NASA/CR-2006-214526

User-Defined Material Model

for Progressive Failure Analysis

Norman F. Knight, Jr.

General Dynamics – Advanced Information Systems, Chantilly, Virginia

December 2006
 The NASA STI Program Office . . . in Profile

Since its founding, NASA has been dedicated to the • CONFERENCE PUBLICATION. Collected
advancement of aeronautics and space science. The papers from scientific and technical
NASA Scientific and Technical Information (STI) conferences, symposia, seminars, or other
Program Office plays a key part in helping NASA meetings sponsored or co-sponsored by NASA.
maintain this important role.
• SPECIAL PUBLICATION. Scientific,
The NASA STI Program Office is operated by technical, or historical information from NASA
Langley Research Center, the lead center for NASA’s programs, projects, and missions, often
scientific and technical information. The NASA STI concerned with subjects having substantial
Program Office provides access to the NASA STI public interest.
Database, the largest collection of aeronautical and
space science STI in the world. The Program Office is
• TECHNICAL TRANSLATION. English-
also NASA’s institutional mechanism for
language translations of foreign scientific and
disseminating the results of its research and
technical material pertinent to NASA’s mission.
development activities. These results are published by
NASA in the NASA STI Report Series, which
Specialized services that complement the STI
includes the following report types:
Program Office’s diverse offerings include creating
custom thesauri, building customized databases,
• TECHNICAL PUBLICATION. Reports of organizing and publishing research results ... even
completed research or a major significant phase providing videos.
of research that present the results of NASA
programs and include extensive data or For more information about the NASA STI Program
theoretical analysis. Includes compilations of Office, see the following:
significant scientific and technical data and
information deemed to be of continuing
• Access the NASA STI Program Home Page at
reference value. NASA counterpart of peer-
http://www.sti.nasa.gov
reviewed formal professional papers, but having
less stringent limitations on manuscript length
• E-mail your question via the Internet to
and extent of graphic presentations.
[email protected]
• TECHNICAL MEMORANDUM. Scientific
• Fax your question to the NASA STI Help Desk
and technical findings that are preliminary or of
at (301) 621-0134
specialized interest, e.g., quick release reports,
working papers, and bibliographies that contain
minimal annotation. Does not contain extensive • Phone the NASA STI Help Desk at
analysis. (301) 621-0390

• CONTRACTOR REPORT. Scientific and • Write to:

technical findings by NASA-sponsored NASA STI Help Desk
contractors and grantees. NASA Center for AeroSpace Information
7115 Standard Drive
Hanover, MD 21076-1320
NASA/CR-2006-214526

User-Defined Material Model

for Progressive Failure Analysis

Norman F. Knight, Jr.

General Dynamics – Advanced Information Systems, Chantilly, Virginia

National Aeronautics and

Space Administration

Langley Research Center Prepared for Langley Research Center

Hampton, Virginia 23681-2199 under NASA Order No. NNL06AA48T

December 2006
 The use of trademarks or names of manufacturers in the report is for accurate reporting and does not
constitute an official endorsement, either expressed or implied, of such products or manufacturers by the
National Aeronautics and Space Administration.

Available from:

NASA Center for AeroSpace Information (CASI) National Technical Information Service (NTIS)
7115 Standard Drive 5285 Port Royal Road
Hanover, MD 21076-1320 Springfield, VA 22161-2171
(301) 621-0390 (703) 605-6000
 User-Defined Material Model
for Progressive Failure Analysis

Norman F. Knight, Jr.

General Dynamics – Advanced Information Systems
Chantilly, VA

Abstract
An overview of different types of composite material system architectures and a brief review of
progressive failure material modeling methods used for structural analysis including failure
initiation and material degradation are presented. Different failure initiation criteria and material
degradation models are described that define progressive failure formulations. These progressive
failure formulations are implemented in a user-defined material model (or UMAT) for use with
the ABAQUS/Standard1 nonlinear finite element analysis tool. The failure initiation criteria
include the maximum stress criteria, maximum strain criteria, the Tsai-Wu failure polynomial,
and the Hashin criteria. The material degradation model is based on the ply-discounting
approach where the local material constitutive coefficients are degraded. Applications and
extensions of the progressive failure analysis material model address two-dimensional plate and
shell finite elements and three-dimensional solid finite elements. Implementation details and use
of the UMAT subroutine are described in the present paper. Parametric studies for composite
structures are discussed to illustrate the features of the progressive failure modeling methods that
have been implemented.

1
ABAQUS/Standard is a trademark of ABAQUS, Inc.

i
This page left blank intentionally.

ii
 Table of Contents
Abstract ............................................................................................................................................ i
Table of Contents........................................................................................................................... iii
Introduction..................................................................................................................................... 1
Composite Materials ....................................................................................................................... 2
Laminated Composites ............................................................................................................... 2
Woven-Fabric Composites ......................................................................................................... 3
Carbon-Carbon Composites........................................................................................................ 5
Constitutive Models ........................................................................................................................ 7
Two-Dimensional Material Model ............................................................................................. 8
Three-Dimensional Material Model ......................................................................................... 10
Implications for UMAT ............................................................................................................ 11
Failure Initiation Criteria .............................................................................................................. 13
Maximum Stress Criteria .......................................................................................................... 15
Maximum Strain Criteria .......................................................................................................... 15
Tsai-Wu Failure Polynomial..................................................................................................... 16
Hashin Failure Criteria.............................................................................................................. 18
Damage Progression Models ........................................................................................................ 19
Ply-Discounting Approach ....................................................................................................... 19
Internal State Variable Approach ............................................................................................. 21
UMAT Implementation ................................................................................................................ 22
Numerical Results and Discussion................................................................................................ 27
Demonstration Problems........................................................................................................... 27
Open-Hole-Tension Coupons ................................................................................................... 30
Concluding Remarks..................................................................................................................... 34
References..................................................................................................................................... 36
Appendix A – Description of ABAQUS UMAT Arguments....................................................... 45
Appendix B – LS-DYNA MAT58 Input Data.............................................................................. 48
Appendix C – MLT Continuum-Damage-Mechanics Model....................................................... 50
Appendix D – Representative ABAQUS/Standard Input File for the Demonstration Problems . 57

iii
This page left blank intentionally.

iv
 Introduction
Advanced composite structures are a challenge to analyze. This challenge is driven by the
evolution of new material systems and novel fabrication processes. Analytical procedures
available in commercial nonlinear finite element analysis tools for modeling these new material
forms are often lagging behind the material science developments. That is to say, analytical
models implemented in the analysis tools must frequently be “engineered” for new material
systems through the careful extension, within the known limits of applicability, of existing
material models. When the materials technology matures sufficiently, verified analytical models
can be developed, validated, and become available in the commercially available tools. In the
interim, some commercial finite element analysis tool developers have responded to user needs
by providing a capability to embed user-defined special-purpose finite elements and material
models.
This need for a capability to explore user-defined material models exists – particularly for
composite structures. Composite material systems have unique characteristics and capabilities
not seen in metals. Characteristics such as being brittle rather than ductile, being directional in
stiffness and strength, and being open to different fabrication architectures. These characteristics
enable advanced design concepts including structural tailoring, multifunctional features, and
performance enhancements.
Another consideration for the material model developer is the intended application and target
finite element analysis tool. Applications involving an explicit transient dynamic nonlinear tool
impose different requirements on the material model developer than applications involving
implicit (or quasi-static) nonlinear tools. In explicit formulations, only the current stress state is
needed to evaluate the current internal force vector in order to march the transient solution
forward in time. In implicit (or quasi-static) formulations, the current stress state and a
consistent local tangent material stiffness matrix are needed to form the internal force vector for
the residual computation and to form the tangent stiffness matrix for the Newton-Raphson
iteration procedure.
The objective of the present paper is to describe a user-defined material model for progressive
failure analysis that is applicable to composite laminates and focuses on implicit, quasi-static
applications. This material model accounts for a linear elastic-brittle bimodulus material
behavior as seen in some refractory composites such as reinforced carbon-carbon material used
on the wing leading edges of the Space Shuttle Orbiter. Several options for detecting failure
initiation and subsequent damage progression are provided in this user-defined material model.
The material model can be applied even when only laminate-level or “apparent” material data
rather than lamina-level data are available. Implementation of the material model for progressive
failure analysis is demonstrated using the user-defined material model or UMAT-feature
provided in ABAQUS/Standard nonlinear finite element analysis tool [1]. A description of the
UMAT subroutine calling arguments based on the ABAQUS/Standard documentation is given in
Appendix A for convenience.
The present paper is organized in the following way. First, a brief overview of composite
materials is presented to define terminology. Second, the basic equations for constitutive
modeling in two- and three-dimensions are described. Next, the failure initiation and material
degradation models proposed for use with laminated composites are defined. Then, the
implementation of the material model for traditional progressive failure analyses is described as
1
a UMAT subroutine for ABAQUS/Standard. Numerical results and discussion are presented for
uniaxial demonstration problems using two materials and for 16-ply open-hole-tension coupons
using two different stacking sequences. Concluding remarks are given in the final section of the
paper.

Composite Materials
Composite materials can be defined as materials comprised of more than one material, such as a
bimetallic strip or as concrete (e.g., see Refs. [2-8]). However, composite materials are more
commonly defined as materials formed using fiber-reinforced materials combined with some
matrix material (e.g., graphite fibers embedded in an epoxy resin system). Typical fiber
materials include graphite and glass, and typical matrix materials include polymer, metallic,
ceramic, and phenolic materials. Composite material systems are available as unidirectional or
textile laminas that are used to design laminated composite structures. Subsequent processing
can be done to create refractory composites where the matrix material is converted into its basic
constitute material (i.e., phenolic to basic carbon). Background material and issues for
laminated, woven-fabric, and carbon-carbon composites are briefly described in the next
subsections.

Laminated Composites
Laminated composite structures are formed by stacking (or laying up) unidirectional lamina with
each layer having different properties and/or different orientations in order to form a composite
laminate of a given total thickness. The elastic analysis of such composite structures is well
developed [2-9]. Typically composite structures are thin structures that are modeled using two-
dimensional plate or shell elements formulated in terms of stress resultants and based on an
assumed set of kinematical relations. Hence, pre-integration through the shell thickness is
performed and the constitutive relations are in terms of resultant quantities rather than point
stresses and strains. Integrating the constitutive relations through the thickness generates the so-
called “ABD” stiffness coefficients where the “A” terms denote membrane, the “B” terms denote
membrane-bending coupling, and the “D” terms denote bending stiffness coefficient matrices,
respectively. However in some cases, composite structures are relatively thick and require three-
dimensional modeling and analysis.
Material characterization of laminated composite structures has been demonstrated to be quite
good, and failure prediction models for damage modes associated with fiber failure in polymeric
composites have also been successful. In laminated composite structures, fiber-dominated failure
modes (e.g., axial tension) are better-understood and analyzed than matrix-dominated failure
modes (e.g., axial compression, transverse tension). Methods to predict failure initiation (e.g., see
Refs. [10-11]) and to perform material degradation remain active areas of research. Material
degradation can be performed using a ply-discounting approach or an internal state variable
approach based on continuum damage mechanics. Most composite failure analysis methods
embedded within a finite element analysis tool perform a point-stress analysis, evaluate failure
criteria, possibly degrade material properties, and then continue to the next solution increment
(e.g., see Refs. [13-32]). Application of progressive failure analysis methods to postbuckled
composite structures continues to challenge analysts due to the complexity and interaction of
failure modes with the nonlinear structural analysis strategies (e.g., see Refs. [16-21, 24, 29, 30]).

2
Damage associated with matrix cracking, the onset of delaminations, and crack growth is very
difficult to predict for laminated composite structures. Analytical methods for predicting matrix
cracking, delamination, and crack growth are still immature although current research associated
with decohesive models appears promising (e.g., see Refs. [30, 33-35]). Failure analysis
procedures based on the use of overlaid two-dimensional meshes of shell finite elements [24],
the use of computed transverse stress fields from the equilibrium equations [16], and the use of
decohesion or interface elements [30, 33-35] have met with limited success.

Woven-Fabric Composites
Woven-fabric composite structures can be formed by laying up woven-fabric layers or by
weaving through the total thickness. Fabric layers usually have interwoven groups of fibers in
multiple orientations. Plain- and satin-weave fabrics have warp and weft (0-degree and 90-
degree, respectively) yarns only. Woven-fabric composites are different from unidirectional-
based laminated composites, and different analysis methods are needed. Often, however, plain-
and satin-weave fabrics of a given thickness are modeled as an equivalent set of two
unidirectional layers (i.e., [0/90]) where each unidirectional layer has half the fabric layer
thickness. Alternatively, a single fabric layer is modeled as an “equivalent” single layer of the
same thickness and having equal in-plane longitudinal and transverse moduli (E11=E22). This
equivalence is a modeling simplification and does not accurately represent the micromechanics
of the fabric composite.
Woven-fabric composites are two-dimensional composites, which exhibit three-dimensional
characteristics, and exhibit good mechanical properties and better impact resistance than
laminated composites formed from unidirectional tape. Material characterization of woven-fabric
composites is also well understood for predicting equivalent elastic material properties based on
three-dimensional unit-cell models (e.g., see Refs. [36-38]). Nonlinear material behavior is more
difficult to predict due in part to the complexity of the fabric-weave geometry and its influence
on failure mechanisms. Three-dimensional models at the same scale as the fabric architecture,
such as those shown by Whitcomb [39] and Glaessgen et al. [40], are required to study woven-
fabric material behavior. Aitharaju and Averill [41] presented an alternate approach of
comparable accuracy for estimating the effective stiffness properties that is more computational
efficient than three-dimensional analysis models.
Ishikawa and Chou [42] present three two-dimensional models for analyzing the stiffness and
strength of woven-fabric composites but do not address damage modeling. Results from these
elastic models are compared with test and reported to be in good agreement. The three models
are the mosaic model, the fiber undulation model and the bridging model. Each model has
benefits and capabilities for woven-fabric composites. However, they are not available in
commercial finite element codes.
Raju and Wang [43] present a formulation for woven-fabric composites based on classical
laminate theory. The extension to woven-fabric composites determines the global engineering
properties based on the fiber and matrix constituent material properties and on the fabric
geometry. Good agreement is shown for the mechanical properties between results obtained
using their method and test results.
Naik [44] developed a general-purpose micro-mechanics analysis tool called TEXCAD [45] for
textile composites based on the repeating unit-cell approach. TEXCAD can be used to predict

3
thermoelastic material properties, damage initiation and growth, and strength for different textile
architectures. This tool enables a systematic evaluation of different textile configurations and
has been correlated with experimental data. Correlation with elastic material data is better than
correlation with experimentally determined damage and failure metrics.
Chung and Tamma [46] describe various methods used with woven-fabric composite to
determine bounds on elastic stiffness terms. Various approaches are described and compared.
Their goal is to provide a bridge between methods to generate stiffness terms based on unit-cell
models and macroscopic methods based on a homogenization of the material. While attractive,
this approach is still at the research level with limited applications demonstrated.
Swan [47]
Kwon and Altekin [48] describe a multi-level, micro/macro-approach for woven-fabric
composite plates. A computational procedure for their multilevel approach is described and
demonstrated using standard problems. Comparisons for stiffness, strength and failure are
reported to be in good agreement with other predictions. Direct comparison for postbuckled
panels is needed.
Nonlinear material models including failure analysis for textile composites are not well
developed. In 1995, Cox [49] presented a concise summary of different failure models for textile
composites. Within a finite element analysis at a structural level, the failure analysis and damage
progression models are essentially extensions of techniques developed for laminated composite
structures. Damage modeling for textile composites has been the subject of much research (e.g.,
see Refs. [50-61]. Much of the early work has focused on failure analysis and damage
accumulation for unit-cell models. Extending these unit-cell concepts to a macroscopic
structural level has received attention as well (e.g., see Refs. [52, 56-60]).
Blackketter et al. [52] incorporate the three-dimensional elastic modeling and analysis features
with an inelastic constitutive response for modeling damage propagation in woven-fabric
composites. The fabric weave geometry is explicitly modeled with three-dimensional finite
elements and a maximum stress theory is evaluated to determine the onset of local material
failure. Degrading the elastic mechanical properties according to one of the six different failure
modes imposes material degradation. Good correlation is shown between analysis and test for
specimen-level tests.
Schwer and Whirley [58] analyzed a three-dimensional woven textile composite panel subjected
to impact loading using the explicit nonlinear transient dynamics finite element tool DYNA3D
from the Lawrence Livermore National Laboratory. Laminate properties are available rather
than lamina data, and hence, an alternate material modeling approach based on a two-
dimensional resultant shell element formulation is developed. Failure initiation is detected using
the Tsai-Wu failure polynomial [10] using the outer fiber stresses obtained from the stress
resultants of the shell element. Damage modeling is accomplished using a damage parameter
that varies from zero (no damage) to unity (complete damage) and multiplies the stress
component.
Tabiei et al. [59, 60] describe additional examples of finite element modeling and analysis of
textile composite structures. A computational micro-mechanical model for woven composites is
presented and implemented within the explicit nonlinear dynamics finite element analysis tool

4
LS-DYNA2 (commercial version of DYNA3D from Lawrence Livermore National Laboratory).
The constituent material properties and fabric geometric definition are accounted for in the
material definition. The failure model is based on the maximum principal stress. Stiffness
degradation follows the Blackketter et al. [52] approach. Nonlinearity of the in-plane shear
stress-shear strain response is also included. Again, good results are presented for a validation
problem.

Carbon-Carbon Composites
Carbon-carbon composites consist of carbon fibers reinforced in a carbon matrix. Fitzer and
Manocha [61] provide further background information. Initially carbon-carbon composites were
used only on ablative structures (e.g., nozzles, re-entry surfaces) on aerospace systems. Later
carbon-carbon composites were recognized as a candidate material for the Space Shuttle
program. They exhibited lightweight, high thermal-shock resistance, low coefficient of thermal
expansion, high stiffness, and retention of strength at high temperatures. These properties make
carbon-carbon composites a good candidate for high-temperature structural applications such as
the wing leading edge of the Space Shuttle Orbiter. While carbon-carbon composites have found
applications in other areas, they are still predominantly used by in the aerospace industry.
The basic composition for a carbon-carbon composite is a laminate of graphite-impregnated
fabric, further impregnated with phenolic resin and layered one ply at a time, in a unique mold
for each part, then cured, rough-trimmed, and inspected. The part is then packed in calcined
coke and fired in a furnace to convert the phenolic resin to its basic carbon, and its density is
increased by one or more cycles of furfuryl alcohol vacuum impregnation and firing. This multi-
stage process of densification and pyrolysis converts the phenolic to the carbon matrix and
increases the mass density and strength of the carbon-carbon part.
The carbon-carbon substrate carries the load and exhibits good strength and stiffness at high
temperatures. It also exhibits a low coefficient of thermal expansion. However it is subject to
oxidation and exposed surfaces are typically coated. Choice of coating material depends on the
application. For the Space Shuttle application, the outer layers of the carbon substrate are
converted into a layer of silicon carbide. The silicon-carbide coating protects the carbon
substrate from oxidation. Unfortunately, craze cracks form in the outer layers because the
thermal expansion rates of silicon carbide and the carbon substrate differ. As a result, coated
carbon-carbon composites exhibit different stress-strain responses in tension and compression, as
indicated in Figure 1. The differences in initial elastic moduli and ultimate strength allowable
values are subsequently denoted using the notation E 0t,c and X t,c , respectively. Such a material
is referred to as a bimodulus material. Basic constitutive relations for a bimodulus material are
presented by Ambarysumyan [62], Tabaddor [63], Jones [64], Bert [65], Reddy and Chao [66],
and Vijayakumar and Rao [67].
Often engineering analysis models of the carbon-carbon structure are employed to examine
structural-level response characteristics. Detailed micromechanics analyses correlated with test
data are rare. Limited information is available on the fracture behavior, damage mechanisms,
and failure modes of carbon-carbon composites. However, some basic work on carbon-carbon

2
LS-DYNA is a registered trademark of Livermore Software Technology Corporation (LSTC).

5
composites is available in the open literature (e.g., see Refs. [61, 68-81]). For example, Piat and
Schnack [78] proposed a hierarchical modeling approach to obtain bounds on the mechanical
properties, Hatta et al. [80] determined the transition from Mode I to Mode II fracture for
compact-tension specimens was related to the cross-ply volume fractions, and early work [82]
from the Space Shuttle program examined the influence of substrate oxidation on the mechanical
properties of reinforced carbon-carbon material.
Reinforced carbon-carbon or RCC is a two-dimensional plain-weave carbon-carbon textile
composite with its outer layers coated to prevent oxidation. The RCC material is used on the
wing leading edge (WLE) structural subsystem and two other locations on the Space Shuttle
Orbiter [83]. The WLE RCC panels provide aerodynamic performance and thermal protection
for the wing internal structure. Mechanical properties for the refractory composite constituents
(carbon-carbon substrate and coating materials) are generally not available. Basic RCC material
characterization is described by Curry et al. [84]. Data are given for the “apparent” mechanical
properties (that is, for the laminate rather than for the carbon-carbon substrate and/or coating
layer). Wakefield and Fowler [82] reported a relationship between the substrate and coating
properties. They postulated that the coating modulus ECoat is a scalar (α) multiple of the
substrate modulus ESubstrate. This factor α was a function of whether the material is loaded in
tension or compression:
ECoat = αESubstrate (1)
where α equals 0.33 for tension and 1.37 for compression. Hence, even if lamina properties
were available, the coating properties still exhibit a bimodulus characteristic that is carried over
to the laminate. In addition, the in-plane shear modulus G will be assumed to be independent of
the sign of the shear strain, and the in-plane shear stress-shear strain relationship will be assumed
to be linear. A nonlinear in-plane normal stress behavior as well as Hahn-type in-plane shear
nonlinearity (see Refs. [85, 86]) can be incorporated as part of a UMAT material model.
A bimodulus elastic orthotropic material model for a modern composite material is not readily
available in most commercial general-purpose finite element tools. In some cases, various
options can be combined or semi-automated procedures developed to address bimodulus
composite material systems. For example in a linear stress analysis, an iterative process to
assign material properties can be used that defines the appropriate material properties based on
the local principal strains. Such an iterative process is typically manually performed or may be
semi-automated, and the process needs to be repeated for each different loading case since the
local strain state changes. For a nonlinear analysis that accounts for material degradation and
damage, a semi-automatic material iteration approach is not viable. In such cases, user-defined
material models can be developed for certain commercial finite element tools (e.g.,
ABAQUS/Standard). Alternatively, “engineering” models (e.g., use only the tensile modulus,
use an average modulus, or some other engineering approximation for the material response) can
be developed that approximate the mechanical behavior provided these approximations are
defined and limitations are understood by the end-users (or stakeholders) of the analysis results.
As part of the Return-to-Flight effort, impact damage assessments were performed for the Space
Shuttle Orbiter’s wing leading edge (e.g., see Refs. [87-99]). The NASA-Boeing impact damage

6
threshold team investigated the debris impact damage threshold3 of the RCC panels (see Refs.
[89, 91-99]). Their studies used the LS-DYNA explicit nonlinear transient dynamics finite
element analysis tool [100] to simulate the impact event and subsequent structural response of
the RCC panel. In the LS-DYNA analyses, the basic material model being used to model RCC
material is Material Type 58 (or simply MAT58), which is described in Refs. [101, 102].
MAT58 is based on the continuum-damage-mechanics model of Matzenmiller, Lubliner, and
Taylor [101], and the LS-DYNA implementation and extensions are based on Schweizerhof,
Weimar, Münz, and Rottner [102]. MAT58 is also referred to as MAT58 and is only available
for selected two-dimensional shell finite elements in LS-DYNA – not three-dimensional solid
finite elements. A summary description of the user input for MAT58 based on the LS-DYNA
Users Manual [100] is presented in Appendix B for convenience.
The continuum-damage-mechanics (CDM) model of Matzenmiller, Lubliner, and Taylor [101]
for laminated, unidirectional, orthotropic, polymeric composites uses a statistical distribution of
defects and strength, Hashin–like failure criteria [12] as a “loading criteria”, and three state
variables for damage evolution. This model is often referred to as the MLT model.
Schweizerhof et al. [102] described an extension of this model for fabric materials based on
equating failure modes and strength allowable values associated with the two in-plane directions.
Applications of the MAT58 material model to laminated composite plate impact problems are
available in Refs. [102-106]. Williams and Vaziri [103] apparently had an early implementation
of the MLT model for evaluation. Since then, variations of the MLT CDM material model have
been under development and reported in the literature.
The present paper describes the development of a user-defined material model (or UMAT) for
the ABAQUS/Standard nonlinear finite element analysis tool [1] to represent the basic nonlinear
material response including progressive failure. This material model is consistent with the
available material data and accounts for its unique bimodulus characteristic. The philosophy
adopted for this effort is similar to that used for the material model developed by Schwer and
Whirley [58] for a three-dimensional woven textile composite including damage and failure that
exploited only the known “apparent” values of the laminate. The next sections describe the
constitutive models used for two-dimensional and three-dimensional stress analysis

Constitutive Models
The constitutive models in a quasi-static stress analysis relate the state of strain to the state of
stress. These relations may be different depending on the kinematic assumptions of the
formulation (e.g., different through-the-thickness assumptions for the displacement fields).
However, in performing point strain analyses, the state of stress at a point is desired and is
readily computed once the point strains are determined through the use of the kinematics
relations. Dimensional reduction may result in one or more of the stress components being
assumed to be zero. For two-dimensional models, C0 plate and shell kinematics models have five
non-zero components and typically account at least for constant non-zero transverse shear

3
Threshold is defined in terms of the kinetic energy of the impacting debris. Impact threshold is defined as the
threshold kinetic energy level associated with the formation of through-thickness penetration of the RCC. Damage
threshold is defined as the threshold kinetic energy level associated with the onset of NDE-detectible damage – a
much more difficult computational task.

