Journal of Composite Materials

http://jcm.sagepub.com/

COMM Toolbox: A MATLAB toolbox for micromechanical analysis of composite materials

CT McCarthy and TJ Vaughan
Journal of Composite Materials 2012 46: 1715 originally published online 21 October 2011
DOI: 10.1177/0021998311423984
The online version of this article can be found at:
http://jcm.sagepub.com/content/46/14/1715

Published by:
http://www.sagepublications.com

On behalf of:

American Society for Composites

Additional services and information for Journal of Composite Materials can be found at:
Email Alerts: http://jcm.sagepub.com/cgi/alerts
Subscriptions: http://jcm.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://jcm.sagepub.com/content/46/14/1715.refs.html

>> Version of Record - Jul 8, 2012

OnlineFirst Version of Record - Oct 21, 2011
What is This?

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

JOURNAL OF
COMPOSITE
M AT E R I A L S

Article

COMM Toolbox: A MATLAB toolbox for

micromechanical analysis of composite
materials

Journal of Composite Materials

46(14) 17151729
! The Author(s) 2011
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0021998311423984
jcm.sagepub.com

CT McCarthy and TJ Vaughan

Abstract
The COmposite MicroMechanics (COMM) Toolbox is a design tool, developed in MATLAB, which provides efficient
pre- and post-processing capabilities for micromechanical analyses of composite materials with finite element analysis.
The COMM Toolbox has automated all manual tasks associated with micromechanical analyses of composite materials,
providing a simple, convenient environment for creating, submitting, monitoring, and evaluating results from micromechanical analyses. The interactive pre-processing capability currently enables a variety of fiber distributions to be generated, allowing either nonuniform or regular fiber arrangements for both low and high fiber volume fractions to be analyzed
under mechanical and/or thermal loading. The functionality of the above features has been demonstrated by carrying out
a case study examining the effect of fiber volume fraction on material behavior. Importantly, the COMM Toolbox means
that advanced multiscale modeling concepts, which determine material behavior based on the physical interactions of the
constituent phases, are likely to gain more widespread exposure from potential academic or industry-based interests.
The COMM Toolbox is freely available to the community and can be obtained by contacting the corresponding author.

Keywords
polymer matrix composites, micromechanics, fracture, finite element analysis, MATLAB design tool

Introduction
High-strength ber-reinforced composites have not
been as widely used within industry as their structural
performance would suggest. A great deal of this can be
attributed to the intense testing and certication procedures applied to composite structures, coupled with the
high cost and complexity associated with their processing techniques. The emergence of advanced multiscale
predictive capabilities potentially provides material and
structural designers with enhanced knowledge and
understanding of how composite material systems
behave, which could help to reduce time between
research phases and the marketplace. However, the
inherent complexity associated with such modeling
strategies forms a barrier with respect to their advancement from academic interests to practical industrybased implementation.
Already, it has been shown that the eld of computational micromechanics can provide detailed predictions of microscale damage and deformation
processes in ber-reinforced composite materials.13
These strategies predict material behavior by representing the ber and matrix phases discretely in the form of

a representative volume element (RVE) and applying

macroscopically uniform loading. While access to
high-performance parallel computing resources allows
multiber statistically equivalent RVEs to be solved
relatively quickly, the pre- and post-processing stages
of micromechanical analyses can remain a complicated
time-consuming task. The pre-processing stage can
require strict mesh generation criteria and the manipulation of large amounts of data to dene suitable nodesets
for the correct implementation of periodic boundary
conditions. Meanwhile, the post-processing stage can
involve the analysis of local stress and strain elds to
determine homogenized properties of the material. For
the experienced analyst, pre- and post-processing times
can obviously be reduced, however, to advance the use
Materials and Surface Science Institute (MSSI) and the Department of
Mechanical, Aeronautical and Biomedical Engineering, University of
Limerick, Limerick, Ireland
Corresponding author:
CT McCarthy, Materials and Surface Science Institute (MSSI) and the
Department of Mechanical, Aeronautical and Biomedical Engineering,
University of Limerick, Limerick, Ireland
Email: conor.mccarthy@ul.ie

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

1716

Journal of Composite Materials 46(14)

of multiscale concepts toward practical industry-based

implementation, analysis techniques need to be simplied. As a result, the COmposite MicroMechanics
(COMM) Toolbox was developed in MATLAB and its
function is to provide ecient pre- and post-processing
capabilities for micromechanical analyses of composite
materials. The COMM Toolbox automates all manual
tasks associated with micromechanical analyses of composite materials, providing a simple, robust, convenient
environment for creating, submitting, monitoring, and
evaluating results from micromechanical analyses.
The COMM Toolbox allows for the prediction of local
intra-ply deformation processes in ber-reinforced composites, such as ber-matrix debonding and matrix
yielding, under both thermal and/or mechanical loading
(Note: much of the details regarding the computational
framework used to predict such local microscale damage
process may be found in the authors previous
contribution3).
This paper outlines the development and main
operation of the COMM Toolbox, together with a
description of commands which were used to interface
between MATLAB and ABAQUS. A case study, which
examines the eect of ber volume fraction on the
transverse failure surface, has been carried out to highlight the functionality of the newly developed COMM
Toolbox. The ndings provide novel insight into the
behavior of composite materials under loading in the
transverse plane. The COMM Toolbox is freely available to the community and can be obtained by contacting the corresponding author.

