International Journal of Structural Stability and Dynamics

Vol. 10, No. 4 (2010) 683
703

.c World Scienti¯c Publishing Company
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

#
DOI: 10.1142/S0219455410003683

NUMERICAL TREATMENT OF DAMAGE PROPAGATION

IN AXIALLY COMPRESSED COMPOSITE
AIRFRAME PANELS
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

CLAUDIO BALZANI* and WERNER WAGNER

Karlsruhe Institute of Technology (KIT)
Institute for Structural Analysis
Kaiserstraße 12, 76131 Karlsruhe, Germany
*[Link]@[Link]

Received 8 January 2009

Accepted 19 May 2009

In the design phase of stringer-sti®ened composite airframe panels, it is a key issue to exploit
material reserves as far as possible to create lighter and safer aircraft. A recent approach is to
apply postbuckling design — standard for metallic panels — also to composite parts. This work
focusses on the development of a simulation procedure which accurately predicts the post-
buckling response of composite panels while accounting for damage propagation. For this
purpose we employ a robust shell element formulation which allows for arbitrary stacking
sequences as well as a variable location of the reference plane. A ply discount model is incor-
porated to account for intralaminar damage growth. The cohesive zone approach is
implemented in a so-called interface element to predict interlaminar damage growth, respective
skin
stringer separation. The numerical model is validated via a numerical example with
experimental evidence.

Keywords: Composite panel; damage growth; postbuckling design.

1. Introduction
Carbon-¯ber/epoxy composites set up an ideal material class for airframe panels,
since they combine high sti®ness and strength with low weight. Thus there is huge
potential for reducing the weight of aircraft and, with that, for increasing their e±-
ciency when metallic parts are replaced by their composite counterparts. In particu-
lar, the fuselage bene¯ts from composite lightweight design. Fuselage panels tend to
buckle under axial compression due to their slenderness. However, the load can
usually be increased signi¯cantly beyond the buckling point. Unfortunately, although
it has been standard for metallic panels for many years, postbuckling design is not
utilized for composite panels. To close this gap the project named COCOMAT1,2 was
launched in 2004; it aimed to set up a related experimental database, as well as to

683
 684 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

develop numerical procedures which could be used for postbuckling design. These
should be able to capture the postbuckling response of composite airframe panels
while taking into account damage initiation and propagation. This paper presents
the authors' contribution to the project; they were in charge of developing a ¯nite-
element-based numerical tool.
Damage processes in composite laminates are very complex, ranging from intra-
laminar damage, e.g. ¯ber fracture or matrix cracking, to interlaminar damage, i.e.
delamination. The COCOMAT project further focussed on skin
sti®ener debonding,
which is treated as a special case of delamination.
Quadrilateral shell elements are employed for the simulation of the laminates in
this work. A material law has been implemented to account for intralaminar damage.
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

This so-called ply failure model is based on a ply-by-ply failure analysis where — in
each ply — the di®erent failure types are predicted via applicable failure criteria, in
our case extended Hashin3 criteria after Ref. 4. Following detection of damage a
selective degradation model is applied. In contrast to Refs. 4
6, the degradation
model in this paper directly reduces the entries of the elasticity matrix and not the
engineering constants. The choice of the respective entries depends on the predicted
failure type. The degradation model relies on constant knockdown factors providing
a simple and e±cient derivation of the tangential sti®ness matrix. E®orts have been
made7
9 to establish continuum damage mechanics models which gradually reduce
the elastic properties. However, the derivation of the tangent is much more com-
plicated and more material properties have to be determined in advance.
Interlaminar damage, i.e. delamination or skin
sti®ener debonding, is described
via so-called interface elements (see e.g. Refs. 10
14), in which a cohesive law can be
implemented very easily. The cohesive zone model goes back to Dugdale15 and
Barenblatt;16 the latter established a theory of molecular forces of cohesion acting at
a crack front. These serve to eliminate in¯nite stresses at the crack tip and to
establish smooth closing of the crack faces. In combination with interface elements
the cohesive forces further serve to hold the crack faces together as long as the
interface is intact. Several cohesive laws have been proposed. The most widely used is
a bilinear one (see e.g. Refs. 14, 17
19), with an initial linear penalty sti®ness fol-
lowed by linear degradation. However, the exponential one proposed by de Andr
es
et al.20 and Ortiz and Pandol¯,21 based on the universal binding law of Rose et al.,22
is computationally more e±cient, as has been shown in Ref. 23. Hence, the latter is
employed in this work. The original version from Refs. 20 and 21 is edited in such a
way that only mode I tension, and not mode I compression, triggers delamination.
Furthermore, a simple penalty contact formulation is added which avoids the inter-
penetration of the crack faces.
After a detailed derivation of the approach for intra- and interlaminar damage in
Secs. 2 and 3, a validation example is presented in Sec. 4 which highlights the
applicability of the proposed concept for simulations of composite airframe panels in
the deep postbuckling region. Two versions of a single-sti®ener °at panel under axial
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 685
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

compression — an intact version and a predamaged version — are investigated and

compared with experiments performed by de Zarate Alberdi and Rebollo.24

2. Ply Failure Approach

This section describes the approach to modeling the intralaminar elastic behavior as
well as the corresponding damage processes. Damage within a ply may occur as ¯ber
fracture, transverse matrix cracking, or combined ¯ber
matrix shear failure. The
approach utilized in this work is very close to continuum damage mechanics (CDM),
since the e®ect of a crack on the mechanical behavior of the structure is of interest
and not the location of a discrete crack. However, no gradual degradation is taken
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

into account but a selective and constant knockdown of the material sti®nesses. This
is what we call the ply discount model. First, some aspects of the shell elements used
for the simulations as well as the elasticity model employed are pointed out brie°y,
followed by a more detailed derivation of the damage model.

