Efficient Modeling of Damage Processes in

Heterogeneous Materials under Impact Load

Rajesh S. Kumar, Mark R. Gurvich
United Technologies Research Center
411 Silver Lane, East Hartford, CT 06108, USA
KumarRS@utrc.utc.com , GurvicMR@utrc.utc.com

Abstract: A general computational approach for modeling of damage in heterogeneous materials

and structures under dynamic load is the focus of the current work. The approach covers
processes of both damage initiation and damage propagation. In addition, multiple damage modes
are considered to capture complex mechanisms of material failure. The proposed approach is
based on selection of two types of expected damage mechanisms, namely, interfacial and internal
damage. Interfacial damage is modeled by cohesive elements, while internal damage is analyzed
using progressive damage and failure material definitions. Computational implementation in
Abaqus/Explicit is illustrated on example of a laminated composite structure. Challenges of robust
solutions under impact loading are considered in detail, and efficiency of the implementation is
demonstrated. Finally, validation of the generated results is presented via comparison with
independent LS-DYNA-based solutions.
Keywords: Composite, Heterogeneous Materials, Impact, Fracture, Damage, Crack, Cohesive
Elements, Progressive Failure.

1. Introduction
Robust modeling of complex structures under impact loading is, probably, one of the most
difficult areas in contemporary FEA for practical engineering applications. Efficient
computational analysis of damage processes presents significant challenges especially, for
complex crack patterns. Thus, Abaqus/Explicit seems to be an invaluable tool in modeling of
damage processes under impact load due to its advanced capabilities to capture both the dynamic
nature of loading as well as material failure.
In the case of heterogeneous materials and structures, complexities of internal microstructure and
expected damage patterns can require considerable modeling efforts and/or computationally
expensive solutions. Modeling of such problems, therefore, has a risk to be reduced into either
extremely time-consuming analysis or too simplified statements with limited practical value. Thus,
development of computationally efficient, yet physically adequate modeling approaches for
heterogeneous materials is valuable in engineering decision making process especially when
dealing with complex and large composite and/or hybrid structures.
The objective of this study is to develop a general engineering approach for modeling damage in
complex heterogeneous materials and structures under impact load and demonstrate its efficient

2010 SIMULIA Customer Conference 1

computational implementation on a representative example. The developed methodology is
expected to capture damage initiation, damage propagation, and the complexity of sophisticated
damage network. An additional objective is numerical validation of obtained representative results
via independent FEA solutions.

2. Approach
Damage processes in heterogeneous materials such as laminated composite materials (Figure 1a)
are complex and spans multiple length-scales. Conceptually, damage initiates at the scale of fiber
in the form of matrix cracking, fiber-matrix interfacial debonding, and fiber fracture. These
microscopic damage entities coalesce to form macroscopic damage within the plies such as
longitudinal and transverse ply cracks. In addition, in laminated material system, interfacial
damage may occur between the plies leading to delamination. The damage evolution processes are
inherently coupled, for example, a delamination typically forms when a transverse ply crack tip
reaches the ply interface. Modeling all of these damage processes in an explicit manner is
computationally prohibitive. Thus, typically, individual damage mechanism is considered in
isolation with the assumption that the coupling is not important.
In the present approach, all expected damage patterns at the simulation length-scale are divided
into two basic categories. The first one defines major damage with well-defined spatial locations
and growth directions. In the case of heterogeneous materials, interfacial boundaries are obvious
candidates for such damage (Figure 1b). Since topology and geometry of expected interfacial
damage are perfectly described, they can be easily modeled in FEA by cohesive elements (CE).
The second category defines possible lower length-scale damage mechanisms within the bulk
material. These internal damage (Figure 1c) can cover potentially the entire volume of the
material, and their locations, orientations and/or geometries are not known a priori. In the case of
internal damage, therefore, cohesive elements are less efficient. It is suggested to use Progressive
Damage (PD) models for FEA of the internal damage as shown in Figure 1d. Different definitions
of PD models may be used in principle. For isotropic elasto-plastic materials, ductile or shear PD
material models can be used to describe damage initiation and evolution. For anisotropic linear
elastic materials such as fiber-reinforced polymer matrix composites, PD model (Hashin damage
initiation and evolution) can be applied. For more complex anisotropic non-linear elastic materials
such as ceramic matrix composites a user defined subroutine for PD would be required, since
available Abaqus capabilities do not support them so far (up to v6.9-1).
The described definition of expected damage patterns as two physically-connected but
independently modeled crack networks provides a very convenient way to reduce the complexity
of analyses. Firstly, existing FEA meshes can be used to add both interfacial and internal damage.
Secondly, possible ambiguity of damage growth is significantly mitigated as major interfacial
cracks are easily recognized and taken into account. Finally, the sophisticated nature of multiple
cracks is expected to be predicted, so subjectivism in their analysis will be mitigated. The
approach can be easily applied for a wide range of materials or structures as illustrated, for
example, for fiber-matrix microstructure (Figure 2a) and T-shape composite joint (Figure 2b).

