strength and failure mechanisms of composite laminates subject to localised transverse loading
TRANSCRIPT
Composite Structures 69 (2005) 301–314
www.elsevier.com/locate/compstruct
Strength and failure mechanisms of composite laminatessubject to localised transverse loading
Gordon Kelly *, Stefan Hallstrom
Division of Lightweight Structures, Department of Aeronautical and Vehicle Engineering, Kungl Tekniska Hogskolan, S-100 44 Stockholm, Sweden
Available online 21 August 2004
Abstract
The behaviour of composite laminates subject to transverse load introduction has been investigated experimentally and numer-
ically. The effect of the specimen size, stacking sequence and material system on the failure load was determined experimentally and
the failure modes examined through fractographic analysis. Damage was found to initiate at low load levels, typically 20–30% of the
failure load. The dominant initial failure mode was matrix intralaminar shear failure which occurred in sub-surface plies. The dam-
age developed into a network of intralaminar and interlaminar shear cracks. Two different macromechanical failure modes were
identified, fastener pull-through failure and global collapse of the laminate. The internal damage and ultimate failure mode were
found to depend upon the laminate stacking sequence and resin system.
A three-dimensional finite element model was developed to analyse the stress distribution within the laminate and predict first-ply
failure. The results from the finite element model were found to be in general agreement with the experimental observations.
� 2004 Elsevier Ltd. All rights reserved.
Keywords: Composite laminate; Localised transverse loading; Failure; Strength; Experimental investigation; Finite element analysis
1. Introduction
Concentrated transverse loading in composite lami-
nates is commonly avoided due to the poor interlaminar
and through-thickness strength of the materials. How-
ever, with the increasing use of such materials in load
bearing structures has come a need to develop design
principles for both in-plane and out-of-plane loading.
In-plane load introduction has been investigated inten-sively in the form of mechanically fastened joints. A
wealth of literature exists concerning experimental, ana-
lytical and numerical analysis of parameters which influ-
ence the behaviour of composite laminates loaded
in-plane. The behaviour of composite laminates subject
to out-of-plane loading has not been investigated or
0263-8223/$ - see front matter � 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2004.07.008
* Corresponding author.
E-mail address: [email protected] (G. Kelly).
understood to the same degree. Limited published liter-ature regarding the transverse load bearing capacity of
composite laminates exists. The few studies published
concern the selection of countersunk fasteners [1], fas-
tener pull-through failure associated with post-buckled
structures [2–4] and quasi-static modelling of transverse
impact [5,6].
Chen and Lee [1] investigated the behaviour of car-
bon fibre/epoxy (AS4/3501-6) cross-ply laminates sub-ject to bending loads. Rectangular laminates were
loaded in three-point bending by countersunk fasteners
with different head geometries. The authors conducted
a progressive failure analysis of the laminates with fail-
ure being predicted based on the maximum stress theory
and the Ye delamination criterion [7]. The effect of fric-
tion between the fastener and the hole was found to be
significant together with the bending stiffness of the lam-inate. Reasonable agreement was found between the
experimental and predicted failure loads.
composite plate
Fastenerload attachment
restraining plate
washer
Fig. 1. Schematic of the experimental set-up.
302 G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314
Waters and Williams [2] conducted an experimental
investigation of bolt push through failure which was a
possible failure mode in skin-stiffener attachments in
aerospace structures at the time. The study considered
the response and failure modes of composite laminates
loaded by a countersunk fastener. Circular specimensof diameter 25.4 and 50.8 mm were clamped at the outer
boundary and a centrally located fastener was used to
apply the transverse load. A series of material systems
were investigated including combinations of different
carbon and kevlar fibres with thermoset and thermo-
plastic matrices. A number of conclusions were drawn
from the study. The effect of specimen size, or boundary
diameter, was not found to significantly affect the loadcarrying capability of the laminates. The behaviour of
laminates made with brittle matrices were improved
through the use of transverse stitching with the initial
failure load found to be dependent on the stitching pat-
tern. It was also apparent from the results presented in
the report that the load capability was not strongly
dependent upon the laminate stacking sequence. How-
ever, the effect of the laminate stacking sequence wasnot addressed explicitly.
Banbury and Kelly [3] conducted an experimental
investigation of fastener pull-through failure in lami-
nates made from woven fabric and unidirectional tape
pre-preg material. The experimental set-up was consist-
ent with [2] with clamped circular specimens of diameter
31.8 mm being centrally loaded via a 3.97 mm diameter
fastener. The effects of the specimen and fastener geom-etry on the failure load were determined and the failure
mechanisms analysed. The observed failure modes were
matrix shear fracture and tensile failure which were
found to be dependent upon the bending stiffness of
the laminate. Matrix shear cracking was found to occur
in almost all tests while tensile failure occurred in lami-
nates with reduced bending stiffness. The internal dam-
age pattern in the laminates was likened to thatobserved in laminates subject to low velocity impact.
Failure was attributed to a critical level of matrix strain
which was independent of the laminate and fastener
geometry and material type for a given resin system.
Banbury et al. [4] conducted a two-dimensional axi-sym-
metric finite element analysis to simulate the fastener
pull-through tests in [3]. A progressive damage model
based on the maximum strain criterion was used to pre-dict failure in the laminates although no consideration
was given to interlaminar failure. The results from the
model were consistent with the experimental observa-
tions and the predicted failure loads were shown to be
in good agreement with the experimentally determined
values.
There are currently no widely accepted design rules or
established methods to accurately predict the structuralperformance of composite laminates subject to concen-
trated transverse load introduction. The aim of this
investigation is to provide a greater understanding of
the mechanisms which govern the structural behaviour
and failure of composite laminates under such loading
conditions.
2. Experimental investigation
The main objectives of the experimental work were to
study the effect of selected parameters on the structural
behaviour of laminates subject to transverse loading and
to investigate the micromechanical failure modes and
development of damage leading to ultimate laminate
failure. An experimental investigation was undertakenwhere the effects of the laminate stacking sequence, lam-
inate thickness, specimen size and material system were
considered.
The experimental apparatus used in the current study
consisted of a loading fixture and a restraining fixture.