Development of the COMM Toolbox

The COMM Toolbox was developed in MATLAB
and its function is to provide ecient pre- and post-processing capabilities for micromechanical analyses of
composite materials. The COMM Toolbox uses
ABAQUS4 as its nite element solver, since advanced
nonlinear and fracture modeling techniques are readily
available. The COMM Toolbox operates through a
series of graphical user interfaces (GUIs) that carry
out all tasks involved in micromechanical analyses.
The core tasks include microstructure generation,
preparation of input les (either linear or nonlinear),
job submission, and evaluating results (homogenization). In addition, the COMM Toolbox provides the
capability for the statistical analysis of ber distributions, submission of jobs to parallel computing
resources, and active job monitoring. This section
describes the operation of the main GUIs associated
with the COMM Toolbox.
The COMM Toolbox can be initialized through the
main MATLAB window by selecting Start !
Toolboxes ! Micromechanics Analysis ! COMM
Toolbox.

Main COMM Toolbox GUI

The main GUI of the COMM Toolbox is shown in
Figure 1. Its primary function is to generate ber distributions, through the use of a number of algorithms
that have been widely used in micromechanical

Figure 1. Main graphical user interface (GUI) of the COMM Toolbox.

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

McCarthy and Vaughan

1717

Figure 2. Fiber distributions available in the COmposite MicroMechanics (COMM) Toolbox: (a) Nearest Neighbor Algorithm
(b) Hard-Core Model (c) Hexagonal Periodic Array, and (d) Square Periodic array.

investigations.5,6 It also provides access to further analysis options available in the COMM Toolbox, such as
pre- and post-processing of nite element models. The
components of the main COMM Toolbox GUI have
been identied in Figure 1 and these are described as
follows:

(a). Distribution options. This drop-down menu

allows the user to select the type of ber distribution
to be generated. Currently, the ber distributions available in the COMM Toolbox are the Nearest Neighbor
Algorithm (NNA),6 the Hard-Core Model (HCM)7 and
both Hexagonal and Square Periodic arrays. The
NNA, shown in Figure 2(a), allows high ber volume
fraction, statistically equivalent RVEs to be generated.
The widely used HCM, shown in Figure 2(b), allows
nonuniform ber arrangements of up to 54% ber
volume fraction to be generated. Meanwhile, both
Hexagonal and Square Periodic arrays, shown in
Figure 2(c) and (d), respectively, provide simplied
ber arrangements which can be computationally inexpensive, compared to either the NNA or HCM.

Figure 3. Nearest Neighbor Algorithm (NNA) properties

graphical user interface (GUI).

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

1718

Journal of Composite Materials 46(14)

Figure 4. Graphical user interface (GUI) for statistical analysis of fiber distribution.

(b) RVE Controls. The RVE Controls allow the user to

dene the parameters of the RVE being generated. In
the case of the NNA, the user must specify the RVE
size and, if required, can select the NNA Properties
push button to initialize the NNA Properties GUI
which provides further options regarding the ber distribution generated by the NNA. The NNA uses experimentally measured nearest neighbor distribution
functions to dene the interber distances and an
experimentally measured diameter distribution function
to assign ber diameters in the RVE.6 The NNA
Properties GUI, shown in Figure 3, allows the user to
modify the parameters describing the diameter distribution function and the 1st and 2nd nearest neighbor
distribution functions, meaning the NNA can adapt to
create ber distributions that have lower or higher ber
volume fractions. Upon execution of the Generate
push button located on the Main COMM Toolbox
GUI, shown in Figure 1, these parameters are used to
generate an RVE which is subsequently plotted in the
RVE Display window, also shown in Figure 1.
In the case of the HCM, the user must specify the
RVE size and the desired ber volume fraction of the
RVE. The user may modify the ber diameter by selecting Edit ! HCM properties from the main menu bar
(shown in Figure 1). Like before, upon execution of the
Generate push button, these parameters are used to

generate an RVE microstructure and this is plotted in

the RVE display window.
Meanwhile, for either the Hexagonal or Square
Periodic arrays, the user must also specify the ber
volume fraction and upon execution of the Generate
push button, the user is prompted to enter the length of
arrays in both x- and y-directions. The resulting ber
array is plotted in the RVE Display window, as before.

(c) RVE Display. As explained above, the RVE generated by the COMM Toolbox is displayed in this
window. By selecting File ! Save from the main
menu bar, the gure in the RVE display window may
be saved as either an Enhanced Meta File (.emf), a
Bitmap (.bmp), or a MATLAB Figure le (.g).
(d) Statistical analysis. The Statistical Analysis push
button initializes the Statistical Analysis GUI, as shown
in Figure 4. The function of the Statistical Analysis
GUI is to comprehensively characterize the ber
arrangement in the RVE using a number of statistical
functions. The COMM Toolbox determines the 1st and
2nd nearest neighbor distribution functions, the radial
distribution function, and the second-order intensity
function of an RVE and displays these functions in
the Statistical Analysis GUI, as shown in Figure 4.
These functions characterise both the short and long

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

McCarthy and Vaughan

1719

Figure 5. (a) Laminate properties graphical user interface (GUI) and (b) material properties GUI.

range interaction of bers in the RVE and further

details regarding their formulation may be found in
Ref. 6. Using the Export push button, shown in
Figure 4, the COMM Toolbox exports this data to a
Microsoft Excel worksheet, providing an ecient
method for the statistical analysis of ber distributions.

