Article

Finite Element Analysis and Experimental

Characterisation of SMC Composite Car Hood
Specimens under Complex Loadings
Josh Kelly, Edward Cyr ID
and Mohsen Mohammadi *
Marine Additive Manufacturing Centre of Excellence (MAMCE), University of New Brunswick,
3 Bailey Drive, P.O. Box 4400, Fredericton, NB E3B 5A1, Canada; josh.kelly@unb.ca (J.K.); ecyr@unb.ca (E.C.)
* Correspondence: Mohsen.Mohammadi@unb.ca; Tel.: +1-506-458-7104; Fax: +1-506-453-5025




Received: 7 July 2018; Accepted: 28 August 2018; Published: 4 September 2018


Abstract: Composite materials have recently been of particular interest to the automotive industry
due to their high strength-to-weight ratio and versatility. Among the different composite materials
used in mass-produced vehicles are sheet moulded compound (SMC) composites, which consist
of random fibres, making them inexpensive candidates for non-structural applications in future
vehicles. In this work, SMC composite materials were prepared with varying fibre orientations and
volume fractions (25% and 45%) and subjected to a series of uniaxial tensile and flexural bending
tests at a strain rate of 3 × 10−3 s−1 . Tensile strength as well as failure strain increased with the
increasing fibre volume fraction for the uniaxial tests. Flexural strength was found to also increase
with increasing fibre percentage; however, failure displacement was found to decrease. The two
material directions studied—longitudinal and transverse—showed superior strength and failure
strain/displacement in the transverse direction. The experimental results were then used to create
a finite element model to describe the deformation behaviour of SMC composites. Tensile results were
first used to create and calibrate the model; then, the model was validated with flexural experimental
results. The finite element model closely predicted both SMC volume fraction samples, predicting the
failure force and displacement with less than 3.5% error in the lower volume fraction tests, and 6.6%
error in the higher volume fraction tests.

Keywords: discontinuous reinforcement; glass fibres; uniaxial tensile tests; three-point bending tests;
mechanical properties; finite element modelling

1. Introduction
In recent years, the technology to manufacture composites has significantly improved, and in
turn, composite materials are becoming popular in applications such as panel componentry in
automotive and aerospace industries. In particular, fibrous composites are of utmost interest because
they can provide favourable mechanical properties and incredible weight savings, far exceeding
common materials such as metals and polymers. Composite materials combine the lightweight
characteristics of polymers and the high strength characteristics of fibres for reinforcement (generally
glass, carbon, or natural fibres). Typically, the constituents of composite materials are inexpensive,
and the advancement of manufacturing processes has continued to improve the cost effectiveness
of these materials. The ability to be formed into virtually any desired shape is another attraction
for both the automotive and aerospace industries. In addition, other benefits of these materials
are their lightweight capabilities, corrosion resistance, chemical stability, high stiffness, and high
strength-to-weight ratio [1,2].
The modeling of composites is a difficult task; however, several different approaches have been
used recently presented in a comprehensive review by Liu and Zheng [3]. A common approach is to

J. Compos. Sci. 2018, 2, 53; doi:10.3390/jcs2030053 www.mdpi.com/journal/jcs

J. Compos. Sci. 2018, 2, 53 2 of 24

use continuum-based mechanics for developing constitutive models to predict stiffness degradation
and damage evolution. Damage evolution and loss of stiffness models can be related on a macroscale
to cracks and voids within the composite matrix, and in turn provide updated mechanical properties.
Many researchers have contributed to the framework around progressive failure analysis through
the use of continuum damage mechanics [4–9]. Their work includes the use of the damage tensors,
free energy, thermodynamics, and power dissipation to predict the damage evolution of a composite.
Schapery [6], Murakami and Kamiya [7], Hayakawa et al. [8], and Olsson and Ristinmaa [10] built off
of this work by introducing a second-order or fourth-order damage tensor to improve the isotropic and
anisotropic models to predict damage evolution and stiffness degradation. These new models made
use of damage evolution laws to address previous shortcomings. Matzenmiller [11], Maa and Chen [12],
and Camanho et al. [13] all developed thermodynamic models to further describe composite damage
evolution and stiffness degradation. These models further developed the relationship between the
damage tensor and the resulting forces, stresses, and strains within the material, and made use of the
known composite failure modes to aid in predicting damage. Perreux et al. [14] and Liu and Zheng [15]
utilised three damage tensors to describe the three main failure modes of composites: fibre breakage,
matrix cracking, and shear failure, where they used these tensors to define damage evolution. However,
it was found that elastic/damage coupling constitutive models may not provide adequate information
regarding damage initiation and evolution; therefore, damage/plasticity coupling nonlinear models
were developed [3]. This is to account for the nonlinear behaviour that occurs in the plastic region,
and its effect on the damage properties. A notable nonlinear elasticity–plasticity/damage-coupling
constitutive model was proposed by Lin and Hu [16]. This model, in conjunction with a mixed failure
criterion, was used to predict composite behaviour.
A second approach to modelling composites is to utilise damage initiation and evolution criteria
based on phenomenological material models, fracture mechanics, and failure modes. Composites may
undergo many possible failure modes before the ultimate fracture of the material. Damage initiation
and evolution criteria aid in determining when and how the damage begins to occur, and how the
properties of the material evolve afterwards [3]. Popular failure criteria, which provide the framework
for much of the work done in this field, includes Tsai and Wu [17], Hashin [18], Hoffman [19],
and Yamada-Sun [20] criteria. Many of these failure criterions are able to predict when failure begins
on the first layer of the composite, but are unable to identify the root failure modes. Liu and Zheng [15]
and Padhi et al. [21] addressed this issue by separating the three stress components to give a better
understanding and identify the failure modes. New failure theories have also being developed; some
notable theories are the strain invariant failure theory [22–24], the multi-continuum theory [25,26],
and the micromechanics-based failure theory [27,28].
It should be noted that, while effective, these theories are not as popular as the previous
failure criteria above. The virtual crack closure technique (VCCT) that was proposed by Rybicki
and Kanninen [29,30] and was based off of the early work of Irwin [31] is a prevalent method for
determining the mixed failure modes/parameters at the tip of a crack. VCCT is more effective at
calculating crack tip fracture mechanics than other methods such as the J-integral [32], which uses
linear-elastic fracture mechanics, because VCCT can be used on three-dimensional (3D) structures
and has low mesh requirements [33]. However, VCCT falls short in predicting failure initiation and
evolution. To counter this, a cohesive zone concept was proposed by Dugdal [34] and Barenblatt [35],
which describes a discrete failure across the material separation plane. This concept was further
developed by Hutchinson and Suo [36], Tvergaard and Hutchinson [37], Allen and Searcy [38],
Camanho et al. [39], and Xie and Waas [40], leading to the cohesive zone theory. This theory describes
the area in front of the crack tip, where the fracture properties are influenced by the upper and lower
layers, while controlled by the cohesive traction–displacement discontinuity relationships [3]. Shin
and Wang [41] further improved the cohesive zone model to predict the notch strength and damage
evolution of anisotropic composites, while Rami [42] proposed a micromechanical cohesive zone model
based on a 3D representative unit cell. Sabiston et al. [43] proposed a model based on functionally
J. Compos. Sci. 2018, 2, 53 3 of 24

graded interface theory [44] in order to overcome the computational inefficiency of cohesive zone
theory. They have used the model to predict the quasi-static and moderate strain rate behaviour of
composites [45].
One of the greatest tasks of modelling composites is to take the damage models listed above and
effectively integrate them into finite element software. Often times, damage progression models of
composites rely on element failure to update the progression of damage within the laminate. This also
means that the material properties have to be recalculated simultaneously, while holding the same
load to remain in equilibrium, causing instability in some simulations due to numerical convergence
issues [3]. This problem was later resolved by Tay et al. [22,23,46], at which point the element
failure method was proposed. This method allowed for more rapid convergence and computational
robustness by manipulating the nodal forces of elements to simulate damage, while leaving the material
stiffness unaffected; although subsequently, this method struggles to predict damage evolutions due to
stiffness degradation. Gao and Bower [47] also tackled the convergence issue by introducing a viscous
constant into Xu and Needlman’s [48] cohesive model, which effectively solved the issue. Another
popular algorithm is the arc-length algorithm proposed by Riks [49], Ramm [50], and Crisfield [51].
This method was found to produce very accurate simulations, but was time-consuming in comparison
to previous models.
Since the interface between the matrix and fibre is on the microscopic scale, to simulate the actual
microscopic failure mechanisms on a macroscopic scale would take far too long to compute. For that
reason, a multiscale approach would be a useful effective tool [3]. Bednarcyk and Arnold [52] used
micromechanics and produced finite element codes to provide an outline for the multiscale stochastic
analysis of progressive failure in composite laminates. This was used to predict the nonlinear response
of laminates. This differed from others by incorporating a parameter for the fibre strength randomness
on multiple scales so that accurate microscopic failure could be achieved. Van Der Meer and Sluys [53]
evaluated the existing continuum material models used for micromodelling the damage of composites,
and found that due to the homogenisation of the matrix, the prediction of matrix failure was insufficient.
They proposed two models: a continuum damage model and a softening plasticity model to combat
this issue. Robbins and Ready [54] reduced the computational effort in multiscale progressive damage
and global failure simulations by proposing an adaptive kinematics-based multiscale model that makes
use of a hierarchical finite element method to reduce computation time.
More recently, the extended finite element method (XFEM) has been developed [55–58].
This model allows a crack to propagate in any plane of the composite by defining solid continuum
elements, but also provides enriched element freedoms so that if failure criterion is satisfied in
any direction, the crack will be able to propagate. Guidault et al. [59] used this method to develop
a multiscale XFEM to determine crack propagation using a domain decomposition method to efficiently
demonstrate the crack’s global and local effects. Huynh and Belytschko [60] modelled the fracture
properties of fibre-reinforced composites using XFEM by meshing the fracture zone independently of
the fibre and matrix interface and crack morphology.
The main objective of this work is to accurately model the deformation behaviour of an SMC
composite car hood with randomly oriented glass fibres at room temperatures. This affordable SMC
composite provides favourable mechanical properties for the automotive industry in non-structural
applications. To collect all of the necessary material data for the study, a series of uniaxial tensile
and flexural bending tests were performed on specimens from the flat surface of the hoods, using an
INSTRON universal testing machine and an extensometer. The tests were performed at a strain rate
of 3 × 10−3 s−1 , which correlates to a low speed car accident [61]. The experimental results from the
uniaxial tensile tests were used to develop and calibrate a phenomenological material model in LS
DYNA, using MAT card 24. The finite element model was then validated by simulating the three-point
bending test, and relating the results to the flexural experimental data.
J. Compos. Sci. 2018, 2, 53 4 of 24