2 2010 SIMULIA Customer Conference

 Expected
interfacial
damage

a b

Cohesive
element
layers
…
Expected
internal
…
Elements with
damages
… progressive
damage and
c d … failure

Figure 1. (a) Schemes of heterogeneous material, (b) expected interfacial damage, (c)
expected internal damage, and (d) FEA implementation

Expected
internal
damages

Expected
interfacial
damages b
a

Figure 2. Examples of damage in heterogeneous materials: (a) micro-structure of fiber-

reinforced composites, and (b) T-shape composite joint.

2010 SIMULIA Customer Conference 3

3. Representative Example
In order to illustrate the developed approach of damage formation and growth in heterogeneous
materials under dynamic load, a representative curved composite panel shown in Figure 3 is
considered. The dimensions of the panel are L = W = 200 mm, H = 25 mm, and h = 5 mm. The
two bottom ends of the panel are assumed to be fixed, and a rigid hemispherical impactor of mass
0.1 kg impacts the center of the panel from below (+Y direction). The panel is made of
unidirectional plies with the layup of [0/90]10 with the fibers in 0 degrees plies aligned along the
global Z-direction and fibers in the 90 degrees plies follow the curvature of the panel. The 20th ply
of the panel comes into the contact with the impactor first (Figure 3). The following representative
material characteristics of plies are used in the simulations: elastic properties (E11 = 150 GPa; E22
= E33 = 10 GPa; ν12 = ν13 = 0.25; ν23 = 0.45; G12 = G13 = 6 GPa; G23 = 4 GPa); strength properties
(XT = XC = 1400 MPa; YT = 40 MPa; YC = 250 MPa; SL = ST = 60 MPa); ply fracture energy (GfT
= GfC = 14 MPa⋅mm; GmT = 0.2 MPa⋅mm; GmC = 7 MPa⋅mm); and density = 1.6e-9 tonnes/mm3.
A FE model of the panel is developed within Abaqus/CAE. The plies are individually modeled so
that cohesive elements may be inserted at each of the nineteen interfaces between the plies. The
plies are modeled using continuum shell (SC8R) or solid (C3D8R) elements with one element
through the thickness for each ply. The number of elements along the X- and Z- directions is 80.
Furthermore, the mesh is biased such that higher mesh density is obtained in the center where the
impactor hits the panel (Figure 4). The mesh is selected based on a limited mesh-sensitivity
analysis and overall computational resources.
In the models that allow for interfacial damage mechanism, zero thickness CEs (COH3D8) are
embedded between each ply. The CEs are such that they share nodes with the elements in the
layers above and below them. The cohesive constitutive behavior is specified via triangular
traction-separation relationship. The parameters specifying the traction-separation relationships
are: normal and shear mode elastic stiffness, En = Es = 2.0e6 MPa, normal mode interface strength,
Tn = 50 MPa, shear mode interface strength, Ts = 80 MPa, normal mode fracture energy, Gnc = 0.3
MPa⋅mm, and shear mode fracture energy, Gsc = 1.5 MPa⋅mm. The mixed mode fracture behavior
in the CEs is specified using Benzeggagh-Kenane (BK) criterion with power, η = 1.8. The
interaction between the impactor and the panel as well as between the plies of the panel after
interfacial failure is specified through the general contact definition in Abaqus/Explicit.
Furthermore, all contact interactions are considered frictionless. The analyses are conducted in
double precision using Abaqus/Explicit code running on a Linux x86-64 machine. Two different
impact velocity, 10 m/s and 50 m/s, are considered for the analyses.

