Interlaminar fracture behaviour of laminated composites

R. Rikards
Institute of Computer Analysis of Structure, Faculty of Civil Engineering, Riga Technical University, LV-1658 Riga, Latvia
Abstract
The interlaminar fracture behavior of composite laminate under full range of in-plane loading conditions has been
investigated. Loading conditions from pure mode I through various mixed mode I/II ratios up to pure mode II have
been generated by the aid of the proposed compound version of the compact tension shear specimen. From the
experimentally measured critical loads the mode I, mode II and the various mixed mode I/II critical energy release
rates at crack initiation have been determined by the aid of the nite element method and the modied virtual crack
closure integral method. Based on these results the parameters for a fracture criterion for the composite under
consideration have been determined. 7 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Laminated composites; Interlaminar fracture toughness; Mixed mode
1. Introduction
Interlaminar fracture behavior of laminated poly-
meric composites has widely been investigated. A brief
review of these investigations was given in Refs. [1,2].
The main problem in predicting the failure of lami-
nated composite materials is the evaluation of interla-
minar fracture toughness properties under mixed mode
loading conditions. Various approaches have been
used to develop test specimens for combined or mixed
mode loading conditions, however extensive research
mainly has been conducted on the mode I and mode II
interlaminar fracture behavior of composite materials.
Double cantilever beam (DCB) and end notched ex-
ure (ENF) specimens have been employed for mode I
and mode II tests, respectively. Also in the mixed
mode tests mainly beam type specimens were used in
order to obtain mixed mode I/II critical energy release
rates [35].
But still the disadvantage of all beam type specimens
is that for pure mode I, pure mode II and for dierent
mixed mode I/II ratios dierent beam type specimens
have to be used. For example, to establish the fracture
criterion for delamination in Ref. [6] four dierent
beam specimens (DCB, ENF, CLS and modied ENF)
were used. In order to obtain reliable results for inter-
laminar fracture toughness under dierent combined
or mixed mode loading conditions starting from pure
mode I to pure mode II it is desirable that only one
type of specimen would be required. Because this was
almost impossible to achieve by using beam type speci-
mens also other types of test specimens have been
developed. One other type is the so called Arcan speci-
men, which rstly was employed for the combined
loading of composites [7], then for fracture tests of iso-
tropic materials [8] and also for mixed mode fracture
tests of composites [9,10]. Another type of specimen
for studying the interlaminar fracture properties of
laminated composites is the Iosipescu specimen [11].
The most recent development is a compound version
of the compact tension shear (CTS) specimen which
Computers and Structures 76 (2000) 1118
0045-7949/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved.
PII: S0045- 7949( 99) 00148- 0
www.elsevier.com/locate/compstruc
E-mail address: rikards@latnet.lv (R. Rikards).
covers all in-plane mixed mode loading conditions
starting from pure mode I through any mixed mode I/
II ratio up to pure mode II loading. The CTS speci-
men was proposed in Ref. [12,13] for fracture tests of
isotropic materials under such general in-plane loading
conditions. In the present paper the compound version
of the CTS specimen is used for mode I, mixed mode
I/II and mode II studies of the interlaminar fracture
toughness properties of a glass/epoxy composite. For
pure mode I and pure mode II loading conditions the
results obtained by the CTS tests have been compared
with those obtained by the beam tests. A detailed
description of the beam tests was given in Ref. [1].
2. Experimental
2.1. Material
For the production of the composite material to be
considered in this investigation glass bers from the
company Owens Corning Ltd and a ductile epoxy
matrix with an ultimate strain of 7.7% at failure in
tension (Araldite epoxy resin from Ciba Geigy) were
used. The glass/epoxy composite investigated in the
present paper is similar to the material investigated
previously [1]. In Ref. [1] the brittle epoxy matrix with
an ultimate strain of 3.3% at failure in tension was
used. In the present paper a more ductile matrix is
employed in order to reduce the inuence of imperfec-
tions since in the CTS test the width of the composite
specimen is only 10 mm. For the beam tests the speci-
men width is 25 mm, i.e. the same as previously [1]. In
Ref. [1] the glass bers sized with polyethylene or
silane coupling agents were tested whereas in the pre-
sent investigation industrial bers with a standard sur-
face treatment are used.