J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 4 of 24

2. Experimental Methodology
2. Experimental Methodology
2.1. Material and Manufacturing Process
2.1. Material and Manufacturing Process
The materials used for this investigation are car hoods from a sheet moulded compound composite.
Two setsThe materials used
of specimens for car
cut from thishoods
investigation are and
were tested car hoods
analysed;from a sheet
both moulded
sets had the same compound
glass fibre
composite. Two
reinforcement, sets material,
matrix of specimens andcut frombut
fillers, carin
hoods werequantities.
different tested and analysed;
One set ofboth setshad
hoods hadathe25%
same glass fibre reinforcement, matrix material, and fillers, but in different quantities.
volume fraction of glass fibres, and a PEEK (polyether ether ketone) matrix with 795 phr (per hundred One set of
hoods
resin) had a 25%
of CaCO volume fraction of glass fibres, and a PEEK (polyether ether ketone) matrix with
3 and 2 phr of MgO as fillers. The second set had a 45% volume fraction of glass fibres
and795 phr (per
a PEEK hundred
matrix withresin)
730 phrof CaCO 3 and 2 phr of MgO as fillers. The second set had a 45% volume
of CaCO 3 and 3 phr of MgO as fillers.
fraction of glass fibres and a PEEK
The car hoods were manufactured at the matrix with 730 phr of CaCO
Fraunhofer Project3 and 3 phr
Centre forofComposites
MgO as fillers. Research at
The car hoods were manufactured at the Fraunhofer Project Centre for Composites Research at
Western University [62], using a sheet moulding compound process. For this process, a film is rolled
Western University [62], using a sheet moulding compound process. For this process, a film is rolled
out onto a platform, where it awaits the premixed polymer resin before going through an extruder
out onto a platform, where it awaits the premixed polymer resin before going through an extruder
and then being distributed onto the film. As the film and resin are carried down the platform, several
and then being distributed onto the film. As the film and resin are carried down the platform, several
spools of glass fibre are connected to a cutter where the fibres are cut and distributed into the resin
spools of glass fibre are connected to a cutter where the fibres are cut and distributed into the resin
below. Once
below. Once thethe
fibres are
fibres distributed
are distributedthroughout
throughout thethe resin, an additional
resin, an additionallayer
layerofoffilm
filmisisplaced
placed onto
onto
of the resin. Following that, the material goes through a compounder to seal the constituents
of the resin. Following that, the material goes through a compounder to seal the constituents together. together.
TheThe
nownowsingular
singular sheet exits
sheet exitsthe
thecompounder
compounderand andis
is cut
cut to length. A
to length. Arobotic
roboticarmarmthen
thentakes
takes the
the sheet
sheet
andand
places it onto a hydraulic press, where it is pressed to the final desired
places it onto a hydraulic press, where it is pressed to the final desired shape. shape.

2.2.2.2.
Specimen Preparation
Specimen Preparation
Both tensile
Both tensileandandflexural
flexuralspecimens
specimenswere were prepared
prepared for the study.
for the study. TheThetensile
tensilespecimens
specimens were
were
machined
machined to to
ASTM
ASTMstandard
standardD638-14
D638-14[63],[63], while
while the
the flexural specimenswere
flexural specimens weremachined
machinedtotoASTM ASTM
standard
standardD790-15
D790-15 [64].
[64].Dimensional
Dimensionalinformation
information for for the tensile
tensile and
andflexural
flexuralspecimens
specimenscan canbebe found
found
in Figure
in Figure1A,B,
1A,B,respectively.
respectively.Both Bothtypes
typesof ofspecimens
specimens had had similar materialpreparation;
similar material preparation;full-sized
full-sized
SMCSMC composite
composite carcarhoods
hoodswere
werecut cutononaaband
band saw
saw toto yield
yield four mm ××305
305 mm
four 305 305mm mmplaques
plaques each.
each.
These
These plaques
plaques were
were fastenedtotoa asheet
fastened sheetofofplywood
plywood and and placed
placed into
into aa computer
computernumerical
numericalcontrol
control
(CNC)
(CNC) to to machine
machine thethe tensileand
tensile andflexural
flexuralspecimens.
specimens. EachEach hood
hood provided
providedboth bothlongitudinal
longitudinal and
and
transverse
transverse specimens,
specimens, as seen
as seen in Figure
in Figure 2; although
2; although the figure
the figure only only displays
displays tensile
tensile specimens,
specimens, it
it should
should be noted that the flexural specimens were cut from the plaques
be noted that the flexural specimens were cut from the plaques in the same fashion. ‘Longitudinal’ in the same fashion.
and‘Longitudinal’
‘transverse’ in and
this‘transverse’
work are in in this work are
reference in reference
to the car hood,tonotthethe
carfibre,
hood,due
not to
thethe
fibre, due to the of
randomness
randomness of
the fibre directions. the fibre directions.

Figure
Figure 1. 1.(A)
(A)Schematic
Schematicofofthe
thetensile
tensile specimen
specimen shape
shape and
and dimensions
dimensionsbased
basedon
onASTM
ASTMD638-14
D638-14
standard [63]; (B) Schematic of the flexural specimen shape and dimensions based on
standard [63]; (B) Schematic of the flexural specimen shape and dimensions based on ASTM ASTM D790-15
D790-15
standard
standard [64].[64].
J. Compos. Sci. 2018, 2, 53 5 of 24
J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 5 of 24
J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 5 of 24

Figure 2. Depiction of the position of rectangular sheets and tensile specimens cut from aa full-sized
full-sized
Figure 2. Depiction
Figure 2. Depiction of
of the
the position
position of
of rectangular
rectangular sheets
sheets and
and tensile
tensile specimens
specimens cut
cut from
from a full-sized
car hood.
car hood.

2.3.
2.3. Uniaxial
Uniaxial Tensile Testing
Tensile Testing
2.3. Uniaxial Tensile Testing
Following
Following specimenpreparation,
preparation,uniaxial
uniaxial tensile
teststests
werewere performedusingusing the INSTRON
Following specimen
specimen preparation, tensile
uniaxial tensile tests performed
were performed the INSTRON
using model
the INSTRON
model
1332. 1332. The INSTRON machine and associated grips for the uniaxial tests can be found in Figure 3A.
modelThe INSTRON
1332. The INSTRONmachine and associated
machine grips
and associated for for
grips thethe
uniaxial tests
uniaxial testscan
canbebefound
foundin
in Figure 3A.
Figure 3A.
The
The INSTRON
INSTRON collected
collected stress
stress data
data throughout
throughout the
the test,
test, and
and was
was used
used in
in conjunction
conjunction with
with a
a 50-mm
50-mm
The INSTRON collected stress data throughout the test, and was used in conjunction with a 50-mm
extensometer
extensometer to accurately
accurately recordthethe displacementmeasurements.
measurements. Theextensometer
extensometer was attached
extensometer to to accuratelyrecord
record thedisplacement
displacement measurements.The The extensometer was attached
was to
attached
to
thethe specimens
specimens using
using rubber
rubber bands,
bands, as as seen
seen in in Figure
Figure 3B.3B.
to the specimens using rubber bands, as seen in Figure 3B.

Figure 3. (A) INSTRON model 1332 used to perform the uniaxial tensile and flexural tests in this work;
Figure 3.
3. (A) INSTRON
(A)failed
INSTRON model
model 1332 used to
to perform the
the uniaxial tensile
tensile and
and flexural
flexural tests
tests in
in this
this work;
Figure
(B) A tensile specimen in1332 used
clamps with perform uniaxial
an extensometer attached. work;
(B) A tensile failed specimen in clamps with an extensometer attached.
(B) A tensile failed specimen in clamps with an extensometer attached.
2.4. Flexural Testing
2.4. Flexural
2.4. Flexural Testing
Testing
Flexural testing was performed on the INSTRON model 1332 as well, utilising two fixed
Flexural testing
Flexural testingwas
wasperformed
performedonison
the the INSTRON
INSTRON model model
1332 1332
well,asutilising
well, utilising
fixedtwo fixed
supports and one loading nose, which shown in Figure 4. In thisasexperiment, thetwo
loading supports
nose was
supports and
and one down one
loading loading
nose, nose, which is shown in Figure 4. In this experiment, the loading nose was
pressed onto thewhich is shown
middle in Figurespecimen
of the flexural 4. In thisuntil
experiment,
fracture.the
Theloading
loading nose was
nose pressed
force was
pressed down onto the middle of the flexural specimen until fracture. The loading nose force was
recorded by the INSTRON, and the displacement of the nose was used to determine the flexural
recorded by the INSTRON, and the displacement of the nose was used to determine the flexural
strain.
strain.
J. Compos. Sci. 2018, 2, 53 6 of 24

down onto the middle of the flexural specimen until fracture. The loading nose force was recorded by
the
J. INSTRON,
Compos. Sci. 2018,and the PEER
2, x FOR displacement
REVIEW of the nose was used to determine the flexural strain. 6 of 24

Figure
Figure 4.
4. Three-point
Three-point bending
bending experimental
experimental setup
setup with fractured specimen.
with fractured specimen.

2.5. Scanning Electron Microscopy

2.5. Scanning Electron Microscopy
The fractured specimens were analysed under a scanning electron microscope (SEM) to further
The fractured specimens were analysed under a scanning electron microscope (SEM) to further
investigate fracture initiation, and help determine the fracture modes of the material. This work was
investigate fracture initiation, and help determine the fracture modes of the material. This work was
performed in the Microscopy and Microanalysis facility located at the University of New Brunswick
performed in the Microscopy and Microanalysis facility located at the University of New Brunswick
(UNB), using a JEOL 6400 SEM system (JEOL USA, Inc., Peabody, MA, USA). Specimens were
(UNB), using a JEOL 6400 SEM system (JEOL USA, Inc., Peabody, MA, USA). Specimens were prepared
prepared using the methods presented in a study by Mohammadi and Salimi [65]. Samples of 25%
using the methods presented in a study by Mohammadi and Salimi [65]. Samples of 25% and 45%
and 45% volume fraction from both the uniaxial and flexural tests were analysed to better understand
volume fraction from both the uniaxial and flexural tests were analysed to better understand fibre and
fibre and matrix interactions.
matrix interactions.
3.
3. Experimental
Experimental Results
Results and
and Discussion
Discussion

3.1. Uniaxial Tensile Test

Test Results
Results
Uniaxial tensile
tensiletests
testswere
wereperformed
performed onontwotwo different
different sets sets of specimens
of specimens cut SMC
cut from fromcarSMC car
hoods,
hoods,
with 25%with
and25% 45%andvolume45%fraction
volumeoffraction of fibres respectively,
fibres respectively, in two separate in two separate longitudinal
orientations, orientations,
longitudinal
and transverse, andyielding
transverse,
four yielding
sets of data.fourEach
sets of data. Each
the tests wasof repeated
the tests was repeated
a minimum ofaeight
minimum
times of
to
eight
ensuretimes to ensureThe
repeatability. repeatability.
results for Theeachresults
data setforwere
eachentered
data setinto
were entered software
MATLAB into MATLAB (MATLABsoftware
and
(MATLAB and Statistics
Statistics Toolbox ReleaseToolbox
2012b, The Release 2012b, The
MathWorks INC.,MathWorks
Natick, MA, INC.,
USA),Natick,
which MA,
wasUSA),
used which was
to average
used to average
the results, producingthe results,
a singleproducing
stress–strain a single
curve stress–strain
for each data curve
set. The fortests
each dataperformed
were set. The tests were
at a strain
performed
rate of 3 × 10at −a3strain
s−1 (all
rateofof
the3 ×uniaxial
10−3 s−1 tensile
(all of the uniaxial
results withtensile
error barsresults
canwith error in
be found bars can be found
Appendix A).
in Appendix
Figure 5A). displays the experimental results from the uniaxial tensile tests in the longitudinal
Figure 5 displays
and transverse directionsthe for
experimental
both the 25% results
and from the uniaxial
45% volume fractiontensile tests in the
specimens. longitudinal
First, analysingand the
transverse
longitudinal directions
results, fromfor both the 25%
the figure, andbe45%
it can volume
clearly fraction
seen that specimens.
the 45% volumeFirst, analysing
fraction specimens the
longitudinal results, from the figure, it can be clearly seen that the 45%
displayed superior strength, a higher failure strain, and modulus. This is because the higher fibre volume fraction specimens
displayed
content leadssuperior strength,
to a decrease a higher
of the mean failure strain,
glass fibre and which
length, modulus. This isleads
eventually because the higher
to higher fibre
fibre–fibre
content leads to a decrease of the mean glass fibre length, which eventually
interaction and higher strength [66]. The 25% volume fraction specimens fractured on average at leads to higher fibre–fibre
interaction and MPa
a stress of 53.7 higher andstrength
averaged [66].a The 25%strain
failure volume fraction
of 1.10% specimens
strain, while fractured on average
the 45% volume at a
fraction
stress of 53.7
specimens MPa and
fractured averaged
on average at aa stress
failureof strain
89.0 MPa of 1.10% strain, while
and averaged thestrain
a failure 45% of volume
1.25% fraction
strain.
specimens fractured on average at a stress of 89.0 MPa and averaged a failure strain of 1.25% strain.
J. Compos. Sci. 2018, 2, 53 7 of 24
J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 7 of 24

Figure 5. Uniaxial tensile stress–strain curves for the longitudinal and transverse
transverse directions.
directions.

