Numerical Simulation of

Mixed-mode Progressive
Delamination in Composite Materials

P. P. CAMANHO1,*,y, C. G. DAVILA2 AND M. F. DE MOURA1

1
DEMEGI, Faculdade de Engenharia, Universidade do Porto,
4200-465, Porto- Portugal
2
Analytical & Computational Methods Branch,
NASA Langley Research Center, Hampton,
VA 23681-2199, USA

(Received June 14, 2002)

(Revised January 14, 2003)

ABSTRACT: A new decohesion element with the capability of dealing with crack
propagation under mixed-mode loading is proposed and demonstrated. The element
is used at the interface between solid finite elements to model the initiation and
non-self-similar growth of delaminations in composite materials. A single relative
displacement-based damage parameter is applied in a softening law to track the
damage state of the interface and to prevent the restoration of the cohesive state during
unloading. The softening law is applied in the three-parameter Benzeggagh-Kenane
mode interaction criterion to predict mixed-mode delamination propagation. To
demonstrate the accuracy of the predictions, steady-state delamination growth is
simulated for quasi-static loading of various single mode and mixed-mode
delamination test specimens and the results are compared with experimental data.

KEY WORDS: delamination, interlaminar damage, decohesion elements.

INTRODUCTION

NTERLAMINAR DAMAGE (DELAMINATION) is one of the predominant forms of failure in

I many laminated composites systems, especially when there is no reinforcement in the
thickness direction. Delamination as a result of impact or a manufacturing defect can
cause a significant reduction in the compressive load-carrying capacity of a structure. The
stress gradients that occur near geometric discontinuities such as ply drop-offs, stiffener
terminations and flanges, bonded and bolted joints, and access holes promote

*Author to whom correspondence should be addressed.

y
Visiting Scientist, Mechanics & Durability Branch, NASA Langley Research Center, Hampton, VA
23681-2199, U.S.A.

Journal of COMPOSITE MATERIALS, Vol. 37, No. 16/2003 1415

0021-9983/03/16 141524 $10.00/0 DOI: 10.1177/002199803034505
2003 Sage Publications

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1416 P. P. CAMANHO ET AL.

delamination initiation, trigger intraply damage mechanisms, and may cause a significant
loss of structural integrity.
The fracture process in high performance composite laminates is quite complex, involving
not only delamination, but also intralaminar damage mechanisms (e.g. transverse matrix
cracking, fibre fracture). For effective predictive capabilities, failure analysis tools for the
different failure modes are required.
The simulation of delamination in composites is usually divided into delamination
initiation and delamination propagation. Delamination initiation analyses are usually
based on the point or average stress criteria proposed by Whitney and Nuismer [1] for
notched laminates. Kim and Soni [2] used the average stress failure criterion [1] to predict
the delamination onset load in carbon fibre reinforced plastics under in-plane tensile and
compressive loading, in which the tensile interlaminar normal stress,
zz, is predominant.
Good agreement between the predictions and the experimental results were obtained using
one ply thickness as the characteristic distance. When the interlaminar shear stresses,
xz
and
yz, are also present, criteria such as the quadratic interaction of the interlaminar
stresses in conjunction with a characteristic distance are normally used [3,4]. Such a
procedure has been used in the prediction of delamination onset loads in composite bolted
joints [3], and in post buckled dropped-ply laminates [4]. The characteristic distance is an
averaging length that is a function of geometry and material properties, so its
determination always requires extensive testing.
Most analyses of delamination growth are based on the Fracture Mechanics approach,
and evaluate the energy release rate G [514]. Variational methods have been used
by Schoeppner and Pagano [5] to describe stress fields in the vicinity of cracks, and
to obtain accurate predictions of the energy release rate for cross-ply laminates. Sun
and Jih [6] have investigated Modes I and II energy release rates for interfacial
cracks between dissimilar isotropic solids. The energy release rate approach was
later extended by Sun and Manoharan [7] to analyse interfacial cracks between two
orthotropic solids.
An alternative method to evaluate the energy release rate is the virtual crack closure
technique (VCCT), proposed by Rybicki and Kanninen [15]. The VCCT technique is
based on the assumption that when a crack extends by a small amount, the energy released
in the process is equal to the work required to close the crack to its original length. The
Modes I, II, and III components of the energy release rate, GI , GII and GIII respectively,
can then be computed from the nodal forces and displacements obtained from the solution
of a finite element model, assuming self-similar delamination growth. The approach is
computationally effective since the energy release rate can be obtained from only one
analysis. Although valuable information concerning the onset and the stability of
delamination can be obtained using the VCCT, its use in the simulation of delamination
growth may require complex moving mesh techniques to advance the crack front when the
local energy release rate reaches a critical value [16]. Furthermore, an initial delamination
must be defined and, for certain geometries and load cases, the location of the
delamination front might be difficult to determine.
The use of decohesion elements placed at the interfaces between laminae can overcome
some of the above difficulties. Decohesion elements are based on a DudgaleBarenblatt
cohesive zone approach [17,18], which can be related to Griffiths theory of fracture when
the cohesive zone size is negligible when compared with characteristic dimensions,
regardless of the constitutive equation [19]. These elements use failure criteria that
combine aspects of strength-based analysis to predict the onset of the softening process at

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1417

the interface between laminae, and Fracture Mechanics to predict delamination

propagation. The main advantage of the use of decohesion elements is the capability to
predict both onset and propagation of delamination without previous knowledge of the
crack location and propagation direction. Therefore, non-self-similar delamination
growth, where the delamination front changes its shape throughout the loading history,
can be predicted.
Decohesion elements can be divided into two main groups: continuous interface
elements and point decohesion elements. Different types of continuous decohesion
elements have been proposed, ranging from zero-thickness volumetric elements connecting
solid elements [20,21], finite-thickness volumetric elements connecting shell elements
[22], and line elements [23,24]. Point decohesion elements are identical to non-linear
spring elements connecting nodes [25,26]. A common feature of most of the previously
developed decohesion elements is the absence of an interaction criterion for the prediction
of softening onset under mixed-mode loading, and the use of linear or quadratic
interaction criteria of the energy release rate components for the prediction of
delamination propagation. However, experimental evidence shows that for some resins
(e.g. epoxies) linear or quadratic interaction criteria are unable to express the dependence
of the fracture toughness on the mode ratio [27]. Under mixed-mode loading conditions,
the propagation of delamination should be predicted using physically sound criteria,
able to predict the variation of the mixed-mode fracture toughness as a function of the
mode ratio.
The objective of the current work is to develop zero-thickness volumetric decohesion
elements able to capture delamination onset and growth under mixed-mode loading
conditions in composite structural components. A quadratic interaction between the
tractions is proposed to predict softening onset. A criterion able to capture the mixed-
mode fracture toughness under different mode ratios is used to predict delamination
propagation. The accuracy of the proposed formulation is assessed by simulating
delamination propagation in double cantilever beam (DCB), end-notch flexure (ENF) and
mixed-mode bending (MMB) test specimens, and comparing the predictions with
experimental data.