Unidirectionally (UD) reinforced glass ber plates
were produced through winding technology and in the
production of the composite the 68-69 TEX (g/1000
m) roving was used. The plates contained a starter
crack, which was introduced by a Teon lm (thick-
ness 40 mm) placed between the central plies. The total
number of plies was 12 in order to get a plate thick-
ness of approximately 3 mm. The laminates were pro-
duced by following the standard cure cycle
recommended by Ciba Geigy. In order to improve the
quality of the plates and to reduce the void content,
the plates were placed in a vacuum before curing. The
specimens (composite strips of 10 mm width) were cut
with a diamond wheel and kept at room temperature
until testing (238C and 50% of relative humidity). All
tests were performed at the same conditions. The ber
volume content of the specimens was 50%. In calcu-
lations the composite strip is considered to be a homo-
geneous transversally isotropic material.
2.2. Compound version of the CTS specimen
The CTS specimen (see Fig. 1) which was proposed
in Refs. [12,13] for mixed mode fracture investigations
of isotropic materials can also be used for laminated
composites with some modications. Such a compound
version of the CTS specimen (see Fig. 2) is used in the
present paper in order to investigate the interlaminar
fracture toughness of laminated composites. The edge
cracked composite strip is glued into glassmat/epoxy
Fig. 1. Loading of the CTS specimen.
Fig. 2. Geometry of the compound version of the CTS speci-
men: width of the specimen t = 10 mm, dimensions of the
specimen are given in mm.
R. Rikards / Computers and Structures 76 (2000) 1118 12
blocks. These blocks again are glued into Aluminium
end blocks in order to carry the applied loads. The in-
termediate glassmat/epoxy blocks must be of about the
same modulus of elasticity as the transverse modulus
of the composite strip in order to reduce the inuence
of the composite strip/block interface on the stress
eld at the crack tip.
The forces F
i
(i = 1, 2, . . . ,6) acting on the holes of
the CTS specimen (see Fig. 2) can be calculated by the
relations [12]
F
3
= F
4
= P
_
1
2
cos a
c
b
_
F
2
= F
5
= Psin a
F
1
= F
6
= P
_
1
2
cos a
c
b
_
(1)
where a is the loading angle (see Fig. 1). In the nite
element analysis these forces are applied at the corre-
sponding nodes of the nite element mesh.
The advantage of the compound version of the CTS
specimen is that with the same test specimen all in-
plane loading conditions can be achieved. In the pre-
sent paper experiments have been carried out for pure
mode I (a = 08), pure mode II (a = 908) and for the
mixed mode loading conditions with a = 758, 788, 818,
848 and 878 which means for mixed mode ratios,
where mode II is dominant.
It should be noted that for testing isotropic ma-
terials an increment of Da = 158 for increasing the
loading angle stepwise from a = 08 (pure mode I) to
a = 908 (pure mode II) was reasonable and sucient
(see Fig. 1). But for investigations of the interlaminar
fracture behavior of composites it is of practical inter-
est to achieve higher mode II to mode I ratios exper-
imentally. Therefore, an additional loading device has
been designed and applied, by which Da = 38 was rea-
lized in the experiments for 758 < a < 908: The highest
mode II to mode I ratio, which has been achieved in
the present investigation, was about 60 (see Table 1).
For the experiments the testing machine Zwick 1446
was used. The specimens were tested by loading the rig
with a constant displacement rate of 0.5 mm/min and
the critical load P
c
and the corresponding displacement
in the direction of the load P have been measured. In
all cases unstable crack growth was observed. By using
the compound version of the CTS specimen 28 lami-
nated composite samples were tested at all for various
mixed mode ratios.