2.1. Elastic shell concept

For the modeling of the laminate we employ quadrilateral shell elements which are
state-of-the-art. Thus the formulation is not derived in detail. Nevertheless, the basic
features are summarized in the sequel.
The shells are based on a plane shell formulation with Reissner
Mindlin kin-
ematics, and thus transverse shear is taken into account. The utilized shell elements are
folded plates which exhibit combined membrane/plate load-carrying capacities. Fol-
lowing from that the normal in-plane stresses and strains, as well as the in-plane and
the two out-of-plane shear stresses and strains, enter into the shell formulation. The
normal stress in the thickness direction and the corresponding normal strain are neg-
lected, providing a good approximation for the mechanical behavior of thin structures.
The shell element is formulated in a Lagrangian setting. Herein, the second
Piola
Kirchho® stress tensor S and the Green
Lagrangian strain tensor E are
employed, reduced by the through-thickness normal contributions, providing a
geometrically nonlinear formulation. However, quadratic terms are considered only
in the in-plane entries accounting for the essential nonlinear e®ects. The formulation
holds for moderate rotations, which is su±cient for the investigated applications and
improves its e±ciency.
It is well known that shell elements su®er from transverse shear locking e®ects. An
e®ective methodology to overcome related problems is application of the assumed
natural strains (ANS) approach, developed by Dvorkin and Bathe.25 Herein, in
addition to the displacement-based approximation, special assumptions are made for
the distribution of the transverse shear strains.
The constitutive equations rely on a layered elasticity model with a variable
location of the reference plane. For the variable reference plane, o®sets between layer
midplanes and the reference plane are considered as input parameters. For each ply a
 686 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

transversely isotropic material law is applied with plane stress assumptions enhanced
by the transverse shear parts. The corresponding material matrix in terms of
engineering constants has the well known form
2 3 2 3
C 011 C 012 0 0 0 E 1 E 2
12 0 0 0
6 7 6 7
6 0 7 6  7
6 C 12 C 022 0 0 0 7 6 E
E 
0 0 0 7
6 7 6 2 12 2
7
6 7 6 7
6 7
C0 ¼ 6 0 0 C 044 0 0 7¼6 6 0 0 G 0 0 7:
7
6 7 6 12
7
6 7 6 7
6 0 0 C 055 7 6 0 0 0 G 0 7
6 0 0 7 4 12
5
4 5
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

0 0 0 0 0
C 66 0 0 0 0 G23
ð2:1Þ

Herein,  is a suitable shear correction factor which is usually  ¼ 5=6. The in-plane
shear modulus is denoted as G12 , while the transverse/transverse shear modulus is
G23 . The set of dependent parameters is given by
E1 E2 E2
E 1 ¼ ; E 2 ¼ ;  ¼ 1

212 ; ð2:2Þ
  E1

where E1 and E2 are the Young moduli parallel and perpendicular to the ¯ber
direction and
12 is the major in-plane Poisson ratio. Obviously, the material matrix
in Eq. (2.1) is related to the material coordinates, which means that the subscripts 1
and 2 characterize the ¯ber direction and the in-plane transverse direction while the
subscript 3 denotes the thickness direction. However, in the most general case the
material coordinates do not coincide with the global coordinates to which the shell
element is related. Thus the constitutive equations have to be transformed by a
suitable transformation matrix which is straightforward. Since the damage model
refers to the local material coordinates, it seems to be su±cient to give only the
material matrix in the local form.
For the layered elasticity model, we apply the above material matrix in each ply
while accounting for the ply o®sets from the reference plane in the bending parts.
After transformation of the single ply contributions into global coordinates, these are
superimposed to generate a global material matrix of the laminate.

2.2. Damage model

The intralaminar damage model provided in the following must agree with the meso-
level elasticity model as well as with the shell concept outlined previously, and is thus
based on a plywise failure analysis while taking into account the assumption that no
normal strains and stresses in the thickness direction occur. A ply failure damage model
should generally provide answers to two questions. First, at which state of loading
does damage occur? Second, how does the material behave after damage initiation?
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 687
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

The answer to the ¯rst question is obtained by calculating the stresses with
respect to the material coordinates ply by ply and subsequently checking for failure
by evaluating suitable failure criteria. The behavior of the material after damage
initiation is obtained by using an applicable degradation model. Then the FE
solution has to be reiterated to ¯nd an equilibrium state.
Before going into details about the damage model, let us introduce a suitable
notation for the second Piola
Kirchho® stresses by S ¼ ½S11 ; S22 ; S12 ; S13 ; S23  T ,
where S11 and S22 are the normal stresses parallel and perpendicular to the ¯ber
direction and S12 , S13 , and S23 are the in-plane and out-of-plane shear stresses,
respectively. The Green-Lagragian strains are de¯ned analogously, i.e. E ¼ ½E11 ;
E22 ; 2E12 ; 2E13 ; 2E23  T . Let us further denote the corresponding strength parameters
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

according to Table 1. These strength parameters are those obtained by uniaxial tests.
The identi¯cation of damage initiation in this work builds upon the Hashin failure
criteria3 and the extension made by Goyal et al.4 for ¯ber/matrix shear failure. For
¯ber fracture the failure criterion reads
!2 !2
hS11 i h
S11 i
fF ¼ þ : ð2:3Þ
R tjj R cjj

Herein and in the following, hi ¼ 12 ð þ j  jÞ are the Macauley brackets and j  j
denotes the norm of . It should be realized that the ¯ber fracture failure criterion is
equal to the well known maximum stress criterion. In the case of transverse tension,
i.e. S22  0, the failure criterion for transverse matrix cracking has the form

2
2
S22 S23 2
S 12 þ S 132
t
fM ¼ þ þ ; ð2:4Þ
R t? R?? R 2jj?

while in the case of transverse compression, i.e. S22 < 0, the criterion for transverse
matrix cracking reads

2 
2
2
S R c? S22 S23 2
S 12 þ S 13
2
c
fM ¼ 22c
1 þ þ þ 2 : ð2:5Þ
R? 2R?? 2R?? R?? R jj?