The loading fixture comprised of a steel cylinder which
was fastened to the load cell of the tensile testing ma-
chine and to the test specimen via a protruding headedtitanium fastener (Hi-Torque / 6 mm). The fastener
was finger tightened to exclude any possible effects of
the clamping load on the failure load and mode. Stain-
less steel washers of thickness 1 mm and outer diameter
12 mm were used together with the fasteners. The test
specimens were simply supported at the outer boundary
by a steel restraining plate with a circular opening.
Restraining plates with circular openings of diameter40, 80 and 120 mm were manufactured to allow investi-
gation of specimen size effects. The restraining plate was
attached to the base of the fixture by four bolts located
at the corners of the plate. The experimental set-up is
illustrated in Fig. 1. The experiments were performed
using an Instron universal test machine (Instron 4505)
under displacement control at a rate of 1.0 mm/min.
The load and transverse displacement of the fastenerwere recorded continuously during testing using a PC
data acquisition system. A series of specimens were
instrumented with strain gauges in order to measure
the strain field in the laminate at different locations dur-
ing the test.
Table 1
Test specimen configurations
Notation Stacking sequence Resin Thickness (mm)
½0=45=90=�45�es [0/45/90/�45]s Epoxy 1.68
½0=45=90=�45�es2 [0/45/90/�45]s2 Epoxy 3.42
½0=90=45=�45�es [0/90/45/�45]s Epoxy 1.48
½0=90=45=�45�vs [0/90/45/�45]s Vinylester 2.42
½0=90�vs2 [0/90]s2 Vinylester 2.42
½0=90=45=�45�vs2 [0/90/45/�45]s2 Vinylester 4.68
G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314 303
The outer boundary diameter of the test specimens
was simply supported in the current study while the pre-
vious studies [2–4] of transverse load introduction con-
sidered specimens with a clamped outer boundary. Asimply supported boundary was chosen to be represent-
ative of a concentrated load introduction in a larger
structure. In contrast to previous studies [2,3], a range
of specimen sizes were investigated.
A fractographic investigation was undertaken in
order to determine micromechanical failure modes
occurring within the specimens. Several tests were
stopped at sub-critical load levels and the specimens un-loaded to allow for microscopic investigation. The spec-
imens were sectioned along planes parallel to the fibre
directions, polished and examined using an optical
microscope.
2.1. Materials and specimen preparation
Two different material systems were considered forthe investigation. The first system consisted of carbon
fibre non-crimp fabric and epoxy matrix (T700/Shell
Epicote LV828) with laminates being manufactured
using the resin transfer moulding (RTM) technique.
The second system was a carbon fibre non-crimp fabric
and vinylester matrix (T700/Jotun 9100) with the lami-
nates being manufactured using the vacuum infusion
technique. The laminate configurations used in theinvestigation are listed in Table 1.
Square specimens were machined from the laminates,
10 mm wider than the diameter of the restraining plate,
using a diamond tipped saw. A centrally located hole of
diameter 6 mm was machined in the specimens using a
dagger drill which was used in combination with a back-
ing plate to limit delamination at the exit side of the
laminate.
3. Results
3.1. Load–displacement behaviour
The load–displacement behaviour of the test speci-
mens subject to transverse load introduction was foundto be different, depending on the effective flexural stiff-
ness of the laminates. The effective flexural stiffness in-
creases with increase in the laminate thickness and
reducing boundary diameter. The load–displacement
relationship for specimens with high flexural stiffness
(low /) was slightly non-linear in the beginning followedby an almost linear load–displacement relation until fail-
ure. Specimens with lower flexural stiffness illustrated
non-linear deformation behaviour with a significant in-
crease in stiffness evident at larger displacements. Typi-
cal load–displacement graphs for the ½0=45=90=�45�esand ½0=90=45=�45�vs specimens are illustrated in Fig.
2(a) and (b) respectively.
The ½0=45=90=�45�es laminate exhibited a number ofminor load drops as the load increased. The minor load
drops were typically between 0.05 and 0.1 kN and the
resulting damage did not appear to have a significant ef-
fect on the global stiffness of the specimen. The minor
load drops occurred more frequently as the laminate ap-
proached final failure. The load levels at which the
minor load drops occurred were found to be independ-
ent of the specimen diameter. A similar independenceof the specimen size on the load when first cracking oc-
curred was noted by Chen et al. [8] in a study of biaxial
flexure of rectangular cross-ply laminates and by Cap-
rino et al. [6] who considered transverse loading of
CFRP plates via an indenter. The first audible fracture
within the specimens occurred at load levels significantly
lower than the first minor load drops visible on the
load–displacement curve. The first audible fractureoccurred at approximately 20–25% of the ultimate load.
Audible fractures were heard periodically as the load
was increased. At higher load levels, the audible frac-
tures coincided with minor load drops visible on the
load–displacement curve. At final failure, a number of
smaller load drops preceded a sudden loss of all load
bearing capacity. In the current study, failure of the
specimen was deemed to occur at the first major loaddrop illustrated by point (F) in Fig. 2(a) and (b). The
load drop is attributed to a major loss of stiffness or
pull-through failure of the laminate. The ultimate failure
mode of each of the ½0=45=90=�45�es laminates was fas-
tener pull-through failure where the fastener penetrated
through the laminate.
Similar load–displacement behaviour was noted for
the ½0=45=90=�45�es2 laminates. Slightly different behav-iour was noted for the specimens with /boundary = 40
mm which illustrated a marked reduction in stiffness at
0 2 4 6 8 10 12 140
1
2
3
4
5
6
7
8
Displacment (mm)
Load
(kN
)
φ=40 mm φ=80 mm φ=120 mm
F
First audible cracking
0 2 4 6 8 10 12 140
1
2
3
4
5
6
7
8
Displacment (mm)
Load
(kN
)
φ=40 mm φ=80 mm φ=120 mm
First audible cracking
Minor load drop
F
(b)(a)
Fig. 2. Typical load–displacement curve for the quasi-isotropic 8-ply laminates: (a) ½0=45=90=�45�es laminates, (b) ½0=90=45=�45�vs laminates.
304 G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314
approximately 45% of the failure load. However, the
specimens continued to carry load beyond this point
but exhibited a reduced stiffness. The reduction in
stiffness was attributed to internal damage in thelaminate.
The ½0=90=45=�45�es laminate illustrated similar char-
acteristics to the ½0=45=90=�45�es laminate. However, the
load drops were slightly larger (0.1–0.2 kN) and were
more visible on the load–displacement curve. The first
audible fractures occurred at 25% of the failure load
and the ultimate failure mode of the ½0=90=45=�45�eslaminates was fastener pull-through failure.