3. Finite element solution
The nite element model of the specimen consists of
2681 quadrilateral plane strain elements with four
nodal points (linear approximation of displacements
within the element) and there are at all 7310 degrees of
freedom (DOF). The loads on the specimen are applied
according to Fig. 2 and are calculated by the ex-
pressions (1). The nite element analysis was per-
formed by the program ABAQUS [14].
For the calculation of the energy release rates G
I
and G
II
the Modied Virtual Crack Closure Integral
(MVCCI) method is used. Then the separated strain
energy release rates are obtained by only one calcu-
lation (MVCCI or 1C method) for the actual crack
length a as proposed in Ref. [15]
G
1C
I
(a) =
1
tDa
1
2
_
F
i
y
(a)Du
i1
y
(a)
_
G
1C
II
(a) =
1
tDa
1
2
_
F
i
x
(a)Du
i1
x
(a)
_
(2)
where F
i
x
(a) and F
i
y
(a) are the nodal point forces at
the crack tip node i in x and y directions, respectively,
while u
i1
x
(a) and u
i1
y
(a) are the relative nodal point
displacements of the opposite crack faces at node i 1
in x and y directions, respectively. Therefore u
i1
x
(a) is
the relative crack sliding displacement, while u
i1
y
(a) is
the relative crack opening displacement at a distance
Da behind the crack tip and t = 10 mm is the width of
the specimen. In the nite element model a crack incre-
ment of Da = 0:5 mm was used.
The energy release rates have been calculated by
using the following elastic properties for the dierent
parts of the compound CTS specimen. For the trans-
versally isotropic glassmat/epoxy blocks the properties
in the plane 12 (see Fig. 2) are as follows
E
1
= E
2
= 5:7 GPa; n
12
= 0:35; G
12
= E
1
=2(1 n
12
)
The transversally isotropic glassmat/epoxy blocks are
randomly reinforced in plane 12. For the Aluminium
blocks (isotropic material) Young's modulus and Pois-
son's ratio are given by
E = 70 GPa; n = 0:30
For the composite strip (see Fig. 2) the elastic proper-
ties are as follows
Table 1
Mixed mode ratios of the combined CTS specimen under
dierent loading angles a
a 08 158 308 458 608 758 788 818 848 878 908
G
II
=G
I
0 0.001 0.045 0.136 0.416 2.0 3.17 5.80 13.0 59.9 o
R. Rikards / Computers and Structures 76 (2000) 1118 13
E
1
= 40:7 GPa; E
2
= 10:27 GPa; G
12
= 4:2 GPa;
n
12
= 0:27; n
23
= 0:29
Composite strip is a transversally isotropic material,
where 1 is the ber direction and 2 and 3 are the direc-
tions transverse to the bers. It should be noted that
the ber direction coincides with the direction of the
crack (see Fig. 2).
By using these properties the nite element solution
for the compound CTS specimen with the initial crack
a = 45 mm was obtained. The energy release rates
have been calculated by expressions (2). The calculated
mixed mode ratios of the compound version of the
CTS specimen are presented in Table 1. It is seen that
for loading angles close to 908 very high mode II to
mode I ratios can be achieved. On the other hand for
loading angles between 08 and 158 very high mode I to
mode II ratios can be created. Therefore, by using the
compound version of the CTS specimen all mixed
mode ratios starting from pure mode I to pure mode
II can be achieved and tested.