In the
In the transverse
transverseresults,
results,the
thesame
same trend
trend as as
thethe longitudinal
longitudinal direction
direction is displayed,
is displayed, where where
the 45%the
45% volume fraction specimens fractured at a higher stress, while achieving
volume fraction specimens fractured at a higher stress, while achieving a greater fracture strain and a greater fracture strain
and modulus.
modulus. It should
It should be noted
be noted that thethattransverse
the transverse direction
direction exhibitsexhibits
greatergreater
failurefailure
stress, stress,
failure failure
strain,
strain, and modulus when compared to the longitudinal specimens
and modulus when compared to the longitudinal specimens of the same volume fraction. of the same volume fraction.
This is This
due
is due
to to the alignment
the alignment of moreofglass
more glassinfibres
fibres in the transverse
the transverse directiondirection
during the during the manufacturing
manufacturing process,
process,toleading
leading more fibreto more
contentfibre content concentration
concentration and higher
and higher strength thanstrength than thedirection,
the longitudinal longitudinal
see
direction, see studies by Fu et al. [66] and Kuriger et al. [67]. The 25%
studies by Fu et al. [66] and Kuriger et al. [67]. The 25% volume fraction specimens fractured volume fraction specimens on
fractured on average at a stress of 72.4 MPa and averaged a failure strain
average at a stress of 72.4 MPa and averaged a failure strain of 1.30% strain, while the 45% volumeof 1.30% strain, while the
45% volume
fraction fraction
specimens specimens
fractured fractured
on average at on average
a stress at a stress
of 131.7 MPa and of 131.7 MPaaand
averaged averaged
failure a failure
strain of 1.50%
strain of 1.50% strain. The mechanical properties along with the anisotropic
strain. The mechanical properties along with the anisotropic behaviour of the SMC composite material behaviour of the SMC
composite
that material that
was investigated inwas
thisinvestigated in this study
study is comparable withisthe
comparable
similar SMC withcomposite
the similarproduced
SMC composite
by the
produced by the Wetlay process earlier
Wetlay process earlier and studied by Lu [68]. and studied by Lu [68].

Test Results
3.2. Three-Point Bending Test Results
The three-point
three-pointbending
bending tests
tests were were performed
performed with with the car
the same same
hood carspecimen
hood specimen volume
volume fractions,
fractions,
25% 25% and
and 45%, andmaterial
45%, and materiallongitudinal
directions, directions, longitudinal
and transverse. andAdditionally,
transverse. Additionally,
the flexural tests the
flexural
were tests were
performed performedof
a minimum a eight
minimum timesoftoeight
ensuretimes to repeatability,
their ensure their repeatability,
and MATLAB andwas MATLAB
used to
was usedthe
average toresults.
averageThese
the results.
tests wereThese tests were
performed atperformed at therate
the same strain same × 10−
of 3strain 3 s−of
rate 1 or
3 ×a 10 −3 s−1 or a
velocity of
velocity
10 mm/min of 10(all
mm/min (all of the
of the flexural flexuralresults
bending bending results
with errorwith
barserror bars
can be can be
found in found
Appendix in Appendix
B). B).
Figure 6 presents the experimental results from from the three-point
three-point bending
bending tests in the longitudinal
transversedirections
and transverse directionsforforboth
both thethe
25%25%andand
45%45% volume
volume fraction
fraction specimens.
specimens. It is noted
It is noted that the that the
same
same as
trend trend
wasas wasforseen
seen for the uniaxial
the uniaxial tensile
tensile tests cantests can be observed
be observed with thetests.
with the flexural flexural
First,tests. First,
focussing
focussing
on on the longitudinal
the longitudinal tests, the tests, the 45%fraction
45% volume volume specimens
fraction specimens had ahigher
had a vastly vastlyforce
higher at force at
failure,
failure,was
which whichmorewas
thanmore than
double double
that seen in that
theseen
25% in the 25%
volume volume
fractions, butfractions,
failed at abut failed
lower at a lower
displacement.
displacement.
The 25% volume The 25% volume
fraction specimens fraction specimens
on average onat
failed average
a forcefailed
of 40.1atN,
a force
and hadof 40.1 N, and had
an average an
failure
average failureofdisplacement
displacement 30 mm. The 45% of 30volume
mm. The 45% volume
fraction specimensfraction specimens
on averaged on averaged
failed at a force failed
of 87.1atN, a
forcehad
and of 87.1 N, andfailure
an average had andisplacement
average failure of 26displacement
mm. This trend of 26 mm.
with theThis trendproperties
flexural with the of flexural
SMC
properties of SMC composites was seen previously in the studied performed by Thomason and Vlug
[69], where higher fibre content led to higher flexural strength.
J. Compos. Sci. 2018, 2, 53 8 of 24

composites was seen previously in the studied performed by Thomason and Vlug [69], where higher
J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 8 of 24
fibre content led to higher flexural strength.

Figure
Figure 6. 6. Three-point
Three-point bendingforce
bending forcedisplacement
displacementcurves
curvesfor
forthe
thelongitudinal
longitudinal and transverse
transverse directions.
directions.

The flexural behaviour

The flexural behaviourof of thethe transverse
transverse samples
samples wassimilar
was very very to similar to those
those seen in theseen in the
longitudinal
longitudinal
tests. The 45% volume fraction results fractured at a higher force and a lower displacement thanlower
tests. The 45% volume fraction results fractured at a higher force and a those
displacement
seen in the 25% than those seen
volume in the
fraction 25% volume
results. However, fraction results. However,
the difference in failurethe difference
force between inthe
failure
25%
force between the 25% and 45% volume fraction specimens is less in the transverse
and 45% volume fraction specimens is less in the transverse results, while the failure displacement results, while the
failure displacement difference is greater. The transverse results displayed higher failure
difference is greater. The transverse results displayed higher failure forces than the longitudinal results, forces than
the longitudinal
while the fracture results, while the remained
displacements fracture displacements
relatively the remained
same. During relatively the same. During
the manufacturing the
process,
manufacturing process, some of the fibres can be aligned in the transverse direction
some of the fibres can be aligned in the transverse direction and out of plane due to the randomness and out of plane
due
andtosmall
the randomness
length of the and smallSince
fibres. lengthflexural
of the fibres. Since
tests are flexural tests are
a combination a combination
of tension of tension
and compression,
and compression, some of these fibres are under compressive loads, where
some of these fibres are under compressive loads, where a higher stiffness than the longitudinal a higher stiffness than the
longitudinal samples was achieved [69]. This was further verified using the
samples was achieved [69]. This was further verified using the SEM images. The 25% volume fraction SEM images. The 25%
volume
specimensfraction specimens
on averaged onataveraged
failed a force of failed
59.5 N,atand
a force
had an ofaverage
59.5 N,failure
and had an average
displacement of failure
30 mm.
displacement of 30 mm. The 45% volume fraction specimens on averaged failed
The 45% volume fraction specimens on averaged failed at a force of 105.7 N, and had an average failure at a force of 105.7 N,
and had an average failure displacement of 24 mm. Similar to the uniaxial tensile
displacement of 24 mm. Similar to the uniaxial tensile tests, the results obtained for the flexural tests tests, the results
obtained
along withforthe
the mechanical
flexural tests along with
behaviour of the
the mechanical
material were behaviour
close toof thethe material
results were close
presented by Lu to the
[68]
results presented by Lu [68] using the Wetlay
using the Wetlay process for manufacturing SMC composites. process for manufacturing SMC composites.

3.3.
3.3.Scanning
ScanningElectron
ElectronMicroscopy
MicroscopyAnalysis
Analysis
AA scanning
scanningelectron
electronmicroscopy
microscopy(SEM) (SEM) investigation
investigation waswasputputforth to look
forth deeper
to look into into
deeper the
mechanisms
the mechanisms of fracture and distinguish
of fracture and distinguishthe complicated failurefailure
the complicated modesmodes
undergone by short
undergone byfibre-
short
reinforced composites.
fibre-reinforced SEM analysis
composites. was performed
SEM analysis on both
was performed onthe tensile
both and flexural
the tensile samples
and flexural post-
samples
fracture. This helped
post-fracture. to determine
This helped if theifsame
to determine failure
the same modes
failure occur
modes in uniaxial
occur tension
in uniaxial as they
tension do in
as they do
bending. Initial
in bending. observations
Initial displayed
observations displayed that there
that therewere
weresigns
signsofof
fibre
fibrebundling,
bundling,which
whichled
ledtotomatrix
matrix
debonding,
debonding,and andsubsequently,
subsequently, failure.
failure.
Figure 7 displays the SEM images from fractured tensile specimens. Due to the randomness of
the fibres and the small focus of the SEM, both the transverse and longitudinal results have negligible
differences, and therefore, the same conclusions can be drawn. Figure 7A,B depict the cross-section
of a 25% and 45% volume fraction specimen fracture surface, respectively, while Figure 7C,D show
the top cross-sections of 25% and 45% volume fraction samples, respectively. In Figure 7A, bundles
of fibres are clearly depicted, where fibre bundling is one of the leading causes of fibre/matrix
J. Compos. Sci. 2018, 2, 53 9 of 24