The ½0=90=45=�45�vs laminates exhibited similar
load–displacement behaviour to the ½0=45=90=�45�eslaminates. Audible fracture could be heard from the
specimens at load levels around 25% of the failure load
although no detectable loss of stiffness was observed as
illustrated in the load–displacement curves (see Fig.
2(b)). The ½0=90=45=�45�vs specimens continued to frac-
ture as the load was increased and exhibited a series ofminor load drops prior to the first major load drop.
The slope of the curve following the load drops re-
mained unaltered indicating that no appreciable loss of
global stiffness occurred. Final failure of the laminate
was characterised by a sudden major drop in load. Fail-
ure occurred through global collapse of the laminate
with interlaminar fractures extending to the boundary
Epoxy matrix laminates
Fig. 3. Schematic of obse
of the test specimens. Compressive failure of the lami-
nate at a plane orthogonal to the fibre orientation of
the surface plies. The failure patterns of the epoxy and
vinylester matrix laminates are illustrated schematicallyin Fig. 3.
Visual inspection at failure of the vinylester speci-
mens showed that the inter-ply damage extended to
the edge of the specimen. The laminates were shown
to be capable of carrying additional load past this failure
point, however, the post-failure behaviour was consid-
ered to be beyond the scope of this study. The magni-
tude of the load drop at failure was found to bedependent upon the specimen size with specimens of lar-
ger diameter experiencing a larger drop in load. The lar-
ger drop occurred as a result of the greater about of
elastic energy accumulated in the larger diameter speci-
mens before failure.
The behaviour of the ½0=90=45=�45�vs2 laminates were
found to be similar to the behaviour of the
½0=45=90=�45�es2 laminates with a distinct reduction instiffness occurring at approximately 45% of the failure
load for the /boundary = 40 mm specimen. The marked
reduction in stiffness was not evident in the larger diam-
eter specimens. A more sparse distribution of load drops
were found on the load–displacement curves of the
½0=90=45=�45�vs2 laminates in comparison to the
½0=45=90=�45�es2 laminates.
Vinylester matrix laminates
rved failure modes.
G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314 305
Comparison of the ½0=90=45=�45�vs and ½0=90�vs2 spec-imens illustrated no significant difference in the load–
displacement behaviour. The behaviour was found to
be similar with the exception that the damage propa-
gated along different paths as a result of the different
stacking sequence.
3.2. Strain distribution
A series of specimens were instrumented with strain
gauges as shown in Fig. 4. Uniaxial strain gauges of
gauge length 2 mm were placed in orthogonal directions
on both sides of the laminate in order to investigate
bending behaviour and the radial strain field aroundthe hole. Strain measurements from the ½0=45=90=�45�es(/boundary = 80 mm) case are illustrated in Fig. 5(a)–
(d). The deformation behaviour of the ½0=45=90=�45�eslaminate was shown to alter as the load increased and
damage developed. The radial strain on the top surface
of the ½0=45=90=�45�es laminate is illustrated in Fig. 5(a)
and (b). The radial strain was initially tensile as a result
of the laminate bending. However as the load was in-creased, local deformation occurred around the load
introduction and the radial strains became non-symmet-
ric. At 40% of the failure load, the radial strain close to
the hole in the 90� direction became compressive. This
corresponded to the plane of the laminate with the lowest
bending stiffness. The same behaviour did not occur in
the 0� direction until 60% of the failure load. The change
in sign of the radial strain indicated a shift from bendingto membrane deformation behaviour in the specimen.
5 mm
7 mm
8 mm
T0_R1
T0_R2
T0_R3
T90
_R1
T90
_R2
T90
_R3
T
Orientation of surface ply
Loading direction
B
Fig. 4. Strain gauge locations on instrumented specimens.
The non-linear deformation was apparent when consid-
ering the underside of the laminate. The initially com-
pressive radial strains changed sign at load levels of
25% and 40% of the failure load in the 90� and 0� direc-tions respectively. At 60% of the failure load, the tensile
radial strain close to the hole was significantly larger inthe 90� direction (�rr = 0.45%) in comparison to the 0�direction (�rr = 0.25%). The difference between the radial
strains in the 90� and 0� directions continued to increase
as the load was increased. The unequal strain distribu-
tion around the hole indicated that damage initiation
may occur at a preferred location, or weakest plane.
3.3. Effect of the specimen size
The effect of the specimen size on the specimen failure
load is illustrated in Fig. 6. The failure load of the
½0=45=90=�45�es laminates was found to be independent
of the specimen diameter. The ultimate failure mode
was similar for each of the specimens and the result
was deemed reasonable given the local nature of the fail-
ure mode. The experiments produced repeatable resultswith the variation in strength being relatively constant
for each of the specimen types. The ½0=45=90=�45�es2laminates illustrated a very similar behaviour with regard
to specimen size. However, the laminates with a diameter
of 40 mm were found to withstand slightly higher loads
with the load decreasing by 15% as the boundary diam-
eter was increased to 80 mm. The failure load remained
unchanged for the specimens with a boundary diameterof 120 mm. The failure loads of the ½0=90=45=�45�es lam-
inates were also found to be independent of specimen
size. The failure loads showed a similar trend to the
½0=45=90=�45�es laminates, with the load levels being
lower as a result of the reduced laminate thickness.
Both the ½0=90=45=�45�vs and the ½0=90�vs2 laminates
were found to behave in a similar manner with regard
to the specimen size. This supported the theory thatthe same failure mode was responsible for damage prop-
agation and ultimate failure of the laminates. The failure
load was found to increase by approximately 30% as the
diameter increased from 40 to 80 mm with no additional
increase in failure load being noted for the /boundary =
120 mm laminates. The failure load of the ½0=90=45=�45�vs2 laminates showed a similar trend to the
½0=90=45=�45�vs laminates although the failure loadwas shown to increase with specimen size for the range
of diameters selected in the current study.
The differing dependence of the failure load on the
specimen size was assumed to be a result of the different
failure modes occurring in the laminates.
