generally, the Mode II interlaminar fracture toughness test was conducted in three
steps: 1) precracking and marking (or just marking for tests with un-precracked
specimens), 2) compliance calibration, and 3) crack initiation test. For compliance
calibration and crack initiation, the test was conducted under displacement control with a
constant displacement rate of 0.5 mm/min for loading and unloading. Details of the
testing procedures are as follows:
(1) Before any testing, the specimens slated for precracking were precracked in
Mode I by driving a thin blade into the manufactured crack created by the embedded
thin film. The specimen was clamped completely across the width at the position of
the intended crack front. The blade was taken out from the specimen before the
specimen was unclamped.


Fig. 4-7: A photograph of the ENF test setup with precracked test configuration
110

(2) To obtain the compliance versus crack length relationship, compliance
calibration procedures were conducted before the crack initiation test at crack lengths
of a
0
, a
0
10, and a
0
5 mm (a
0
is the initial crack length based on the insert film)
for test configuration A, and at crack lengths of a
p
+ 9, a
p
+ 6, a
p
+ 3, a
p
and a
p
- 3
mm for test configuration B. Marks with increment of a were made on the edge of
the specimen for locating the corresponding positions of the rollers for compliance
calibration at each crack length, as shown schematically in Figure 4-8. The crack
length increment, a, was 5 mm for ENF test configuration A, and 3 mm for ENF test
configuration B. For each compliance calibration test, the specimen was loaded to a
load point displacement of about 50% - 60% of the estimated critical displacement
and then unloaded. Five initial crack lengths were achieved by sliding the specimen in
the bending fixture in the longitudinal direction of specimen. Care was taken to
prevent crack onset during the compliance calibration procedures.
(3) For the crack initiation test, the specimen was placed in the bending fixture as
shown schematically in Figure 4-4 for ENF test configuration A and Figure 4-5 for
ENF test configuration B. The specimen was loaded until crack initiation was
observed, and then unloaded.



a
a
p


Fig. 4-8: Markings on ENF specimen edge for compliance calibration
111

4.2.2 Mode II quasi-static test results
4.2.2.1 Load-displacement curves
Typical load vs. displacement plots for ENF crack initiation tests with un-
precracked and precracked specimens are shown in Figure 4-9 and Figure 4-10,
respectively. These two plots show some differences between the un-precracked
specimen with shorter initial crack length (a
0
/L 0.5), and precracked specimen with
longer initial crack length (a
p
/L 0.7).


y = 441.81x - 40.558
R
2
= 0.9999
0
200
400
600
800
1000
1200
0 0.5 1 1.5 2 2.5 3
Displacement, mm
L
o
a
d
,

N
Loading
Unloading
P
cr

Fig. 4-9: A representative load vs. displacement plot for ENF crack onset test (Specimen
2-2, un-precracked, a
0
/L 0.5)
112

In the un-precracked cases tested, the amount of load drop immediately after
crack initiation ranged from 497 to 731 N. The unstable crack in some tests grew to a
point that was about 10 mm past the center loading roller. In the test for which the load-
displacement plot is shown in Figure 4-9, immediately after the unstable crack onset the
load dropped about 630 N. Additionally, before the crack onset critical point, the loading
curve is particularly linear (except for the initial nonlinear load take-up portion). Hence
for the un-precracked configuration, the critical point for crack onset critical SERR
calculation is unambiguously the maximum load point.
In the precracked cases tested, the amount of load drop immediately after crack
initiation ranged from 52 to 232 N. The unstable crack growth typically stopped
somewhere near the center loading roller. In one test, the crack growth was quasi-stable,
showing both unstable and stable crack growth behaviors before the crack observed
through the telescope reached the center loading roller, and the load vs. displacement
curve for this test is shown in Figure 4-11. Additionally, in contrast to the un-precracked
specimens, nonlinearity in the load-displacement curve was observed for the precracked

Linear fit: y = 399.47x - 15.871
R
2
= 0.9987
0
100
200
300
400
500
600
700
800
0.0 0.5 1.0 1.5 2.0
Displacement, mm
L
o
a
d
,

N
P_max load
P_nonlinear
P_5%offset
5% Compliance
offset

Fig. 4-10: A load vs. displacement plot for ENF crack onset test (Specimen 2-5,
precracked, test configuration B)
113

specimens. In most pre-cracked specimens, the maximum load occurred before the
compliance increased 5% relative to the initial compliance, and after nonlinearity became
obvious. Hence, the critical point for crack onset critical SERR was defined as the
maximum load point in precracked specimens, according to JIS standard (JIS K 7086
1993).
In Figure 4-10, Figure 4-11 and the later plots in which the 5% compliance offset
construction line is shown, the 5% compliance offset construction line is defined by
assigning the x-intercept and slope as follows. Firstly, the initial compliance straight line
is constructed by fitting a straight line through the load vs. displacement test data where
the load-displacement relation is observed to be linear (usually from the point where
displacement is greater than 0.24 mm in an ENF test or 0.22 mm in an SLB test to the
point fracture initiates). Then, the 5% compliance offset straight line is constructed in
such a way that, it intersects with the x-axis at the same point as the initial compliance
straight line and has a slope which is 1/1.05 times that of the initial compliance line. The
construction method is shown schematically in Figure 4-12.

Linear fit: y = 391.6 x - 29.158
R
2
= 0.999
0
100
200
300
400
500
600
700
0.0 0.5 1.0 1.5 2.0 2.5
Displacement, mm
L
o
a
d
,

N
P_max
P_5% offset
P_nonlinear
5% compliance
offset

Fig. 4-11: A quasi-stable load vs. displacement plot for ENF crack onset test (Specimen
2-6, precracked)
114

4.2.2.2 Compliance calibration
Some typical load-displacement plots for a compliance calibration test are shown
in Figure 4-13. The compliance value at each initial crack length was obtained by taking
the inverse of the slopes of straight lines best-fitted to the linear portion of loading curves
(where displacement is greater than 0.24 mm or so). Similar compliance calibration
procedures were conducted for both ENF quasi-static specimens and fatigue specimens.
Hence, the results are presented together.


L
o
a
d
,

P
2 1
1 1
1.05
S S
=

Fig. 4-12: Construction method for the 5% compliance offset line
115

Two forms of polynomial were used to correlate compliance and crack length.
Firstly, it is predicted by classical plate theory that compliance of an ENF specimen is
given by Eq. (2.30). Hence, a polynomial in the form of Eq. (4.1) was used to fit the
compliance vs. crack length data.
Parameters A and B were determined from a linear least squares fit of the C vs. a
3
plot.
Parameter B is only related to the flexural modulus of the specimen. The ratio A/B is only
related to half span of the bending fixture, L. The first compliance calibration method is
denoted by CC 1) in the following discussion. A plot of C(8bh
3
) vs. a
3
for all specimens
is given in Figure 4-14. The parameters determined are listed in Table 4-3.

Linear fit (a=43.7 mm):
y = 295.61x 21.618, R
2
= 0.9988
Linear fit (a=40.7 mm):
y = 321.36x 20.274, R
2
= 0.9985
Linear fit (a=37.9 mm):
y = 351.22x 21.559, R
2
= 0.9985
Linear fit (a=35.0 mm):
y = 383.56x 22.888, R
2
= 0.9988
Linear fit (a=32.0 mm):
y = 412.71x 26.888, R
2
= 0.9986
0
50
100
150
200
250
300
350
400
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Dis placement,mm
L
o
a
d
,

N
a=43.7
a=40.7
a=37.9
a=35.0
a=32.0

Fig. 4-13: Load vs. displacement curves for compliance calibration (precracked
Specimen 2-6, at initial crack lengths of a = 32.0, 35.0, 37.9, 40.7 and 43.7 mm)
3 3
) 8 ( Ba A bh C + = (4.1)
116



0.0E+00
1.0E-15
2.0E-15
3.0E-15
4.0E-15
5.0E-15
6.0E-15
7.0E-15
8.0E-15
-2.0E-05 0.0E+00 2.0E-05 4.0E-05 6.0E-05 8.0E-05 1.0E-04 1.2E-04
Crack length, a
3
, m
C
(
8
b
h
3
)
,

(
m
5
/
N
)
2-1 2-2 2-3
2-4 2-5 2-6
2-7 5-3 5-4
5-5
3 3
) 8 ( Ba A bh C + =

Fig. 4-14: C(8bh
3
) vs. a
3
plots for all ENF specimens
117

Secondly, another polynomial in the form of Eq. ((4.2) was used to produce
compliance vs. crack length relationship as well. Parameters, C
0
, C
1
, C
2
, and C
3
, were
determined by a cubic least squares fit. The second compliance calibration method is
denoted by CC 2) in the following discussion.
The second plot made for compliance calibration is shown in Figure 4-15. Parameters
determined by the second compliance calibration methods are listed in Table 4-4.
Table 4-3: Parameters A and B, determined by CC 1) for ENF specimens
Specimen A, m
4
/N B, m/N A/B, m
3
R
2

2-1 2.56E-15 3.32E-11 7.69E-05 0.992
2-2 3.01E-15 3.16E-11 9.55E-05 0.998
2-3 2.67E-15 2.93E-11 9.11E-05 0.996
Un-
precracked
2-4 2.37E-15 3.15E-11 7.54E-05 0.999
STD * 2.69E-16 1.61E-12 1.01E-05 0.003
Mean 2.65E-15 3.14E-11 8.47E-05 0.996
Statistical
results
COV ** 10.15% 5.12% 11.87% 0.33%
Specimen A, m
4
/N B, m/N A/B, m
3
R
2

2-5 2.58E-15 2.89E-11 8.94E-05 1.000
2-6 2.78E-15 2.96E-11 9.40E-05 0.999
2-7 2.57E-15 3.01E-11 8.53E-05 0.999
5-3 2.62E-15 2.98E-11 8.79E-05 0.999
5-4 2.74E-15 3.36E-11 8.15E-05 1.000
Precracked
5-5 2.42E-15 3.00E-11 8.07E-05 1.000
STD 1.31E-16 1.69E-12 5.06E-06 0.0003
Mean 2.62E-15 3.03E-11 8.65E-05 0.999
Statistical
results
COV 4.99% 5.56% 5.85% 0.03%
Note: * standard deviation
** coefficient of variation, STD/COV
2 3
0 1 2 3
C C C a C a C a = + + + (4.2)
118



1.0E-06
1.5E-06
2.0E-06
2.5E-06
3.0E-06
3.5E-06
4.0E-06
0 0.01 0.02 0.03 0.04 0.05
Crack length, a, m
C
o
m
p
l
i
a
n
c
e
,

C
,

m
/
N
2-2 2-3 2-4
2-1 2-5 2-6
2-7 5-3 5-4
5-5
2 3
0 1 2 3
C C C a C a C a = + + +

Fig. 4-15: A C vs. a plot for compliance calibration, by Eq. (2.39)
119


Comparing the two compliance calibration methods, CC 1) the compliance
calibration by Eq. (4.1) and CC 2) by Eq. (2.39), the second method fits the experimental
data better based on R
2
values. Since there are two more parameters in Eq. (2.39) than
Eq. (4.1), the relation by CC 2) should be better fitted to test data. However, the two
additional parameters (C
1
and C
2
) are not involved in any physical property of the ENF
specimen. The better fit provided by CC 2) might be a result of including the errors of the
crack length measurement, the effect of the diameter of the loading and supporting pins,
effect of friction and others, into the final compliance calibration relation. With these
additional effects incorporated, it is uncertain whether the resulting CC relation is
reflective of the compliance vs. crack length relation of the ENF specimen or the
compliance vs. crack length relation of the whole loading system. Additionally, large
specimen-to-specimen variation exists in the parameters, C
0
to C
3
, determined by CC 2).
This creates difficulties in comparing the CC relation from different specimens and
evaluating the accuracy of this compliance calibration method. On the other hand,
Table 4-4: Parameters C
0
, C
1
, C
2
, and C
3
, determined by CC 2) for ENF specimens

Specimen C
0
, m/N C
1
, 1/N C
2
, 1/(Nm) C
3
, 1/(Nm
2
) R
2

2-1 2.07E-06 -1.15E-05 -2.38E-04 3.57E-02 0.998
2-2 2.43E-06 -6.20E-05 2.83E-03 -1.95E-02 1.000
2-3 1.55E-06 2.19E-05 -5.17E-04 2.15E-02 1.000
Un-precracked
2-4 1.31E-06 5.67E-05 -2.53E-03 5.83E-02 1.000
STD 5.07E-07 5.05E-05 2.21E-03 3.27E-02
Statistical
results
Mean 1.84E-06 1.28E-06 -1.13E-04 2.40E-02 0.999
Specimen C
0
, m/N C
1
, 1/N C
2
, 1/(Nm) C
3
, 1/(Nm
2
) R
2

2-5 -1.15E-06 2.40E-04 -6.68E-03 8.03E-02 1.000
2-6 1.30E-05 -9.09E-04 2.45E-02 -1.99E-01 1.000
2-7 6.10E-06 -3.43E-04 8.83E-03 -5.49E-02 1.000
5-3 1.39E-05 -9.20E-04 2.29E-02 -1.69E-01 1.000
5-4 1.13E-05 -7.66E-04 2.06E-02 -1.60E-01 1.000
Precracked
5-5 5.35E-06 -2.90E-04 7.43E-03 -4.30E-02 1.000
STD 5.75E-06 4.54E-04 1.20E-02 1.05E-01
Statistical
results
Mean 8.08E-06 -4.98E-04 1.29E-02 -9.10E-02 1.000

120

calculating SERR using the compliance calibration relation directly involves the
differentiation of the compliance to crack length, dC/da, instead of compliance, C, at a
certain crack length. Hence, the accuracy of the dC/da vs. a relation derived from the
compliance calibration relation is also important. It is hard to judge the accuracy of this
relation derived from CC 2) since large specimen-to-specimen variation exists in the
parameters and no physical meaning is involved in these parameters. Therefore, so far the
compliance calibration method CC 2) does not show obvious advantage to CC 1). For
convenience and a more repeatable compliance calibration relation, it is suggested to use
the relation determined by CC 1). For fatigue tests, it is more convenient to use the first
approach CC 1) to predict crack length based on the compliance obtained during certain
cycles and hence it was used for fatigue tests in this investigation.
4.2.2.3 G
IIc
onset values
The G
IIc
onset values were calculated by three methods: the classical plate theory
(CPT) using Eq. (4.3) and two compliance calibration (CC) methods using Eq. (4.5) and
Eq. (4.6), respectively.
In the previous two equations, a
1
is crack length calculated for the critical point, by
Eq. (4.4). The initial crack length, a
0
, is replaced by the precracked crack length, a
p
, if a
precracked specimen is used. C
0
is the compliance of the initial elastic portion, which is
taken as the inverse of the slope of the initial linear portion of the load-displacement
curve; C
1
is the compliance at the critical point; P
c
is the load at the critical point.

The expressions for SERR using two compliance calibration relations are:
2 2
1 0
IIc 3 3
0
9
2 (2 3 )
c
a P C
G
b L a
=
+

(4.3)
1/ 3
3 3 1 1
1 0
0 0
2
1
3
C C
a a L
C C

= +


(4.4)
121

where in the previous two equations, B is a parameter determined by CC 1), Eq. (4.1),
and C
1
, C
2
and C
3
are parameters determined by CC2) , using Eq. ((4.2).
The results of the G
IIc
values at crack onset using three methods are shown in
Figure 4-16. Comparing the average G
IIc
onset values for precracked and un-precracked
specimens as shown in Figure 4-16, the values calculated by the two compliance
calibration methods are very similar, however, for precracked specimens, the average
value calculated by CPT is about 10% higher than that by CC methods. The average G
IIc

value of un-precracked specimens is about 44% higher than that of precracked specimens
by CPT method, and about 60% higher by CC methods. The numerical values of G
IIc
for
crack onset are given in Appendix B, Table B-9. From the ENF test results found from
the literature (Appendix A, Table A-2), the G
IIc
values based on an insert film for
carbon/brittle epoxy materials range from 300 to1500 J/m
2
, while the G
IIc
values based on
a Mode I precrack range from 250 to 1000 J/m
2
. During this investigation, the average
G
IIc
values based on an insert film determined by different calculation methods ranged
from 782 to 801 J/m
2
, while the average G
IIc
values based on a Mode I precrack ranged
from 498 to 545 J/m
2
. Compared to results in the literature, the Mode II quasi-static
fracture toughnesses of the investigated material are in the middle of the common range
for carbon/brittle epoxy materials.

2
2
IIc 0 3
3
2 8
c
P B
G a
b bh
= (4.5)
2
2
IIc 1 2 0 3 0
( 2 3 )
2
c
P
G C C a C a
b
= + + (4.6)
122

4.2.2.4 A short Mode II fracture resistance curve
For the quasi-stable ENF test for which the load-displacement curve is shown in
Figure 4-11, a short Mode II fracture resistance curve can be generated based on the load-
displacement curve, the compliance calibration results and the critical SERR value at
crack onset. The methods for fracture toughness and crack length calculations are
discussed next.
From Classical plate theory, the compliance of an ENF specimen can be written in
the form of Eq. (4.7).