(e) Laminate properties. The Laminate Properties

push button initializes the Laminate Properties GUI,
shown in Figure 5(a). The function of the Laminate
Properties GUI is to determine eective properties of
the RVE stored in the current workspace. In the
Analysis Options panel, the user can designate the components of the laminate properties to be determined by
selecting the check boxes shown in Figure 5(a). To
determine E11, E22, E33, and G23, the COMM
Toolbox carries out a two-dimensional generalized
plane strain analysis, while to determine G12 and G13
a three-dimensional model is used. By default, the constituent material properties used by the COMM
Toolbox are set to those of HTA carbon ber and
6376 epoxy matrix, which were taken from Ref. 3.
However, the user may modify the constituent material
properties by clicking the Material Properties push
button located on the Laminate Properties GUI,
shown in Figure 5(a). The Material Properties GUI,
shown in Figure 5(b), allows the ber properties to be
dened as either isotropic, in terms of Youngs modulus
and Poissons ratio, or orthotropic, in terms of the
engineering constants. Meanwhile, the matrix properties are assumed to be isotropic, again dened in terms
of Youngs modulus and Poissons ratio. The Material
Properties GUI (Figure 5(b)) also contains a number of
further features that are used for dening thermal and
nonlinear material behavior, which will be discussed
later in section Pre-processing ABAQUS Input Files
for nonlinear analysis.
Upon the execution of the Submit push button,
shown in Figure 5(a), the COMM Toolbox carries

out a series of tasks to determine eective properties

of the RVE. The process requires a suitable mesh to
be generated, the application of periodic boundary conditions, submission of input les to the ABAQUS
solver, and post-processing of data to evaluate results.
These tasks are carried out by the COMM Toolbox as
automated background processes that are shown schematically in Figure 6, while their implementation is
described in detail below.
For the RVE shown in Figure 6(a), the COMM
Toolbox generates a Python script (job-name.py) containing a series of scripting interface commands that
create the nite element model. The Python script is
executed from the MATLAB environment using the
command, dos(abaqus cae noGUI job-name.py),
which initializes the ABAQUS Python interpreter.
This is carried out as a background process and the
Python scripting interface commands create the RVE
geometry, assign ber/matrix section properties, generate a mesh, and write the details to an ABAQUS Input
File (job-name.inp), as shown in Figure 6(b) and (c).
Periodic boundary conditions, similar to those
described by Van der Sluis et al.,8 are applied to the
RVE to ensure a macroscopically uniform stress/displacement eld exists across the boundaries of each RVE.
These consist of a series of tie constraints that require
corresponding nodes on each opposing face of the RVE
to undergo identical displacements. The periodic
boundary conditions can be expressed in terms of the
nodal displacement vector, u, such that,
uNorth  un4 uSouth

uEast  un2 uWest

where the subscripts North, South, East, and West

correspond to nodes situated on each edge of the RVE
and subscripts n2 and n4 correspond to the control
nodes, which are located at each of corner of the

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

1720

Journal of Composite Materials 46(14)

Figure 6. Tasks carried out by the COmposite MicroMechanics (COMM) Toolbox to determine effective properties for an
representative volume element (RVE).

RVE, as shown in Figure 6(d). From the ABAQUS

Input File (job-name.inp), the nodal co-ordinates
(stored under the *Node heading) are imported to the
MATLAB workspace and the COMM Toolbox creates
a series of nodesets and constraint equations to prescribe periodic boundary conditions to the RVE, as
shown by Figure 6(d). The COMM Toolbox writes
these details to the ABAQUS Input File (job-name.inp)
and executes it from the MATLAB environment using
the command dos(abaqus job job-name.inp), which
submits the job to the ABAQUS solver. Upon job completion, a further Python script is used to extract results
from the ABAQUS output database le (job-name.odb),
shown in Figure 6(e), and the COMM Toolbox calculates the eective properties, as shown in Figure 6(f).
All tasks shown in Figure 6 are carried out by the
COMM Toolbox as background processes and require
no further input from the user. Once eective properties
have been determined, they are displayed in the listbox
on the Laminate Properties GUI, shown in Figure 5(a).

(f) ABAQUS analysis. The ABAQUS Analysis push

button located on the Main COMM Toolbox GUI,
shown in Figure 1, initializes the ABAQUS Analysis
GUI which allows the user create ABAQUS Input
Files for nonlinear micromechanical analysis. This feature is described in more detail in section Pre-processing
ABAQUS Input Files for nonlinear analysis.