3.4. Fractographic investigation of failure modes
The load–displacement behaviour of the laminates
did not provide any clear indication of damage initiation
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
6Lo
ad (
kN)
Strain (%)
T-0-R1T-0-R2T-0-R3
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
6
(Loa
d kN
)
Strain (%)
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
6
Load
(kN
)
Strain (%)-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
0
1
2
3
4
5
6
Load
(kN
)
Strain (%)
(a) (b)
(c) (d)
B-0-R1B-0-R2B-0-R3
T-90-R1T-90-R2T-90-R3
B-90-R1B-90-R2B-90-R3
Fig. 5. Radial strain distribution in [0/45/90/�45]s epoxy laminate: (a) top surface, 0� orientation; (b) top surface, 90� orientation; (c) bottom surface,
0� orientation; (d) bottom surface, 90� orientation.
20 40 60 80 100 120 1400
5
10
15
20
Fai
lure
Loa
d (k
N)
φ
[0/90/45/–45]
[0/45/90/–45]
[0/90/45/–45]
[0/90/45/–45] s2v
[0/90] s2v
[0/45/90/–45] se
se
s2e
sv
boundary (mm)
Fig. 6. Effect of specimen size on the ultimate failure load.
306 G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314
and the development of damage within the laminates
prior to failure. It was also evident from visual inspec-
tion of the specimens that two different final failure
modes occurred, namely fastener pull-through failure
and global laminate collapse. It was considered neces-
sary to identify the failure mechanism responsible for
damage initiation, the location and the manner in whichthe damage propagated.
The planes which were examined are shown in Fig. 7.
The observations made for each laminate type are dis-
cussed in the following sections. The individual plies in
the laminates are numbered from top to bottom with
ply 1 being the outermost ply on the top side of the lam-
inate. The term off-axis plies refers to plies where the
fibre orientation does not coincide with the directionof the sectioned plane. This convention is adopted
throughout the following sections.
A B C
D
O
+45o
-45o
0o
90o
Fibre directions
Fig. 7. Sectioned planes for the fractographic investigation.
G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314 307
3.4.1. [0=45=90=�45]esn laminates
The first audible fractures occurred in the ½0=45=90=�45�es specimens at load levels approximately 20–
25% of the failure load. There was no indication of dam-
age on the load–displacement curve at this load level
and inspection of optical micrographs taken of speci-
mens exposed to this load level revealed only limited
damage. A minor intralaminar shear failure in the sub-
surface ply (ply 7) beneath the edge of the washer onplane O–D was noted together with limited tensile frac-
tures in the resin rich layer on the top surface of the lam-
inate. After 50% of the failure load, more significant
structural damage was apparent in the form of trans-
verse tensile failures in the top surface of the laminate.
On plane O–D, the transverse tensile cracks propagated
through the upper 0� (ply 1) and 45� (ply 2) plies and
were arrested at the interface of the fibre bundles. Typ-ical transverse tensile failures are shown in Fig. 8. To-
gether with the intralaminar matrix shear failure, this
failure mode was assumed to be responsible for damage
initiation in the ½0=45=90=�45�es laminates. No indica-
tion of the transverse tensile failures occurring at 50%
of the failure load were visible on the load–displacement
curves. The fracture location correlated with the radial
Fig. 8. Micrograph of ½0=45=90=�45�es epoxy laminate, /boundary = 40
mm loaded to 50% of the failure load (plane O–D).
strain in the 90� direction which was tensile at this load
level. There was no apparent damage at this load level in
the other planes examined.
At 90% of the failure load, a number of different
micromechanical failure modes were found to be present
in all of the examined planes. Fibre kinking and matrixintralaminar shear cracking was found directly under
the edge of the washer. The most significant intralami-
nar shear cracking was found along planes O–A and
O–C. Along plane O–A, inclined shear cracking had
propagated through plies 3–6 which were orientated
off-axis to the sectioned plane. Matrix shear cracking
was also evident in the upper plies along the plane O–
C. Inclined intralaminar shear cracks and in-plane inter-laminar shear cracks were found to exist between plies 1
and 3 with the damage extending approximately 10 mm
from the edge of the hole. Limited damage was noted in
plane O–B in the specimens inspected. The dominant
failure mode in the O–D plane was transverse tensile
failure of the matrix in the top ply (ply 1). The trans-
verse tensile cracks were most apparent in the region
adjacent to the edge of the washer.Final failure of the ½0=45=90=�45�es specimens in-
volved the fastener penetrating through the laminate.
The through-thickness shear cracks which propagated
conically upward from the edge of the washer resulted
in local collapse with the material unable to withstand
the transverse load. No significant in-plane damage
was noted in the failed specimens except in the upper
plies where splitting occurred causing a �volcano� likefailure topology. The behaviour was found to be consist-
ent for each of the specimen diameters.
The first audible fractures in the ½0=45=90=�45�es2laminates occurred at 25–30% of the failure load. The
initial damage occurred in the form of matrix intralami-
nar shear cracking in sub-surface plies at the edge of the
washer. Damage was most evident along plane O–D. At
50% of the failure load, internal damage within thelaminate was more pronounced. The specimen with a
boundary diameter of 40 mm illustrated a distinct reduc-
tion in stiffness at this load level. Inspection of the spec-
imen revealed damage in all of the examined planes.
Single shear cracks through off-axis plies were evident
in planes O–B and O–D. The most significant damage
was however noted in planes O–A and O–C. A network
of inclined intralaminar shear cracks was visible to-gether with interlaminar shear failures which occurred
at different interfaces through the thickness. A micro-
graph of plane O–A is shown in Fig. 9. The matrix
intralaminar shear cracking which initiated in sub-sur-
face plies beneath the edge of the washer, was shown
to have propagated at an angle of 45� to the applied
load. The cracks propagated through fibre bundles be-
fore reaching the interface between the fibre bundleand the matrix. Upon reaching the fibre bundle inter-
faces, the cracks changed direction and continued to
Fig. 9. Micrograph of ½0=45=90=�45�es2 laminate, /boundary = 40 mm
loaded to 50% of the failure load (plane O–A).
308 G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314
propagate along the interfaces before being arrested.
Similar fracture patterns were also found above themid-plane of the laminate where inclined shear cracks
propagated through off-axis plies and formed interlami-
nar cracks at fibre bundle interfaces. Very limited crack-
ing was found in the resin rich areas which were present
between fibre bundles, with the cracks tending to prop-
agate through the off-axis fibre bundles. The extent of
the damage in the larger diameter specimens was not
as pronounced in comparison to the 40 mm diameterspecimens. Intralaminar shear failure was shown to be
present in most planes with the most significant damage
occurring in the O–D plane. Final failure of the
½0=45=90=�45�es2 laminates was found to occur in a sim-
ilar manner as in the ½0=45=90=�45�es laminates with the
fastener penetrating through the laminate.