0
100
200
300
400
500
600
700
800
900
1000
2-1 2-2 2-3 2-4 2-5 2-6 2-7
Specimen No.
G
I
I
c
,

J
/
m
2
by CPT
by CC 1)
by CC 2)
average by CPT
average by CC 1)
average by CC 2)
un-precracked precracked

Fig. 4-16: G
IIc
onset values
3 3
3
3
1
2 3
' '
8
L a
C A B a
E bh
+
= = +
(4.7)
123

Thus, A and B are only related to half span length L, specimen dimensions, b and h, and
Youngs modulus in the longitudinal direction, E
1
. They are given by,
The crack length can thus be predicted by, Eq. (4.9)
The SERR can be written as Eq. (4.10),
Substituting Eq. (4.9) into Eq. (4.10), results in,
Substituting C = /P into Eq. (4.11) and rearranging terms, yields,
Substituting load and displacement at the critical crack onset point, P
cr
and
cr
, yields,
Denoting the propagated critical SERR by G
IIR
, and dividing Eq. (4.12) by Eq. (4.13),
results in,
Parameter A is related to the compliance calibration parameter A (Eq. (4.1)) by,
3
3
1
2
'
8
L
A
E bh
= ,
3
1
3
'
8
B
E bh
=
(4.8)
1/ 3
'
'
C A
a
B

=


(4.9)
2
2
3 '
2
II
P
G B a
b
= (4.10)
( )
2/ 3
2 2
2/ 3
1/ 3
3 ' 3
' ' '
2 ' 2
II
P C A P
G B B C A
b B b

= =


(4.11)
( ) ( )
2 1/ 3
2/ 3
2/ 3
1/ 3 2 3
3 3 '
' / ' '
2 2
II
P B
G B P A P A P
b b
= = (4.12)
( )
1/ 3
2/ 3
2 3
3 '
'
2
IIc cr cr cr
B
G P A P
b
= (4.13)
( )
( )
2/ 3
2 3
IIR
2/ 3
2 3
IIc
'
'
cr cr cr
P A P
G
G
P A P


=


(4.14)
3
'
8
A
A
bh
= (4.15)
124

As a result, an expression for propagated critical SERR is given by,
The crack length can be calculated based on instantaneous load and displacement by,
Eq. (4.17)
where parameters B is related to the compliance calibration parameter B (Eq. (4.1)) by,
Eq. (4.18)
Using Eq. (4.16) and Eq. (4.17), a short fracture resistance curve based on the
load vs. displacement data for the quasi-stable ENF test was generated as shown in
Figure 4-17. The crack length was calculated based on the compliance calibration relation
1). From Figure 4-17, the fracture resistance in Mode II changes slightly with crack
extension.
2/ 3
2 3 3
IIR IIc 2/ 3
2 3 3
/(8 )
/(8 )
cr cr cr
P A bh P
G G
P A bh P



=



(4.16)
1/ 3 1/ 3
' / '
' '
C A P A
a
B B

= =


(4.17)
3
'
8
B
B
bh
= (4.18)
125

On the other hand, if a constant Mode II fracture resistance with crack extension
is assumed, a theoretical load vs. displacement curve after crack initiation can be
predicted based on Eq. (4.14). Load P can be solved as a function of displacement and
material Mode II fracture toughness, G
IIR
. Assuming G
IIR
equals crack initiation value
G
IIc
, constant G curves generated for an ENF specimen from Eq. (4.14) is shown in
Figure 4-18.

0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
30 35 40 45 50
Crack length, a, mm
G
I
I
R
/
G
I
I
c

Fig. 4-17: Mode II resistance curve (Specimen 2-6, precracked)
126

4.2.2.5 Special issues on the ENF test
(1) Load drop immediately after crack initiation
The steep slope of the constant G curve (G = G
IIc
) after crack initiation as shown
in Figure 4-18 somewhat explains the large amount of load drop at the crack initiation
point of an ENF test. The steep slope indicates that a small amount of crack growth
greatly reduces the stiffness of specimen and thus initiates a significant load drop. The
theoretical prediction assumes the crack continuously advances in infinite small
increments as shown in the double dot centerline in Figure 4-19. However, in reality, the
crack advances in bigger increments than theoretical prediction, which results in a load-
displacement curve as shown in the blue dashed line in Figure 4-19. If the crack advances
in even bigger increments, then it could result in a load vs. displacement curve close to a
real test as shown in the sold line in Figure 4-19.

0
100
200
300
400
500
600
0.0 0.5 1.0 1.5 2.0 2.5
Displacement, mm
L
o
a
d
,

N
P_max
P_5% offset
P_nonlinear
G = G
IIc
G = 1.1G
IIc

Fig. 4-18: A load vs. displacement plot for ENF test with constant G curves shown
127


(2) Crack instability after crack initiation
It was predicted by theory solution that the crack growth in an ENF specimen is
unstable when a/L< 0.7. However, during this investigation, even when using precracked
specimens for which a
0
/L> 0.7, the crack growth immediately after onset was unstable.
This instability can be explained by the crack growth rate for a quasi-static ENF test
derived from CPT theories. Derived from Eq. (4.10), alternative expressions for Mode II
SERR could be,
Solving for displacement from Eq. (4.19), one obtain,


Fig. 4-19: A schematic of load vs. displacement curves for different states of crack
growth
( )
2 2 2
2 2 2
2 2
3
3 ' 3 ' 3 '
2 2
2 ' '
IIR
P
G B a B a B a
b bC
b A B a

= = =
+

(4.19)
( )
3
IIR
' '
2
3 '
A B a
bG
B a

+
=
(4.20)
128

Assuming a constant G
IIR
with crack extension, results in,
and, Eq. (4.22)
In a displacement controlled test with constant loading speed (d/dt=0.5 mm/min),
da/d is proportional to the crack propagation speed assuming the crack advance in
infinite small increments as assumed theoretically. Substituting
3
'/ ' 2 / 3 A B L = , obtained
from Eq. (4.8), d/dt=0.5 mm/min, and using
3
'/(8 ) B bh =3.03E-11 and G
IIR
=500 J/m
2
,
the calculated da/d is plotted against crack length as shown in Figure 4-20. It is
predicted that the crack propagation speed near a/L= 0.7 is extremely high. The high
speed crack propagation, or high sensitivity of crack growth in response to the loading
displacement near a/L= 0.7, could be the cause of an observed unstable crack growth
immediately after crack initiation during a precracked ENF test.
IIR IIR
2 3
2 2 ' ' '
2 ' 2
3 ' 3 '
bG bG B d A A
B a a
da B a B a

= + = +


(4.21)
IIR
3
1
2 ' '
2
3 '
da
d bG B A
a
B a

=

+



(4.22)
129

4.3 The Mode II Fatigue IFT Testing
The Mode II fatigue IFT tests were conducted with precracked ENF specimens as
shown in Figure 4-5. The aim of the tests is to determine the crack growth length versus
loading cycles and the crack growth rate with respect to cycles versus the SERR relation.
4.3.1 Mode II fatigue test method
Mode II fatigue tests were conducted with ENF precracked specimens, following
the same procedures as was applied in the ENF quasi-static tests. The same compliance
calibration procedures used for the ENF quasi-static tests were applied as well for the
fatigue tests.

0
5
10
15
20
25
30
35
0.6 0.7 0.8 0.9 1
a/L
d
a
/
d
t
,

m
m
/
s
e
c
a
0
/L

Fig. 4-20: Plot of crack growth rate in a quasi-static ENF test versus normalized crack
length
130

The fatigue test was carried out in displacement control. The critical displacement
for crack onset in a quasi-static test at the initial length range tested,
IIcr
, is
approximately 1.6 mm. The loading point displacement was cycled between maximum
displacement (
IImax
) and minimum displacement (
IImin
) using a sinusoidal wave. Testing
parameters for the fatigue test are summarized in Table 4-5.
The maximum cycling displacement used for this fatigue test was approximately
75% of the estimated critical opening displacement for ENF quasi-static test at the same
initial crack length. As a result, the maximum SERR to critical SERR ratio (G
IImax
/ G
IIc
)
at the beginning of the fatigue test is approximately 0.56, where the critical SERR is the
estimated critical SERR for crack onset in a quasi-static ENF test. The test was set up in a
way such that the initial crack length to half span ratio, a
p
/L, for the precracked specimen
is greater than 0.693. For a fatigue test with a constant maximum loading point
displacement, the maximum SERR is predicted by Eq. (4.23) (Classical Plate Theory),
and its trend with crack extension is shown in Figure 4-21.

Table 4-5: Testing parameters for ENF fatigue tests
Displacement ratio: R 0.2
Loading frequency: f 10 Hz
Maximum displacement:
IImax
1.2 mm
Minimum displacement:
IImin
0.24 mm
Estimated critical
displacement of static test:

IIcr
1.6 mm


( )
( )
2
2
3 3
2
,max ,max
2 2
3 3
2 3
2 3
p
II II
IIc IIcr p
L a
G
a
G a
L a

+

=

+

(4.23)
131

In Figure 4-21, the normalized maximum SERR (F
1
*G
IImax
/ G
IIc
) is plotted against
normalized crack length (a/L) for a constant displacement amplitude fatigue test as
predicted by classical plate theory.

The maximum to critical SERR ratio (G
IImax
/ G
IIc
)
occurs at the point where a/L 0.693. When a/L > 0.693, we have dG
IImax
/da < 0 and
hence stable crack growth is predicted till the crack reaches the loading line where a/L =1,
assuming the fracture resistance of material is not decreasing as crack length increases.
To obtain the crack length as a function of the number of loading cycles, two
methods were used: visual measurement and compliance calibration. A slower loading
frequency of 0.086 Hz was used for loading cycle # 1, 100, 500, 1000, 1500, 2000, 2500,
3000, 4000, 5000, 6000, 7000, and etc. Load-displacement data were recorded for each
slow cycle. A load-displacement curve was obtained for every slow cycle. Compliance
was obtained by taking the inverse of the slope of the load-displacement curve. The crack
length was then calculated using the compliance-crack length relationship obtained from
a compliance calibration done earlier on the specimen. Additionally, crack growth was

0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
a/L
F
1
*
G
I
I
m
a
x
/
G
I
I
c
Predicted stable crack
growth range

Fig. 4-21: Maximum Mode II SERR (G
IImax
) vs. normalized crack length plot for an ENF
test with fixed displacement amplitude (F
1
is a factor related to initial crack length and
maximum opening displacement)
132

measured visually by a Questar QM-1 long distance telescope during each slow cycle.
The ENF test was stopped when the crack grew close to but not beyond (about 3-5 mm
away from) the loading line.
4.3.2 Mode II fatigue test results
The results of compliance calibration of ENF specimens were presented in
Section 4.2.2. In this section, the compliance calibration relation obtained was used for
crack length calculation, and compared to the crack length measured visually. Crack
growth rate was related to the maximum SERR by the modified Paris law, and
exponents were found to be high in comparison to literature.
4.3.2.1 Crack growth
Crack growth measured visually and calculated by compliance calibration is
shown in Figure 4-22. The crack length calculated by compliance calibration is about
between 0 and 3 mm larger than that by visual measurement. Possible reasons for this
could be: a) the crack length was hard to measure visually when the specimen was loaded
in Mode II and hence large error could exist for visual measurement; b) the crack length
is measured visually on one specimen edge, and it could be more or less than the crack
length calculated based on the compliance, which is theoretically an average crack length
across the width.
133

A microscopic view of an ENF specimen edge when the same specimen was
subject to different loading conditions is shown in Figure 4-23 (a) to (c). Comparing (a)
to (b) and (c), the crack was hard to discriminate from the surface when the specimen was
loaded in Mode II and crack surfaces were forced in contact with each other. The reason
one can observe the crack in Mode I opening is that the crack appears to be a black line
on a white background. However, in Mode II, one can observe the crack only because of
the relative shear displacement of the crack surfaces, which is very small even far away
from the crack tip when the specimen is loaded. As a result, for Mode II loading, it is
difficult to visually detect the crack.

30
32
34
36
38
40
42
44
46
48
50
0 2,000 4,000 6,000 8,000 10,000
Cycle No., N
C
r
a
c
k

l
e
n
g
t
h
,

a
,

m
m
5-3, by CC
5-4, by CC
5-5, by CC
5-3, by visual
5-5, by visual

Fig. 4-22: Crack growth for ENF fatigue specimens
134




(a) Crack opened in Mode I
Fig. 4-23: A microscopic view of a specimen edge while the specimen was in different
loadings (Specimen 5-3)
135


~1 mm

(b) specimen loaded to the mean loading point displacement of a Mode II fatigue cycle

(c) specimen loaded to the maximum loading point displacement of a Mode II fatigue
cycle
Fig. 4-23 (continued): A microscopic view of a specimen edge while the specimen was in
different loadings (Specimen 5-3)
136

After the ENF fatigue tests, the specimens were split apart by hand. Photographs
of crack surfaces of three specimens are shown in Figure 4-24. While the crack front
shapes created by Mode I precrack appear to be irregular, the crack front shapes created
by Mode II fatigue test are consistently slightly convex relative to the front created by the
insert film. This indicates that there is a trend that the crack length on the edge of
specimen is smaller than the average crack length across width. Additionally, it could be
easily examined that the crack front shape is not always perfectly symmetric about the
center-width of the specimen as the crack proceeds progressively. Due to the local
property of the material, the crack can proceed faster or slower on one specimen edge at
certain stages of a test compared to the progress elsewhere across the width. Only
focusing on the progress of crack growth on one specimen edge, the visual measurement
of crack growth can be somewhat noisy, particularly if crack growth from one
measurement point to the next is very small.
137


Crack surface created by Mode I precrack
Crack surface created by an ENF fatigue test

Crack surface created during precrack
Crack surface created during the fatigue test
Crack surface created by Mode I precrack
Crack surface created by an ENF fatigue test

Crack surface created by Mode I precrack
Crack surface created by an ENF fatigue test
Fig. 4-24: Photographs of fracture surfaces
138

4.3.2.2 da/dN - G
IImax
plots
The crack growth rate, da/dN, with crack growth calculated by compliance
calibration, was plotted against Mode II maximum strain energy release rate, G
IImax
, as
shown in Figure 4-25. Additionally, da/dN, with crack growth measured visually, was
plotted against G
IImax
for two tests, as shown in Figure 4-26. The power law relation
between da/dN and G
IImax
, ( )
II
II IImax
/
n
da dN B G = , was used to fit the data from each ENF
specimen. With the crack growth lengths calculated from compliance calibration, the
exponent for the power law relation was found to be in the range of 7.7 to 12.8. With the
crack growth lengths measured visually, the exponent was in the range of 7.1 to 6.8. In
the literature, the exponents of this power law relation for ENF fatigue tests with
carbon/epoxy or glass/epoxy material mostly range from 4.3 to 6.5 (Appendix Table A-3).
However, in one case, this value was found to be between 13 and 15 for an unidirectional
carbon/epoxy material (Hojo et al. 2006). Compared to results in the literature, the
exponent for the material investigated is somewhat high. From the literature, for the same
material the exponent of the power law relation for Mode I is usually higher than that for
Mode II. This is also the case for the investigated material. Additional crack growth rate
(da/dN) vs. maximum SERR (G
IImax
) plots for individual specimens are given in
Appendix B from Figure B-9 to Figure B-13.

