Professional Documents
Culture Documents
PII: S0020-7683(15)00293-0
DOI: http://dx.doi.org/10.1016/j.ijsolstr.2015.06.029
Reference: SAS 8832
Please cite this article as: Shojaei, A., Li, G., Tan, P.J., Fish, J., Dynamic Delamination in Laminated Fiber
Reinforced Composites: A Continuum Damage Mechanics Approach, International Journal of Solids and
Structures (2015), doi: http://dx.doi.org/10.1016/j.ijsolstr.2015.06.029
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Dynamic Delamination in Laminated Fiber Reinforced Composites: A
Continuum Damage Mechanics Approach
Abstract
Delamination causes a significant reduction in the load carrying capacity of Fiber Reinforced
Polymer (FRP) composites, which is a major concern to the aerospace and automotive
industries. High performance FRPs are often subjected to dynamic loadings of different
energy densities in their service life when strain rate, stress triaxiality, temperature, and mode
of fracture can have a significant knock-down effect on the interfacial strength of plies in a
laminate. This paper develops a predictive tool, in a Continuum Damage Mechanics (CDM)
framework, to assess delamination damage in FRPs under dynamic loading; the model takes
into account the effects of dynamic energy density, mixed-mode fractures and temperature.
The CDM model is formulated based on the Fracture Mechanics (FM) of decohesion and an
advantage of the proposed model is that nearly all of the material parameters can be obtained
directly through calibration to experimental data rather than numerical curve fitting of
simulation results. The developed scheme is coded into a commercial FEA package
(ABAQUS) through user-defined subroutines and the fidelity of the model is assessed by
comparison with existing experimental data in the literature. The validated model is used to
where it will be shown that the extent of delamination damage through the thickness of the
FRP structure is dependent upon different wave propagation scenarios. The proposed model
provides a design platform for damage assessment caused by dynamic delamination and may
be a useful tool for designing FRP composites with a greater impact tolerance.
1
1. Introduction
The common failure modes encountered in FRPs are delamination, transverse matrix
cracking, fiber fracture, and fiber-matrix interfacial debonding (Voyiadjis et al. , Chaboche et
al. 2001, Le Quang and He 2008, Horstemeyer and Bammann 2010, Azizi et al. 2011, Doghri
et al. 2011, Kruch and Chaboche 2011, Voyiadjis et al. 2011, Brassart et al. 2012, Li and
Shojaei 2012, Shojaei et al. 2012, Voyiadjis et al. 2012a, Voyiadjis et al. 2012b, Voyiadjis et
al. 2012c, Shojaei and Li 2013, Shojaei et al. 2014, Shojaei et al. 2014). Delamination is
particularly pertinent to composite lay-ups with resin-rich interlaminar layers that have not
been reinforced in the thickness direction. Due to the lower mechanical strength of the
interlaminar layers, compared to their adjacent reinforced plies, interlaminar delamination can
occur in any of the three basic modes shown schematically in Fig. 1. In reality, the state of the
stress in the interlaminar layer is three-dimensional (3D) and delamination occurs under
impact loading and free-edge stresses. A wealth of literature on the modeling of delamination
damage in FRPs have already existed; only a selection of recent pertinent ones will be
reviewed here. Readers are referred to (Hutchinson 1982, Wisnom 2012, Liu et al. 2013, Park
and Paulino 2013) for a more complete review on recent developments in cohesive modeling
methodologies.
Figure 1 Schematic of the three basic modes of loading, viz. (a) Mode I (b) Mode II and (c) Mode III, for
inter-ply delamination.
Different approaches are followed to simulate delamination in composites, and they can be
broadly classified into three categories, viz. (a) Virtual Crack Closure Techniques (VCCT),
2
(b) Cohesive Zone Models (CZMs) (Li and Chandra 2003, Zhang and Paulino 2005, Xu and
Lu 2013), and (c) CDM models. The VCCT technique is based on Irwin’s 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 (Krueger et al. 1999). Whilst this
method is computationally efficient, there are numerical difficulties associated with it, say,
utilizing re-meshing techniques, identification of the delamination front after each crack ‘pop-
up’ and prescribing the delamination path. Cohesive zone models, on the other hand, treat the
fracture formation as a gradual phenomenon in which the separation of the crack surfaces
takes place across an extended crack tip, or cohesive zone, resisted by cohesive tractions
(Needleman 1987a, Needleman and Tvergaard 1987b). Different types of cohesive elements
have been proposed to model cohesive fracture with the finite element method, ranging from
zero-thickness volumetric elements connecting solid elements (De Moura et al. 1997), finite-
thickness volumetric elements connecting shell elements (Reedy et al. 1996), to line elements
(Chen et al. 1999). (Kubair and Geubelle 2003) have argued that intrinsic cohesive
delamination models may result in numerical instabilities and give rise to unrealistic results as
the crack opening velocity becomes negative at the cohesive zone tip. A decohesion element,
capable of dealing with crack propagation under mixed-mode loading, has been proposed and
its efficacy has been demonstrated by (Benzeggagh and Kenane 1996, Kenane and
Benzeggagh 1997, Camanho et al. 2003). The main disadvantages of using cohesive elements
in FEA are (1) difficulties associated with obtaining converged solutions (Alfano and
Crisfield 2001), and (2) the a priori identification and prescription of likely fracture path(s).
(Corigliano 1993) discussed the difficulties concerning the use of interface models in
numerical analyses. (Schellekens and De Borst 1993) simulated the free edge delamination of
uniaxially stressed layered specimens, using non-linear FEA. CDM models have been
(Allix and Ladevèze 1992, Li et al. 1998, Zou et al. 2002, Li et al. 2006, Maimi et al. 2008).
The CDM framework outperforms the conventional VCCT and CZM by (1) CDM utilizes
regular solid elements whilst CZM or VCCT requires that specific types of element to be used
3
for meshing the fracture path, (2) CDM can be coupled with plasticity theories to provide a
more realistic representation of the fracture mechanisms in ductile materials, and (3) fracture
path in CDM approach evolves naturally based upon damage dissipative energies while in the
The topic of dynamic crack growth has been investigated by many researchers including
(Glennie 1971, Glennie 1971a, Freund and Hutchinson 1985, Tvergaard and Hutchinson
1996, Landis et al. 2000, Chen and Ghosh 2012). In general, higher crack velocity results in
higher strain rate hardening effect within the fracture process zone, leading to increased
toughness of the interface (Wei and Hutchinson 1997). Dynamic crack growth along the
(Tzaferopoulos and Panagiotopoulos 1993, Brunner 2000, Corigliano and Allix 2000,
Espinosa et al. 2000, Hashagen and de Borst 2000, Sprenger et al. 2000, Coker et al. 2003,
Khan and Khraisheh 2004, Mariani and Corigliano 2005, Khan and Farrokh 2006, Ahmed
and Sluys 2014). (Slepyan 2010) has studied the non-uniform dynamic crack growth in a
homogeneous isotropic elastic medium subjected to the action of remote oscillatory loads and
(Remmers et al. 2008) carried out FE analyses, using cohesive segments, of the dynamic
impact loads has been investigated by (Rubio-Gonzalez and Mason 2000) and X-FEM has
been utilized by (Grégoire et al. 2007) to study the dynamic crack growth problem.