Fig. 10. Intralaminar and interlaminar shear cracking in the
½0=90=45=�45�vs laminate, /boundary = 40 mm at 50% of failure load
(plane O–A).
3.4.2. [0=90=45=�45]es laminates
The first audible fractures in the ½0=90=45=�45�es lam-
inates were recorded at approximately 25% of the failure
load. There was no indication of damage on the load–
displacement curve and inspection of the laminates
loaded to this load level revealed only minor transverse
tensile cracking on the upper surface of the laminate (ply
1). Transverse tensile failure was regarded as the first ply
failure mode for these laminates.At 50% of the failure load, more significant damage
was observed in the laminates. The most notable dam-
age was on plane O–B where intralaminar shear fracture
was found in ply 6 close to the edge of the washer. Sim-
ilar intralaminar shear cracks were observed on plane
O–D where cracks had propagated through ply 6 and
through plies 4 and 5.
At 90% of the failure load, the most significant dam-age was found on planes O–B and O–D. The damage
was consistent with the damage pattern observed at
50% of the failure load. On plane O–B, the most signif-
icant damage was an interlaminar shear crack between
plies 5 and 6. Intralaminar cracks branched from the
main crack through ply 5 in the region close to the edge
of the washer. Transverse tensile fractures were also ob-
served in ply 7 which was orientated at 90� to the sec-
tioned plane. Damage was also observed on plane O–
C where transverse tensile fractures were observed on
the upper surface (ply 1). There were also interlaminar
shear cracks which had propagated between plies 1
and 2 and between plies 2 and 3.Final failure of the ½0=90=45=�45�es laminates oc-
curred with the fastener penetrating through the lami-
nate. The results were found to be consistent for all
specimen diameters.
3.4.3. [0=90=45=�45]vsn laminates
Damage initiation in the ½0=90=45=�45�vs laminates
occurred at approximately 20% of the failure load.The initial failure was found to be similar as in the
½0=45=90=�45�esn laminates with inclined intralaminar
shear cracking occurring in sub-surface plies beneath
the edge of the washer. At 50% of the failure load, dam-
age was found to be significant in all of the examined
planes. In plane O–A, interlaminar shear failures were
found to extend 10 mm from the edge of the hole. Inter-
laminar cracks were found to be most prominent be-tween the +45� and �45� plies (3 and 4 and 5 and 6).
The interlaminar shear cracks were connected by in-
clined intralaminar shear cracks through off-axis plies
(4 and 5) in the centre of the laminate as shown in
Fig. 10. Similar fractures were noted on each of the ob-
served planes. Additional damage was noted on plane
O–D with interlaminar failure occurring between plies
1 and 2. The interlaminar cracks appeared to have devel-oped from transverse tensile matrix cracks which oc-
curred in the top ply (ply 1) in this plane.
At 90% of the failure load, interlaminar cracks were
found to extend to the free edge of the specimen along
several planes. The failure modes were found to be con-
sistent with those observed at 50% of the failure load. An
additional failure mode, namely shear failure in the
upper plies (1–3) was observed. The failures were foundto occur toward the outer boundary of the specimen.
Fig. 11 illustrates the failures which were present in off-
axis plies on plane O–C. An interlaminar shear crack be-
tween the plies parallel (ply 4) and perpendicular (ply 3)
to plane O–C was also evident. The damage pattern was
found to be similar for each of the specimen diameters.
Fig. 11. Shear failure in the upper plies of the ½0=90=45=�45�vslaminate /boundary = 40 mm at 90% of the failure load (plane O–C).
G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314 309
At final failure, the interlaminar shear cracks were
found to have propagated to the boundary of the spec-
imen on all of the examined planes. Visual inspection re-
vealed compressive collapse of the laminate along the
plane O–D at one side of the hole. The collapse of the
laminate was most likely a secondary failure modewhich occurred as a result of instability due to the loss
of bending stiffness associated with the extensive interla-
minar shear failures. The failure modes were consistent
for each of the specimen diameters.
The ½0=90=45=�45�vs2 laminates illustrated a similar
development of damage to the ½0=90=45=�45�vs lami-
nates. The initial failure mode was inclined intralaminar
shear failure which occurred in the sub-surface plies atthe edge of the washer. As the load was increased, the
damage developed through intralaminar shear cracks
in off-axis plies as shown in Fig. 12. A staircase of shear
cracks extending from the edge of the washer through
the thickness of the laminate was evident in all of the in-
spected planes. The staircase form of the intralaminar
and interlaminar shear fractures was also reported by
Banbury and Kelly [3]. The most prominent failuremode was interlaminar shear failure in the plane of the
laminate where cracks propagated along the interface
between the +45� and �45� plies. Interlaminar failure
along the interface between ply 2 (90�) and ply 3
Fig. 12. Intralaminar shear cracking in the ½0=90=45=�45�vs2 laminate
at final failure /boundary = 40 mm (plane O–D).
(+45�) was also evident, especially on plane O–D. In
the larger diameter specimens, transverse tensile failures
occurred close to the hole in the 0� plies in the centre of
the laminate on plane O–D. More pronounced indenta-
tion and fibre kinking at the edge of the washer was evi-
dent in the ½0=90=45=�45�vs2 laminates in comparison tothe ½0=90=45=�45�vs laminates. Final failure of the
½0=90=45=�45�vs2 laminates was found to occur in a sim-
ilar manner to the ½0=90=45=�45�vs laminates with com-
pressive collapse occurring on plane O–D.
3.4.4. [0=90]vs2 laminates
The initiation of damage in the ½0=90�vs2 laminates was
consistent with the ½0=90=45=�45�vs laminates with ini-tial audible fractures occurring at 20–25% of the failure
load. The initial failure mode was also found to be sim-
ilar with matrix intralaminar shear failure occurring in
sub-surface plies. At 50% of the failure load, damage
was evident on all of the examined planes. On plane
O–B, which is parallel to the fibre direction of ply 1,
interlaminar shear cracks were found to have propa-
gated between plies 1 and 2 and between plies 6 and 7.The interlaminar cracks extended approximately 10
mm from the edge of the hole. Limited damage was
found on plane O–C with only minor intralaminar shear
cracks found in ply 6 close to the edge of the washer.