DECOHESION ELEMENT FORMULATION

Element Kinematics

The zero-thickness decohesion element with 8-nodes shown in Figure 1 is proposed to

simulate the resin-rich layer connecting the laminae of a composite laminate. The
constitutive equation of zero-thickness decohesion elements is established in terms of
relative displacements (also called displacement discontinuities [28]) and tractions
across the interface.
The definition of the relative displacements for an element with a general orientation in
space is obtained using a procedure based on the works of Ahmad et al. [29] and Beer [30].
The vector defining the relative displacement in global coordinates,
, can be obtained as
a function of the displacement of the points located on the top and bottom surface of the
element, u 
i and ui respectively:

i u 
i  ui 1

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1418 P. P. CAMANHO ET AL.

Figure 1. Zero-thickness decohesion element.

The top and bottom displacements are obtained as:

u
i Nk uki , k 2 ftop nodesg 2
u 
i Nk uki , k 2 fbottom nodesg 3

where u 
ki , uki are the displacements in the i direction of the k top and bottom nodes of the
element, respectively. Nk are standard Lagrangian shape functions. By defining:


Nk , k 2 ftop nodesg
N k 4
Nk , k 2 fbottom nodesg

then the relative displacements defined in (1) can be written:

i N k uki 5

For a general element shape and alignment, the normal and tangential relative
displacements must be determined in local coordinates. The tangential plane at a given
point is spanned by two vectors, v and v, obtained by differentiating the global position
vector with respect to the natural (local) coordinates:

vi xi,  , vi xi,  6

Defining an isoparametric element, the global position vector is obtained as:

xi Nk xki 7

The components of the two vectors that define the tangential plane can be written as:

vi Nk xki ,  Nk,  xki 8

vi Nk xki ,  Nk,  xki 9

Although v and v are, generally, not orthogonal to each other, their vector product
defines a surface normal. Therefore, the local normal coordinate vector is obtained as:

v v
v kv
v k1 10

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1419

The tangential coordinates are then obtained as:

vs v kv k1 11
vt vn
v s 12

The components of vn, vs, and vt represent the direction cosines of the local coordinate
system to the global coordinate system, thus defining the transformation tensor si .
Using (5), the relative displacements can then be obtained in local coordinates as:

s si
i si N k uki Bsik uki 13

The constitutive operator of the decohesion element, Dsr, relates the element tractions,  s,
to the element relative displacements, r:

s sr Dsr r 14

where sr is the Kronecker delta.

The coefficients Dsr of the element constitutive operator can be used to simulate
elaborate mechanical behaviours, including the mechanics of interfacial decohesion and
crack propagation, and will be discussed later1. An important characteristic of the
proposed method is that, unlike thin continuum elements (degenerate continuum
elements), the stiffness of the interface before softening onset, referred here as the
penalty stiffness, is not a function of the discretization, but is defined by the coefficients
Dsr. Daudeville et al. [31] have proposed the definition of the penalty stiffness as a function
of the interface thickness, t, and elastic moduli of the interface (E3, G13
and G23) as: D33 E3/t, D11 2G13/t, D22 2G23/t.
Considering  as the mid-surface area in the deformed configuration, the decohesion
element stiffness matrix and internal load vector can be obtained from the principle of
virtual work:
Z
ds s d  fki duki 0 15


From (13) and for a geometrically non-linear problem [32]:

Z  
@Bspj
ujp Bsik s d  fki 0 16
 @uki

The first term of (16) represents the decohesion element internal load vector. From (13)
and (14):
Z  
@Bspj
sr Dsr Brvz ujp Bsik duzv fki 17
 @uki
Kkizv uzv fki 18

1
For simple contact elements Dsr are the penalty parameters: Dsr 0 if 3 > 0, and D33 K if 3  0.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1420 P. P. CAMANHO ET AL.

The decohesion element stiffness matrix is then:

Z  
 @Bspj
Kkizv sr Dsr Brvz ujp Bsik d 19
 @uki

where the deformed differential area of the mid-surface of the decohesion element, d, is
related with the undeformed differential area of the mid-surface of the decohesion element,
d0, as: d kv
v kd0.
Care must be exercised in the choice of the integration scheme used to obtain the
stiffness matrix and the internal load vector. Several investigations have shown the
superiority of using NewtonCotes integration techniques over traditional Gaussian
integration techniques in decohesion elements [3337]. Using eigenmode analysis of the
element stiffness matrices it has been shown that Gaussian integration can cause undesired
spurious oscillations of the traction field when large traction gradients are present over an
element [35,36].
Another relevant issue related with the integration scheme is the number of integration
points used. Analyses of problems involving crack propagation and softening behaviour
have shown that the use of full integration was superior to the use of reduced integration
schemes [38]. However, Alfano and Crisfield [39] have shown that for linear 4-node
decohesion elements increasing the number of integration points from 2 (corresponding to
full integration) to 20 results in an increase of spurious oscillations in the loaddisplace-
ment curve and, consequently, in a less robust solution algorithm. For the above reasons,
NewtonCotes full integration [40] is used in the decohesion element proposed here.

Proposed Constitutive Equation

SINGLE-MODE DELAMINATION
The need for an appropriate constitutive equation in the formulation of the decohesion
element is fundamental for an accurate simulation of the interlaminar cracking process. It
is considered that there is a process zone or cohesive zone ahead of the delamination
tip. Figure 2 represents the cohesive zone in specimens loaded in pure Mode II

(a) (b)

Figure 2. Pure mode constitutive equations: (a) Mode II or Mode III; (b) Mode I.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1421