4. Failure criterion
Experimentally measured critical loads P
c
have been
employed in order to calculate the separated critical
energy release rates G
IC
and G
IIC
: It is of interest to
compare the results obtained by the compound version
of the CTS specimen with the results obtained by the
beam tests. The same laminated composite has been
tested by using the double cantilever beam (DCB) spe-
cimen (four samples) and the end notched exure
(ENF) specimen (four samples). The mode I critical
energy release rates G
IC
at crack initiation obtained by
the DCB test and the mode II critical energy release
rates G
IIC
obtained by the ENF test have be compared
with the corresponding values obtained by the CTS
test (four samples for the mode I test and three
samples for the mode II test). The mode I critical
energy release rates obtained from these tests are as
follows: G
IC
= 0:22720:0836 kJ/m
2
(CTS test) and
G
IC
= 0:15120:0115 kJ/m
2
(DCB test). The mode II
critical energy release rates are as follows: G
IIC
=
1:65820:491 kJ/m
2
(CTS test) and G
IIC
= 1:62020:218
kJ/m
2
(ENF test). For both the beam and the CTS
tests the experimental error was calculated at 95%
condence level. The data reduction for the DCB and
ENF tests have been performed by using the linear
beam analysis. The details of these calculations have
been presented in Ref. [1]. The CTS test was analyzed
by the nite element method and the critical energy
release rates was calculated by the formulae (2).
Due to large scatter in the experimental results the t-
test for comparing mean values of two normal distri-
butions was employed. Using the t-test for comparing
the DCB experiment results (group 1, number of
samples in the rst group n
1
= 4) and the mode I CTS
experiment results (group 2, number of samples in the
second group n
2
= 4) the test fails. Therefore, the pure
mode I DCB and CTS test results dier at a 95% con-
dence level. This is due to large scatter in the pure
mode I test results. The same test for comparing the
ENF experiment results (group 1, number of samples
n
1
= 4) and the mode II CTS experiment results
(group 2, number of samples n
2
= 3) gives the conse-
quence that the pure mode II ENF and CTS test
results not dier signicantly at a 95% condence
level. This is because the scatter in the pure mode II
test results is signicantly smaller. The mixed mode
CTS test results are to be with larger scatter for the
loading conditions when mode I is dominant, and
smaller for the loading conditions when mode II is
dominant. Further for determination the fracture (fail-
ure) criterion all experimental results (the pure and the
mixed mode CTS test as well as the beam test at all
36 experimental points) are used in order to reduce the
inuence of the large scatter on the parameter esti-
mation.
In the literature [1621] a number of mixed mode
fracture criteria have been proposed for describing the
interlaminar fracture of composites. These criteria are
similar to those used for isotropic materials. The frac-
ture criteria can be formulated for dierent crack tip
fracture parameters like stress intensity factor (SIF) or
energy release rate for which the critical fracture par-
ameters (K
C
or G
C
) can be determined experimentally.
Previously, the stress intensity factors were mainly
used as fracture parameters but for composites the
energy release rates, which also are used in the present
paper, seems to be more useful and common.
In Ref. [16] Wu has proposed the following empiri-
cal relationship for the mixed mode fracture evaluation
of balsa wood or glass ber reinforced plastic
K
I
K
IC

_
K
II
K
IIC
_
2
= 1 (3)
Later this criterion was used in [17] for the evaluation
of mixed mode fracture of graphite/epoxy composites
and the parameters K
IC
and K
IIC
for criterion (3) were
obtained by the method of least squares.
The mixed mode fracture criterion can also be writ-
ten in a more general form [18,19]
_
K
I
K
IC
_
m

_
K
II
K
IIC
_
n
= 1 (4)
The quadratic form of this criterion was used in Ref.
[9] in order to evaluate the interlaminar fracture beha-
vior of graphite/epoxy under mixed mode loading con-
R. Rikards / Computers and Structures 76 (2000) 1118 14
ditions. In Eq. (4) the energy release rates can be used
instead of the stress intensity factors
_
G
I
G
IC
_
m

_
G
II
G
IIC
_
n
= 1 (5)
In Ref. [20] it was referred that for graphite/thermo-
plastic materials the best-t criterion was found with
the exponents m = 1 and n = 3=2 in Eq. (5). For the
graphite/epoxy material presented in Ref. [20] a lower
bound (i.e., most of the data lies outside of the failure
surface) was obtained with m = 1 and n = 1 in Eq. (5).