Pre-processing ABAQUS Input Files for nonlinear

analysis
The COMM Toolbox can create nonlinear ABAQUS
Input Files using the ABAQUS Analysis GUI, shown
in Figure 7. The ABAQUS Analysis GUI may be initialized using the ABAQUS Analysis push button
located on the Main COMM Toolbox GUI, shown in
Figure 1. Its function is to allow ecient pre-processing
of ABAQUS Input Files for two-dimensional generalized plane strain analysis and provide a convenient job
submission system for the ABAQUS solver. The
ABAQUS Analysis GUI enables the prediction of
ber-matrix debonding using a cohesive zone model
at ber-matrix interfaces, while matrix yielding may
currently be predicted using Mohr-Coulomb plasticity
theory. The ABAQUS Analysis GUI allows the user
dene various parameters of a micromechanical
model, for example material properties, mesh density,
or boundary conditions, using the ABAQUS Toolbar,
as shown in Figure 7. The main components of the
ABAQUS Analysis GUI are identied in Figure 7
and these are described as follows:

(a) Material properties. The Material Properties

push button initializes the Material Properties GUI
which enables the user to dene the thermal and
mechanical properties of the constituent materials, as
shown in Figure 5(b). The ber properties may be

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

McCarthy and Vaughan

1721

Figure 7. ABAQUS analysis graphical user interface (GUI) for pre-processing files for nonlinear analysis.

dened as either isotropic, in terms of Youngs modulus

and Poissons ratio, or orthotropic, in terms of the
engineering constants. If a thermal analysis is being
carried out, values for either isotropic or orthotropic
thermal expansion coecients may be dened.
Meanwhile, the matrix properties are assumed to be
isotropic and nonlinear behavior may be specied by
selecting the Mohr-Coulomb plasticity option, which
has been widely used to describe behavior of epoxy
matrices.13 The Mohr-Coulomb criterion states that
yielding will occur on a given plane when the shear
stress () exceeds the cohesive stress of the material
plus the frictional force acting along the failure plane,
such that
 c  n tan

where c is the cohesion yield stress, rn is the normal

stress acting on the failure plane, and is the angle of
internal friction. The Mohr-Coulomb option requires
the user to enter the cohesion stress (c) and friction
angle () of the material, as shown in Figure 5(b).
For all the parameters above, default values are initially
displayed which describe the behavior of HTA carbon
bers and 6376 epoxy matrix.3

(b) Mesh properties. The Mesh Properties push

button initializes the Mesh Properties GUI, shown in
Figure 8(a). The Mesh Properties GUI allows the user
to dene the global seed size used in mesh generation.

It also allows a cohesive section to be included at the

ber-matrix interface to predict the onset and evolution
of ber-matrix debonding. This is ordinarily a complicated, time-consuming task as it requires an individual
section, with correct material orientations, to be
assigned around each ber. The behavior of the cohesive section is governed by a standard bi-linear traction
separation law, which relates the separation displacement between the top and bottom faces of the cohesive
element to the traction vector acting upon it.
Figure 8(b) shows the initial response of the traction
separation model is assumed to be linear elastic. Failure
of the interface initiates when either the normal stress
(tn ) or the shear stress (ts ) exceed the predened normal
(ton ) or shear (tos ) strengths. The evolution of damage at
the interface is controlled by a linear softening curve
meaning that once the failure stress is exceeded, the
element stiness reduces linearly until complete failure
at a specied normal (fn ) or shear (fs ) displacement, as
shown in Figure 8(b). This eective displacement at
failure (f ) determines the rate of damage in the element
and this was dened in terms of the fracture energy, ,
which corresponds to the area under the traction
separation curve. The user denes the behavior of the
cohesive section in terms of the ber-matrix interface
strength, fracture energy, and initial elastic stiness for
both normal and shear directions (i.e. Mode I and
Mode II type fracture), as shown in Figure 8(a). The
user may also modify the viscosity coecient, which is
used to aid the convergence of cohesive sections.

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

1722

Journal of Composite Materials 46(14)

Figure 8. (a) Mesh Properties graphical user interface (GUI) and (b) traction-separation law-governing behavior of cohesive
elements.

Figure 9. (a) Loads/boundary conditions graphical user interface (GUI) (b) boundary conditions applied to an representative volume
element (RVE).

(c) Loads/boundary conditions. The Loads/Boundary

Conditions push button initializes the Loads/
Boundary Conditions GUI, shown in Figure 9(a),
which enables the user to prescribe a thermal and/or
mechanical loading step for the given analysis. For a
thermal load, the user may apply a temperature change,

T, which is applied globally to each node in the
model, as shown in Figure 9(b). For a mechanical
load, the user can prescribe either a transverse normal
displacement (dn), transverse shear displacement (ds), or
combined transverse normal/shear displacements, as
shown in Figure 9(b). These displacements are applied
to active control nodes, n2 and n4, as shown in Figure
9(b). The Load/Boundary Conditions GUI also allows

the user to congure the time incrementation settings

for each analysis step, such as the maximum number of
increments and the increment size.

(d) Analysis outputs. The Analysis Outputs push

button initializes the Analysis Outputs GUI, shown in
Figure 10. This allows the user to select a number of
eld or history output requests for the current analysis.
By default, the COMM Toolbox outputs stress, strain,
and displacement components for an analysis, however,
the Analysis Outputs GUI enables the user to request
supplementary eld outputs, such as plastic strain components, cohesive element variables, or element temperature values, depending on the type of analysis

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

McCarthy and Vaughan

1723

being carried out. By default, the history output of the

reaction forces and spatial displacements at the control
nodes is stored and written to the ABAQUS data le
(job-name.dat), as shown in Figure 10. This data is
required for a number of post-processing features available in the COMM Toolbox, which are described in
more detail in section Post-Processing ABAQUS
Results.

