The current issue and full text archive of this journal is available at

www.emeraldinsight.com/1573-6105.htm

Composite
Continuum description of damage laminates
and failure of composite
laminates based on viscous
regularization 525
Received 24 August 2013
ShiYang Zhao and Pu Xue Revised 19 November 2013
School of Aeronautics, Northwestern Polytechnical University, Xi’an, China 12 February 2014
15 April 2014
Accepted 10 May 2014
Abstract
Purpose – The purpose of the paper is to improve the calculability of a continuum damage failure
model of composite laminates based on Tsai-Wu criteria.
Design/methodology/approach – A technique based on viscous regularization, a characteristic
element length and fracture energies of fiber and matrix are used in the model.
Findings – The calculability of the material model is improved. The modified model can predict the
behavior of composite structure better.
Originality/value – The convergence problem and the mesh softening problem are main concern in
the calculability of numerical model. In order to improve the convergence, a technique based on viscous
regularization of damage variable is used. Meanwhile, characteristic element length and fracture
energies of fiber and matrix are added into the damage constitutive equation to reduce the mesh
sensitivity of numerical results. Finally, a laminated structure with damages is implemented using a
User Material Subroutine in ABAQUS/Standard. Mesh sensitivity and value of viscosity are discussed.
Keywords Convergence, Strain softening, Composite laminates, Continuum damage,
Material models, Mesh sensitivity
Paper type Research paper

1. Introduction
Some modification of a continuum damage failure model of composite laminates were
given in an earlier paper (Zhao et al., 2013), in order to improve its calculability. The
continuum damage model for composites was used to predict the damage onset and
damage evolution of material point under static loading in their paper. The model
(Zhao et al., 2013) contains the damage onset criteria based on Tsai-Wu criterion and
a nonlinear damage evolution law proposed by Batra et al. (2012) and Linde et al.
(2004). By defining an internal parameter, material properties of fiber and matrix were
degradable, respectively. The constitutive model was described in detail (Zhao et al.,
2013). The model proposed in their paper should have a good calculability, and it will
be used under various engineering loading conditions.
The calculability includes the convergence problem and the mesh softening
problem. The convergence problem is known to arise from an elastic snap-back
instability, which occurs just after the stress reaches the peak strength of the material
when using the damage model to simulate crack first nucleates or damage start (Gao
and Bower, 2004). In implementing a numerical calculation using Newton-Raphson
iteration to solve the nonlinear equilibrium equations, the radius of convergence of the Multidiscipline Modeling in Materials
Newton-Raphson scheme reduces to zero at the point of instability. In an explicit and Structures
Vol. 10 No. 4, 2014
pp. 525-536
r Emerald Group Publishing Limited
This work is financial supported by the National Natural Science Foundation of China under 1573-6105
grant no. 11072202. DOI 10.1108/MMMS-08-2013-0057
MMMS scheme, the solution quickly diverges from the equilibrium path and leads to
10,4 unphysical predictions (Gao and Bower, 2004). Various approaches have been tried to
resolve these convergence problems. Crisfield (1981) used Riks method to follow the
unstable branch of the solution during the snap-problems. A small value of viscosity
was introduced to the damage constitutive model by Crisfield (1981), Maimi et al.
(2007a, b), and Duvaut and Lions (1976).
526 In the crack band model (Bazant and Oh, 1983), it was assumed that the failure
process zone was represented by a damaged finite element zone, and the established
model was used for softening constitutive equations. In order to reduce mesh
sensitivity of numerical results, characteristic element length and fracture energies of
fiber and matrix were included in the damage constitutive equation (Bazant and Oh,
1983; Linde et al., 2004; Lapczyk and Hurtado, 2007; Wang et al., 2009).
In this paper, we use the technique based on viscous regularization of the damage
variable (a generalization of the Duvaut-Lions regularization) (Crisfield, 1981) and the
crack band model to improve the calculability of the failure model in the paper Zhao
et al. (2013). Finally, the modified model is implemented as a User Material Subroutine
(UMAT). A numerical analysis is conducted for a panel with a circular hole in the
center under tensile loading using Abaqus/Standard. Two finite element models are
used to analyze mesh sensitivity of the model. Finally, the influence of the value of
viscosity is discussed.