(Figure 2(a)) and in pure Mode I (Figure 2(b)). Figure 2 also illustrates the constitutive
behaviour for pure Modes I, II, and III loading. For pure Modes I, II or III loading the bi-
linear softening constitutive behaviour represented in Figure 2 is used. A high initial
stiffness ( penalty stiffness, K ) is used to hold the top and bottom faces of the
decohesion element together in the linear elastic range (Point 1 in Figure 2). For pure
Mode I, II or III loading, after the interfacial normal or shear tractions attain their
respective interlaminar tensile or shear strengths (Point 2 in Figure 2), the stiffnesses are
gradually reduced to zero. The onset displacements are obtained as: o3 N=K, o2
S=K and o1 T=K, where N is the interlaminar tensile strength, and S and T are the
interlaminar shear strengths.
The concept of cohesive zone was initially proposed by Barenblatt [18] and using such a
concept the singularity at the crack tip is removed. Physically, the cohesive zone represents
the coalescence of crazes in the resin rich layer located at the delamination tip and reflects
the way by which the material loses load-carrying capacity. Ungsuwarungsri and Knauss
[41] considered that if the size of the process zone is narrow compared to the size of the
specimen, a softening material behaviour confined to a thin layer adjacent to the crack
plane is a realistic scheme for the simulation of the cohesive zone and crack growth.
Needleman [42] considered that cohesive zone models are particularly attractive when
interfacial strengths are relatively low compared with that of the adjoining material, as is
the case in composite laminates, because the high stress gradients associated with cracks in
homogeneous bodies do not develop, and standard finite elements can be used.
It has been shown that cohesive zone approaches can be related to Griffiths theory of
fracture if the area under the tractionrelative displacement relation is equal to the
corresponding fracture toughness [43] (see Figure 2), regardless of its shape. Furthermore,
Alfano and Crisfield [39] have shown that when the relative displacements o3 and 3f
shown in Figure 2 are coincident (corresponding to a sudden load drop to zero) a perfectly
brittle fracture is simulated. A model for brittle fracture must be able to capture the high
stress gradients at the crack tip with sufficiently fine mesh densities or singular elements.
The area under the tractionrelative displacement curves, corresponding to the
respective (Mode I, II or III) fracture toughness (GIC, GIIC and GIIIC respectively), defines
the final relative displacements,  f3 ,  f2 and  f1 corresponding to complete decohesion:

Z  f3
3 d3 GIC 20
0
Z f2
2 d2 GIIC 21
0
Z  f1
1 d1 GIIIC 22
0

The final displacements are then obtained as:  f3 2GIC =N,  f2 2GIIC =S and  f1
2GIIIC =T:
Once a crack is unable to transfer any further load (Points 4 and 5 in Figure 2), all the
penalty stiffnesses revert to zero. However, it is necessary to avoid the interpenetration of
the crack faces. The contact problem is addressed by reapplying the normal penalty
stiffness when interpenetration is detected.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1422 P. P. CAMANHO ET AL.

In order to formulate the complete constitutive equation, the unloading behaviour must
be defined. It is considered that a softening point unloads towards the origin, as shown in
Figure 2. Using the Macauley operator defined as:


0, x<0
hxi 23
x, x0

the loading condition can be formulated in terms of a variable defined as the maximum
relative displacement, max, suffered by the point:

Mode II or III: max

i maxfmax
i , ji jg, i 1, 2 24
Mode I: max
i maxfmax
3 , 3 g, with max
3 0 25

and using a loading function, F, defined as:


ji j  max
i
Mode II or III: Fji j  max
i , i 1, 2 26
ji j  max
i
 
max 3  max
3
Mode I: F3  3 , with max
3 0 27
3  max
3

Using max in the constitutive equation, the irreversibility of damage is taken into account.
This is shown in Figure 2: if the relative displacement decreases, the point unloads
elastically towards the origin with a reduced, secant, stiffness (point 3 in Figure 2).
The irreversible, bi-linear, softening constitutive behaviour for single-mode loading
shown in Figure 2 can be defined as [23,39,44]:
8
>
< Ki , max
i  0i
i 1  di Ki , 0i < max <  fi 28
>
:
i
0, max
i  fi
 fi max
i  0i
di , i 1, 2, 3; di 2 0, 1 29
max
i  fi  0i

In order to avoid interpenetration of the crack faces, the following condition is introduced:

3 K3 , 3  0 30

The properties required to define the interfacial behaviour are the penalty stiffness , K, the
corresponding fracture toughnesses, GIC, GIIC and GIIIC, and the corresponding
interlaminar normal tensile or shear strengths, N, S or T respectively.

MIXED-MODE DELAMINATION
In structural applications of composites, delamination growth is likely to occur under
mixed-mode loading. Therefore, a general formulation for decohesion elements dealing
with mixed-mode delamination onset and propagation is also required.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1423

Damage onset prediction

Under pure Mode I, II or III loading, the onset of damage, corresponding to the
nucleation of microcavities in brittle solids [45], or the formation of crazes in polymers
[41], can be determined simply by comparing the traction components with their respective
allowables. However, under mixed-mode loading damage onset and the corresponding
softening behaviour may occur before any of the traction components involved reach their
respective allowables, which is an issue that is usually neglected in the formulation of
decohesion elements. Cui et al. [46] have highlighted the importance of the interactions
between interlaminar stress components when predicting delamination. It was shown that
poor results are obtained using the maximum stress criterion. Therefore, a mixed-mode
criterion accounting for the effect of the interaction of the traction components in the
onset of delamination is proposed here.
It is assumed that the onset of damage can be predicted using the quadratic failure
criterion [46], considering that compressive normal tractions do not affect delamination
onset, and using the operator defined in (23):

 
h3 i 2 2 2 1 2
1 31
N S T

This criterion has been successfully used to predict the onset of delamination in previous
investigations [3,4,47].
The total mixed-mode relative displacement m is defined as:

q q
m 21 22 h3 i2 2shear h3 i2 32

where shear represents the norm of the vector defining the tangential relative displace-
ments of the element.
Using the same penalty stiffness in Modes I, II and III, the tractions before softening
onset are:

i Ki , i 1, 2, 3 33

Assuming S T, the single-mode relative displacements at softening onset are:

N
o3 34
K
S
o1 o2 oshear 35
K

For an opening displacement 3 greater than zero, the mode mixity ratio
is defined as:

shear

36
3

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1424 P. P. CAMANHO ET AL.

The mixed-mode relative displacement corresponding to the onset of softening, om , is

obtained by substituting Equations (32)(36) into (31) and solving for m, which gives:
8 s
>
< o o 1
2
o 3 1 , 3 > 0
m o1 2
o3 2 37
>
:
oshear , 3  0

Clearly, single-mode loading is a particular case of the proposed formulation, as

om o3 for
0 (Mode I), and om oshear for 3 0 (or when
! 1, shear mode).