Table 1. Notation of the uniaxial intralaminar strength parameters.

Symbol Description

R tjj Normal tensile strength parallel to the ¯bers

R cjj Normal compressive strength parallel to the ¯bers
R t? Normal tensile strength transverse to the ¯bers
R c? Normal compressive strength transverse to the ¯bers
Rjj? Shear strength parallel/transverse to the ¯bers
R?? Shear strength transverse/transverse to the ¯bers
 688 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

The criterion for combined ¯ber/matrix shear failure according to Ref. 4 for both
tension and compression in the ¯ber direction is given by the expression
!2
h
S11 i S2 þ S2
fFM ¼ c þ 12 2 13 : ð2:6Þ
R jj R jj?

The degradation model is in a way similar to that employed by Goyal et al.4 How-
ever, in Ref. 4, the engineering constants are selectively reduced, depending on the
damage type. This can lead to an inconsistent double degradation on the o®-diagonal
entries of the elasticity matrix in the event of matrix cracking, since both the
transverse Young modulus and the Poisson ratio are reduced. In contrast to that, we
directly reduce the entries of the elasticity matrix. The corresponding damage model
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

can be formulated in a thermodynamically consistent manner, in a similar way to

CDM; see e.g. Refs. 26
28. It further avoids the above-mentioned double degra-
dation. Let us introduce a Helmholtz free energy density of the form
X
5
¼ ð1
di Þ i;
0
ð2:7Þ
i¼1

where 0i ¼ 2 1
E T C 0i E are purely elastic contributions. Herein,  is the mass density
in the Lagrangian setting and di are suitable scalar-valued damage parameters. It
can be seen that the strain energy density consists of ¯ve superimposed parts. The
single elasticity matrices are de¯ned by
2 3 2 3 2 3
C 011 0 0 0 0 0 0 0 0 0 0 C 012 0 0 0
6 7 6 7 6 7
6 0 0 0 0 07 6 0 C0 0 0 0 7 6C0 0 0 07
6 7 6 22 7 6 12 0 7
6 7 6 7 6 7
C1 ¼ 6 0 ; C ¼ 6 0 0 0 0 07 C 03 ¼ 6 0 0 07
0 0 0 07 7; 7;
0 0
6 7 2 6 6 0 0
6 7 6 7 6 7
4 0 0 0 0 05 60 0 0 0 07 6 0 0 0 07
4 5 4 0 5
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

2 3 2 3
0 0 0 0 0 0 0 0 0 0
6 7 6 7
60 0 0 0 07 60 0 0 0 0 7
6 7 6 7
6 7 6 7
C 04 ¼ 6
6 0 0 C 0
44 0 0 7
7; C 0
5 ¼ 60
6 0 0 0 0 7 7; ð2:8Þ
6 7 6 7
6 0 0 0 C0 0 7 60 0 0 0 0 7
4 55 5 4 5
0 0 0 0 0 0 0 0 0 C 066

with C 0mn according to (2.1). From these it follows that 01 is the decoupled normal
contribution parallel to the ¯ber direction, 02 the decoupled normal contribution
transverse to the ¯ber direction, 03 the part accounting for the normal coupling, i.e. the
lateral contraction, 04 the contribution of the parallel/transverse shear, and 05 that of
the transverse/transverse shear. We start from the second law of thermodynamics,
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 689
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

which, in the case of isothermal conditions and vector/matrix notation, reads

: :
D :¼ S T E
  0: ð2:9Þ
:
The time derivative ðÞ ¼ dðÞ=dt of the free energy density is de¯ned by the
expression
: X5
@ 0i : X 5
¼ ð1
di Þ E
0 _
i di: ð2:10Þ
i¼1
@E T
i¼1

Substitution of (2.10) into (2.9) yields

" #
X5
@ 0i : X5
D¼ S
 T
ð1
di Þ E þ  0 _
i di  0: ð2:11Þ
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

i¼1
@E T i¼1

According to the principles of rational continuum mechanics, the dissipation inequality

:
must hold for every admissible process, and thus for every E and d_i . Therefore, the
parenthesized term in (2.11) must be equal to zero, which directly leads to
X 5
@ 0i
S¼ ð1
di Þ ; ð2:12Þ
i¼1
@E
@ 0
where  @Ei ¼ C 0i E provide the undamaged elastic stresses collected in the vector S 0 .
Obviously, the superposition of the ¯ve contributions leads to the well-known elasticity
law
S 0 ¼ C 0 E; ð2:13Þ
where the superscript 0 denotes purely elastic contributions and C 0 is de¯ned as in
(2.1). The stresses of the damaged material are obtained via the damage parameters di .
The dissipation inequality gives the following additional information for these damage
parameters:
X
5
 0 _
i di  0: ð2:14Þ
i¼1

It holds that  0i  0 by de¯nition. Consequently, it has to be ensured that the

evolution of the damage parameters is nonnegative, i.e. d_i  0. Hence, the degradation
model has to be history-dependent in order to avoid healing of the material at damaged
sites. For this purpose we introduce a set of scalar-valued internal state variables i ,
which is de¯ned by

 k1 ¼ maxf k
1
1 ; fF g;

 k2 ¼ maxf k
1
2 ; fF ; fM g;

 k3 ¼ maxf k
1
3 ; fF ; fM g; ð2:15Þ

 k4 ¼ maxf k
1
4 ; fF ; fFM g;

 k5 ¼ maxf k
1
5 ; fF g:
 690 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