2. Modified damage material model

2.1 Basic frame of damage materials model
The elasticity relationship for the damaged material is given by:

fsg ¼ ½Cd feg ð1Þ

where [Cd] is the effective damage elasticity matrix; {e} is strain vector; and {s} is
stress vector.
The effective damage elasticity matrix [Cd] is shown in Equation (2):
2 3
ð1  edÞC11 ð1  edÞð1  ldÞC12 ð1  edÞC13 0 0 0
6 ð1  ldÞC22 ð1  ldÞC23 0 0 0 7
6 7
6 7
6 C33 0 0 0 7
½Cd  ¼ 6
6
7
6 ð1  edÞð1  ldÞC44 0 0 7
7
6 7
4 symmetric 0 C55 0 5
0 0 C66

ð2Þ

where C11, C12, C13, C22, C23, C33, C44, C55, C66 are the components of the elasticity in the
undamaged state, d is damage variable, e and l are two internal variables, which will be
defined in Section 2.2.

2.2 Definition of damage variable

The damage variable is defined based on the equivalent strain. A quadratic strain
polynomial formulation based on Tsai-Wu failure criteria is proposed, which is used to
detect failure initiation in both tension and compression. The equivalent strain of the Composite
material point in the damage model, g, is as: laminates
!
1 2 1 1
g12 ¼ f ; c f ; t ðe11 Þ þ  e11
e11 e11 e11f ; t e11f ; c
! !2
1 2 1 1 1 ð3Þ 527
2
g2 ¼ m; c m; t ðe22 Þ þ m; t  m; c e22 þ f
ðe12 Þ2
e22 e22 e22 e22 e12
g 2 ¼ g12 þ g22

where eij is the strain component, ef11; t ¼ sf11; t =E11 , ef11; c ¼ sf11; c =E11 , em; t m; t
22 ¼ s22 =E22 ,
em; c m; c f f f;t f;c
22 ¼ s22 =E22 , e12 ¼ T12 =E12 , and Eij are material properties, s11 ( s11 ) is the fiber
m; t m; c
tensile (compressive) strength, s22 and s22 is the matrix tensile (compressive)
f
strength, T12 is the shear strength, respectively. When g exceeds 1.0, the failure will
initiate.
Internal variables e and l are defined, through the variables g1, g2 , e ¼ g1/g, and
l ¼ le. e and l are introduced to describe the damage distribution of the fiber and
matrix, respectively, which are used for describing continuous degradation of stiffness.
The mesh sensitivity is a common problem of the constitutive models with strain
softening response. In order to minimize the mesh sensitivity of the numerical results,
the characteristic length Lc was introduced into the fracture energy-based damage
evolution law by some researchers (Bazant and Oh, 1983; Linde et al., 2004; Lapczyk
and Hurtado, 2007; Wang et al., 2009). Based on their work, the modified damage
variable is defined as follows:

1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
d ¼ 1  expð C11 C22 =Gf Gm ef11; t em; t c
22 ðg  1ÞL Þ; gX1 ð4Þ
g

where ef11; t , em; t

22 , C11, C22 are defined in Section 2.1. Gf and Gm are the fracture energy of
the fiber and matrix, respectively. g is the equivalent strain of the material point. When
gX1, damage will happen. Lc is the characteristic element length associated with the
material point of the finite element, and is supplied by Abaqus/UMAT directly. For
membranes and shells, it is a characteristic length in the reference surface (Simulia,
2011). Characteristic element length is a typical length of a line across an element for a
first-order element and is the half of the same typical length for a second-order element.