(e) Input file preview. This listbox displays a preview

of the ABAQUS Input File which the COMM Toolbox
will create, as shown in Figure 7. The contents are
updated each time a model parameter is edited using
the ABAQUS Toolbar menu, that is upon completion
of any of the steps (a)-(d), described above. For example, once the user edits the material properties using the
Material Properties GUI, shown in Figure 5(b), the
contents of the ABAQUS Input File are automatically

Figure 10. Analysis outputs graphical user interface (GUI).

updated to include these parameters under the

*Material option in the Input File Preview, as shown
in Figure 7. Once the user has dened all the parameters describing the model behavior, the Create
push button is clicked and the COMM Toolbox
prompts the user to enter a job name and subsequently
creates the ABAQUS Input File (job-name.inp). The
COMM Toolbox follows a similar procedure to that
described in section Main COMM Toolbox GUI (e) to
prescribe periodic boundary conditions and create the
ABAQUS Input File.

(f) Submission options. The Submission Options panel

allows the selection of any ABAQUS Input File from
the current directory using the listbox on the ABAQUS
Analysis GUI, shown in Figure 7. The input le can be
submitted to run on either the local machine or parallel
computing resources. In the case of the Local Machine,
the analysis is carried out in the current working directory and the ABAQUS Input File (job-name.inp) is submitted, upon execution of the Submit push button, to
the ABAQUS solver by the COMM Toolbox through
the DOS command, dos(abaqus job job-name
interactive).
In the case of Parallel Resources being selected, the
job is submitted to run on the AMPS parallel computing cluster located at the University of Limerick. This
capability was implemented by remotely connecting the
COMM Toolbox to the AMPS parallel computing cluster using an open source Secure Shell (SSH) Toolbox,
which was developed for MATLAB by Nehrbass et al.9
The SSH Toolbox uses the widely recognized PuTTY
SSH client to connect between the Windows and Linuxbased systems. This allows the execution of commands
from the COMM Toolbox to the remote host.
Upon execution of the Submit push button located
on the ABAQUS Analysis GUI, shown in Figure 11,
the COMM Toolbox connects to the AMPS cluster, as

Figure 11. Submitting jobs to parallel computing resources using the COMM Toolbox.

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

1724

Journal of Composite Materials 46(14)

Figure 12. Graphical user interface (GUI) for post-processing Abaqus output database files.

Figure 13. Fiber distributions generated by the COmposite MicroMechanics (COMM) Toolbox for (a) Vf 20%, (b) Vf 40%, and
(c) Vf 59%.

shown in Figure 11(a). This requires the user to provide

authentication, as shown in Figure 11(b). Finally, the
user then denes the number of nodes and processors
the job requires, as shown in Figure 11(c), and these
details are written to a torque (submission) script. The
COMM Toolbox then copies the ABAQUS Input File
to the AMPS cluster drive and executes the torque
script which launches the job on the AMPS queuing
system. This submission process can be easily modied
for submission to other remote parallel computing
resources by changing the SSH connection properties,
that is the SSH host and user names. These may be
modied by selecting Edit ! SSH Properties on the
main menu bar on the ABAQUS Analysis GUI,
shown in Figure 11.

This feature is discussed in detail in section PostProcessing Abaqus Results.

(g) Results. This push button initializes the ABAQUS

(a) Output database directory. This listbox displays

Results GUI which enables the user to eciently postprocess results from micromechanical analyses.

all the ABAQUS output database les (job-name.odb)

located in the current directory. Upon execution of the

Post-processing ABAQUS results

The Results push button located on the ABAQUS
GUI, shown in Figure 7, initializes the ABAQUS
Results GUI, shown in Figure 12. Its function is to
eciently post-process results from micromechanical
analyses and also to monitor the progress of active
jobs. The ABAQUS Results GUI allows the user
access ABAQUS output database les, display, and/
or extract stressstrain response data from an analysis
and view the increment summary of an active job. The
main components of the ABAQUS Results GUI are
identied in Figure 12 and may be described as follows,

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

McCarthy and Vaughan

1725

Open push button, the COMM Toolbox opens the

selected output database le in the ABAQUS/CAE
window (not shown).

(b) Results options. The Results Options panel allows

the user to access results from the selected ABAQUS
output database le through either the Plot or
Export options, shown in Figure 12 as push buttons.
Results are available as stressstrain data determined in
either the 2- or 3-direction from the active control
nodes, n2 and n4, which were identied in Figure 9(b).
The Plot push button displays the stressstrain
response from a given node on the adjacent axes
(shown in Figure 12(c)), while the Export push
button automatically exports the selected data to a
Microsoft Excel worksheet (job-name.xls). The
COMM Toolbox uses the ABAQUS data le (jobname.dat) to access results, which contains the nodal
outputs of reaction forces and displacements. As the
ABAQUS data le is updated throughout an analysis,
this enables the user to obtain a live update of the
current macroscopic stress state in a model.