Asymptotic crack tip stress-strain fields have been developed by (Zhu and Hwang 2008) for
the case of dynamic fractures. Rate-dependent models for damage and plastic deformation of
brittle and ductile materials under dynamic loading have also been extensively investigated
(Zuo et al. 2010, Shojaei et al. 2013). The stability problem of a dynamically propagating
crack has been studied by (Obrezanova et al. 2002, Obrezanova et al. 2002). Bazant and co-
workers have studied the size effect of cohesive cracks (Zi and Bažant 2003, Caner and
Bažant 2009). Despite the developments, nearly all the delamination models for FRPs in the
literature consider only a subset of possible failure mechanisms. Their approach is typically
4
limited to using traction separation laws (cohesive laws) or CDM models that have been
developed based on quasi-static experimental data (Ouyang and Li 2009b, Ouyang and Li
2009c). One of the problems associated with these simplified approaches is that they do not
account for the complete state of the applied stress when investigating the delamination
mechanism. In other words, in dynamic problems with different energy densities, the applied
deviatoric and hydrostatic stress waves may be significant and need to be accounted for in
order to accurately predict fracture (Bao and Wierzbicki 2004, Hooputra et al. 2004, Bao and
Wierzbicki 2005, Dey et al. 2007, Crowell et al. 2012). It is conventional to classify dynamic
problems into the low and high energy density cases where different deformation and damage
mechanisms may operate in each case (Shojaei et al. 2013). In the case of high dynamic
energy density problems, the induced hydrostatic stress may be several times the material
strength and the temperature, due to the dissipative mechanisms, may reach the melting
temperature of the material (Eftis and Nemes 1991, Eftis and Nemes 1996, Eftis et al. 2003).
Furthermore, when the compressive shock waves induced by the loading is reflected from any
free surfaces as tensile waves, delamination damage can occur by spalling (Eftis et al. 2003;
Shojaei, et al. 2013). In the case of lower dynamic energy densities, the shear stress is the
dominant driving force for the delamination damage (Chen and Ghosh 2012) even though
hydrostatic stress could still have a significant effect on the delamination process. In this
paper, a comprehensive delamination model is developed within the CDM framework that
accounts for the influence of shear stress and hydrostatic stress effects, as well as the fracture
mode-mixity effects, on the delamination of FRPs. Several attempts have been made to
level, including homogenization techniques (Jain and Ghosh 2008, Fish 2013, Shojaei and Li
2013, Shojaei and Fish 2015). Instead of dealing with microscale and macroscale bridging
complications, the present work focuses on developing continuum level damage models with
enhanced failure criteria that can capture progressive delamination in FRPs and eliminate the
phenomenological mechanism based CDM model, and the thermodynamic consistency is not
5
discussed herein. The main contributions of the present work include, (i) a rate and
temperature sensitive CDM model is formulated for the delamination analysis of FRPs, (ii)
the developed model is capable of capturing mode-I, mode-II and mixed-mode delaminations,
and (iii) the dynamic loading effect on the progressive delamination is considered.
The present paper is structured as follows. The kinematics of the problem is set out in Section
2 and the fracture mechanics of decohesion is presented in Section 3. The CDM model for
density are incorporated into the CDM model. Section 6 concerns the computational
implementation of the proposed model, and the simulations and results are discussed in
Section 7.
2. Kinematics
Both inelastic deformation and damage mechanisms are concurrently active during the
delamination process of laminated composites. In this work it is assumed that the inelastic
deformation effect on the non-linear mechanical response of each reinforced ply during the
delamination process is negligible. Thus, only damage mechanisms in the interface remain the
main source of non-linearity in softening and post-softening behavior during the delamination
process. Furthermore, in order to simplify the formulation of the delamination process, the
plasticity and damage mechanisms within each reinforced ply are assumed to have negligible
effect on the interlaminar delamination process. Hence, the only damage mechanism,
considered herein, is due to the delamination of the interface that connects two adjacent
elastic plies. The developed delamination model may be used in conjunction with the
It is assumed that the infinitesimal deformation mechanism governs the delamination process
within the fracture process zone. The additive decomposition of the total strain rate, , into
its elastic, , and damage,
, strain rate components holds as follows
6
=
+
. (1)
Appendix A elaborates the physics behind Eq. (1). The constitutive relations for
and
will need to be specified. Nucleation and growth of microcracks lead to a ‘deterioration’
of the macroscopic properties of the interlaminar layer either in the form of added elastic
flexibility (diminution of the elastic stiffness) or increased material anisotropy (Lemaitre and
Dufailly 1987). Hence, it is convenient to write the constitutive equation for the damaged
=
(2)
where is the fourth-order time-dependent damaged elastic stiffness tensor (to be defined
given by
= −
(3)
where
= ∫
is the damaged elastic strain tensor. It remains to specify an appropriate
evolution law for the damage parameters and to formulate the added compliance tensor using
interface is unloaded, the magnitude of the residual strains may not be visible at macroscopic
scale; although, decohesion and debonding processes will induce the residual strains at
microscale within the fracture process zone. In other words, one may consider the localized
strains at the crack tip to justify the existence of the damage strain rate tensor
.
The strength and thickness of the interface layer that connects two reinforced plies are
determined by many factors including depth of the diffused resin, the intermolecular van der
Waals bonds, environmental effects and temperature. Figure 2(a) shows a schematic of the
7
atomistic level decohesion and progressive plastic deformation in the fracture process zone as
the crack-tip advances. The effect of cohesive layer thickness on traction-separation relations
applications the thickness of cohesive layer is negligible compared to the thickness of plies.
composites occurs over a surface of vanishing thickness, allowing the use of a cohesive
formulation. The cohesive-zone methodology is particularly useful when the crack path in
known a priori – which in this study is along the two connecting plies. Figure 2(b) shows a
schematic of the typical decohesion relation where separation of adjacent plies is related to
the magnitude of the cohesive traction across the interface layer. The work of separation,
which is the work needed to create a unit area of fully developed crack, is given by
(Barenblatt 1959)
Γ =
(4)
with
and the stress and relative displacement across the fracture process zone. In the
current application, the three important parameters in the CZM are peak separation stress
∗,
critical separation distance and work of separation or fracture energy given by Eq. (4).