2.3 Viscous regularization of damage variable

In the numerical implementation, the stresses are updated according to Equation (1)
and the Jacobian matrix is obtained by differentiating Equation (1):
   
qs qCd
¼ ½Cd  þ : feg ð5Þ
qe qe
where {s} is the stress vector, [Cd] is the effective damage elasticity matrix, {e} is the
strain vector, fqs=qeg is the Jacobian matrix.
In Equation (5), the Jacobian matrix has two terms. The second item is anisomerous,
which will lead to bad performance and large memory usage of the damage model.
MMMS When material softening happens, the convergence problem happens. In order
10,4 to overcome it, the technique based on viscous regularization of the damage
variable was proposed, as given by Duvaut-Lions viscosity model (Duvaut and
Lions, 1976).
Using the technique of the viscous regularization, we do not need directly to use the
damage variable calculated from the aforementioned damage evolution laws; instead,
528 the damage variable are regularized via the following equations:
.v 1
d ¼ ðd  d v Þ ð6Þ
Z
where d is the damage variable calculated according to the damage evolution laws
presented; dv is the regularized damage variable used in the calculations of the damage
elasticity matrix and the Jacobian matrix. Z is the viscosity parameter controlling the
rate at which the regularized damage variable, dv, approaches the value of the true
damage variables d. To update the regularized damage variable at time t0 þ Dt, the
above equation is discretized in time as follows:
Dt Z
dv jt0 þDt ¼ djt0 þDt þ dv j ð7Þ
Z þ Dt Z þ Dt t0

It can be shown that:

qd v Dt
¼ ð8Þ
qd Z þ Dt

2.4 Jacobian matrix

The Jacobian matrix is needed in the user subroutine. Through Equation (8), the
Jacobian matrix of the modified damage model is:
       
qs qCd qd v qCd qe qg1 qe qg2 Dt
¼ ½Cd  þ : feg  þ : feg   þ  
qe qd v qe qe qg1 qe qg2 qe Z þ Dt
ð9Þ
h i h i
qCd qCd
The matrices qdv and qe are:
2 3
eC11 ð2eld v  1ÞC12 eC13 0 0 0
6 lC22 lC23 0 0 07
  6 7
qCd 6 0 0 0 07
¼6
6
7 ð10Þ
qd v
6 ð2eld v  1ÞC44 0 077
4 symmetric 0 0 05
0 0 0
2 3
d v C11 ð1  2eÞðdv Þ2 C12 d v C13 0 0 0
6 dv C22 d v C23 0 0 07
  6 7
qCd 6 0 0 0 07
¼6
6
7 ð11Þ
qe 6 ð1  2eÞðd v Þ2 C44 0 077
4 symmetric 0 0 05
0 0 0
The parameters in Equations (9-11) have been defined previously in Section 2.1 and Composite
Section 2.2. laminates
3. Flowchart of UMAT
The modified material model using the technique of the viscous regularization of the
damage variable is implemented through a UMAT, which is linked to the finite element
software ABAQUS (Simulia, 2011). The failure initiation criteria, damage degradation 529
model of material properties, and stress updating procedure, and so on, should be
included in the UMAT, shown in Figure 1. The damage variable is stored in the states
variable. There are three major steps in this flowchart. First, strain increment is read
by the UMAT in order to be used in the calculation of the strain and stress. Second, the
failure criteria are used and the damage degradation model of the material properties is
applied if failure occurs. Finally, the state variable and Jacobian matrix are updated.

4. Application and discussion

In order to verify that the modified damage model has a better calculability, a damaged
composite laminate with a circular hole in the center under tensile loading is implemented
using Abaqus/Standard with the user subroutine UMAT.

4.1 The finite element model of the laminate

The laminate consists of 24 plies in a [(01/901)6]s lay-up (Chang and Chang, 1987; Chang
et al., 1991). The material of the models is T300/1034-C graphite epoxy. Each ply has

Start: initialization

Calculus of strain tensor

Apply failure criteria

N
Failure?