There are no fibres aligned parallel to plane O–C which
most probably influenced the damage pattern. On plane
O–D, very limited damage was evident with only minor
fractures in the bottom ply. The damage on plane O–Ashowed similar characteristics to that found on plane O–
C but in addition, interlaminar shear failure was evident
between plies 1 and 2 in the region above the edge of the
washer.
At increased load levels, the failure modes remained
unchanged but the extent of the in-plane and through-
thickness damage increased significantly. The ½0=90�vs2laminates exhibited more pronounced inclined intrala-minar shear failures in comparison to the ½0=90=45=�45�vs laminates. At final failure, the inclined intralami-
nar cracks extended through the thickness of the lami-
nate which resulted in numerous compressive shear
buckling failures along the planes parallel to the fibre
directions. Permanent out-of-plane deformation of the
specimens was found to occur as a result of the
through-thickness shear failures.
3.5. Effect of the laminate stacking sequence and resin
system
The laminate stacking sequence was shown to affect
the failure modes and the propagation of damage within
the laminates. Microscopic inspection of failed lami-
nates where ply angles between adjacent plies was 90�revealed widespread delamination. The delamination
was not as pronounced in laminates with 45� between
Table 3
Strength properties of the carbon fibre/epoxy lamina
Xt (MPa) Xc (MPa) Yt (MPa) Yc (MPa) S
1250 730 60 200 70
310 G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314
adjacent plies. The dominant failure mode in these lam-
inates was inclined interlaminar shear cracking. Similar
damage patterns were reported by Banbury and Kelly
[3] who investigated laminates with [0/90/45/�45]ns and
[0/90]8s stacking sequences. Again it can be noted that
the difference in ply orientation between adjacent plieswas 90� at several locations within the laminate.
The difference in ply orientation within the laminates
causes high interlaminar stresses which in turn promote
delamination. Tao and Sun [9] investigated the influence
of ply orientation on delamination in composites plates
made from carbon fibre epoxy pre-preg (AS4/3501-6)
subject to mode II fracture loading. The critical strain
energy release rates of 0�/h� interfaces were determinedexperimentally and the interlaminar fracture toughness
was found to decrease as the off-axis angle (h) increased.After initial failure, transverse loading imposes a mixed
mode fracture load on the laminates with the dominant
mode being mode II. Thus, the results of Tao and Sun,
while based on different materials, can be used to under-
stand the development of damage in the laminates after
initial failure. Similar correlation between stacking se-quence and delamination was reported by Liu [10]
who investigated impact induced delamination. The
delamination area in [0�/h�] laminates subject to trans-
verse impact was shown to increase with increasing h.A theoretical impact damage model by Clark [11] sup-
ported the hypothesis of delamination occurring prefer-
entially at certain interfaces.
The resin system appeared to have the most signifi-cant effect with regard to the final failure mode. Ulti-
mate failure of the epoxy matrix laminates resulted in
the fastener penetrating through the laminate. The
epoxy matrix has a lower failure strain (�f = 1.5%) in
comparison to the vinylester matrix (�f = 11%) and this
was reflected by the internal damage within the lami-
nates where a large network of inter- and intralaminar
cracks was present in the epoxy laminates. In contrast,the ultimate failure of the vinylester matrix laminates
was global collapse where in-plane damage propagated
to the specimen boundary. The damage in the vinylester
laminates was characterised by delamination between
0�/90� interfaces which propagated in-plane. The com-
plex network of smaller fractures under the fastener
head was not evident in the vinylester laminates. After
initial failure of the laminates, the strength appears tobe governed by the interlaminar strength properties.
The results from this study show that the stacking se-
quence influences the mode of failure within the lami-
nate and in particular the extent of in-plane damage.
Table 2
Elastic properties of the carbon fibre/epoxy lamina
E11 (GPa) E22 (GPa) E33 (GPa) G12 (GPa) G13 (GPa
98 7.8 7.8 4.7 4.7
The ultimate failure mode appeared to be governed by
the resin system with the failure mode being shown to
be independent of laminate stacking sequence for
the configurations tested. No conclusive relation be-
tween the stacking sequence/resin and the ultimate fail-
ure load could be deduced from the performedexperiments.
4. Finite element modelling
Finite element analysis was used to investigate the
stress distribution within ½0=45=90=�45�es laminates
when subject to a concentrated transverse load. Thestress distribution around a hole in the laminate is mul-
ti-axial and thus a three-dimensional finite element
model was deemed necessary. Finite element models of
the laminates were developed using the ABAQUS [12] fi-
nite element software. Each ply of the laminate was
modelled using one layer of quadratic solid brick ele-
ments (C8D20). Each ply of the finite element model
was modelled as an orthotropic solid. The in-planeproperties of the lamina were experimentally determined
[13] and the through-thickness properties estimated val-
ues. The material properties which have been used in the
model are presented in Tables 2 and 3.
The load was applied to the model in the form of a
prescribed displacement on the lower surface of the lam-
inate representing the washer. The washer has been as-
sumed to be rigid due to the difference in stiffnessbetween the washer/bolt head and the thin laminate
plate. The circular boundary of the plate was simply
supported i.e. only being prevented from moving in
the transverse direction. A half section of the finite ele-
ment model is illustrated in Fig. 13. It was evident from
the experimental results that the laminates underwent
large deformation and thus the calculations were per-
formed using large deformation analysis.The results from the finite element model were com-
pared with the strain measurements recorded during
the tests. A comparison of the surface strains is illus-
trated in Fig. 14. The strains from the finite element
analysis are shown to compare reasonably well with
the measured strains.
) G23 (GPa) m12 (GPa) m13 (GPa) m23 (GPa)
3.2 0.34 0.34 0.44
Fig. 13. Finite element model of the laminate subject to transverse
load introduction.
0 5 10 15 20 25 30 35 40–0.8
–0.6
–0.4
–0.2
0
0.2
0.4
0.6
0.8
Str
ain
(%)
Radial position (mm)
FE Top 90FE Top 0FE Bot 90FE Bot 0Exp Top 90Exp Top 0Exp Bot 90Exp Bot 0
Fig. 14. Comparison of predicted and experimental strains (P = 0.7
kN).