(c) Stressstrain response. As explained above, the

stressstrain response determined from one of the
active control nodes is displayed here.

(d) Status file preview. This listbox enables the user to

monitor the progress of active jobs by displaying a preview of the ABAQUS status le (job-name.sta). This le
stores the status of increment summaries, which are
written by ABAQUS during an analysis, revealing the
current stage of the analysis.

COMM Toolbox case study: effect of fiber

volume fraction on transverse behavior
To demonstrate the features of the newly developed
COMM Toolbox, a case study was carried out examining the eect of ber volume fraction on the behavior of
a carbon ber/epoxy (HTA/6376) composite under
loading in the transverse plane. To consider the eect
of ber volume fraction (Vf) on the transverse behavior,
the COMM Toolbox was used to generate a number of
microstructures which had ber volume fractions of Vf
20%, Vf 40%, and Vf 59%. The NNA6 was
used to generate the microstructures for Vf 59%.
The NNA6 uses experimentally measured nearest
neighbor distribution functions to dene the interber
distances and an experimentally measured diameter distribution function to assign ber diameters in the RVE.
This allows statistically equivalent ber distributions of
a carbon ber/epoxy (HTA/6376) composite to be generated, enabling an accurate representation of the local
microscopic stress state.6 Meanwhile, the HCM was

used to generate microstructures for both Vf 20%

and Vf 40%. The HCM has been widely used to
generate lower ber volume fraction microstructures
(less than approximately Vf 50%) for composite
materials.1,5,7 It represents the bers as a set of nonoverlapping inclusions that have been randomly distributed within the domain. As was carried out for the
NNA,6 a minimum inhibition distance of 0.1 mm
between neighboring bers was specied in the HCM
to allow for adequate discretisation in the interber
region. While this section analyses the eect of ber
volume fraction on the mechanical behavior of the
material, a statistical analysis has also been carried
out in order to characterize the ber distributions that
are generated by the NNA (for high ber volume fraction microstructures) and by the HCM (for low ber
volume fraction microstructures).

Statistical analysis
To characterize the ber distributions produced by the
NNA and HCM, 20 microstructures were generated
using the COMM Toolbox at each ber volume fraction (i.e. Vf 20%, Vf 40%, and Vf 59%). These
measured 165  165 mm and a resulting microstructure
for each ber volume fraction is shown in Figure 13.
The Statistical Analysis GUI, discussed in section Main
COMM Toolbox GUI (d), was used to extract the 1st
and 2nd nearest neighbor distribution functions for
these models. Nearest neighbor distribution functions
detail the short-range interaction between bers by analyzing the distance between each ber and their nth
closest neighbour.10 Shown respectively in Figures
14(a) and (b) are the mean values of the 1st and 2nd
nearest neighbor distribution functions for each ber
volume fraction considered.
For the highest ber volume fraction, that is, Vf
59%, the 1st and 2nd nearest neighbor distributions
exhibit narrow ranges and high peaks occur at distances
of 7 mm and 7.2 mm, respectively. This is an obvious
result of the high ber volume fraction as there is limited space available for bers to position themselves. In
contrast, for the lowest ber volume fraction, that is, Vf
20%, the 1st and 2nd nearest neighbor distribution
functions show no observable peak and each distribution exhibits a wide range, highlighting that bers have
no tendency, or indeed obligation, to cluster close to
one another. Meanwhile, for Vf 40%, the characteristics of the nearest neighbor distribution functions are
intermediate to those of the highest (i.e. Vf 59%)
and lowest (i.e. Vf 20%) ber volume fractions
shown. It was shown in Ref. 11 that, under transverse
shear loading, the proximity of neighboring bers had a
signicant eect on the interfacial stress state. This suggests that the large dierence in the 1st and 2nd nearest

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

1726

Journal of Composite Materials 46(14)

Figure 14. (a) 1st Nearest neighbor distribution function and (b) 2nd Nearest neighbor distribution function for distributions
generated.

neighbor distribution functions observed here for each

ber volume fraction could aect the failure behavior
under loading in the transverse plane.

Mechanical analysis
To examine the eect of ber volume fraction on the
material behavior, a combined transverse compressive
and shear loading regime was used to predict failure
envelopes for the material. The COMM Toolbox was
used to generate an RVE measuring 66  66 mm at
each ber volume fraction under consideration (i.e. Vf
20%, Vf 40%, and Vf 59%). The behavior of
the constituent materials is the same as that described
in Ref. 3. Thus, the HTA bers are assumed to be linear
elastic while the 6376 matrix behaves as an elastic-plastic solid where yielding behavior is described by MohrCoulomb plasticity theory, which was outlined in section Pre-processing ABAQUS Input Files for nonlinear
analysis (a), where the internal friction angle and cohesion stress for the 6376 epoxy matrix are 26  and
c 82 MPa,3 respectively. Meanwhile, as described in
section Pre-processing ABAQUS Input Files for nonlinear analysis (b) a cohesive section was included at
the ber-matrix interface to predict the eects of
ber-matrix debonding. For simplicity, it has been
assumed for this analysis that the interfacial normal
strength was equal to the interfacial shear strength,
that is, ton tos , and the fracture energies were the
same for Mode I and Mode II type failures, that is,
I II (a similar assumption has previously been
made in Ref. 1, 3, and 12).
For this combined loading case, a transverse shear
load was imposed on the RVEs by applying complimentary horizontal and vertical displacements (ds/2)
to the active control nodes n4 and n2, respectively.
Meanwhile, the transverse compressive load
was imposed by applying a horizontal normal displacement (n ) at the active control node n2, as shown in