Herein, the superscript k characterizes the current load step. For the initial values of
the internal variables, it holds that  0i ¼ 0. In the following i ¼  ki , and hence the
superscript for the current load step is omitted. The damage parameters may now be
de¯ned as
hi
1i
di ¼ d i ; ð2:16Þ
i
1
where d i are constant knockdown factors. To be mathematically correct we de¯ne
hi
1i
i
1 ¼ 0 for i
1 ¼ 0. Generally, the knockdown factors can be di®erent for the
di®erent damage parameters, but for simplicity we apply uniform values, i.e.
d 1 ¼ d 2 ¼ d 3 ¼ d 4 ¼ d 5 ¼ 0:99. This means that the initial elasticity matrix entries
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

are reduced to 1% of their initial values. It can be seen from (2.16) that the damage
parameters are di ¼ 0 if the respective damage did not occur yet and are di ¼ d i
otherwise. The damaged constitutive law can now be written explicitly as
S ¼ CE; ð2:17Þ
with the damaged elasticity matrix given by
2 3
ð1
d1 ÞC 011 ð1
d3 ÞC 012 0 0 0
6 7
6 ð1
d ÞC 0 ð1
d ÞC 0 7
6 3 12 2 22 0 0 0 7
6 7
6 7
C¼6 0 0 ð1
d4 ÞC 044 0 0 7: ð2:18Þ
6 7
6 7
6 0 0 0 ð1
d4 ÞC 055 0 7
4 5
0 0 0 0 ð1
d5 ÞC 066

In terms of CDM it does generally not pose a problem to include gradual softening
functions within the damage parameters, but then critical energy release rates would be
required which are not available to the authors. However, since carbon/epoxy com-
posites break in a brittle manner (except in shear mode), application of a constant and
simply stress-strength-based degradation scheme seems to be physically sound and
makes it numerically less expensive. With regard to shearing, the present model may be
extended using CDM to account for the nonlinear behavior which is generally under-
stood to be due to microcrack formation prior to collapse. Nevertheless, including
critical energy release rates requires regularization techniques to avoid mesh-dependent
solutions.29 The e®ect of mesh dependency could not be observed in the simulations
presented later on and hence does not seem to be signi¯cant when using constant
degradation.
For visualization purposes we introduce relative failure indices (RFIs) for the
di®erent failure modes. These are de¯ned by the ratio between the accumulated
thickness of plies damaged in the respective failure mode and the overall thickness of
the laminate. It can thus have values between 0 (not damaged) and 1 (completely
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 691
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

damaged). A value of 0.8, for instance, means that 80% of the cross-section is
damaged in the particular failure mode.

3. Interlaminar Failure Approach

As already mentioned, the approach to describing interlaminar damage growth
employs the cohesive zone model.15,16 While Dugdale15 considered a plastic zone ahead
of a crack tip, we follow Barenblatt's theory of molecular forces of cohesion acting at a
crack front.16 Anyway, both models introduce an extended crack tip which is called the
fracture process zone (FPZ). Within this region a softening process takes place. The
cohesive zone model is implemented in a special ¯nite element type, a so-called
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

interface element. In this work we employ the approach with an initial zero thickness
which relates interfacial tractions, the cohesive forces, to relative displacements, which
are also referred to as separations. The element formulation is described in the sequel,
followed by a description of a special cohesive law.

3.1. Interface element

In this subsection the ¯nite element formulation of the interface element used is
derived. Figure 1 illustrates one interface element in the reference and the current
con¯guration. The interface element consists of two quadrilateral surfaces; the top
surface is characterized by the superscript þ and the bottom surface by the super-
script
. Consequently, the node numbers 1; 2; 3; 4 are associated with the bottom
surface and the node numbers 5; 6; 7; 8 are associated with the top surface. Both
con¯gurations are parametrized in the Euclidean space spanned by the global rec-
tangular coordinate vectors e1 ; e2 ; e3 . A local rectangular basis system is spanned by
the vectors A1 ; A2 ; A3 , where A1 ; A2 are in-plane basis vectors and A3 is the director
on the initial con¯guration. The origin of the local basis system is located in the

Fig. 1. Reference (left) and current (right) con¯guration of one interface element.
 692 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

element's center, whose position vector in global coordinates reads X gc . The

element's center is calculated via the expression

1X 4
1X 8
1X 8
X gc ¼ X gI ¼ X gI ¼ X g; ð3:1Þ
4 I¼1 4 I¼5 8 I¼1 I

where I denotes a particular node number and X gI are the nodal position vectors in
global coordinates. Position vectors in the initial con¯guration of material points
located on the top and bottom surface are denoted as X þg and X
g in global
coordinates. These are initially coincident. The transformation matrix
T :¼ Ai  ej ¼ ½A1 ; A2 ; A3  T ð3:2Þ
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

transforms the global position vectors into a local description via

2 þ3 2
3
X1 X1
6 7 6 7
X þ ¼ T T ðX þg
X gc Þ ¼ 6 þ7
T
g
4 X 2 5; X ¼ T ðX
X c Þ ¼ 4 X 2 5;
g 6
7 ð3:3Þ
X 3þ X 3

where X þ and X
denote the position vectors on the top and the bottom surface in
local coordinates, respectively. The real, virtual, and incremental nodal displacement
vectors are given by
2 3 2 3 2 3
uI uI uI
6 7 6 7 6 7
vI ¼ 4 vI 5; vI ¼ 6 7 6
4 vI 5; vI ¼ 4 vI 5:
7 ð3:4Þ
wI wI wI

Utilizing bilinear shape functions NI ¼ 14 ð1 þ I Þð1 þ I Þ, with  2 ½
1; þ1 and
 2 ½
1; þ1 denoting natural coordinates and I 2 f
1; þ1g and I 2 f
1; þ1g the
nodal values of the natural coordinates, for both geometry and displacements, i.e.
following the isoparametric concept, we obtain
X
8 X
4
Xþ ¼ NI X I ; X
¼ NI X I ð3:5Þ
I¼5 I¼1

for the approximations of the position vectors and

X
8 X
8 X
8
uþ ¼ NI v I ; u þ ¼ NI vI ; u þ ¼ NI vI ;
I¼5 I¼5 I¼5
ð3:6Þ
X
4 X
4 X
4
u
¼ NI v I ; u
¼ NI vI ; u
¼ NI vI
I¼1 I¼1 I¼1

for the approximation of the real, virtual, and incremental displacement vectors on the
top and the bottom surface, respectively. As previously mentioned, the interface
element relates interfacial tractions to separations which are the relative displacements
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 693
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

between the top and the bottom surface as given by the expression
X
8 X
4 X
8
½u ¼ u þ
u
¼ N I vI
NI v I ¼ BI v I : ð3:7Þ
I¼5 I¼1 I¼1