significant plastic zone develops (Tvergaard and Hutchinson 1992, Hutchinson and Evans
2000). However, if the size of this plastic zone is small compared to other characteristic
material parameter that can be measured experimentally (Tvergaard and Hutchinson 1992,
Hutchinson and Evans 2000); otherwise, another fracture framework is needed to describe the
8
Figure 2 (a) Fracture energy dependence on cohesive and plastic dissipative works, and (b) Normal or shear
traction-separation curve for Embedded Process Zone (EPZ) at the crack tip. The EPZ under normal
opening mode-I is depicted in which denotes the limiting fracture energy, is the fracture energy
without any plasticity dissipation and indicates the radius of the plastically affected area.
plastic deformation prevails in the vicinity of the crack-tip so that Γ → 0 and Γ dominates
the decohesion response. Two cohesive behaviors are assumed to be in operation at the
interface; they are modeled by the normal (mode-I) and shear (mode-II and –III) cohesive
curve is constituted from two distinguishable sections including linear elastic, and
# = $ ∗ # − # .
- 0
*1 − , for # > #
# − #
(5)
critical energy release rate for mode-I fracture. In a similar vein, the relationship between
9
%8 8 for 8 < 8
7=9 ∗ 8 − 8
-: 0
7 *1 − , for 8 > 8
8 − 8
(6)
where 1; is the non-linearity exponent; %8 the shear stiffness; 8 (=7 ∗ /%8 ) and 8 (=2566 /
7 ∗ ) are the initiation and critical displacements, respectively, and 566 is the critical energy
release rate for mode-II fracture. Mixed-mode fracture is considered in Section 4.2 by
introducing the mode-mixity effect on the normal and tangential cohesive responses. Due to
the fact that the interlaminar layer is constructed from a thin resin rich layer, it is not
unreasonable to assume that the interlaminar layer is isotropic. In other words, the
interlaminar layer is made of bulk resin material without any reinforcement and it can be
considered isotropic.
In a CDM framework, the density of micro-flaws within the interlaminar layer is represented
by a damage parameter (Kachanov 1958, Murakami 1988, Haddag et al. 2007, Haddag et al.
2009, Voyiadjis et al. 2012a). Following (Sayers and Kachanov 1991), the state of anisotropic
damage is incorporated within a second rank damage tensor, < . Figure 1 shows the
Cartesian coordinate axes adopted to describe the geometry of the problem. If assuming that
delamination damage occurs only by normal opening (mode-I) and/or shear sliding (mode-II
or mode-III), then the active damage parameters in the interlaminar layer are the through-
thickness normal damage == and the shear damages =; and =2 . Hence, the damage tensor
reduces to
0 0 2=
< = > 0 0 ;= ?.
2= ;= ==
(7)
Since the interlaminar layer between two adjacent plies is assumed to be isotropic (see
Section 3), it follows immediately that only one shear damage model needs to be defined to
10
calculate 2= and ;= . To avoid complications associated with size scale effects on the
decohesion process, the thickness of the interlaminar layer needs to be large compared to the
size of the plastic zone at the tip of the delamination crack. The damage parameters will be
Two approaches exist to link the damaged and undamaged material properties, viz.
equivalence of strain energy densities and principle of strain equivalence between the
damaged and fictitious effective configurations (Voyiadjis et al. 2012a, Voyiadjis et al.
2012c). Based on the principle of strain equivalence, the relationships between the
@= = @A B1 − ==C,
where @A and D̅ are the elastic tensile and shear moduli of an undamaged interface,
respectively. A simple procedure is proposed here to calibrate the normal and shear damage
parameters in Eq. (8) against known normal and shear traction separation laws that are widely
available in the literature for different FRPs. Upon the initiation of damage, the parameters in
Eq. (8) are computed directly from known traction-separation relationships as follows
== = 1 − H
FG
G G
,
2= = ;= = 1 − H ,
I
(9)
J J
where # and 8 are the normal and tangential displacements of the interface; whilst
# K# L
and 7K8 L are computed directly from Eqs. (5) and (6), respectively. It is worth noting that the
coupling effects between normal and shear damages are ignored in Eq. (9) and mode-I and –II
fractures are treated separately; however, after the shear and normal moduli of the interface
are updated, other properties may be affected because of the relationships between the elastic
constants.
11
4.2. Mixed-mode delamination
In mixed-mode fractures, the normal and shear separations are concurrently active. For any
combination of in-plane crack deformation modes, the relative amount of mode–I and –II
loading within the cracked edge zone may be represented by a phase angle M based on the
|7 |
M = lim Rarctan W YZ
Q→
#
(10)
where |7| = [; \ \ is the effective shear stress in the interface where \ K=
− =
L
= 2
angle typically ranges between – ^/2 ≤ M ≤ ^/2 (Stringfellow and Freund 1993).
Hutchinson and Evans (2000) have shown that in the case of highly negative or positive phase
angles, i.e. M < −^/6 or M > ^/4, a tenfold increase in fracture resistance compared to the
model-I toughness is possible (Hutchinson and Evans 2000). To account for mixed-mode
fracture of the interlaminar layer, the phase angle must be incorporated into the traction-
separation relationships of the interlaminar layer given in Eq. (5) (for pure model-I, i.e. M =
0) and into Eq. (6) (for pure mode-II fracture, i.e. M = +^/2 ).
equivalent normal and tangential traction-separation laws for – M2 < M < where – M2 is a
b
;
lower bound value dependent on the materials (Freund and Suresh 2003). Ji et al. (Ji et al.
2011, Ji et al. 2012a, Ji et al. 2012b, Ji et al. 2013) have shown experimentally that fracture
proposed that the equivalent normal and shear traction-separation laws be obtained by a
12
normal strength (
∗), normal initiation (# ) and normal critical (# L displacements in Eqs.
A ∗ = K
efee −
∗ LM +
∗ ,
d ∗
b
where
efee
∗
, hKefeeL
, and hKefeeL
are the mixed-mode-I/II normal strength, initiation and
critical displacements, respectively, to be obtained from the experimental results of (Ji et al.
2012b). In a similar vein, the shear traction separation parameters in Eq. (6) are replaced with
the following
7̅ ∗ = K7 ∗ − 7efee
∗ L
jM − k + 7efee
d b ∗
b d
,
where 7efee
∗
, lKefeeL
, and lKefeeL
are the mixed-mode (I-II) shear strength, initiation and
critical displacement, respectively. Eqs. (11) and (12) provide a simple approach to account
for the effects of phase angle on the fracture resistance of the interlaminar layer and it will be
It is worth noting that the phase angle for mixed-mode fractures may depend on the dynamic
loading rates as well. The rate dependency of the phase angle will be considered by the
To complete the CDM formulation, a damage initiation criterion must be specified. The
criterion used will depend on whether the interlaminar layer is initially intact or whether there
is a pre-existing crack of a structural length scale. A mixture of stress-based, mF , and fracture
mechanics-based, mn , damage criterion will be adopted here. If the interlaminar layer is
13
initially intact, then a stress-based damage initiation criterion o , based on a quadratic
q== ; 7̂2; ; 7̂2= ;
o = mF = p ∗ r + p ∗ r + p ∗ r ≤ 1
7 7
(13)
where
q==, 7̂2; and 7̂2= are the normal (out-of-plane) and in-plane shear stresses, respectively.