139



Fig. 4-25: Crack growth rate against Mode II maximum SERR plot, with crack growth
calculated by compliance calibration
140




Fig. 4-26: Crack growth rate against Mode II maximum SERR plot, with crack growth
measured visually
141


Chapter 5

Mixed-mode I/II Interlaminar Fracture Toughness Testing
Mixed-mode I/II interlaminar fracture toughness (IFT) tests under quasi-static and
cyclic loading were conducted with single-leg-bending (SLB) specimens. The Mode II to
total strain energy release rate (SERR) ratio (G
II
/G
T
) achieved with mid-thickness
delaminated SLB specimens was about 0.43. The mixed-mode interlaminar fracture
toughness for crack onset was characterized under this mode ratio, and the fracture
resistance curve was obtained based on the calculated crack length. SLB fatigue tests
were conducted in displacement control with displacement ratio R = 0.2. The results
show that the exponent of the power law relation ( )
i
n
i
G B dN da
max
/ = is higher for
mixed mode loading than that for Mode I or Mode II loading.
5.1 Material, Specimen and Test Configuration
The specimens used in this investigation are from two flat carbon/epoxy panels of
[0]
12
lay-up as shown in Figure 4-1 and Figure 5-1. The mixed-mode I/II IFT tests were
conducted with SLB specimens. The SLB specimen geometry and notations are shown in
Figure 5-2. Quasi-static tests were conducted with four SLB specimens, of which
Specimens 3-1 and 3-3 are from the panel shown in Figure 4-1 and Specimens 3-2 and 3-
3 from the panel shown in Figure 5-1. All the specimens were tested without precracking.
Fatigue tests were conducted with three SLB specimens (Specimen 6-1, 6-2, and 6-3 in
Figure 5-1). The dimensions of the SLB specimens in terms of length width thickness
are approximately 150 25.4 3.9 mm, with an initial artificial crack created by the
embedded thin film of 13 m thickness and ~50.8 mm long. Specific dimensions of each
SLB specimen tested quasi-statically and cyclically are listed in Appendix Table B-6 and
Table B-7, respectively.
142







00.0005 Teflon film
0-degree fiber direction
Specimen 3-2
Specimen 3-4
Specimen 6-3
Specimen 6-2
Specimen 6-1

Fig. 5-1: The SLB panel diagram
143

The SLB test is essentially an un-symmetric three-point-bending test using a
three-point bending fixture as the loading apparatus. Part of one delaminated leg of the
specimen was removed before testing so that a crack opening displacement can be
applied to the other leg. The SLB test configuration is shown in Figure 5-3 and test
dimensions are listed in Table 5-1. For the three-point bending fixture used in this
investigation, the three loading and supporting pins are attached to the bending fixture by
three springs, so that the loading pin can rotate about its centerline and the two supporting
pins can roll along the longitudinal direction of the bending fixture.

l

b
2
h

a 0
L
L
Loading
direction
Support
Support

Fig. 5-2: A schematic of SLB specimen geometry and test configuration
144


5.2 The Mixed-mode I/II Quasi-static IFT Testing
The goal of the SLB quasi-static test is to characterize the critical SERR value for
crack onset and growth, and also obtain the load and displacement at the critical onset
point. The test setup used for a SLB test is similar to the setup for an ENF test; however,
while for a quasi-static ENF test the crack growth was usually unstable at crack onset, the
crack growth was stable but very fast for a quasi-static SLB test. Since the mode ratio
G
II
/G
Tc
is about 0.43 for the SLB test configuration used and the Mode II fracture
toughness is about 5-8 times higher than the Mode I fracture toughness, the SLB test is
considered to be more governed by Mode I behavior than Mode II behavior.

Loading roller
r
1
Test specimen
Loading
direction
c
c
Left supporting
roller
Right supporting
roller
r
2
r
2
L L
Spacer
Loading roller
r
1
Test specimen
Loading
direction
c
c
Left supporting
roller
Right supporting
roller
r
2
r
2
L L
Spacer
Fig. 5-3: SLB test configuration
Table 5-1: SLB test configuration dimensions
Notation Section of test specimen measured Dimension, mm (in.)
L Half span length of bending fixture 63.5 (2.5)
r
1
Radius of loading roller 6.4 (0.25)
r
2
Radius of supporting rollers 3.2 (0.125)
c Overhang 12.8 (0.5)

145

5.2.1 Mixed-mode I/II quasi-static test method
A photograph of the SLB test set up is shown in Figure 5-4. The SLB specimen
was placed in a three point bending fixture and loaded by a 13.5 kN (3 kip) MTS
machine. The load was measured by a 0.448 kN (100 lb) capacity load cell. Loading
point displacement was measured by the MTS LVDT built into the actuator.
Due to the special geometry of the SLB specimen, certain attention should be paid
to the fabrication of the specimen. Before testing, the majority of the lower cracked
region needs to be removed using a water-cooled diamond abrasive cut-off wheel. The
specimen was marked at a certain distance (~ 35 mm) from the delaminated end on the
top surface before cutting. To prevent accidental cutting of the upper leg, a thin blade
was inserted between the upper and lower crack region, and pushed forward until it went
close to the desired cutting point. Care was taken to ensure the crack growth did not
occur during the cutting process.
Compliance calibration procedures were conducted before the crack initiation test
at crack lengths of a
0
, a
0
3 mm, and a
0
6 mm (a
0
is the initial crack length of the

MTS built-in load
cell (stationary)
0.44 kN (100 lb)
load cell
SLB Specimen
Bending fixture
Not shown:
-Servo-hydraulic actuator
-long-distance
instrumented stage
microscope
MTS built-in load
cell (stationary)
0.44 kN (100 lb)
load cell
SLB Specimen
Bending fixture
Not shown:
-Servo-hydraulic actuator
-long-distance
instrumented stage
microscope
Fig. 5-4: A photograph of SLB test set up
146

crack initiation test). For the compliance calibration procedure and the crack initiation
test, a constant displacement rate of 0.5 mm/min was used for loading and unloading.
Marks were made on the edge of the specimen to locate the positions of the rollers
for compliance calibration tests at each crack length, as shown in Figure 5-5. The
distance between two neighboring marks (a) was approximately 3 mm and was
measured through the instrumented telescope before testing. Additionally, the initial
crack length was measured before testing. For each compliance calibration test, the
specimen was loaded to a load point displacement of about 60% of the estimated critical
displacement and then unloaded. Five initial crack lengths were achieved by sliding the
specimen in the bending fixture in the longitudinal direction of specimen.
For the crack initiation test, the specimen was placed in the bending fixture as
shown schematically in Figure 5-3. The specimen was loaded until the crack propagated
for about 20 mm, and then unloaded.




a
0
a
Fig. 5-5: Markings on SLB specimen edge
147

5.2.2 Mixed-mode I/II test results
5.2.2.1 Load-displacement curves
Load-displacement curves from four SLB quasi-static tests are shown in Figure 5-
6. In a manner similar to Mode I and Mode II IFT tests, the typical loading curve shows
increasing load up to its maximum value followed by decreasing load with crack
extension. However, unlike Mode I and Mode II testing, a slight nonlinear behavior
before the maximum load point was observed for the mixed mode tests, as shown in the
closer view of the loading curve near the critical crack onset point for one test in
Figure 5-7.






0
50
100
150
200
250
0.0 0.5 1.0 1.5 2.0
Displacement, mm
L
o
a
d
,

N
3-2 3-3
3-4 3-1

Fig. 5-6: Load vs. displacement curve for SLB quasi-static tests
148

The slight non-linear behavior seen in the loading curve before load decreases
raises the question on the definition of the critical crack onset point. Four alternate
definitions are typical for defining the critical point in an interlaminar fracture toughness
test. They are: 1) the nonlinear point, where nonlinear behavior of load-displacement
curve is first observed; 2) the visual onset point, where crack onset is visually observed;
3) the maximum load point, where the maximum load occurs; and 4) the 5% compliance
offset point, where the compliance increases by 5%. For the material tested in this
investigation, the visual crack onset point was very close to the maximum load point;
however, it is difficult to capture the exact point of visual crack onset and discriminate it
from the maximum load point. From the nonlinear point to the maximum load point, the
compliance had increased by a small amount e.g., in the case shown in Figure 5-7, the
compliance increased by 0.7% from the nonlinear point to the maximum load point. This
may be due to a local crack growth within the center of specimen. It is estimated that the
maximum load point is the closest to the point where global crack growth onset among
all the critical points could be defined. For the material tested, it was always the case that
the maximum load was reached before the compliance increased by 5% relative to the

y = 166.4x - 3.8998
R
2
= 0.9999
150
170
190
210
230
250
1.1 1.2 1.3 1.4
Displacement, mm
L
o
a
d
,

N
Max. load point
Fitted straight line
5% compliance
increase construction
line
Nonlinear point

Fig. 5-7: Load-displacement plot near the critical onset point (Specimen 3-2)
149

initial compliance. Therefore, the maximum load point was determined to be suitable for
the critical crack onset point definition for the material investigated.
5.2.2.2 Compliance calibration
Compliance calibration procedures were conducted for each specimen at five
initial crack lengths of a
0
, a
0
3 mm, and a
0
6 mm. Two relations were used to
correlate compliance vs. crack length data. One is a third order polynomial with four
terms, as expressed by Eq. (5.1), denoted by CC method (1) in the later discussion,
and the other one is also a third order polynomial but omits the first and second order
terms, as expressed by Eq. (5.2), denoted by CC method (2) in the later discussion, which
is in a similar form to compliance predicted by classical plate theory.
In these equations, C
SLB
is the compliance of a SLB specimen and a is the crack length.
Parameters q
0
, q
1
, q
2
, and q
3
are parameters determined fitting a third order polynomial
by least squares in the C vs. a plot. Parameters
1
and
2
are parameters determined by a
least square fit of a straight line in the C(8bh
3
) vs. a
3
plot. The plot for Eq. (5.1) is show
in Figure 5-8, while that for Eq. (5.2) is shown in Figure 5-9 for all the available SLB
data.

3
3
2
2 1 0
a q a q a q q C
SLB
+ + + = (5.1)
( )
3
2 1
3
8 a bh C
SLB
+ = (5.2)
150



4.0E-06
4.5E-06
5.0E-06
5.5E-06
6.0E-06
6.5E-06
7.0E-06
7.5E-06
8.0E-06
8.5E-06
3.0E-02 3.5E-02 4.0E-02 4.5E-02 5.0E-02
a , m
C
,

m
/
N
3-2 3-3 3-4
3-1 6-1 6-2
6-3
3
3
2
2 1 0
a q a q a q q C + + + =

Fig. 5-8: C vs. a plot for all SLB specimens

0.0E+00
4.0E-15
8.0E-15
1.2E-14
0.0E+00 2.0E-05 4.0E-05 6.0E-05 8.0E-05 1.0E-04 1.2E-04
a
3
, m
3
C
(
8
b
h
3
)
,

m
5
/
N
3-2 3-3 3-4
3-1 6-1 6-2
6-3
( )
3
2 1
3
8 a bh C
SLB
+ =

Fig. 5-9: C(8bh
3
) vs. a
3
plot for all SLB specimens
151

From the two compliance calibration plots, one can see that both expressions fit
the data very well based on the R
2
values. Theoretically, the shape of each fitted curve in
the C-a plot is related to each specimens flexural modulus in the longitudinal direction
of specimen, E
1f
, the span length (2L) and specimen dimensions (b, 2h). For the second
C(8bh
3
) vs. a
3

plot, according to classical plate theory(Eq. (5.3)), the slope of each fitted
straight line is only related to longitudinal flexural modulus, E
1f
, and x-intercept of the
straight line is only related to support distance, 2L.
Hence, it is reasonable that the data from different test specimens look more collapsed in
the second plot than in the first one since in the first plot the curve shape depends on
additional specimen dimensions. Furthermore, the variation in the slope of different
curves in the second plot indicates the variation in flexural modulus from specimen to
specimen. The coefficients determined for the first and second compliance calibration
relation are listed in Table 5-2 and Table 5-3, respectively. Since it is more convenient to
use Eq. (5.2) for crack length calculations based on compliance, Eq. (5.2) was used for
crack growth calculation for fatigue tests.


3
1
3 3
8
7 2
bh E
a L
C
f
+
=
(5.3)
Table 5-2: Coefficients determined for compliance calibration by Eq. (5.1) (Load was
measured in Newton and displacement was measured in meter.)





Specimen q
0
q
1
q
2
q
3
R
2

3-1 1.61E-05 -1.01E-03 2.65E-02 -1.87E-01 0.998
3-2 2.16E-05 -1.43E-03 3.69E-02 -2.70E-01 1.000
3-3 -1.69E-05 1.55E-03 -3.86E-02 3.67E-01 1.000
3-4 2.27E-05 -1.50E-03 3.92E-02 -2.94E-01 1.000
6-1 -1.97E-05 1.76E-03 -4.44E-02 4.14E-01 1.000
6-2 5.53E-06 -1.65E-04 4.61E-03 2.19E-03 1.000
6-3 6.52E-06 -2.19E-04 5.36E-03 3.81E-03 1.000
STD 1.90E-05 1.46E-03 3.74E-02 3.16E-01
Mean 4.89E-06 -1.33E-04 4.01E-03 5.60E-03 1.000

152

5.2.2.3 Mixed-mode I/II critical strain energy release rate (G
Tc
) onset value
The mixed-mode I/II critical SERR for SLB specimen was calculated by classical
plate theory and two compliance calibration methods with compliance vs. crack length
relation by Eq. (5.1) and Eq. (5.2), respectively. The expression used for total critical
SERR calculations are written as Eq. (5.4) by classical plate theory, Eq. (5.5) by CC
method (1) and Eq. (5.6) by CC method (2).
where, C
0
is the compliance of specimen at the initial crack length.
The G
Tc
results by these three methods are shown in Figure 5-10. Numerical
results are shown in Appendix B Table B-10. The average G
Tc
value is 185 J/m
2
by CPT,
185 J/m
2
by CC method (1), and 181 J/m
2
by CC method (2). Within the three methods
for SERR calculation investigated, the G
Tc
value by CPT method yields the least
Table 5-3: Coefficients determined for compliance calibration by Eq. (5.2) (Load was
measured in Newton and displacement was measured in meter.)
Specimen
1

2

1
/
2
R
2

3-1 4.90E-15 6.36E-11 7.70E-05 0.998
3-2 4.89E-15 6.68E-11 7.33E-05 1.000
3-3 5.06E-15 6.88E-11 7.35E-05 0.999
3-4 5.39E-15 6.51E-11 8.29E-05 1.000
6-1 4.84E-15 6.14E-11 7.89E-05 0.999
6-2 5.30E-15 6.60E-11 8.03E-05 1.000
6-3 5.24E-15 6.92E-11 7.57E-05 1.000
Mean 5.09E-15 6.58E-11 7.74E-05 0.999
STD 2.22E-16 2.80E-12 3.55E-06
COV 4.4% 4.2% 4.6%

2 2
0 0
Tc 3 3
0
21
2 (2 7 )
c
a P C
G
b L a
=
+

(5.4)
2
2
Tc 1 2 3
( 2 3 )
2
c
P
G q q a q a
b
= + + (5.5)
2
2 2
Tc 3
3
2 8
c
P
G a
b bh

= (5.6)
153

variation (4.8%) from specimen to specimen, and the CC (2) method results in the
maximum variation (8.7%).
A summary of the Mode I, Mode II, and Mixed-mode fracture toughness at crack
onset by one or two representative calculation method(s) is given in Table 5-4. In this
table, the fracture toughness shown at each mode ratio is the average toughness for all the
specimens tested at this mode ratio during this investigation. Mode I and Mode II SERR
in the mixed mode test was calculated using the mode ratio G
II
/G
T
= 0.43 and average
fracture toughness G
T
= 185J/m
2
. It was obtained by Asp et al. (Asp et al. 2001) that: for
a carbon/epoxy material, the Mode I and Mode II toughness (G
Ic
, G
IIc
) were 260 J/m
2
and
1002 J/m
2
, respectively, and the mixed mode fracture toughness (G
TC
) was 447 J/m
2
at
the mode ratio G
II
/G=0.5. Comparing to their results, the fracture toughnesses of the
currently investigated material are low.
Two possible mixed-mode delamination failure criterion for the investigated
carbon/epoxy material are proposed based on the test results at the mode ratio
G
II
/G
T
=0.43. At this mode ratio, Reeders linear mixed mode failure criteria by Eq. (2.63)
gives fairly good prediction of the failure locus, as shown in Figure 5-11. Choosing m =
2.5 in Eq. (2.66), the B-K law also gives a good prediction of the failure locus at the

0
50
100
150
200
250
3-1 3-2 3-3 3-4
Specimen
G
T
c
,

J
/
m
2
by CPT
by CC (1)
by CC (2)

Fig. 5-10: G
Tc
values determined from SLB tests
154

investigated mode ratio, as shown in Figure 5-12. However, to decide which criteria gives
a better prediction of the failure locus at all mode ratios, more mixed-mode test results at
other mode ratios are needed. Summarizing the current IFT test results for the AS4/3501-
6 carbon/epoxy material, Mathews and Swanson (Mathews and Swanson 2007) found
that the Reeders linear law and power law by Eq. (2.64) gave good representation of the
test data. Also, using the B-K law to represent the failure locus, they found the parameter
m to be 4.78 by performing a least squares fit.