4. Damage Formation and Growth under Dynamic Loading

Conditions
Under impact loading condition both interfacial and internal damage mechanisms occur
simultaneously within the material or structure. However, in many circumstances only interfacial
damage mechanism is modeled and the internal damage mechanism is ignored or vice-versa,
without proper justification. Thus, in the following, two cases are considered. In the first case, the

4 2010 SIMULIA Customer Conference

 y
x

z 90
0

H
h
Interfaces: st
1 ply
W 18/17
…
L 19/18 th
20/19 19 ply
th
20 ply

Figure 3. Scheme of considered example and local segment of corresponding FEA model.

Solid or
shell
elements

…
Cohesive
elements

Rigid
impactor Expected pattern of
interfacial damages
y
x

Figure 4. Details of FEA implementation.

internal damage modes are ignored and only interfacial damage mode is considered. Subsequently,
both the damage modes are considered together in order to understand the coupling between the
damage mechanisms and to assess whether ignoring certain damage modes in the analyses is
justified.

2010 SIMULIA Customer Conference 5

4.1 Single Damage Mechanism
In this case the focus is on a single damage mechanism in isolation. This mechanism includes only
interfacial damage processes between the plies. Thus, the plies are modeled as linear orthotropic
elastic material without any internal damage evolution within them; the entire damage process
consists of interfacial damage between the plies. The initiation and evolution of interfacial damage
between the layers is modeled using cohesive elements in Abaqus/Explicit. Zero thickness CEs
with matching nodes are embedded between each layer. Damage within CEs initiate once the
maximum traction-based damage initiation criterion is met. Upon damage initiation the traction
decreases until it reaches zero. Once the traction has become zero or, correspondingly, the damage
parameter has a reached 1.0, an interfacial separation (delamination) has occurred at that material
point.
Figure 5 shows the time history of damage within the CEs at the interface closest to the impactor
(interface between plies 19 and 20) for the impact velocity of 10 m/s. (the same legends for SDEG
are applied to other figures, unless otherwise noted). It is seen that a small delamination is formed
fairly early in the loading history, but it does not grow significantly. The damage in the
surrounding CEs continues to grow with time until it reaches a steady state. However, it remains
below 1.0 signifying that delamination has not formed. Similarly, at the impact velocity of 50 m/s
the damage within the same interface was found to be considerably larger, as expected. For both
impact velocities, the interfacial damage reduces in size as one move to the ply interfaces away
from the interface 20/19. The time-history of the impact force for both impact velocities is shown
in Figure 6. In these figures, a comparison is made with the cases where both interfacial and
internal damage mechanisms are ignored (i.e., a purely elastic solution) and where both damage
mechanisms are considered (to be discussed next). It is evident that interfacial damage mechanism
reduces the peak impact force when compared with purely elastic solution that ignores any
damage mechanisms.
4.2 Multiple Damage Mechanisms
The case of multiple damage mechanisms is considered next. In this case, the internal damage
mechanisms occurring within the plies (matrix cracks, fiber breakage etc.) are modeled using the
progressive damage model for fiber-reinforced composite materials in Abaqus/Explicit. In this
model a Hashin failure criterion is used for damage initiation in various modes such as
longitudinal tensile and compressive (fiber direction) and transverse tensile and compressive
(matrix direction). Once the damage initiation has occurred an energy-based criterion is used for
damage evolution in various modes. In Abaqus/Explicit complete failure of a material point occurs
when fiber mode damage variables reach a user-specified maximum value (or a default value of
1.0). Furthermore, the element is considered fully damaged and removed from the mesh once the
above criterion is satisfied at all the section points at one location within the element [1].
In this analysis a number of damage parameters are of interest. For interfacial damage the scalar
damage parameter SDEG indicates the degree of damage within the CEs at ply interfaces. In
addition, damage parameters representing internal damage within the plies are of interest. Here the
damage parameter DAMAGESHR is shown for illustration. This damage parameter combines the
different fiber and matrix damage modes and can be considered as an effective scalar internal
damage parameter. Figure 7 gives the time history of interfacial damage parameter at ply interface