Delamination propagation prediction

The criteria used to predict delamination propagation under mixed-mode loading
conditions are usually established in terms of the energy release rate and fracture
toughness. There are standardized test methods to obtain the Modes I and II interlaminar
fracture toughness. The Double Cantilever Beam (DCB) specimen is used for Mode I. The
End-Notched Flexure (ENF) or the End-Loaded Split (ELS) test specimens are used for
Mode II. For Mixed-modes I and II, the Mixed-Mode Bending (MMB) test specimen is
used. The Edge Crack Torsion (ECT) [48] test can be used to measure the Mode III
interlaminar fracture toughness, GIIIC. However, there is no reliable mixed-mode
delamination failure criterion incorporating Mode III because there is no mixed-mode
test method available incorporating Mode III loading. Therefore, most of the failure
criteria proposed for delamination growth are established for Mixed-modes I and II
loading only. For these reasons, and following Lis work [9,10], the concept of energy
release rate for shear loading, Gshear GII GIII, is used here.
For mixed-mode loading the dependence of the fracture toughness on mode ratio must
be accounted for in the formulation of decohesion elements. The relation between the
mixed-mode interlaminar fracture toughness and the fracture surfaces of unidirectional
laminates has been thoroughly examined using scanning electron microscope analyses
[27,49]: for epoxy composites under pure Mode I loading the fracture surface is flat
indicating cleavage fractures, whereas under pure Mode II loading the fracture surface of
epoxy composites exhibit hackles having an orientation of approximately 45 with respect
to the fibre direction. Under Mixed-modes I and II the mechanisms are more complex,
including both cleavage paths and hackles [27,49].
The most widely used criterion to predict delamination propagation under mixed-mode
loading, the power law criterion, is established in terms of an interaction between the
energy release rates [50]:
   
GI  GII 
1 38
GIC GIIC
Reeder [27] performed MMB tests to measure the Mixed-modes I and II interlaminar
fracture toughness of composites, and obtained valuable experimental data to assess the
several criteria proposed to predict delamination growth. The power law criterion
obtained from (38) with  1 was found to be suited to predict failure of thermoplastic
PEEK matrix composites because the results were comparable to the predictions made
using criteria with two additional independent parameters. However, the power law
criterion failed to accurately capture the dependence of the mixed-mode fracture
toughness on the mode ratio occurring in epoxy composites using both  1 and  2.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1425

In order to accurately account for the variation of fracture toughness as a function of

mode ratio in epoxy composites, the mixed-mode criterion proposed by Benzeggagh and
Kenane [49] is used here (BK criterion). This criterion is expressed as a function of the
Mode I and Mode II fracture toughnesses, and a parameter  obtained from MMB tests at
different mode ratios:
 
GII
GIC GIIC  GIC GC , with GT GI GII 39
GT

If Mode III loading occurs the criterion is:

 
Gshear 
GIC GIIC  GIC GC , with GT GI Gshear 40
GT

Figure 3 shows the predictions of the power law and BK criteria for composites using a
tough epoxy resin (IM7/977-2), a brittle epoxy resin (AS4/3501-6), and a thermoplastic
resin (AS4/PEEK). The figure also includes the average of the experimental results
obtained by Reeder [27] at different mode ratios (discrete points shown in Figure 3) and
the  and  values used for each material [27].
By using three parameters, GIC, GIIC and , the BK criterion represents the mixed-
mode fracture toughness over the range of mode mixities, whereas the two-parameter
power law criterion that is usually implemented in decohesion elements can lead to
inaccurate results over a large range of mode ratios. Furthermore, using the power law
criterion with  2 in epoxy based composites, values of mixed-mode fracture toughness
higher than GIIC occur in the interval 0.9  GII/GT<1.0. Experimental evidence (Figure 3)
shows that the highest mixed-mode fracture toughness of epoxy composites occurs for
GII/GT 1, i.e., Gmax
C GIIC . Therefore, using of the power law criterion with  2 in
2
GC (kJ/m )
Experimental Failure =2
data criteria
AS4/3501-6
IM7/977-2
AS4/PEEK
=0.63

=1
=1
=1.39 =2

=1.75 =2 =1

GII/GT

Figure 3. Mixed-mode fracture toughness.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1426 P. P. CAMANHO ET AL.

epoxy based composites can lead to unconservative predictions in a small range of mode
mixity ratios.
Figure 3 shows that both the power law criterion using  1, and the BK criterion
provide reasonable results for the prediction of the mixed-mode fracture toughness of the
AS4/PEEK composites. The power law criterion with  2 is clearly inadequate to predict
the mixed-mode fracture toughness of the AS4/PEEK composites.
Considering that the BK criterion is able to capture the mixed-mode fracture
toughness over a comprehensive range of mode mixities for different composite materials,
this criterion is implemented in the decohesion element for the prediction of delamination
propagation.
The energy release rates corresponding to total decohesion are obtained from:

K3mf 3mo
GI 41
2
Kshear
m
f shear o
m
Gshear 42
2

where 3mo and shear

m
o
are respectively the normal and shear relative displacements
corresponding to the onset of softening under mixed-mode loading, and 3mf and shear
m
f

are respectively the normal and shear relative displacements corresponding to the total
decohesion under mixed-mode loading.
From (32) and (36):

om mf
3mo p ; 3mf p 43
1
2 1
2

mo
mf
shear
m
o
p ; shear
m
f
p 44
1
2 1
2

Using the Equations (43) and (44) in (41) and (42) the energy release rate components
corresponding to total decohesion can be established in terms of mf , mo and
. Substituting
in the mixed-mode failure criterion, (40), and solving the equation for mf , the mixed-mode
relative displacement corresponding to total decohesion is obtained as:
8   2  
< 2 GIC GIIC  GIC
>

, 3 > 0
f
m Km
o
q 1
2 45
>
:
 f1 2  f2 2 , 3  0

Single-mode loading is a particular case of the proposed formulation, as mf  f3 for

0
(Mode I) and mf  fshear for 3 0 (or when
! 1, shear mode).