0 2 4 6 8 10 120
1
2
3
4
5
6
7
8
9
10
Displacement (mm)
Load
(kN
)
Experiment: φ=40mmExperiment: φ=80mmExperiment: φ=120mm Predicted: φ=40mmPredicted: φ=80mmPredicted: φ=120mm
Fig. 15. Comparison of predicted and experimental load–displacement
behaviour.
G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314 311
Through knowledge of the stress field within the lam-
inate, the load at which first-ply failure occurred could
be predicted together with the location of the failure
within the laminate. The strength based failure crite-
rion proposed by Hashin [14] has been used to predict
first-ply failure within the laminate. The criterion distin-
guishes between fibre and matrix failure modes in ten-
sion and compression and requires measurableproperties of unidirectional laminae as input. The
three-dimensional form of Hashin�s criteria is given in
Eqs. (1)–(4),
Fibre tensile fracture, r11P 0:
r11
X t
� �2
þ r212 þ r2
13
S2
� �P 1 ð1Þ
Fibre compressive fracture, r11< 0:
� r11
X c
� �P 1 ð2Þ
Matrix tensile or shear cracking failure, r22 + r33 P0:
ðr22 þ r33Þ2
Y 2þ r2
12 þ r213 þ r2
23 � r22r33
S2P 1 ð3Þ
t
Matrix compressive or shear cracking failure, r22 +
r33 < 0:
1
Y c
Y c
2S
� �2
� 1
" #ðr22 þ r33Þ þ
ðr22 þ r33Þ2
4S2
þ r212 þ r2
13 þ r223 � r22r33
S2P 1 ð4Þ
where Xt is the longitudinal tensile strength, Xc is the
longitudinal compressive strength, Yt is the transverse
tensile strength, Yc is the transverse compressive
strength and S is the shear strength. The rij terms arethe components of the stress tensor. The experimentally
determined strengths of a unidirectional ply of the car-
bon fibre epoxy laminate which have been used in the
calculation of first-ply failure are given in Table 3. The
failure criterion was implemented in ABAQUS in
the form of a user-subroutine (USDFLD) where each fail-
ure equation is defined as a field variable. The load is ap-
plied incrementally with the field variables (failurecriteria) being evaluated at each increment.
4.1. Results from the finite element analysis
The results from the finite element analysis are illus-
trated in Figs. 15 and 16. The load versus displacement
behaviour of the laminates with different boundary
diameters are shown in Fig. 15. The predicted load–dis-placement behaviour is shown to be in good agreement
with the experimentally determined behaviour. The
inclusion of damage propagation in the finite element
model could possibly improve the predictions further.
The results from the simulations were analysed
through plotting the values of the maximum Hashin
failure index at each tangential location (h) within the
0 20 40 60 80 100 120 140 160 1800
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Has
hin
failu
re In
dex
θ (degrees)
Ply 1Ply 2Ply 3Ply 4Ply 5Ply 6 Ply 7Ply 8MAX
0 20 40 60 80 100 120 140 160 1800
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Has
hin
failu
re In
dex
θ (degrees)
Ply 1Ply 2Ply 3Ply 4Ply 5Ply 6 Ply 7Ply 8MAX
0 20 40 60 80 100 120 140 160 1800
0.2
0.4
0.6
0.8
1
1.2
Has
hin
failu
re In
dex
θ (degrees)
Ply 1Ply 2Ply 3Ply 4Ply 5Ply 6 Ply 7Ply 8MAX
0 20 40 60 80 100 120 140 160 1800
0.2
0.4
0.6
0.8
1
1.2
1.4
Has
hin
failu
re In
dex
θ (degrees)
L8FC
L6FC
L1MT
L7FC
L7MC
L8FCL1
MT
L7MC
L6MC
L8FC
L8MC
L6FC
MC - Matrix compressionMT - Matrix tensionFC - Fibre compressionFT - Fibre tensionL - Ply number n
P=0.5kN
P=1.5kN
P=1.0kN
(a) (b)
(c) (d)
Fig. 16. Failure mode map for the ½0=45=90=�45�es , / = 80 mm specimen at a radius R = 6.5 mm from the hole centre: (a) ply failure indices (P = 0.5
kN), (b) ply failure indices (P = 1 kN), (c) ply failure indices (P = 1.5 kN), (d) failure mode map.
312 G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314
laminate. An example of the analysis is illustrated in Fig.
16(a)–(d) for the case of the / = 80 mm laminate.
The four Hashin failure criteria (Eqs. (1)–(4)) were
applied to the individual plies of the laminate at each
load increment and the maximum value of the failure
index plotted as shown in Fig. 16(a)–(c). A failure mode
map was generated by plotting the maximum failure
index at each tangential location for a given load asshown in Fig. 16(d). At low loads (P = 0.5 kN), the most
highly stressed plies are the top ply (ply 1) where the ma-
trix tensile failure mode is dominant, and the sub-sur-
face plies beneath the washer (plies 6 and 7) where
matrix compression or shear failure is dominant. As
the load increased to 1 kN, the dominant failure modes
remained unchanged although the matrix tensile mode
was active over a larger area. At the predicted failureload (P = 1.5 kN), a more complex mode map exists
with the dominant failure modes being matrix tensile
failure at the top surface (ply 1), fibre compression fail-
ure in the bottom two plies (7 and 8) and matrix com-
pression failure in the sub-surface plies beneath the
washer (plies 6 and 7). The failure modes illustrated in
the failure mode map correlate well with those observed
in the fractographic analysis of the laminates. The failure
mode map also provides an indication of how efficient
the laminate is designed for carrying a specific load.
The predicted first-ply failure loads for the laminates
are presented in Table 4. The predictions are in generalagreement with the experimental values, however the
predicted load levels for the laminates with /boundary =
40 and 80 mm are slightly higher than those
observedexperimentally. A possible reason for the over
predicted failure loads could be that the lamina failure
strengths are ultimate values and thus initial damage
may occur at lower load levels. The initial failure mode
for the larger diameter specimens (/boundary = 80, 120mm) is transverse tensile failure which precedes intrala-
minar shear failure in the sub-surface plies. This change
in initial failure mode was also evident from the fracto-
graphic analysis.