1 =

IIc
II
Ic
I
G
G
G
G

(2.63)
( )
m
T
II
Ic IIc Ic Tc
G
G
G G G G

+ = (2.66)
Table 5-4: A summary of Mode I, Mode II, and Mixed-mode I/II fracture toughness at
crack onset





G
Tc
, by CPT: 185 J/m
2

G
I
: 105.45 J/m
2

G
II
: 79.55 J/m
2

G
Ic,
for the non-precracked case
by MBT: 125 J/m
2

G
I
/G
Ic
for the SLB test using
G
Ic
by MBT, non-precrack:
0.84
G
Ic,
for the precracked case
by MCC: 131 J/m
2

G
I
/G
Ic
for the SLB test using
G
Ic
by MCC, precrack:
0.80
G
IIc,
for the non-precracked case
by CPT: 782 J/m
2

G
II
/G
IIc
for the SLB test using
G
IIc
by CPT, non-precrack:
0.10
G
IIc,
for the precracked case
by CPT: 545 J/m
2

G
II
/G
IIc
for the SLB test using
G
IIc
by CPT, precrack:
0.15
by CC 1): 498 J/m
2

G
II
/G
IIc
for the SLB test using
G
IIc
by CC 1), precrack:
0.16

155





0.0
0.2
0.4
0.6
0.8
1.0
0.0 0.2 0.4 0.6 0.8 1.0
G
I
/G
Ic
G
I
I
/
G
I
I
c
nonprecrack nonprecrack
precrack precrack by CPT)
precrack precrack by CC
G
Ic
G
IIc
Reeder's
failure locus

Fig. 5-11: The Reeders linear mixed mode failure locus and test data

0
100
200
300
400
500
600
700
800
900
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
G
II
/G
T
G
T
c
,

J
/
m
2
G
IIc
, nonprecrack
G
IIc
, precrack by
CPT
G
IIc
, precrack by
CC
m=1
m=2.5
test data

Fig. 5-12: The B-K Law failure locus and test data
156

5.2.2.4 Mixed-mode fracture resistance curve
The mixed-mode fracture resistance curve can be generated based on the load-
displacement curve, the compliance calibration results and the critical SERR value at
crack onset. The methods for fracture toughness and crack length calculations are
discussed next.
The crack length is solved from the compliance calibration relation. Starting from
the expression for compliance (Eq. (5.3)) from classical plate theory, and equating the
compliance expression to a third order polynomial that can be obtained from compliance
calibration, yields,
Solving for crack length, a, we have,
where ) 8 /(
3
1 2
bh A = and ) 8 /(
3
2 2
bh B = .
1
and
2
are coefficients obtained from
compliance calibration (Eq. (5.2)).
The load-displacement behavior of the SLB specimen after crack initiation can be
derived from classical plate theory. Substituting Eq. (5.8) and C = /P into the expression
for SERR by CPT (Eq. (5.4)), and rearranging terms, the -P relation is obtained as in
Eq. (5.9).
If a material demonstrates constant mixed-mode fracture toughness (G
TR
) as crack
extends, post-initiation load-displacement curves with various G
TR
values can be
generated based on Eq. (5.9). By observing the intersection of constant G
TR
curves with
the loading curve from a test, the trend of G
TR
as crack extends can be estimated. In
Figure 5-13, the constant G
TR
curves were plotted along with the loading curve for
3
2 2
3
1
3 3
8
7 2
a B A
bh E
a L
C
f
+ =
+
=
(5.7)
3 / 1
2
2


=
B
A C
a (5.8)
( )
3/ 2
2 3 TR
2 1/ 3
2
2
3
bG
P A P
B


=


(5.9)
157

Specimen 3-4. The loading curve of Specimen 3-4 intersects with constant G
TR
curves of
increasing order as crack extends, beginning with the G
TR
= G
Tc
curve and ending with
the G
TR
= 1.4 G
Tc
curve. This indicates that the fracture toughness of the Specimen 3-4
was increasing with crack extension and had increased by about 40% at the end of test.
The critical SERR expression without involving crack length can be derived
based on Eq. (5.9). Normalizing Eq. (5.9) by the values at the critical point, Eq. (5.10) is
determined. With the crack length calculated by the compliance calibration relation,
Eq. (5.11), a fracture resistance curve can be generated for an SLB test.
where, G
Tc
is critical SERR value for crack onset, and G
TR
is SERR as crack extends.
cr

and P
cr
are the displacement and load at crack onset, respectively.

0
50
100
150
200
250
0.0 0.5 1.0 1.5 2.0 2.5
Displacement, mm
L
o
a
d
,

N
G
TR
=G
Tc
G
TR
=1.1G
Tc
G
TR
=1.4G
Tc

Fig. 5-13: Load vs. displacement plot for Specimen 3-4 with constant G
TR
curves shown

3
2
2
3
2
2
2 / 3
cr cr cr
Tc
TR
P A P
P A P
G
G


(5.10)
( )
( )
3 / 1
3
2
3
1
3 / 1
2
2
8 /
8 / /


=
bh
bh P
B
A C
a


(5.11)
158

Fracture resistance curves were obtained for four SLB tests as shown in Figure 5-
14. In Figure 5-14, one can see that for all the specimens the fracture resistance curve
rises slightly as the crack length increases. The fracture toughness increases slightly in
the first 10 mm crack extension. After about 20 mm crack extension, the fracture
toughness had increased approximately 11-53% for three of the specimens. For
Specimen 3-1, the crack growth at the end of the test was less than 20 mm.
5.3 Mixed-mode I/II Fatigue IFT Testing
5.3.1 Mixed-mode fatigue test method
The mixed mode fatigue crack propagation test was conducted using SLB
specimens. The test setup for the fatigue test is the same as for quasi-static test shown in
Figure 5-4. The SLB specimen was placed in a three point bending fixture and loaded by
a 13.5 kN (3 kip) MTS machine. The load was measured by a 0.448 kN (100 lb) capacity

0
50
100
150
200
250
300
350
30 40 50 60 70
Crack length, a, mm
G
T
R
,

J
/
m
2
3-1
3-2
3-3
3-4

Fig. 5-14: Fracture resistance curves for SLB specimens
159

load cell. Loading point displacement was measured by the MTS LVDT built into the
actuator.
Before the fatigue test, the same compliance calibration procedures as used for the
SLB quasi-static tests were applied for SLB fatigue specimens as well. The fatigue test
was carried out in displacement control. The loading point displacement was cycled
between maximum displacement (
max
) and minimum displacement (
min
) using a
sinusoidal wave. Testing parameters for the fatigue test are summarized in Table 5-5.
The maximum cycling displacement used for the fatigue tests was approximately
88% of the estimated critical opening displacement for an SLB quasi-static test at the
same initial crack length. As a result, the maximum SERR to critical SERR ratio (G
max
/
G
Tc
) at the beginning of the fatigue test is approximately 0.78, where the critical SERR is
the estimated critical SERR of crack onset in a quasi-static SLB test.
To obtain the crack length as a function of the number of loading cycles, two
methods were used: visual measurement, and compliance calibration. A slower loading
frequency of 0.0093 Hz was used for loading cycle # 1, 100, 200, 500, 1000, 1500, 2000,
2500, 3000 and so on. Load-displacement data were recorded for each slow cycle. A
displacement-load plot was obtained for every slow cycle and compliance was obtained
from the slope of a straight line fit by least-squares. Crack length was then calculated by
a compliance-crack length relationship obtained from a compliance calibration performed
earlier on the specimen. Additionally, crack growth was measured visually by a Questar
QM-1 long distance telescope when the loading point displacement reached the
maximum value of each slow cycle.
Table 5-5: Testing parameters for SLB fatigue tests
Displacement ratio: R 0.2
Loading frequency*: f 10 Hz
Maximum displacement:
max
1.12 mm
Minimum displacement:
min
0.224 mm
Estimated critical displacement for
crack onset of quasi-static test:

cr
1.27 mm

* Except for periodic crack growth and compliance measurements, where f = 0.0093 Hz.

160

5.3.2 Mixed-mode fatigue test results
5.3.2.1 Crack growth
Crack growth was observed on one specimen edge and back calculated by a
compliance calibration relation derived from Eq. (5.2), and written as Eq. (5.12).
Crack growth vs. number of cycles by visual measurement and compliance calibration
method for all fatigue SLB specimens is shown in Figure 5-15.
As shown in Figure 5-15, the crack growth predicted by compliance calibration is
always more than that measured visually. For Specimen 6-1, after cycle No. 2000 (after
which cycle crack growth was visually observed), the crack length predicted by
compliance calibration was about 2.8-3.5 mm greater than that measured visually. For
3 / 1
2
1
2
3
) 8 (

C
bh
a (5.12)

0.0
2.0
4.0
6.0
8.0
10.0
12.0
0 20,000 40,000 60,000 80,000 100,000
Cycle No., N
C
r
a
c
k

g
r
o
w
t
h
,

a
,

m
m
6-1, by CC
6-1, by visual
6-2, by CC
6-2, by visual
6-3, by CC
6-3, by visual

Fig. 5-15: Crack growth of SLB fatigue specimens
161

Specimen 6-2, after cycle No. 1500, the crack growth predicted by compliance
calibration was about 1.7-2.4 mm greater than crack growth measured visually. For
Specimen 6-3, it was observed that after cycle No. 1500, the crack length predicted by
compliance calibration was approximately 2.3-2.7 mm greater than crack length
measured visually.
Possible explanations for this difference between crack growth by visual
measurement and compliance calibration are proposed as follows. Firstly, for the visual
measurements, because the crack tip created during the SLB test is very sharp, how well
the tip can be seen depends strongly on crack opening displacement and on the thickness
of the edge coating. Assuming that the crack can be visually detected through a
microscope when the distance between the crack surfaces is more than a critical value, d
c
,
the real crack length will be larger than the crack length visually measured, by an amount

a
. If the visual measurement of crack length is always taken at the maximum crack
opening displacement (
max
) of a fatigue loading cycle, which is constant through a
fatigue test, the error of visual measurement,
a
, is increasing as the crack length increases,
as shown schematically in Figure 5-16. Additionally, from this sketch as the opening
displacement (
max
) increases the error of visual measurement (
a
) decreases.

a
1
a
2

a1

a2
d
c
d
c
a
1
< a
2

a1
<
a2
Crack tip
position 1
Crack tip
position 2
Crack can not be visually
detected
a
1
a
2

a1

a2
d
c
d
c
a
1
< a
2

a1
<
a2
Crack tip
position 1
Crack tip
position 2
Crack can not be visually
detected
Fig. 5-16: A sketch of opening crack
162

In the SLB fatigue test, the crack opening displacement is very small, and thus
there should be a certain length of crack (
a
) behind the crack tip that can not be visually
detected. Additionally, if crack closure occurs at the crack tip, the error of visual
measurement is even greater. However, for the compliance calibration procedure, the
crack tip position can be accurately measured because of the different color of the insert
film defining the initial crack.
Secondly, because of the curvature of the crack front shape created by a SLB test,
the crack length is greatest at the mid-width of the specimen as shown in Figure 5-17.
While the visual method gives a measure of crack growth on the edge, the compliance
calibration method predicts the average crack length across the specimen width.
5.3.2.2 Crack growth rate (da/dN) vs. maximum SERR (G
Tmax
) plots
The crack growth rate was calculated based on the crack growth measured
visually and predicted by the compliance calibration method. The maximum SERR was
calculated by the compliance calibration method according to Eq. (5.6). In Figure 5-18,
the crack growth rate calculated with crack length obtained using the compliance
calibration method is plotted against maximum total SERR. Additionally, in Figure 5-19,
the crack growth rate calculated with crack length by visual measurement is plotted

Crack surface created by an
SLB fatigue test
Crack surface created by an
SLB fatigue test
Crack front at
the end of a test

Fig. 5-17: Fracture surfaces of a SLB specimen
163

against maximum total SERR. More da/dN-G
max
plots for individual specimens are given
in Appendix B, Figure B-14 to Figure B-19. In these graphs, the power law relation,
given in Eq. (2.75), was used to correlate crack growth rate and maximum total SERR.


( )
i
n
i
G B
dN
da
max
= (2.75)


Fig. 5-18: A da/dN vs. G
max
plot for SLB specimens (crack length by compliance
calibration method)
164

As shown in Figure 5-18 and Figure 5-19, a power law relation can fit the da/dN-
G
max
test data from individual specimens very well. This indicates that among all the
fatigue crack growth models presented in literature review from Eq. (2.79) to Eq. (2.81),
the modified version of the Russell and Streets model, ( )
max
/
n
da dN B G = , can provide
a good fit to crack growth rate versus SERR data for the investigated material system.
The exponent of this modified Paris law for mixed-mode I/II, which is in the range from
16 to 22, is higher than that for Mode I (11-13) and Mode II (7-13). It was found by Asp
et al. (Asp et al. 2001) that, for a carbon/epoxy material, the exponent of modified Paris
law was 5.5 under Mode I loading, 4.4 under Mode II loading, and 6.3 under mixed-mode
loading at the mode ratio G
II
/G
T
= 0.5. Compared to their results, the exponent of the
modified Paris law for the currently investigate material is very high under loadings at
various mode ratios.


Fig. 5-19: A da/dN vs. G
max
plot for SLB specimens (crack length by visual
measurement)
165


Chapter 6

Finite Element Modeling of Crack Propagation in DCB Specimens
Two-dimensional (2D) and three-dimensional (3D) finite element models of the
DCB specimen were built with Abaqus using the virtual crack closure technique (VCCT)
and/or cohesive elements, in order to further analyze delamination behavior and evaluate
the experimental results. Geometries, loadings, modeling techniques, and results of three
2D models and three 3D models are presented in this chapter. Performing the crack
propagation analysis by the finite element method (FEM) is the opposite of conducting an
IFT test as described in Figure 6-1. That is, in an IFT test, the critical SERR (fracture
toughness) is characterized based on certain information obtained from the test, such as,
the load vs. displacement response and crack growth vs. displacement. However, in a
crack propagation analysis with VCCT, the fracture resistance property of the material
(critical SERR) is part of the input information of the finite element model. Based on this
input information, the load vs. displacement response and crack growth are predicted by
performing finite element analysis.



Fig. 6-1: Finite element analysis and the IFT test
166

6.1 Two-dimensional Modeling of the DCB Specimen
6.1.1 Geometry, loading and boundary conditions of 2D models
The geometry of the 2D models is shown in Figure 6-2. The specimen length, l, is
100 mm and thickness is 3.9 mm. The initial crack is located at the mid-thickness of
specimen. Equal but opposite displacements are applied to the ends of the two
delaminated beams. During analysis, the displacement is increased linearly with the
loading time. The end opposite to the loading end is restricted from moving in the
thickness and longitudinal directions of the specimen.
6.1.2 Modeling techniques for 2D models
The objectives of 2D modeling are to: 1) compare the experimentally measured
crack lengths with those predicted by FE models; 2) compare the load-displacement
behavior predicted by FE models with the experimental results; 3) determine the effect(s)
of element size and release tolerance for VCCT; and 4) obtain detailed stress distribution
near the crack tip, through the thickness of the specimen. To fulfill these purposes, three
2D models were built under the plane strain condition with differences in element size,
release tolerance, and/or fracture interface type. In 2D Models #1 and #2, the fracture

/2, P
/2, P
Fig. 6-2: Geometry and boundary conditions of 2-dimensional DCB models
167

interface incorporates only the VCCT interaction, i.e., the VCCT subroutine for Abaqus
controls the debonding of the fracture surfaces by comparing the current and critical
SERR. However, 2D Model #2 uses a more refined mesh and a smaller release tolerance
compared to Model #1. Thus, the first two models were designed to reveal the effect(s) of
release tolerance and element size. In 2D Model #3, the same meshing scheme as in
Model #1 is used; however, in addition to the VCCT interaction, the fracture interface in
2D Model #3 incorporates a cohesive layer which continues to apply tensile forces
between the pre-bonded nodes after a debonding reaction is requested by the VCCT
subroutine. From the experimental results of one DCB test specimen (Specimen 1-2), the
fracture toughness for the investigated material slightly increased within the first 10 mm
crack extension, and stabilized after that. This toughening behavior of the material
could be a result of fibers bridging between crack surfaces. The bridging fibers can
transmit stresses across the crack faces, and thus absorb a portion of applied energy
before breaking and/or pulling out of the crack faces. It is assumed that the mechanical
behavior of the bridging fibers can be simulated by the cohesive elements, and thus in the
third model, a cohesive layer is added to bond the fracture interaction surfaces. The load
vs. displacement response predicted by the third model can then be compared with those
given by the first two models and experimental results. For convenience, the fracture
resistance curve of the DCB test was idealized as a bi-linear curve, as shown in Figure 6-
3. To implement the increasing resistance curve to the FE model, the user needs to
specify spatially varying critical SERRs, i.e., critical SERR value is specified from node
to node along the slave fracture surface.
168


The main features of the three 2D models are summarized as follows:
a) 2D Model #1:
Initial crack length is 51 mm;
The element size in the length direction is about 1 mm; 4 elements comprise
the thickness direction for each delamination leg;
Maximum crack opening displacement is 8 mm;
Release tolerance is 0.01 (the release tolerance is a parameter of the VCCT
subroutine for controlling the debonding of nodes, more details are available
in Section 2.5.2);
Assumed elastic properties of composite material are: (E
11
was back-
calculated from the DCB test results, and other parameters, which were
considered to be less important for the current model, are found from the
literature for similar type of material (Krueger and OBrien 2001).)
E
11
= 110.0 GPa; E
22
= 10.16 GPa; E
33
= 10.16 GPa;

12
=
13
= 0.3;
23
= 0.436;
G
12
= G
13
= 4.6 GPa; G
23
= 3.54 GPa.
Mode I fracture toughness linearly increases from 110 N/m to 130 N/m in the
first 10 mm crack extension, as shown in Figure 6-3.