Y
Apply degradation law

Corrected constitutive tensor

Updating stress tensor

Updating Jacobian matrix
Updating state variable

Figure 1.
END Flowchart of UMAT
MMMS a thickness of 0.132 mm. The length of the laminate is 200.0 mm, and the width is
10,4 25.4 mm. The diameter of the hole is 6.35 mm. The plate is loaded in tension in the
length direction by displacement boundary conditions. Only one-fourth of the laminate
needs to be modeled, with appropriate symmetric boundary conditions applied. The
finite element analysis is performed under static analysis. In order to simulate the
stress concentration around the area of the circular hole accurately, finer mesh is used
530 in the area near the hole as shown in Figure 2. Eight-node linear, reduced-integration
elements, C3D8R, are used in the finite element model. Number of elements in A-model
is 14856, and number of elements in B-model is 17,126. The properties of material
are shown in Table I. The subscripts “T” and “C” refer to tension and compression,
respectively.
The technique based on viscous regularization of the damage variable is
implemented to improve the convergence of the damage model. However, since the
characteristic length calculation is based only on the element geometry without taking
into account the real cracking direction, some level of mesh sensitivity remains
(Simulia, 2011). Elements with an aspect ratio close to one recommended by Abaqus/
Analysis (Simulia, 2011) (Abaqus mesh technology) can reduce mesh sensitivity
further. Therefore, the finite element model uses elements with an aspect ratio close
to one, shown in Figure 2(a) and labeled as A-model. The other finite element model
in Figure 2(b) is labeled as B-model which uses elements with anomalous shape in the

(a) Displacement
Y–Symmetry. loading

X–Symmetry

Y
Z
Z–Symmetry at the back
X face of the laminate

(b) X–Symmetry Y–Symmetry.

Z
Y
Figure 2. Displacement
X Z–Symmetry at the back loading
Finite element model of face of the laminate
laminate with damage
Notes: (a) A-model; (b) B-model

E1 (GPa) 146.9 G23(GPa) 5.78 Y C (MPa) 268.2

E2 (GPa) 11.38 G12 ¼ G13 (GPa) 6.185 Y T (MPa) 66.54
E3 (GPa) 12.38 S12 ¼ S1 (MPa) 133.8 ZC (MPa) 280
Table I. u12 ¼ u13 0.3 S23 (MPa) 100 ZT (MPa) 52
Material properties XT (MPa) 1,730.6 XC (MPa) 137
left area of Figure 2(b). The quality of mesh partition of B-model is not good. B-model is Composite
used to validate that A-model is very effective to reduce mesh sensitivity further. laminates
4.2 Choosing a proper value of viscosity
When using the technique based on viscous regularization of the damage variable,
care must be taken when determining the value of the viscosity since a large value
of viscosity may cause a noticeable delay in the degradation of the stiffness. 531
Figure 3 and Table II give the stress – time curves and failure strength of A-model
for different values of viscosity parameter, Z, labeled as A  Z ¼ 2eð4Þ,
A  Z ¼ 3eð4Þ, A  Z ¼ 5eð4Þ. The experimental result was given in the paper
Chang and Chang (1987). The smaller the viscosity, the more abrupt the failure and
the lower the failure strength. In Figure 3, the failure curve of A  Z ¼ 2eð4Þ is
discontinuous when damage reaches a certain value. So the convergence in this case
is poor. The failure curve of A  Z ¼ 3eð4Þ is continuous until the full failure of the
laminates. Therefore, the convergence of the model at A  Z ¼ 3eð4Þ is noticeably
improved. The unrealistic energy due to viscous regularization (available as output
variable ALLCD) and the strain energy of the model (ALLSE) at A  Z ¼ 3eð4Þ are
shown in Figure 4. When A  Z ¼ 3eð4Þ, the energy dissipated due to the viscosity is

450
A–  =2e–4
400
A–  =3e–4
350
A–  =5e–4
300
Stress/Mpa

250
200
150
100
50 Figure 3.
0 The stress-time
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 curves of A-model
Time/s

Z ¼ 2e(4) Z ¼ 3e(4) Z ¼ 5e(4) Experimental results Table II.

Strength for the viscosity
367.9 MPa 375.6 MPa 402.2 MPa 358 MPa parameter in A-model

1,600
1,400 ALLSE–A– =3e–4
1,200 ALLCD–A– =3e–4
Energy (mJ)

1,000
800
600 Figure 4.
400 Unrealistic energy
200 (ALLCD) and strain
0 energy (ALLSE) at
0 0.005 0.01 0.015 0.02 0.025 0.03 AZ ¼ 3e(4)
Time (s)
MMMS relatively small, and this does not obviously affect the results. Therefore, a viscosity of
10,4 0.0003 is chosen and used for the study of element selection.