However, for that graphite/epoxy material the best-t
criterion was found of the exponential form [20]
G
I
G
II
e
c1Mc2
c
3
= 0 (6)
with
M =

1 G
II
=G
I

E
1
=E
2
_
_
and where c
1
, c
2
, and c
3
are empirical parameters and
E
1
, E
2
are Young's moduli of the composite.
In Ref. [21] it was proposed to use a non convex cri-
terion for the mixed mode failure of composites. It was
shown that for PEEK-matrix/carbon-bre composites
a concave criterion [21] gives a good agreement with
the experimental data for the crack propagation
values.
In the present paper the approximation of the results
obtained by the CTS test are based on the criterion
(5). Two variants of the criterion for the exponents
m = 1, n = 2 or m = 2, n = 2 have been analyzed. The
approximations have been performed by the method of
least squares in the polar coordinates G
I
= rcos j,
G
II
= rsin j and the cost function f (x) has been mini-
mized
f(x) =

N
i=1
_
r(x) r
i
_
2
(7)
Here r(x) is the criterion (5) in polar coordinates, x =
[G
IC
,G
IIC
] contains the parameters of optimization, r
i
are the experimental values of the critical energy
release rates in the polar coordinates and N is the total
number of the experimental points (in our case
N = 36, i.e. 28 points were obtained by the CTS test
and in addition eight points were obtained by the
beam tests). The results of the minimization of the cost
function (7) are presented in Table 2, where f
+
is the
value of the function f (x) in the point of optimum
( f
+
= f (x
+
)). For the lower value of the cost function
the approximation is better. It is seen that a better ap-
proximation is obtained for the exponents m = 1 and
n = 2 in the criterion (5). Both approximations used
here are shown in Fig. 3.
Further a more detailed analysis is performed for
the results with m = 1 and n = 2 in criterion (5). In
Fig. 3 a rather big scatter of the experimental values of
the interlaminar fracture toughness can be seen. So, it
is of interest to calculate the condence limits for the
criterion. In Fig. 4 the 95% condence limits are pre-
sented and also the mean values of the critical energy
release rates obtained experimentally for various mixed
mode ratios are shown. It can be observed that the
mean values for all mixed mode ratios are within the
95% condence limits of the criterion (5) for the expo-
nents m = 1 and n = 2.
By introducing a comparative energy release rate G
V
the fracture criterion (5) for the exponents m = 1 and
n = 2 can be written in the form [22]
G
V
=
G
I
2

1
2

G
2
I
4(a
1
G
II
)
2
_
RG
IC
(8)
Here a new material constant a
1
is introduced
Table 2
Parameters of the criterion (5) obtained by the method of
least squares
Criterion G
IC
(kJ/m
2
) G
IIC
(kJ/m
2
) f
+
m = 1; n = 2 0.175 1.561 0.878
m = 2; n = 2 0.156 1.503 1.193
Fig. 3. Approximation of the mixed mode experimental data
for criterion (5) with the exponents m = 1; n = 2 and m = 2;
n = 2:
R. Rikards / Computers and Structures 76 (2000) 1118 15
a
1
=
G
IC
G
IIC
(9)
which denotes the mode I to mode II fracture tough-
ness ratio. From Table 2 it is seen that for the compo-
site investigated in the present paper a
1
= 0:112: So,
the energy required to drive a mode II crack is about 9
times higher than that for a mode I crack.
Also it is of interest to compare the energy release
rates under dierent loading angles a of the CTS speci-
men for a constant value of the load P = 1000 N (see
Fig. 1). In Fig. 5 the comparative energy G
V
(8), the
mode II energy release rate G
II
and the total energy
release rate G
T
= G
I
G
II
are presented as functions of
the loading angle a: The mode I energy release rate G
I
practically is equal to the comparative energy G
V
from
a = 08 to a = 608: Therefore, for a wide range of the
loading angles the mode I energy release rate is domi-
nant. However, for the loading angles from a = 808 to
a = 908 (see Table 1) mode II loading conditions at the
crack tip are dominant.