For a partially fractured interlaminar layer, the stress-based criterion in Eq. (13) may lead to
numerical instabilities and/or difficulties associated with resolving the stress field near the
crack tip. An alternative criterion based on a quadratic interaction of the energy release rates
nt 2
ntt 2
nttt 2
nt ntt nt nttt ntt nttt
o = mn< = p r +p r +p r +p × r+p × r+p × r
5tu 5ttu 5tttu 5tu 5ttu 5tu 5tttu 5ttu 5tttu
(14)
≤w
566 and 5666 are the critical energy release rates in mode I, II and III, respectively; and n
(i=I,II, and III) are energy release rates computed based on the traction-displacement
n6 = ∫
# dy==,
n666 = ∫ 7 dy;=,
where y== , y;= and y2= are the normal and tangential displacements; and
# and 7 are the
traction-separation laws of Eqs. (5) and (6). The computational strategy will be described in
Section 6.
14
It is worthwhile noting that the main objective of introducing two separate damage initiation
and propagation criteria in this work is to minimize the stress singularity effects that may
have a profound effect on mesh dependency of the results. The damage mechanisms are
multiscale phenomena that start from micro-cracks/-voids and eventually they form
CDM models is the element removal process in which elements with excessive damage are
removed from analysis. The element removal may result in stress singularity points, viz.
stress approaches to infinity upon mesh refinement steps. The transition between the stress,
i.e. Eq. (13), and fracture energy, i.e. Eq. (14), based criteria can be either smooth or sudden.
In the case of smooth transition, two damage thresholds may be utilized where at lower
damage levels, e.g. less than 0.05, the stress based criterion controls the damage
initiation/propagation while at higher levels of damages, e.g. higher than 0.25, the fracture
based criterion is used. In other words at lower damage levels, it is assumed that the softening
due to damage is negligible and the stress singularity effect is not dominant and the stress
based criterion can effectively capture the damage initiation and propagation. While in the
case of higher damage levels the fracture based criterion is utilized to capture the damage
initiation and propagation that is less sensitive to stress singularity effects. In the midrange
damage level, viz. 0.05 to 0.25, a mixed criterion can be designed to provide a smooth
transition between the two criteria. Obviously the lower and higher damage levels are
material parameters that can be estimated using bi-material elasticity problem solution in
which the elastic moduli difference is calculated for inducing a stress singularity at the
boundary, e.g. see (Williams, and Zak, 1962, Sator and Becker 2011).
Delamination in FRPs is affected by the physical and chemical composition of the distinct
phases, the operating temperature, chemical environment, rate of loading, rate of cracking,
and other factors. It is postulated that the dynamic energy density of a system has a significant
influence on the delamination process and is related to the level of stress triaxiality λ (∶=
15
Σ/|τ| where Σ is the hydrostatic part of the applied stress, and |7| is the effective shear stress).
A schematic showing how critical displacement varies with λ is given in Figure 3 (Bao and
Wierzbicki 2005, Shojaei et al. 2013). Different loading regimes can be defined depending on
λ as follows:
- Regime (I) - { < −{2 : High level of compressive hydrostatic stress prevents micro-flaws
- Regime (II) - −{2 < { < 0: The hydrostatic stress, whilst still negative, is sufficiently
microcracking with friction. The microcrack growth rate equation by Deng and Nemat-Nasser
(1992) and Shojaei et al. (2013) takes into account the effects of frictional sliding.
- Regime (III) - 0 < { < {; : The hydrostatic stress is tensile in this regime, allowing the
microcrack surfaces to separate and propagate with negligible sliding friction. Shear forces
remain the dominant mechanism that drives the damage process because of the weak shear
strength.
- Regime (IV) - {; < { < {= : High level of tensile hydrostatic stress facilitates void
- Regime (V) - { > {= : Compressive stress wave is reflected from the free surface as a tensile
wave. If the magnitude of the reflected tensile wave is higher than the tensile strength of the
hydrostatic stress controls the damage process and appropriate equations of states must be
16
Figure 3 Effect of the applied dynamic energy density (stress triaxility level) on the critical failure
displacement.
The effects of stress triaxiality are incorporated into the cohesive laws by modifying the
critical normal #̅ and shear 8̅ displacements in Eqs. (11) and (12). In regimes (II) and
(III), the new critical normal #̿ and shear 8̿ displacements, double over-bar denotes post
where subscript “#” is either “n” or “t”. If stress triaxiality falls between {; < { < { , a
where “#” is the same as above. The proposed phenomenological approach enables dynamic
energy density to be incorporated into the delamination model and it will be shown later to
compare well with experimental data. From the mathematical point of view, Eqs. (16) and
(17) result in very high critical strain (no failure) in the case of { → −|{2 | and the critical
strain is reduced by marching from −|{2 | to { (resulting in higher damage levels). For
17
more experimental justifications, the interested reader may refer to Johnson and Cook (1985)
Strain rate can have a significant influence on the ultimate tensile and shear stresses (Huang et
al. 2009, Amini et al. 2010). Likewise, elevated temperature may also give rise to material
instability; in particular, under dynamic loading of high dynamic energy content (Deng and
Nemat-Nasser 1992). The effects of strain rate and elevated temperature are accounted for by
adjusting the interfacial mechanical strength in Eqs. (11) and (12). Following (Johnson and
∗ =
A∗ × 1 + log j k × 1 − ∗
| |
7̿ = 7̅∗ × 1 + log j k × 1 − ∗
| |
(18)
∗
where || = [ is the applied strain rate; is a reference strain rate;
A∗ and 7̅∗ are,
;
=
respectively, the normal and shear interface strengths measured at reference strain rate ; T*
is the homologous temperature to be defined in Eq. (20); and C and are material parameters
widely available in the literature (Johnson and Cook 1985, Steinberg 1996). One may note
that Johnson-Cook model has originally been developed for metals, and this model
provides an efficient way for including the rate and temperature effects into the
constitutive modeling framework. In this work the performance of the JC model has
experimentally investigated and it is shown that this model can be used for capturing
rate dependent cohesive strength of FRPs. In a similar vein, the quasi-static normal and
# and shear %
shear stiffness of the interface must also be adjusted to dynamic normal % 8
stiffness as follows
# = %
% # × 1 + ′ log j| |k × 1 − ∗
′
(19)
18
8 = %
% 8 × 1 + ′ log j| |k × 1 − ∗
′
# =
A∗ /#̅ and %
where % 8 = 7̅∗ /8̅ are, respectively, the reference normal and shear
stiffness measured at the reference strain rate . The homologous temperature, T*, in Eqs.
where (K) is the current temperature, 8# (K) is the transition temperature and 8 is
the melting temperature. Extensive testing is needed to accurately establish the coefficients ′
and ′ for various delamination scenarios. Because experimental data regarding temperature
effect is not currently available, only the strain rate effect is considered in this work. The
investigation by the authors and the results will be reported in a forthcoming paper.