7
stresses through the thickness, while C1 plate and shell kinematics models have only three stress
components and neglect transverse stresses. For three-dimensional models, all six stresses and
six strains are evaluated. The subsequent sections describe the constitutive relations for these
kinematics models. Additional details on laminated composite structures are provided in Refs.
[2-9]. The engineering material constants in subsequent equations are the linear elastic
mechanical property values ( E11 , E22 , E33 , G12 , G13 , G23 , ν 12 ,ν 13 , and ν 23 ) .

Two-Dimensional Material Model

In two dimensions, the strain-stress relations for a typical linear elastic orthotropic material for
shell elements based on C0 continuity requirements implemented within ABAQUS are written
as:
⎧ ε11 ⎫ ⎡S110 S120 0 0 0 ⎤⎧ σ11 ⎫
⎪ ⎪ ⎢ 0 ⎥⎪ ⎪
⎪⎪ε22 ⎪⎪ ⎢S21 S22 ⎥⎪⎪σ 22 ⎪⎪
0
0 0 0
{ε} = ⎨ γ12 ⎬ = ⎢ 0 0 0
S44 0 0 ⎥⎨σ12 ⎬ = [S 0 ]{σ } (2)
⎪γ ⎪ ⎢ ⎥⎪ ⎪
⎥⎪σ13 ⎪
0
⎪ 13 ⎪ ⎢ 0 0 0 S55 0
⎪⎩γ 23 ⎪⎭ ⎢⎣ 0 0 0 0 S66 ⎦⎩σ 23 ⎪⎭
0 ⎥⎪

For C1 shell elements, the terms associated with the transverse or interlaminar shear effects (i.e.,
terms associated with γ13 and γ23) are eliminated. It is clear from Eq. 2 that the normal
components are coupled to one another, while the shear components are completely uncoupled.
The compliance coefficients Sij0 in this matrix are defined in terms of the elastic engineering
material constants (see Ref. [1], Vol. II, p. 10.2.1-3). Namely, the non-zero terms are:
1 1
S110 = 0
; S22 = ;
E11 E 22
ν 21 ν12
S120 = S21
0
=− =−
; (3)
E 22 E11
1 1 1
0
S44 = 0
; S55 = 0
; S66 =
G12 G13 G23
Hence, the stress-strain relations can be obtained directly from the compliance relations given in
Eq. 2 as:
⎧σ 11 ⎫ ⎡ C11 0 0
C12 0 0 0 ⎤⎧ ε11 ⎫
⎪ ⎪ ⎢ 0 ⎥⎪ ⎪
σ
⎪⎪ 22 ⎪⎪ ⎢C 21 C 22
0
0 0 0 ⎥⎪ε 22 ⎪
⎪ ⎪
[ ] {ε} = [C 0 ]{ε}
−1
{σ } = ⎨σ 12 ⎬ = ⎢ 0 0
0 C 44 0 0 ⎥⎨ γ 12 ⎬ = S
0
(4)
⎪σ ⎪ ⎢ ⎥⎪ ⎪
⎪ 13 ⎪ ⎢ 0 0 0 0
C 55 0 ⎥⎪ γ 13 ⎪
⎪⎩σ 23 ⎪⎭ ⎢⎣ 0 0 0 0 0 ⎥⎪γ ⎪
C 66 ⎦⎩ 23 ⎭

Note that the order of the stress and strain terms is the convention used by ABAQUS. The
stiffness coefficients Cij0 in this matrix are defined in terms of the elastic material constants (see
Ref. [1], Vol. II, p. 10.2.1-3). Namely,

8
 E11 E
C110 = 0
; C22 = 22 ;
Δ Δ
ν12 E 22 ν 21 E11
C120 = C21
0
= = ;
Δ Δ
(5)
0 0 0
C44 = G12 ; C55 = G13 ; C66 = G23

Δ = 1− ν12ν 21
and based on the reciprocity relation for plane elasticity given by:
E11ν 21 = ν 12 E22 (6)
In this approach, the material is modeled implicitly through the thickness by using the kinematics
assumptions of the two-dimensional shell elements (i.e., in-plane strains vary linearly through
the shell thickness and the transverse normal strain is zero).
Similarly, the stress-strain relations for a typical linear elastic orthotropic material for shell
elements based on the C1 continuity requirements are written as:
⎧σ11 ⎫ ⎡C110 C120 0 ⎤⎧ε11 ⎫
⎪ ⎪ ⎢ 0 ⎥⎪ ⎪
{σ } = ⎨σ 22 ⎬ = ⎢C21 0
C22 0 ⎥⎨ε22 ⎬ = [C 0 ]{ε} (7)
⎪σ ⎪ ⎢ 0 0 ⎥⎪ ⎪
⎩ 12 ⎭ ⎣ 0 C44 ⎦⎩γ12 ⎭
where the Cij0 coefficients were defined in Eq. 5.

At each planar Gaussian integration point within each shell element surface plane (e.g., 2×2
Gaussian integration points for a fully integrated two-dimensional shell element), material points
through the entire thickness (referred to as section points in ABAQUS) are evaluated to
determine the through-the-thickness state of strain and to estimate the trial stresses (i.e., point
stresses and strains). For a homogeneous orthotropic material, these values are defined as:
5 5
K11ts = G13 t; K 22
ts
= G23 t; K12ts = 0 (8)
6 6
where 5/6 represents the shear correction factor for an isotropic member, G13 and G23 are the
elastic transverse shear moduli on the 13- and 23-planes, respectively, and t is the thickness of
the shell. Shear correction factors for orthotropic laminates can be computed using the approach
presented by Whitney [107].
ABAQUS, however, does not allow the user-defined material models to modify the constitutive
0 0
terms associated with the transverse shear strains (i.e., C55 and C66 ). The transverse shear
stiffness terms are defined only once and only outside of the UMAT subroutine by using the
keyword command *TRANSVERSE SHEAR STIFFNESS followed by the input of three
numerical values: the transverse shear stiffness in each planar direction and the coupling
transverse shear stiffness. As a result, the UMAT subroutine for either two-dimensional shell
element formulation (C0 or C1) is the same. No failure can be detected or imposed on the
transverse shear components, and the constitutive matrix returned by the UMAT subroutine is
always, and only, a 3×3 matrix.

9
Three-Dimensional Material Model
In three dimensions, the strain-stress relations for a typical linear elastic orthotropic material
implemented within ABAQUS are written as:
⎧ ε11 ⎫ ⎡S110 S120 S130 0 0 0 ⎤⎧σ11 ⎫
⎪ ⎪ ⎢ 0 ⎥⎪ ⎪
⎪ε22 ⎪ ⎢S21 0 ⎥⎪σ 22 ⎪
0 0
S22 S23 0 0
⎪ε33 ⎪ ⎢S310 0 0
0 ⎥⎪σ 33 ⎪
⎥⎨ ⎬ = [S ]{σ }
S32 S33 0 0
{ε} = ⎨ ⎬ = ⎢ 0
(9)
⎪ γ12 ⎪ ⎢ 0 0 ⎥⎪σ12 ⎪
0
0 0 S44 0
⎪ γ13 ⎪ ⎢ 0 0 0 0 0
S55 0 ⎥⎪σ13 ⎪
⎪ ⎪ ⎢ 0 ⎥
⎪ ⎪
⎩γ 23 ⎭ ⎣ 0 0 0 0 0 S66 ⎦⎩σ 23 ⎭
From Eq. 9, it is clear that the normal components are coupled to one another, while the shear
components are completely uncoupled. The compliance coefficients Sij0 in the matrix given in
Eq. 9 are defined in terms of the elastic engineering material constants (see Ref. [1], Vol. II, p.
10.2.1-3). Namely, the non-zero terms of Sij0 are:
1 1 1
S110 = 0
; S22 = 0
; S33 = ;
E11 E 22 E 33
ν 21 ν12
S120 = S21
0
=− =− ;
E 22 E11
ν 31 ν13
S130 = S31
0
=− =− ; (10)
E 33 E11
ν 32 ν 23
0
S23 = S32
0
=− =−
;
E 33 E 22
1 1 1
0
S44 = 0
; S55 = 0
; S66 =
G12 G13 G23
Hence, the stress-strain relations can be obtained directly from the compliance relations (Eq. 9)
as:
⎧σ11 ⎫ ⎡C110 C120 C130 0 0 0 ⎤⎧ ε11 ⎫
⎪ ⎪ ⎢ 0 ⎥⎪ ⎪
⎪σ 22 ⎪ ⎢C21 C22 0 ⎥⎪ε22 ⎪
0 0
C23 0 0
⎪σ ⎪ ⎢C 0 C 0 0
0 ⎥⎪ε33 ⎪
⎥⎨ ⎬ = [S ] {ε} = [C ]{ε}
C33 0 0 0 −1
{σ } = ⎨ 33 ⎬ = ⎢ 31 32 0
(11)
⎪σ12 ⎪ ⎢ 0 0 ⎥⎪ γ12 ⎪
0
0 0 C44 0
⎪σ13 ⎪ ⎢ 0 0 0 0 0
C55 0 ⎥⎪ γ13 ⎪
⎪ ⎪ ⎢ 0 ⎥
⎪ ⎪
⎩σ 23 ⎭ ⎣ 0 0 0 0 0 C66 ⎦⎩γ 23 ⎭
Note that the order of the stress and strain terms is the convention used by ABAQUS. The
stiffness coefficients Cij0 in this matrix are defined in terms of the elastic material constants (see
Ref. [1], Vol. II, p. 10.2.1-5). Namely, the non-zero terms are:

10
 (1 − ν 23ν 32 ) E11 (1 − ν 13ν 31 ) E 22 (1 − ν 12ν 21 ) E33
C110 = 0
; C 22 = ; C 330 = ;
Δ Δ Δ

(ν 12 + ν 13ν 32 ) E 22 (ν 21 + ν 31ν 23 ) E11

C120 = C 21
0
= = ;
Δ Δ

(ν 13 + ν 12ν 32 ) E33 (ν 31 + ν 21ν 23 ) E11

C130 = C 310 = = ;
Δ Δ
(12)
(ν 23 + ν 13ν 21 ) E33 (ν 32 + ν 31ν 12 ) E 22
0
C 23 = C 320 = = ;
Δ Δ

0
C 44 = G12 ; C 550 = G13 ; C 660 = G23

Δ = 1 − ν 12ν 21 − ν 23ν 32 − ν 13ν 31 − 2ν 21ν 32ν 13

and the reciprocity relations for three-dimensional elasticity are:
E11ν 21 = ν 12 E22

E11ν 31 = ν 13 E33 (13)

E22ν 32 = ν 23 E33
In the three-dimensional approach, the laminate is modeled explicitly through the thickness by
using a series of three-dimensional solid elements through the entire laminate thickness. Within
each solid element, each of the material points (e.g., a 2×2×2 array of Gaussian integration
points) through the element thickness is evaluated to determine the state of strain and estimate
the trial stresses.

Implications for UMAT

The ABAQUS/Standard is a nonlinear finite element analysis tool [1]. In an ABAQUS/Standard
analysis, the analyst defines various analysis “steps” to perform. Step 1 may be a nonlinear
stress analysis for one set of boundary conditions; Step 2 may be a nonlinear stress analysis for a
different set of boundary conditions, and so forth. Within each “step”, a number of solution
increments may be performed depending on the type of analysis for that solution step. ABAQUS
uses the concept of an accumulated “pseudo-time” to cover all types of solution steps within a
complete analysis and a separate “pseudo-time” variable for each individual solution step. In a
quasi-static analysis, “pseudo-time” is related to the load factor. In the present analyses, only a
single solution step is employed – quasi-static nonlinear analysis through the keyword *STATIC.
Within the solution step, the pseudo-time variable is used to scale the applied loads and
displacements. The analyst specifies the starting solution increment, the maximum value of the
scaling parameter (also called the solution step period or duration), the minimum solution
increment value, and the maximum solution increment value. These factors are referred to as the

11
solution increment size factors in later sections. If automatic solution step size is permitted the
solution process within a solution step will begin with the initial solution increment value. Then
depending on convergence ease or difficulty, the solution increment size may increase to a
maximum value defined by the maximum solution increment value or decrease to a minimum
value defined by the minimum solution increment value. The analysis for each solution step
ends when the accumulated “pseudo-time” reaches the solution step period. Typical ABAQUS
keyword input records for a geometrically nonlinear static analysis are:
**STEP 1
*STEP, NLGEOM, INC=10000
*STATIC
0.10, 1.0, 1.0e-04, 0.10
ABAQUS/Standard is based on an incremental formulation where each increment in
displacement generates an increment of strain. For the k+1th solution increment, the strains may
be written as:
k +1,i k +1,i
{ε} = {ε} + {Δε}
k,∞
(14)
where {ε}k,∞ represents the strains from the previous kth converged solution increment (denoted
by letting the iteration index i go to infinity or i → ∞ ), {Δε}k +1,i represents the increment of strain
from the previous kth converged step to the ith iteration of the current k+1th solution increment,
and {ε}k +1,i represents the estimate of the strains for the ith iteration of the current k+1th solution
increment. The recovery of the stress state involves two aspects of the constitutive relations:
those stresses computed using the reference deformation state defined by the previous converged
solution and the increment of stress computed using the current local tangent state. As a result,
the stress-strain relations are written as:

{σ }k +1,i = {σ ({ε }k ,∞ )} { (
+ Δσ {ε } )}
k ,∞ k +1,i k +1,i

k +1.i

{(
= σ {ε }
k ,∞
)}k ,∞ ⎡ ∂Δσ ⎤
+⎢ {Δε }k +1,i (15)
⎣ ∂Δε ⎥⎦

{(
= σ {ε }
k ,∞
)}k ,∞
+ [J ({ε })]
k +1,i
{Δε }k +1,i
{ ( )} represents the stress state at the previous k converged solution increment
k,∞
Here σ {ε}k,∞ th

(denoted by the superscript ∞), {σ ({ε} )} represents the trial stress state for the i iteration of
k +1,i
k +1,i th

the current k+1th solution increment, and [J ({ε})]

k +1,i
represents the trial local tangent stiffness
matrix (or trial local material Jacobian matrix) for the ith iteration of the current k+1th solution
increment. The trial stress state and the trial local tangent stiffness matrix need to represent a
consistent set of relations. To obtain a consistent set, these terms are derived based on the strain
energy density functional U written as:

12
 1 1
{σ } {ε} = σ rsε rs
T
U= (16)
2 2
where the stress terms are expressed as functions of the strains through the stress-strain
constitutive relations. For a linear elastic material, the stress-strain constitutive relations, as
given in Eq. 11, are expressed as:
{σ } = [C 0 ]{ε} (17)

The first variation of the strain energy density functional U given by Eq. 16 with respect to the
strain terms gives the expressions for the stress terms as:
∂U
σ rs = (18)
∂ε rs

The second variation of the strain energy density functional given by Eq. 16 with respect to the
strain terms gives the local tangent stiffness coefficients Jrsmn, which may be expressed in
compact matrix form [J ({ε})] as:
∂ 2U ∂σ rs
J rsmn = = ⇒ [J ({ε})] (19)
∂ε rs∂ε mn ∂ε mn

In the nonlinear iteration process, the next solution goes from a converged solution increment k
(denoted by letting the iteration index i go to infinity or i → ∞ ) to the solution for the ith iteration
of solution increment k+1. The local tangent stiffness coefficients are determined as:

⎜ rs
⎟
(
⎛ ∂σ ⎞k +1,i ∂ (σ rs ) + (Δσ rs )
=
k,∞ k +1,i

= ⎜
)
⎛ ∂Δσ ⎞k +1,i
rs
⎟ (20)
⎝ ∂ε mn ⎠ (
∂ (ε mn ) + (Δε mn )
k,∞ k +1,i
)
⎝ ∂Δε mn ⎠

These two sets of relations need to be provided by the UMAT subroutine after being called by
ABAQUS/Standard; namely, the trial stress estimates, Eq. 18, and the trial tangent constitutive
coefficients, Eq. 20, for the ith iteration of the current k+1th solution increment. For a linear
elastic brittle material, the trial local tangent stiffness coefficients are independent of the current
strain values and are defined using the local secant stiffness approach from the ply-discounting
material degradation process. For a material that exhibits a nonlinear relationship in terms of
strains, special consideration must be taken in defining these terms.
The trial stress estimates and the trial constitutive coefficients are returned to ABAQUS/Standard
from the UMAT subroutine at each material point and incorporated into the computations that
generate the internal force vector for the residual vector (or force imbalance vector) and the
tangent stiffness matrix for a particular iteration and solution increment. The next section
describes failure initiation (or failure detection) criteria commonly used with laminated
composite structures.

Failure Initiation Criteria

Having described the basic constitutive models for two- and three-dimensional formulations for
the recovery of stresses at a material point in a composite structure, the failure initiation criteria
implemented in this UMAT subroutine for ABAQUS/Standard are now described. When only
“apparent” material data are available (i.e., only laminate-level material data), failure initiation
criteria are defined in a manner consistent with the available data. As such, the through-the-

13
thickness material definition is smeared over the laminate and any lamina properties are
essentially those of identical orthotropic layers with E1 and E2 being equal to the corresponding
effective engineering values (Ex and Ey) obtained from a laminate analysis. An average lamina
thickness is defined as the local total laminate thickness divided by the number of apparent ply
layers.
The present UMAT subroutine implementation for failure initiation includes four common
criteria; namely, the maximum stress criteria, the maximum strain criteria, the Tsai-Wu failure
polynomial [10, 18, 19, 28], and the Hashin criteria [12]. Note that two- and three-dimensional
versions of this UMAT subroutine are required depending on whether the finite element model
of the structural component involves two-dimensional shell elements or three-dimensional solid
elements, and both versions are included within this single UMAT subroutine. Even though
transverse shear behavior is not treated explicitly within an ABAQUS/Standard UMAT
subroutine, their failure criteria are defined herein for completeness.
At each material point in the finite element analysis model, ABAQUS/Standard provides the
strain state referenced to a specific coordinate system. The default coordinate system is the
global coordinate system of the finite element model. Frequently, this default coordinate system
is not aligned with the material coordinate system for the composite lamina. For simple
geometries, an alternate reference orientation can be specified using ABAQUS keyword input:
** Align fiber direction (1-direction) with global Y
*ORIENTATION,NAME=FIBER
0.,-1.,0.,+1.,0.,0.
3,0.
In this example, the original global Y axis is aligned with the fiber direction and hence, an
alternate reference frame is needed to transform the reference coordinate system to have a new
X′ axis aligned with the fiber direction. This transformation is defined by specifying two points
(three values each) on the input record after the *ORIENTATION keyword. The first three values
give the global coordinates of a point that when combined with the origin defines the new X′
axis. The second set of three values defines a point that when combined with the origin defines
the new Y′ axis. This alternate coordinate system has the X′ axis aligned with the 1-direction of
the material (fiber direction). Then the stacking sequence of the laminate can be specified
relative to this alternate reference coordinate system. For example, the following input records
define a four-layer laminate with a [0 / 45/ 90 /− 45] stacking sequence, a single integration point
in each layer, a layer thickness of 0.00645 inches, and reference a material named T800H39002
within the ABAQUS input:
*SHELL SECTION, COMPOSITE, ELSET=PLATE1, ORIENTATION=FIBER
0.00645, 1, T800H39002, 0.0
0.00645, 1, T800H39002, +45.0
0.00645, 1, T800H39002, 90.0
0.00645, 1, T800H39002, -45.0
where the first entry is the layer thickness, the second entry is the number of integration points
for this layer, the third entry is the material name for the material in this layer, and the last entry
is the orientation angle (in degrees) relative to a rotation about the normal using the coordinate
systems defined on the *SHELL SECTION keyword record (i.e., in this example it is FIBER). As a
result, the strains enter the UMAT subroutine referenced to this alternate coordinate system.