4.3 Damage analysis of the laminate

Damage distribution of the laminate with the circular hole under static loading is
shown in Figure 5. Because of the orthogonal characteristic of the laminate in
532 a [(01/901)6]s lay-up, the damage distributing of all 01 plies is the same, and the damage
distributing of all 901 plies is also the same. The state variable of SDV1 is set for the
damage variable of the material model modified in Section 2.2.
In the 01 layer, the process of the damage propagation is shown in Figure 5(a)-(c).
The damage propagates along a line, which is orthogonal to the loading direction.
And, there is almost no damage in the 901 fiber-reinforced epoxy layer before sudden
fracture, shown in Figure 5(d). The results are in good agreements with the
experimental process (Chang and Chang, 1987; Chang et al., 1991).

(a)
SDV1
(Avg: 75%)
+9.995e – 01
+9.162e – 01
+8.329e – 01
+7.496e – 01
+6.663e – 01
+5.830e – 01
+4.997e – 01
+4.164e – 01
+3.332e – 01
+2.499e – 01
+1.666e – 01
+8.329e – 02
+0.000e + 00

(b)
SDV1
(Avg: 75%)
+9.997e – 01
+9.164e – 01
+8.331e – 01
+7.498e – 01
+6.665e – 01
+5.832e – 01
+4.999e – 01
+4.165e – 01
+3.332e – 01
+2.499e – 01
+1.666e – 01
+8.331e – 02
+0.000e + 00

(c)
SDV1
(Avg: 75%)
+1.000e + 00
+9.167e – 01
+8.333e – 01
+7.500e – 01
+6.667e – 01
+5.833e – 01
+5.000e – 01
+4.167e – 01
+3.333e – 01
+2.500e – 01
+1.667e – 01
+8.333e – 02
+0.000e + 00

(d)
SDV1
(Avg: 75%)
+0.000e + 00
+0.000e + 00
+0.000e + 00
+0.000e + 00
+0.000e + 00
+0.000e + 00
+0.000e + 00
+0.000e + 00
+0.000e + 00
Figure 5. +0.000e + 00
+0.000e + 00
+0.000e + 00
Damage process +0.000e + 00
of the laminate
Notes: (a) 0° ply; (b) 0° ply; (c) 0° ply; (d) 90° ply
4.4 Mesh sensitivity Composite
The mesh quality in B-model is not good, because it is intend to sue B-model to validate laminates
that A-model effectiveness. Figure 6 gives the finally damage distributing of B-model.
Damage distributing at the bottom area of Figure 6 has some difference with that
of Figure 5(c). However, it is almost the same as that of A-model. Figure 7 gives the
stress-time results of B-model for different values of viscosity parameter, Z, labeled
as B  Z ¼ 1eð4Þ, B  Z ¼ 5eð4Þ, B  Z ¼ 2eð3Þ, B  Z ¼ 3eð3Þ. When the 533
value of viscosity parameter is small ( B  Z ¼ 5eð4Þ ) and the energy dissipated due
to viscosity is relatively small, the predicted strength shown in Table III is 381.7 MPa,
which is larger than that of A-model at Z ¼ 3eð4Þ. Mesh sensitivity in B-model is
still existent and serious. Therefore it is verified that the calculability of the numerical
model proposed here will be also affected by mesh. Meanwhile, the work validates that
elements with an aspect ratio close to one can reduce mesh sensitivity further.
The failure curves of B  Z ¼ 5eð4Þ, B  Z ¼ 2eð3Þ are discontinuous when
damage reaches a certain value. So the convergence of the model in B-model is poor.
The convergence of B  Z ¼ 3eð3Þ is noticeably improved. However, failure strength
is 468 MPa and its error of strength 30.7 percent. In Figure 8, the viscous energy
dissipated due to the use of viscosity (ALLCD) is relatively big, and it markedly affects
the results. Therefore, the effect of viscosity parameter in B-model is not well. And the
effective value of viscosity parameter is related to a specific FE mesh. The work also

SDV1
(Avg: 75%)
+1.080e + 00
+9.901e – 01
+9.001e – 01
+8.101e – 01
+7.201e – 01
+6.301e – 01
+5.401e – 01
+4.501e – 01
+3.601e – 01 Figure 6.
+2.700e – 01
+1.800e – 01 Damage distributing
+9.001e – 02 of 01 ply in B-model
+0.000e + 00

500 B– =1e–4
B– =5e–4
400
B– =2e–3
Stress/Mpa

300 B– =3e–3

200

100
Figure 7.
0 The stress-time
0 0.01 0.02 0.03 0.04 0.05 curves of B-model
Time/s

Z ¼ 1e(4) Z ¼ 5e(4) Z ¼ 2e(3) Z ¼ 3e(3) Table III.