6. Implementation Aspects
Figure 4 shows a flow chart of the numerical implementation into a commercial FEA package
subroutines. The dynamic implicit integration scheme, see (ABAQUS, 2011), with refined
mesh, together with enhanced hourglass meshes, are utilized. The developed damage
variables are updated incrementally at each material point as field variables and the updated
values are utilized to reduce the elastic tensile and shear moduli. Once the damage parameter
reaches to its critical value in one of the integration points, the corresponding stiffness is set
to a small value to artificially remove the respective element from the model. This element
deletion technique keeps the mass of the system unchanged. As discussed by Lemaitre and
Dufailly (1987) the critical elastic modulus should be defined based on the material system.
19
In dynamic problems with progressive damage, stiffness reduction of elements (due to the
known as ill-posed constitutive relations (de Borst and Sluys 1991, Sluys and de Borst 1994,
Shojaei et al. 2013). This latter phenomenon results in strong mesh dependencies in the FEA
results where the dissipated energies decrease with repeated mesh refinement. There are a few
approaches, proposed in the literature, to avoid pathological mesh dependency of the CDM
results, such as non-local continua, see (Shojaei et al., 2013) for more details. In the case of
non-local methods, a gradient enhanced theory is required to remove the ill-posedness of the
problem via integration schemes that involve characteristic lengths, see (Voyiadjis et al.
2014) for example. Due to the fact that the proposed CDM model is a local theory, an
alternative approach is adopted to compensate for the ill-posedness of the boundary value
problem, where the softening part of the CDM constitutive law is defined based upon a stress-
displacement relation. The energy dissipated during the damage process is specified per unit
area, not per unit volume. Thus, the damage dissipation energies (consistent with critical
energy release rate parameters in fracture mechanics) are used to compute the displacement at
which full material damage occurs. This formulation ensures that the correct amount of
energy is dissipated and alleviates the mesh dependency (ABAQUS 1011). Consequently, the
y== , y2= and y;= in Eq. (15), are computed based on energy descriptions, i.e. damage
discusses more in detail the computational difficulties facing ill-posed constitutive relations.
The mesh convergence study is provided in Appendix B to show the efficiency of the
20
Figure 4 FEA implementation flowchart for the delamination damage model
Experimental investigations into Mode-I delamination were carried out by (de Morais et al.
2002, Ji et al. 2010, Ji et al. 2013) using Double Cantilever Beam (DCB) specimens. End-
Notched Flexure (ENF) specimens were used to investigate the mode-II interlaminar fractures
by (Pereira et al. 2004, Ji et al. 2011, Ouyang et al. 2011, Ji et al. 2012a). Mode-III
delamination was studied using Edge Crack Torsion (ECT) samples for carbon/epoxy
laminates by (de Morais et al. 2009). Mixed mode I/II fracture for single-leg bending (SLB)
specimens was studied by (Ouyang et al. 2011, Ji et al. 2012a). In this section, the
previously reported by (Ji et al. 2011, Ji et al. 2012a, Ji et al. 2012b, Ji et al. 2013).
The adhesive, LOCTITE Hysol 9460, that was used has a high peel strength, good impact
resistance, and good fatigue resistance. According to the manufacturer the adhesive agent has
an elastic modulus of 2.76 GPa, tensile strength of 30.3 MPa and failure elongation of 3.5%.
Prefabricated plain woven glass fabric reinforced epoxy laminates are used to manufacture
the Double Cantilever Beam (DCB) specimen for mode-I experiments, and the End Notched
Flexure (ENF) specimen for mode-II experiments. Single Leg Bending (SLB) samples made
from steel plates bonded with Hysol 9460 were used for mixed mode-I/II tests. The DCB
21
specimen is made of two 3.1 mm thick slabs, with a width and length of 25.4 mm and 254.0
mm, respectively, and an initial crack length of 25mm. Each slab is made of 17 layers of
continuous glass woven fabric with epoxy resin (Ji et al. 2013). The ENF specimens, made of
the same material system, are 20mm wide and 150mm long, with a total thickness of 3 mm
for each laminate and an initial crack length of 35 mm (Ji et al. 2011). Table 1 summarizes
the material properties of the DCB, and ENF samples. For the SLB samples, the steel plate
used for the upper adherent is 6.35 mm thick, 25.4 mm in width and has a length of 304.80
mm; whilst the lower adherent has a length of 279.40 mm (Ji et al. 2012b). For each test,
performed on an MTS 810 machine, the local deformations are captured using a Sony XCD-
CR90 high resolution CCD camera. The test data will be used to validate the results of the
Table 2 lists the material parameters for the shear and normal traction-separation relationships
used in different mode-mixity scenarios. It is worth noting that only two parameters 12 and 1;
are obtained from numerical curve fitting techniques. Fig. 5 shows a 3D representation of
how the shear traction separation relationship varies with phase angle. The experimental data
for the ENF specimens (M = 90 ) and for the SLB specimens (M = 45 ) are plotted in the
Table 1 Material parameters for epoxy laminates used in DCB and ENF tests
Table 2 Traction-separation material parameters obtained from DCB, ENF and SLB
∗
∗ f ¡¢ ¡£ ¡¢K f L
experiments
∗ ∗
f
(DCB) (ENF) (SLB) (SLB) (DCB) (ENF) (SLB)
(MPa) (MPa) (MPa) (MPa) (µm) (µm) (µm)
22
Figure 6 shows a 3D representation for the normal traction-separation responses for various
phase angles. The DCB data is for M = 0 and SLB test data is for M = 45 . Good
agreements between simulations and experiments are observed for both normal and shear
traction-separations. The effects of stress triaxiality on the critical normal and shear
displacements are given in Fig. 7. Simulation results in this graph confirm experimental
observations that high negative stress triaxialities delay failures while highly positive stress
triaxialities result in accelerated failure mechanisms (Bao and Wierzbicki 2004). According to
the experimental observation the following values for the boundaries between different
regimes in Fig. 3 are chosen: {2 = −1/3, {; = 2 and { = 10, see for example (Bao and
Wierzbicki 2004). One may notice that {= is ignored herein and a linear correlation is used in
Eq. (17) from {; to { . Low to medium range strain rate tests were carried out to capture
normal (DCB test) and shear (ENF test) interfacial strengths. As shown in Fig. 8, the strain
rate dependency of the interfacial strengths can be effectively captured using Eq. (18). The
measured at this strain rate. The effect of temperature rise is not considered in the current
simulations, as it requires extensive experimental studies. The simulation results for the
normal = and shear 2= (=;= ) damage parameters are given in Figs. (9) and (10),
respectively, which shows how the extent of delamination damage varies with phase angle.