Figure 7 displays the SEM images from fractured tensile specimens. Due to the randomness of
the fibres and the small focus of the SEM, both the transverse and longitudinal results have negligible
differences, and therefore, the same conclusions can be drawn. Figure 7A,B depict the cross-section
of a 25% and 45% volume fraction specimen fracture surface, respectively, while Figure 7C,D show
the top cross-sections of 25% and 45% volume fraction samples, respectively. In Figure 7A, bundles of
fibres are clearly
J. Compos. Sci. 2018,depicted, where
2, x FOR PEER fibre bundling is one of the leading causes of fibre/matrix debonding
REVIEW 9 of 24
in reinforced composites. As the fibres bundle, their available bonding surface area is reduced;
debonding
therefore, their in ability
reinforced composites.
to adhere to theAsmatrix
the fibres bundle, [11].
is reduced their Figure
available 7Bbonding
displayssurface area type
the same is
reduced; therefore, their ability to adhere to the matrix is reduced [11]. Figure
of fibre bundling, but due to the higher percentage of fibres, they are more concentrated together, 7B displays the same
type of fibre bundling, but due to the higher percentage of fibres, they are more concentrated
forming larger bundles than those seen previously. It is clear that higher fibre content led to higher
together, forming larger bundles than those seen previously. It is clear that higher fibre content led
fibre–fibre interactions, and bundling yielding higher strength [66]. This fibre bundling also led to the
to higher fibre–fibre interactions, and bundling yielding higher strength [66]. This fibre bundling also
fibre/matrix debonding of a group of fibres at once (Figure 7B), which is known as surface debonding,
led to the fibre/matrix debonding of a group of fibres at once (Figure 7B), which is known as surface
rather than the individual separation of fibres (Figure 7A), which is known as fibre bundle splitting.
debonding, rather than the individual separation of fibres (Figure 7A), which is known as fibre
Figure 7C,D both demonstrate the randomness of the fibre matrix, displaying that in just one small
bundle splitting. Figure 7C,D both demonstrate the randomness of the fibre matrix, displaying that
high magnification
in just one small high frame, multiple fibre
magnification directions
frame, multipleare found
fibre in both
directions areimages.
found inThe
bothpresence
images. ofThe out
of presence
plane-oriented
of out of fibres due to the fibres
plane-oriented manufacturing
due to the process can be process
manufacturing seen incanthese
be two
seen figures,
in these which
two
was confirmed by Thomason and Vlug [69] earlier. Although both images
figures, which was confirmed by Thomason and Vlug [69] earlier. Although both images are very are very similar in this
front, it can
similar inbe seen
this that itthe
front, 45%
can bevolume fraction
seen that the 45%sample
volumehas afraction
much higher
sampleconcentration of fibres,
has a much higher
as expected.
concentration of fibres, as expected.

Figure
Figure 7. 7.
SEMSEMofof Tensile(A)
Tensile (A)Fractography
Fractography25%
25% volume
volume fraction
fraction(VF)
(VF)atat50
50μm;
µm;(B)
(B)Fractography
Fractography45%
45%
VF at 50 μm; (C) Top Cross-Section 25% VF 100 μm; (D) Top Cross-Section 45% VF
VF at 50 µm; (C) Top Cross-Section 25% VF 100 µm; (D) Top Cross-Section 45% VF 100 µm. 100 μm.

Figure 8 displays the SEM images of fractured flexural specimens. Figure 8A,B show the
fractured surface cross-sections of 25% and 45% volume fraction samples, respectively, while Figure
8C,D show the top cross-sections of the fracture surface for 25% and 45% volume fraction samples,
respectively. In Figure 8A,B, the same type of fibre bundling and fibre/matrix debonding that was
experienced in the tensile specimens can be observed, although in this image, it was more common
J. Compos. Sci. 2018, 2, 53 10 of 24

Figure 8 displays the SEM images of fractured flexural specimens. Figure 8A,B show the fractured
surface cross-sections of 25% and 45% volume fraction samples, respectively, while Figure 8C,D show
the top cross-sections of the fracture surface for 25% and 45% volume fraction samples, respectively.
In Figure 8A,B, the same type of fibre bundling and fibre/matrix debonding that was experienced in
the tensile specimens can be observed, although in this image, it was more common to find bundles
of J.fibres debonding
Compos. Sci. 2018, 2, xfrom other
FOR PEER fibre bundles, rather than individual debonding fibres. This is
REVIEW 10 due
of 24 to
the complex nature of the flexural test, which ensures that a combination of tension and compression
This is due to the complex nature of the flexural test, which ensures that a combination of tension and
is present, leading to out of plane oriented-fibre bundles to debond [69]. The only difference seen
compression is present, leading to out of plane oriented-fibre bundles to debond [69]. The only
between the two images in Figure 8A,B is the fibre percentages; the general failure mechanics appear
difference seen between the two images in Figure 8A,B is the fibre percentages; the general failure
the same. Figure 8C,D display the top cross-sections of the specimens, and similar to observations from
mechanics appear the same. Figure 8C,D display the top cross-sections of the specimens, and similar
the tensile images, these figures display multiple fibre directions in just a small surface area, which
to observations from the tensile images, these figures display multiple fibre directions in just a small
demonstrates the fibre direction randomness that is seen throughout the entire part. Here, we also see
surface area, which demonstrates the fibre direction randomness that is seen throughout the entire
fibres
part.that have
Here, webeen
also seebent and that
fibres demonstrate
have beencurvature, although these
bent and demonstrate curvedalthough
curvature, fibres still appear
these curvedto be
debonding
fibres stillin largertobundles
appear of fibres,
be debonding ratherbundles
in larger than individually. Thisthan
of fibres, rather is because both ends
individually. This isofbecause
the short
fibres
both ends of the short fibres cannot be fully stressed in comparison with long fibres, leading to further
cannot be fully stressed in comparison with long fibres, leading to fibre bundling and fibre
debonding
bundling [70].and further debonding [70].

Figure
Figure 8. 8.SEM
SEMof
ofFlexural
Flexural (A)
(A) Fractography
Fractography25%25%VFVF
at 50 μm;µm;
at 50 (B) (B)
Fractography 45% VF
Fractography 45%at VF
50 μm;
at 50(C)
µm;
Top Cross-Section 25% VF 100 μm; (D) Top Cross-Section 45% VF 100 μm.
(C) Top Cross-Section 25% VF 100 µm; (D) Top Cross-Section 45% VF 100 µm.

3.4. Remarks on Fracture

Through careful analysis of the SEM images provided along with the flexural and tensile test
results, it can be stated with confidence that the main mode of failure for this SMC composite material
has been fibre bundling, leading to matrix/fibre debonding. These findings agree with other studies
on this topic in literature, see Matzenmiller et al. [11], Fu et al. [66], Kuriger et al. [67], Lu [68],
J. Compos. Sci. 2018, 2, 53 11 of 24

3.4. Remarks on Fracture

Through careful analysis of the SEM images provided along with the flexural and tensile test
results, it can be stated with confidence that the main mode of failure for this SMC composite material
has been fibre bundling, leading to matrix/fibre debonding. These findings agree with other studies on
this topicSci.
J. Compos. in2018,
literature,
2, x FORsee Matzenmiller
PEER REVIEW et al. [11], Fu et al. [66], Kuriger et al. [67], Lu [68], Thomason 11 of 24
and Vlug [69], Baxter [70] and Aboudi et al. [71]. This can help explain the differences between the
between
failure the failure
stress stressseen
and strains andbetween
strains seen
the 25% between
and 45%the 25%
volumeandfraction
45% volume fraction results.
experimental experimental
In the
results. In the 25% volume fraction specimens, there is more individual
25% volume fraction specimens, there is more individual fibre separation seen; therefore, the fibres fibre separation seen;
therefore,
are the fibresatare
not all breaking notThe
once. all 45%
breaking
volume at once.
fractionThe 45% volume
specimens fraction
see groups specimens
(bundles) see groups
of fibres failing
(bundles) of fibres failing at once, making these failures more catastrophic,
at once, making these failures more catastrophic, and likely the specimen fractured as soon as one and likely the specimen
fractured
of as soonseparations
these bundle as one of thesebeganbundle separations
to occur. The 25% began to occur.
volume Thesamples
fraction 25% volume fraction
are not able samples
to reach
are not able to reach as high of a stress, since more stress is being put on each
as high of a stress, since more stress is being put on each fibre, making them fail at lower stressesfibre, making them fail
at lower stresses and lower strains, while the 45% volume fraction samples
and lower strains, while the 45% volume fraction samples are able to reach higher stress and higher are able to reach higher
stress and
strains, since higher
there strains,
are more since
fibresthere are morethe
to distribute fibres
loadtoanddistribute
keep thethe load and
specimen keepHowever,
intact. the specimenonce
intact. However, once separation
separation occurs it is a rapid failure. occurs it is a rapid failure.

4. Finite Element Model Development

4.1. Model Development and Calibration

The model was developed using commercial FE software LS DYNA (LS-DYNA R8.0, LSTC, LSTC,
Livermore, CA, USA) and experimental data from the uniaxial tensile tests. For each materialmaterial type,
type,
i.e., volume fraction and direction, a different model was created, since the material properties differ
for each direction and volume fraction. This work began by 3D modelling the tensile specimen using
Autodesk Inventor, and importing that 3D model into LS-PrePost as an IGS file. file. The specimen was
then meshed using solid elements with three layers of elements through the thickness. The mesh was
chosen so that each element was approximately 1 mm. A mesh sensitivity analysis was conducted to
verify the choice of element size, as presented
presented in Section 4.2. Figure 9 shows the tensile specimen used
with nearly constant element size
size used throughout.
used throughout.

Figure
Figure 9.
9. Tensile specimen element
Tensile specimen element mesh.
mesh.

Several material models were tested to predict the deformation of the SMC composite, but
Several material models were tested to predict the deformation of the SMC composite, but material
material model (MAT_24) Linear Piecewise Plasticity was chosen, as it was the most effective by
model (MAT_24) Linear Piecewise Plasticity was chosen, as it was the most effective by allowing for
allowing for experimental stress–strain data to be a material input. Other material models considered
experimental stress–strain data to be a material input. Other material models considered include:
include: (MAT_6) Viscoelastic, (MAT_19) Strain Rate Dependent Plasticity, (MAT_22) Composite
Damage, (MAT_64) Rate Sensitive Power law Plasticity, and (MAT_76) General Viscoelastic. While
these models produced encouraging results, the unique deformation pattern of the composite
material required a material model that was able to precisely capture the stress–strain behaviour, and
for that reason, MAT_24 was chosen. MAT_24 utilises both elastics properties and effective plastic
J. Compos. Sci. 2018, 2, 53 12 of 24

(MAT_6) Viscoelastic, (MAT_19) Strain Rate Dependent Plasticity, (MAT_22) Composite Damage,
(MAT_64) Rate Sensitive Power law Plasticity, and (MAT_76) General Viscoelastic. While these models
produced encouraging results, the unique deformation pattern of the composite material required
a material model that was able to precisely capture the stress–strain behaviour, and for that reason,
MAT_24 was chosen. MAT_24 utilises both elastics properties and effective plastic stress–strain curves
as inputs for calculating material responses. The model utilises the elastic properties with elastic
equations to calculate the initial material response, but as the material reaches the yield point and
transitions from elastic to plastic, the material model transitions to the inputted effective plastic curve
to calculate the plastic response.
After the specimen was assigned a material mesh (aligned with the x-coordinate direction),
the boundary conditions were put in place to replicate the uniaxial test. The x = 0 end of the specimen
was prescribed nodal constraints to prevent rotation or displacement in any direction. At the x = L end,
element nodes were given a degree of freedom in the x-direction only, and assigned a nodal velocity to
correlate to the testing strain rate of 3 × 10−3 s−1 .
The model material properties were the next task in the simulation process. For this, the density,
Young’s modulus, Poisson’s ratio, yield stress, and failure strain were measured and prescribed as
material inputs. The density, Poisson’s ratio, and failure strain were extracted from experimental
test data. Due to the viscoelastic nature of the material, and the non-distinct yield point possessed
by thermosets [72], Young’s modulus and yield properties could not accurately be determined from
experimental testing alone; therefore, they had to be calibrated using the finite element method (FEM).
This was done by beginning with calculated estimates and iteratively adjusting the Young’s Modulus
and yield the stress to best match the experimental data. For this type of model, the plastic deformation
can be simulated very accurately, but it requires a logical elastic deformation zone as defined by the
linear plastic model. This type of elastic deformation is not present in SMC composites, as there is no
distinct elastic to plastic transition, as would be seen on a typical stress–strain curve. This is the reason
that iteration was needed to determine the Young’s modulus and yield stress. The Young’s modulus
was increased gradually, while the yield stress was progressively set to occur at earlier strain values,
so that the combinations of the two would result in a smooth transition from elastic deformation to
the user-defined plastic deformation. This resulted in less accuracy in the elastic portion of the curve,
but resulted in a much more accurate plastic deformation zone, which is of greater interest to the
study. Each model was calibrated separately, and the resulting properties can be found in Table 1.
Material orientations are designated with an ‘L’ for longitudinal or ‘T’ for transverse (in reference to
the direction of the hood and not fibre directions), along with the volume fraction of fibres.