0.06
0.07
0.08
0.09
0.10
0.11
0.12
0.13
0.14
50 60 70 80 90 100
a , mm
G
I
c
,

N
/
m
m

Fig. 6-3: Idealized Mode I fracture toughness with crack extension for Specimen 1-2
169


b) 2D Model #2:
Initial crack length is 50.5 mm;
Refined mesh with 0.5 mm element size in the length direction; 4 elements
comprise the thickness direction for each delamination leg;
Maximum crack opening displacement is 8 mm;
Release tolerance is 0.002;
The same elastic and fracture toughness material properties as 2D Mode #1
are used.

c) 2D Model #3
The fracture surfaces are bonded by a cohesive layer in addition to the VCCT
contact interaction;
Initial crack length is 51 mm;
The same meshing discretization as 2D Model #1 is used;
Release tolerance is 0.01, as in 2D Model #1;
Assumed fracture toughnesses of the laminates material are: G
Ic
= 110 N/m;
G
IIc
= 600 N/m; G
IIIc
= 1200 N/m.

The elastic properties of the cohesive layer material are specified in terms
of the traction separation response with stiffness values. To ensure the cohesive
elements will have no adverse effect on the stable time increments, it is suggested
to choose,
e c
K K 1 . 0 = , where K
c
is the cohesive element stiffness, and K
e
is the
stiffness of surrounding material. Based on this principle, elastic properties of the
cohesive layer chosen are:
E = 1.016 GPa; G
1
= G
2
= 1.016 GPa
where E is the Youngs modulus and G
1
and G
2
are shear moduli of the bulk
material of the cohesive elements.

170

The above elastic properties are used to construct diagonal terms in the
stiffness matrix as described in Section 2.5.3.1. The stiffness values are chosen
using the same scheme as used in the cohesive element application example in
Abaqus Benchmark Manual. In this example the stiffness values are chosen to be
1/10 that of the stiffness E
2
of the composite material.

The quadratic traction-interaction failure criterion is selected for damage
initiation in the cohesive elements; and a mixed-mode, energy based damage
evolution law based on a B-K law criterion is selected for damage propagation.
Ultimate normal stress (t
n
0
) and shear stress (t
s
0
) are:
t
n
0
= 0.55 MPa, t
s
0
= 0.55 MPa
(A range of ultimate normal and shear stress values (0 - 7.5 MPa) were tried for
3D Model #3. The values listed above were found to result in closest load vs.
displacement response to the experimental data.)

The B-K form of damage evolution criterion was used, and the parameters
are:
G
n
C
= 110 N/m, G
s
C

= 600 N/m, = 1.
Here, G
n
C
and G
s
C
are the critical fracture energies required to cause failure in the
normal, and the first shear directions, respectively. They are specified to
approximately equal to the Mode I and Mode II fracture toughness of the
composite material respectively. is the parameter for B-K form of damage
evolution criterion (details on damage evolution criterion of cohesive elements are
available in Section 2.5.3.3).
6.1.3 Meshing of 2D models
All 2D models were meshed with 4-node bilinear plane strain elements (CPE4).
For 2D Model #1 and #3, a coarser mesh was used with a total element number of 800 as
shown in Figure 6-4. For 2D Model #2, a refined mesh is used with a total element
171

number of 1600, as shown in Figure 6-5. The cohesive layer for 2D Model #3 was
meshed with COH2D4 elements.

6.1.4 Results of 2D models
6.1.4.1 Crack propagation
The bond state of nodes within the slave fracture surface can be written as an
output parameter by VCCT for Abaqus. As a result, the crack front can be identified after
each loading increment, and crack growth versus opening displacement curve can be
obtained after a crack propagation analysis. The crack growth results obtained for each
model is shown in Figure 6-6, along with experimental results obtained using visual
measurements on one edge of the specimen.


Fig. 6-4: Mesh configuration for 2D Model #1 and #3


Fig. 6-5: Mesh configuration for 2D Model #2
172

Comparing the results between finite element models (Figure 6-6), the
experimental and FE crack growth curves almost coincide with each other. However, the
FE result of crack growth is about 2-4 mm more that obtained from the experiment. This
could be explained by the curved crack front usually created during a DCB test. As
summarized in Chapter 3, it is commonly observed during DCB tests that the crack
length in the center-width of specimen is 2 to 5 mm more than that on the edge. It is the
crack growth on the specimen edge that can be measured during experiments. However,
for the 2D FE models presented here, the plane strain condition is assumed and the crack
growth is therefore the same across specimen width. The crack growth predicted by 2D
FE models should be close to the average crack growth across specimen width.
Therefore, the FE models are expected to over-predict the crack growth relative to the
visual experimental results.

40
50
60
70
80
90
0 2 4 6 8
Opening displacement, , mm
C
r
a
c
k

l
e
n
g
t
h
,

a
,

m
m
IFT Test
2D Model #1
2D Model #2
2D Model #3

Fig. 6-6: Crack growth versus opening displacement from test data and 2D finite element
modeling
173

6.1.4.2 Load-displacement curve
The load vs. displacement curve is indicative of the residual strength of a DCB
specimen as crack size grows. The load vs. displacement curves obtained from 2D FE
models are plotted in Figure 6-7 and along with the curve from the DCB test, which the
FE models are simulating.
From Figure 6-7, the following conclusions can be made:
1) The load vs. displacement curve predicted by any of the 2D models is not
smooth after crack initiation, as was previously pointed out in (Krueger 2008).
However, the curve is smoother in 2D Model #2 which has finer mesh and smaller
release tolerance than the other FE models. A stick-slip behavior is also observed in
the load-displacement curves of FE models, similar to that of a test specimen. This
behavior of FE models can be explained as follows: the slip (steep load drop)
portion is observed immediately after a pair of bonded nodes has just been released

0
5
10
15
20
25
30
35
40
45
0 2 4 6 8
Applied opening displacement, , mm
L
o
a
d
,

P
,

N
IFT test
2D Model #1
2D Model #2
2D Model #3

Fig. 6-7: Load vs. displacement curves from test data and 2D finite element modeling
174

after the release criterion has been met. As the opening displacement further increases,
the load increases as well, until the applied SERR of the node at crack front reaches
the release criterion again. During this linear loading period, a stick (linearly
increasing) behavior of the load-displacement curve is observed. Comparing the
post-initiation load vs. displacement curve from 2D Model #1 and that from Model
#2, it can be concluded that the amplitude of load drop and increase encountered in
one stick-slip step is related to the element size and node release tolerance used by
VCCT.
2) The FEM prediction by 2D Models #1 and #2 agrees well with the
experimental results. This indicates that the toughening behavior of this material
could be better modeled by specifying spatially varying critical SERR than using a
cohesive layer.
3) Using a cohesive layer combining with the VCCT interaction, 2D Model #3 at
first over-predicts the load shortly after crack initiation and then as loading proceeds
it coincides with the experimental data.
6.1.4.3 Stress distribution
To observe the distribution of stresses over the thickness of a DCB specimen, values
of the stresses in the longitudinal and thickness direction are plotted for the area around
the crack tip as shown in Figure 6-8 and Figure 6-9. Stress concentration was observed in
the area surrounding the crack tip.

175



Crack tip

Fig. 6-8: Contour plot of the stress in the longitudinal direction (
xx
), in MPa, around the
crack tip (2D Model #2)
176

6.2 Three-dimensional Modeling of the DCB Specimen
While the 2-dimensional (2D) models described above give good predictions of
load vs. displacement behavior of the DCB specimen with relatively small computation
cost, some details are ignored because of the plane strain assumption. Based on this
assumption, it is expected that the 2D finite element analysis results agree well with beam
theory, which is also based on the plane strain assumption. To simulate SERR
distribution along specimen width and thus analyze finite width effects, three 3-
dimensional (3D) models with straight and curved initial crack front shapes and/or
different fracture interaction schemes were built.


Crack tip
Fig. 6-9: Contour plot of the stress in the thickness direction (
22
), in MPa, around the
crack tip (2D Model #2)
177

6.2.1 Geometry, loading and boundary conditions of 3D models
The 3D model geometry is shown in Figure 6-10. Specimen length (l) is 100 mm
and initial crack length (a
0
) is about 50 mm. Specimen width (b) is 25 mm and total
thickness is 3.9 mm. The initial crack is located at the mid-thickness position of the
specimen. Loading is applied by specifying the linearly increasing opening displacements,
/2, at the ends of initially delaminated arms. At the other end, the specimen is restricted
from moving in the width (y) and thickness (z) directions.
6.2.2 Modeling techniques for 3D models
Crack propagation analysis using VCCT is a numerically intensive and expensive
process. The VCCT subroutine in Abaqus actively checks the bond state of nodes within
the fracture interfaces, and hence tracks the crack front along the fracture region. Only
nodes along the delamination front are allowed to debond during any given increment.
Because of the nonlinear property of the crack propagation analysis, very small time

I
n
i
ti
a
l
c
r
a
c
k
l
e
n
g
t
h
, a
0
S
p
e
c
i
m
e
n
le
n
g
t
h
, l
/2
/2
Fracture
interface
x
y
z
W
i
d
t
h
,

b

Fig. 6-10: Geometry of 3D models of the DCB specimen
178

increments are required for convergence of the solution. The crack propagation is
realized by the debonding of nodes along fracture interface. To avoid losing stability
during the analysis, the crack needs to propagate sequentially from node to node through
small time increments. In addition, the element size along the crack front needs to be
small enough so that a dramatic change of energy will not occur if the one or several
pairs of nodes debond(s) in the next increment. Since the 3D analysis requires a small
time increment and fine mesh, the crack propagation analysis simulation is very time
consuming. Thus, efficiency and convergence are two major considerations for 3D
models. For convergence of the solution after the crack initiates, stabilization techniques
are required. For crack propagation analysis using VCCT, the following stabilization
technique can be used (Abaqus 2007):
(i) Contact stabilizationapplied across only selected contact pairs and used to
control the motion of two contact pairs while they approach each other in multi-
body contact. Damping is applied when bonded contact pairs debond and move
away from each other.
(ii) Automatic or static stabilizationapplied to the motion of entire model,
commonly in models that exhibit statically unstable behavior such as buckling.
(iii) Viscous regulationonly applied to the nodes on contact pairs that have just de-
bonded. The viscous regulation damping causes the tangent stiffness matrix of
the softening material to be positive for sufficiently small time increments.
After the application of stabilization, it is expected that the crack propagation behavior
may not be physically correct, having been modified by the damping forces. It is
suggested in the Abaqus Manual to plot the damping energy and compare the results to
the strain energy in the model. If the damping parameter has been set properly, the value
of damping energy should appear to be a small portion of the total strain energy.
To simulate the crack propagation of specimen with an artificial defect, 3D Model
#1 was designed to have an initially straight crack front, as in a DCB test specimen with a
crack defined by an insert film. In a DCB test specimen, as the crack propagates, a
convex crack front forms under Mode I loading. However, to experimentally track the
crack front shape of the DCB test specimen in real time is very difficult. Hence, it is
179

interesting to find out whether the finite element method with VCCT is capable of
simulating the change in crack front shape as the crack propagates, and 3D Model #1 was
built for this purpose. To reveal the effect of the initial crack front shape on the
development of the subsequent crack front and compare with 3D Model #1, 3D Model #2
was built with an initially curved crack front. While with 3D Model #1, one design
initiative is to examine whether the initially straight crack front will become curved after
the crack propagates, with 3D Model #2, the initiative is to examine whether the opposite
will occur, i.e., the initially curved crack front will become flatter as the crack
propagates. In 3D Model #1 and 3D Model #2, only the fracture interaction defined by
VCCT was applied to the fracture surfaces and a constant fracture toughness of each
mode was specified. To determine the effect(s) of material toughening, in 3D Model #3,
an adhesive layer was added to bond the fracture surfaces in order to simulate the
toughening behavior. The release tolerance for VCCT was set to be a relatively larger
value (0.2) for all 3D models compared to 2D models to facilitate convergence of the
solutions. Details of the 3D models are listed as follows:
a) 3D Model #1
Initially straight crack front; the initial crack length is 50.5 mm;
Elastic properties of composite materials are:
E
11
= 110.0 GPa; E
22
= 10.16 GPa; E
33
= 10.16 GPa;

12
=
13
= 0.3;
23
= 0.436;
G
12
= G
13
= 4.6 GPa; G
23
= 3.54 GPa;
Fracture toughness: G
Ic
= 110 N/m; G
IIc
= 600 N/m; G
IIIc
= 1200 N/m. No
toughening is assumed as the crack grows;
For global automatic stabilization, the damping factor used is 2E-8.

b) 3D Model #2
Initially curved crack front. Initially, crack length at width-center of specimen
is 54 mm, and on the edge is 50.5 mm;
Element size along crack propagation direction in the region where fracture can
occur is about 0.5 mm;
180

Material constitutive and fracture properties are as described for 3D Model #1.
No toughening is assumed as the crack grows;
For global automatic stabilization, the damping factor used is 2E-8.

c) 3D Model #3
Initially straight crack front shape; Initial crack length is 50 mm;
Fracture interfaces are bonded by a cohesive layer in addition to the definition
of VCCT fracture interaction;
Elastic properties of the cohesive layer are:
E = 1.016 GPa; G
1
= G
2
= 1.016 GPa
The above cohesive element properties are the same as 2D Model #3.
Ultimate normal stress and shear stresses in two directions are:
t
n
0
= 0.75 MPa, t
s
0
= 0.75 MPa, t
t
0
= 0.75 MPa
These values were specified to be somewhat higher than those in 2D Model #3
to reduce the amount of load drop resulting from energy consumption by
stabilization.
The B-K form of damage evolution criterion was used, and the parameters are:
G
n
C
= 110 N/m, G
s
C

= 600 N/m, G
t
C

= 1200 N/m, = 1.
Here, G
n
C
, G
s
C
and G
t
C
are the critical fracture energies required to cause
failure in the normal, the first and second shear directions, respectively. They
are specified to approximately equal to the Mode I, Mode II and Mode III
fracture toughness of the composite material respectively. is the parameter
for B-K form of damage evolution criterion.
Material is toughened as the crack grows.
Constitutive properties are as described for the other 3D Models;
For global automatic stabilization, the damping factor used is 2E-11.
181

6.2.3 Meshing of 3D models
For all 3D models, the composite material was modeled with 8-node linear brick,
reduced integration elements (C3D8R). For 3D Model #1 (Figure 6-11), the total number
of elements is 15120. For 3D Model #2, shown in Figure 6-12, the total number of
elements is 16254. For 3D Model #3, as shown in Figure 6-13, a coarser mesh was used
with total element number of 13450, including 1000 cohesive elements. The cohesive
layer for 3D Model #3 was meshed with COH3D8 elements.


x, length
y, width
x, length
z,
thickness
Initial crack front
Fig. 6-11: Mesh configuration for 3D Model #1
182




x, length
y, width
x, length
z,
thickness
3.5 mm
Crack front Initial crack front
Fig. 6-12: Mesh configuration for 3D Model #2

x, length
y, width
x, length
z,
thickness
Cohesive layer
Fig. 6-13: Mesh configuration for 3D Model #3
183

6.2.4 Results of 3D Models
6.2.4.1 Crack front shape observation
The crack front shapes obtained from 3D Models #1 and #2 are compared to that
of a DCB test specimen next. In a DCB test specimen, the crack fronts are outlined by the
dashed lines in Figure 6-14, repeated here for convenience. In a FE model, the crack front
of a DCB specimen is implied from the bond state of nodes on the bottom and top crack
surfaces. This bond state can be requested as an output parameter for each node in the
bottom crack surface during an Abaqus analysis, and varies from 0 (fully debonded) to 1
(fully bonded). The bond states of the crack surfaces at several loading increments of 3D
Model #1 and 3D Model #3 are shown in Figure 6-14 and Figure 6-15, respectively. In
both figures, the red color indicates fully debonded area, and blue color indicates fully
bonded area, and the crack front is the intersection between them.