Table 4
Predicted first-ply failure loads of ½0=45=90=�45�es laminates
/boundary (mm) Predicted first-ply failure load (kN) Failure mode Experimental first-ply failure load (kN)
40 1.69 Matrix shear (ply 7, h = 135�) 1.25–1.58
80 1.50 Matrix tensile (ply 1, h = 90�) 1.2–1.4
120 1.02 Matrix tensile (ply 1, h = 90�) 1.1–1.6
h = 0 is the orientation of the surface plies.
G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314 313
5. Conclusions
The behaviour of composite laminates subject to
transverse load introduction was investigated experi-mentally and numerically. The load–displacement
behaviour was found to depend on the effective flexural
rigidity of the specimens with the response becoming
non-linear at large deformations. The load at which
the first audible cracking occurred was found to be
20–30% of the failure load. The damage which resulted
from the initial cracking was not visible on the load–dis-
placement curve and thus did not result in any noticea-ble loss of stiffness. At higher load levels, the load–
displacement curve illustrated minor load drops which
were the result of damage accumulation.
The initial failure mode of the laminates was identi-
fied through fractographic analysis of specimens loaded
to sub-critical load levels. The initial failure mode was
predominantly intralaminar matrix shear failure which
occurred in sub-surface plies beneath the edge of thewasher. Transverse tensile failures were found in the
outermost plies of the laminates with low flexural rigid-
ity but these failures were not pronounced and not
deemed to cause the loss of structural integrity. The ini-
tial damage propagated both through the thickness and
in the plane of the laminate in the form of intralaminar
and interlaminar shear failures. In-plane damage was
found to be more widespread in laminates where adja-cent plies were oriented at 90� to one another. Further
research is required in order to quantify the effects of
the stacking sequence.
Two different ultimate failure modes were identified,
namely fastener pull-through failure and global collapse
of the laminate. Fastener pull-through failure occurred
in the epoxy matrix laminates, where inclined intralami-
nar shear cracks propagated through the thickness ofthe laminate. Global collapse was found to occur in the
vinylester matrix laminates where ultimate failure oc-
curred after interlaminar cracks had propagated to the
boundary of the specimen. The different failure modes
was assumed to be related to a critical level of matrix
strain. The vinylester matrix had a significantly higher
failure strain than the epoxy matrix used in this study.
The effect of the specimen size on the failure load wasinvestigated for each test configuration. Laminates
which failed by fastener pull-through failure showed lit-
tle dependence on the specimen size, however laminates
which failed by global collapse showed a stronger
dependence with the load capacity increasing with the
increasing specimen diameter.
A three-dimensional finite element model was devel-
oped to study the behaviour of the ½0=45=90=�45�es lam-inates subject to transverse concentrated loading. The
predicted non-linear load–displacement behaviour of
the laminates was shown to be in good agreement with
the experimental results. The prediction of the failure ini-
tiation load and location was also found to show reason-
able agreement with experimentally determined values.
Based on the results obtained in this investigation, it
is concluded that the ultimate failure of transverse loadintroductions occurs by a process of damage accumula-
tion. The interaction of intralaminar and interlaminar
cracks plays an important role in the damage develop-
ment within the laminates. Interlaminar failure has been
shown to be a prominent failure mode in the laminates
and its effect should be included in the analysis of trans-
verse load introductions after first-ply failure. It is re-
commended to avoid transverse concentrated loadingwhere possible in composite laminates given that the
first-ply failure occurs at 20–25% of the ultimate failure
load. The fact that the laminates accumulate damage at
such low load levels will have an important bearing on
the fatigue strength of such load introductions.
The present investigation has highlighted that further
research is required in order to gain a more thorough
understanding of the behaviour of composite laminatessubject to such loading and to develop suitable guide-
lines for design.
Acknowledgements
This work has been financially supported by the
Commission of the European Union through GrowthProject TECABS (Technologies for Carbon Fibre Rein-
forced Modular Automotive Body Structures) and by
the Swedish Foundation for Strategic Research through
the national Swedish research program �Integral VehicleStructures�.
References
[1] Chen WH, Lee S-S. Numerical and experimental failure analysis
of composite laminates with bolted joints under bending loads. J
Compos Mater 1995;29(1):15–36.
314 G. Kelly, S. Hallstr€om / Composite Structures 69 (2005) 301–314
[2] Waters WA, Williams JG. Failure mechanisms of laminates
transversely loaded by bolt push through. NASA Technical
Memorandum 87603. 1985.
[3] Banbury A, Kelly DW. A study of fastener pull-through failure
of composite laminates. Part 1: experimental. Compos Struct
1999;45(4):241–54.
[4] Banbury A, Kelly DW, Jain LK. A study of fastener pull-through
failure of composite laminates. Part 2: failure prediction. Compos
Struct 1999;45(4):255–70.
[5] Caprino G, Langella A, Lopresto V. Elastic behaviour of circular
composite plates transversely loaded at the centre. Composites
Part A 2002;33:1191–7.
[6] Caprino G, Langella A, Lopresto V. Prediction of first failure
energy of circular carbon fibre reinforced plastic plates loaded at
the centre. Composites Part A 2003;34:349–57.
[7] Ye L. The role of matrix resin in delamination onset and growth
in composites. Compos Sci Technol 1988;33:257–77.
[8] Chen AS, Matthews FL, Sims GD. The effect of support
conditions on the performance of CFRP plates subject to biaxial
flexure. Compos Sci Technol 1998;58:613–21.
[9] Tao J, Sun CT. Influence of ply orientation on the delamination in
composite plates. J Compos Mater 1998;32(2):1933–47.
[10] Liu D. Impact-induced delamination––a view of bending stiffness
mismatching. J Compos Mater 1998;22:674–91.
[11] Clark G. Modelling of impact damage in composite laminates.
Composites 1999;20(3):209–14.
[12] Hibbet, Karlsson, Sorensen Inc. Abaqus/Standard User�s Manu-
als, V6.2. Pawtucket, RI, USA, 2001.
[13] Truong CT, Lomov SV, Verpoest I. The mechanical properties of
multi-axial multi-ply carbon fabric reinforced epoxy composites.
In: Proceedings of the 10th European conference on composite
materials (ECCM-10), Brugge, Belgium, 2002.
[14] Hashin Z. Failure criteria for unidirectional fibre composites. J
Appl Mech 1980;47:329–44.