Table 1. Material Model Inputs.

Density Young’s Modulus Yield Stress Failure Strain

Volume Fraction Poisson’s Ratio
(kg/m3 ) (GPa) (MPa) (mm/mm)
L25% 1910 10.3 0.30 5.5 0.98%
L45% 1910 17.1 0.30 5.5 1.15%
T25% 1910 12.55 0.30 5.4 1.13%
T45% 1910 21.4 0.30 5.94 1.34%

The results for the longitudinal and transverse uniaxial tensile simulations can be found in
Figures 10 and 11, respectively. Both the 25% and 45% volume fraction’s actual and predicted
stress–strain data are presented on the same plot. Here, both the longitudinal and transverse uniaxial
tensile tests are exhibiting the same type of behaviour. The 25% volume fraction simulations are
accurate for the majority of the simulation, and in particular at higher strains, but there is minor
overprediction prior to 0.5% strain, with a maximum error of 2.3 MPa and 4.3 MPa for the longitudinal
and transverse directions, respectively. The same result is seen in both the 45% volume fraction
simulations, although there is a larger overprediction at the low strain region of the curve (maximum
error of 5.7 MPa and 9 MPa for the longitudinal and transverse directions, respectively) compared
J. Compos. Sci. 2018, 2, 53 13 of 24

to the 25% volume fraction simulations, and the curves converge between 0.6% and 0.7% strain.
The longitudinal and transverse 25% volume fraction simulations had root mean square (RMS) errors
of 0.99 MPa and 1.97 MPa, respectively, while the longitudinal and transverse 45% volume fraction
simulations
J.
J.Compos.
Compos.Sci. had2,2,RMS
Sci.2018,
2018, errors
xx FOR
FOR PEER of 2.98 MPa and 4.00 MPa, respectively.
PEER REVIEW
REVIEW 13
13 of
of24
24

Figure
Figure 10.
Figure 10. Uniaxial
10. Uniaxial tension
Uniaxial tension experimental
tension experimental and
experimental and simulated
and simulated stress–strain
simulated stress–strain curves
stress–strain curves for
curves the
for the
for longitudinal
the longitudinal direction.
longitudinaldirection.
direction.

Figure
Figure
Figure 11.
11. Uniaxialtension
11.Uniaxial
Uniaxial tensionexperimental
tension experimentaland
experimental andsimulated
and simulated stress–strain
stress–straincurves
simulated stress–strain curvesfor
curves forthe
for thetransverse
the transversedirection.
transverse direction.
direction.

An
Anexplanation
explanationforforthe
theseparation
separationat atlow
lowstrains
strainscan
canbe
bethat
thatthe
thematerial
materialmodel
modelused
usedis isaaplasticity
plasticity
model,
model, which
which assumes
assumes that
that there
there is
is aa clear
clear and
and sudden
sudden elastic–plastic
elastic–plastic transition,
transition, but
but that
that isis not
not the
the
case
case for
for this
this SMC
SMC composite.
composite. TheThe transition
transition from
from elastic
elastic to
to plastic
plastic for
for this
this material,
material, similar
similar toto
thermosets,
thermosets, is is gradual
gradual and
and cannot
cannot bebe defined
defined by
by aa single
single parameter.
parameter. The
The plastic
plastic model
model also
also assumes
assumes
that
that the
the work
work hardening
hardening is
is due
due to
to slip,
slip, while
while the
the SMC
SMC material
material deviates
deviates from
from pure
pure elastic
elastic due
due to
to
J. Compos. Sci. 2018, 2, 53 14 of 24
J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 14 of 24

separation seen in the

An explanation forelastic portion of
the separation atthose simulations.
low strains can be that Calibration
the material of the models
model usedwasis aiteratively
plasticity
conducted
model, which to determine
assumes that thethere
inputismaterial properties
a clear and suddenso that the hightransition,
elastic–plastic strain regionbutof theiscurve
that not the would
case
be more accurate, since this is the part of the curve that is of the most interest.
for this SMC composite. The transition from elastic to plastic for this material, similar to thermosets,
The experimental
is gradual and cannot beand simulated
defined valuesparameter.
by a single for ultimate Thetensile
plasticstrength
model also (UTS) and fracture
assumes that thestrain
work
are presented
hardening in Table
is due to slip,2.while
Fromthe these
SMCresults,
material it can be seen
deviates from that theelastic
pure fracture
duestrains from alland
to debonding of the
simulations matched
breaking of fibres the experimental
causing results well,
damage that weakens and the ultimate
the material. Additionally,tensile strength
the also matched
elastic properties are
well to within
greater 2% error.
in the 45% volume fraction models, explaining why there is a greater separation seen in the
elastic portion of those simulations. Calibration of the models was iteratively conducted to determine
Table 2.properties
the input material Tensile Model Results.
so that the UTS:
high ultimate tensileofstrength,
strain region the curveRMS: root mean
would be moresquare.
accurate, since
thisVolume
is the part of the
UTS (MPa) curve that
UTS (MPa) is of the most interest.𝜺Fracture 𝜺Fracture RMS Error
The experimental and simulated % values
Difference
for ultimate tensile strength % Difference
(UTS) and fracture strain
Fraction. Experiential Simulation Experimental Simulation (MPa)
are presented
L25% in Table
53.7 2. From
54.3 these results,
1.10% it can be seen
1.10% that the
1.10% fracture strains
0% from all
0.99 the
of
simulations
L45% matched
89 the experimental
90.7 results
1.90% well, and the
1.25% ultimate tensile
1.25% strength
0% also matched2.98 well
to within
T25% 2% error.72.4 71.7 1.00% 1.30% 1.30% 0% 1.97
T45% 131.7 131 0.50% 1.50% 1.50% 0% 4
Table 2. Tensile Model Results. UTS: ultimate tensile strength, RMS: root mean square.

4.2.Volume
Mesh Sensitivity
UTS (MPa) UTS (MPa) εFracture εFracture RMS Error
% Difference % Difference
Fraction Experiential Simulation Experimental Simulation (MPa)
Mesh sensitivity analysis was performed on the simulation in order to eliminate the effects of
L25% 53.7
mesh size on the
L45% 89
computed 54.3
90.7
1.10%
results. Figure
1.90%
1.10%
12 shows1.25% 1.10%
the three mesh
1.25%
0%
sizes tested
0%
0.99
in this work.
2.98
The
stress–strain
T25% curves
72.4 were found
71.7 not to be sensitive to1.30%
1.00% mesh size, 1.30%
but the failure0%strain was.1.97
Table 3
T45% the comparison
displays 131.7 of 131 0.50% strength1.50%
ultimate tensile and failure1.50% 0% mesh tested.
strain for each 4 The

chosen mesh was determined according to failure strain’s mesh-dependency [73].

4.2. Mesh Sensitivity
Table 3. Mesh Variability Results.
Mesh sensitivity analysis was performed on the simulation in order to eliminate the effects
of mesh size on the computed results. Figure UTS
Mesh Size (MPa)the𝜺three
12 shows Fracture mesh sizes tested in this work.

The stress–strain curves were found not Mesh

Rough to be sensitive
54.9to mesh1.13%
size, but the failure strain was. Table 3
displays the comparison of ultimate tensile
Model strength and
Mesh 54.4failure 1.10%
strain for each mesh tested. The chosen
mesh was determined according to failure
Fine Meshstrain’s mesh-dependency
54.4 1.10% [73].

Figure 12. (A)

Figure 12. (A) Rough
Rough Mesh;
Mesh; (B)
(B) Model
Model Mesh;
Mesh; (C)
(C) Fine Mesh.
Fine Mesh.
J. Compos. Sci. 2018, 2, 53 15 of 24

Table 3. Mesh Variability Results.

Mesh Size UTS (MPa) εFracture

Rough Mesh 54.9 1.13%
Model Mesh 54.4 1.10%
Fine Mesh
J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 54.4 1.10% 15 of 24

4.3.
4.3. Overview
Overview
This
This model
model makes
makes use
use of
of experimental
experimental tensile
tensile results
results of
of specimens
specimens that
that were
were cut
cut from SMC car
from SMC car
hoods in two different material orientations: longitudinal and transverse. A series of specimens
hoods in two different material orientations: longitudinal and transverse. A series of specimens were were
tested
tested for each material
for each material orientation;
orientation; those
those results
results were
were then
then averaged
averaged and
and used
used to produce aa finite
to produce finite
element model. A finite model was created for each of the material orientations, so that
element model. A finite model was created for each of the material orientations, so that in future in future
works, an
works, an anisoptropic
anisoptropic model
model can be applied
can be applied toto the
the entire
entire car
car hoods,
hoods, while
while using
using these
these models
models asas
inputs for the
inputs for the longitudinal
longitudinal and
and transverse
transverse orientations.
orientations.

5. Model Validation and Further Discussion

Discussion

5.1. Flexural
5.1. Flexural Model
Model Setup
Setup
To validate
To the simulation
validate the simulationframework
frameworkpresented
presentedearlier,
earlier,each
eachuniaxial
uniaxial tensile
tensile model
model waswas used
used to
to simulate the associated material response in three-point bending tests. The three-point
simulate the associated material response in three-point bending tests. The three-point bending bending
simulations used
simulations used cylindrical
cylindrical supports
supports and
and aa loading
loading nose
nose meshed
meshed as as rigid
rigid parts
parts to
to govern
govern the
the bend
bend
deformation, rather than the nodal constraints used in the tension simulations. The model was
deformation, rather than the nodal constraints used in the tension simulations. The model was
assembled in
assembled in an
an Autodesk
Autodesk Inventor
Inventor prior
prior to
to being
being imported
imported into
into LS-DYNA
LS-DYNA PrePost
PrePost (Livermore
(Livermore
Software Technology
Software Technology Corp.,
Corp., Livermore,
Livermore, CA,
CA, USA)
USA) and
and meshed
meshed with
with solid
solid elements.
elements. The
The same
same mesh
mesh
size was used as the ones in the uniaxial tensile simulations. The meshed flexural model can be seen
size was used as the ones in the uniaxial tensile simulations. The meshed flexural model can be seen
in Figure
in Figure 13.
13.