6 2010 SIMULIA Customer Conference

 Time = 0 Time = 0.5e-4 Time = 1.0e-4

Time = 1.5e-4 Time = 2.0e-4 Time = 2.5e-4

Time = 3.0e-4 Time = 4.0e-4 Time = 5.0e-4

z
x
Time = 6.0e-4 Time = 10.0e-4

Figure 5. Time evolution of interfacial damage parameter (SDEG) at ply interface 20/19
[impact velocity, v = 10 m/s; without internal damage within the plies]

2010 SIMULIA Customer Conference 7

 a b

Figure 6. Time history of overall impact force for impact velocity, v = 10 m/s (a) and 50 m/s
(b): (1) elastic solution, (2) only interfacial damage, and (3) interfacial and internal ply
damage

20/19 for the case of 10 m/s impact velocity. For the same impact velocity and at the end of
simulation, the interface damage parameter at various ply interfaces is shown in Figure 8. The
time history of internal damage parameter in the 19th ply is shown in Figure 9 and the value of the
same parameter within various plies at the end of the simulation is shown in Figure 10. Note that
Figure 9 is a zoomed-in view of the impact region.
A number of observations can be made from these figures. First, the interface damage and
delamination patterns are similar to those observed in the case where no internal damage
mechanisms were considered (Figure 5). Secondly, the size of delamination and associated
interface damage is larger when compared with the case without internal damage. It may also be
seen that some perforation of plies has occurred in the impact region for the impact velocity of 10
m/s. The ply damage is, however, more localized than the interfacial damage. Also note that both
interface damage and ply internal damage reduce in size and intensity as one traverse away from
the impact location in the +Y direction. Furthermore, the internal damage reduces more rapidly in
the Y-direction than the interfacial damage.
The time history of the impact force for both impact velocities is superimposed with the earlier
results in Figure 6. Note that for the case of 50 m/s impact velocity the analysis did not complete
due to excessive distortion of the continuum shell elements in the close proximity of the impact
site. For some reason one of the section points of the excessively distorted element did not reach
the condition for complete damage, preventing that element from deletion. In any case, it may be
noted that for both impact velocities consideration of multiple damage mechanisms leads to a
lower peak impact force when compared with the case where only interfacial damage mechanism
was considered. This is intuitively expected as multiple damage mechanisms allow for more
damage pathways leading to larger damage dissipation and overall compliant structure.

8 2010 SIMULIA Customer Conference

 Time = 0 Time = 0.5e-4 Time = 1.0e-4

Time = 1.5e-4 Time = 2.0e-4 Time = 2.5e-4

Time = 3.0e-4 Time = 3.5e-4 Time = 4.0e-4

Time = 4.5e-4 Time = 5.0e-4 Time = 5.852e-4

Figure 7. Time evolution of interfacial damage parameter (SDEG) at ply interface 20/19
[impact velocity, v = 10 m/s; with interfacial and internal ply damage]

2010 SIMULIA Customer Conference 9

 20/19 interface 19/18 interface 18/17 interface

17/16 interface 16/15 interface 15/14 interface

14/13 interface 13/12 interface 12/11 interface

11/10 interface 10/9 interface 9/8 interface

Figure 8. Distribution of interfacial damage parameter (SDEG) at various ply interfaces at

time = 5.852e-4 s [impact velocity, v = 10 m/s; with interfacial and internal ply damage]

10 2010 SIMULIA Customer Conference

 Time = 0 Time = 0.5e-4

Time = 1.0e-4 Time = 1.5e-4

Time = 2.0e-4 Time = 3.0e-4

Time = 4.0e-4 Time = 5.852e-4

Figure 9. Time evolution of internal ply damage parameter (DAMAGESHR) in ply 19

[impact velocity, v = 10 m/s; with interfacial and internal ply damage]

2010 SIMULIA Customer Conference

11
 th th th
20 ply 19 ply 18 ply

th th th
17 ply 16 ply 15 ply

th th th
14 ply 13 ply 12 ply

z
th th x
11 ply 10 ply

Figure 10. Distribution of internal ply damage parameter (DAMAGESHR) in ply 19 at time
= 5.852e-4 s [impact velocity, v = 10 m/s; with interfacial and internal ply damage]

12 2010 SIMULIA Customer Conference

The effect of impact velocity on interfacial damage for the case where internal ply damage is not
considered can be noted from Figure 11. As expected, higher velocity would lead to more
interface damage. For the case where internal ply damage is included in the analysis, it is expected
that the interface damage would follow similar trend.