For sake of clarity different colors are used in Figs. 5, 6, 9, 10 in which the meaning of each
23
Figure 5 Experimental and simulation results for shear traction-separation in which ENF and
SLB test data are compared to the simulated results. Experimental data are from (Ji et al. 2011,
Ji et al. 2012b). Colors black and blue show the simulation results for the mixed-mode and mode-
II, respectively. Other colors show the transition between mode-II and mixed-mode.
Figure 6 Experimental and simulation results for normal traction-separation in which DCB and
SLB test data are compared to simulated results. Experiment data are from (Ji et al. 2012b, Ji et
al. 2013). Colors green and blue show the simulation results for the mode-I and mixed-mode,
respectively. Color spectrum illustrates the transition between mode-I and mode-II.
24
To investigate the performance of the developed framework for dynamic fracture of the
laminated FRPs, the delamination CDM model is introduced into a commercial FEA package,
ABAQUS, through user-defined subroutines VUMAT. At first stage, the DCB and ENF tests
are modeled to compare the FEA and experimental outputs for the sake of validation and
conducting the mesh convergence studies. The dimensions of the DCB and ENF samples
have already been mentioned, and an interlaminar layer with thickness of 0.5 mm is
considered for both DCB and ENF models. The C3D8R reduced integration elements with
enhanced hourglass are used in both models. Plies are meshed by the same element type and
as mentioned before they are modeled as elastic medium for sake of simplicity (see paragraph
before Eq. (1) for more details). Fig. 11 compares the traction separation response of DCB
and ENF models to their respective experimental data. It is worth noting that the FEA
example herein provides a performance demonstration for the proposed CDM model. It is
now well-understood that eight-node, one-point quadrature element, e.g. C3D8R, which has
reduced integration and hourglass control, has very poor behavior in bending-dominated
problems. The interested reader may refer to (Fish and Belytschko, 2007) for more details on
alternative enhanced finite element formulations which are highly recommended to use in
Figure 7 Effect of stress triaxiality on the dynamic delamination response of the material system
25
Figure 8 Comparison of experimental data and results of numerical simulations for the
variations of the normal and shear interfacial strengths with strain rate
Figure 9 Normal damage parameter for various phase angles. The color spectrum shows the
Figure 10 Shear damage parameter for various phase angles. The color spectrum shows the
26
Figure 11 Experimental and FEA results for (a) DCB, (b) ENF tests. The experimental setups are
shown in red boxes in each graph, refer to (Ji et al. 2013) for details of experiments.
The performance of the developed scheme in the case of a dynamic loading condition is
investigated next. A laminated FRP under projectile impact is studied. The same material
properties, considered for DCB and ENF models, are assumed for the plies and interface. The
12 in which the numerically measured load (F) and experimentally characterized delamination
shown in Fig. 13. Each laminated substrate, consists of 20 plies, is 2 mm thick and
interlaminar layers with 0.5 mm thickness are considered between them. The laminated FRP
is assumed to be clamped all around external edges. The element type and size should be
chosen in accordance to the mesh convergence studies for the DCB and ENF models. The
mesh convergence study has been conducted for the DCB sample in Appendix B and the
element size related to the converged results are utilized for the case of projectile impact
simulation. A rigid spherical projectile with 10 mm radius and 0.2 kg mass is impacted onto
the center of the system with initial velocity of 100 m/s. The compressive and tensile wave
propagations in the system are studied and the dynamic delamination is investigated in each
interface. Fig. 14(b) to 14(d) illustrate stress wave propagations across the thickness and
transverse directions. General contact between rigid spherical surface of the projectile and the
composite plate is prescribed where frictionless tangential and hard normal (no penetration)
contact properties are assigned to the model. In the case of Explicit Integration methods in
27
which the displacement equilibrium solution at time “t + ∆t” is based on using the conditions
of equilibrium at time “t”; a stable integration requires that ∆t is smaller than a critical value
∆t ¬Q . In the case of ∆t > t ¬Q the integration is unstable that means errors resulting from the
numerical integration or round off are too high that affects the response calculations. The
critical time step ∆t ¬Q is correlated to the mass, stiffness and damping properties of the
structure. An approximation to the stability limit can be expressed in terms of the smallest
transit time of a dilatational wave across any of the elements in the mesh ∆t ¬Q ≅ D°±h /c²
where D°±h is the smallest element dimension and c² is the dilatational wave speed.
Figure 12 DCB simulation results (a) von-Mises stress distribution, and (b) Progressive
delamination damage in DCB specimen. The Load (F) -Delamination lengths (Lf) data correlate
well with the observed experiments. The color spectrum shown represents various damaged
states in which the red color indicates fully damaged (damage variable=1.0) and the blue color
28
Figure 13 Schematic for a special orthotropic composite lay-up configuration with the
Figure 14 A laminate impacted by a rigid spherical projectile with V0=100 (m/s), (a) schematic of
a quarter (X and Y cut planes are shown) of the FEA model and boundary conditions (B.C.), (b)
reflection of the projectile after impacting the laminate, (c) propagation of stress waves at time
step 1e-4, and (d) propagation of compressive wave through the thickness at time step 1e-3 .
Fig. 15(a) shows the stress triaxiality distribution on the center line, see Fig. 14(a), of the
laminate for different time steps. The local coordinate system (1,2,3), shown in Fig. 14(b), is
used to plot Fig. 15. It is worth noting that upon incidence of the compressive and tensile
29
waves with interlaminar layers, a portion of the wave is transmitted to the next ply while
another portion is reflected back. The noises in the stress triaxiality distributions in Fig. 15(a)
are due to these interactions. Fig. 15(b) depicts the state of the normal damage distribution in
the interlaminar layers. It is obvious from Fig. 15(b) that due to the compressive wave
propagation the normal delamination damage is quite small in interlaminar layers near the
impacted face. Upon reflection of the compressive wave from the free surface (bottom of the
laminate), the effect of normal delamination becomes more significant and results in mode-I
dominated delamination of plies near the free edges. Fig. 15(c) depicts the state of the shear
damage for different interlaminar layers. One may argue that the thickness of the interlaminar
layer in laminated FRPs might be less than 0.5 mm. The effect of the interlaminar layer
thickness on the delamination responses can be taken into account through sensitivity
analysis.