Figure
Figure 13.
13. Three-point
Three-point bending
bending meshed model.
meshed model.

The supports and loading nose were assigned the same mesh size as the flexural specimen to
The supports and loading nose were assigned the same mesh size as the flexural specimen to
ensure smooth and more stable simulations [74]. Three surface-to-surface contacts were defined: one
ensure smooth and more stable simulations [74]. Three surface-to-surface contacts were defined: one
between the loading nose and the flexural specimen, and one between each of the supports and the
between the loading nose and the flexural specimen, and one between each of the supports and the
flexural specimen. The friction coefficients at each contact are summarised in Table 4, as determined
flexural specimen. The friction coefficients at each contact are summarised in Table 4, as determined
from the literature [71]. The flexural specimen was not constrained beyond contact with the supports
and the load applied by the nose. Each of the supports was fixed, while the loading nose had
translational freedom in the y-direction only.

Table 4. Friction Coefficients.

J. Compos. Sci. 2018, 2, 53 16 of 24

from the literature [71]. The flexural specimen was not constrained beyond contact with the supports
and the load applied by the nose. Each of the supports was fixed, while the loading nose had
translational freedom in the y-direction only.

Table 4. Friction Coefficients.

Flexural Specimen Static Friction Coefficient Dynamic Friction Coefficient

Loading Nose/Specimen 0.3 0.2
Supports/Specimen 0.3 0.2

Although the objective of simulating the flexural response of the material with the previously
developed tensile model was validation, some modifications for the definitions of contact still needed
to be made. First, the previous model did not have any contact; therefore, both contact and global
dampening needed to be added to the model. Second, a failure criterion based on failure strain was
added to the tensile model to accommodate failure. The two types of tests (uniaxial tensile and flexural)
activate failure mechanisms at different strain levels. In addition, the presence of global damping
adds computational stiffness to the model, which affects the onset of failure. This can be understood
because of the more complex state of strain in the flexural case, and the local behaviour of strain
in the matrix [75]. Therefore, the failure strain was recalibrated for the flexural model. The failure
strain values for both the tensile and flexural model can be found in Table 5. From this table, it can
be seen that the 25% volume fraction models had a greater difference between the failure strains
in the tensile model versus the flexural model. This was because the 45% volume fraction models
were found to have more vibration noise at the beginning of the simulation and needed additional
damping to smoothen the initial contact. Furthermore, there was some sliding that occurred between
the supports and the specimen. This frictional force needed dampening as well to smoothen the noise
that aroused from the contact and sliding, and also to capture the real bending phenomenon during
simulation [76]. This can be confirmed from Figure 4, where the specimen had gradually been sliding
along the supports throughout the flexural tests.

Table 5. Failure Strain for the Tensile and Flexural Models.

Volume Fraction Tensile Flexural

L25% 0.98% 1.95%
L45% 1.25% 1.65%
T25% 1.13% 1.63%
T45% 1.34% 1.34%

5.2. Flexural Test Simulations

The results for the flexural longitudinal and transverse simulations are shown in Figures 14 and 15,
respectively. The longitudinal result for the 25% volume fraction matched the experimental data the
best, having minor overprediction prior to 15 mm of displacement, (maximum error of 4.1 N) and
converging thereafter. The transverse 25% volume fraction simulation results were not as accurate as
the longitudinal results, having a maximum error of 5.7 N happening in the beginning of the flexural
test, where the contact condition is not in its steady condition. However, the flexural simulations
were very accurate in terms of predicting the failure displacement and force. The curves did converge
once at approximately 12 mm of displacement, and did not converge again until fracture. There was
a similar result seen in both of the 45% volume fraction simulations, which were much less accurate
than those seen in the 25% volume fraction simulations. The 45% volume fractions had a larger
separation at the beginning of the force displacement curve, having a maximum error of 17.6 N and
19.1 N for the longitudinal and transverse directions, respectively. It is clear that the 45% volume
fraction simulations experienced a much higher tangent modulus. The 45% volume fraction curves do
J. Compos. Sci. 2018, 2, 53 17 of 24

later converge once, and the resulting failure displacement is accurate, but the failure force experienced
an increased error compared to that seen in other simulations, differing by 4.2% and 6.6% for the
longitudinal and transverse models, respectively. The RMS error was calculated over several points
through the curves, and was found to be 1.53 N and 3.39 N for the 25% volume fraction longitudinal
and transverse directions, respectively, while the RMS error for the 45% volume fraction simulations
was 6.22 N and 6.74 N for the longitudinal and transverse directions, respectively.
Figure 16 displays a fractured flexural specimen from the longitudinal 25% volume fraction
simulation. This figure captures the specimen just as it begins to fracture in the centre of the piece (as
seen in Figure 4), while the colour scale displays that the part failed at an equivalent Von Mises stress
of approximately 462,MPa,
J. Compos. Sci. 2018, which
x FOR PEER translates to a failure force of approximately 40 N.
REVIEW 17 of 24
J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 17 of 24

Figure 14. Three-point bending experimental and simulated stress–strain curves for the longitudinal direction.
14. Three-point
FigureFigure bending
14. Three-point experimental
bending experimentaland
andsimulated stress–straincurves
simulated stress–strain curvesforfor
thethe longitudinal
longitudinal direction.
direction.

Figure 15. Three-point bending experimental and simulated stress–strain curves for the transverse direction.
Figure 15. Three-point bending experimental and simulated stress–strain curves for the transverse direction.
Figure 15. Three-point bending experimental and simulated stress–strain curves for the transverse direction.
J. Compos. Sci. 2018, 2, 53 18 of 24
J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 18 of 24

Figure
Figure 16.
16. Simulated
Simulated fractured
fractured longitudinal 25% volume
longitudinal 25% volume fraction
fraction specimen.
specimen.

The flexural simulation and experimental results are presented in Table 6. The displacement at
The flexural simulation and experimental results are presented in Table 6. The displacement
failure was mostly accurate in all of the simulation results, differing between the simulation and
at failure was mostly accurate in all of the simulation results, differing between the simulation and
experimental value by 2.5% or less in all of the simulations. The failure force was the less accurate
experimental value by 2.5% or less in all of the simulations. The failure force was the less accurate
parameter, but the results were positive still. The longitudinal failure force results differed by 1.5%
parameter, but the results were positive still. The longitudinal failure force results differed by 1.5% and
and 4.2% for the 25% and 45% volume fraction models, respectively, while the transverse results
4.2% for the 25% and 45% volume fraction models, respectively, while the transverse results differed
differed by 3.5% and 6.6% for the 25% and 45% volume fraction models, respectively. This model
by 3.5% and 6.6% for the 25% and 45% volume fraction models, respectively. This model demonstrates
demonstrates accuracy similar to those of the other predictive models that have been used to predict
accuracy similar to those of the other predictive models that have been used to predict composite
composite material behaviour [77–79].
material behaviour [77–79].
Table 6. Flexural Model Results.
Table 6. Flexural Model Results.
Volume Failure Force Failure Force Displacement Displacement RMS Error
Volume
Fraction. Failure Force Failure Force % Difference Displacement Displacement % Difference RMS Error
(N) Experiential (N) Simulation % Difference (mm) Experimental (mm) Simulation % Difference (N)
Fraction (N) Experiential (N) Simulation (mm) Experimental (mm) Simulation (N)
L25%
L25% 40.1
40.1 39.5
39.5 1.50%
1.50% 30
30 30
30 0%
0% 1.53
1.53
L45% 87.1 83.4 4.20% 26 26.3 1.20% 6.22
L45% 87.1 83.4 4.20% 26 26.3 1.20% 6.22
T25% 59.5 57.4 3.50% 30 30.7 2.30% 3.39
T25%
T45% 59.5
105.7 57.4
98.7 3.50%
6.60% 30
24 30.7
24.6 2.30%
2.50% 3.39
6.74
T45% 105.7 98.7 6.60% 24 24.6 2.50% 6.74
5.3. Remarks on the Accuracy of the Results
5.3. Remarks on the Accuracy of the Results
From the results above, it is clear that the model becomes less accurate in simulations with higher
forces.From Thethe results above,
simulations it is clear
at higher that
forces the
are model
also those becomes
that havelesshigher
accurate in simulations
volume with higher
fraction percentages,
forces.
and with The simulations
higher volumeatpercentages,
higher forces are is
there also those that
a greater have
chance ofhigher volume
fibre-related fraction
failure percentages,
modes to occur:
and with higher volume percentages, there is a greater chance of fibre-related
either fibre bundling and kinking, or fibre rupture. These failure modes can cause localised failure, failure modes to occur:
either fibre
which leadsbundling
to the partand kinking,
failing or fibre
before rupture.
it reaches These failure
its ultimate tensile modes
strain,can cause the
making localised failure,
properties of
which leads
higher to the part failing
fibre composites before it reaches
less predictable its ultimate
[71]. Another tensile
potential strain,
source of making the properties
error is that the bending of
higher fibre composites less predictable [71]. Another potential source of
model was generated from a tensile finite model, which means that the tensile properties of the materialerror is that the bending
model
are was generated
accounted from
for, but the a tensile finite
compression model,is which
behaviour means that is
not. Compression the tensile properties
a significant portion of the
of the
material are accounted for, but the compression behaviour is not. Compression
three-point bending test, as the failure stress could come from the top or bottom of the specimen [70]. is a significant portion
ofthe
If thefracture
three-point bending
happens test,
at the as theoffailure
bottom stress could
the flexural comethen
specimen, from thethe top orproperties
tensile bottom ofgovern
the specimen
failure,
[70].ifIfitthe
but fracture
fractures onhappens at the bottom
the top section of thecompression
of the part, flexural specimen,
would thenbe thetheinstigator
tensile properties
of failure.govern
failure,
Thebut if it fractures
main source ofon the top section
inaccuracy of the part, compression
seen throughout these simulationswouldisbethe thefirst
instigator
half ofofthefailure.
force
The maincurves
displacement sourceinofthe
inaccuracy
45% volume seenfraction
throughout these simulations
simulations. While thereiswas the separation
first half ofseen
the atforce
the
displacement
beginning curves
of every in the
curve, the45% volume seen
separation fraction
in thesimulations.
45% volume While there
fraction was separation
simulations seen at the
was significantly
beginning
greater. of every
From Figure curve,
8, it canthe separation
clearly be seenseen that in theare
there 45%many volume fraction
different simulations seen
fibre orientations was
significantly greater. From Figure 8, it can clearly be seen that there are many different fibre
orientations seen throughout the material, and as the specimen is loaded, these fibres begin to
J. Compos. Sci. 2018, 2, 53 19 of 24

throughout the material, and as the specimen is loaded, these fibres begin to distribute the load
amongst each other. In the experiments, these fibres would not all take the same load [70], and they
would not all be in the most favourable loading direction [67], but the simulation cannot take this into
account. It was assumed that the fibres would all be taking the same load throughout the simulation.
However, this is not the case in reality, and for this reason, there is a large separation at the beginning
of the curve. It can be seen that as the material deforms, the simulation converges towards the
experimental data curve. This is because the material input given to the model would overpredict the
force at the beginning of the simulation due to the uneven loading seen at the initial stages. The fibres
and matrix fail during material deformation [11], where this weakens the composite. It can be seen
that the model takes this failure into consideration as the simulation curves quickly decline and begin
to match the experimental data. However, it is the model’s inability to account for the lack of fibres
loaded at the beginning of the simulations [70] causing the inaccuracy.
Overall, the model was successful, being able to predict the deformation of the 25% volume
fraction specimens with at most a 3.5% error (in reference to failure strain and failure force), and the
45% volume fraction specimens with at most a 6.6% error. A micromechanics model as developed
earlier for long fibre-reinforced composites by Sabiston et al. [43,45] based on the random characteristics
of the SMC composites can help to capture these errors mainly for the longitudinal and transverse 45%
volume fraction samples.