14
The stresses in the material coordinate system are computed within the current UMAT
subroutine using the given strains and the material stiffness coefficients at each material point.
Having these stresses, failure criteria are evaluated and material degradation is imposed as
warranted. In the following subsections, the different failure criteria implemented within the
current UMAT subroutine are described. Then, in a separate section of the present paper, the
material degradation approach is described.

Maximum Stress Criteria

The maximum stress criteria are simple non-interacting failure criteria that compare each stress
component against the corresponding material ultimate strength allowable value. The maximum
stress criteria for three-dimensional problems are written as:
σ11 σ11
e1t = for σ11 ≥ 0; e1c = for σ11 ≤ 0
XT XC
σ 22 σ 22
e2t = for σ 22 ≥ 0; e2c = for σ 22 ≤ 0
YT YC
(21)
σ 33 σ 33
e =
t
3 for σ 33 ≥ 0; e = c
3 for σ 33 ≤ 0
ZT ZC
τ12 τ 23 τ13
e4 = ; e5 = ; e6 =
S12 S23 S13
where each stress component is compared to its allowable value (either tension or compression).
t,c
If that ratio ei exceeds unity, then failure initiation has occurred for that stress component at
that material point and material degradation will be performed. That is, each component may
fail independently at different load levels, and the maximum stress failure flags f i MS are set
accordingly:
⎧ 0 for e t,c ≤ 1
f iMS = ⎨ i
for i = 1, 6 (22)
⎩±1 for ei > 1
t,c

The failure flags take on a positive value for tensile-related failures and a negative value for
compressive-related values. Shear-related failures have positive values for the failure flag. The
UMAT subroutine then assigns the solution increment number to nonzero failure flags that are
subsequently archived by ABAQUS for post-processing. In this way, an indication of failure
progression (by increment number) can be visualized.

Maximum Strain Criteria

The maximum strain criteria are simple non-interacting failure criteria that compare the strain
components against the corresponding material ultimate strain allowable value. The maximum
strain criteria for three-dimensional problems are written as:

15
 ε11 ε11
e1t = for ε11 ≥ 0; e1c = ult for ε11 ≤ 0
ε11T
ult
ε11C

ε 22 ε 22
e2t = for ε 22 ≥ 0; e2c = ult for ε 22 ≤ 0
ε 22T
ult
ε 22C
(22)
ε 33 ε 33
e3t = ult for ε 33 ≥ 0; e3c = ult for ε 33 ≤ 0
ε 33T ε 33C
γ 12 γ 23 γ
e4 = ; e5 = ult ; e6 = 13
γ 12
ult
γ 23 γ 13
ult

where each strain component is compared to its allowable value (either tension or compression).
t,c
If that ratio ei exceeds unity, then failure initiation has occurred for that strain component at
that material point and material degradation will be performed. That is, each component may
fail independently at different load levels, and the maximum strain failure flags f i ME are set
accordingly:
⎧ 0 for e t,c ≤ 1
f iME = ⎨ i
for i = 1, 6 (23)
⎩±1 for e i >1
t,c

Tsai-Wu Failure Polynomial

The Tsai-Wu failure polynomial [10] is an interacting failure criterion since all stress
components are used simultaneously to determine whether a failure at a material point has
occurred or not. The three-dimensional form of the Tsai-Wu failure polynomial for three-
dimensional problems is written as:
φ = F1σ 11 + F2σ 22 + F3σ 33 + F11 (σ 11 )2 + F22 (σ 22 )2 + F33 (σ 33 )2
+ 2 F12σ 11σ 22 + 2 F23σ 22σ 33 + 2 F13σ 11σ 33 (24)
+ F44 (σ 13 ) + F55 (σ 23 ) + F66 (σ 12 )
2 2 2

In this expression, the linear terms (Fi) account for the sign of the stresses and the quadratic
terms (Fij) define an ellipsoid in the stress space [10]. The values of the polynomial coefficients,
Fi and Fij, are dependent on the material ultimate strength allowable values as given by:
1 1 1 1 1 1
F1 = − ; F2 = − ; F3 = −
XT X C YT YC ZT ZC
1 1 1
F11 = ; F22 = ; F33 =
XT XC YT YC Z T ZC
1 1 1
F44 = ; F55 = ; F66 = (25a)
(S13 ) (S23 ) (S12 )
2 2 2

1 1 1 1
F12 = − ; F13 = −
2 XT XC YT Y 2 X T XC Z T Z
C C
1 1
F23 = −
2 YT YC ZT Z
C

Note that other definitions of the cross terms in Eq. 25a (F12, F13, and F23) do appear in the
literature; however, these definitions are implemented in the present UMAT subroutine. The
magnitude of the interaction terms are constrained by the inequality [10]:

16
 Fii F jj − Fij2 ≥ 0 (25b)

where summation is not implied by repeated indices. Since the Tsai-Wu failure model is an
interacting failure model that provides only a single condition for local material failure (i.e.,
where φ>1 denotes failure initiation), identifying the mode of failure requires a different
approach for the material degradation step in the progressive failure model from that used with
the maximum stress criteria. Reddy and Reddy [18] proposed using the relative contribution of
each stress component to the failure polynomial as a strategy to define the failure mode. Singh et
al. [19] proposed a slightly different strategy for failure mode identification using the Tsai-Wu
model. Huybrechts et al. [28] incorporated the out-of-plane shear terms into the failure
polynomial and achieved better correlation with test. A dominant component of the failure
polynomial is identified, and the corresponding failure indicator is set. Terms in the failure
polynomial involving stress component products are distributed equally between the associated
failure polynomial terms. This procedure is defined by writing the failure polynomial φ as a sum
of six components for a three-dimensional problem as given by:
6
⎧≤ 1 ... No failure
φ = φ1 + φ2 + φ3 + φ4 + φ5 + φ6 = ∑ φi = ⎨ (26)
i =1 ⎩> 1 ... Failure
Failure initiation occurs when φ exceeds a value of unity. The individual components φi are
defined by collecting like terms in Eq. 24. For three-dimensional problems, this grouping is
given by:
φ1 = F1σ11 + F11(σ11) + F12σ11σ 22 + F13σ11σ 33
2

φ 2 = F2σ 22 + F22 (σ 22 ) + F12σ11σ 22 + F23σ 22σ 33

2

φ 3 = F3σ 33 + F33 (σ 33 ) + F13σ11σ 33 + F23σ 22σ 33

2

(27a)
φ 4 = F66 (σ12 )
2

φ 5 = F44 (σ13 )
2

φ 6 = F55 (σ 23 )
2

For two-dimensional C1 problems, this grouping is given by:

φ1 = F1σ 11 + F11(σ 11) + F12σ 11σ 22
2

φ 2 = F2σ 22 + F22 (σ 22 ) + F12σ 11σ 22

2
(27b)
φ 3 = F66 (σ 12 )
2

Once the value of the failure polynomial φ exceeds unity, the individual components φi expressed
by Eq. 27a for three-dimensional problems (or Eq. 27b for two-dimensional problems) are used
to define the failure indices. Each failure polynomial component is assigned to a failure index in
that:
ei = φ i for i = 1, 2,..., 6 when φ ≥ 1 (28)
The largest value of the failure indices ei defines the failure mode as imax. That is,

17
 ⎧0 for ei where i ≠ imax
f iTW = ⎨ for i = 1, 6 (29)
⎩1 for ei where i = imax
No distinction is made between tension and compression values for the normal stress
components in evaluating the Tsai-Wu failure polynomial terms.

Hashin Failure Criteria

The Hashin failure criteria [12] are also interacting failure criteria as the failure criteria use more
than a single stress component to evaluate different failure modes. The Hashin criteria were
originally developed as failure criteria for unidirectional polymeric composites, and hence,
application to other laminate types or non-polymeric composites represents a significant
approximation. Usually the Hashin criteria are implemented within a two-dimensional classical
lamination approach for the point-stress calculations with ply discounting as the material
degradation model. Failure indices for the Hashin criteria are related to fiber and matrix failures
and involve four failure modes. Additional failure indices result from extending the Hashin
criteria to three-dimensional problems wherein the extensions are simply the maximum stress
criteria for the transverse normal stress components. The failure modes included with the Hashin
criteria [12] are:
• Tensile fiber failure – for σ 11 ≥ 0
⎛ σ ⎞ σ 2 + σ 2 ⎧> 1 failure
2

(e )
t 2
1 = ⎜ 11 ⎟ + 12 2 13 = ⎨
⎝ XT ⎠ S12 ⎩≤ 1 no failure
(30)

• Compressive fiber failure – for σ 11 < 0

⎛ σ ⎞ ⎧> 1 failure
2

(e )c 2
1 = ⎜ 11 ⎟ = ⎨
⎝ X C ⎠ ⎩≤ 1 no failure
(31)

• Tensile matrix failure – for σ 22 + σ 33 > 0

(σ + σ 33 )
2
(σ 2
− σ 22σ 33 ) σ122 + σ132 ⎧> 1 failure
(e 2)
t 2
= 22
YT2
+
23
2
S23
+
S122
=⎨
⎩≤ 1 no failure
(32)

• Compressive matrix failure – for σ 22 + σ 33 < 0

⎡⎛ ⎤⎛
Y ⎞ σ + σ 33 ⎞ (σ 22 + σ 33 )
2 2

()
2
e2c = ⎢⎜ C ⎟ −1⎥ ⎜ 22 ⎟ +
⎢⎣⎝ 2S23 ⎠ ⎥⎦ ⎝ YC ⎠
2
4S23
(33)
+
(
σ 23
2
− σ 22σ 33 )+ σ 12
2
+ σ 13
2 ⎪⎧>1 failure
=⎨
⎩⎪≤1 no failure
2 2
S23 S12

• Interlaminar normal tensile failure – for σ 33 > 0

⎛ σ ⎞2 ⎧⎪>1 failure
()
2
e3t = ⎜ 33 ⎟ = ⎨ (34)
⎝ ZT ⎠ ⎪⎩≤1 no failure

18
 • Interlaminar normal compression failure – for σ 33 < 0
⎛ σ ⎞2 ⎧⎪>1 failure
()
2
e3c = ⎜ 33 ⎟ = ⎨ (35)
⎝ ZC ⎠ ⎪⎩≤1 no failure

In these failure criteria, lamina strength allowable values for tension and compression in the
lamina principle material directions (fiber or 1-direction and matrix or 2-direction) as well as the
in-plane shear strength allowable value are denoted by XT, XC, YT, YC, and S12, respectively, ZT
and ZC are the transverse normal strength allowable values in tension and compression,
respectively, and S13 and S23 are the transverse shear strength allowable values. In Eq. 30-35, the
in-plane normal and shear stress components are denoted by σij (i, j=1, 2). The failure indices
for the transverse normal stress component σ33 are based on the maximum stress criteria, as
indicated by Eqs. 34 and 35. Finally, failure indices for tension and compression are then
compared to unity to determine whether failure initiation is predicted.
t,c
If any failure index ei exceeds unity, then failure initiation has occurred for that strain
component at that material point and material degradation will be performed. That is, each
failure mode may fail independently at different load levels, and the Hashin failure flags f i H are
set accordingly:
⎧ 0 for e t,c ≤ 1
f iH = ⎨ i
for i = 1, 3 (36)
⎩±1 for ei > 1
t,c

Damage Progression Models

The previous two sections of the present paper described the constitutive models and the failure
initiation criteria implemented in the current UMAT subroutine. This section presents the
damage progression strategies that provide material degradation and also the heuristics
associated with the material degradation models for these failure initiation criteria. Various
material degradation models have been proposed and demonstrated for laminated composite
structures (e.g., see Refs. [13-35]). Ochoa and Reddy [4] present an excellent discussion of
progressive failure analyses. These models may be generally categorized into two main groups:
heuristic models based on a ply-discounting material degradation approach and models based on
a continuum-damage-mechanics (CDM) approach using internal state variables. Representative
models for the ply-discounting material degradation approach are implemented within the
current UMAT subroutine; however, both approaches are described subsequently in the present
paper.

Ply-Discounting Approach
Traditionally, ply-discounting material degradation models are based on the degradation (or
discounting) of the elastic material stiffness coefficients by a value βI that, in essence, generates
a diminishing stiffness value (i.e., approaches zero as the number of solution increments from
failure initiation increases) for the ith stress component. In this strategy, the ith diagonal entry of
the elastic constitutive matrix C ii0 is set equal to βI multiplied by C ii0 , and the other row and
column entries of the elastic constitutive matrix [C0] are also degraded in a similar manner.
Another strategy to ply discounting involves degrading the elastic mechanical properties of the
material directly (i.e., assumption of vanishing elastic moduli after failure initiation is detected)

19
and then re-computing the local material stiffness coefficients using the degraded mechanical
properties. In this strategy, special care needs to be given to maintain symmetry in the degraded
constitutive matrix.
The application of material degradation implemented in this UMAT subroutine is related to the
rate of material degradation once failure initiation is detected. Instantaneous or single-step
degradation degrades the material stiffness coefficients only once by the degradation factor.
Typical values for the degradation factor can range from a very small value (e.g., 10-6) to a large
value (e.g., 0.8) for this type of material degradation. Recursive degradation successively
degrades the material stiffness coefficients in a gradual manner and thereby avoids some of the
numerical convergence issues associated with an instantaneous local change in material stiffness.
Specifying recursive degradation with a near-zero degradation factor is nearly equivalent to
specifying instantaneous degradation with the same near-zero factor.
Methods based on recursive material degradation have been implemented to minimize the
computational impact of localized changes in material stiffness terms, and recursive degradation
is included in the current UMAT subroutine implementation. Artificial viscous damping has also
been used to improve the convergence behavior of the nonlinear solution procedure in quasi-
static analyses [15, 29]. Examples of the ply-discounting approach and related computational
details are presented in Refs. [16-21, 29]. The ply-discounting approach based on degrading the
constitutive matrix coefficients including options for either recursive or single-step
(instantaneous) material degradation (controlled by the parameter RECURS) has been
implemented into the current UMAT subroutine and is available as a user-specified option.
For the ply-discounting models, the analyst selects the failure initiation criterion (PDA), the
material degradation factor (β), and the type of degradation (instantaneous or recursive). The
failure initiation criteria implemented in this UMAT are the maximum stress criteria (PDA=1),
maximum strain criteria (PDA=2), Tsai-Wu failure polynomial (PDA=3) [10, 18, 19, 28], and the
Hashin criteria (PDA=4) [12]. The form of the material degradation is selected by the parameter
RECURS. When RECURS equals zero, instantaneous degradation is imposed. When RECURS
equals unity, recursive degradation is imposed. Material degradation is also dependent on the
failure mode (tension, compression, or shear) and independent degradation factors can be
specified in this UMAT subroutine (i.e., Dgrd(1) for tension, Dgrd(2) for compression, and
Dgrd(3) for shear). The material degradation factor implemented in this UMAT subroutine
multiplies the material stiffness coefficients directly rather than multiplying the engineering
material properties themselves. By multiplying the stiffness coefficients directly, the issue of
maintaining the reciprocity relations for degraded engineering material properties is avoided.
The combined use of recursive degradation (RECURS=1) and a fractional degradation factor
(e.g., β=0.5) can provide a gradual degradation of material stiffness over several solution
increments in an attempt to minimize numerical convergence difficulties attributed to near
singularities in the stiffness matrix caused by localized material failures. In the present paper,
the degradation factor β will be shown to have a significant effect on the failure prediction. The
numerical studies reported here use a degradation factor of 0.5 as a default value for β. This
value gives successive reductions in the stiffness coefficients by a factor of two on each solution
increment after initial failure has been detected. While somewhat arbitrary, this value has been
shown to give good convergence behavior for the overall nonlinear solution algorithm for
progressive failure simulations.

20
For the failure initiation criteria described previously, six failure indices ei are evaluated for the
three-dimensional analysis models. Failure initiation is defined when one or more of these
failure indices reach or exceed unity. The material degradation rules for the ply-discounting
approach are defined in Table 1 for the maximum stress and maximum strain failure criteria, in
Table 2 for the Tsai-Wu failure polynomial criterion, and in Table 3 for the Hashin criteria. For
the various criteria, the material degradation rules follow the heuristics available in the literature.
Typically when a failure is detected in a particular mode, say when the first failure flag is
nonzero, then the local material stiffness associated with the fiber direction is degraded.
This approach to material degradation may lead to very conservative predictions for interacting
failure criteria. For example, the tensile fiber failure mode given by Eq. 30 for the Hashin [12]
criteria in the present UMAT subroutine implementation is will result in a degradation of the
fiber related stiffness coefficients as well as the in-plane shear coefficients (see Table 3).
However, if a first term in Eq. 30 does not cause failure initiation (maximum stress comparison)
but the entire term does exceed unity, then the failure mode could be interpreted as a fiber-matrix
shearing failure mode wherein only the in-plane shear stiffness coefficients are degraded. No
material degradation rules were presented by Hashin [12].
In this UMAT subroutine, the ply-discounting material degradation models are based on
discounting (or degrading) the terms of the elastic stiffness constitutive matrix (i.e.,
CijDegraded = β I Cij0 ). Degradation factors βI are defined for three failure types: tension-failure
degradation factor βT, compression-failure degradation factor βC, and shear-failure degradation
factor βS. Off-diagonal terms in the constitutive coefficient matrix along the same row and
column are also degraded in the same manner. Material degradation can be performed either
only once when failure initiation is detected (single-step approach), or it can be done recursively
on each solution increment after failure initiation is detected (recursive approach). Recursive
material degradation typically provides a more “gentle” process of degrading material stiffness
data and potentially can improve convergence characteristics of the solution procedure compared
to an “abrupt” single-step degradation approach using near zero values for the degradation
factors. In addition, once a failure mode is detected that failure mode is not checked at that
material point again; however, recursive degradation of the material stiffness coefficients will
continue to be applied. Material degradation continues until the degradation factor reaches a
specified minimum value and then is held constant at that minimum value (currently set as 10-30).
At subsequent solution increments, other failure modes within a given failure criteria are
evaluated at that material point and potentially could lead to a subsequent failure in a different
mode.

Internal State Variable Approach

Continuum-damage-mechanics (or CDM) models generally describe the internal damage in the
material by defining one or more internal state variables. Regardless of the damage state, these
CDM models still represent the material as continuum having smooth, continuous field
equations. Krajcinovic [108] described CDM “as a branch of continuum solid mechanics
characterized by the introduction of a special (internal) field variable representing, in an
appropriate (statistical) sense, the distribution of microcracks locally.” Talreja [109] was one of
the earliest to propose such a model for laminated composites along with Chang and Chang [15].
Ladevèze and LeDantec [110], Shahid and Chang [111], and Barbero and Lonetti [25] have
additional CDM models. CDM models express the constitutive relations in a manner similar to

21
the elastic constitutive relations given previously, except that the coefficients (i.e., either
compliance coefficients, stiffness coefficients, or the mechanical properties themselves) are
functions of one or more internal state variables. CDM models also generally require additional
material data as part of the material property definition. For example, one may require a material
strength allowable value to be defined as a function of an internal state variable, which may
themselves be dependent on laminate stacking sequence. As such, their use requires even more
care than material degradation models based on heuristic rules as used in ply-discounting
material degradation models. Krajcinovic [112] provides additional details on CDM methods for
damage modeling.
One approach is to incorporate the statistical nature of the material into the constitutive relations.
Matzenmiller et al. [101] proposed such a model (called the MLT model) based on the use of a
Weibull function [113] to describe the statistical nature of internal defects and the ultimate
strength of a fiber bundle within a composite lamina. Creasy [114] developed a different model
based on the Weibull function, and Moas and Griffin [21] used Weibull functions in their
degradation model. In the MLT model, the Weibull factor m is a primary control parameter that
affects the strain-energy density at a given material point. By adjusting the value m, the degree
of strain softening in the post-ultimate region is controlled. Nonlinearity in the pre-ultimate
region is controlled by the independent specification of the ultimate stress value, the ultimate
strain value, and a parameter similar to a material modulus. Hahn and Tsai [115] considered the
post-ultimate behavior of symmetric cross-ply laminates. Their results indicate that a gradual
degradation of the cross-ply stiffness terms explains the bilinear stress-strain behavior that
occurred prior to total failure. Furthermore, they suggested that any failed lamina continues to
carry its failure load in the post-ultimate range until the laminate fails (see Refs. [115-117]).
Hence, Schweizerhof et al. [102] incorporated a stress limit factor that sets the minimum stress
value in the post-ultimate regime of the stress-strain curve.
The Matzenmiller et al. [101] formulation as extended by Schweizerhof et al. [102] is available
in LS-DYNA [100] as material MAT58. The MLT continuum-damage-mechanics or CDM
approach for progressive failure analysis is described in Appendix C for one-dimensional
problems. Parametric studies for a quasi-static one-dimensional uniaxial test case are also
described in Appendix C to illustrate the MLT approach.

UMAT Implementation
Having a user-defined material feature is becoming an increasingly important capability for
commercial finite element software systems. Developers of commercial finite element analysis
tools frequently provide entry points for material data through special features and well-
documented subroutine calling arguments. Within the ABAQUS/Standard finite element tool,
this feature is referred to as a UMAT subroutine (see Ref. [1], Volume III, p. 23.2.29-1). The
user is given various computational results from the element routines and solution procedure that
are passed to the UMAT subroutine through the calling arguments and may be used to determine
new trial values for the stresses and constitutive coefficients based on the current deformation
state. It is important to realize that the UMAT subroutine actually defines a trial stress state for
the end of the increment and as such, damage can evolve as the nonlinear solution procedure

22
iterates to find the solution for the current increment.4 These new trial values are returned to the
main program for additional computations (i.e., element stiffness matrices and internal force
vectors). The reader is referred to Appendix A for a description of the ABAQUS/Standard
UMAT subroutine calling arguments, which is included for convenience and completeness.
The material models described in the previous sections have been implemented as a UMAT
subroutine for ABAQUS/Standard to describe the mechanical behavior of a laminated composite
material system including progressive failure analysis. Different options are provided within the
UMAT subroutine. This UMAT subroutine may be used with either two-dimensional shell
elements or three-dimensional solid elements. It may be used for linear elastic bimodulus
response only (PDA=0). In addition, it may be used for progressive failure analysis based on
point-stress analysis, failure criteria assessment (i.e., maximum stress criteria (PDA=1),
maximum strain criteria (PDA=2), Tsai-Wu failure polynomial (PDA=3) or Hashin criteria
(PDA=4)), and material degradation based on the ply-discounting approach.
For all values of the parameter PDA, this UMAT subroutine computes several strain-energy-
density-related parameters at each element integration point. The use of strain-energy density as
a computational sensor in progressive failure analyses is not new. Strain-energy density can be
used as a scalar parameter to identify highly stressed regions in the structure. In addition,
damage initiation and propagation has been related to strain-energy density – perhaps more so
for metal structures than composite structures. The total strain-energy density U Total at a material
point corresponding to the current strain level is computed first and is approximated by the total
area under the nonlinear stress-strain curve up to the current strain state ε˜ij (and corresponding
stress state σ˜ ij ) as given by:
ε~ij

U Total = ∫ σ ij dε ij (37)
0

This integral is evaluated at each material point during the incremental solution process using the
trapezoidal rule for numerical integration to the current strain state. The recoverable strain-
energy density U Recover based on a secant-modulus approach is computed next and represents the
triangular-shaped area bounded between the strain axis and intersection of the current strain level
with the stress-strain curve. That is,
1
U Recover = σ˜ ijε˜ij (38)
2
The modulus of toughness can be defined as the failure strain-energy density U Failure and is equal
to the area under the stress-strain curve up to material failure. For a brittle material, this failure
strain energy density can be related to the critical strain-energy release-rate Gc of the material
and the width of the fracture-process zone WFPZ by the following relationship:

4
Alternatively, the ABAQUS/Standard user-defined field approach using the USDFLD subroutine explicitly
defines the stress state at the beginning of the increment and maintains that stress state over the increment. Such an
approach is generally used in explicit solution procedures and in implicit procedures wherein the solution increment
size is limited to be a small value.