Figure 15 (a) Distribution of the stress triaxiality across the thickness of the laminate on the
center line for various step times, (b) normal damage distribution in different interlaminar
30
The actual strain hardening/softening rate dependency in dynamic crack growth problems
occurs within the fracture process zone, where crack propagation velocity governs the
localized strain rates at the crack tip. Thus, it is instructive to discuss the relationship between
crack opening/sliding displacement rates, local strain rate ̅ at the crack tip within the
Embedded Process Zone (EPZ), which is depicted in Fig. 2b, and is called EPZ strain rate
hereinafter; and macroscale strain rate which occurs outside the EPZ zone, e.g. in adjacent
plies. The macroscopic strain rate , which is governed by continuum deformation
mechanisms in the medium, may remain finite outside the EPZ; although, the localized EPZ
strain rate ̅ may exceed 1000 s-1 due to the speed of crack advance (Glennie 1971, Glennie
1971a). The crack propagation may reach velocities in the order of 50 to 300 m.s-1 in the case
the EPZ strain rate via dilatational u2 and shear u; stress wave speeds in the medium
(Glennie 1971, Glennie 1971a). One possible correlation between normal # and sliding
2 2
|̅| ≅ ³´ j Gk + p1 − r j1 − | | k j Jk
| {| |{| |{|
{\µ¶·· u1 {\µ¶·· {1 u2
(21)
where ³ is a constant to be obtained experimentally. Eq. (21) implies that |̅| depends on the
dynamic energy density content of the loading. In the case of shear-dominated loading, i.e.
|{| → 0, the shear wave dominates the deformation mechanism in the interface and the shear
stress wave governs EPZ strain rate. For highly compressive stresses || → 1, and →1
|| ||
.
for tensile stresses. Thus, the dilatational waves and normal crack opening rates affect the
EPZ strain rate. In an isotropic medium, the longitudinal and shear wave speeds can be
estimated through relations between elastic modulus, Poisson’s ratio and density of the
system (Glennie 1971, Glennie 1971a). Bridging the decohesion mechanisms in microscale
31
One may note that only delamination damage is considered in this work; while matrix
cracking, other types of damage and/or plastic deformation of plies may have a profound
effect on integrity of the system. For a full-scale health monitoring of the laminated FRP, the
present delamination model needs to be coupled with proper descriptions of plasticity and
damage of the plies. The plastic and damage processes in plies influence the delamination
damage mechanisms, and different damage distributions may be expected compared to that
shown in (15). This is the subject of the forthcoming paper by the authors where the matrix
cracking, fiber/matrix debonding and matrix plasticity are also considered in addition to
delamination damage to monitor the integrity of the laminated FRP under dynamic loading
conditions.
8. Conclusion
Dynamic delamination in FRPs is investigated in this work within the CDM framework.
delamination model for FRPs. Particularly the effects of the mixed-mode fractures and
dynamic energy density on the delamination responses are considered. The strain rate effects
on the strength and stiffness of the interlaminar layers are also studied in which empirical
relations are established for dynamic stiffness and strength of the interfaces. Comparing to the
existing cohesive models in the literature, the framework is developed in such a way that most
of the material parameters are directly obtained from the experimental testing that minimizes
the numerical curve fitting tasks. The developed CDM framework may be considered as an
alternative to the classical cohesive element in Finite Element Analysis (FEA) problems. The
fracture path in the developed CDM approach evolves naturally and conventional mesh types
can be utilized to mesh the FEA model; whilst in the case of discrete cohesive elements the
fracture paths need to be predefined to use specific types of cohesive element in that region.
In other words, in the case of discrete cohesive elements the fracture paths are always
predefined by the shape of two mating surfaces in which a single cohesive element is placed
between those mating surfaces. Consequently, the fracture path is always the same as the
32
shape of two bonding surfaces. In the case of CDM, multiple elements through the thickness
are allowed that result in natural fracture path evolution based on dissipative energy release
rates. Thus, the delamination fracture path in CDM models is not anymore dependent on the
shape of the two adjacent layup surfaces. One may note that still material properties of the
layups and interlaminar layers are assigned separately. Introduction of the developed scheme
into a commercial FEA package, through user-defined subroutines, provide a powerful design
tool to analyze the delamination in FRPs. The performance of the proposed scheme is shown
to compare well with existing experimental data in the literature, which demonstrates that the
Acknowledgment
between NASA and the Louisiana Board of Regents under contract NASA/LEQSF (2011-
14)-Phase3-05. This study was also partially supported by the NSF, under grant number
REFERENCES
Ahmed, A. and L. J. Sluys (2014). "A phantom node formulation for modeling
coupled adiabatic–isothermal cracking in FRP composites." Computer Methods in
Applied Mechanics and Engineering 278(0): 291-313.
Alfano, G. and M. A. Crisfield (2001). "Finite element interface models for the
delamination analysis of laminated composites: mechanical and computational
issues." International Journal for Numerical Methods in Engineering 50(7): 1701-
1736.
Allix, O. and P. Ladevèze (1992). "Interlaminar interface modelling for the prediction
of delamination." Composite Structures 22(4): 235-242.
Azizi, R., C. F. Niordson and B. N. Legarth (2011). "Size-effects on yield surfaces for
micro reinforced composites." International Journal of Plasticity 27(11): 1817-1832.
Bao, Y. and T. Wierzbicki (2004). "On fracture locus in the equivalent strain and
stress triaxiality space." International Journal of Mechanical Sciences 46(1): 81-98.
33
Bao, Y. and T. Wierzbicki (2005). "On the cut-off value of negative triaxiality for
fracture." Engineering Fracture Mechanics 72(7): 1049-1069.
Bouvard, J. L., J. L. Chaboche, F. Feyel and F. Gallerneau (2009). "A cohesive zone
model for fatigue and creep–fatigue crack growth in single crystal superalloys."
International Journal of Fatigue 31(5): 868-879.
Coker, D., A. J. Rosakis and A. Needleman (2003). "Dynamic crack growth along a
polymer composite–Homalite interface." Journal of the Mechanics and Physics of
Solids 51(3): 425-460.
34
Corigliano, A. (1993). "Formulation, identification and use of interface models in the
numerical analysis of composite delamination." International Journal of Solids and
Structures 30(20): 2779-2811.
Dey, S., T. Børvik, X. Teng, T. Wierzbicki and O. S. Hopperstad (2007). "On the
ballistic resistance of double-layered steel plates: An experimental and numerical
investigation." International Journal of Solids and Structures 44(20): 6701-6723.
35
Eftis, J. and J. A. Nemes (1991). "Evolution equation for the void volume growth rate
in a viscoplastic-damage constitutive model." International Journal of Plasticity 7(4):
275-293.
Fish, J. (2013). Practical Multiscaling. United Kingdom, John Wiley & Sons.
Haddag, B., F. Abed-Meraim and T. Balan (2009). "Strain localization analysis using
a large deformation anisotropic elastic–plastic model coupled with damage."
International Journal of Plasticity 25(10): 1970-1996.
Hooputra, H., H. Gese, H. Dell and H. Werner (2004). "A comprehensive failure
model for crashworthiness simulation of aluminium extrusions." International Journal
of Crashworthiness 9(5): 449-464.