Constitutive equation for mixed-mode loading

The constitutive equation for mixed-mode loading is defined by the penalty parameter
K, the damage evolution function d, the mixed-mode relative displacements corresponding
to damage initiation and total decohesion, om and mf respectively, as:

s Dsr r , with: 46

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1427
8
>
>  sr K, max
m  o
 m
>
>
< h3 i 
sr 1  d K dK s3 , mo < max
m <  fm
Dsr 3 47
>
>
>
> h3 i
: s3 3r K, max m   fm
3

 fm max o
m  m
d , d 2 0, 1 48
max f o
m  m  m

where sr is the Kronecker delta. It is worth noticing that Equation (47) provides the
stiffness to control interpenetration of the crack faces of the decohesion element.
In order to define the loading and unloading conditions, the variable maximum
mixed-mode relative displacement, maxm , and the loading function, F, are defined as:

max
m max fmax
m , m g 49
 
max m  max
m
Fm  m 50
m  max
m

The mixed-mode softening law presented above is a single-variable response similar to the
bi-linear single-mode law illustrated in Figure 2, defined by a damage evolution law (48),
maximum mixed-mode relative displacement, (49), and loading function (50). Only one
variable, the maximum relative displacement variable maxm , is used to track the damage at
the interface. By recording the highest value attained by m, the unloading response is as
shown in Figure 2. The relative displacements for initiation and ultimate failure are
functions of the mode mixity
, the material properties, and the penalty stiffness.
The mixed-mode softening law can be illustrated in a single three-dimensional map by
representing Mode I on the 03 plane, and shear mode on the 0shear plane, as
shown in Figure 4. The triangles 0N  f3 and 0S fshear are the bi-linear response in

2
)

Figure 4. Mixed-mode softening law.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1428 P. P. CAMANHO ET AL.

Mode I and in shear mode respectively. In this three-dimensional map, any point on the
0shear 3 plane represents a mixed-mode relative displacement. The relative displace-
ment corresponding to softening onset, om , is calculated using (37), and the relative
displacement corresponding to delamination propagation,  fm , is calculated using (45).

Solution Method for the Non-linear Problem

The softening nature of the decohesion element constitutive equation causes difficulties
in obtaining a converged solution for the non-linear problem. Furthermore, high penalty
values can lead to large unbalanced forces and shoot the iteration process beyond its
radius of convergence. Crisfield et al. [51] found that when using the NewtonRaphson
method under load (with the arc-length method) or displacement control, the iterative
solutions often failed to converge. In order to obtain convergence, a line search
procedure with a negative step length was proposed. Other methods such as modified
cylindrical arc-length method [52] and the nodal crack opening displacement (COD)
control have been proposed [28,37].
The NewtonRaphson method and a line-search algorithm is used to solve the non-
linear problem. Therefore, the material tangent modelar matrix [23], DT, must be defined as:

@Dsr
DTsrzv 51
@uzv

The scalar components of the decohesion element tangent modular matrix are:
8
>
< K fm om w
2 f
srw 3 Bwvz Fm  max
m , om < max
m <  fm
DTsrzv max o
m  m  m
m 52
>
:
0, max
m  om Vmax
m   fm

The index v is a global degree of freedom, z represents node numbers, and w, s, r are local
degrees of freedom. The function srw(3) is defined as:
  
  h3 i  h3 i 
csrw 3 sr 1  s3 1  3w 3w 53
3 3

For unloading conditions, Fm  max T

m 0, and Dsrzv 0:

SIMULATION OF DELAMINATION IN FIBRE-REINFORCED COMPOSITES

The decohesion element proposed here was implemented in the ABAQUS finite element
code [53] as a user-written element subroutine (UEL). To verify the element under
different loading conditions, the DCB test, the ENF test, and MMB tests are simulated.
The DCB test consists of pure Mode I delamination. The ENF tests measure pure Mode II
interlaminar fracture toughness, and the MMB measures the interlaminar fracture
toughness under Mixed-modes I and II loading. In the absence of Mode III loading,
Gshear GII . To investigate the accuracy of the proposed formulation in the simulation of
delamination, DCB, ENF and MMB simulations are conducted for APC2/PEEK,

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1429

a thermoplastic matrix composite material, and the numerical predictions are compared
with experimental data.

Mode I, Mode II, and Mixed-modes I and II Delamination Growth for AS4/PEEK
Composite Laminates

The 8-node decohesion element developed is used to simulate DCB, ENF and MMB
tests in a 24-ply unidirectional AS4/PEEK (APC2) carbon fibre reinforced composite. The
main advantages of the MMB test method [54,57] are the possibility of using virtually the
same specimen configuration as for Mode I tests, and the capability of obtaining different
mixed-mode ratios, ranging from pure Modes I to II, by changing the length c of the
loading lever shown in Figure 5.
The experimental tests were performed by Reeder and Crews [55] at different GII/GT
ratios (0.0, 0.2, 0.5, 0.8 and 1.0), ranging from pure Mode I loading to pure Mode II
loading2. The specimens are 102 mm-long, 25.4 mm-wide, with two 1.56 mm-thick arms.
The material properties are shown in Table 1, and a penalty stiffness K 106 N/mm3 is
used. The initial delamination length of the specimens (a0) and the mixed-mode fracture
toughness obtained experimentally [55] are shown in Table 2. It is worth noticing that the
initial delamination length for the ENF test specimen, a0 39.3 mm, leads to stable
propagation of the delamination. For the material and geometry used here, it was shown
by Reeder and Crews [55] that for initial delamination lengths smaller than 35 mm
unstable propagation occurs, because G exceeds GC as the delamination grows.
Models using 150 decohesion elements along the length of the specimens, and 4
decohesion elements along the width, are created to simulate the ENF and MMB test

Figure 5. MMB test specimen.

Table 1. Properties for AS4/PEEK [55,58].

E11 E22 E33 G12 G13 G23 12 13
122.7 GPa 10.1 GPa 5.5 GPa 3.7 GPa 0.25
23 GIC GIIC N S
0.45 0.969 kJ/m2 1.719 kJ/m2 80 MPa 100 MPa

2
The loaddisplacement relations experimentally determined for the different mode ratios are not reported in [55].
This information was provided by the author of the experiments [58].

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1430 P. P. CAMANHO ET AL.

cases. The initial size of the delamination is simulated by placing open decohesion
elements along the length corresponding to the initial delamination of each specimen (see
Table 2). These elements are capable of dealing with the contact conditions occurring for
Mode II or Mixed-modes I and II loading, therefore avoiding interpenetration of the
delamination faces. The model of the DCB test specimen uses 102 decohesion elements
along the length of the specimen.
The different GII/GT ratios are simulated by applying different loads at the middle and
at the end of the test specimen. The determination of the middle and end loads for each
mode ratio is presented in Appendix A. The experimental results relate the load to the
displacement of the point of application of the load P in the lever (load-point
displacement, Figure 5). Since the lever is not simulated, it is necessary to determine the
load-point displacement from the displacement at the end and at the middle of the
specimen, using the procedure described in Appendix B.
The BK criterion is used to predict delamination propagation, and the parameter
 2.284 is calculated by applying the least-squares fit procedure proposed in Appendix C
to the experimental data shown in Table 2.
Figure 6 shows the loaddisplacement relation predicted by the numerical model and
the experimental data for all the test cases simulated, and Table 3 shows the comparison
between the predicted and experimentally determined maximum loads.