23
 Gc
U Failure = (39)
WFPZ
The width of the fracture-process zone WFPZ can be set by the user as an input parameter to the
UMAT subroutine. In finite element analyses, the value of this parameter is frequently related to
the smallest dimension of a finite element in the local region of interest, and in these analyses, it
is assumed to be 0.2 inches. Hence for any value of strain, the strain-energy density loss U Loss is
given by the difference between the total and recoverable strain-energy densities. An energy-
based damage estimate Denergy is then expressed as the ratio of the strain-energy density loss and
the failure strain energy density:
U Loss U − U Recover
Denergy = = Total (40)
U Failure U Failure
When this energy-based damage estimate Denergy approaches a value of unity, complete failure at
that material point has occurred. This approach to estimating damage appears to be applicable
for strain-softening material models and follows an approach used by Hillerborg [95, 96] for
concrete.
The current version of the UMAT subroutine requires the user to specify fifty-five input values
defining material property values and analysis options. The input variables to this UMAT
subroutine are listed in Table 4. Output from this UMAT subroutine includes a set of solution-
dependent variables or SDVs that are computed at each material point in the finite element model
and returned on exit from this UMAT subroutine: eight variables for two-dimensional shell
elements and fourteen variables for three-dimensional solid elements. The solution-dependent
variables listed in Table 5 are assigned within this UMAT subroutine and stored for post-
processing. For this UMAT subroutine, the solution-dependent variables represent the
degradation factors for each stress component, the failure flags for each component, the total
strain energy density, and the energy-based damage estimate for each solution increment of the
nonlinear analysis. The failure flags are defined as the increment number when first failure is
detected. Hence, a plot of the failure flags can be used to give an indication of the damage
progression. Solution-dependent variables can be selected for output to the ABAQUS solution
database (*.odb file) for subsequent post-processing of the solution using ABAQUS VIEWER5.
Note that for user-defined material models, ABAQUS/Standard actually does not distinguish
between C0 and C1 shell elements – only the in-plane stress and strain states can be treated within
the UMAT subroutine (i.e., all two-dimensional shell elements are treated as C1 shell elements,
transverse shear components are not available to the user).
Applying the current UMAT subroutine to structures composed of multiple materials requires a
capability to handle multiple sets of material properties where each set is associated with
different local thickness value or a different material type. Material properties are required to be
specified by the user for each thickness value including the tensile modulus ET, the compression
modulus EC, the in-plane shear modulus G, Poisson’s ratio ν, the tensile material strength
allowable value XT, the compressive material strength allowable value XC, and the in-plane shear

5
ABAQUS VIEWER is a trademark of ABAQUS, Inc.

24
material strength allowable value S. Essentially each thickness value would define a separate
material number or name within the UMAT subroutine and the analysis input data.
The UMAT material property data for a representative T800/3900-2 graphite epoxy composite
material named T800H39002 are defined using the following records:
** ==========================================================================================
** ==========================================================================================
** UMAT Property Data Definitions
** props(1-8):E11t,E22t,E33t,E11c,E22c,E33c,G12,G13,
** props(9-16):G23,nu12,nu13,nu23, Xt, Yt, Zt, Xc,
** props(17-24):Yc,Zc,S12,S13,S23,Eps11T,Eps22T,Eps33T,
** props(25-32):Eps11C,Eps22C,Eps33C,Gam12,Gam13,Gam23,Eps11Tmx,Eps22Tmx,
** props(33-40):Eps33Tmx,Eps11Cmx,Eps22Cmx,Eps33Cmx,Gam12mx,Gam13mx,Gam23mx,GIc,
** props(41-48):FPZ,SlimT,SlimC,SlimS,weibull(1),weibull(2),weibull(3),weibull(4),
** props(49-55):weibull(5),weibull(6),Dgrd(1),Dgrd(2),Dgrd(3),RECURS,PDA
** ==========================================================================================
** T300H/3900-2 Material 1 with total thickness = 0.00645 inches per ply
*MATERIAL, NAME=T800H39002
*USER MATERIAL, CONSTANTS=55
2.320E+07, 1.300E+06, 1.300E+06, 2.320E+07, 1.300E+06, 1.300E+06, 0.900E+06, 0.900E+06,
0.500E+06, 2.800E-01, 2.800E-01, 2.800E-01, 4.120E+05, 8.720E+03, 8.720E+03, 2.250E+05,
2.430E+04, 2.430E+04, 1.376E+04, 1.376E+04, 7.644E+03, 0.178E-01, 0.671E-02, 0.671E-02,
0.970E-02, 0.187E-01, 0.187E-01, 0.153E-01, 0.153E-01, 0.153E-01, 1.000E-01, 1.000E-01,
1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 0.860E+00,
2.000E-01, 0.800E+00, 0.800E+00, 0.800E+00, 1.000E+00, 1.000E+00, 1.000E+00, 1.000E+00,
1.000E+00, 1.000E+00, 5.000E-01, 5.000E-01, 5.000E-01, 1.000E+00, 4.000E+00
*DEPVAR
8
** ==========================================================================================
** ==========================================================================================

Note that the only difference for applying this UMAT subroutine to two-dimensional or three-
dimension problems is the last keyword *DEPVAR. This keyword defines the number of
solution-dependent variables or SDVs for the problem. For two-dimensional finite elements, it is
8, and for three-dimensional finite elements, it is 14.
Material-property data and state-variable data from an ABAQUS execution are passed as calling
arguments to the UMAT subroutine. In addition, input variables to UMAT from an ABAQUS
execution include the total strains from the previous solution increment k and the current iterative
increment of strains (increment k+1) for a given material point in the structure. Given these
values, the principal strains can be computed for this material point in the structure and then
control is transferred to the appropriate branch within the UMAT subroutine (i.e., tension branch
or compression branch). That is, on entry to the UMAT subroutine, an estimate of the current
total strains for the current iteration is determined:

(ε ) = (εij ) + (Δεij )
k +1 k k +1
ij (41)

Using the total strains from the previous solution increment, the first invariant of the strain tensor
is computed, which represents the volumetric strain from the previous solution increment,
⎧εxx + εyy for 2D
Iε = trace(εij ) = ⎨ (42)
⎩εxx + εyy + εzz for 3D
The sign of the first invariant of the strain tensor determines the overall stress state at that
material point (i.e., either tensile or compressive). This scalar quantity is invariant with regard to
coordinate system and is used within the UMAT subroutine to determine whether the material

25
point is in a state of tension or compression. Based on the sign of the first invariant of the strain
tensor, the appropriate branch of the material response curve is selected. That is,
⎧≥ 0 Tension
Iε ⎨ (43)
⎩< 0 Compression
Once on the appropriate branch, new trial constitutive coefficients and new trial stresses are
determined. These stresses are then used to evaluate failure initiation criteria (if the input
variable PDA is non-zero). If failure initiation is detected, material degradation is imposed using
the ply-discounting approach if PDA equals 1, 2, 3 or 4. The constitutive relations, the failure
initiation criteria, and the material degradation models needed for both three-dimensional solid
elements and two-dimensional shell elements in ABAQUS/Standard are described earlier in the
present paper.
The implementation process for the three-dimensional solid elements follows the next series of
steps. For each element integration point, these basic steps are followed:
1. Call the user-defined subroutine UMAT with the previous strain vector, the iterative
increment in strains, and trial values for the constitutive matrix at this material point (i.e.,
the element integration point through the thickness).
2. Compute the current total strains for this iteration by summing the total strains from the
previous increment and the corresponding iterative increments of strain.
3. Using the current total strains, compute the first invariant of the strain tensor at that
material point. Based on sign of the first invariant of the strain tensor, set the engineering
material properties to either their tension or compression values. Then compute the trial
constitutive matrix.
4. Compute the trial stresses using the trial constitutive matrix and the total strains at this
material point.
5. Perform failure initiation check using a point-stress analysis approach for either:
a. Evaluate the maximum stress failure criteria (PDA=1) and determine whether any
material failure initiated. If so, perform material degradation. Or,
b. Evaluate the maximum strain failure criteria (PDA=2) and determine whether any
material failure initiated. If so, perform material degradation. Or,
c. Evaluate the Tsai-Wu failure polynomial (PDA=3) and determine whether any
material failure initiated. If so, perform material degradation. Or,
d. Evaluate the Hashin criteria (PDA=4) and determine whether any material failure
initiated. If so, perform material degradation.
6. Perform material degradation using a ply-discounting approach – If material failure is
detected, then degrade the material properties by degrading the entries in the constitutive
matrix, rather than degrading the engineering properties, so that the appropriate stress
component will approach zero after failure initiation. If the degradation factors are set to
zero, then instantaneous degradation to zero will occur. Recursive degradation is
imposed when the variable RECURS is set to unity, while instantaneous (single or one-
time) degradation is imposed when RECURS equals zero.
26
 7. Re-compute the trial constitutive matrix and the trial stresses, update the solution-
dependent variables, and return to ABAQUS.
The input data to the UMAT subroutine include the strain state and the estimate of the
constitutive coefficients based on user data defined in the property array PROPS (see Table 6)
and on the computed data defined in the solution-dependent variable array STATEV (see Table
7). The output variables passed back to ABAQUS/Standard from the UMAT subroutine includes
the trial stresses, the trial constitutive matrix coefficients, and updated values of the solution-
dependent variable array.
The overall nonlinear solution strategy is essentially unchanged from the default approach in
ABAQUS/Standard. However, it may be necessary to disable the extrapolation feature for the
next solution increment in order to enhance numerical convergence. This is accomplished using
the ABAQUS keyword command *STEP EXTRAPOLATION=NO.

Numerical Results and Discussion

Example problems have been solved using this user-defined material model, and selected results
are described in the present paper. First, a basic demonstration problem is described to illustrate
the basic features of the user-defined material model. Then a tension-loaded coupon strip with a
center circular hole is described, and results compared with existing test data. Results are
presented in terms of xy-plots of results recovered from the history variables and in terms of
contour plots of results recovered from the field variables stored in the solution database (*.odb
file). Output of the history variables to the solution database occurs after every ten solution
increments or at the end of the simulation provided the maximum solution factor was obtained.
If convergence difficulties are encountered, say during increment 39, and the solution process
stops, then history variable output only up to the increment 30 is stored in the solution database.
However, field variable output is stored for every solution increment. Convergence of the
solution process can be monitored using the *.sta file provided by ABAQUS.

Demonstration Problems
To demonstrate the various features of this user-defined material model implemented as a
UMAT subroutine within ABAQUS/Standard, a two-dimensional rectangular panel, shown in
Figure 3, subjected to uniform extension is solved using a single 4-node flat quadrilateral finite
element. A right-hand coordinate system is used with positive z directed out of the paper and
toward the reader. This problem simulates a uniaxial response. The dimensions of the panel are
4-inches by 8-inches and a total thickness of 0.2 inches. The loading is an applied edge
displacement in the long (positive x) direction of the panel. The boundary conditions provide
restraint from rigid body motion and permit contraction in the shorter (y) dimension to prevent
the development of stresses transverse to the loading direction (i.e., develop a uniaxial stress
state). A representative ABAQUS/Standard input file for this problem set is given in Appendix
D.
Two different isotropic materials are considered as indicated in Figure 4. Both materials have the
following tensile properties: 2.0 Msi for the elastic modulus, 0.25 for Poisson’s ratio, and 7,000
psi for the ultimate longitudinal tensile strength. The first material has a linear elastic isotropic
brittle response with an ultimate longitudinal tensile strain value of 0.0035 in./in. The second
material has a nonlinear elastic isotropic brittle response with an ultimate longitudinal tensile

27
strain value of 0.006 in./in. The only difference in input data for these two material systems is
the ultimate tensile strain values. For the linear elastic brittle material, the ultimate tensile strain
is computed using the elastic modulus and the ultimate tensile stress value. For the nonlinear
elastic material, the ultimate strain is specified independent of the ultimate stress value and the
linear elastic modulus. These problems demonstrate the basic functionality of the UMAT
subroutine and illustrate the progressive failure analysis options provided. Typical input records
for the UMAT subroutine for these demonstration problems are given in Appendix D.
The first problem solved involves the linear elastic brittle material system (see Figure 4) and
considers several of the progressive failure analysis options provided in the UMAT subroutine.
The basic stress-strain response is shown in Figure 5 for different progressive failure analysis
options. All solutions employ instantaneous material degradation (RECURS=0), a material
degradation factor of 10-6, and a fixed solution increment size factor of 0.01. The linear elastic
response (PDA=0) as well as the response predicted using the stress-based failure criteria
(PDA=1, maximum stress criteria; PDA=3, Tsai-Wu failure polynomial; PDA=4, Hashin criteria)
and using the strain-based criteria (PDA=2, maximum strain criteria) each provide identical
solutions for the uniaxial tension case. Stress-based and strain-based failure criteria predict the
same response for linear elastic brittle material systems.
Other response quantities are shown in Figure 6. The resultant axial force as a function of
applied strain is shown in Figure 6a. The peak axial force is 5,600 pounds and is predicted by all
methods. The material degradation factor as a function of applied strain is shown in Figure 6b.
In these cases, instantaneous material degradation (RECURS=0), and a degradation factor of 10-6
were used in all cases; hence the factor starts with a value of unity (no damage) and changes to
10-6 once failure is detected (solution increment 35). The strain energy density U Total at a
material point (see Eq. 37) is shown in Figure 6c and represents the area under the stress-strain
curve. For the elastic case, the value continues to increase regardless of the strain level. For the
progressive failure analyses, the strain energy density continually increases until a value of 12.55
in.-lb/in.3 is attained and then remains constant due to the material failures and the material
degradation factor. Due to the solution increment size, this computed strain energy density value
is slightly larger than the theoretical maximum value at ultimate of 12.25 in.-lb/in.3 (i.e.,
computed using 1 2 X T ε11ultT ). The energy-based damage parameter Denergy (see Eq. 40) as a
function of applied strain is shown in Figure 6d. It maintains a zero value until failure initiation
and then exhibits a step increase to a constant value for the remainder of the simulation. The
value of Denergy is dependent upon the size of the fracture process zone WFPZ, which typically is
related to a characteristic length of an element. This energy-based damage parameter appears to
have limited use for the ply-discounting progressive failure analysis models applied to linear
elastic brittle materials. However, application to strain softening materials needs to be assessed.
The sensitivity of the progressive failure analysis results to the maximum solution increment size
factor is indicated in Figure 7. Using the maximum stress criteria (PDA=1) and instantaneous
degradation (RECURS=0) with a degradation factor of 10-6, results are presented for three
different solution increment size factors; namely, 0.01, 0.05, and 0.10. For the smallest factor
value, a near vertical reduction in stress from 7,000 psi to zero is achieved with little increase in
strain. For the value of 0.05, the same peak value is achieved; however, zero stress occurs for a
somewhat larger strain than 0.0035 in./in. (ultimate strain value). For the value of 0.1, the peak
stress is below the ultimate stress value because the next solution increment would result in
28
exceeding the failure criteria. These results indicate that when the solution increment size is too
large, failure initiation may be predicted prematurely. However, the use of a very small value
can increase the overall computational cost of the simulation. Solution increment size has an
influence on the progressive failure simulation depending on its value.
The effect of recursive material degradation for various values of the degradation factor is seen
by comparing Figures 8a and 8b for the stress-strain response, Figures 9a and 9b for the material
degradation factor, and Figures 10a and 10b for the strain energy density. The linear elastic
results (PDA=0) are presented on the figures for comparison. Again all progressive failure
analysis results are computed using the linear elastic isotropic brittle material, the maximum
stress criteria (PDA=1), and a solution increment factor of 0.01.
The influence of different values of the degradation factor β combined with instantaneous
degradation (RECURS=0) is shown in Figures 8a, 9a, and 10a. Since the degradation factor is
only applied once at failure initiation, the degradation occurs at the same stress level; however,
the post-ultimate stress-strain response shown in Figure 8a is very dependent on the degradation
factor. Values larger than 0.01 appear to result in nonzero stresses within the model that increase
as the applied strain increases. The variation of the degradation factor as a function of applied
strain is shown in Figure 9a (i.e., a single application of the degradation factor). The strain
energy density variation shown in Figure 10a indicates a significant increase in total strain
energy density even after failure initiation when the degradation factor is larger than 0.01.
Hence, the degradation factors should be very small when used with instantaneous degradation.
The influence of different values of the degradation factor β combined with recursive
degradation (RECURS=1) is shown in Figures 8b, 9b, and 10b. In these cases, the degradation
factor is recursively applied on each solution increment after failure initiation. The degradation
factor is first applied at failure initiation, which occurs at the same stress level for all values of β.
Subsequent solution steps recursively apply the degradation factor, and the material constitutive
stiffness coefficients are continually degraded. In addition, the post-ultimate stress-strain
response shown in Figure 8b is very similar in all cases with only minimal stress levels occurring
as the applied strain increases. The variation of the degradation factor as a function of applied
strain is shown in Figure 9b (i.e., recursive application of the degradation factor). The strain
energy density variation shown in Figure 10b indicates only a modest increase in total strain
energy density even after failure initiation when the degradation factor is larger than 0.1. Hence,
the degradation factor used with recursive degradation can be an order of magnitude larger than
the value used with instantaneous degradation.
The second problem solved involves the nonlinear elastic isotropic brittle material system (see
Figure 4) and considers several of the progressive failure analysis options provided in the UMAT
subroutine. The predicted stress-strain response for the single element model is shown in Figure
11 for different progressive failure analysis options. This UMAT subroutine considers the pre-
ultimate response to be linear elastic in all cases. All solutions employ instantaneous material
degradation (RECURS=0), a material degradation factor of 10-6, and a fixed solution increment
size factor of 0.01. The linear elastic response (PDA=0) as well as responses predicted using the
stress-based failure criteria (PDA=1, maximum stress criteria) and using the strain-based criteria
(PDA=2, maximum strain criteria) provide different solutions for the uniaxial tension case of a
nonlinear elastic material.

29
 Stress-based and strain-based failure criteria predict different responses for the nonlinear elastic
brittle material system. Stress values predicted by a linear elastic model exceed the ultimate
stress level as the applied strain level increases – response is elastic and no failures are ever
indicated. Stress value for failure initiation predicted by the maximum stress criteria is predicted
correctly; however, failure initiation is predicted to occur at a significantly lower strain level than
the ultimate strain allowable value. Failure initiation occurs once the nominal failure strain
based on the elastic modulus and ultimate stress allowable value is reached. Stress values
predicted by the maximum strain criteria exceed the ultimate stress level as the applied strain
level increases; however, failure initiation is indicated at the specified ultimate strain value (i.e.,
0.006 in./in.). These results indicate that a material model is needed to address nonlinear pre-
ultimate stress-strain behavior, and one such model is the MLT model described in Appendix C.
The effect of recursive material degradation for various values of the degradation factor is shown
in Figure 12 for the stress-strain response. Again all results are computed using the nonlinear
elastic isotropic brittle material, the maximum strain criteria (PDA=2), recursive degradation,
and a solution increment factor of 0.01. These results indicate that the stress-strain response is
minimally affected by the size of the degradation factor when recursive degradation is used.
Failure initiation occurs at the same stress level in all cases.
These results demonstrate the features and capabilities of the UMAT subroutine described in the
present paper for a simple uniaxial problem modeled with a single 4-node finite element. For a
linear elastic isotropic brittle material, all failure criteria predict the same behavior.
Instantaneous degradation has been shown to require a smaller degradation factor than when
recursive degradation is used. For a nonlinear elastic isotropic brittle material, stress-based
criteria under predict the strain level for failure initiation and strain-based criteria over predict
the stress level for failure initiation. In all cases, the pre-ultimate stress-strain behavior is
assumed to be linear elastic and any nonlinear elastic effects are ignored in the current
implementation. Based on these results, the UMAT subroutine is applied next to a more
complex analysis problem involving many finite elements, many layers of material, general
quadrilateral finite element shapes, a combined stress state, and a local stress gradient.