Approximated crack front
Fig. 6-14: Crack surfaces of a quasi-static DCB test specimen
184



Increment #650:
Increment #200:
After crack initiation, increment #1:
Fig. 6-14: Crack fronts predicted by 3D Model #1
185

In the following, the crack front shape predicted by 3D Model #1 and Model #3
are compared. For 3D Models #1 and #3, they both have an initially straight crack front;
however, material toughening is simulated by adding a cohesive layer bonding the
fracture surfaces. In Figure 6-14, which is obtained from 3D Model #1, it is shown that
the crack initiates at the center-width region of specimen. As the crack advances, the
crack front shape nearly stays straight. In Figure 6-15, which is obtained from 3D Model

Increment #1240:
Increment #300:
After crack initiation, increment #1:
Increment #1240:
Increment #300:
After crack initiation, increment #1:
Fig. 6-15: Crack fronts predicted by 3D Model #3
186

#3, it is shown that the crack initiates from center width region as well. In contrast to 3D
Model #1, as crack advances, the crack front becomes slightly convex relative to the left
end in 3D Model #3. This might indicate that a different crack front shape during crack
propagation might be caused by the toughening behavior of material as crack grows,
which is modeled by adding an adhesive layer between the crack surfaces in 3D Model
#3. Another difference between 3D Model #1 and #3 is that the element size in the width
direction is relatively smaller in 3D Model #1. Comparing the FE results to the
experimental results, the convexity predicted by FE analysis is less than that shown in the
DCB test specimen.
On the other hand, 3D Model #2 was built with an initially curved crack front
with no toughening behavior assumed. The contour plot of bond states of crack surfaces
at several increments are shown in Figure 6-16. It is shown that in this case the crack
initiates at the specimen edges. As the loading proceeds, the curved crack front tends to
grow flatter compared to the original crack front shape.
187





Increment #660:
Increment #300:
After crack initiation, increment #100:
Fig. 6-16: Crack fronts predicted by 3D Model #2
188

6.2.4.2 Load vs. displacement curve behavior
The load-displacement curves from the three 3D models are plotted in Figure 6-17
along with those from two 2D models. From Figure 6-17, the following conclusions can
be made:
i) The initial stiffness of the specimen (slope of the initial linear portion of the
curve) is slightly smaller in 3D models than in 2D models. This difference
might be a result of different constraint conditions, i.e.the plane strain
condition is assumed for 2D models. The initial stiffness for 3D Model #2 is
the smallest among all models because the curved crack front makes the
average initial crack length longer for 3D Model #2 than for other models.
ii) For all 3D models, a sudden load drop occurs at the crack initiation point.
This could be because damping is suddenly applied to the system
immediately after crack initiation. Global stabilization was used for all 3D
models. As a result, the damping energy increased as the crack grew. The
total strain energy of whole model and damping energy versus loading time
is plotted in Figure 6-18 for all 3D models. It is shown that the damping
energy is always less than 25% that of total strain energy. Immediately after
crack initiation, the load for 3D Model #1 or Model #3 is about 5 N lower
than that for 2D models. As the damping energy increases with crack
propagation in 3D models, the difference in load between 3D models and 2D
models increases.
iii) Comparing the result from 3D Model #3 to 3D Model #1, the post-initiation
load-displacement curve for 3D Model #3 appears to be shifted upward from
that of 3D Model #1, which indicates a material toughening in 3D Model #3.
This is contributed by the adhesive layer added between the crack surfaces in
3D Model #3.

189



0
5
10
15
20
25
30
35
40
45
0 0.5 1 1.5 2 2.5 3 3.5
Opening displacement, , mm
L
o
a
d
,

P
,

N
3D Model #1
3D Model #2
3D Model #3
2D Model #3
2D Model #2

Fig. 6-17: Load vs. displacement curves from 2D and 3D finite element analyses
190

6.2.4.3 Stress distribution
The stress distributions of all 3D models are very similar. As an example, the
distribution of longitudinal stress (
11
) in the specimen is plotted in Figure 6-19.
Total strain energy, 3D Model #1
Damping energy, 3D Model #1
Total strain energy, 3D Model #2
Damping energy, 3D Model #2
Total strain energy, 3D Model #3
Damping energy, 3D Model #3

Fig. 6-18: Comparison of damping energy to total strain energy for 3D models (The
loading speed is 2 mm/sec. Unit for time is second and for energy is Nmm.)
Crack
initiation
Crack
initiation
191

6.3 Conclusions
The virtual crack closure technique (VCCT) was used to model the crack
propagation behavior of the DCB specimen using Abaqus/Standard V6.7. In two-
dimensional models, two methods were used to simulate the slightly toughening behavior
of DCB specimens. For the first method, critical SERR value was specified as a spatially
varying material property. For the second method, a cohesive layer was used to bond the
fracture interaction surfaces in addition to the constant fracture toughness specified by
VCCT. The results show that the load-displacement curve predicted by the first method
match the experimental data better. For three-dimensional modeling, a stabilization
technique has to be used in order to obtain a converged solution. Possibly because of the
energy consumed by stabilization, an accurate prediction of load-displacement behavior
was not obtained. In terms of crack front shape, one 3D model predicts a slight convexity

Fig. 6-19: Distribution of longitudinal stress (
11
) in a 3D DCB specimen
192

of the crack front shape after crack propagation. However, compared to experimental
results, the curvature is less than that of a DCB test specimen.

193


Chapter 7

Conclusions and Recommendations
7.1 Mode I Interlaminar Fracture Toughness Characterization
Mode I interlaminar fracture toughness tests under quasi-static and cyclic loadings
were conducted with the Double Cantilever Beam (DCB) specimens. For the investigated
carbon/epoxy material system, from quasi-static tests critical SERR values onset from a
thin insert film (G
Ic
) determined by different calculation methods ranged from 110 to 150
J/m
2
. Compared to the results from the literature, G
Ic
of the investigated material is near
the low end of the common range for carbon/brittle epoxy materials. Based on a
comparison of alternative methods of creating the initial crack for Mode I testing, to base
the G
Ic
value on a short Mode I precrack is suggested for the investigated material system
for three reasons. First, there was larger variation in the critical SERR value based on
an insert film than that based on a Mode I precrack. In addition, in several cases the
critical SERR value based on an insert film was higher (unconservative) than that based
on a short Mode I precrack. Such behavior could be attributed to a resin rich pocket of
material existing just ahead of the inert film. Second, when using either the modified
beam theory or modified crack compliance method to calculate G
Ic
, the value based on
initiation at an insert film could be inaccurate because of the dissimilarity in crack front
shapes formed by an insert film and by the Mode I loading. Third, compared to the
resistance curves obtained for other types of unidirectional laminated material in the
literature, a relatively flat fracture resistance curve was obtained with crack extension for
the material system investigated, (the increase in toughness after 50 mm crack extension
is less than 40% of the G
Ic
value). The often-cited disadvantage of creating a much
higher critical SERR value based on a Mode I precrack does not exist for the material
system investigated. Based on these three arguments, the critical SERR value onset from
a short Mode I precrack is considered to represent a more repeatable and accurate crack
194

onset fracture toughness value for the currently investigated material system. Based on
quasi-static test results, the modified beam theory and modified compliance calibration
method are compared. These theories gave very similar results for the compliance
calibration and SERR value except for the results related to crack growth from the insert
film.
For Mode I fatigue tests, the modified Paris law ( ( )
I
n
ax I
G B dN da
Im
/ = ) was used
to fit the experimentally determined crack growth rate per cycle versus the maximum
applied Mode I SERR. Comparing to the exponents found in literature, which are in the
range of 3.6 to 15, the exponents found in this investigation (11.12 and 12.29) are near
the high end of the common range. For structural design, this material characteristic
implies that a small error in the crack driving force prediction results in a larger than
usual error in the crack growth prediction. Hence, in this case, it might be suitable to
consider a no-growth design criteria, which uses the threshold SERR value, G
Ith
, as the
fatigue fracture toughness for design purposes.

7.2 Mode II Interlaminar Fracture Toughness Characterization
The Mode II interlaminar fracture toughness tests under quasi-static and cyclic
loadings were conducted with the End Notched Flexure (ENF) specimen. For the quasi-
static tests, un-precracked and precracked ENF specimens were used. For Mode II testing,
the crack growth is usually unstable after onset if the common ENF test configuration is
used. Furthermore, even if an alternative test configuration were to be able to produce
stable crack growth, it is very hard to visually measure crack extension accurately under
mode II loading by ENF because of the extended crack tip being held tightly closed.
Therefore, the critical SERR value at crack onset is of main interest. During this
investigation, the average G
IIc
values based on an insert film determined by different
calculation methods ranged from 782 to 801 J/m
2
, while the average G
IIc
values based on
a Mode I precrack ranged from 498 to 545 J/m
2
. Compared to results in the literature, the
195

Mode II quasi-static fracture toughnesses of the investigated material are in the middle of
common range for carbon fiber composites made with brittle epoxies. The average G
IIc

value of un-precracked specimens is about 44% higher than that of precracked specimens
according to the classical plate theory method, and about 60% higher by the compliance
calibration methods. For compliance calibration, the third order polynomial relation in
the form of
3 3
) 8 ( Ba A bh C + = is recommended.
As was done in the Mode I fatigue tests, the Modified Paris law was fit to the
experimentally determined crack growth rate versus the maximum SERR for each Mode
II specimen. With the crack growth lengths calculated from compliance calibration, the
exponent for the power law relation was found to be in the range of 7.7 to 12.8. With the
crack growth lengths measured visually, the exponent was in the range of 7.1 to 6.8.
Compared to results in the literature (4.3-15), the exponent for the material investigated is
somewhat high. From the literature, for the same material the exponent of power law
relation for Mode I is usually higher than that for Mode II. This is also the case for the
investigated material.

7.3 Mixed Mode I/II Interlaminar Fracture Toughness Characterization
The mixed Mode I/II interlaminar fracture toughness (IFT) tests under quasi-static
and cyclic loading were conducted with single-leg-bending (SLB) specimens. The Mode
II to total strain energy release rate (SERR) ratio achieved with mid-thickness
delaminated SLB specimens was about 0.43. Given this mode ratio G
II
/G
Tc
and the fact
that the Mode II fracture toughness is about 5-8 times higher than the Mode I fracture
toughness, the SLB test results can be expected to be dominated by Mode I behavior.
From quasi-static tests, critical SERR values at crack onset from an insert film (G
Tc
)
calculated by three methods were in the range of 181 to 185 J/m
2
. The fracture resistance
curves were obtained for SLB tests with the crack length calculated from compliance
calibration methods. In all cases, the fracture toughness increased slightly with crack
196

extension. In particular, after about 20 mm crack extension, the fracture toughness
increased by about 11 to 53%.
Among all the fatigue crack growth models presented in literature review for
mixed mode crack growth in fatigue loading, the model given by ( )
max
/
n
da dN B G =
provides a good fit to crack growth rate per cycle versus maximum SERR data for the
investigated material system. The exponent of this modified Paris law for mixed-mode
I/II, which is in the range from 16 to 22, was higher than that for either Mode I (11-13) or
Mode II (7-13).

7.4 Preliminary Finite Element Modeling Results
The virtual crack closure technique (VCCT) was used to model the crack
propagation behavior of the DCB specimen using ABAQUS/Standard V6.7. In two-
dimensional (2D) models, by specifying the critical SERR value as a spatially varying
material property, a good match of load-displacement curves between the finite element
modeling results and experimental data was obtained. Meanwhile, crack length versus
opening displacement curves by 2D modeling and experiment were similar. However,
the model over predicts the crack length by 2-4 mm, compared to the crack length
measured on the specimen edge during the DCB test. For three-dimensional modeling,
stabilization technique(s) should be used in order to obtain a converged solution. Because
of the energy consumed by stabilization, an accurate prediction of load-displacement
behavior was not obtained. In terms of the crack front shape, the 3D model with an
initially straight crack front predicts a slight convexity of the crack front shape after crack
propagation. However, compared to experimental results, the curvature is less in the
model. For future work, the stabilization technique of 3D modeling needs to be improved
to give better prediction of load versus displacement behavior. Also, it is of interest to
study how the material properties are related to the curvature created by Mode I loading,
so that the inaccuracy of SERR calculations and compliance calibrations resulting from
197

the curved crack front shape can be corrected. Further, the accuracy of SERR calculations
for Mode II and mixed Mode I/II specimens and mode decomposition for mixed-mode
specimens predicted by classical theories should be examined by finite element methods.

198


Bibliography

Abaqus (2007). Abaqus Users Manual, Version 6.7.
Anderson, T. L. (2005). Fracture Mechanics Fundamentals and Applications, third
edition, Tayler & Francis Group, Boca Raton.
ASD-STAN preEN 6033. (1995). "Aerospace series carbon fibre reinforced plastics test
method determination of interlaminar fracture toughness energy mode I - GIc."
AeroSpace and Defense Industries Association of Europe - Standardization.
ASD-STAN prEN 6034. (1995). "Aerospace series carbon fibre reinforced plastics test
method determination of interlaminar fracture toughness energy mode II-GIIc."
AeroSpace and Defence Industries Association of Europe-standardization.
Asp, L., Sjgren, A., and Greenhalgh, E. (2001). "Delamination growth and thresholds in
a carbon/epoxy composite under fatigue loading." Journal of Composites
Technology and Research, 23(2), 55-68.
ASTM D5528-01. (2002). "Standard test method for Mode I interlaminar fracture
toughness of unidirectional fiber-reinforced polymer matrix composite." In:
ASTM Annual book of ASTM standard, vol. 15.03, American Society for Testing
and Materials, Philadelphia, 254-263.
ASTM D6115-97. (1997(R2004)). "Standard test method for mode I delamination growth
onset of unidirectional fiber reinforced polymer matrix composites." In: ASTM
Annual book of ASTM standard, vol. 15.03., American Society for Testing and
Materials., Philadelphia, 314-319.
ASTM D6671. (2006). "Standard test method for mixed mode I-mode II interlaminar
fracture toughness of unidirectional fiber reinforced polymer matrix composites."
In: ASTM Annual book of ASTM standard, vol 15.03., American Society for
Testing and Materials, Philadelphia, 377-390.
Benzeggagh, M., and Kenane, M. (1996). "Measurement of mixed-mode delamination
fracture toughness of unidirectional glass/epoxy composites with mixed-mode
bending apparatus." Composite Science and Technology, 56, 439-449.
199

Blanco, N., Gamstedt, E., Asp, L., and Costa, J. (2004). "Mixed-mode delamination
growth in carbon-fibre composite laminates under cyclic loading." International
Journal of Solids and Structures, 41, 4219-4235.
Brunner, A. (2000). "Experimental aspects of mode I and mode II fracture toughness
testing of fibre-reinforced polymer-matrix composites." Computer Methods in
Applied Mechanics and Engineering, 185(2-4), 161-172.
Brunner, A., Blackman, B., and Williams, J. (2006). "Calculating a damage parameter
and bridging stress from G
Ic
delamination tests on fiber composites." Composite
Science and Technology, 66, 785-795.
Bureau, M., Perrin, F., Denault, J., and Dickson, J. (2002). "Interlaminar fatigue crack
propagation in continuous glass fiber/polypropylene composites." International
Journal of Fatigue, 24, 99-108.
Carlsson, L., Gillespie, J., and Pipes, R. (1986). "On the analysis and design of the end
notched flexure specimen for mode II testing." Journal of Composite Materials,
20, 594-604.
Chambers, J. R. (2003). "Concept to reality: Contributions of the NASA Langley
research center to U.S. Civil aircraft of the 1990s."
<http://oea.larc.nasa.gov/PAIS/Concept2Reality/composites.html>.
Daniel, I. M., and Ishai, O., eds. (2006). Engineering Mechanics of Composite Materials,
second edition, Oxford University Press, New York.
Davidson, B., Altonen, C., and Polaha, J. (1996). "Effect of stacking sequence on
delamination toughness and delamination growth behavior in composite end-
notched flexure specimens." In: Composite Materials: Testing and Design
(Twelfth Volume), ASTM STP 1274, R. B. Deo and C. R. Saff, eds., American
Society for Testing and Materials, Philadelphia, 393-413.
Davidson, B., Fariello, P., Hudson, R., and Sundararaman, V. (1997). "Accuracy
assessment of the singular-field-based mode-mix decomposition procedure for the
prediction of delamination." In: Composite Materials: Testing and Design, vol. 13,
ASTM STP 1242, American Society for Testing and Materials, Philadelphia, 109-
128.
Davidson, B., Gharibian, S., and Yu, L. (2000). "Evaluation of energy release rate-based
approaches of predicting delamination growth in laminated composites."
International Journal of Fracture, 105, 343-365.
Davidson, B., and Koudela, K. (1999). "Influence of the mode mix of precracking on the
delamination toughness of laminated composites." Journal of Reinforced Plastics
and Composites, 18(15), 1408-1414.
200