Criterion (5) with m = 1 and n = 2 or expression (8)
can be written through the critical load of the struc-
ture. In the case of linear response for all parts of the
structure the relations between the energy release rates
and the load P can be given by
G
I
= g
1
P
2
: G
II
= g
2
P
2
(10)
Here the parameters g
1
and g
2
depend on the geometry
and material properties of the structure. These par-
ameters can be obtained by the nite element solution.
In the case of the CTS test the parameters g
1
and g
2
also depend on the loading angle a: For dierent load-
ing angles (see Table 1) these parameters were obtained
by the nite element analysis. By substituting ex-
pressions (10) into criterion (5) with the exponents
m = 1 and n = 2 the critical loads P
c
of the structure
can be calculated from the equation
_
g
2
G
IIC
_
2
P
4
c

g
1
G
IC
P
2
c
1 = 0 (11)
Through substituting into this equation the interlami-
Fig. 4. Criterion (5) with the exponents m = 1; n = 2 and
95% condence limits.
Fig. 5. Energy release rates G
V
, G
II
and G
T
= G
I
G
II
versus
loading angle, a, of the CTS specimen for the load P = 1000
N.
Fig. 6. Critical load, P
c
, versus loading angle, a, of the CTS
specimen.
R. Rikards / Computers and Structures 76 (2000) 1118 16
nar fracture toughness values G
IC
= 0:175 kJ/m
2
and
G
IIC
= 1:561 kJ/m
2
(see Table 2) of the glass/epoxy
composite the critical loads of the CTS specimen under
dierent loading angles can be calculated. In Fig. 6 the
predicted values obtained from Eq. (11) are compared
with the experimental results and a good agreement is
found. It should be noted that a failure criterion can
be presented through the energy release rates (5),
through the comparative energy (8) or through critical
loads (11). But for the analysis of mixed mode interla-
minar crack propagation in laminated composite struc-
tures the failure criterion in the form of Eq. (8) most
convenient to use.
5. Conclusions
In this paper a compound version of the CTS speci-
men is proposed for the investigation of critical energy
release rates of laminated polymeric composites. A
wide range of mixed mode ratios have been obtained
by changing the loading angle of the specimen. The
highest mixed mode ratio obtained in the present ex-
periments was G
II
=G
I
= 60: The interlaminar fracture
characteristics of the glass/epoxy composite have been
investigated. It was shown that in the case of pure
mode II loading conditions there are no statistically
signicant dierences among the results obtained by
the beam (ENF) and the CTS tests. In the case of pure
mode I loading conditions due to large scatter there
are dierences among the beam (DCB) and CTS test
results. The energy required to drive the crack in the
pure mode II loading condition was found to be about
9 times higher than that required under pure mode I
loading conditions and the energies required for mixed
mode I/II loading conditions are between those for the
pure mode loading conditions. The results of the
mixed mode experiments were used in order to obtain
the parameters for the failure criterion. It was shown
that a better approximation can be achieved with a lin-
ear and a quadratic term in the criterion, where the
mode I energy release rate is dominant.
Acknowledgements
This study was performed in the framework of a
project on design of interfaces of glass ber reinforced
composites. The project was performed in cooperation
between the Universities of Riga, Kassel and Pader-
born. The authors gratefully acknowledge the nancial
support by the VW Foundation (Hannover, Germany)
through grant I/70 665. This work was partly sup-
ported also through Grant No. 96.0504 by the Latvian
Council of Science. Thanks are also due to NATO
ADVANCED STUDY INSTITUTE and FUNDA-
CAO PARA A CIENCIA E A TECNOLOGIA, POR-
TUGAL for their generous support of this paper
through the travel grant to the conference.
References
[1] Rikards R, Korjakin A, Buchholz F-G, Bledzki AK,
Wang H, Wacker G. Interlaminar fracture toughness of
GFRP inuenced by ber surface treatment. J
Composite Mat 1998;32:152859.