6. Conclusions
Multiple series of both uniaxial tensile and flexural bending tests were performed on SMC
composite specimens from an SMC car hood to quantify the mechanical characteristics of the material.
The study produced a collection of results for the 25% and 45% volume fraction samples for both
the longitudinal and transverse directions, where both the tensile and flexural tests were performed
at a strain rate of 3 × 10−3 s−1 . Scanning electron microscopy was performed to study the fracture
surface and aid in determining the primary failure modes of the composite. These results were used
to determine the mechanical properties for the material, so that a finite element model could be
developed. The tensile results were used to develop and calibrate the finite element model to describe
the deformation of the SMC composite. The flexural results were then used to validate the model and
predict the bending behaviour of these composites. The following conclusions can be drawn from the
current work based on the experiments and simulations performed in this study:

• Throughout the tensile tests, the higher volume fractions led to a higher UTS and failure strain,
while the transverse direction was stronger than the longitudinal direction. The flexural test saw
similar results, with the higher volume fraction specimens reaching a higher failure force, but the
failure displacement was less than that seen in the lower volume fraction tests; however, again,
the transverse direction proved to be stronger than the longitudinal direction.
• The SEM images taken lead to the conclusion that fibre bundling leading to matrix debonding is
the primary failure mode of fracture in both the 25% and 45% volume fraction specimens, for the
tensile and flexural tests.
• A linear piecewise plasticity material model was able to simulate very well the tensile stress–strain
data for calibration, and provide good results. The model was then validated by simulating the
three-point bending tests along with the experimental flexural data. The lower volume fraction
simulation was more accurate than the higher volume fraction simulation, with a maximum error
of 3.5% for the 25% volume fraction tests, and 6.6% for the 45% volume fraction tests. The reason
that the 45% volume fraction simulations are less accurate is that the model cannot account for
the uneven loading of fibres experienced at the beginning of the flexural tests. This is an area that
should be investigated further in future work using micromechanics-based models.

Author Contributions: Investigation: J.K.; Validation: E.C., Supervision: M.M.

J. Compos. Sci. 2018, 2, 53 20 of 24

Acknowledgments:
J.J.Compos.
Compos.Sci. 2018,2,2,xThe
Sci.2018, xFOR
FORAuthors would like to thank Natural Sciences and Engineering Research Council
PEERREVIEW
PEER REVIEW of
20ofof
20 24
24
Canada (NSERC) grant number RGPIN-2016-04221 and New Brunswick Innovation Foundation (NBIF) grant
number RAI 2016-108 for providing sufficient funding to execute this work. In addition, the authors would like
totothank
to
thankJeff
thank JeffWood
Jeff Wood
Wood(Western
(WesternUniversity)
University)and
(Western University)
andFraunhofer
FraunhoferProject
and Fraunhofer
ProjectCentre
Centrefor
Project Centre
forComposites
CompositesResearch
for Composites
ResearchininLondon,
Research in
London,
London,
Ontario
Ontariofor
Ontario for providing
forproviding
providing the
thethe composite
composite
composite car
car car hoods.
hoods.
hoods. Finally
Finally
Finally the authors
the authors
the authors would
wouldwould like
like thelike the thank technicians
thetechnicians
thank thank technicians Brian
Brian
Brian Guidry,
BobGuidry,
Guidry, Bob MacAllister,
Bob MacAllister,
MacAllister, and Vinceand and Vince Boardman,
Vince Boardman,
Boardman, without
without
without their their
helptheir help this
help this
this project project
project
would would
notwould not be possible.
not be possible.
be possible.
Conflicts
Conflicts
Conflicts of
ofInterest:
ofInterest:The
Interest:Theauthors
The declare
authors
authors no
declare
declare conflict
no
no ofofof
conflict
conflict interest.
interest.
interest.

Appendix
AppendixA.
Appendix A.Uniaxial
A. UniaxialTensile
Uniaxial TensileTests
Tensile Testswith
Tests withError
with ErrorBars
Error Bars
Bars

Figure
Figure A1.
Figure A1.
A1. Longitudinal
Longitudinal 25%
Longitudinal 25%
25%andand
45%
and 45%
45% 33×10
3× ×10−−3
103−3s−
ss−1−11Results
Resultswith
withStandard DeviationError
StandardDeviation
Deviation ErrorBars.
Error Bars.
Bars.

Figure A2. Transverse 25% and 3×

45% − 3−3s− 1 Results with Standard
Figure A2.
Figure Transverse
A2. 25%
Transverse and
25% 45%
and 45% 33×10
×10
10−3 ss−1−1 StandardDeviation
Results with Standard DeviationError
Deviation ErrorBars.
Error Bars.
Bars.
 J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 21 of 24
J. J. Compos.Sci.
Compos. Sci.2018,
2018,2,2,53x FOR PEER REVIEW 2121
ofof
2424

AppendixB.B.
Appendix
Appendix B.Flexural
FlexuralBending
Flexural BendingTests
Bending Testswith
Tests withError
with ErrorBars
Error Bars
Bars

FigureA3.
Figure
Figure A3.Longitudinal
A3. LongitudinalResults
Longitudinal Resultswith
Results withStandard
with StandardDeviation
Standard DeviationError
Deviation ErrorBars.
Error Bars.
Bars.

Figure
Figure A4.Transverse
FigureA4.
A4. TransverseResults
Transverse Resultswith
Results withStandard
with StandardDeviation
Standard DeviationError
Deviation Error
Error
Bars.
Bars.
Bars.

References
References
References
1. Jayasree,
1.1. Jayasree,N.A.;
Jayasree, N.A.;Airale,
N.A.; Airale,A.G.;
Airale, A.G.;Ferraris,
A.G.; Ferraris,A.;
Ferraris, A.;Messana,
A.; Messana,A.;
Messana, A.;Sisca,
A.; Sisca,L.;L.;
Sisca, L.;Carello,
Carello,M.
Carello, M.Process
M. Processanalysis
Process analysisfor
analysis forstructural
for structural
structural
optimisation
optimisation
optimisation of
of of thermoplastic
thermoplastic
thermoplastic composite
composite
composite component
component
componentusingusing the
the building
using building
the building block
blockblock approach.
approach. Compos.
approach. Compos.
Compos. Part B
Part B 2017,
Part B
2017,
126, 126, 119–132.
119–132.
2017, 126, 119–132.
[CrossRef]
2. Friedrich,
2.2. Friedrich,
Friedrich,K.; K.; Almajid,A.A.
K.;Almajid,
Almajid, A.A.Manufacturing
A.A. Manufacturingaspects
Manufacturing aspectsofof
aspects ofadvanced
advancedpolymer
advanced polymercomposites
polymer compositesfor
composites forautomotive
for automotive
automotive
applications.
applications.Appl.
applications. Appl. Compos.
Appl.Compos. Mater.
Compos.Mater. 2013,
2013,20,
Mater.2013, 20, 107–128.
20,107–128.
107–128.[CrossRef]
3. Liu,
3.3. Liu,P.F.;
Liu, P.F.;
P.F.; Zheng,
Zheng,
Zheng, J.Y.
J.Y.
J.Y. Recent
Recent
Recent developments
developments
developments on
onon damage
damage
damage modeling
modeling
modeling and
and
and finite
finite element
element
finite element analysisfor
analysis
analysis forcomposite
for composite
composite
laminates:AA
laminates:
laminates: Areview.
review.Mater.
review. Mater.Des.
Mater. Des.2010,
Des. 2010,31,
2010, 31,3825–3834.
31, 3825–3834.[CrossRef]
3825–3834.
4. Lemaitre,
4.4. Lemaitre, J.; Chaboche, J.L. Mechanics of Solid
Lemaitre, J.; Chaboche, J.L. Mechanics of SolidMaterials;
J.; Chaboche, J.L. Mechanics of Solid Materials;Cambridge
Materials; CambridgeUnitversity
Cambridge UnitversityPress:
Unitversity Press:Cambridge,
Press: Cambridge,UK,
Cambridge, UK,1990.
UK, 1990.
1990.
5. Ristinmaa,
5.5. Ristinmaa,
Ristinmaa,M.; M.; Ottosen,
M.;Ottosen,
Ottosen, N.S.N.S. Viscoplasticity based
Viscoplasticity based
N.S. Viscoplasticity on
basedon a additive
onaaadditive split
additivesplit of
split the
of of conjugated
thethe forces.
conjugated
conjugated Eur. J.
Eur.Eur.
forces.
forces. Mech.
J.
J. Mech.
A/Solids
Mech. 1998,1998,
A/Solids
A/Solids 1998, 17, 207–235.
17, 207–235.
17, 207–235. [CrossRef]
J. Compos. Sci. 2018, 2, 53 22 of 24