36
Huang, Y., D. Meyer and S. Nemat-Nasser (2009). "Damage detection with spatially
distributed 2D Continuous Wavelet Transform." Mechanics of Materials 41(10):
1096-1107.
Ji, G., Z. Ouyang and G. Li (2012a). "Local Interface Shear Fracture of Bonded Steel
Joints with Various Bondline Thicknesses." Experimental Mechanics 52(5): 481-491.
Ji, G., Z. Ouyang and G. Li (2012b). "On the interfacial constitutive laws of mixed
mode fracture with various adhesive thicknesses." Mechanics of Materials 47(0): 24-
32.
Ji, G., Z. Ouyang, G. Li, S. I. Ibekwe and S. S. Pang (2010). "Effects of Adhesive
Thickness on Global & Local Mode-I Interfacial Fracture of Bonded Joints."
International Journal of Solids and Structures 47: 2445-2458.
37
Khan, S. M. A. and M. K. Khraisheh (2004). "A new criterion for mixed mode
fracture initiation based on the crack tip plastic core region." International Journal of
Plasticity 20(1): 55-84.
Li, G. and A. Shojaei (2012). "A viscoplastic theory of shape memory polymer fibres
with application to self-healing materials." Proceedings of the Royal Society A:
Mathematical, Physical and Engineering Sciences 468(2144): 2319-2346.
Li, H. and N. Chandra (2003). "Analysis of crack growth and crack-tip plasticity in
ductile materials using cohesive zone models." International Journal of Plasticity
19(6): 849-882.
Li, S., S. R. Reid and P. D. Soden (1998). "A continuum damage model for transverse
matrix cracking in laminated fibre–reinforced composites." Philosophical
Transactions of the Royal Society of London. Series A: Mathematical, Physical and
Engineering Sciences 356(1746): 2379-2412.
Li, S., S. R. Reid and Z. Zou (2006). "Modelling damage of multiple delaminations
and transverse matrix cracking in laminated composites due to low velocity lateral
impact." Composites Science and Technology 66(6): 827-836.
Liu, J., J. Li and B. Wu (2013). "The Cohesive Zone Model for Fatigue Crack
Growth." Advances in Mechanical Engineering 2013: 16.
38
Mariani, S. and A. Corigliano (2005). "Impact induced composite delamination: state
and parameter identification via joint and dual extended Kalman filters." Computer
Methods in Applied Mechanics and Engineering 194(50–52): 5242-5272.
Ouyang, Z. and G. Li (2009c). "Cohesive Zone Model Based Analytical Solutions for
Adhesively Bonded Pipe Joints under Torsional Loading." International Journal of
Solids and Structures 46(5): 1205-1217.
Reedy, E. D., F. J. Mello and T. R. Guess (1996). Modeling the Initiation and Growth
of Delaminations in Composite Structures SAND95-3070 New Mexico, CA, USA,
Sandia National Laboratories.
39
Rubio-Gonzalez, C. and J. J. Mason (2000). "Dynamic stress intensity factors at the
tip of a uniformly loaded semi-infinite crack in an orthotropic material." Journal of
the Mechanics and Physics of Solids 48(5): 899-925.
Sayers, C. M. and M. Kachanov (1991). "A simple technique for finding effective
elastic constants of cracked solids for arbitrary crack orientation statistics."
International Journal of Solids and Structures 27(6): 671-680.
Shojaei, A., A. Dahi Taleghani and G. Li (2014). "A continuum damage failure model
for hydraulic fracturing of porous rocks." International Journal of Plasticity 59(0):
199-212.
Shojaei, A., G. Li, J. Fish and P. J. Tan (2014). "Multi-scale Constitutive Modeling of
Ceramic Matrix Composites by Continuum Damage Mechanics." International
Journal of Solids and Structures 51(23-24): 4068-4081.
Slepyan, L. I. (2010). "Dynamic crack growth under Rayleigh wave." Journal of the
Mechanics and Physics of Solids 58(5): 636-655.
40
softening model." Computer Methods in Applied Mechanics and Engineering 185(2–
4): 123-139.
41
Xu, Q. and Z. Lu (2013). "An elastic–plastic cohesive zone model for metal–ceramic
interfaces at finite deformations." International Journal of Plasticity 41(0): 147-164.
Zhu, X.-K. and K.-C. Hwang (2008). "Asymptotic crack-tip fields for dynamic crack
growth in compressive power-law hardening materials." International Journal of
Solids and Structures 45(13): 3644-3659.
Zi, G. and Z. P. Bažant (2003). "Eigenvalue method for computing size effect of
cohesive cracks with residual stress, with application to kink-bands in composites."
International Journal of Engineering Science 41(13–14): 1519-1534.
Zou, Z., S. R. Reid, S. Li and P. D. Soden (2002). "General expressions for energy–
release rates for delamination in composite laminates." Proceedings of the Royal
Society of London. Series A: Mathematical, Physical and Engineering Sciences
458(2019): 645-667.
Zuo, Q. H., D. Disilvestro and J. D. Richter (2010). "A crack-mechanics based model
for damage and plasticity of brittle materials under dynamic loading." International
Journal of Solids and Structures 47(20): 2790-2798.
In most of the classical FEA simulations utilizing a stable material model is essential before
conducting a finite element analysis. Drucker stability criteria ensures that the material model
is stable
> 0 (A.1)
It is worth noting that Drucker’s stability law is not a thermodynamic law, and many
materials show clear signs that they are not stable in the sense of Drucker’s stability check
that the violation of Drucker's stability postulation results in volumetric locking in FEA
element formulation, and failure of linear elasticity theory; unless an additional length– or
time–scale is specified in the constitutive relations (de Borst et al. 1999). The material
instability is a part of a progressive failure simulation and dealing with the material instability
42
is unavoidable in the case of delamination analysis. In this work a length scale is included into
The damage mechanisms, i.e. microcracking and microvoiding, induce residual strains due to
mismatch between non-damaged and damaged stiffness (Bouvard et al. 2009). As depicted in
Fig. A1, there are two elastic strains that can be computed based upon either damaged E or
undamaged E elastic stiffness. In present work, the total strain in Eq. (1) is decomposed into
recoverable damaged elastic strain ϵº and inelastic damage strain ϵ² . One may note that the
name damaged elastic strain indicates that the recoverable elastic strain is computed based
Figure A1. Typical stress straincurve in the case pure elastic and damage deformation
mechanisms (no plasticity)
at full damage from damage dissipation energy. The convergence study to support this
43
argument is provided in this appendix. Fig. B1 compares the traction separation response of
DCB models to the experimental data. The mesh convergence study is conducted by
systematically halving the mesh sizes and it reveals that fine meshes ensure the convergence
of the results. Only three mesh numbers are shown in Fig. B1 for DCB sample, and the
corresponding mesh sizes for each case are as follows: Mesh#1: 0.1 mm, Mesh#2: 0.025 mm,
44