Strength for the viscosity
381.7 MPa 396.1 MPa 441.6 MPa 468 MPa parameter in B-model
MMMS validates that the convergence of the damage material model is also affected by mesh
10,4 partition.

4.5 Numerical result

In Section 4.2, viscosity of 0.0003 is recommended. A-model has been validated to be
a good and effective model in Section 4.4. Therefore, A-model using a viscosity of
534 0.0003 is chosen to study the damage process of the laminate. The experimental value
is 358 MPa (Chang and Chang, 1987), and maximum stress for the model with viscous
regularization is 375.6 MPa. The error of the modified model is 4.9 percent.

4.6 The laminate of [(451/451)6]s

The laminate of [(451/451)6]s is also from the reference (Chang and Chang, 1987;
Chang et al., 1991). The model of [(451/451)6]s are the same to that of [(01/901)6]s in
Section 4.1. The laminate of [(451/451)6]s will be loaded under tensile direction.
Figure 9 gives the stress-time results of [(451/451)6]s for different values of
viscosity parameter, Z, labeled as Z ¼ 5eð5Þ, Z ¼ 1eð4Þ, Z ¼ 2eð4Þ. In Figure 9,
the failure curve of Z ¼ 5eð5Þ is discontinuous when damage reaches a certain value.
So the convergence at Z ¼ 5eð5Þ is poor. The failure curve at Z ¼ 1eð4Þ is
continuous until the full failure of the laminates. Therefore, the convergence of
the material model is noticeably improved. Table IV gives the failure strength of

3,500
3,000 ALLSE–B– =3e–3
2,500
ALLCD–B– =3e–3
Energy (mJ)

2,000
1,500
Figure 8.
1,000
Unrealistic energy
(ALLCD) and strain 500
energy (ALLSE) at 0
BZ ¼ 3e(3) 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
–500
Time (s)

160 =5e–5
140
=1e–4
120
=2e–4
Stress/Mpa

100
80
60
40

Figure 9. 20
The stress-time 0
curves of [(451/451)6]s 0 0.01 0.02 0.03 0.04
Time/s

Table IV.
Strength for the Z ¼ 5e(5) Z ¼ 1e(4) Z ¼ 2e(4) Experimental results
viscosity parameter
in [(451/451)6]s 132 MPa 135 MPa 138 MPa 127 MPa
[(451/451)6]s for different values of viscosity parameter. The experimental value of Composite
[(451/451)6]s is 127 MPa (Chang and Chang, 1987), and failure strength for the model laminates
with Z ¼ 1eð4Þ is 135 MPa. The error is 6.3 percent. Meanwhile, the value of viscous
parameter, Z ¼ 1eð4Þ, is very small. Therefore, the calculability of the numerical
model for the laminate of [(451/451)6]s is also good.

5. Conclusion 535
A modified three dimensional damage model of composite laminates is proposed to
improve convergence and mitigate mesh sensitivity of the finite element model. The
modified model is implemented as a UMAT in Abaqus/Standard. The composite
laminates with a circular hole in the center under tensile loading is simulated. The
main work and conclusions are as follows:
(1) A technique based on viscous regularization of the damage variable is used
which can improve the convergence of damage models. Care must be exercised
to choose an appropriate value of the viscosity, and a viscosity of 0.0003 is
used here.
(2) Fracture energies of fiber and matrix and a characteristic element length are
added into the damage constitutive equation which can reduce mesh sensitivity
of numerical results. Meanwhile, mesh partition using elements with an aspect
ratio close to one is a very important and effective method to reduce mesh
sensitivity. The convergence of the damage material model is also affected by
mesh partition.
(3) The curve for the model with viscous regularization is glossier and the time
step computed is longer. When material softening happens, the damage
material model still can work. The numerical results verify that the modified
damage model has the better capacity of predicting the behavior of composite
structure and better calculability of the damage model.