Open-Hole-Tension Coupons
Tension-loaded coupons with a center circular hole are examined and compared with available
test data [120]. Each coupon is 9-in. long and 1-in. wide. The hole diameter is 0.25 in. The
laminate is a 16-ply T300H/3900-2 graphite-epoxy with an average ply thickness of 0.00645 in.
for a 0.1032-inch total laminate thickness. Two stacking sequences are analyzed in the present
paper: a [(0/90)4]s cross-ply laminate and a [(0/45/90/-45)2]s quasi-isotropic laminate. Three
open-hole-tension tests were reported in Ref. [120] for the cross-ply laminate giving a net-
section average strength of 124.1 ksi, which equates to an average failure load of 9,605 lb. Two
open-hole-tension tests were reported in Ref. [120] for the quasi-isotropic laminate giving a net-
section average strength of 100.9 ksi, which equates to an average failure load of 7,810 lb.
Progressive failure analyses of these two laminates are reported in Ref. [120] using a continuum-
damage-mechanics model based on crack density as an internal state variable. The predicted net-
section strength values (failure load values) of 122.5 ksi (9,482 lb) for the cross-ply laminate and
110.0 ksi (8,514 lb) for the quasi-isotropic laminate are given in Ref. [120].
Material data for this graphite-epoxy system are given in Ref. [120] as: E11=23.2 Msi,
E22=E33=1.30 Msi, G12=G13=0.90 Msi, G23=0.50 Msi, ν12=ν13=ν23=0.28, XT=412 ksi,
30
YT=ZT=8.72 ksi, XC=225 ksi, YC=ZC=24.3 ksi, S12=S13=13.76 ksi, S23=7.64 ksi, and GIc=0.86
in.-lb/in.2 The allowable strains are obtained by dividing the strength allowable values by the
elastic modulus for that component. The tensile and compressive elastic moduli are the same,
and the response is assumed to be linear elastic up to the ultimate strength values of the material.
Hence, the progressive failure analyses performed using either stress- or strain-based non-
interacting failure criteria should provide the same solution for this UMAT subroutine. The
UMAT input records for the lamina modeling approach are listed in the UMAT Implementation
section of the present paper.
As mentioned earlier in the present paper, plain-weave textile composites are often modeled by
treating each fabric layer as a pair of [0/90] cross-ply layers with each layer having half the
thickness of the single fabric layer. For the present 16-ply cross-ply laminate, the effective
engineering properties were calculated and used as the “apparent” mechanical properties for the
entire 16-ply laminate. These effective or apparent properties for the 0.1032-inch-thick laminate
are: Ex=Ey=12.29 Msi, νxy=0.029, Gxy=0.90 Msi, XT=YT=210.4 ksi, XC=YC=124.7 ksi, and
Sxy=Sxz=13.76 ksi. The corresponding ultimate normal strains are 0.0171 in./in. in tension, and
0.0101 in./in. in compression, and the ultimate shearing strain is 0.0153 in./in. The input records
for the UMAT subroutine based on smeared or effective engineering properties of this system
are:
** ==========================================================================================
** ==========================================================================================
** UMAT Property Data Definitions
** props(1-8):E11t,E22t,E33t,E11c,E22c,E33c,G12,G13,
** props(9-16):G23,nu12,nu13,nu23, Xt, Yt, Zt, Xc,
** props(17-24):Yc,Zc,S12,S13,S23,Eps11T,Eps22T,Eps33T,
** props(25-32):Eps11C,Eps22C,Eps33C,Gam12,Gam13,Gam23,Eps11Tmx,Eps22Tmx,
** props(33-40):Eps33Tmx,Eps11Cmx,Eps22Cmx,Eps33Cmx,Gam12mx,Gam13mx,Gam23mx,GIc,
** props(41-48):FPZ,SlimT,SlimC,SlimS,weibull(1),weibull(2),weibull(3),weibull(4),
** props(49-55):weibull(5),weibull(6),Dgrd(1),Dgrd(2),Dgrd(3),RECURS,PDA
** ==========================================================================================
** T800H/3900-2 graphite epoxy with 0.00645 inches per ply
** Smeared laminate values based on effective engineering properties Ex and Ey
** for the [(0/90)_4]_s laminate
*MATERIAL, NAME=SMEARED16
*USER MATERIAL, CONSTANTS=55
1.229E+07, 1.229E+07, 1.300E+06, 1.229E+07, 1.229E+07, 1.300E+06, 0.900E+06, 0.900E+06,
0.500E+06, 2.970E-02, 2.970E-02, 2.970E-02, 2.104E+05, 2.104E+05, 8.720E+03, 1.247E+05,
1.247E+05, 2.430E+04, 1.376E+04, 1.376E+04, 7.644E+03, 0.171E-01, 0.171E-01, 0.671E-02,
0.101E-01, 0.101E-01, 0.187E-01, 0.153E-01, 0.153E-01, 0.153E-01, 1.000E-01, 1.000E-01,
1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 0.860E+00,
2.000E-01, 0.000E+00, 0.000E+00, 0.000E+00, 1.000E+00, 1.000E+00, 1.000E+00, 1.000E+00,
1.000E+00, 1.000E+00, 5.000E-01, 5.000E-01, 5.000E-01, 1.000E+00, 3.000E+00
*DEPVAR
8
** ==========================================================================================
** ==========================================================================================

Progressive failure analyses are performed using these apparent or smeared properties with a
single equivalent layer and compared with the predictions obtained using a layer-by-layer
laminate modeling approach.
The finite element analysis model for this problem is the same as the analysis model used in Ref.
[29] and is shown in Figure 13. A right-hand coordinate system is used with positive z directed
out of the paper and toward the reader. A total of 80 4-node S4R5 shell elements are distributed
around the perimeter of the hole, and ten rings of elements are distributed between the hole
boundary and the edge of the coupon (i.e., the one-inch-square region in the vicinity of the hole).
The entire coupon is modeled using a total of 1,200 4-node S4R5 shell elements.

31
Boundary conditions indicated on Figure 13 were imposed on opposite ends of the finite element
models to simulate the clamped edge and the edge with an imposed uniform longitudinal (y-
direction) displacement δ. In the analyses, the bottom edge is fully restrained (i.e.,
u = v = w = θ x = θ y = θ z = 0 ), while the top edge is restrained except for the applied
displacement (i.e., u = w = θ x = θ y = θ z = 0 and v = δ ). The end displacement is specified
and incrementally increased (tensile loading), and the nonlinear solution process of
ABAQUS/Standard is under displacement control.
Nonlinear analyses are performed using a full Newton-Raphson procedure without extrapolation
for the next solution increment. The initial factor for the solution increment size is set at 0.005
with a minimum value specified as 0.0001 and a maximum value specified as 0.02. Automatic
solution increment size control is permitted. Solution increment size may be adjusted by the
nonlinear solution strategy implemented in ABAQUS/Standard. and twenty equilibrium
iterations can occur before the solution step size is reduced. In addition, progressive failure
analyses were performed using the various failure criteria based on the value of PDA. The
progressive failure analyses utilize the user-defined material model described in the present
paper and implemented as a UMAT subroutine within the ABAQUS/Standard nonlinear finite
element tool.
Progressive failure simulations are performed on the cross-ply open-hole-tension coupon using
different failure criteria (different values of PDA). The applied tension load as a function of end
displacement for different failure criteria are summarized in Figure 14. The short-dash line
indicates the average failure load from three tests reported in Ref. [120]. The long-dash line
represents the linear elastic response from the finite element analysis. Progressive failure
analysis results were generated using the nominal lamina material data, a single integration point
through the thickness of each lamina within the 16-ply laminate, and recursive degradation with
a degradation factor β of 0.5. Results obtained using two different values of the maximum
increment size factor for the nonlinear analysis are indicated in Figure 14 – 0.02 in Figure 14a
and 0.01 in Figure 14b. For both values of this factor, the predicted progressive failure responses
are similar. The maximum stress criteria and the maximum strain criteria predict nearly the same
failure load (i.e., peak load) of approximately 11,200 pounds. The Tsai-Wu failure polynomial
predicts a lower failure load (10,263 pounds) due the interactive nature of the failure criterion
since the material system is a linear elastic brittle material. The Hashin criteria under predict the
failure load by nearly 30% due to the degradation model associated with tensile fiber failure, also
referred to as fiber-matrix shear, see Eq. 30. As the maximum solution increment size factor is
reduced, the peak load predictions obtained from the other failure criteria are reduced except for
the Hashin criteria, which increased to 7,129 pounds. By reducing the maximum solution
increment size, a smooth response is still predicted, and numerical drift in the simulation due to
smaller solution increments is minimal.
The influence of the maximum increment size factor is shown in Figure 15a for the maximum
strain criteria and in Figure 15b for the Hashin criteria. Recursive degradation with a
degradation factor β of 0.5 is used in all cases. Four values of the maximum increment size
factor are considered. As the factor is reduced in size for the maximum strain criteria, the peak
load prediction is also reduced as shown in Figure 15a. However, the load value corresponding
to local fiber failure near the hole is the same for all values – approximately 7,000 pounds. For
the Hashin criteria, the peak load prediction generally were unchanged with decreasing
32
maximum solution increment size. However, for the smallest value considered (0.001), the peak
load prediction is reduced by approximately 30%. The maximum solution increment size factor,
which is used in the nonlinear solution strategy, apparently influences the post-ultimate behavior
of the simulation, and its influence on the response needs to be established for each new analysis
effort.
The influence of the degradation factor β is shown in Figure 16a for the maximum strain criteria
and in Figure 16b for the Hashin criteria. Recursive degradation and a maximum solution
increment size factor of 0.02 are used in all cases. Five values of the degradation factor are
considered for both criteria. As the degradation factor β is reduced, the peak load prediction for
the maximum strain criteria is also reduced as indicated in Figure 16a; however, the load value
corresponding to the initiation of local fiber failure near the hole is the same for all values of β.
As the degradation factor β is reduced, the peak load prediction for the Hashin criteria is also
reduced as indicated in Figure 16b. When recursive degradation is used, the degradation factor
and the solution increment size have a combined effect on the predicted progressive failure
response. For small values of the degradation factor, the local material stiffness coefficients
degrade in value at a faster rate than when larger degradation factors are used and tend to
simulate instantaneous degradation with a near zero degradation factor.
The influence of the maximum solution increment size factor is shown in Figure 17 for the
maximum strain criteria using instantaneous degradation with a degradation factor β of 10-6.
Three values of the maximum increment size factor are considered. As the factor is reduced, the
peak load prediction is essential the same as the peak load predicted using recursive degradation
with a degradation factor of 0.001. Reducing the maximum solution increment size factor has
little effect on the progressive failure response prediction when using instantaneous degradation
with a very small value of the degradation factor (i.e., β=10-6).
Peak failure loads are summarized in Table 6 for different criteria using recursive degradation
with a degradation factor β of 0.5 and a maximum increment size factor of 0.01. The maximum
stress (PDA=1) and maximum strain (PDA=2) criteria give essentially the same result since the
material is a linear elastic brittle material, and these predicted failure loads correlate well with
the reported average test value. The Tsai-Wu failure polynomial (PDA=3) predicts a similar
behavior and onset of damage; however, a lower failure load is predicted. The Hashin criteria
(PDA=4) under predict the peak load by nearly 25%.
Progressive failure simulations performed on the cross-ply open-hole-tension coupon using
different failure criteria are also performed using the apparent or smeared material properties for
the cross-ply laminate. These progressive failure analysis results are generated using the
effective engineering material data, three integration points through the laminate thickness, and
recursive degradation with a degradation factor β of 0.5. The maximum solution increment size
factor is 0.01 for the nonlinear analysis. The results shown in Table 6 are consistent with the
results obtained using the cross-ply laminate (ply-by-ply) modeling approach with a single
integration point in each layer; however, information on lamina failure mode is not recoverable.
In addition, results generated for the case of none integration points through the laminate
thickness agree with those generated using only three integration points.
Progressive failure simulations performed on the quasi-isotropic open-hole-tension coupon using
different failure criteria are summarized in Figure 18. The short-dash line indicates the average
failure load from two tests [120]. The long-dash line represents the linear elastic response from
33
the finite element analysis. These progressive failure analysis results are generated using the
nominal lamina material data, a single integration point through the thickness of each lamina,
recursive degradation with a degradation factor β of 0.5, and a maximum solution increment size
factor of 0.01. The maximum stress (PDA=1) and maximum strain (PDA=2) criteria again give
essentially the same result because the material is a linear elastic brittle material. The Tsai-Wu
failure polynomial (PDA=3) predicts a similar behavior and failure load since the laminate is
quasi-isotropic. The Hashin criteria (PDA=4) under predict the peak load. This under (or
conservative) prediction of the peak load by the Hashin criteria is due to the tensile-fiber failure
mode defined in Eq. 30 and associated degradation rules. Alternate failure criteria and
degradation rules are given by Chang et al. [120].
The predicted peak failure loads for the open-hole-tension coupons are summarized in Table 6.
For the cross-ply laminate case, the peak failure loads predicted by the maximum stress,
maximum strain, and Tsai-Wu criteria are within 10% of the average failure load from test. The
Hashin criteria gave very conservative predictions (lower by more than 25%). Similar trends are
obtained for the cross-ply case using smeared or apparent material data. For the quasi-isotropic
laminate case, the peak failure loads predicted by the maximum stress, maximum strain, and
Tsai-Wu criteria are within 2% of the average failure load from test. The Hashin criteria again
gave conservative predictions (lower by 23%).

Concluding Remarks
A user-defined material model for laminated composite materials and structures is described in
the present paper. The material model is leveraged on previous work in progressive failure
analysis of laminated polymeric composites. In addition, extensions of these models are
provided for bimodulus materials. Progressive failure analysis options are provided for different
point-stress methods with traditional failure initiation and material degradation models as well as
from a linear elastic model. These models are implemented within an UMAT subroutine and are
executed using the ABAQUS/Standard nonlinear finite element tool.
Each of the progressive failure analysis models is described in the present paper. For the ply-
discounting models, the failure initiation criteria are described as well as the procedure for
material degradation. Material degradation is achieved through a set of degradation factors
dependent on whether the failure mode is tension, compression or shear driven. Material
degradation is applied to the material stiffness coefficients rather than the elastic engineering
mechanical properties to maintain symmetry in the local material stiffness matrix. Material
degradation can be applied instantaneously or recursively in the ply-discounting models.
Recursive degradation in combination with fractional degradation factors is found to provide
reliable progressive failure solutions.
The features of the progressive failure analysis models implemented in this UMAT subroutine
are illustrated using a single two-dimensional shell finite element as a uniaxial demonstration
problem. The characteristics of stress-based and strain-based ply-discounting models are
illustrated. For material systems that exhibit nonlinear stress-strain behavior in the pre-ultimate
regime, strain-based failure initiation models are recommended. Having a capability to represent
the nonlinear pre-ultimate material response is a continuing need that is met in part by the MLT
formulation.

34
In addition, progressive failure simulations are performed for two 16-ply open-hole-tension
coupons. One coupon has a cross-ply lamination scheme, and the other coupon has a quasi-
isotropic lamination scheme. This material exhibited a linear elastic orthotropic brittle behavior.
The form of degradation, the degradation factor, and the solution increment size are found to
influence the progressive failure predictions. The maximum stress and maximum strain criteria
predicted nearly the same peak failure loads and nonlinear response. The Tsai-Wu failure
polynomial predicted lower peak failure loads but still in good agreement with test results. The
Hashin criteria typically under predicted the peak failure loads by approximately 25% for a
nominal value of the degradation factor (β=0.5) when combined with recursive degradation.
This result is tied to the material degradation model in Table 3 for the tensile fiber failure mode
given by Eq. 30. This expression typically implies a fiber-matrix shear failure, and material
degradation should only be applied to the in-plane shear coefficients [120]. However, typical
material degradation implementations associated with the Hashin criteria, including the present
UMAT subroutine, degrade the in-plane normal coefficients as well as the in-plane shear term
when this failure mode is detected, which leads to a very conservative failure prediction.

35
 References
1. Anon., ABAQUS/Standard User’s Manual, Version 6.5, ABAQUS, Inc., Providence, RI,
2004.
2. Jones, Robert M., Mechanics of Composite Materials, McGraw-Hill Book Company,
New York, 1975. Second Edition, Taylor and Francis, Philadelphia, PA, 1999.
3. Tsai, Stephen W. and Hahn, H. Thomas, Introduction to Composite Materials,
Technomic Publishers, Westport, CN, 1980.
4. Ochoa, O. and Reddy, J. N., Finite Element Analysis of Composite Laminates, Kluwer
Academic Publishers, Dordrecht, The Netherlands, 1992.
5. Silverman, Edward M. (compiler), Composite Spacecraft Structures Design Guide,
NASA CR 4708, Part 1, March 1996.
6. Silverman, Edward M. (compiler), Composite Spacecraft Structures Design Guide,
NASA CR 4708, Part 2, March 1996.
7. Hyer, Michael H., Stress Analysis of Fiber-Reinforced Composite Materials, McGraw-
Hill, Boston, 1998.
8. Herakovich, Carl T., Mechanics of Fibrous Composites, John Wiley and Sons, New
York, 1998.
9. Kelly, A. and Zweben, C. (editors), Comprehensive Composite Materials, Elsevier
Science Ltd., 2000.
10. Tsai, S. W. and Wu, E. M., “A General Theory of Strength for Composite Anisotropic
Materials,” Journal of Composite Materials, Vol. 5, January 1971, pp. 58-80.
11. Rosen, B. W. and Zweben, C. H., Tensile Failure Criteria for Fiber Composite
Materials, NASA CR-2057, August 1972.
12. Hashin, Z. “Failure Criteria for Unidirectional Fiber Composites,” ASME Journal of
Applied Mechanics, Vol. 47, No. 2, June 1980, pp. 329-334.
13. Lee, J. D., “Three Dimensional Finite Element Analysis of Damage Accumulation in
Composite Laminate,” Computers and Structures, Vol. 15, No. 3, 1982, pp. 335-350.
14. Ochoa, O. O. and Engblom, J. J., “Analysis of Failure in Composites,” Composites
Science and Technology, Vol. 28, 1987, pp. 87-102.
15. Chang, F. K. and Chang, K. Y., “A Progressive Damage Model for Laminated
Composites Containing Stress Concentrations,” Journal of Composite Materials, Vol.
21, No. 9, 1987, pp. 834-855.
16. Engelstad, S. P., Reddy, J. N., and Knight, N. F., Jr., “Postbuckling Response and
Failure Prediction of Graphite-Epoxy Plates Loaded in Compression,” AIAA Journal,
Vol. 30, No. 8, August 1992, pp. 2106-2112.
17. Krishnamurty, A.V. and Reddy, J. N., “Compressive Failure of Laminates and
Delamination Buckling: A Review,” The Shock and Vibration Digest, Vol. 25, No. 3,
March 1993, pp. 3-12.

36
18. Reddy, Y. S. and Reddy, J. N., “Three-Dimensional Finite Element Progressive Failure
Analysis of Composite Laminates under Axial Compression,” Journal of Composite
Technology and Research, Vol. 15, No. 2, Summer 1993, pp. 73-87.
19. Singh, S. B., Kumar, A., and Iyengar, N. G. R., “Progressive Failure of Symmetrically
Laminated Plates under Uni-axial Compression,” Structural Engineering and
Mechanics, Vol. 5, No. 4, 1997, pp. 433-450.
20. Sleight, D. W., Knight, N. F., Jr., and Wang, J. T., “Evaluation of a Progressive Failure
Analysis Methodology for Laminated Composite Structures,” AIAA Paper No. 97-
1187, presented at the AIAA/ASME/ASCE/AHS/ASC 38th Structures, Structural
Dynamics, and Materials Conference, Kissimmee, FL, April 7-10, 1997.
21. Moas, E. and Griffin, O. H., Jr., “Progressive Failure Analysis of Laminated Composite
Structures,” AIAA Paper No. 97-1186, presented at the AIAA/ASME/ASCE/AHS/ASC
38th Structures, Structural Dynamics, and Materials Conference, Kissimmee, FL, April
7-10, 1997
22. Soden, P. D., Hinton, M. J., and Kaddour, A. S., “A Comparison of the Predictive
Capabilities of Current Failure Theories for Composite Structures,” Composite Science
and Technology, Vol. 58, 1998, pp. 1225-1254.
23. Huybrechts, S., “Failure of Composite Structures,” in Advances in Composite Materials
and Mechanics, Arup Maji (editor), ASCE, 1999, pp. 34-41.
24. Dávila, C. G., Ambur, D. R., and McGowan, D. M., “Analytical Prediction of Damage
Growth in Notches Composite Panels Loaded in Compression,” Journal of Aircraft,
Vol. 37, No. 5, September-October 2000, pp. 898-905.
25. Barbero, E. J. and Lonetti, P., “An Inelastic Damage Model for Fiber Reinforced
Laminates,” Journal of Composite Materials, Vol. 36, No. 8, 2002, pp. 941-962.
26. Lin, W.-P. and Hu, H.T., “Nonlinear Analysis of Fiber-Reinforced Composite
Laminates Subjected to Uniaxial Tensile Load,” Journal of Composite Materials, Vol.
36, No. 12, 2002, pp. 1429-1450.
27. Lin, W.-P. and Hu, H.T., “Parametric Study on the Failure of Fiber-Reinforced
Composite Laminates under Biaxial Tensile Load,” Journal of Composite Materials,
Vol. 36, No. 12, 2002, pp. 1481-1503.
28. Huybrechts, S., Maji, A., Lao, J., Wegner, P., and Meink, T., “Validation of the
Quadratic Composite Failure Criteria with Out-of-Plane Shear Terms,” Journal of
Composite Materials, Vol. 36, No. 15, 2002, pp. 1879-1888.
29. Knight, N. F., Jr., Rankin, C. C., and Brogan, F. A., “STAGS Computational Procedure
for Progressive Failure Analysis of Laminated Composite Structures,” International
Journal of Non-Linear Mechanics, Vol. 37, No. 4-5, 2002, pp. 833-849.
30. Goyal, V. K., Jaunky, N., Johnson, E. R., and Ambur, D. R., “Intralaminar and
Interlaminar Progressive Failure Analyses of Composite Panels with Circular Cutouts,”
AIAA Paper No. 2002-1745, presented at the AIAA/ASME/ASCE/AHS/ASC 43rd
Structures, Structural Dynamics, and Materials Conference, Denver, CO, April 22-25,
2002.

37
31. Hinton, M., Kaddour, A. S., and Soden, P. D., Failure Criteria in Fibre-Reinforced-
Polymer Composites: The World-Wide Failure Exercise, Elsevier Science Publishing
Company, 2004.
32. Dávila, C. G., Camanho, P. P., and Rose, C. A., “Failure Criteria for FRP Laminates,”
Journal of Composite Materials, Vol. 39, No. 4, February 2005, pp. 323-345.
33. Chen, J., Crisfield, M., Kinloch, A. J., Busso, E. P., Matthews, F. L., and Qiu, Y.,
“Predicting Progressive Delamination of Composite Material Specimens via Interface
Elements,” Mechanics of Composite Materials and Structures, Vol. 6, No. 4, Oct.-Dec.
1999, pp. 303-317.
34. Camanho, P. P., Dávila, C. G., and Ambur, D. R., Numerical Simulation of
Delamination Growth in Composite Materials, NASA TP 2001-211041, August 2001.
35. Dávila, C. G., Camanho, P. P., and de Moura, M. F., “Mixed-Mode Decohesion
Elements for Analyses of Progressive Delamination,” AIAA Paper No. 2001-1486,
presented at the AIAA/ASME/ASCE/AHS/ASC 42nd Structures, Structural Dynamics,
and Materials Conference, Seattle, WA, April 16-19, 2001.
36. Poe, C. C., Jr. and Harris, C. E. (editors), Mechanics of Textile Composites Conference,
NASA CP-3311, Parts 1 and 2, October 1995.
37. Cox, B. and Flanagan, G., Handbook of Analytical Methods for Textile Composites,
NASA CR-4750, March 1997.
38. Poe, C. C., Jr., Dexter, H. B., and Raju, I. S., “Review of the NASA Textile Composites
Research,” Journal of Aircraft, Vol. 36, No. 5, September-October 1999, pp. 876-884.
39. Whitcomb, J. D., Three-Dimensional Stress Analysis of Plain Weave Composites,
NASA TM-101672, November 1989.
40. Glaessgen, E. H., Pastore, C. M., Griffin, O. H., and Birger, A., “Geometrical and Finite
Element Modelling of Textile Composites,” Composites: Part B, Vol. 27B, No. 1, 1996,
pp. 43-50.
41. Aitharaju, V. R. and Averill, R. C., “Three-Dimensional Properties of Woven-Fabric
Composites,” Composite Science and Technology, Vol. 59, 1990, pp. 1901-1911.
42. Ishikawa, T. and Chou, T.-W., “Stiffness and Strength Behaviour of Woven Fabric
Composites,” Journal of Material Science, Vol. 17, 1982, pp. 3211-3220.
43. Raju, I. S. and Wang, J. T., “Classical Laminate Theory Models for Woven Fabric
Composites,” Journal of Composites Technology and Research, Vol. 16, No. 4, October
1994, pp. 289-303.
44. Naik, R. A., “Failure Analysis of Woven and Braided Fabric Reinforced Composites,”
Journal of Composite Materials, Vol. 29, No. 17, 1995, pp. 2334-2363.
45. Naik, Rajiv, TEXCAD – Textile Composite Analysis for Design, Version 1.0 User’s
Manual, NASA CR-4639, December 1994.
46. Chung, P. W. and Tamma, K. K., “Woven Fabric Composites – Developments in
Engineering Bounds, Homogenization and Applications,” International Journal for
Numerical Methods in Engineering, Vol. 45, No. 12, 1999, pp. 1757-1790.