The virtual and incremental separations, ½u and ½u, are calculated analogously.
The operator matrix BI , introduced in (3.7), is de¯ned by
(

1 1  I  4;
BI :¼ I NI 1; with I :¼ ð3:8Þ
1 5  I  8;

where I is an interconnection operator di®ering between nodes located on the top or

Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

the bottom surface, and 1 is a 3  3 identity matrix. The displacement jump across
the interface, namely the separations, is a strainlike quantity, while the interfacial
tractions t are a stresslike quantity, which will be determined via a cohesive law
containing constitutive equations for the traction vector in the next subsection. We
postulate the existence of a cohesive free energy function denoted as . Thus,
according to the second law of thermodynamics, the tractions and the traction
linearization matrix C are de¯ned by

@ @t @2
t¼ ; C¼ ¼ : ð3:9Þ
@½½u @½½u @½½u2

A straightforward ¯nite element formulation yields the residual vector r eifI and the
element tangential sti®ness matrix K eif TIK on the element level, given by
Z Z
r eifI ¼ B TI t dA; K eif TIK ¼ B TI CBK dA: ð3:10Þ
e0 e0

Herein,  e0 is the domain of the midplane of one element. For the integration of
Eq. (3.10), a numerical integration scheme is employed. It has been reported in
Ref. 30 that Gauss integration can lead to spurious oscillations in the stress ¯eld and
Newton
Cotes integration has been recommended.11,17,18,31,32 Alfano and Cris¯eld19
examined an increase of Newton
Cotes integration points from 2 to 20 in one
direction and reported an increase of the oscillatory e®ects. Following from this a
3  3 Newton
Cotes integration is employed in this work.

3.2. Cohesive law

This subsection describes in detail the cohesive law used within this work. This law is
based on the universal binding law of Rose et al.,22 which has also been used in
Refs. 20 and 21. However, some modi¯cations have been made, i.e. incorporating a
penalty contact formulation and avoiding damage that occurs due to compressive
normal tractions, namely in the case of contact forces. A more prominent cohesive
law is the bilinear model,14 but it has been shown in Ref. 23 that the present model
 694 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

performs more stably at the point of delamination initiation due to the absence of a
kink in the traction
separation response.
The cohesive law is formulated in terms of a scalar-valued e®ective separation
denoted as ", which is de¯ned by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
" :¼ h½½u3i 2 þ  2 ð½½u12 þ ½u ½ 22 Þ ¼ ½u
½  T C½½u; ð3:11Þ

where the subscripts 1; 2; 3 denote the local coordinates A1 ; A2 ; A3 , respectively, and

the interconnection matrix C is de¯ned by
2 2 3
 0 0
6 7
6 0 2 0 7
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

C :¼ 66 7: ð3:12Þ
7
4  5
h½½u3i
0 0
½u
½ 3
To be mathematically precise, we de¯ne hi= ¼ 0 for  ¼ 0. The Macauley brackets
in Eqs. (3.11) and (3.12) take into account only normal opening, i.e. mode I tension.
This feature is missing in Refs. 20 and 21. The e®ective separation is thus the norm of
the separation vector in the case of mode I tension and is the total shear separation,
"s , otherwise. The parameter  assigns di®erent weights for mode I and the shear
modes. To make the cohesive law history-dependent, i.e. to avoid restoration of
cohesion, we introduce an internal variable,
n :¼ maxf"; n
1 g; ð3:13Þ
where n denotes the actual load step. The internal variable thus tracks the maximum
e®ective separation in the loading history. To distinguish loading from unloading, a
loading function is further introduced which reads
h"
n
1 i
F :¼ : ð3:14Þ
"
n
1
The loading function is equal to 1 in the case of loading and vanishes otherwise. In
the following, RI and RII ¼ RIII ¼ Rs denote the single mode I interlaminar tensile
strength and the shear mode strength, respectively. It should be noted that it is
assumed here that the mode I interlaminar tensile strength is equal to the intrala-
minar tensile strength transverse to the ¯ber direction. Furthermore, equal strengths
for mode II and mode III are assumed due to lack of experimental information about
mode III.
Figure 2 shows the e®ective traction
separation behavior. The e®ective traction
is denoted as t. It can be seen that the tractions increase mildly nonlinearly up to the
peak value RI with the corresponding e®ective separation " 0 . This is the point which
correlates with delamination initiation. Beyond this point a distinct softening
behavior can be observed and the e®ective traction approaches zero at some point.
The point at which the e®ective traction t 0 is interpreted as the point of complete
decohesion. Following for example Ref. 33, the interface unloads linearly to the
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 695
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

Fig. 2. E®ective traction
separation law.