Davidson, B., Kruger, R., and Konig, M. (1995a). "Three dimensional analysis and
resulting design recommendations for unidirectional and multidirectional end-
notched flexure test." Journal of Composite Materials, 29(16), 2108-2133.
Davidson, B., and Sundararaman, V. (1996). "A single leg bending test for interfacial
fracture toughness determination." International Journal of Fracture, 78, 193-210.
Davidson, B. D., Hu, H., and Schappery, R. A. (1995b). "An analytical crack tip element
for layered elastic structures." Journal of Applied Mechanics, 62, 294-305.
Davidson, B. D., Krger, R., and Knig, M. (1995c). "Three-dimensional analysis of
center-delaminated unidirectional and multidirectional single-leg bending
specimens. ." Composites Science and Technology, 54, 385-394.
Davies, P., Moulin, C., and Kausch, H. (1990). "Measurement of GIC and GIIC in
Carbon/Epoxy Composites." Composites Science and Technology, 39, 193-205.
Davies, P., Blackman, B. R. K., and Brunner, A. J. (1998). "Standard test methods for
delamination resistance of composite materials: current status." Applied
Composite Materials, 5, 345-364.
Davies, P., Blackman, B. R. K., and Brunner, A. J. (2001). "Mode II delamination." In:
Fracture Mechanics Testing Methods for Polymers, Adhesives and Composites, D.
R. Moore, A. Pavan, and J. G. Williams, eds., Elsevier Science, Oxford, UK, 307-
334.
Davies, P., Casari, P., and Carlsson, L. (2005). "Influence of fiber volume fraction on the
interlaminar fracture toughness of glass/epoxy using the 4ENF specimen."
Composites Science and Technology, 65, 295-300.
Davies, P., Ducept, F., Brunner, A. J., Blackman, B. R. K., and de Morais, A. B. (1996).
"Development of a standard mode II shear fracture test procedure." ECCM-CTS3,
Woodhead Publishing, 9.
de Morais, A. B., and Pereira, A. B. (2007). "Application of the effective crack method to
mode I and mode II interlaminar fracture of carbon/epoxy unidirectional
laminates." Composites: Part A, 38, 785-794.
Ducept, F., Davies, P., and Gamby, D. (1997). "An experimental study to validate tests
used to determine mixed-mode failure criteria of glass/epoxy composites."
Composites: Part A, 28A, 719-729.
Gregory, J., and Spearing, M. (2005). "Constituent and composite quasi-static and fatigue
fracture experiments." Composites: Part A, 36, 665-674.
201

Gustafson, C., and Hojo, M. (1987). "Delamination fatigue crack growth in unidirectional
graphite/epoxy laminates." Journal of Reinforced Plastics and Composites, 6, 36-
52.
Hashemi, S., Kinloch, A. J., and Williams, J. G. (1990a). "The effects of geometry, rate
and temperature on the mode I, mode II, and mixed-mode I/II interlaminar
fracture of carbon-fiber/poly(ether-ether ketone) composites." Journal of
Composite Materials, 24, 918-956.
Hashemi, S., Kinloch, A. J., and Williams, J. G. (1991). "Mixed-mode fracture in fiber-
polymer composite laminates." In: Composite materials: fatigue and fracture,
ASTM STP 1110, T. K. O'Brien, ed., American Society for Testing and Materials,
Pheladephia, 143-168.
Hashemi, S., Kinloch, J., and Williams, J. G. (1990b). "Mechanics and Mechanisms of
delamination in a poly(ether sulphone)-fiber composites." Composites Science
and Technology, 37, 429-462.
Hojo, M., Tanaka, K., Gustafson, C. G., and Hayashi, R. (1987). "Effect of stress ratio on
near-threshold propagation of delamination fatigue cracks in unidirectional
CFRP." Composites Science and Technology, 29, 273-292.
Hojo, M., Gustafson, C., and Tanaka, K. (1994). "Effect of matrix resin on delamination
fatigue crack growth in CFRP laminates." Engineering Fracture Mechanics, 49(1),
35-47.
Hojo, M., Kageyama, K., and Tanaka, K. (1995). "Prestandardization study on mode I
interlaminar fracture toughness test for CFRP in Japan." Composites, 26(4), 243-
255.
Hojo, M., Matsuda, S., Ochiai, S., Tsujioka, N., and Nakanishi, Y. (2001). "Mode II
interlaminar properties under static and fatigue loadings for CF/epoxy laminates
with different fiber-surface treatment." Advanced Composite Material, 10(2,3),
237-246.
Hojo, M., Matsuda, S., Tanaka, M., Ochiai, S., and Murakami, A. (2006). "Mode I
delamination fatigue properties of interlayer-toughened CF/epoxy laminates."
Composites Science and Technology, 66, 665-675.
Hojo, M., Ando, T., Tanaka, M., Adachi, T., Ochiai, S., and Endo, Y. (2006a). "Modes I
and II interlaminar fracture toughness and fatigue delamination of CF/epoxy
laminates with self-same epoxy interleaf." International Journal of Fatigue, 28,
1154-1165.
202

Hug, G., Thevenet, P., Fitoussi, J., and Baptiste, D. (2006). "Effect of the loading rate on
mode I interlaminar fracture toughness of laminated composite." Engineering
Fracture Mechanics, 73, 2456-2462.
ISO 15024. (2002). "Fiber-reinforced composites - Determination of mode I interlaminar
fracture toughness, Gic, of unidirectional fiber reinforced materials."
Irwin, G. R. (1956). "Onset of fast crack propagation in high strength steel and aluminum
alloys." Sagamore Research Conference Proceedings, vol 2, 289-305.
JIS K 7086. (1993). "Testing methods for interlaminar fracture toughness of carbon fiber
reinforced plastics." Japanese Standards Association.
Kalbermatten, T. D., Jggi, R., FLeler, P., Kausch, H. H., and Davies, P. (1992).
"Microfocus radiography studies during mode I interlaminar fracture tests on
composites." Journal of Materials Science Letters, 11(9), 543-546.
Kim, B., and Mayer, A. (2003). "Influence of fiber direction and mixed-mode ratio on
delamination fracture toughness of carbon/epoxy laminates." Composites Science
and Technology, 63(5), 695-713.
Kinloch, A. J., Wang, Y., and Williams, J. G. (1993). "The mixed-mode delamination of
fiber composite materials." Composites Science and Technology, 47, 225-237.
Krueger, R. (2002). "The virtual crack closure technique: history, approach and
applications." ICASE Report.
Krueger, R. and OBrien, T. K. (2001). A shell/3D modeling technique for the analysis
of delaminated composite laminates. Composites: Part A, 32, 25-44.
Martin, R., and Davidson, B. (1999). "Mode II fracture toughness evaluation using four-
point bend end-notched flexure test." Plastic Rubber Composites, 28(8), 432-437.
Martin, R., Elms, T., and Bowron, S. (Year). "Characterization of mode II delamination
using the 4ENF." Proceedings of the 4th European conference on composites
materials: testing and standardization, Institute of Materials, London, 161-170.
Mathews, M., and Swanson, S. (2007). "Characterization of the interlaminar fracture
toughness of a laminated carbon/epoxy composite." Composites Science and
Technology, 67, 1489-1498.
Matsubara, G., Ono, H., and Tanaka, K. (2006). "Mode II fatigue crack growth from
delamination in unidirectional tape and satin-woven fabric laminates of high
strength GFRP." International Journal of Fatigue, 28, 1177-1186.
Meo, M., and Thieulot, E. (2005). "Delamination modeling in a double cantilever beam."
Composite Structures, 71, 429-434.
203

Murri, G. B., and Martin, R. H. (1993). "Effect of initial delamination on mode I and
mode II interlaminar fracture toughness and fatigue fracture threshold." In:
Composite Materials: Fatigue and Fracture, 4th vol., ASTM STP 1156, W. W.
Stinchcomb and N. E. Ashbaugh, eds., American Society for Testing and
Materials, Philadelphia, 239-256.
Nakai, Y., and Hiwa, C. (2002). "Effects of loading frequency and environment on
delamination fatigue crack growth of CFRP." International Journal of Fatigue, 24,
161-170.
O'Brien, T. (1998a). "Interlaminar fracture toughness: the long and winding road to
standardization." Composites Part B, 29B, 57-62.
O'Brien, T. K. (1998b). "Composite interlaminar shear fracture toughness, G
IIc
: Shear
measurement or sheer myth?" In: Composite Materials: Fatigue and Fracture, 7th
vol., ASTM STP 1330, R. B. Bucinell, ed., American Scociety for Testing and
Materials, Philadephia, 3-18.
O'Brien, T. K., Murri, G. B., and Salpekar, S. A. (1989). "Interlaminar shear fracture
toughness and fatigue thresholds for composite materials." In: Composite
Materials: Fatigue and Fracture, 2nd vol., ASTM STP 1012, P. A. Lagace, ed.,
American Scociety for Testing and Materials, Philadelphia, 222-250.
Ozdil, F., and Carlsson, L. (1999). "Beam analysis of angle-ply laminate DCB
specimens." Composites Science and Technology, 59, 305-315.
Ozdil, F., Carlsson, L., and Davies, P. (1998). "Beam analysis of angle-ply laminate end-
notched flexure specimens." Composites Science and Technology, 58, 1929-1938.
Paris, P. (Year). "Fatigue-An Interdisciplinary Approach." Tenth Sagamore Conference,
Syracuse University Press, New York, 107.
Polaha, J., Davidson, B., Hudson, R., and Pieracci, A. (1996). "Effects of mode ratio, ply
orientation and precracking on the delamination toughness of a laminated
composite." Journal of Reinforced Plastics and Composites, 15, 141-172.
Qiao, P., Wang, J., and Davalos, J. (2004). "Analysis of tapered ENF specimen and
characterization of bonded interface fracture under mode II loading."
International Journal of Solids and Structures, 40, 1865-1884.
Ramkumar, R., and Whitcomb, J. (1985). "Characterization of mode I and mixed-mode
delamination growth in T300/5208 graphite/epoxy." In: Delamination and
debonding of materials, ASTM STP 876, American Society for Testing and
Materials, Philadelphia, 315-335.
204

Reeder, J. R. (1993). "A bilinear failure criterion for mixed-mode delamination." In:
Composite materials: testing and design, 11th vol., ASTM STP 1206, J.
Camponechi, E. T., ed., American Society for Testing and Materials, Philadelphia,
303-322.
Reeder, J. R., and Crews, J. (1990). "Mixed-mode bending method for delamination
testing." AIAA Journal, 28(7), 1270-1276.
Reeder, J. R., and Crews, J. J. H. (1991). "Nonlinear analysis and redesign of the mixed-
mode bending delamination test." NASA Technical Memorandum 102777.
Reeder, J. R., and Crews, J. J. H. (1992). "Redesign of the mixed-mode bending
delamination test to reduce nonlinear effects." Journal of Composites Technology
& Research, 14, 12-19.
Rhee, K. Y. (1994). "Characterization of delamination behavior of unidirectional
graphite/PEEK laminates using cracked lap shear (CLS) specimens." Composite
Structures, 29, 918-956.
Russell, A., and Street, K. (1989). "Predicting interlaminar fatigue crack growth rates in
compressively loaded laminates." In: In: Delamination and debonding of
materials, ASTM STP 1012, American Society for Testing and Materials,
Philadelphia, 162-178.
Russell, A. J., and Street, K. N. (Year). "A constant G test for measuring mode I
interlaminar fatigue crack growth rates." Composite Materials: Testing and
Design (Eighth Conference), ASTM STP 972, Philadelphia, 259-277.
Russell, A. J., and Street, K. N. (1985). "Moisture and Temperature Effects on the
Mixed-mode Delamination Fracture of Unidirectional Graphite/Epoxy." In:
Delamination and Debonding of Materials, ASTM STP W. S. Johnson, ed.,
American Society of Testing and Materials, Philadelphia, 349-370.
Schn, J., Nyman, T., Blom, A., and A., A. (2000). "A numerical and experimental
investigation of delamination behaviour in the DCB specimen." Composites
Science and Technology, 60, 173-184.
Schuecker, C., and Davidson, B. (2000). "Evaluation of the accuracy of the four point
bend end-notched flexure test for mode II delamination toughness determination."
Composites Science and Technology, 60, 2137-2146.
Shivakumar, K., Chen, H., Abali, F., Le, D., and Davis, C. (2006). "A total fatigue life
model for mode I delaminated composite laminates." International Journal of
Fatigue, 26, 33-42.
205

Singh, S., and Partridge, I. K. (1995). "Mixed-mode fracture in an interleaved carbon-
fiber/epoxy composite." Composites Science and Technology, 55, 319-327.
Szekrnyes, A., and Jzsef, U. (2006). "Comparison of some improved solutions for
mixed-mode composite delamination coupons." Composite Structures, 72, 321-
329.
Szekrnyes, A., and Uj, J. (2005). "Mode-II fracture in E-glass/Polyester Composite."
Journal of Composite Materials, 39(19), 1747-1768.
Szekrnyes, A., and Uj, J. (2007). "Over-leg bending test for mixed-mode I/II
interlaminar fracture in composite laminates." International Journal of Damage
Mechanics, 16, 5-33.
Tanaka, K., and Tanaka, H. (1997). "Stress-ratio effect on mode II propagation of
interlaminar fatigue cracks in graphite/epoxy composites." In: Composite
Materials: Fatigue and Fracture, 6th vol., ASTM STP 1285, E. A. Armanios, ed.,
American Society for Testing and Materials, Philadelphia, 126-142.
Tanaka, K., Kageyama, K., and Hojo, M. (1995). "Prestandardization study on mode II
interlaminar fracture toughness test for CFRP in Japan." Composites, 26, 256-267.
Tanaka, K., Yuasa, T., and Katsura, K. (1998). "Continuous mode II interlaminar fracture
toughness measurement by over notched flexure test." Testing and
Standardization: Proceedings of the 4th European Conference on Composites,
European Society of Composite Materials, Lisbon, Portugal, 171-179.
Tracy, G., Feraboli, P., and Kedward, K. (2003). "A new mixed mode test for
carbon/epoxy composite systems." Composites: Part A, 34(11), 1125-1131.
VCCT for Abaqus Manual. (2007). "User's Guide, Version 1.3-2."
Vinciquerra, A., Davidson, B., Schaff, J., and Smith, S. (2002). "Determination of the
mode II fatigue delamination toughness of laminated composites." Journal of
Reinforced Plastics and Composites, 21(7), 663-677.
Walsh, T. F., and Bakis, C. E. (1995). "The effect of high-temperature degradation on the
mode I critical strain energy release rate of a graphite/epoxy composite." Journal
of Composites Technology & Research, 17(3), 228-234.
Wang, H., and Vu-Khanh, T. (1996). "Use of the end-loaded-split test to study stable
fracture behavior of composites under mode-II loading." Composite Structures, 36,
71-79.
206

Wang, J., and Qiao, P. (2003). "Fracture toughness of wood-wood and wood-FRP bonded
interfaces under mode II loading." Journal of Composite Materials, 37(10), 875-
897.
Wilkins, D. J., Eisenmann, J. R., Camin, R. A., Margolis, W. S., and Benson, R. A.
(1982). "Characterizing delamination growth in graphite-epoxy." In:
Delamination and Debonding of Materials, ASTM STP 775, American Society for
Testing and Materials, Philadelphia, 315-335.
Williams, J. G. (1989). "End corrections for orthotropic DCB specimens." Composites
Science and Technology, 35(4), 367-376.
207

Appendix A

IFT Test Results From Literature
Specimen dimension
Material Fiber Matrix
Loading
speed,
mm/min
Length,
L, mm
Thickness,
2h, mm
Width,
b, mm
Initial
crack
length,
a
0
, mm
Insert type Precrack G
Ic
(J/m
2
) Source
UD FRP E-glass Epoxy 1 150 5 20 35 and 50
Polypropylene
8 m
NO
a
0
= 35 mm, NL:243, 5%:
317, AE:246; (initiation)
a
0
= 50 mm, NL:268, 5%:
343, AE:178; (initiation)
(Ducept et al.
1997)
UD FRP E-glass M11 2 150 6 20 23-55 Teflon film N/A
Initiation value:
118.022.72
(Benzeggagh
and Kenane
1996)
UD FRP Carbon Epoxy 1-5 4.2 20
PTFE 60 and
12.6 m;
aluminum foil
20 and 40 m
with and
without
100-200 (initiation),
150-250 (propagation)
(Davies et al.
1990)
UD FRP
HR
carbon
(T300)
Epoxy
(914)
0.5 190 4.16 20 30 N/A N/A 165 (initiation) (Hug et al. 2006)



Table A-1: DCB tests results from literature
208

Table A-1. Continued
Specimen dimension
Material Fiber Matrix
Loading
speed,
mm/min
Length,
L, mm
Thickness,
h, mm
Width,
b, mm
Initial
crack
length,
a
0
, mm
Insert type Precrack G
Ic
(J/m
2
) Source
IM7 8552 22010 (initiation)
HTA7 6376 2303 (initiation) UD FRP
T300 914
220 3.12 20 50
Release film
A4000, 10 m

11210 (initiation)
(Schn et al.
2000)
UD FRP IM7 8551-7A 3.2 25 40
PTFE 25.4
and 12.7 m
None and
wedge open
220-284 (25.4 m insert);
144-176 (12.7 m insert)
(Walsh and
Bakis 1995)
[0]
6
4.38 NO
28242 (initiation)
[30]
5
7.3 NO
21414 (initiation);
500 (steady state)
[45]
5

E-Glass
Isophthalic
Polyester
0.76
7.3
20 33
Polypropylene
8 m
NO
17632 (initiation);
300 (steady state)
(Ozdil and
Carlsson 1999)
See
Figure A-1
Ciba-
Geigy
graphite
Epoxy,
R6376
0.75 4.06 25.4 57.2 Teflon
with and
without
See Figure A-2
(Polaha et al.
1996)
209


















mean
150
200
250
300
350
400
G
Ic
,
J/m
2
G
Ic
, of non-precracked DCB specimens
[ ] 0
+
+
15

+
-
15

+
+
30
c
D

+
-
30
c
D

+
+
30
t
B

+
-
30
t
B


mean
150
200
250
300
350
400
G
Ic
,
J/m
2
G
Ic
, of precracked DCB specimens
[ ] 0
+
+
15

+
-
15

+
+
30
c
D

+
-
30
c
D

+
+
30
t
B

+
-
30
t
B


Fig. A-2: G
Ic
values of different layups from (Polaha et al. 1996).