Table 2. Experimental data [55,58].

GII/GT 0% (DCB) 20% 50% 80% 100% (ENF)
GC(kJ/m2) 0.969 1.103 1.131 1.376 1.719
a0 (mm) 32.9 33.7 34.1 31.4 39.3

Figure 6. Predicted and experimental loaddisplacement relation for the different mode ratios.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1431

Table 3. Experimental [55,58] and numerical maximum loads.

GII/GT (%) Pmax (experimental, N ) Pmax (predicted, N ) Error (%)
0 (DCB) 147.1 155.3 5.6
20 108.1 99.9 8.3
50 275.4 274.5 0.3
80 518.7 502.0 3.2
100 (ENF) 734.0 697.6 5.0

It can be concluded that a good agreement between the numerical predictions

and the experimental results is generally obtained. The largest difference ( 8.3%)
corresponds to the case of an MMB test specimen with GII/GT 20%. This fact is not
surprising, since the largest difference between the fracture toughness experimentally
measured and the one predicted using the BK criterion occurs for GII/GT 20% (see
Appendix C).
The numerical predictions obtained for the DCB test specimen show a more
pronounced load drop in the fracture zone than the one obtained experimentally. This
fact can be related to the increase of the Mode I fracture toughness with increasing crack
lengths. This R-curve behaviour, due to fibre bridging in the unidirectional DCB specimen,
is not taken into account in the formulation of the element. This is not a major drawback,
because in the multi-directional composite laminates typically used fibre bridging is not as
pronounced as in a unidirectional composite laminates.

CONCLUDING REMARKS

A method for the simulation of progressive delamination based on decohesion elements

was presented. Decohesion elements are placed between layers of solid elements that open
and shear in response to the loading situation. The onset of damage and the growth of
delamination can be simulated without previous knowledge about the location, the size, or
the direction of propagation of the delaminations. A softening law for mixed-mode
delamination that can be applied to any interaction criterion was proposed. The
constitutive equation proposed uses a single variable, the maximum relative displacement,
to track the damage at the interface under general loading conditions. The material
properties required to define the element constitutive equation are the interlaminar
fracture toughnesses, the penalty stiffness, and the strengths. The BK interaction law
requires additionally a material parameter  that is determined from standard
delamination tests.
Three examples were presented that test the accuracy of the method. Simulations of the
DCB and ENF test represent cases of single-mode delamination. The MMB tests were
simulated at various proportions of Modes I and II loading conditions, and the BK
criterion for delamination propagation was used. The examples analysed are in good
agreement with the test results and they indicate that the proposed formulation can predict
the strength of composite structures that exhibit progressive delamination. Although the
examples presented in this work were obtained for composite specimens containing
pre-existing delaminations, the formulation can be extended to composite structures

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1432 P. P. CAMANHO ET AL.

without any pre-existing defect. Future work will investigate the application of the method
proposed here in the simulation of the fracture of composite structural components
without any pre-existing delamination.

ACKNOWLEDGEMENTS

The authors wish to acknowledge Dr. James Reeder for providing the experimental
data, and Dr. Pedro Areias for the useful discussions. The financial support of the
Portuguese Foundation for Science and Technology (FCT) under the project POCTI/
EME/43525/2000 is gratefully acknowledged.

APPENDIX A: DETERMINATION OF MIDDLE AND

END LOAD IN MMB TESTS

The length of the lever used in the MMB experimental test, c, was obtained taking into
account the weight of the lever. Since the lever is not simulated in the numerical models,
the lengths corresponding to the different mode ratios need to be calculated.
The mode mixity ratio, , is defined as a function of the energy release rates as:

GII GI 1  
 [ 54
GI GII GII 

The relation between Modes I and II energy release rates in an MMB test specimen is
[55,59]:

 
GI 4 3c  l 2 l
, for c  55
GII 3 c l 3

From (54) and (55), the length of the lever can be obtained as a function of the mode
mixity ratio and specimen length (2l ):
s
  !
1 1
l 3 1
2 
c s
  56
1 1
3 3
2 

Table 4. Lengths of the lever.

GII/GT 20% 50% 80%
c (numerical, mm) 109.4 44.4 28.4
c (experimental, mm) 97.4 42.2 27.6

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1433

Table 4 shows the lengths of the lever for each mode mixity used in the numerical model
and in the experiments.
Neglecting the weight of the lever, the middle and end loads, Pm and Pe respectively, are
obtained as a function of the total load P as [55]:
 
cl
Pm P 57
l
c
Pe P 58
l
Therefore:

Pm c 1
59
Pe c

From (59) and (56):

p
Pm 6 31  
8 p 60
Pe 3 9 8 31  

Using (60) the relation between the load at the middle and at the end of the specimen is
obtained for the different mode ratios. The results are shown in Table 5.
The pure mode load components have been shown [55] to be given as a function of the
total load applied to the specimen by:
 
3c  l
PI P 61
4l
 
cl
PII P 62
l

From (56), (61) and (62):

 p
1   2 31   P
PI 63
13  1
 p
8 6 31   P
PII 64
3 13  1
The Modes I and II loads for each case are shown in Table 6.

Table 5. Ratio between middle and end loads.

GII/GT 20% 50% 80%
Pm/Pe 1.46 2.14 2.79

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1434 P. P. CAMANHO ET AL.

Table 6. Mode I and Mode II load components.

GII/GT 20% 50% 80%
PI 1.37P 0.41P 0.17P
PII 3.15P 1.87P 1.56P

APPENDIX B: DETERMINATION OF THE LOAD-POINT DISPLACEMENT

The information available from the MMB test relates the load to the displacement of the
load-point (Figure 5). Since the lever is not simulated in the numerical models, it is
necessary to calculate the load-point displacement using the information available from
the numerical models of the MMB test specimens.
The load-point displacement, LP , is obtained from the pure mode displacement
components, I and II , as [60]:
   
3c  l cl
LP I II 65
4l l

Using simple beam theory, Mi and Crisfield [59] calculated the Mode II displacement
component, II , as a function of the displacement at the middle of the MMB test
specimen, M :

1
II M I ; for a < l 66
4

Therefore:
    
3c  1 cl 1
LP I M  I ; for a<l 67
4l l 4

The values of I and M are computed from the numerical model. The load-point
displacement is then calculated using (67), being this procedure valid only for crack
lengths (a) smaller than half the length of the specimen (l).