38
47. Swan, C. C., “Unit Cell Analysis of 3-D Graphitic Textile-Reinforced Polymer Matrix
Composites,” in Advances in Composite Materials and Mechanics, Arup Maji (editor),
ASCE, 1999, pp. 65-75.
48. Kwon, Y. W. and Altekin, A., “Multilevel, Micro/Macro-Approach for Analysis of
Woven-Fabric Composite Plates,” Journal of Composite Materials, Vol. 36, No. 12,
April 2002, pp. 1005-1022.
49. Cox, B., Failure Models for Textile Composites, NASA CR-4686, August 1995.
50. Svensson, N. and Gilchrist, M. D., “Modelling of Failure of Structural Textile
Composites,” Computational Mechanics, Vol. 26, No. 3, September 2000, pp. 223-228.
51. Bolotina, K. S., Murzakhanov, G. Kh., and Schugorev, V. N., “Damage and Fracture
Mechanisms of Textile Composites,” Mechanics of Composite Materials, Vol. 36, No.
1, 2000, pp. 45-48.
52. Blackketter, D. M., Walrath, D. E., and Hansen, A. C., “Modeling Damage in a Plain
Weave Fabric-Reinforced Composite Material,” Journal of Composites Technology and
Research, Vol. 15, No. 2, Summer 1993, pp. 136-142.
53. Karayaka, M. and Kurath, P., “Deformation and Failure Behavior of Woven Composite
Laminates,” Journal of Engineering Materials and Technology, Vol. 116, No. 2, April
1994, pp. 222-232.
54. Kollegal, M. G. and Sridharan, S., “Compressive Behavior of Plain Weave Lamina,”
Journal of Composite Materials, Vol. 32, No. 15, 1998, pp. 1334-1355.
55. Adeyemi, N. B., Shivakumar, K. N., and Avva, V. S., “Delamination Fracture
Toughness of Woven-Fabric Composites under Mixed-Mode Loading,” AIAA Journal,
Vol. 37, No. 4, April 1999, pp. 517-520.
56. Gao, F., Boniface, L., Ogin, S. L., Smith, P. A., and Greaves, R. P., “Damage
Accumulation in Woven-Fabric CFRP Laminates under Tensile Loading: Part 1.
Observations of Damage Accumulation,” Composite Science and Technology, Vol. 59,
1999, pp. 123-136.
57. Gao, F., Boniface, L., Ogin, S. L., Smith, P. A., and Greaves, R. P., “Damage
Accumulation in Woven-Fabric CFRP Laminates under Tensile Loading: Part 2.
Modelling the Effect of Damage on Macro-Mechanical Properties,” Composite Science
and Technology, Vol. 59, 1999, pp. 137-145.
58. Schwer, L. E. and Whirley, R. G., “Impact of a 3D Woven Textile Composite Thin
Panel: Damage and Failure Modeling,” Mechanics of Composite Materials and
Structures, Vol. 6, No. 1, 1999, pp. 9-30.
59. Tabiei, A. and Jiang, Y., “Woven Fabric Composite Material Model with Material
Nonlinearity for Nonlinear Finite Element Simulation,” International Journal of Solids
and Structures, Vol. 36, 1999, pp. 2757-2771.
60. Tabiei, A. and Ivanov, I., “Materially and Geometrically Non-Linear Woven Composite
Micro-Mechanical Model with Failure for Finite Element Simulations,” International
Journal of Non-Linear Mechanics, Vol. 39, No. 2, March 2004, pp. 175-188.

39
61. Fitzer, E. and Manocha, L. M., Carbon Fiber, Carbon Reinforcements and
Carbon/Carbon Composites, Springer, New York, 1998.
62. Ambartsumyan, S. A., “The Basic Equations and Relations of the Different-Modulus
Theory of Elasticity of an Anisotropic Body,” Mechanics of Solids, Vol. 4, No. 3, 1969,
pp. 48-56. Originally in Russian in Akademiia Nauk SSSR, Izvestiia, Mekhanika
Tverdogo Tela, Vol. 4, No. 3, pp. 51-61, 1969.
63. Tabaddor, F., “Constitutive Equations for Bimodulus Elastic Materials,” AIAA Journal,
Vol. 10, No. 4, April 1972, pp. 516-518.
64. Jones, R. M., “Stress-Strain Relations for Materials with Different Moduli in Tension
and Compression,” AIAA Journal, Vol. 15, No. 1, January 1977, pp. 16-23.
65. Bert, C. W., “Models for Fibrous Composites with Different Properties in Tension and
Compression,” ASME Journal of Engineering Materials and Technology, Vol. 99, No.
4, October 1977, pp. 344-349.
66. Reddy, J. N. and Chao, W. C., “Finite-Element Analysis of Laminated Bimodulus
Composite-Material Plates,” Computers and Structures, Vol. 12, 1980, pp. 245-251.
67. Vijayakumar, K. and Rao, K. P., “Stress-Strain Relations for Composites with Different
Stiffnesses in Tension and Compression,” Computational Mechanics, Vol. 2, No. 3,
1987, pp. 167-175.
68. Guess, T. R. and Hoover, W. R., “Fracture Toughness of Carbon-Carbon Composites,”
Journal of Composite Materials, Vol. 7, January 1973, pp. 2-20.
69. Perry, J. L. and Adams, D. F., “An Experimental Study of Carbon-Carbon Composite
Materials,” Journal of Materials Science, Vol. 9, No. 11, 1974, pp. 1764-1774.
70. Pollock, P. B., “Tensile Failure in 2-D Carbon-Carbon Composites,” Carbon, Vol. 28,
No. 5, 1990, pp. 717-732.
71. Gupta, V., Anand, K., and Dartfort, D., “Failure Mechanisms of Laminated Carbon-
Carbon Composites – Part I: Under Uniaxial Compression Loading,” Acta Metallurgica
et Materialia, Vol. 42, No. 3, 1994, pp. 781-795.
72. Anand, K., Gupta, V., and Dartfort, D., “Failure Mechanisms of Laminated Carbon-
Carbon Composites – Part II: Under Shear Loading,” Acta Metallurgica et Materialia,
Vol. 42, No. 3, 1994, pp. 797-809.
73. Anand, K., Gupta, V., and Y-He, M., “A Numerical Study of the Compression and
Shear Failure of Woven Carbon-Carbon Laminates,” Journal of Composite Materials,
Vol. 29, No. 18, 1995, pp. 2446-2463.
74. Ochoa, O. O., Aikens, Q. Y., and Engblom, J. J., “An Integrated Numerical and
Statistical Model for Woven 2D Carbon-Carbon Composites,” in Mechanics of
Composites at Elevated and Cryogenic Temperatures, S. N. Singhal, W. F. Jones, and
C. T. Herakovich (editors), ASME AMD Vol. 108, 1991, pp. 157-169.
75. Ashley, T. H. and Ochoa, O. O., “Structural Response of Oxidation-Resistant
Carbon/Carbon Composites,” Composite Science and Technology, Vol. 59, 1999, pp.
1959-1967.

40
76. Chapman, C. D. and Whitcomb, J. D., “Strategy for Modeling Eight-Harness Satin
Weave Carbon/Carbon Composites Subjected to Thermal Loads,” AIAA Paper No.
AIAA-96-1520-CP, presented at the AIAA/ASME/ASCE/AHS/ASC 37th Structures,
Structural Dynamics, and Materials Conference, Salt Lake City, UT, April 15-17, 1996.
77. Aubard, X., Cluzel, C., Guitard, L., and Ladevèze, P., “Damage Modeling at Two
Scales for 4D Carbon/Carbon Composites,” Computers and Structures, Vol. 78, 2000,
pp. 83-91.
78. Piat, R. and Schnack, E., “Hierarchical Material Modeling of Carbon/Carbon
Composites,” Carbon, Vol. 41, 2000, pp. 2121-2129.
79. Siron, O. and Lamon, J., “Damage and Failure Mechanisms of a 3-Directional
Carbon/Carbon Composite under Uniaxial Tensile and Shear Loads,” Acta Material,
Vol. 46, No. 18, 1998, pp. 6631-6643.
80. Hatta, H., Denk, L., Watanabe, T., Shiota, I., and Aly-Hassan, M. S., “Fracture
Behavior of Carbon-Carbon Composites with Cross-ply Lamination,” Journal of
Composite Materials, Vol. 38, No. 17, September 2004, pp. 1479-1494.
81. Hatta, H., Aoi, T., Kawahara, I., Kogo, Y., and Shiota, I., “Tensile Strength of Carbon-
Carbon Composites: I – Effect of C-C Density,” Journal of Composite Materials, Vol.
38, No. 19, October 2004, pp. 1667-1684.
82. Wakefield, R. M. and Fowler, K. R., “A Method for Determining Structural Properties
of RCC Thermal Protection Material,” AIAA Paper No. 78-869, presented at the 2nd
AIAA/ASME Thermophysics and Heat Transfer Conference, Palo Alto, CA, May 24-
26, 1978.
83. Gordon, M. P., Leading Edge Structural Subsystem and Reinforced Carbon-Carbon
Reference Manual, Boeing Report No. KSC-KLO-98-008, October 19, 1998.
84. Curry, D. M., Scott, H. C., and Webster, C. N., “Material Characterization of Space
Shuttle Reinforced Carbon-Carbon,” in The Enigma of the Eighties: Environment,
Economics, Energy, Vol. 24, Part 2, SAMPE, 1979, pp. 1524-1539.
85. Hahn, H. T., “Nonlinear Behavior of Laminated Composites,” Journal of Composite
Materials, Vol. 7, No. 2, April 1973, pp. 257-271.
86. Hahn, H. T., “A Note on Determination of the Shear Stress-Strain Response of
Unidirectional Composites,” Journal of Composite Materials, Vol. 7, No. 3, July 1973,
pp. 383-386.
87. Gehman, Jr., H. W., et al., Columbia Accident Investigation Board Report, Volume 1,
NASA, Washington, DC, August 2003.
88. Fahrenthold, E. P. and Park, Y.-K., “Simulation of Foam Impact Effects on
Components of the Space Shuttle Thermal Protection System,” AIAA Paper No. 2004-
940, presented at the AIAA Aerospace Sciences Meeting, Reno, NV, January 5-8, 2004.
Also available in Journal of Spacecraft and Rockets, Vol. 42, No. 2, 2005, pp. 201-207.
89. Melis, M., Carney, K., Gabrys, J., Fasanella, E., and Lyle, K., “A Summary of the
Space Shuttle Tragedy and the Use of LS-DYNA in the Accident Investigation and

41
 Return to Flight Efforts,” in Proceedings of the Eighth International LS-DYNA Users
Conference, Dearborn, Michigan, May 2-4, 2004.
90. Park, Y.-K. and Fahrenthold, E. P., “Simulation of Hypervelocity Impact Effects on
Reinforced Carbon-Carbon,” Journal of Spacecraft and Rockets, Vol. 43, No. 1,
January-February 2006, pp. 200-206.
91. Fasanella, E. L., Lyle, K. H., Gabrys, J., Melis, M., and Carney, K., “Test and Analysis
Correlation of Foam Impact onto Space Shuttle Wing Leading Edge RCC Panel 8,” in
Proceedings of the Eighth International LS-DYNA Users Conference, Dearborn,
Michigan, May 2-4, 2004.
92. Carney, K., Melis, M., Fasanella, E., Lyle, K., and Gabrys, J., “Material Modeling of
the Space Shuttle Leading Edge and External Tank Materials for Use in the Columbia
Accident Investigations,” in Proceedings of the Eighth International LS-DYNA Users
Conference, Dearborn, Michigan, May 2-4, 2004.
93. Gabrys, J., Schatz, J., Carney, K., Melis, M., Fasanella, E., and Lyle, K., “The Use of
LS-DYNA in the Columbia Accident Investigation and Return to Flight Activities,” in
Proceedings of the Eighth International LS-DYNA Users Conference, Dearborn,
Michigan, May 2-4, 2004.
94. Goldberg, R. K. and Carney, K., “Modeling the Nonlinear, Strain Rate Dependent
Deformation of Woven Ceramic Matrix Composites with Hydrostatic Stress Effects
Included,” in Proceedings of the Eighth International LS-DYNA Users Conference,
Dearborn, Michigan, May 2-4, 2004.
95. Jackson, K. E., Fasanella, E. L., Lyle, K. H., and Spellman, R. L., The Influence of
Mesh Density on the Impact Response of a Shuttle Leading-Edge Panel Finite Element
Simulation, NASA TM-2004-213501, November 2004.
96. Stockwell, A. E., The Influence of Model Complexity on the Impact Response of a
Shuttle Leading-Edge Panel Finite Element Simulation, NASA CR-2005-213535,
March 2005.
97. Fasanella, E. L., Jackson, K. E., Lyle, K. H., Jones, L. E., Hardy, R. C., Spellman, R. L.,
Carney, K. S., Melis, M. E., and Stockwell, A. E., “Dynamic Impact Tolerance of the
Shuttle RCC Leading Edge Panels Using LS—DYNA, AIAA Paper No. 2005-3631,
presented at the 41st AIAA/ASME/SAE/ASEE Joint Propulsion Conference and
Exhibit, July 10-13, 2005, Tucson, AZ.
98. Jackson, K. E., Fasanella, E. L., Lyle, K. H., and Spellman, R. L., “A Mesh Refinement
Study on the Impact Response of a Shuttle Leading-Edge Panel Finite Element
Simulation,” in Proceedings of the Ninth International LS-DYNA Users Conference,
Dearborn, Michigan, June 4-6, 2006.
99. Carney, K. S., Benson, D. J., Du Bois, P., and Lee, R., “A High Strain rate Model with
Failure for Ice in LS-DYNA,” in Proceedings of the Ninth International LS-DYNA
Users Conference, Dearborn, Michigan, June 4-6, 2006.
100. Anon., LS-DYNA Users Manual – Volume II (Material Models, References and
Appendices), Version 960, Livermore Software Technology Corporation, Livermore,
CA, March 2001.

42
101. Matzenmiller, A., Lubliner, J., and Taylor, R. L., “A Constitutive Model for Anisotropic
Damage in Fiber-Composites,” Mechanics of Materials, Vol. 20, No. 2, April 1995, pp.
125-152.
102. Schweizerhof, K., Weimer, K., Münz, Th., and Rottner, Th., “Crashworthiness Analysis
with Enhanced Composite Material Models in LS-DYNA – Merits and Limits,” in
Proceedings of the 5th International LS-DYNA Users Conference, Southfield, MI,
September 21-22, 1998.
103. Williams, K. and Vaziri, R., “Finite Element Analysis of the Impact Response of CFRP
Composite Plates,” in Proceedings of Tenth International Conference on Composite
Materials – Volume V: Structures, A. Poursartip and K. Street (editors), Woodhead
Publishing Limited, 1995, pp. 647-654.
104. Williams, K. V., Vaziri, R., Floyd, A. M., and Poursartip, A., “Simulation of Damage
Progression in Laminated Composite Plates using LS-DYNA,” in Proceedings of the 5th
International LS-DYNA Users Conference, Southfield, MI, September 21-22, 1998.
105. Delfosse, D. and Poursartip, A., “Energy-Based Approach to Impact Damage in CFRP
Laminates,” Composites Part A, Vol. 28A, 1997, pp. 647-655.
106. Nandlall, D, Williams, K., and Vaziri, R., “Numerical Simulation of the Ballistic
Response of GRF Plates,” Composite Science and Technology, Vol. 58, No. 9, 1998,
pp. 1463-1469.
107. Whitney, J. M., “Shear Correction Factors for Orthotropic Laminates under Static
Load,” ASME Journal of Applied Mechanics, Vol. 40, March 1973, pp. 302-304.
108. Krajcinovic, D., “Continuum Damage Mechanics,” Applied Mechanics Reviews, Vol.
37, No. 1, January 1984, pp. 1-6.
109. Talreja, R., “Modeling of Damage Development in Composites using Internal Variable
Concepts,” in Damage Mechanics in Composites, A. S. D. Wang and G. K. Haritos
(editors), ASME AD-Vol. 12, 1987, pp. 11-16.
110. Ladevèze, P. and LeDantec, E., “Damage Modeling of the Elementary Ply for
Laminated Composites,” Composites Science and Technology, Vol. 43, 1992, pp. 257-
267.
111. Shahid, I. and Chang, F.-K., “Accumulative Damage Model for Tensile and Shear
Failures of laminated Composite Plates,” Journal of Composite Materials, Vol. 29, No.
7, July 1995, pp. 926-981.
112. Krajcinovic, D., Damage Mechanics, North Holland Publishing, Amsterdam, 1996.
113. Weibull, W., “A Statistical Distribution Function of Wide Applicability,” ASME
Journal of Applied Mechanics, Vol. 18, September 1951, pp. 293-297.
114. Creasy, T. S., “Modeling Analysis of Tensile Tests of Bundled Filaments with a
Bimodal Weibull Survival Function,” Journal of Composite Materials, Vol. 36, No. 2,
2002, pp. 183-194.
115. Hahn, H. T. and Tsai, S. W., “On the Behavior of Composite Laminates after Initial
Failures,” Journal of Composite Materials, Vol. 8, July 1974, pp. 288-305.

43
116. Petit, P. H. and Waddoups, M. E., “A Method of Predicting the Nonlinear Behavior of
Laminated Composites,” Journal of Composite Materials, Vol. 3, January 1969, pp. 2-
19.
117. Matzenmiller, A. and Schweizerhof, K. “Crashworthiness Simulations of Composite
Structures – A First Step with Explicit Time Integration,” in Nonlinear Computational
Mechanics – State of the Art, P. Wriggers and W. Wagner (editors), Springer-Verlag,
Berlin, 1991, pp. 642-670.
118. Hillerborg, A., “The Theoretical Basis of a Method to Determine the Fracture Energy
GF of Concrete,” Materiaux et Constructions (Materials and Structures), Vol. 18, No.
106, 1985, pp. 291-296.
119. Hillerborg, A., “Numerical Methods to Simulate Softening and Fracture of Concrete,”
in Fracture Mechanics of Concrete: Structural Application and Numerical Calculation,
G. C. Sih and A. DiTommaso (editors), Martinus Nijhoff Publishers, Dordrecht, 1985,
pp. 141-170.
120. Chang, F.-K., Qing, X., Sun, H.-T., and Yan, Y., Damage Tolerance-Based Design of
Bolted Composite Joints, Final Report under Contract No. TZ-370444-07LHN,
Department of Aeronautics and Astronautics, Stanford University, 2000.

44
 Appendix A – Description of ABAQUS UMAT Arguments
The main purpose of a user-defined material model is to return a trial stress state and trial values
for the constitutive coefficients for the given strain state. Input values to an ABAQUS/Standard
UMAT subroutine [1] include the strain state from the previous solution increment (STRAN
array) and the increment in strain from that point to the current iteration cycle (DSTRAN array).
Then UMAT returns the current strain state (STRAN on exit), trial stress state (STRESS array)
and trial constitutive coefficients (DDSDDE array), and updated solution-dependent variables set
by the user (STATEV array). These trial values are computed based on the strain state from the
previous solution increment, the increment in strain from that point to the present iteration cycle
of the current solution increment, and the trial constitutive coefficients corresponding to the
present iteration cycle of the current solution increment.
The calling sequence for UMAT subroutine based on the ABAQUS/Standard documentation [1]
is:
SUBROUTINE UMAT(STRESS,STATEV,DDSDDE,SSE,SPD,SCD,RPL,DDSDDT,
1 DRPLDE,DRPLDT,STRAN,DSTRAN,TIME,DTIME,TEMP,DTEMP,PREDEF,DPRED,
2 CMNAME,NDI,NSHR,NTENS,NSTATV,PROPS,NPROPS,COORDS,DROT,PNEWDT,
3 CELENT,DFGRD0,DFGRD1,NOEL,NPT,LAYER,KSPT,KSTEP,KINC)

The arguments in the UMAT subroutine calling sequence are defined as followed:
• STRESS – stress vector at the beginning of the solution increment on entry and updated
to the trial stress vector at the present iteration of the current increment; dimensioned
NTENS
• STATEV – array of solution-dependent variables saved from the previous solution
increment at each material point and updated on exit to the trial values at the present
iteration for the current solution increment; defined by the UMAT developer and equal to
NSTATV in number
• DDSDDE – matrix of secant constitutive coefficients dimensioned NTENS × NTENS
• SSE – strain energy density at a material point
• SPD – specific plastic dissipation at a material point
• SCD – specific creep or viscous dissipation at a material point
• RPL – volumetric heat generation per unit time
• DDSDDT – variation of the stress increments with respect to temperature; dimensioned
NTENS
• DRPLDE – variation of RPL with respect to strain increment; dimensioned NTENS
• DRPLDT – variation of RPL with respect to temperature increment, dimensioned NTENS
• STRAN – total strain from the previous solution increment on entry; total strain for
present iteration of the current increment on exit; dimensioned NTENS

45
• DSTRAN – strain increment from previous solution increment to present iteration of the
current solution increment; dimensioned NTENS
• TIME – array with two entries; TIME(1) is the value of step time at beginning of
current increment; TIME(2) is the value of total time at the beginning of the current
increment
• DTIME – time increment or step size in a static analysis
• TEMP – temperature at the beginning of the increment
• DTEMP – increment of temperature
• PREDEF – array of interpolated values of predefined field variables at this material point
at the start of this increment
• DPRED – array of increments of predefined field variables
• CMNAME – name of material within this UMAT subroutine
• NDI – number of direct stress terms: 2 for shell elements, 3 for solid elements
• NSHR – number of shear stress terms: 1 for shell elements, 3 for solid elements
• NTENS – size of the stress or strain component array; dimensioned NDI+NSHR
• NSTATV – number of solution-dependent variable to be archived at each material point
• PROPS – material properties and parameters for this material model; defined and ordered
by the UMAT developer; dimensioned NPROPS
• NPROPS – number of properties and values defined for this material model
• COORDS – array containing the coordinates of this material point; dimensioned 3
• DROT – rotation increment matrix; dimensioned 3×3
• PNEWDT – ratio of suggested new time increment to the time increment being used
• CELENT – characteristic element length
• DFGRD0 – array containing the deformation gradient at the beginning of the increment;
dimensioned 3×3
• DFGRD1 – array containing the deformation gradient at the end of the increment;
dimensioned 3×3
• NOEL – element number for this material point
• NPT – surface integration point number within the NOEL element
• LAYER – layer number within the total laminate for this material point
• KSPT – section point within the current layer in the laminate; continuous numbering of
the layer integration points throughout the laminate from bottom layer to top layer

46
 • KSTEP – ABAQUS analysis step number; typically constant and equal to unity
• KINC – ABAQUS solution increment number in the analysis

Typically a user-defined material model is defined within a single FORTRAN file, commonly
named UMAT.f; however, a user may choose to develop the material model using modular
programming and have several subroutines that are called by the UMAT subroutine as an overall
driver subroutine. It is the user’s responsibility to insure that the subroutine names are unique.
ABAQUS/Standard UMAT documentation recommends a starting letter of “k” for the name of
each user-supplied subroutine called by the UMAT subroutine (e.g., kDGRADE instead of just
DGRADE).