origin in a partly damaged state. The cohesive free energy which corresponds to the
e®ective traction
separation law in Fig. 2 has the form
¼ ^ 1 ð"Þ þ ^ 2 ð½½u3Þ:
 ð3:15Þ
The ¯rst term characterizes the decohesion process and the second term accounts for
contact. The two parts of the cohesive free energy read
h  " 
"=" 0 i 1
1 ¼ 1 ð"Þ ¼ eRI " 1
1 þ 0 e
^ 0
; ^ ½u3Þ
2 ¼ 2 ð½  ¼ Kh
½½u3i
 2 ; ð3:16Þ
" 2
where e ¼ exp(1) is the Eulerian number and K is the penalty contact sti®ness.
According to (3.9), the tractions are obtained as the partial derivatives of the
cohesive free energy with respect to the separations, i.e.
@ @ ^ 1 ð"Þ @" @ ^ 2 ð½½u3Þ

t¼ ¼ þ ¼ ½t1 ; t2 ; t3  T ; ð3:17Þ
@½½u @" @½½u @½½u
where the traction vector components t1 ; t2 ; t3 are the tractions in the directions of
the local basis vectors A1 ; A2 ; A3 , respectively. By de¯ning  0 ¼ @  =@", the e®ective
traction vector can be formulated as
0 "
"=" 0
t¼ 1 ¼ eRI e : ð3:18Þ
"0
Substitution of (3.18) into (3.17) yields


t t
t ¼ C½½u þ KC c½u
½ ¼ C þ KC c ½u
½  ¼ D½½u; ð3:19Þ
" "

where the material matrix D and the compression identi¯er matrix C c are de¯ned by
2 3
0 0 0
6 7
t 60 0 0 7
D ¼ C þ KC c ; C c :¼ 6 6
7:
7 ð3:20Þ
" 4  5
h
½½u3i
0 0

½½u3
 696 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

It can be shown34 that utilizing a quadratic interaction criterion of the form

ht3 i 2 t 21 þ t 22
þ ¼1 ð3:21Þ
R 2I R 2s

(see e.g. Ref. 14), the weighting parameter  in (3.11) and (3.12) becomes the ratio
between the interlaminar shear strength and the mode I interlaminar tensile
strength, i.e.  ¼ Rs =RI . Up to now the equations hold for loading. In the case of
un- or reloading, i.e. F ¼ 0 and "  , we apply linear elastic response.33 In this
event the e®ective traction reads
t^ð" ¼ Þ
t¼ ": ð3:22Þ
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]


According to (3.9), the traction linearization matrix is de¯ned by the partial
derivatives of the tractions with respect to the separations, so we obtain
 
@t 00
0
C¼ ¼DþF 1
1

2 ðC½½uÞðC½½uÞ T ; ð3:23Þ
@½½u 

with 0 ¼ t from (3.18), D from (3.20), and 00 1 de¯ned by

1
 "  eRI
"=" 0
00
1 ¼ 1
0 e : ð3:24Þ
" "0
Evidently, the second term of (3.23) is applied only in the case of loading. In the
event of un- or reloading, the linearization matrix is C ¼ D, i.e. the cohesive law is
linear elastic.
In order to bene¯t from fracture mechanics approaches, the area under the
e®ective traction
separation law is equal to the mixed mode fracture toughness Gc .
Utilizing this condition yields
Gc ¼ lim ½ ^ 1 ð"Þ ¼ eRI " 0 ; ð3:25Þ
"!1

an expression which relates the fracture toughness to the e®ective separation and
traction at the delamination onset, " 0 and RI , respectively. Since the fracture
toughness and the strength are physically meaningful material properties, it may
rather be of interest to relate the dependent magnitude " 0 to the independent
material properties, i.e.
Gc
"0 ¼ : ð3:26Þ
eRI
The penalty sti®ness is governed by the postulation of initially the same response in
pure mode I tension and compression. It has been shown34 that in this case the
penalty sti®ness is de¯ned by
ðeRI Þ 2
K¼ ; ð3:27Þ
Gc
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 697
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

an expression which, except for the Eulerian number, includes only independent
material properties.

4. Validation Example
The validation example presented in this section is the simulation of axial com-
pression tests of single-sti®ener panels. The corresponding experiments are reported
in Ref. 24. Two specimen con¯gurations have been tested, an intact version and a
predamaged version. In the predamaged version a Te°on layer has been inserted
between skin and sti®ener in the middle of the specimen in order to observe propa-
gation of this initial debond. The geometry of the specimens and the layups of the
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

skin and the sti®ener blade are depicted in Fig. 3. For an accurate load introduction,
a 50-mm-long potting was applied at both ends. The engineering constants utilized
for the simulations are listed in Table 2. The intra- and interlaminar strength
parameters as well as the interlaminar fracture toughness are given in Table 3.

Fig. 3. Geometry of the single-sti®ener panels (P — potting; TI — te°on insert) and layups of the skin
and the stringer blade.

Table 2. Elastic constants (moduli in N=mm 2 ).

E1 E2
12 [
] G12 ¼ G13 G23

147,300 11,800 0.3 6000 6000

Table 3. Fracture parameters (strengths in N=mm 2 ).

R tjj R cjj R t? R c? Rjj? R?? RI RII Gc (kJ/m 2 )

2379 1365 39 170 102 78 4 20 0.980

 698 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

During the experiments no degradation prior to ¯nal collapse could be observed in

the case of the intact panel. Although after collapse all failure modes have been
present (¯ber rupture, matrix cracking, delamination), we assume that intralaminar
damage was the major contributor which subsequently triggered delamination.
Thus, it does not seem worthwhile to insert interface elements. Final failure was due
to massive damage in the sti®ener near the pottings. In the predamaged version there
was a certain degree of degradation in terms of propagation of the initial defect.
Interface elements have thus been inserted next to the predamage over a length of
50 mm at each side. However, ¯nal collapse also occurred due to intralaminar damage.
The main location of the damage was in the sti®ener in the center of the specimen.
The FE mesh of the intact specimen consists of 5-mm-long elements in the longi-
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

tudinal direction (10 þ 60 þ 10 elements). In the transverse direction the element size
is chosen in such a way that approximately quadratic elements are generated. Thus,
four elements in each stringer °ange and 16 elements in the skin are used. The stringer
blade consists of four elements in its height direction. The FE mesh of the predamaged
version is edited in such a way that the number of elements in the longitudinal
direction is increased in the region of the predamage (in order to capture the right local
buckling shape) and in the regions where interface elements are inserted (in order to
obtain convergence). In these regions the element size is reduced to 2 mm. The shell
reference planes are located on the interior surface of the skin.
In the potting regions all degrees of freedom are ¯xed except for the longitudinal
displacement, which is ¯xed only on the bottom edge. The load is introduced via a
displacement-controlled arclength method.
On the left-hand side of Fig. 4 the applied load is plotted versus the axial
shortening for the intact version. The experimental response is linear elastic up to a
shortening of about 1 mm. Then local skin buckling starts, which is expressed by a
very mild kink in the load-shortening graph. At a shortening of about 1.8 mm, global