Figure 9(b). By varying the ratio of the applied transverse shear/normal displacements (s =n ), a failure
envelope for each ber volume fraction was determined
in the 22  23 stress space.
It was shown in Ref. 12 that failure of the material
under transverse shear loading could be dominated by
ber-matrix interface failure (for a weak ber-matrix
interface) or by matrix yielding (for a strong bermatrix interface). Thus, the analysis carried out here
considers both failure modes. For the rst case, an
interface strength of ton=s 60 MPa and a fracture
energy of
10 J/m2were assumed, meaning that
failure was dominated by ber-matrix interface failure.
Meanwhile, the second case assumes a perfect bermatrix interface, meaning that failure was dominated
by matrix yielding. Constructing a failure surface for
each dierent case (dierent ber volume fractions and
ber-matrix interface strengths) required a large
number of models to be generated. The COMM
Toolbox allowed for the ecient generation of over
40 dierent models for this study using the ABAQUS
Analysis GUI, described in section Pre-processing
ABAQUS Input Files for nonlinear analysis. The
ABAQUS Analysis GUI also allowed for convenient
job submission to the AMPS parallel computing cluster. Meanwhile, the Abaqus Results GUI, discussed in
section Post-Processing ABAQUS Results, was used to
extract the stressstrain response of the RVEs at each
loading ratio.
Figure 15(a) shows the eect of ber volume fraction
on the transverse failure surface when a ber-matrix
interface strength of ton=s 60 MPa and a fracture
energy of
10 J/m2 were assumed. For this interface-dominated failure, it was found that reducing the
ber volume fraction actually results in a slight expansion of the transverse failure surface. Figure 15(a)
shows only a slight dierence between the failure surfaces of 40% and 59% ber volume fraction; however,
the 20% ber volume fraction shows much high failure

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

McCarthy and Vaughan

1727

Figure 15. Effect of fiber volume fraction on transverse failure surface for (a) interface-dominated failure and (b) matrix-dominated
failure.

strengths for all loading ratios. This higher strength for

lower ber volume fractions can be attributed to the
larger interber spacing between neighboring bers.
This was established in Figure 14(a) and (b), where,
for lower ber volume fractions, the 1st and 2nd nearest
neighbor distribution functions showed no observable
peak and exhibited a wide range. It is also shown in
Ref. 11, for the transverse shear loading case, that a
larger interber spacing causes a reduction in the interfacial normal and shear stresses. Thus, in lower ber
volume fraction arrangements, the larger interber spacing eectively delays the onset of debonding, resulting
in higher strengths for each loading ratio. It should also
be noted that due to the larger interber spacing, the
propagation of interface cracks to adjacent bers (due
to stress redistribution) is inhibited and promotes
further yielding in intermediate matrix regions, leading
to higher predicted strengths. In other terms, it is easier
for an interface crack to propagate to a neighboring
interface which is closer to it, than to one which is
further away. This can be seen when the nal plots
from the Vf 20% case and Vf 59% case are
compared for a loading ratio of s =n 2, as shown
in Figure 16(a) and (b), respectively. While the lower
ber volume fraction RVE (Vf 20%) results in a
higher predicted strength under loading in the transverse plane, it must be remembered that it would be
only 35% as strong as the high ber volume fraction
RVE (Vf 59%) when loaded in the ber direction.
Figure 15 (b) shows the eect of ber volume fraction on the transverse failure surface for an assumed
perfect ber-matrix interface, such that failure was
matrix dominated. In contrast to interface dominated
failure (shown in Figure 15 (a)), it was found that reducing the ber volume fraction for matrix dominated
failure results in a contraction of the transverse failure

surface, as shown in Figure 15 (b). This is caused by the

lack of reinforcement in low ber volume fraction congurations, which allows matrix yield paths to propagate more easily. This can be seen in Figure 17(a) and
(b), which shows the deformation in an RVE subject to
a loading ratio of s =n 1 for the Vf 20% and Vf
59% cases, respectively. For Vf 20%, a dominant
shear band forms which is unobstructed by bers,
whereas for Vf 59% the high ber volume fraction
inhibits the propagation of matrix yielding as many of
the yield paths are arrested due to bers obstructing
their path. It is also interesting to note that, while the
low ber volume fraction RVE, shown in Figure 17(a),
shows a dominant yield path and a relatively
small amount of yielding in the rest of the microstructure, the high ber volume fraction RVE, shown in
Figure 17(b), shows that intense yielding is widespread
throughout the microstructure. Thus, it was found
here, that widespread yielding over the microstructure
results in damage being more eectively dissipated over
the entire RVE, leading to a higher predicted strength.