47
 Appendix B – LS-DYNA MAT58 Input Data
The input for material type 58 (MAT58) of LS-DYNA is summarized in this appendix based on
the descriptions given in the LS-DYNA User Manual [100] and is included for the convenience
of the reader. The material name for MAT58 in keyword format is
*MAT_LAMINATED_COMPOSITE_FABRIC. Seven input records are required for this material type;
however, all values do not need to be defined or specified. The entries on the first three records
and the last two records are of primary interest. MAT58 input variables are listed next.
Record 1
• MID – material identification number
• RO – mass density
• EA – Young’s elastic modulus in the longitudinal direction, a
• EB – Young’s elastic modulus in the transverse direction, b
• EC – Young’s elastic modulus in the normal direction, c – not used
• PRBA – Poisson’s ratio
• TAU1 – stress limit for the first nonlinear part of the shear stress vs. shear strain curve
• GAMM1– strain limit for the first nonlinear part of the shear stress vs. shear strain
curve
Record 2
• GAB – elastic shear modulus ab
• GBC – elastic shear modulus bc
• GCA – elastic shear modulus ca
• SLIMT1 – factor to determine the minimum stress limit after stress maximum (fiber
tension)
• SLIMC1 – factor to determine the minimum stress limit after stress maximum (fiber
compression)
• SLIMT2 – factor to determine the minimum stress limit after stress maximum (matrix
tension)
• SLIMC2 – factor to determine the minimum stress limit after stress maximum (matrix
compression)
• SLIMS – factor to determine the minimum stress limit after stress maximum (shear)
Record 3
• AOPT – material axes option
• TSIZE – time step size for automatic element erosion
• ERODS – maximum effective strain for element layer failure. A value of unity would
equal 100% strain
48
 • SOFT – softening reduction factor for strength in the crash front
• FS – failure surface type
Record 4
• XP, YP, ZP – define the coordinates of point p for AOPT=1
• A1, A2, A3 – define the components of vector a for AOPT=2
Record 5
• V1, V2, V3 – define the components of vector v for AOPT=3
• D1, D2, D3 – define the components of vector d for AOPT=2
• BETA – material angle in degrees for AOPT=3
Record 6
• E11C – strain at longitudinal compressive strength, a-axis
• E11T – strain at longitudinal tensile strength, a-axis
• E22C – strain at transverse compressive strength, b-axis
• E22T – strain at transverse compressive strength, b-axis
• GMS – strain at shear strength, ab plane
Record 7
• XC – longitudinal compressive strength
• XT – longitudinal tensile strength
• YC – transverse compressive strength
• YT – transverse tensile strength
• SC – shear strength, ab plane
If the quantity E11C*EA/XC is less than zero, then the input parameters E11C, E11T, E22C,
E22T, and GMS are not interpreted as strains but instead they are interpreted as the Weibull scale
parameters (m) as defined by Matzenmiller et al. [101]; otherwise, the Schweizerhof et al. [102]
approach is followed.6

6
Based on personal communication with Dr. James Day, LSTC, December 16, 2004.

49
 Appendix C – MLT Continuum-Damage-Mechanics Model
For the present study, the continuum-damage-mechanics (CDM) model described by
Matzenmiller et al. [101] and extended by Schweizerhof et al. [102] is referred to as the MLT
model. The MLT model is considered as an optional damage progression approach implemented
within the current UMAT subroutine that can be selected by the user. The MLT model was
originally developed for uniaxial polymeric composites [101], and hence again, application to
other laminate types or non-polymeric composites represents a significant approximation.
Schweizerhof et al. [102] implemented the MLT model within LS-DYNA, and within LS-
DYNA, the MLT model is referred to as Material Type 58 or simply MAT58. Williams and
Vaziri [103] also have evaluated this CDM model for application to the impact of laminated
composite plates; however, the primary references for theoretical and implementation details of
the MLT model are Refs. [101, 102]. The explicit evaluation of failure initiation criteria has a
different meaning for the CDM model since damage evolution is dependent on a set of “internal
state variables”. The MLT internal state variables, denoted by the symbol ωij, are dependent on
the sign of the normal strain terms (i.e., different for tension and compression) and independent
of the sign of the shear strain terms.
The elastic constitutive coefficients themselves, as well as the stress-strain behavior, are
modified by this set of internal state variables. Nonlinearity in the pre-ultimate stress-strain
response is dependent on the ratio of the secant modulus determined from the ultimate values
and the initial linear elastic modulus. The initial elastic modulus is actually a misnomer in that
this value is unrelated to the modulus of the stress-strain curve but instead contributes to defining
the nonlinearity of the stress-strain curve prior to ultimate stress. This modulus ratio also
influences the behavior of the internal state variables as a function of the applied strain. The
MLT stress-strain curve is based on a Weibull-distribution function [113] to represent the stress-
strain nonlinearity through the use of internal state variables for damage evolution.
The internal state variables introduced by the MLT model vary from zero (indicating no damage)
to unity (indicating complete damage). These internal state variables increase from zero as the
local strain level increases and serve as indicators of the local reduction in load carrying
capability of a lamina. This reduction is caused by failure of weaker material or the coalescence
of local defects within the lamina. Within a lamina, some fiber bundles are statistically stronger
while others are weaker than the average fiber bundle, and the ultimate strength for the lamina is
the accumulation of all fiber bundle strengths. As the strain level increases, the weaker fiber
bundles fail and the total number of fiber bundles available to carry load decreases. However,
those fiber bundles that remain are the statistically stronger fiber bundles. This behavior is
accounted for in the MLT model by the Weibull-distribution assumption for the stress-strain
curve. A reduction in the number of load-carrying fiber bundles relates to a reduction in the
effective lamina cross-sectional area. Hence, a possible physical interpretation of the MLT
internal state variables is that they represent the fractional loss of load-carrying capability in a
particular mode or direction within an individual lamina.
The overall approach for the MLT model involves a series of key steps. The first step is
selecting appropriate values of the input parameters that affect the MLT stress-strain
representation. For example, in a one-dimensional uniaxial problem, five sets of parameters (or
nine independent values) need to be assigned by the user: a pseudo initial tensile modulus E0t ,

50
the ultimate tensile strength values Xt,c, ultimate strain values εult t,c
, stress limit factors after
ultimate SLIM or α , and maximum strain limits EROD or εmax . The first three sets of
t,c t,c

parameters can be used to match measured stress-strain behavior up to ultimate including

bimodulus material behavior and nonlinearity in the pre-ultimate response. The last two sets of
parameters can be tuned to match the final damage state – within the approximations of the
analysis model. Thus, an empirically derived material model can be defined – one that matches
the stress-strain response up to the ultimate strength values in tension and compression (i.e., pre-
ultimate response) and is empirically correlated to the post-ultimate response.
Within LS-DYNA, the MLT model is known as MAT58, and the input variables are used to
define a nonlinear, bimodulus stress-strain behavior through the definition of five parameters: an
initial modulus, tension and compression ultimate strength allowable values, and tension and
compression ultimate strain allowable values. A secant modulus can be defined as the ultimate
strength allowable value divided by the ultimate strain allowable value in tension or
compression. The ratio of the initial modulus and this secant modulus influences the behavior of
the internal state variable as a function of the applied strain. In the LS-DYNA implementation,
Schweizerhof et al. [102] also implemented stress limiting factors (SLIMT, SLIMC, SLIMS) and
a maximum strain limit (ERODS) that control the post-ultimate stress-strain response (i.e.,
impose a perfect-plasticity-like stress-strain behavior and a maximum-strain-based element-
removal criterion).
The next steps are associated with a sequence of computations performed for the MLT model.
Based on a computed strain state, a trial stress state is determined using the linear elastic
constitutive relations based on the initial elastic moduli. The Hashin failure criteria [12] are then
evaluated using these trial stresses where the failure indices are considered as “loading” indices.
Once a loading index exceeds unity, then the constitutive terms are evaluated using the internal
state variable approach and new trial stresses are determined. Note that this initial elastic state
may be neglected (or insignificant) if the material system exhibits the onset of damage almost
immediately. Next, when the ultimate strength allowable value has been exceeded, the trial
stresses are limited to minimum values defined by the stress limit factors (SLIM) and the
ultimate strength allowable values based on the Schweizerhof et al. [102] implementation.
These post-ultimate stress limits are maintained until the computed strain values reach specified
maximum strain limits at which time the applicable trial stress components are set to zero and the
values of the state variables ωij become unity. After this occurs, this material point only
minimally contributes to the local material stiffness. Once all through-the-thickness material
points at a specific surface-integration point for two-dimensional finite elements have a value of
unity for the state variable, then that local material is considered “failed.” For two-dimensional
finite elements with a single surface-integration point, such an occurrence would result in the
removal of that finite element from any further computations (this process is often called element
extinction or element erosion). For two-dimensional finite elements with multiple surface-
integration points, all surface-integration points would need to “fail” for the element removal
process to occur. These steps define the basic computational process for the MLT model in an
explicit finite element tool such as LS-DYNA. Similar steps can be applied to three-dimensional
finite elements. However, element extinction is not available for implicit solvers such as
ABAQUS/Standard finite element tool.
As an illustrative example, the MLT material degradation approach is first described using a one-
dimensional uniaxial stress-strain model as illustrated in Figure C1. The one-dimensional
51
uniaxial model is instructive for understanding the role of the internal state variable ω, the elastic
mechanical properties, and the material ultimate strength values. In this one-dimensional
uniaxial model, the stress-strain relations are defined using the MLT model as:
⎧ E t,c (ω )ε for 0 ≤ ε ≤ εmax
t,c

σ =⎨ (C1)
⎩0 for ε > εmax
t,c

where the current total strain level is denoted by ε (could be tensile or compressive), the effective
modulus E t,c (ω ) (in either tension or compression, denoted by the superscript t or c,
respectively) as a function of the internal state variable ω, and the maximum strain value εmax t,c
.
The effective modulus E (ω ) in terms of the internal state variable ω, and the initial elastic
t,c

modulus E 0t,c is given by:

E t,c (ω ) = (1− ω ) E 0t,c (C2)
where the superscripts (or subscripts) “t,c” indicate whether tension or compression is
considered. The internal state variable ω of Eq. C2 is given by:
⎛ 1 ⎞
⎜ ln(β ) ⎡ E 0 ε ⎤ ln( β ) ⎟
t,c t,c t ,c

ω = 1− exp⎜− ⎢ ⎥ ⎟⎟ (C3)
⎜ e ⎣ X t,c ⎦
⎝ ⎠
and
εult
t,c
E 0t,c εult
t,c
β t,c = E 0t,c = = (C4)
X t,c (E t,c )
ult

sec ant
(ε )t,c
ult nominal

The actual ultimate strain allowable in either tension or compression is given by εult
t,c
, and the
nominal ultimate strain allowable computed from the initial elastic modulus and the ultimate
strength allowable is given by:

(ε )
X t,c
t,c
ult nominal = (C5)
E 0t,c
These expressions hold for β t,c greater than unity. The parameter β t,c is a ratio of the initial
elastic modulus to the secant modulus at ultimate based on the specified values of the ultimate
strength allowable X t,c and the ultimate strain εult
t,c
. When the parameter β t,c has a value of unity,
the material response essentially mimics a linear elastic brittle material. As the parameter β t,c
increases above a value of unity, the degree of material nonlinearity (strain-softening behavior)
increases in the pre-ultimate regime of the stress-strain curve up to the ultimate strain value εult
t,c
.
Beyond the ultimate strain value, the stress decreases according to an exponential function (i.e.,
Weibull distribution) defining the internal state variable. The parameter β t,c is related to the
Weibull distribution parameter m t,c as:
1
1
m t,c = ⇒ β t,c = e m
t,c
(C6)
ln(β t,c )

52
Taking the basic uniaxial stress-strain functional relationship normalized by the ultimate strength
allowable and writing it in terms of this Weibull parameter m t,c gives:
⎡ t,c
⎤
⎛ ⎞m
σ ⎢ 1 ⎜ ε ⎟ ⎥ ε
= exp⎢− t,c ⎥ t,c (C7)
X t,c
⎣
⎜
( )
⎢ m e ⎝ ε ult
t,c ⎟
nominal ⎠
⎦
( )
⎥ ε ult
nominal

or written in a slightly different form gives:

⎡ t,c
⎤
⎛ ⎞m
⎢ 1 ⎜ ε ⎟ ⎥E t,cε = (1 − ω )E t,cε = E t,c (ω )ε
σ = exp⎢− t,c
⎟ ⎥ 0
(C8)
⎜ t,c
( )
0
⎢ m e ⎝ ε ult ⎠ ⎥
⎣ nominal
⎦

Matzenmiller et al. [101] proposed this basic form. From this form of the stress-strain
relationship, the nature of the Weibull distribution function is evident for the stress-strain
relationship (Weibull [113]).
Schweizerhof et al. [102] incorporated a stress limit factor α t,c that sets the minimum stress
value in the post-ultimate regime of the stress-strain curve. If the factor α t,c is non-zero, then
the MLT-predicted stress level in the post-ultimate regime of the stress-strain curve is modified
to represent a perfect-plasticity-type response commonly seen in metals. Hence, the stress is
computed using the MLT model until the MLT-predicted stress level in the post-ultimate regime
is reduced to a value below the stress limit value given by ±α t,c X t,c (positive if tensile and
negative if compressive). Below this stress limit value, the stress is held constant and equal to
±α t,c X t,c until the local computed strain exceeds the specified maximum strain value εmax
t,c
. That
is,
⎧ E t,c (ω )ε for 0 ≤ ε ≤ εmax
t,c

⎪ t,c
σ = ⎨±α X t,c for ε ≥ εult t,c
and E t,c (ω )ε < α t,c X t,c and α t,c ≠ 0 (C9)
⎪0 for ε > ε t,c
⎩ max

where the factor α t,c imposes a minimum stress value in the post-ultimate regime of the stress-
strain curve as illustrated in Figure C1. When the factor α t,c is nonzero, the internal state
variable ω is redefined as:
X t,c
ω = 1− α t,c for ε > εult
t,c
and σ < α t,c X t,c (C10)
E t,c ε

Based on these expressions, a simple uniaxial stress-strain curve can be represented that exhibits
nonlinear behavior through the internal state variable ω up to the ultimate strain level (i.e., up to
εult
t,c
). This nonlinear stress-strain response can be different in tension and compression based on
the user-specified values of E 0t,c , X t,c , and εult
t,c
. The degree of material nonlinearity (or strain
softening) in the pre-ultimate range is indicated by the parameter β t,c . As the value of β t,c
increases above unity, the degree of nonlinearity increases. This parameter can be defined in
terms of these user-specified material values (see Eq. C4) or in terms of the Weibull parameter

53
m t,c (see Eq. C6) as originally proposed by Matzenmiller et al. [101]. As such, the definition of
the initial tensile elastic modulus E0t is somewhat arbitrary since the local tangent modulus
varies as a function of strain level, and the initial elastic modulus can be used to tune the pre-
ultimate stress-strain representation empirically.
Beyond the ultimate strain value (i.e., in the post-ultimate range, ε > εult t,c
), the internal state
variable ω is also used to adjust the stress-strain curve according to an exponential decay
function. Once the ultimate stress value (in either tension or compression) X t,c is obtained and
the strain level exceeds the ultimate strain value εult t,c
, the stress level will decrease with
increasing strain values (strain-softening behavior). If the factor α t,c is nonzero, then once the
stress level decreases to a value equal to ±α t,c X t,c (positive if tensile, and negative if
compressive), the stress level is held constant at this value and the internal state variable ω is
computed accordingly until a specified maximum strain value εmax t,c
(or ERODS) is reached. For
strain values larger than this maximum value, the stress is set to zero and the internal state
variable is set to unity. The factor α t,c is therefore defined as the minimum stress limit factor
(SLIMT for tension or SLIMC for compression) for the damaged material in the post-ultimate
regime until the specified maximum strain value εmax t,c
is reached (ERODT for tension or ERODC
for compression).
Representative material behavior is illustrated for two cases of uniaxial tension loading in
Figures C2 and C3. First, an initially linear elastic material with β t , c equal to unity as shown in
Figure C2 is considered wherein the ultimate strain and ultimate strength allowable values are
related by the initial elastic modulus (i.e., X t,c = E 0t,cεult
t,c
). The predicted stress-strain response is
shown in Figure C2a, and the internal state variable ω as a function of strain is shown in Figure
C2b. The behavior for three values of the stress limit factor α t,c (i.e., SLIMT or SLIMC) is
shown in Figure C2: 1.0, 0.8, and 0. In this first case, the pre-ultimate stress-strain behavior is
essentially linear up to the ultimate strain value of 0.0025 in./in. as indicated in Figure C2a.
The internal state variable ω shown in Figure C2b remains zero up to the ultimate strain value.
For strain values larger than the ultimate strain value (i.e., in the post-ultimate regime), the
stress-strain behavior is dependent on β t,c as well as on the stress limit factor. For SLIMT equal
to unity, once the stress level reaches the ultimate strength value, the stress remains constant at
that level until the strain value reaches the user-specified maximum strain limit (i.e., ERODS),
say 0.008 in./in., as shown in this example in Figure C2a. Then the stress becomes zero.
Likewise, the internal state variable ω is initially zero and remains zero until the stress level
reaches the ultimate strength value as shown in Figure C2b. At this value, the internal state
variable increases until the maximum strain value is reached. As the SLIMT value is reduced
from a value of unity, a corresponding reduction in the stress level occurs after the ultimate
strength X t,c has been exceeded. The reduction continues until the stress level in the post-
ultimate regime of the stress-strain curve is reduced to a value given by α t,c X t,c . This minimum
stress value is held constant until the strain reaches the maximum strain (or ERODT). Setting
SLIMT to zero for the β t , c equal to unity case appears to mimic the traditional ply-discounting
material degradation model with instantaneous material degradation for a linear elastic brittle
material, see Figure C2.
54
The second case is the nonlinear strain-softening material with β t,c > 1 as shown in Figure C3
wherein the ultimate strain and ultimate strength allowable values are not related by the initial
elastic modulus but instead have values equal to 0.004 in./in., 5,000 psi and 2 Msi, respectively.
The maximum strain (or ERODT) is set to 0.008 in./in. The results are shown in Figure C3 for
three values of the minimum stress limit factor α t,c (or SLIMT): 1.0, 0.8, and 0. The predicted
stress-strain response is shown in Figure C3a, and the internal state variable ω as a function of
strain is shown in Figure C3b. The nonlinearity of the stress-strain behavior prior to reaching the
ultimate strength allowable is evident in Figure C3a, and the internal state variable ω initially
takes on nonzero values as indicated in Figure C3b.
The effect of different values of the initial modulus E 0t,c on the predicted stress-strain response
for the MLT model and on the internal state variable ω is shown in Figure C4. In these cases,
the ultimate strength Xt,c, ultimate strain εultt,c
, minimum stress limit after ultimate α t,c X t,c , and
maximum strain limit εmax t,c
have values equal to 5,000 psi, 0.004 in./in., 0.0, and 0.008 in./in.,
respectively, while the initial modulus is varies. As the initial modulus E 0t,c increases from 2
Msi to 4 Msi and then 10 Msi (i.e., β t,c = 1.6, 3.2, and 8.0 , respectively, or
m t,c = 0.47, 1.16, and 2.08 ), a shift occurs in the pre-ultimate stress-strain curve shown in Figure
C4a. The pre-ultimate stress-strain response appears to exhibit a stiffer, more nonlinear behavior
prior to reaching the ultimate values. The post-ultimate behavior is also stiffer. Changes in the
initial modulus value have a larger effect on the area under the stress-strain curve in the post-
ultimate region than in the pre-ultimate region of the stress-strain curve. Interestingly the
internal state variable changes its character with increasing values of the initial modulus as
indicated in Figure C4b. For E 0t,c equal to 2 Msi, the internal state variable increases gradually
at low strain levels. As E 0t,c (and therefore β t,c ) increases, the value of the internal state variable
increases rapidly at low strain levels. The effect of the maximum strain value (ERODT) on the
post-ultimate stress-strain curve is also evident in Figure C4.
The variation of the internal state variable ω with different values of the initial modulus E 0t,c is
examined further in Figure C5. In these cases, the ultimate strength Xt,c, ultimate strain εult t,c
,
minimum stress limit after ultimate α X t,c , and maximum strain limit εmax have values of
t,c t,c

5,000 psi, 0.004 in./in., 0.0, and 0.008 in./in., respectively, while the initial modulus E 0t,c (Eo in
Figure C5) is varies. Recall that β t,c represents the ratio of the initial modulus to the secant
modulus at ultimate (see Figure C1). As β t,c approaches unity from above, the response of the
internal state variable ω as a function of strain approaches a step function as shown in Figure C5.
As β t,c increases above unity, the step-function response tends to rotate clockwise and exhibits
smooth transitions near values of zero and unity, and the internal state variable ω begins to
change from zero sooner and increases faster (i.e., at lower strain levels). When the value of β t,c
is approximately two, the internal state variable ω tends to increase even faster for values of
strain smaller than the ultimate strain level, it reaches a value near 0.9 when the strain reaches
the ultimate strain level εult
t,c
, and a value of unity at the maximum strain limit εmax
t,c
.