60 40
experiment experiment
50
current current
30
Applied Load [kN]

Applied Load [kN]

40

30 20

20
10
10

0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2
Axial Shortening [mm] Axial Shortening [mm]

Fig. 4. Load-shortening curves of the intact (left) and the predamaged (right) single-sti®ener compression
test.
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 699
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

stringer-based buckling starts, accompanied by ¯nal failure. It can be seen that the
simulation agrees well in sti®ness, onset of local skin buckling, global load-shortening
behavior, and collapse load. Only the shortening at ¯nal failure is slightly over-
estimated, but the di®erence from the experiment is negligibly small.
The graphs on the right-hand side of Fig. 4 plot the applied load versus the axial
shortening for the predamaged specimens. Also here the initial experimental response
is linear elastic. At a shortening of about 1 mm, local skin buckling starts. At the
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

Fig. 5. Contour plots of the intact con¯guration: RFIs plotted on the deformed shape at a shortening of
2.1 mm (M — matrix cracking; FM — ¯ber
matrix shear failure; F — ¯ber fracture).
 700 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

Fig. 6. Contour plots of the predamaged con¯guration: RFIs plotted on the deformed shape at a
shortening of 1.4 mm (M — matrix cracking; FM — ¯ber-matrix shear failure; F — ¯ber fracture).

same time the initial defect starts to grow. At a shortening of 1.4
1.6 mm, depending
on the specimen number, global stringer-based buckling and ¯nal failure occur
simultaneously. The propagation of the initial skin
stringer debond is expressed by
a distinctive degradation in the load-shortening behavior. Also for the predamaged
version the simulation agrees very well. The initial sti®ness is in between the test results.
Local skin buckling and debond growth also start at a shortening of approximately
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 701
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

1 mm. Then degradation can be observed, followed by ¯nal failure when global
stringer-based buckling starts at a shortening of about 1.4 mm.
Figures 5 and 6 show contours of the RFIs for the three intralaminar failure
modes after collapse, plotted on the deformed con¯gurations of the intact and the
predamaged specimens. It can be observed that the most critical damage is in the
stringer at the potting ends in the intact version and in the stringers in the middle of
the specimen in the predamaged version. This agrees very well with the experimental
observations. The most severe damage is matrix cracking in the intact version
(RFI ¼ 0:89) and ¯ber fracture in the predamaged version (RFI ¼ 0:93). However,
¯ber fracture has the most catastrophic e®ect and leads to collapse of both panels.
The ¯nal debond state can also be observed in the predamaged version, as well as the
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

buckling shapes of the specimens.

5. Conclusions
In this paper a damage model for advanced composites has been derived which is
capable of simulating the postbuckling response of composite structures including
intra- as well as interlaminar damage initiation and growth. For this purpose a
successive ply failure model has been implemented in a geometrically nonlinear shell
element formulation. The ply failure model is able to distinguish between ¯ber
fracture, matrix cracking, and combined ¯ber/matrix shear failure. This feature is
obtained by application of the extended Hashin failure criteria from Ref. 4. A ply
discount model has been applied for the degradation of the elastic behavior of
damaged plies in order to obtain postdamaged solutions. This degradation model in
this work reduces directly the entries of the elasticity matrix instead of the engin-
eering constants which stands in contrast to Refs. 4 and 6. For the description of
delamination the formulation of an interface element has been derived in which a
special cohesive law has been implemented. The cohesive law is governed by smooth
and kinkfree traction
separation relations according to the universal binding
law from Ref. 22. It further accounts for contact by means of a penalty contact
formulation which avoids the interpenetration of the crack faces. Both the intra- and
the interlaminar damage models are history-dependent, leading to nonrecurring
sti®ness degradation. Both an intact and a predamaged con¯guration of a single-
sti®ener panel under axial compression have been utilized for validation purposes.
The present model could correctly predict the available experimental results.
However, more experimental results, for more complex composite structures, may be
needed before ¯nal proof of the model's versatility.

Acknowledgments
This work is supported by the European Commission, Priority Aeronautics and
Space, Contract AST3-CT-2003-502723. The ¯nancial support is greatefully
 702 C. Balzani & W. Wagner
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

acknowledged. The authors also thank the consortium of the COCOMAT project for
successful cooperation and fruitful discussions.

References
1. COCOMAT homepage, [Link]
2. R. Degenhardt, R. Rolfes, R. Zimmermann and K. Rohwer, COCOMAT — improved
material exploitation of composite airframe structures by accurate simulation of post-
buckling and collapse, Compos. Struct. 73(2) (2006) 175
178.
3. Z. Hashin, Failure criteria for unidirectional ¯ber composites, J. Appl. Mech. 47(2) (1980)
329
334.
4. V. K. Goyal, N. R. Jaunky, E. R. Johnson and D. R. Ambur, Intralaminar and inter-
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