Concluding Remarks
This paper has outlined the development of the
Composite Micromechanics (COMM) Toolbox which
provides ecient pre- and post- processing capabilities
for micromechanical analyses of composite materials.
As computational power continues to increase, more
detailed and advanced multiscale modeling strategies
have emerged, which better predict internal damage
mechanisms and their eect on structural behavior.
However, the inherent complexity associated with
such modeling strategies forms a barrier with respect
to their advancement from academic interests to practical industry based implementation. The COMM

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

1728

Journal of Composite Materials 46(14)

Figure 16. Final deformation for a loading ratio of dsdn 2 in representative volume elements (RVEs) where (a) Vf 20% and
(b) Vf 59%.

Figure 17. Final deformation for a loading ratio of ds/dn 1 in representative volume elements (RVEs) where (a) Vf 20% and
(b) Vf 59%.

Toolbox addresses this issue as it provides a simple,

convenient environment for creating complex micromechanical damage models, which facilitate a better
understanding of how local deformation mechanisms
inuence material behavior at higher length scales.
The COMM Toolbox automates all manual tasks
associated with micromechanical analysis of composite
materials. These include microstructure generation, statistical analysis of ber distributions, preparation of

input les (either linear or nonlinear), job submission

to parallel computing resources, active job monitoring,
and homogenization calculations. To demonstrate the
functionality of the COMM Toolbox, a parametric case
studies was carried out which analyzed the eect of
ber volume fraction on behavior under loading in
the transverse plane. It was found that a reduced ber
volume fraction led to an increased predicted strength
when failure was dominated by interface cracking.

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013

McCarthy and Vaughan

1729

In contrast, it was found that a reduced ber volume

fraction led to a reduction in predicted strength when
failure was dominated by matrix yielding.
The COMM Toolbox provides a simple, ecient
predictive modeling capability for composite materials,
based on the physical interactions of the constituent
phases which could guide material system selection or
processing conditions. The intellectual overhead in
learning to use the COMM Toolbox is low and could
be suited to analysts with varying levels of expertise.
Importantly, the COMM Toolbox allows the analyst
to apply advanced nonlinear and fracture modeling
techniques without requiring extensive background
knowledge. In particular, the COMM Toolbox could
prove useful in industrial applications for determining
application-specic properties for composite material
systems by numerically carrying out various parameter
studies on variables such as cure temperature and/or
interface strength, for example. The COMM Toolbox
is freely available to the community and can be
obtained by contacting the corresponding author.
Acknowledgments
This authors wish to acknowledge the funding provided by
the Irish Research Council for Science, Engineering and
Technology (IRCSET) and Science Foundation Ireland
(SFI). Computational work was carried out using the MSSI
AMPS cluster, enabled under the framework of the INSPIRE
programme, funded by the Irish Governments Programme
for Research in Third Level Institutions, Cycle 4, National
Development Plan 20072013.

References
1. Gonzalez C and Llorca J. Mechanical behavior of unidirectional fiber-reinforced polymers under transverse compression: Microscopic mechanisms and modeling. Compos
Sci Technol 2007; 67(13): 27952806.

2. Totry E, Molina-Aldaregu a JM, Gonzalez C and Llorca

J. Effect of fiber, matrix and interface properties on the
in-plane shear deformation of carbon-fiber reinforced
composites. Compos Sci Technol 2010; 70(6): 970980.
3. Vaughan TJ and McCarthy CT. Micromechanical modelling of the transverse damage behaviour in fibre reinforced composites. Compos Sci Technol 2011; 71(3):
388396.
4. ABAQUS 6.9. Hibbett, Karlsson and Sorensen Inc,
Pawtucket, RI, USA, 2009.
5. Trias D, Costa J, Mayugo JA and Hurtado JE. Random
models versus periodic models for fibre reinforced composites. Comput Mater Sci 2006; 38(2): 316324.
6. Vaughan TJ and McCarthy CT. A combined experimental-numerical approach for generating statistically
equivalent fibre distributions for high strength laminated
composite materials. Compos Sci Technol 2010; 70(2):
291297.
7. Buryachenko VA, Pagano NJ, Kim RY and Spowart
JE. Quantitative description and numerical simulation
of random microstructures of composites and their
effective elastic moduli. Int J Solids Struct 2003; 40(1):
4772.
8. Van der Sluis O and Schreurs PJG. Brekelmans WAM
and Meijer HEH. Overall behaviour of heterogeneous
elastoviscoplastic materials: Effect of microstructural
modelling. Mech Mater 2000; 32(8): 449462.
9. Nehrbass J, Samsi S, Chaves JC, Unpingco J, Guilfoos B
and Ahalt S. Interfacing PC-based Matlab directly to
HPC resources. HPCMP Users Group Conference, 2006.
10. Diggle P. Statistical analysis of spatial point patterns. 2nd
ed. London: Arnold, 2003.
11. Vaughan T and McCarthy CT. A micromechanical study
on the effect of intra-ply properties on transverse shear
fracture in fibre reinforced composites. Compos A: Appl
Sci Manuf 2011; 42(9): 12171228.
12. Totry E, Gonzalez C and Llorca J. Failure locus of fiberreinforced composites under transverse compression and
out-of-plane shear. Compos Sci Technol 2008; 68(34):
829839.

Downloaded from jcm.sagepub.com at NATIONAL INST. OF TECHNOLOGY on February 5, 2013