APPENDIX C: EXPERIMENTAL DETERMINATION OF THE

PARAMETER  FOR THE BK DELAMINATION CRITERION

The problem consists in determining  from a set of experimental data using the
polynomial:
   
GII GII
p GIC GIIC  GIC 68
GT GT
 
A least-square fit is proposed. Considering the pair GII =GT j ; GT j as the experi-
mental data and n as the number of data points, the problem can be posed as the

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1435

Figure 7. Prediction of the mixed-mode fracture toughness using the BK criterion over the entire range of
mode ratio.

minimization of:
"   #2
X
n
GII
q GT j  GIC  GIIC  GIC 69
j1
GT j

Considering dq=d 0:
"   #   
Xn
GII GII GII
GT j  GIC  GIIC  GIC ln 0 70
j1
GT j GT j GT j

The value of  is then obtained from the solution of Equation (70). For the experimental
data used (Table 2),  2.284. The application of the BK criterion over the entire range
of mode-mixity ratios is shown in Figure 7.
It should be noticed that only one experimental point was available for each loading
condition, and that the largest difference between the BK criterion and the experimental
results occurs for GII/GT 0.2.

REFERENCES

1. Whitney, J.M. and Nuismer, R.J. (1974). Stress Fracture Criteria for Laminated Composites
Containing Stress Concentrations, Journal of Composite Materials, 8: 253265.
2. Kim, R.Y. and Soni, S.R. (1984). Experimental and Analytical Studies on the Onset of
Delamination in Laminated Composites, Journal of Composite Materials, 18: 7080.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1436 P. P. CAMANHO ET AL.

3. Camanho, P.P. and Matthews, F.L. (1999). Delamination Onset Prediction in Mechanically
Fastened Joints in Composite Laminates, Journal of Composite Materials, 33: 906927.
4. Davila, C.G. and Johnson, E.R. (1993). Analysis of Delamination Initiation in Postbuckled
Dropped-Ply Laminates, AIAA Journal, 31(4): 721727.
5. Schoeppner, G.A. and Pagano, N.J. (1998). Stress Fields and Energy Release Rates in Cross-ply
Laminates, International Journal of Solids and Structures, 11: 10251055.
6. Sun, C.T. and Jih, C.J. (1987). On Strain-Energy Release Rates for Interfacial Cracks in
Bi-Material Media, Engineering Fracture Mechanics, 28(1): 1320.
7. Sun, C.T. and Manoharan, M.G. (1989). Strain Energy Release Rates of an Interfacial Crack
Between Two Orthotropic Solids, Journal of Composite Materials, 23: 460478.
8. Krueger, R. and OBrien, T.K. (2001). A Shell/3D Modeling Technique for the Analysis of
Delaminated Composite Laminates, Composites-Part A, 32: 2544.
9. Li, J. and Sen, J.K. (2000). Analysis of Frame-to-Skin Joint Pull-Off Tests and Prediction of the
Delamination Failure, 42nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics
and Materials Conference, Seattle, WA, USA.
10. Li, J. (2000). Three-Dimensional Effects in the Prediction of Flange Delamination in Composite
Skin-Stringer Pull-Off Specimens, 15th Conference of the American Society for Composites,
Texas, USA.
11. Krueger, R., Cvitkovich, M.K., OBrien, T.K. and Minguet, P.J. (2000). Testing and Analysis of
Composite Skin/Stringer Debonding Under Multi-Axial Loading, Journal of Composite
Materials, 34(15): 12631300.
12. Koning, M., Krueger, R. and Rinderknecht, S. (1995). Numerical Simulation of Delamination
Buckling and Growth, 10th International Conference on Composite Materials (ICCM-10),
Whistler, Canada.
13. Krueger, R. (1999). A Shell/3D Modeling Technique for Delamination in Composite Laminates,
14th Conference of the American Society for Composites, Ohio, USA.
14. Krueger, R., Minguet, P.J. and OBrien, T.K. (1999). A Method for Calculating Strain
Energy Release Rates in Preliminary Design of Composite Skin/Stringer Debonding Under
Multi-Axial Loading, NASA TM-1999-209365.
15. Rybicki, E.F. and Kanninen, M.F. (1977). A Finite Element Calculation of Stress Intensity Factors
by a Modified Crack Closure Integral, Engineering Fracture Mechanics, 9: 931938.
16. Rinderknecht, S. and Kroplin, B. (1994). Calculation of Delamination Growth With Fracture and
Damage Mechanics, Recent Developments in Finite Element Analysis, CIMNE, Barcelona, Spain.
17. Dudgale, D.S. (1960). Yielding of Steel Sheets Containing Slits, Journal of Mechanics and
Physics of Solids, 8:100104.
18. Barenblatt, G.I. (1962). Mathematical Theory of Equilibrium Cracks in Brittle Failure,
Advances in Applied Mechanics, 7.
19. Camanho, P.P., Davila, C.G. and Ambur, D.R. (2001). Numerical Simulation of Delamination
Growth in Composite Materials, NASA-TP-211041.
20. de Moura, M.F., Goncalves, J.P., Marques, A.T. and de Castro, P.T. (1997). Modeling
Compression Failure After Low Velocity Impact on Laminated Composites Using Interface
Elements, Journal of Composite Materials, 31: 14621479.
21. Goyal-Singhal, V., Johnson, E.R., Davila, C.G, and Jaunky, N. (2002). An Irreversible
Constitutive Law for Modeling the Delamination Process Using Interface Elements. 43rd AIAA/
ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, Colorado,
USA.
22. Reddy Jr., E.D., Mello, F.J. and Guess, T.R. (1997). Modeling the Initiation and Growth of
Delaminations in Composite Structures, Journal of Composite Materials, 31: 812831.
23. Chen, J., Crisfield, M.A., Kinloch, A.J., Busso, E.P., Matthews, F.L., and Qiu, Y. (1999).
Predicting Progressive Delamination of Composite Material Specimens Via Interface Elements,
Mechanics of Composite Materials and Structures, 6: 301317.
24. Petrossian, Z. and Wisnom, M.R. (1998). Prediction of Delamination Initiation and Growth
From Discontinuous Plies Using Interface Elements, Composites-Part A, 29: 503515.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