laminar progressive failure analyses of composite panels with circular cutouts, Compos.
Struct. 64(1) (2004) 91
105.
5. A. C. Ori¯ci, Degradation model for the collapse analysis of composite aerospace struc-
tures. Ph.D. thesis, RMIT University, Melbourne, Australia (2007).
6. A. C. Ori¯ci, R. S. Thomson, I. Herszberg, T. Weller, R. Degenhardt and J. Bayandor, An
analysis methodology for failure in postbuckling skin
sti®ener interfaces, Compos.
Struct. 86(1
3) (2008) 186
193.
7. P. Maimí, P. P. Camanho, J. A. Mayugo and C. G. D
avila, A continuum damage model
for composite laminates: Part I — constitutive model, Mech. Mater. 39(10) (2007)
897
908.
8. P. Maimí, P. P. Camanho, J. A. Mayugo and C. G. D
avila, A continuum damage model
for composite laminates: Part II — computational implementation and validation, Mech.
Mater. 39(10) (2007) 909
919.
9. G. Z. Voyiadjis, P. I. Kattan and Z. N. Taqieddin, Continuum approach to damage
mechanics of composite materials with fabric tensor, Int. J. Dam. Mech. 16(3) (2007)
301
329.
10. A. Needleman, An analysis of tensile decohesion along an interface, J. Mech. Phys. Solids
38(3) (1990) 289
324.
11. J. C. J. Schellekens and R. de Borst, Numerical simulation of free edge delamination in
graphite-epoxy laminates under uniaxial tension, in Proc. 6th Int. Conf. Compos. Struct.
(Paisley College of Technology, UK; Sep. 1991), pp. 647
657.
12. O. Allix and P. Ladevèze, Interlaminar interface modeling for the prediction of delami-
nation, Compos. Struct. 22(4) (1992) 235
242.
13. M. Cris¯eld, Y. Mi, G. A. O. Davies and H. B. Hellweg, Finite element methods and the
progressive failure-modelling of composite structures, in Computational Plasticity —
Fundamentals and Applications: Proc. 5th Int. Conf. Comput. Plast., eds. D. R. J. Owen,
E. Oñate and E. Hinton (CIMNE, Barcelona, Spain, 1997), pp. 239
254.
14. P. P. Camanho and C. G. D
avila, Mixed-mode decohesion ¯nite elements for the simu-
lation of delamination in composite materials. Technical Memorandum TM-2002-211737,
NASA, Langley Research Center, Hampton, Virginia, USA (2002).
15. D. S. Dugdale, Yielding of steel sheets containing slits, J. Mech. Phys. Solids 8(2) (1960)
100
104.
16. G. I. Barenblatt, The mathematical theory of equilibrium cracks in brittle fracture, Adv.
Appl. Mech. 7 (1962) 55
129.
17. Y. Mi, M. A. Cris¯eld, G. A. O. Davies and H. B. Hellweg, Progressive delamination using
interface elements, J. Compos. Mater. 32(14) (1998) 1246
1272.
 Numerical Treatment of Damage Propagation in Axially Compressed Composite Airframe Panels 703
by GERMAN NATIONAL LIBRARY OF SCIENCE AND TECHNOLOGY on 06/27/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

18. J. P. Gonçalves, M. F. de Moura, P. T. de Castro and A. T. Marques, Interface element

including point-to-surface constraints for three-dimensional problems with damage
propagation, Eng. Comput. 17(1) (2000) 28
47.
19. G. Alfano and M. A. Cris¯eld, Finite element interface models for the delamination
analysis of laminated composites: mechanical and computational issues, Int. J. Numer.
Meth. Eng. 50 (2001) 1701
1736.
20. A. de Andr
es, J. L. P
erez and M. Ortiz, Elastoplastic ¯nite element analysis of three-
dimensional fatigue crack growth in aluminum shafts subjected to axial loading, Int. J.
Sol. Struct. 36(15) (1999) 2231
2258.
21. M. Ortiz and A. Pandol¯, Finite-deformation irreversible cohesive elements for
three-dimensional crack-propagation analysis, Int. J. Numer. Meth. Eng. 44(9) (1999)
1267
1282.
22. J. H. Rose, J. Ferrante and J. R. Smith, Universal binding energy curves for metals and
Int. J. Str. Stab. Dyn. 2010.10:683-703. Downloaded from [Link]

bimetallic interfaces, Phys. Rev. Lett. 47(9) (1981) 675
678.

23. C. Balzani and W. Wagner, An interface element for the simulation of delamination in
unidirectional ¯ber-reinforced composite laminates, Eng. Fract. Mech. 75(9) (2008)
2597
2615.
24. I. O. de Zarate Alberdi and F. Rebollo, Small degradation specimen test results —
contribution to deliverable D07 (database of degradation models). COCOMAT Technical
Report, AERnnova (2006).
25. E. N. Dvorkin and K.-J. Bathe, A continuum mechanics based four-node shell element for
general non-linear analysis, Eng. Comput. 1(1) (1984) 77
88.
26. L. M. Kachanov, On the creep fracture time, Izv. Akad. Nauk USSR Otd. Tech. 8 (1958)
26
31 (in Russian).
27. D. Krajcinovic, Constitutive equations for damaging materials, J. Appl. Mech. 50 (1983)
355
360.
28. J. Lemaitre, A Course on Damage Mechanics (Springer-Verlag, Berlin, Heidelberg,
New York, 1996).
29. W. A. M. Brekelmans and J. H. P. de Vree, Reduction of mesh sensitivity in continuum
damage mechanics, Acta Mechanica 110 (1995) 49
56.
30. R. de Borst and J. G. Rots, Occurrence of spurious mechanisms in computation of strain-
softening solids, Eng. Comput. 6(4) (1989) 272
280.
31. J. C. J. Schellekens, computational strategies for composite structures. Ph.D. thesis,
Technical University Delft (1992).
32. J. C. J. Schellekens and R. de Borst, On the numerical integration of interface elements,
Int. J. Numer. Meth. Eng. 36 (1993) 43
66.
33. G. T. Camacho and M. Ortiz, Computational modelling of impact damage in brittle
materials, Int. J. Sol. Struct. 33(20
22) (1996) 2899
2938.
34. W. Wagner and C. Balzani, Simulation of delamination in stringer-sti®ened ¯ber-
reinforced composite shells, Comput. Struct. 86(9) (2008) 930
939.