55
For this one-dimensional case, the stress-strain relationship is readily described. For explicit
transient dynamics nonlinear finite element analysis tools (e.g., for ABAQUS/Explicit7 or LS-
DYNA), only this type of relationship is needed as it defines the stress state for the internal force
vector calculations. However, for implicit transient dynamics or quasi-static nonlinear finite
element analysis tools (e.g., ABAQUS/Standard), the internal force vector and the tangent
stiffness matrix are both required, and hence, the Jacobian of the stress-strain relationship must
be formulated. For the one-dimensional case, the Jacobian is a single equation given by:
∂σ
J= (C11)
∂ε
The derivation of a consistent set of stress and tangent stiffness terms is accomplished by
examining the first and second variations of the strain energy density functional U given by:
1
U (ε , ω ) = E(ω )ε 2 (C12)
2
where E(ω) is the instantaneous uniaxial material modulus as a function of the internal state
variable ω and ε is the strain. Then the expressions for a consistent set of the one-dimensional
uniaxial stress and tangent stiffness (Jacobian) terms are defined as:
∂U
σ= (C13)
∂ε
and
∂ 2U ∂σ
J= = (C14)
∂ε 2 ∂ε
Within the UMAT subroutine, the Jacobian matrix for the material model is defined by the
subroutine argument array named ddsdde (see Appendix A). Implicit solvers require both the
internal force vector and the tangent stiffness matrix and introduce significant complexity to the
material model development process applicable to two- and three-dimensional finite element
analyses.

7
ABAQUS/Explicit is a trademark of ABAQUS, Inc.

56
 Appendix D – Representative ABAQUS/Standard Input File
for the Demonstration Problems
The following records define the ABAQUS/Standard finite element model for the demonstration
problem used to illustrate this UMAT subroutine’s capabilities and features. It involves a single
4-node shell element with boundary conditions and loading resulting in a uniaxial uniform stress
condition.
*HEADING
**
*NODE
1, 0., 0.
2, 0., 4.0
3, 8., 0.
4, 8., 4.0
**
*ELEMENT, TYPE=S4R, ELSET=MODEL
1, 1, 3, 4, 2
**
** DEMOMAT1 = linear elastic brittle
** DEMOMAT2 = nonlinear elastic brittle
**
*SHELL SECTION, COMPOSITE, ELSET=MODEL
0.1, 1, DEMOMAT2, 0.0
0.1, 1, DEMOMAT2, 0.0
*TRANSVERSE SHEAR STIFFNESS
1.0E6, 1.0E6
**
** ==========================================================================================
** ==========================================================================================
** UMAT Property Data Definitions
** props(1-8):E11t,E22t,E33t,E11c,E22c,E33c,G12,G13,
** props(9-16):G23,nu12,nu13,nu23, Xt, Yt, Zt, Xc,
** props(17-24):Yc,Zc,S12,S13,S23,Eps11T,Eps22T,Eps33T,
** props(25-32):Eps11C,Eps22C,Eps33C,Gam12,Gam13,Gam23,Eps11Tmx,Eps22Tmx,
** props(33-40):Eps33Tmx,Eps11Cmx,Eps22Cmx,Eps33Cmx,Gam12mx,Gam13mx,Gam23mx,GIc,
** props(41-48):FPZ,SlimT,SlimC,SlimS,weibull(1),weibull(2),weibull(3),weibull(4),
** props(49-55):weibull(5),weibull(6),Dgrd(1),Dgrd(2),Dgrd(3),RECURS,PDA
** ==========================================================================================
** Demo problem material definition #1, total thickness = 0.20 inches
** - Linear elastic brittle
*MATERIAL, NAME=DEMOMAT1
*USER MATERIAL, CONSTANTS=55
2.000E+06, 2.000E+06, 2.000E+06, 4.500E+06, 4.500E+06, 4.500E+06, 0.800E+06, 0.800E+06,
0.800E+06, 2.500E-01, 2.500E-01, 2.500E-01, 7.000E+03, 7.000E+03, 7.000E+03, 1.800E+04,
1.800E+04, 1.800E+04, 6.200E+03, 6.200E+03, 6.200E+03, 3.500E-03, 3.500E-03, 3.500E-03,
4.000E-03, 4.000E-03, 4.000E-03, 7.750E-03, 7.750E-03, 7.750E-03, 0.200E-01, 0.200E-01,
0.200E-01, 0.200E-01, 0.200E-01, 1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 3.000E+01,
2.000E-01, 0.000E+00, 0.000E+00, 0.000E+00, 0.000E+00, 0.000E+00, 0.000E+00, 0.000E+00,
0.000E+00, 0.000E+00, 1.000E-06, 1.000E-06, 1.000E-06, 1.000E+00, 1.000E+00
*DEPVAR
8
** ==========================================================================================
** Demo problem material definition #2, total thickness = 0.20 inches
** - Nonlinear elastic brittle
*MATERIAL, NAME=DEMOMAT2
*USER MATERIAL, CONSTANTS=55
2.000E+06, 2.000E+06, 2.000E+06, 4.500E+06, 4.500E+06, 4.500E+06, 0.800E+06, 0.800E+06,
0.800E+06, 2.500E-01, 2.500E-01, 2.500E-01, 7.000E+03, 7.000E+03, 7.000E+03, 1.800E+04,
1.800E+04, 1.800E+04, 6.200E+03, 6.200E+03, 6.200E+03, 6.000E-03, 6.000E-03, 6.000E-03,
5.800E-03, 5.800E-03, 5.800E-03, 7.750E-03, 7.750E-03, 7.750E-03, 0.200E-01, 0.200E-01,
0.200E-01, 0.200E-01, 0.200E-01, 1.000E-01, 1.000E-01, 1.000E-01, 1.000E-01, 3.000E+01,
2.000E-01, 0.000E+00, 0.000E+00, 0.000E+00, 0.000E+00, 0.000E+00, 0.000E+00, 0.000E+00,
0.000E+00, 0.000E+00, 1.000E-06, 1.000E-06, 1.000E-06, 1.000E+00, 2.000E+00

57
*DEPVAR
8
** ==========================================================================================
**
** Solution step 1
**
*STEP,NLGEOM, INC=10000
*STATIC
0.01, 1.0, 1.0e-04, 0.01
*CONTROLS, PARAMETERS= TIME INCREMENTATION
10,10,10,10,10,, , 10, 10
**
** Boundary conditions
**
*NSET, NSET=UVWDIS
1,
*NSET, NSET=UWDIS
2,
*NSET, NSET=DIS, GENERATE
3, 4, 1
**
** uvwdis
**
*BOUNDARY, OP=MOD
UVWDIS, 1,, 0.
UVWDIS, 2,, 0.
UVWDIS, 3,, 0.
**
** uwdis
**
*BOUNDARY, OP=MOD
UWDIS, 1,, 0.
UWDIS, 3,, 0.
**
** dis - imposed displacement
**
*BOUNDARY, OP=MOD
DIS, 1,, 0.080
**
*OUTPUT, FIELD, FREQ=1
*NODE OUTPUT
U,
*ELEMENT OUTPUT
S,
E,
SDV,
*OUTPUT, HISTORY, FREQ=1
*NODE OUTPUT, NSET=DIS
U,
RF,
*ELEMENT OUTPUT, ELSET=MODEL
E,
S,
SDV,
*NODE FILE, FREQ=1
U,
**
*EL FILE, FREQ=1
E,
S,
*END STEP

58
Table 1. Material degradation factors for three-dimensional models and various failure modes
based on the maximum stress and maximum strain criteria.
Failure Failure Degradation factors for diagonal terms of constitutive matrix
(corresponding row and column off-diagonal entries are also degraded and
index i mode symmetry of the constitutive matrix is enforced)

C11 C22 C33 C44 C55 C66

1 (T) Longitudinal in-plane βT(C11)o

tension

1 (C) Longitudinal in-plane βC(C11)o

compression

2 (T) Transverse in-plane βT(C22)o

tension

2 (C) Transverse in-plane βC(C22)o

compression

3 (T) Transverse out-of-plane βT(C33)o

tension

3 (C) Transverse out-of-plane βC(C33)o

compression

4 In-plane shear βS(C44)o

(12-plane)

5 Transverse shear βS(C55)o

(13-plane)

6 Transverse shear βS(C66)o

(23-plane)

59
Table 2. Material degradation factors for three-dimensional models and various failure modes
based on the Tsai-Wu criteria.
Failure Failure Degradation factors for diagonal terms of constitutive matrix
(corresponding row and column off-diagonal entries are also degraded and
index i mode symmetry of the constitutive matrix is enforced)

C11 C22 C33 C44 C55 C66

1 (T) Longitudinal in-plane βT(C11)o

tension

1 (C) Longitudinal in-plane βC(C11)o

compression

2 (T) Transverse in-plane βT(C22)o

tension

2 (C) Transverse in-plane βC(C22)o

compression

3 (T) Transverse out-of-plane βT(C33)o

tension

3 (C) Transverse out-of-plane βC(C33)o

compression

4 In-plane shear βS(C44)o

(12-plane)

5 Transverse shear βS(C55)o

(13-plane)

6 Transverse shear βS(C66)o

(23-plane)

60
Table 3. Material degradation factors for three-dimensional models and various failure modes
based on the Hashin criteria.
Failure Failure Degradation factors for diagonal terms of constitutive matrix
(corresponding row and column off-diagonal entries are also degraded and
index i mode symmetry of the constitutive matrix is enforced)

C11 C22 C33 C44 C55 C66

1 (T) Longitudinal in-plane βT(C11)o βS(C44)o

tension

1 (C) Longitudinal in-plane βC(C11)o

compression

2 (T) Transverse in-plane βT(C22)o βS(C44)o

tension

2 (C) Transverse in-plane βC(C22)o βS(C44)o

compression

3 (T) Transverse out-of-plane βT(C33)o

tension

3 (C) Transverse out-of-plane βC(C33)o

compression

4 In-plane shear βS(C44)o

(12-plane)

5 Transverse shear βS(C55)o

(13-plane)

6 Transverse shear βS(C66)o

(23-plane)

61
 Table 4. User-defined property data for the UMAT subroutine.
PROPS array Variable Description
entry, i name
1,2,3 Et(i) Initial elastic tension moduli: E11t, E22t, E33t
4,5,6 Ec(i) Initial elastic compression moduli: E11c, E22c, E33c
7,8,9 G0(i) Initial elastic shear moduli: G12, G13, G23
10,11,12 Anu(i) Poisson’s ratios: ν12, ν13, ν23
13,14,15 Xt(i) Ultimate tension stress allowable values in the 1-, 2-, 3-directions
16,17,18 Xc(i) Ultimate compression stress allowable values in the 1-, 2-, 3-directions
19,20,21 S(i) Ultimate shear (12-, 13-, 23-planes) stress allowable values
22,23,24 EpsT(i) Ultimate normal tension strain allowable values in the 1-, 2-, 3-directions
25,26,27 EpsC(i) Ultimate normal compression strain allowable values in the 1-, 2-, 3-
directions
28,29,30 GamS(i) Ultimate shear (12-, 13-, 23-planes) strain allowable values
31,32,33 EpsTx(i) Maximum normal tension strain allowable value in the 1-, 2-, 3-directions;
currently not used
34,35,36 EpsCx(i) Maximum normal compression strain allowable values in the 1-, 2-, 3-
directions; currently not used
37,38,39 GamSx(i) Maximum shear (12-, 13-, 23-planes) strain allowable values; currently not
used
40 GIc Critical strain energy release rate for Mode I fracture
41 FPZ Width of the fracture process zone
42 SlimT Stress limit factor for tension behavior; currently not used
43 SlimC Stress limit factor for compression behavior; currently not used
44 SlimS Stress limit factor for in-plane shear behavior; currently not used
45, 46, 47 Weibl(i) Weibull parameter of MLT model for normal stress components (i=1, 2, 3);
currently not used
48, 49, 50 Weibl(j) Weibull parameter of MLT model for shear stress components (j=4, 5, 6);
currently not used
51 Dgrd(1) Material degradation factor for tension failures, non-zero values result in
degradation after failure initiation; active only when PDA=1 to 4
52 Dgrd(2) Material degradation factor for compression failures, non-zero values result
in degradation after failure initiation; active only when PDA=1 to 4
53 Dgrd(3) Material degradation factor for shear failures, non-zero values result in
degradation after failure initiation; active only when PDA=1 to 4
54 RECURS Flag for selecting the type of material degradation: 0=instantaneous,
1=recursive when PDA=1 to 4
55 PDA Progressive failure analysis option (0=linear, elastic, bimodulus response,
1=use maximum stress criteria, 2=use maximum strain criteria, 3=use Tsai-
Wu polynomial, 4=use Hashin criteria)

62
 Table 5. UMAT-defined solution-dependent variables.
STATEV Solution-
array Dependent Description of Solution-Dependent Variables
entry i Variable
Name
Two-dimensional shell elements, NDV=8
1 dmg(1) Degradation factor for the σ11 stress component
2 dmg(2) Degradation factor for the σ22 stress component
3 dmg(3) Degradation factor for the σ12 stress component
4 fflags(1) Failure flag for first failure mode
5 fflags(2) Failure flag for second failure mode
6 fflags(3) Failure flag for third failure mode
7 SEDtot Total strain-energy density
8 Damage Damage estimate based on energy lost (total minus
recoverable)/(fracture toughness)
Three-dimensional solid elements. NDV=14
1 dmg(1) Degradation factor for the σ11 stress component
2 dmg(2) Degradation factor for the σ22 stress component
3 dmg(3) Degradation factor for the σ33 stress component
4 dmg(4) Degradation factor for the σ12 stress component
5 dmg(5) Degradation factor for the σ13 stress component
6 dmg(6) Degradation factor for the σ23 stress component
7 fflags(1) Failure flag for first failure mode
8 fflags(2) Failure flag for second failure mode
9 fflags(3) Failure flag for third failure mode
10 fflags(4) Failure flag for fourth failure mode
11 fflags(5) Failure flag for fifth failure mode
12 fflags(6) Failure flag for sixth failure mode
13 SEDtot Total strain-energy density
14 Damage Damage estimate based on energy lost (total minus
recoverable)/(fracture toughness)

Note: The degradation solution-dependent variables (SDVs) should be zero until failure
initiation is detected. Once failure initiation has been detected, the degradation SDVs will be
non-zero and approach a value of unity (i.e., complete degradation at that material point). The
failure flag SDVs are the solution increment number when failure initiation at that material point
and for that stress component is detected. Contour plots of the failure flag SDVs can be used to
give an indication of the evolution of the damage progression.

63
 Table 6. Summary of the peak failure loads for the cross-ply and quasi-isotropic 16-layer
graphite-epoxy open-hole-tension coupons using different failure criteria, a maximum solution
increment size of 0.01, and a recursive degradation with a degradation factor β of 0.5.

Peak Failure Load, pounds

Progressive
Cross-ply Cross-ply Quasi-
failure Analysis
(laminate) (smeared) isotropic
(PDA)
(laminate)

Max. Stress (1) 9,531 10,041 7,975

Max. Strain (2) 9,620 9,996 7,941

Tsai-Wu (3) 8,655 9,466 7,869

Hashin (4) 7,129 6,056 5,985

Test Average 9,605 9,605 7,810

[120]

64
 Stress

σTall

ET

Strain

EC
σCall

Figure 1. Typical stress-strain curve for a linear elastic-brittle bimodulus material.

Figure 2. Representative uniaxial stress-strain curve for strain-softening material –

nomenclature.

65
 Figure 3. Geometry and boundary conditions for the demonstration problem.

Figure 4. Stress-strain curves for the two materials studied in the demonstration problems.

66
Figure 5. Stress-strain behavior for different progressive failure analysis criteria implemented
within the UMAT subroutine obtained using instantaneous degradation, a degradation factor of
10-6, and a maximum solution increment size factor of 0.01.

67
 (a) Axial reaction force as a function of applied strain.

(b) Material degradation factor as a function of applied strain.

Figure 6. Selected response parameters for different progressive failure analysis criteria
implemented within the UMAT subroutine obtained using instantaneous degradation, a
degradation factor of 10-6, and a maximum solution increment size factor of 0.01.

68
 (c) Strain energy density as a function of applied strain.

(d) Energy-based damage factor as a function of applied strain.

Figure 6. Concluded.

69
 Figure 7. Stress-strain behavior for the linear elastic brittle material obtained using selected
maximum solution increment size factors, instantaneous degradation, a degradation factor of 10-
6
, and the maximum stress criteria.

70
 (a) Effect of different degradation factors combined with instantaneous degradation.

(b) Effect of different degradation factors combined with recursive degradation.

Figure 8. Stress-strain behavior for the linear elastic brittle material obtained using a maximum
solution increment size factor of 0.01and the maximum stress criteria.

71
 (a) Effect of different degradation factors combined with instantaneous degradation.

(b) Effect of different degradation factors combined with recursive degradation.

Figure 9. Material degradation factor as a function of strain level for the linear elastic brittle
material obtained using a maximum solution increment size factor of 0.01and the maximum
stress criteria.

72
 (a) Effect of different degradation factors combined with instantaneous degradation.

(b) Effect of different degradation factors combined with recursive degradation.

Figure 10. Strain energy density predictions for the linear elastic brittle material obtained using a
maximum solution increment size factor of 0.01and the maximum stress criteria.

73
 Figure 11. Stress-strain behavior predicted using the maximum stress and maximum strain
criteria with instantaneous degradation, a degradation factor of 10-6, and a maximum solution
increment size factor of 0.01 and applied to the nonlinear elastic brittle material.

Figure 12. Stress-strain behavior predicted using the maximum strain criteria with recursive
degradation, a degradation factor of 10-6, and a maximum solution increment size factor of 0.01
and applied to the nonlinear elastic brittle material.

74
(a) Complete model. (b) Close-up view of region near the hole.
Figure 13. Finite element model of the open-hole-tension coupon.

75
 (a) Predictions obtained using a 0.02 value for the maximum solution increment size.

(b) Predictions obtained using a 0.01 value for the maximum solution increment size.
Figure 14. Comparison of different progressive failure analysis models for the 16-ply cross-ply
open-hole-tension coupon using recursive degradation with a degradation factor β of 0.5.

76
 (a) Results using maximum strain criteria (PDA=2).

(b) Results using Hashin criteria (PDA=4).

Figure 15. Influence of maximum solution increment size on the progressive failure predictions
for the 16-ply cross-ply open-hole-tension coupon using recursive degradation with a
degradation factor β of 0.5.

77
 (a) Results using maximum strain criteria (PDA=2).

(b) Results using Hashin criteria (PDA=4).

Figure 16. Influence of degradation factor β on the progressive failure predictions for the 16-ply
cross-ply open-hole-tension coupon using recursive degradation and a maximum solution
increment size of 0.02.

78
Figure 17. Influence of maximum solution increment size on the progressive failure predictions
for the 16-ply cross-ply open-hole-tension coupon using the maximum strain criteria (PDA=2)
and instantaneous degradation with a degradation factor β of 10-6.

Figure 18. Comparison of different progressive failure analysis models for the 16-ply quasi-
isotropic open-hole-tension coupon using recursive degradation with a degradation factor β of
0.5 and a maximum solution increment size of 0.01.
79
Figure C1. Representative uniaxial stress-strain curve for MLT material model – nomenclature.

80
 (a) Predicted stress-strain response.

(b) State variable ω as a function of strain.

Figure C2. Predicted uniaxial response for different values of SLIM when initial and secant
moduli are equal (E0=2 Msi, XT=5,000 psi, and εult=0.0025 in./in., EROD=0.008).

81
 (a) Predicted stress-strain response.

(b) State variable ω as a function of strain.

Figure C3. Predicted uniaxial response different values of SLIM when initial modulus is greater
than secant modulus (E0=2 Msi, XT=5,000 psi, and εult=0.004 in./in., EROD=0.008).

82
 (a) Predicted stress-strain response.

(b) State variable ω as a function of strain.

Figure C4. Predicted uniaxial response for different values of the initial modulus E0
(XT=5,000 psi, and εult=0.004 in./in., SLIM=0.0, EROD=0.008).

83
Figure C5. Effect of the ratio of initial modulus to secant modulus β on the internal state
variable ω as a function of strain (XT=5,000 psi, and εult=0.004 in./in., SLIM=0.0,
EROD=0.008).

84
 Form Approved
REPORT DOCUMENTATION PAGE OMB No. 0704-0188
The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources,
gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this
collection of information, including suggestions for reducing this burden, to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and
Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person
shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number.
PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (From - To)
01- 12 - 2006 Contractor Report
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
User-Defined Material Model for Progressive Failure Analysis
5b. GRANT NUMBER

5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) 5d. PROJECT NUMBER

Knight, Norman F., Jr. NNL06AA48T
5e. TASK NUMBER

Task 10
5f. WORK UNIT NUMBER

732759.07.09
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
REPORT NUMBER
NASA Langley Research Center General Dynamics
Hampton, VA 23681-2199 Advanced Information Systems
14700 Lee Road
Chantilly, VA 20151
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR'S ACRONYM(S)
National Aeronautics and Space Administration NASA
Washington, DC 20546-0001
11. SPONSOR/MONITOR'S REPORT
NUMBER(S)
NASA/CR-2006-214526
12. DISTRIBUTION/AVAILABILITY STATEMENT
Unclassified - Unlimited
Subject Category 39
Availability: NASA CASI (301) 621-0390
13. SUPPLEMENTARY NOTES
Prepared for Langley Research Center under GSA Task Order GS-00F-0067M, NASA Order No. NNL06AA48T.
Langley Technical Monitor: James R. Reeder
An electronic version can be found at http://ntrs.nasa.gov
14. ABSTRACT
An overview of different types of composite material system architectures and a brief review of progressive failure material
modeling methods used for structural analysis including failure initiation and material degradation are presented. Different
failure initiation criteria and material degradation models are described that define progressive failure formulations. These
progressive failure formulations are implemented in a user-defined material model (or UMAT) for use with the
ABAQUS/Standard nonlinear finite element analysis tool. Implementation details and use of the UMAT subroutine are
described in the present paper. Parametric studies for composite structures are discussed to illustrate the features of the
progressive failure modeling methods that have been implemented.

15. SUBJECT TERMS

Progressive failure analysis; Nonlinear analysis; Composite structures

16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF RESPONSIBLE PERSON
ABSTRACT OF
a. REPORT b. ABSTRACT c. THIS PAGE PAGES STI Help Desk (email: [email protected])
19b. TELEPHONE NUMBER (Include area code)
U U U UU 93 (301) 621-0390
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std. Z39.18