References
Batra, R.C., Gopinath, G. and Zheng, J.Q. (2012), “Damage and failure in low energy impact of
fiber-reinforced polymeric composite laminates”, Compos Struct, Vol. 94 No. 2, pp. 540-547.
Bazant, Z.P. and Oh, B.H. (1983), “Crack band theory for fracture of concrete”, Mater Struct,
Vol. 16 No. 3, pp. 155-177.
Chang, F.K. and Chang, K.Y. (1987), “A progressive damage model for laminated composites
containing stress concentrations”, J Compos Mater, Vol. 21 No. 9, pp. 834-855.
Chang, K.Y., Liu, S. and Chang, F.G. (1991), “Damage tolerance of laminated composites
containing an open hole and subjected to tensile loadings”, J Compos Mater, Vol. 25 No. 3,
pp. 274-301.
Crisfield, M.A. (1981), “A fast incremental/iteration solution procedure that handles
‘Snap-Through’”, Comput Struct, Vol. 13 Nos 1/3, pp. 55-62.
Duvaut, G. and Lions, J.L. (1976), Inequalities in Mechanics and Physics, Springer, Berlin.
Gao, Y.F. and Bower, A.F. (2004), “A simple technique for avoiding convergence problems in finite
element simulations of crack nucleation and growth on cohesive interfaces”, Modelling
Simul Mater Sci Eng, Vol. 12 No. 3, pp. 453-463.
Lapczyk, I. and Hurtado, L.A. (2007), “Progressive damage modeling in fiber-reinforced materials”,
Composites Part A: Applied Science and Manufacturing, Vol. 68 No. 11, pp. 2333-2341.
MMMS Linde, P., Pleitner, J., Boer H, D. and Carmore, C. (2004), “Modelling and simulation of fiber metal
laminates”, paper presented at ABAQUS User’s Conference, Boston, MA, pp. 421-439.
10,4
Maimi, P., Camanho, P.P., Mayugo, J.A. and Davila, C.G. (2007a), “A continuum damage model for
composite laminates: part I-constitutive model”, Mech Mater, Vol. 39 No. 10, pp. 897-908.
Maimi, P., Camanho, P.P., Mayugo, J.A. and Davila, C.G. (2007b), “A continuum damage model for
composite laminates: part II-computational implementation and validation”, Mech Mater,
536 Vol. 39 No. 10, pp. 909-919.
Simulia (2011), ABAQUS Analysis User’s Manual V6.11, Simulia, Providence, RI.
Wang, Y.Q., Tong, M.B. and Zhu, S.H. (2009), “Three dimensional continuum damage mechanics
model of progressive failure analysis in fibre-reinforced composite laminates”, AIAA,
pp. 2009-2629.
Zhao, S.Y., Xue, P. and Zhu, J. (2013), “Continuum description of damage and failure for composite
laminates under static loading”, paper presented at the 9th International Conference
on Fracture and Strength of Solids, FEOFS2013, Jeju Island, June 9-13, available at: www.
feofs2013.org/html/main/main.html (accessed February 13, 2013).

About the authors

ShiYang Zhao is studying for his Doctor’s Degree at the School of Aeronautics, Northwestern
Polytechnical University, Xi’an, China. His research interests are mechanical behavior of
composite laminates, finite element method and simulation. ShiYang Zhao is the corresponding
author and can be contacted at: [email protected]
Pu Xue is a Professor at the School of Aeronautics, Northwestern Polytechnical University,
Xi’an, China. Her research interests are material mechanics, mechanical vibrations, dynamic
modeling and simulation.

To purchase reprints of this article please e-mail: [email protected]

Or visit our web site for further details: www.emeraldinsight.com/reprints
Copyright of Multidiscipline Modeling in Materials & Structures (Emerald Group Publishing
Limited) is the property of Emerald Group Publishing Limited and its content may not be
copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for
individual use.