Only Interfacial Interfacial and Internal

Damage Damage Considered

v = 50 m/s
t = 0.5e-4 s

a b z
x

v = 10 m/s
t = 1e-3 s

c d

Figure 11. Interfacial damage parameter (SDEG) at ply interface 20/19 predicted without
and with internal ply damage.

For the impact velocity of 10 m/s the interfacial damage size at interface 20/19 at the end of the
simulation for the case where internal damage is not considered is compared against the case
where internal ply damage is considered in Figure 11c,d. It is seen that consideration of internal
damage within the plies leads to higher interfacial damage. This result is counter-intuitive as one
might expect that ignoring internal damage mechanisms in plies would lead to higher interfacial
damage as more energy is available for forming interfacial damage in this case. However, the
simulation results point to a synergism between the interfacial and the internal damage
mechanisms: by considering internal damage mechanisms in the plies leads to more interfacial
damage. A similar conclusion can be noted for higher impact velocity (50 m/s) as shown in Figure
11a,b.

2010 SIMULIA Customer Conference

13
5. Effect of Element Selection and Comparison with LS-DYNA
Solution
In computational simulations involving commercial FEA codes the analyst is faced with numerous
modeling choices and options. These choices are usually based on analyst’s experience, time
constraints, and available capabilities within the codes, amongst others. For example, in the
present study the choice of continuum shell elements to model the plies was dictated by the lack of
a built-in progressive damage material model for fiber-reinforced composite materials for 3D solid
elements in Abaqus. Even though a damage model for solid element is available as a user material
model (VUMAT) in Abaqus/Explicit, it is different than the Hashin’s progressive damage model.
However, a limited study was conducted on the effect of the element type (continuum shell, SC8R
versus solid, C3D8R) for the case where only interface damage was considered and the results will
be discussed shortly.
Another aspect of computational analyses that is crucial in any engineering environment is
robustness and correctness of the computational solution. In this work the modeling approach and
results are assessed by comparing Abaqus/Explicit and LS-DYNA [2] solution for the case where
only interfacial damage mechanism is allowed in the analysis.

a b

Figure 12. Time history of overall impact force for impact velocity, v = 10 m/s (a) and 50 m/s
(b) when only interfacial damage considered: (1) Abaqus continuum shell, (2) Abaqus solid,
(3) LS-DYNA thick shell, and (4) LS-DYNA solid elements

The results from both these studies are compared in Figure 12 corresponding to impact velocity of
10 m/s and 50 m/s, respectively. It is seen that for v = 10 m/s, both Abaqus solid element and LS-
DYNA solid element leads to almost identical force-time response. The response of continuum
shell (thick shell in LS-DYNA terminology) differs slightly from the solid response. However, all

14 2010 SIMULIA Customer Conference

the response characteristics still appear reasonable; the reason for the difference could be related to
the mesh density, differences in bending behavior of solid versus thick shell, etc. For v = 50 m/s
LS-DYNA solid and thick shell response are almost identical, whereas, Abaqus solid and
continuum shell response show slight difference even though the peak force is identical. At this
stage the reasons for these differences are not clear. It should be worth mentioning that despite the
minor differences in the overall force-displacement response, the interfacial damage patterns are
similar in these cases, giving confidence in the simulation results.

6. Conclusions
An efficient methodology is suggested and demonstrated for impact modeling of complex
heterogeneous materials and structures. Reliable computational validation of the described
methodology is demonstrated by comparison with corresponding independent LS-DYNA
solutions as well as using different types of elements (continuum shell, solid elements). It is shown
that the coupling between the internal and the interfacial damage mechanisms can be very
important in understanding and characterization of actual material failure. The methodology seems
to be a convenient way for numerous engineering applications involving impact behavior.

7. References
1. Abaqus Analysis User’s Manuals, v6.9, Dassault Systèmes SIMULIA Corp., 2009.
2. LS-DYNA User’s Manuals, v971, Livermore Software Technology Corp., 2007.

8. Acknowledgement
The authors thank United Technologies Research Center for permission to publish this paper.

2010 SIMULIA Customer Conference

15