[2] Albertsen H, Ivens J, Peters P, Wevers M, Verpoest I.
Interlaminar fracture toughness of CFRP inuenced by
bre surface treatment. Part I: experimental results.
Comp Sci Technol 1995;54:13345.
[3] Reeder JR, Crews jr JH. Mixed-mode bending method
for delamination testing. AIAA J 1990;28:12706.
[4] Kinloch AJ, Wang Y, Williams JG, Yayla P. The mixed
mode delamination of bre composite materials. Comp
Sci Technol 1993;47:22537.
[5] Zhao S, Ga dke M, Prinz R. Mixed-mode delamination
behavior of carbon/epoxy composites. J Reinf Plast and
Comp 1995;14:80426.
[6] Hwu C, Kao CJ, Chang LE. Delamination fracture cri-
teria for composite laminates. J Comp Mater
1995;29:196287.
[7] Arcan M, Hashin Z, Voloshin A. A method to produce
plane-stress states with application to ber-reinforced
materials. Experimental Mechanics 1978;18:1416.
[8] Banks-Sills L, Arcan M, Bui HD. Toward a pure shear
specimen for K
IIC
determination. Int J Fracture
1983;22:R9R14.
[9] Jurf RA, Pipes RB. Interlaminar fracture of composite
materials. J Comp Mater 1982;16:38694.
[10] Yoon SH, Hong CS. Interlaminar fracture toughness of
graphite/epoxy under mixed mode deformations.
Experimental Mechanics 1990;30:2349.
[11] Bansal A, Kumosa M. Application of biaxial Iosipescu
method to mixed-mode fracture of unidirectional compo-
sites. Int J Fracture 1995;71:13150.
[12] Richard HA, Benitz K. A loading device for the creation
of mixed mode in fracture mechanics. Int J Fracture
1983;22:R55R58.
[13] Richard HA. Crack problems under complex loading. In:
Sih GC, Nisitani H, Ishihara T, editors. Role of Fracture
Mechanics in Modern Technology. Proc. Int. Conf. on
the Role of Fracture Mechanics in Modern Technology,
Fukuoka, Japan, 26 June. Amsterdam: North-Holland,
1987. p. 57788.
[14] Hibbitt, Karlson and Sorensen Inc. ABAQUS User's
Manual, Version 5.3, Providence Rhode Island, 1993.
[15] Rybicki EF, Kanninen MF. A nite element calculation
of stress intensity factors by modied crack closure inte-
gral. Engng Fracture Mech 1977;9:9318.
[16] Wu EM. Application of fracture mechanics to anisotro-
pic plates. J Appl Mech 1967;34:96774.
[17] McKinney JM. Mixed-mode fracture of unidirectional
graphite/epoxy composites. J Comp Mater 1972;6:1646.
R. Rikards / Computers and Structures 76 (2000) 1118 17
[18] Sih GC. Strain-energy-density factor applied to mixed
mode crack problems. Int J Fracture 1974;10:30521.
[19] Di Leonardo G. Fracture toughness characterization of
materials under multiaxial loading. Int J Fracture
1979;15:53752.
[20] Whitney JM. Experimental characterization of delamina-
tion fracture. In: Pagano NJ, editor. Interlaminar
Response Of Composite Materials. Amsterdam: Elsevier,
1989. p. 161250.
[21] Charalambides M, Kinloch AJ, Wang Y, Williams JG.
On the analysis of mixed-mode failure. Int J Fracture
1992;54:26991.
[22] Richard HA. Safety estimation for construction units
with cracks under complex loading. In: Rossmanith HP,
editor. Structural Failure, Product Liability and
Technical Insuarance: Proc. of 2nd Int. Conf., July 13,
1986, Vienna, Austria. Geneva: Inderscience, 1987. p.
42337.
R. Rikards / Computers and Structures 76 (2000) 1118 18