Numerical Simulation of Mixed-mode Progressive Delamination 1437

25. Cui, W. and Wisnom, M.R. (1993). A Combined Stress-Based and Fracture Mechanics-Based
Model for Predicting Delamination in Composites, Composites, 24(6): 467474.
26. Shahwan, K.W. and Waas, A.M. (1997). Non-Self-Similar Decohesion Along a Finite Interface
of Unilaterally Constrained Delaminations, Proceedings of the Royal Society of London, 453:
515550.
27. Reeder, J.R. (1992). An Evaluation of Mixed-Mode Delamination Failure Criteria, NASA-TM-
104210.
28. Allix, O. and Corigliano, A. (1999). Geometrical and Interfacial Nonlinearities in the Analysis of
Delamination in Composites, International Journal of Solids and Structures, 36: 21892216.
29. Ahmad, S., Irons, B.M. and Zienkiewicz, O.C. (1970). Analysis of Thick and Thin Shell
Structures by Curved Finite Elements, International Journal for Numerical Methods in
Engineering, 2: 419451.
30. Beer, G. (1985). An Isoparametric Joint/Interface Element for Finite Element Analysis,
International Journal for Numerical Methods in Engineering, 21: 585600.
31. Daudeville, L., Allix, O. and Ladeve`ze, P. (1995). Delamination Analysis by Damage
Mechanics: Some Applications, Composites Engineering, 5(1): 1724.
32. Pinho, S.T. (2002). Crush Simulation and Energy Absorption of Composite Tubes, MSc Thesis,
Faculty of Engineering, University of Porto, Portugal.
33. Goncalves, J.P., de Moura, M.F., de Castro, P.T. and Marques, A.T. (2000). Interface Element
Including Point-to-Surface Constraints for Three-Dimensional Problems With Damage
Propagation, Engineering Computations, 17(1): 2847.
34. Mi, Y., Crisfield, M.A., Davies, G.A.O. and Hellweg, H.B. (1998). Progressive Delamination
Using Interface Elements, Journal of Composite Materials, 32: 12461273.
35. Schellekens, J.C.J. and de Borst, R. (1991). Numerical Simulation of Free Edge Delamination in
Graphite-Epoxy Laminates Under Uniaxial Tension, Proceedings of the 6th International
Conference on Composite Structures.
36. Schellekens, J.C.J. and de Borst, R. (1993). On the Numerical Integration of Interface Elements,
International Journal for Numerical Methods in Engineering, 36: 4366.
37. Schellekens, J.C.J. (1992). Computational Strategies for Composite Structures, PhD Thesis,
Technical University of Delft, The Netherlands.
38. de Borst, R. and Rots, J.G. (1989). Occurrence of Spurious Mechanisms in Computation of
Strain-Softening Solids, Engineering Computations, 6: 272280.
39. Alfano, G. and Crisfield, M.A. (2001). Finite Element Interface Models for the Delamination
Analysis of Laminated Composites: Mechanical and Computational Issues, International
Journal for Numerical Methods in Engineering, 50: 17011736.
40. Dodes, I.A. (1978). Numerical Analysis for Computer Science, North-Holland, New York, USA.
41. Ungsuwarungsri, T. and Knauss, W.G. (1987). The Role of Damage-Softened Material Behavi-
our in the Fracture of Composites and Adhesives, International Journal of Fracture, 35: 221241.
42. Needleman, A. (1987). A Continuum Model for Void Nucleation by Inclusion Debonding,
Journal of Applied Mechanics, 54: 525531.
43. Rice, J.R. (1968). A Path Independent Integral and the Approximate Analysis of Strain
Concentration by Notches and Cracks, Journal of Applied Mechanics, 31: 379386.
44. Davila, C.G., Camanho, P.P. and de Moura, M.F. (2001). Mixed-Mode Decohesion Elements
for Analyses with Progressive Delamination, 42nd AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics and Materials Conference, Seattle, Washington, USA.
45. Krajcinovic, D. and Fonseka, G.U. (1981). The Continuous Damage Theory of Brittle
Materials, Journal of Applied Mechanics, 48: 809815.
46. Cui, W., Wisnom, M.R. and Jones, M. (1992). A Comparison of Failure Criteria to Predict
Delamination of Unidirectional Glass/Epoxy Specimens Waisted Through the Thickness,
Composites, 23(3): 158166.
47. Mohammadi, S., Owen, D.R.J. and Peric, D. (1998). A Combined Finite/Discrete Element
Algorithm for Delamination Analysis of Composites, Finite Elements in Analysis and Design, 28:
321336.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016

1438 P. P. CAMANHO ET AL.

48. Lee, S.M. (1993). An Edge Crack Torsion Method for Mode III Delamination Fracture Testing,
Journal of Composites Technology & Research, 15(3): 193201.
49. Benzeggagh, M.L. and Kenane, M. (1996). Measurement of Mixed-Mode Delamination
Fracture Toughness of Unidirectional Glass/Epoxy Composites With Mixed-Mode Bending
Apparatus, Composites Science and Technology, 56: 439449.
50. Wu, E.M. and Reuter Jr., R.C. (1965). Crack Extension in Fiberglass Reinforced Plastics, T. &
AM Report No. 275, University of Illinois.
51. Crisfield, M.A., Hellweg, H.B. and Davies, G.A.O. (1994). Failure Analysis of Composite
Structures Using Interface Elements, NAFEMS Conference on Application of Finite Elements
to Composite Materials, London, UK.
52. Hellweg, H.B. (1994). Nonlinear Failure Simulation of Thick Composites, PhD Thesis, Imperial
College of Science and Technology, University of London, UK.
53. Hibbitt, Karlsson and Sorensen. (1996). ABAQUS 6.2 Users Manual. Pawtucket, USA.
54. Crews, J.H. and Reeder, J.R. (1988). A Mixed-Mode Bending Apparatus for Delamination
Testing, NASA-TM-100662.
55. Reeder, J.R. and Crews, J.R. (1990). Mixed-Mode Bending Method for Delamination Testing,
AIAA Journal, 28: 12701276.
56. Reeder, J.R. and Crews, J.H. (1991). Nonlinear Analysis and Redesign of the Mixed-Mode
Bending Delamination Test, NASA-TM-102777.
57. Test Method D6671-01. Standard Test Method for Mixed Mode I-Mode II Interlaminar
Fracture Toughness of Unidirectional Fiber Reinforced Polymer Matrix Composites, American
Society for Testing and Materials (ASTM), West Conshohocken, PA, USA.
58. Reeder, J.R. (2002). Loaddisplacement Relations for DCB, ENF and MMB AS4/PEEK Test
Specimens. Private Communication.
59. Mi, Y. and Crisfield, M.A. (1996). Analytical Derivation of LoadDisplacement Relationship
for the DCB and Proof of the FEA Formulation, Department of Aeronautics Report, Imperial
College of Science and Technology, University of London, U.K.
60. Reeder, J.R. (2000). Refinements to the Mixed-Mode Bending Test for Delamination
Toughness, 15th Conference of the American Society for Composites, Texas, USA.

Downloaded from jcm.sagepub.com at UNIV NEBRASKA LIBRARIES on June 9, 2016