Specimen notation Layup
[ ] 0

[ ]
32
0

+
+
1 5


+ -
- 4 +
15/0/-15/0/15/0 / 15 / 0 / 15 / 0 / 15
s



+
-
15


+ -
- 4 +
- +
+ 4 -
[ 15/0/-15/0/15/0 / 15 / 0 / 15 / 0 / 15/ /
15/0/15/0/-15/0 / 15 / 0 / 15 / 0 / 15]
d


+
+
30
c
D


+ -
- 4 +
30/0/-30/0/30/0 / 30/ 0/ 30/ 0/ 30
s



+
-
30
c
D


+ -
- 4 +
- +
+ 4 -
[ 30/0/-30/0/30/0 / 30 / 0 / 30 / 0 / 30/ /
30/0/30/0/-30/0 / 30 / 0 / 30 / 0 / 30]
d


+
+
30
t
B


+ -
- 12 +
30/0 / 30
s



+
-
30
t
B


+ - - +
- 12 + + 12 -
30/0 / 30/ / 30/ 0 / 30 d


Fig. A-1: Lay-up of specimens, based on (Polaha et al. 1996).
210


Material Fiber Matrix
Loading
speed,
mm/min
Half Span
Length, L,
mm
Thickness,
2h, mm
Width,
b, mm
Initial
crack
length,
a
0
, mm
Insert type Precrack G
IIc
(J/m
2
) Source
UD FRP E-glass Epoxy 1 65 5 20 35
Polypropylene 8
m
no
1460 (NL);
2185 (5%);
1595 (AE)
(Ducept et
al. 1997)
UD FRP Carbon Epoxy 1-5 40 4.2 20
PTFE, 60 and
12.6 m;
aluminum foil,
20 and 40 m
w/wo
~500 (mode
I precrack);
~1000-1500
(starter film)
(Davies et
al. 1990)
[0]
6
4.4 20 25 no 496135
[30]
5
7.3 20 25 no 97671
[45]
5

E-Glass
Isophthalic
polyester
1 50
7.3 20 25
Polypropylene 8
m
no 1485158
(Ozdil et
al. 1998)
PI 25 m folded w/wo 372/1036
PI 12.5 m w/wo 417/480
PI 7.5 m w/wo 257/304
UD CFRP
CF
(Besfight
HTA)
Brittle
epoxy
3 mm
(22 plies)
PTFE 12.5 m w/wo 320/370
PI 25 m folded w/wo 920/890
PI 12.5 m w/wo 560/740
PI 7.5 m w/wo 610/460
Woven
CFRP
CF
(Besfight
T400HB)
Brittle
epoxy
3 mm
(16 plies)
PTFE 12.5 m w/wo 570/640
Toughened
UD CFRP
CF (IM-
600)
Toughened
epoxy
PI 12.5 m w/wo 930/991
UD CFRP AS4 CF PEEK
0.5 50

25 25
PI 7.5 m w/wo
1530/147
7
(Tanaka et
al. 1995)
Table A-2: ENF test results from literature
211

Specimen Fatigue loading Paris' law parameters
Material
Lay-up Specimen G
II
/G
Dimension
(Lb2h)
Initial
crack
length, a
0

Frequency
Load
ratio, R
G
c
**
Coefficient,
B
Exponent,
n
Reference
mm mm Hz J/m
2
mm/cycle
UD
carbon/epoxy,
HTA/6376C

MMELS
*
0.5 150203.3 N/A 2 0.1 447 1.6810
-7
6.28
(Blanco et
al. 2004)
UD
graphite/epoxy,
IM7/977-3
[0]
28
DCB 0 150257.5 50 5 0.1 200 15
(Gregory
and Spearing
2005)
AT400/epoxy [0]
26
ENF 1
170253
span length
100
25 10 0.5 610 6.5 and 4.6
(Hojo et al.
2001)
UD
graphite/epoxy
[0]
26
ENF 1 20025NA 31.75 10 0.1
(Tanaka
1997)
Woven E-
glass/vinyl ester
DCB 0 254385 38 1-4 0.1 350 5.4
(Shivakumar
et al. 2006)
UD
graphite/epoxy,
T300/914C
DCB 0 145N/A8 N/A
0.1, 0.3,
0.5
1.3210
-25
8.81
(Gustafson
and Hojo
1987)
Glass/polypropy
lene
ENF 1
130204,
total span
110
26-51 5 0.1
1.410
-5
-
1.110
-3

4.3-4.8
(Bureau et
al. 2002)
UD
carbon/epoxy
[0]
16

T800H/3900-2
DCB 0 140203 20- 25 10 0.1- 0.5 180 3.6-7.3
(Hojo et al.
2006b)

Table A-3: Fatigue test results from literature
( ) [ ]
s 8 5
0 , 5 // 0
( ) [ ]
s 8 20
0 , 5 // 0
212


Table A-3. Continued
Specimen Fatigue loading Paris' law parameters
Material
Lay-up Specimen G
II
/G
Dimension
(Lb2h)
Initial
crack
length, a
0

Frequency
Load
ratio, R
G
c
**
Coefficient,
B
Exponent,
n
Reference
mm mm Hz J/m
2
mm/cycle
[0]
16

T800H/3900
-2
0.1- 0.5 180 3.6-7.3
UD
carbon/epoxy
[0]
24

UT500/111
DCB 0 140203 20- 25 10
0.2- 0.5 150


6-8
(Hojo et
al.
2006b)
DCB 0 140203 25 160
4.6 and
5.0 UD
CF/epoxy,
UT 500/111

ENF
160203,
total span
100
25
10
0.1 and
0.5
470

13 and 15
(Hojo et
al.
2006a)
DCB 0 1.210
-7
5.5
ENF 1 7.510
-7
4.4
carbon/epoxy
,
HTA/6376C
[0
12
//(
5/0
4
)
S
]
MMB 0.5
150203.1 35 5 0.1
2.510
-8
6.3
(Asp et
al. 2001)

* Mixed mode end load split specimen
** Critical strain energy release rate from static tests
213


Appendix B

Additional Specimen Information and Test Results
1. Dimensions of Specimens


Table B-1: Dimensions of DCB specimens used for quasi-static tests
Specimen
Dimension Notation
1-1
1-2 1-3 1-4 1-5
Length:
l
152.5 152.8 153.2 152.4 152.7
Nominal width: b
24.5 25.2 25.2 25.1 25.1
Nominal thickness:
2h
3.7 3.9 3.9 4.0 3.9
Initial crack length:
a
0

49.3 51.4 52.4 50.2 51.1

Table B-2: Dimensions of DCB specimens used for fatigue tests
Specimen
Dimension Notation
4-2 4-3 4-4 4-5
Length: l 153.0 153.1 153.2 153.3
Nominal width: b 25.1 25.0 25.0 25.1
Nominal thickness:
2h 4.0 4.0 3.9 3.8
Initial crack length:
a
0
51.8 53.0 52.9 50.0

214





Table B-3: Dimensions of ENF specimens (un-precracked, tested in quasi-static tests)
Specimen
Dimension (mm) Notation
2-1 2-2 2-3 2-4
Length:
l 152.5 153.1
152.9 152.6
Nominal width:
b 25.0 25.0
25.0 25.1
Nominal thickness:
2h
3.8 3.9 3.9 3.8
Initial crack length:
a
0

24.6 25.1 23.8 24.2

Table B-4: Dimensions of ENF specimens (precracked, tested in quasi-static tests)
Specimen
Dimension (mm) Notation
2-5 2-6 2-7
Length:
l 152.5 152.6 152.7
Nominal width:
b 25.3 25.5 25.3
Nominal thickness:
2h 3.9 3.9 3.9
Initial crack length
created by insert film:
a
0
32.3 32.1 31.1
Initial crack length after
precracking:
a
p
35.1 35.0 33.6


Table B-5: Dimensions of ENF specimens (precracked, tested in fatigue tests)
Specimen
Dimension (mm) Notation
5-3 5-4 5-5
Length:
l 152.6 152.7 152.7
Nominal width:
b 25.3 25.3 25.5
Nominal thickness:
2h 3.9 3.9 3.9
Initial crack length
created by insert film:
a
0
33 31.1 31.1
Initial crack length after
precracking:
a
p
36.8 34.5 34.7


215




Table B-6: Dimension of SLB quasi-static specimens

SLB, Quasi-static Specimen
Dimension (mm) Notation
3-1 3-2 3-3 3-4
Length:
l 152.6 151.6
153.3 151.7
Nominal width:
b 25.1 25.1
25.2 25.1
Nominal thickness:
2h
3.9 3.9 3.9 3.9
Initial crack length:
a
0

38.4 38.7 38.4 38.2

Table B-7: Dimension of SLB fatigue specimens
SLB, Fatigue Specimen
Dimension (mm) Notation
6-1 6-2 6-3
Length:
l 151.8 151.6
152.2
Nominal width:
b 25.2 25.2
25.0
Nominal thickness:
2h
3.9 3.9 3.9
Initial crack length:
a
0

40.0 38.4 38.9

216


2. Mode I Quasi-static Fracture Toughness at Crack Onset









Table B-8: A summary of Mode I quasi-static test results

Critical point Mode I Critical SERR
Initial
crack
length
Displacement Load
MCC,
initial onset
MBT,
initial onset
Precrack,
by MCC
Specimen
No.
a
0
, mm
cr
, mm P
cr
, N G
Ic
, J/m
2

1-1 49.285 2.824 38.1 120 111 142
1-2 51.441 2.675 40.4 124 111 129
1-3 52.447 2.852 48.7 161 145 118
1-4 50.246 2.750 42.8 136 120 130
1-5 51.080 3.010 44.7 146 137 134
Average 2.822 42.9 137 125 131
STD * 17.0 15.5 8.7
COV ** 12.4% 12.4% 6.7%
* Standard deviation;
** Coefficient of variation, STD/mean.
217

3. Mode I Fatigue Crack Growth Rate vs. Maximum SERR Plots

Fig. B-1: da/dN - G
Imax
plot for Specimen 4-2 (Crack growth measured visually.)

Fig. B-2: da/dN - G
Imax
plot for Specimen 4-2 (Crack growth calculated by compliance
calibration.)
218




Fig. B-3: da/dN - G
Imax
plot for Specimen 4-3 (Crack growth measured visually.)

Fig. B-4: da/dN - G
Imax
plot for Specimen 4-3 (Crack growth calculated by compliance
calibration.)
219




Fig. B-5: da/dN - G
Imax
plot for Specimen 4-4 (Crack growth measured visually.)

Fig. B-6: da/dN - G
Imax
plot for Specimen 4-4 (Crack growth calculated by compliance
calibration.)
220




Fig. B-7: da/dN - G
Imax
plot for Specimen 4-5. (Crack growth measured visually.)

Fig. B-8: da/dN - G
Imax
plot for Specimen 4-5 (Crack growth calculated by compliance
calibration.)
221

4. Mode II Quasi-static Fracture Toughness at Crack Onset






Table B-9: A summary of Mode II quasi-static test results

Critical point Mode II Critical SERR
Initial
crack
length
Displacement Load by CPT by CC (1) by CC (2)
Specimen
No.
a
0
, mm
cr
, mm P
cr
, N G
IIc
, J/m
2

2-1
24.6 2.013 980 733 837
801
2-2
25.1 2.294 1086 808 745
812
2-3
23.8 2.308 1108 851 821
833
2-4
24.2 2.156 1007 736 801
740
STD
57.6 40.1 39.9
Mean 24.4 2.193 1045.3 782 801 797
COV

7.37% 5.01% 5.01%
2-5 35.1 1.624 610.4 576.0 518.0 505
2-6 35.0 1.567 561.6 497.0 436.0 484
2-7 33.6 1.629 630.7 563.0 539.0 511
STD 42.4 54.4 14.2
Mean 34.6 1.607 600.9 545 498 500
COV 7.77% 10.94% 2.84%

222

5. Mode II Fatigue Crack Growth Rate vs. Maximum SERR Plots

Fig. B-9: da/dN - G
IImax
plot for Specimen 5-3 (Crack growth rate was calculated with
crack growth by compliance calibration. The points excluded from linear fit are near the
crack growth arrest domain.)

Fig. B-10: da/dN - G
IImax
plot for Specimen 5-3 (Crack growth rate was calculated with
crack growth by visual measurement. The points excluded from linear fit are taken near
the beginning of the test.)
223




Fig. B-11: da/dN - G
IImax
plot for Specimen 5-4 (Crack growth rate was calculated with
crack growth by compliance calibration. The points excluded from linear fit are taken
near the beginning of the test.)
224



Fig. B-12: da/dN - G
IImax
plot for Specimen 5-5 (Crack growth rate was calculated with
crack growth by compliance calibration. The points excluded from linear fit are taken
near the beginning of the test.)

Fig. B-13: da/dN - G
IImax
plot for Specimen 5-5 (Crack growth rate was calculated with
crack growth by visual measurement. The points excluded from linear fit are taken near
the beginning of the test.)
225

6. Mixed Mode I/II Quasi-static Fracture Toughness at Crack Onset













Table B-10: A summary of Mode I quasi-static test results

Critical point Mixed Mode I/II Critical SERR Initial
crack
length
Displacement Load
by CPT by CC (1) by CC (2)
Specimen
No.
a
0
, mm
cr
, mm P
cr
, N G
Tc
, J/m
2

3-1
38.379 1.249 208.6
177 173 167
3-2 38.718 1.358 218.5 196 207 198
3-3
38.447 1.290 207.9
178 173 181
3-4 38.172 1.325 215.0 187 187 178
Mean 185 185 181
STD 38.4 1.306 212.5 8.9 16.1 12.8
COV.

4.8% 8.7% 7.1%

226

7. Mode II Fatigue Crack Growth Rate vs. Maximum SERR Plots

Fig. B-14: da/dN - G
max
plot for Specimen 6-1 (Crack growth was calculated by
compliance calibration.)

Fig. B-15: da/dN-G
max
plot for Specimen 6-1 (Crack growth was measured visually.)
227




Fig. B-16: da/dN-G
max
plot for Specimen 6-2 (Crack growth was calculated by
compliance calibration.)

Fig. B-17: da/dN - G
max
plot for Specimen 6-2 (Crack growth was measured visually.)
228





Fig. B-18: da/dN-G
max
plot for Specimen 6-3 (Crack growth was calculated by
compliance calibration.)

Fig. B-19: da/dN-G
max
plot for Specimen 6-3 (Crack growth was measured visually.)
229


Appendix C

Non-Technical Abstract
In laminated fiber-reinforced composite materials, high-strength fibers are
combined with a ductile, light-weight matrix to achieve good mechanical properties in
the fiber directions. According to engineering needs, thin layers with various
unidirectional orientations of fibers are selected and bonded together to achieve the
desired properties. With this manufacturing method, high mechanical performance and
great design flexibility are achieved. This type of material is ideal for load bearing
components in many engineering fieldsparticularly those where low weight is
important. One weakness of this type of material is delamination or interlaminar fracture,
where adjacent layers separate from each other. Delamination is especially important
because it could cause a catastrophic loss of compressive strength of a structural
component if undetected in nondestructive inspections. In developing new materials, it is
essential to characterize the materials resistance to delamination, which is named
interlaminar fracture toughness (IFT). The purpose of this investigation is to characterize
the IFT properties of a carbon/epoxy laminated material system in quasi-static and fatigue
loadings. Additionally, it is the goal of this investigation to examine the state-of-the-art
IFT test methods, tailor them to suit the current material, and make recommendations for
standardizing the test methods for similar material systems. According to the relative
displacement between the fracture surfaces, the delamination behavior can be
decomposed into three modes: Mode I, Mode II, and Mode III. The Mode I, Mode II, and
Mixed-mode I/II interlaminar fracture toughnesses of the investigated material were
characterized with the DCB specimen, ENF specimen, and SLB specimen, respectively.
Under Mode I loading, the DCB quasi-static test results showed that for the investigated
material system the fracture toughness is low in comparison to results in the literature.
Under Mode II loading, the quasi-static ENF test results showed that the fracture
toughness for the investigated material system is in the middle of common range. Under
230

fatigue loading, tests in all modes showed that the delamination growth rate decreased
greatly as the crack driving force decreased. The test results can be used to complete the
database of the properties of the investigated material system. Some controversial test
methods found in the literature were examined for the investigated material system, and
recommendations were made to give more repeatable and accurate characterization of
fracture toughness for similar material systems.

You might also like