6. Schapery, R.A. A theory of mechanical behaviour of elastic media with growing damage and other changes
in structure. J. Mech. Phys. Solids 1990, 38, 215–253. [CrossRef]
7. Murakami, S.; Kamiya, K. Constitutive and damage evolution equations of elastic-brittle materials based on
irreversible thermodynamics. Int. J. Mech. Sci. 1997, 39, 473–486. [CrossRef]
8. Hayakawa, K.; Murakami, S.; Liu, Y. An irreversible thermodynamics theory for elastic-plastic-damage
materials. Eur. J. Mech. A/Solids 1998, 17, 13–32. [CrossRef]
9. Qing, H.; Liu, T. Micromechanical analysis of SiC/Al metal matrix composites: Finite element modeling and
damage simulation. Int. J. Appl. Mech. 2015, 7, 1550023. [CrossRef]
10. Olsson, M.; Ristinmaa, M. Damage Evolution in Elasto-plastic Materials–Material Response Due to Different
Concepts. Int. J. Damage Mech. 2003, 12, 115–140. [CrossRef]
11. Matzenmiller, A.; Lubliner, J.; Taylor, R.L. A constitutive model for anisotropic damage in fiber-composites.
Mech. Mater. 1995, 20, 125–152. [CrossRef]
12. Maa, R.H.; Cheng, J.H. A CDM-based failure model for predicting strength of notched composite laminates.
Compos. Part B 2002, 33, 479–489. [CrossRef]
13. Camanho, P.P.; Maimi, P.; Davila, C.G. Prediction of size effects in notched laminates using continuum
damage mechanics. Compos. Sci. Technol. 2007, 67, 2715–2727. [CrossRef]
14. Perreux, D.; Robinet, P.; Chapelle, D. The effect of internal stress on the identification of the mechanical
behaviour of composite pipes. Compos. Part A 2006, 37, 630–635. [CrossRef]
15. Liu, P.F.; Zheng, J.Y. Progressive failure analysis of carbon fiber/epoxy composite laminates using continuum
damage mechanics. Mater. Sci. Eng. A 2008, 485, 711–717. [CrossRef]
16. Lin, W.P.; Hu, H.T. Nonlinear Analysis of Fiber-Reinforced Composite Laminates Subjected to Uniaxial
Tensile Load. J. Compos. Mater. 2002, 36, 1429–1451. [CrossRef]
17. Tsai, S.W.; Wu, E.M. A General Theory of Strength for Anisotropic Materials. J. Compos. Mater. 1971, 5, 58–81.
[CrossRef]
18. Hashin, Z. Failure criteria for unidirectional fiber composites. J. Appl. Mech. 1980, 47, 329–363. [CrossRef]
19. Hoffman, O. The Brittle Strength of Orthotropic Materials. J. Compos. Mater. 1967, 1, 200–207. [CrossRef]
20. Yamada, S.E.; Sun, C.T. Analysis of Laminate Strength and its Distribution. J. Compos. Mater. 1978, 12,
275–285. [CrossRef]
21. Padhi, G.S.; Shenoi, R.A.; Moy, S.S.J.; Hawkins, G.L. Progressive failure and ultimate collapse of laminated
composite plates in bending. Compos. Struct. 1998, 40, 277–291. [CrossRef]
22. Tay, T.E.; Tan, S.; Tan, V.; Gosse, J.H. Damage progression by the finite element-failure method (EFM) and
strain invariant failure theory (SIFT). Compos. Sci. Technol. 2005, 65, 935–944. [CrossRef]
23. Tay, T.E.; Liu, G.; Yudhanto, A.; Tan, V.B.C. A Micro-Macro Approach to Modeling Progressive Damage in
Composite Structures. Int. J. Damage Mech. 2008, 17, 5–29. [CrossRef]
24. Li, R.; Kelly, D. Application of a First Invariant Strain Criterion for Matrix Failure in Composite Materials.
J. Compos. Mater. 2003, 37, 1977–2001. [CrossRef]
25. Mayes, J.S.; Hansen, A.C. Composite laminate failure analysis using multicontinuum theory.
Compos. Sci. Technol. 2004, 64, 379–394. [CrossRef]
26. Mayes, J.S.; Hansen, A.C. A comparison of multicontinuum theory based failure simulation with
experimental results. Compos. Sci. Technol. 2004, 64, 517–527. [CrossRef]
27. Zhu, H.; Sankar, B.V.; Marrey, R.V. Evaluation of Failure Criteria for Fiber Composites Using Finite Element
Micromechanics. J. Compos. Mater. 1998, 32, 766–783. [CrossRef]
28. Ha, S.K.; Jin, K.K.; Huang, Y. Micro-Mechanics of Failure (MMF) for Continuous Fiber Reinforced Composites.
J. Compos. Mater. 2008, 42, 1873–1896.
29. Rybicki, E.F.; Kanninen, M.F. A Finite Element Calculation of Stress Intensity Factors by a Modified Crack
Closure Integral. Eng. Fract. Mech. 1977, 9, 931–938. [CrossRef]
30. Rybicki, E.F.; Schmueser, D.W.; Fox, J. An Energy Release Rate Approach for Stable Crack Growth in the
Free-Edge Delamination Problem. J. Compos. Mater. 1977, 11, 470–488. [CrossRef]
31. Irwin, G.R. Fracture dynaimcs, fracturing of metals. Am. Soc. Met. 1948, 8, 147–203.
32. Rice, J.R. A path independant integral and the approximate analysis of strain concentration by notches and
cracks. J. Appl. Mech. 1968, 2, 379–465. [CrossRef]
33. Krueger, R. Virtual crack closure technique: History, approach, and applications. Appl. Mech. Rev. 2004, 57,
109–144. [CrossRef]
J. Compos. Sci. 2018, 2, 53 23 of 24

34. Dugdale, D.S. Yielding of steel sheets containing slits. J. Mech. Phys. Solids 1960, 8, 100–104. [CrossRef]
35. Barenblatt, I. Mathematical thoery of equilibrium cracks in brittle fracture. Adv. Appl. Mech. 1962, 7, 55–129.
36. Hutchinson, J.W.; Suo, Z. Mixed mode cracking in layered materials. Adv. Appl. Mech. 1992, 29, 63–191.
37. Tvergaard, V.; Hutchinson, J.W. Effect of strain-dependent cohesive zone model on predictions of crack
growth resistance. Pergamon 1995, 33, 261–269. [CrossRef]
38. Allen, D.H.; Searcy, C.R. Numerical Aspects of a Micromechanical Model of a Cohesive Zone. J. Reinf.
Plast. Compos. 2000, 19, 240–249. [CrossRef]
39. Camanho, P.P.; Davila, C.G.; de Moura, M.F. Numerical Simulation of Mixed-mode Progressive Delamination
in Composite Materials. J. Compos. Mater. 2003, 37, 1415–1439. [CrossRef]
40. Xie, D.; Waas, A.M. Discrete cohesive zone model for mixed-mode fracture using finite element analysis.
Eng. Fract. Mech. 2006, 73, 1783–1796. [CrossRef]
41. Shin, C.S.; Wang, C.M. An Improved Cohesive Zone Model for Residual Notched Strength Prediction of
Composite Laminates with Different Orthotropic Lay-ups. J. Compos. Mater. 2004, 38, 713–737. [CrossRef]
42. Rami, H.A. Cohesive micromechanics: A new approach for progressive damage modeling in laminated
composites. Int. J. Damage Mech. 2009, 8, 691–719.
43. Sabiston, T.; Mohammadi, M.; Cherkaoui, M.; Levesque, J.; Inal, K. Micromechanics for a long fibre reinforced
composite model with a functionally graded interphase. Compos. Part B Eng. 2016, 84, 188–199. [CrossRef]
44. Mohammadi, M.; Saha, G.C.; Akbarzadeh, A.H. Elastic field in composite cylinders made of functionally
graded coatings. Int. J. Eng. Sci. 2016, 101, 156–170. [CrossRef]
45. Sabiston, T.; Mohammadi, M.; Cherkaoui, M.; Levesque, J.; Inal, K. Micromechanics based elasto-visco-plastic
response of long fibre composites using functionally graded interphases at quasi-static and moderate strain
rates. Compos. Part B Eng. 2016, 100, 31–43. [CrossRef]
46. Tay, T.E.; Liu, G.; Tan, V.B.; Sun, X.S.; Pham, D.C. Progressive Failure Analysis of Composites.
J. Compos. Mater. 2008, 42, 1921–1967. [CrossRef]
47. Gao, Y.F.; Bower, A.F. A simple technique for avoiding convergence problems in finite element simulations
of crack nucleation and growth on cohesive interfaces. Model. Simul. Mater. Sci. Eng. 2004, 12, 453–463.
[CrossRef]
48. Xu, X.P.; Needleman, A. Numerical Simulations of Fast Crack Growth in Brittle Solids. J. Mech. Phys. Solids
1994, 42, 1397–1434. [CrossRef]
49. Riks, E. An Incremental Approach to the Solution of Snapping and Buckling Problems. Int. J. Solid Struct.
1979, 15, 529–551. [CrossRef]
50. Ramm, E. Strategies for Tracing the Nonlinear Response Near Limit Points. Nonlinear Finite Elem. Anal.
Struct. Mater. 1981, 63–89. [CrossRef]
51. Crisfield, M.A. An arc-length method including line searches and accelerations. Int. J. Numer. Methods Eng.
1983, 19, 1269–1358. [CrossRef]
52. Bednarcyk, B.A.; Arnold, S.M. A Framework for Performing Multiscale Stochastic Progressive Failure Analysis of
Composite Structures; NASA Center for Aerospace Information: Hanover, MD, USA, 2007; pp. 1–16.
53. Van der Meer, F.P.; Sluys, L.J. Continuum Models for the Analysis of Progressive Failure in Composite
Laminates. J. Compos. Mater. 2009, 43, 2131–2156. [CrossRef]
54. Robbins, D.H., Jr.; Reddy, J.N. Adaptive Hierarchical Kintematics in Modeling Progress Damage and Global
Failure in Fiber-reinforced Composite Laminates. J. Compos. Mater. 2008, 42, 43–173. [CrossRef]
55. Belytschko, T.; Black, T. Elastic crack growth in finite elements with minimal remeshing. Int. J. Numer.
Methods Eng. 1999, 45, 601–621. [CrossRef]
56. Moes, N.; Dolbow, J.; Belytschko, T. A finite element method for crack growth without remeshing. Int. J.
Numer. Methods Eng. 1999, 46, 131–181. [CrossRef]
57. Moes, N.; Belytschko, T. Extended finite element method for cohesive crack growth. Eng. Fract. Mech. 2002,
69, 813–833. [CrossRef]
58. Dolbow, J.; Moes, N.; Belytschko, T. Discontinuous enrichment in finite element with a partition of unity
method. Finite Elem. Anal. Des. 2000, 36, 235–260. [CrossRef]
59. Guidault, P.A.; Allix, O.; Champaney, L.; Cornuault, C. A multiscale extended finite element method for
crack propagation. Comput. Methods Appl. Mech. Eng. 2008, 197, 381–399. [CrossRef]
60. Huynh, D.B.; Belytschko, T. The extended finite element method for fracture in composite materials. Int. J.
Numer. Methods Eng. 2008, 77, 214–239. [CrossRef]
J. Compos. Sci. 2018, 2, 53 24 of 24

61. Kelly, J.; Mohammadi, M. Uniaxial tensile behavior of sheet molded composite car hoods with different fibre
contents under quasi-static strain rates. Mech. Res. Commun. 2018, 87, 42–52. [CrossRef]
62. Western University. (2015) Thermoset Sheet Moulding Copound (SMC). Available online: http://www.eng.
uwo.ca/fraunhofer/pdf/FPC_Western_D-SMC_2015.pdf (accessed on 7 July 2018).
63. Standard Test Method for Tensile Properties of Plastics; ASTM D638-14; ASTM International: West Conshohocken,
PA, USA, 2014.
64. Standard Test Methods for Flexural Properties of Unreinforced and Reinforced Plastics and Electrical Insulating
Materials; ASTM D790-15; ASTM International: West Conshohocken, PA, USA, 2015.
65. Mohammadi, M.; Salimi, H.R. Failure analysis of a gas turbine marriage bolt. J. Fail. Anal. Prev. 2007, 7,
81–86. [CrossRef]
66. Fu, S.Y.; Lauke, B.; Mäder, E.; Yue, C.Y.; Hu, X. Tensile properties of short-glass-fiber-and
short-carbon-fiber-reinforced polypropylene composites. Compos. Part A 2000, 31, 1117–1125. [CrossRef]
67. Kuriger, R.J.; Alam, M.K.; Anderson, D.P. Strength prediction of partially aligned discontinuous
fiber-reinforced composites. J. Mater. Res. 2001, 16, 226–232. [CrossRef]
68. Lu, Y. Mechanical Properties of Random Discontinuous Fiber Composites Manufactured from Wetlay Process; Virginia
Tech: Blacksburg, Virginia, 2002.
69. Thomason, J.L.; Vlug, M.A. Influence of fibre length and concentration on the properties of glass
fibre-reinforced polypropylene: 1. Tensile and flexural modulus. Compos. Part A 1996, 27, 477–484. [CrossRef]
70. Baxter, W.J. The strength of metal matrix composites reinforced with randomly oriented discontinuous fibers.
Metall. Trans. A 1992, 23A, 3045–3053. [CrossRef]
71. Aboudi, J.; Arnold, S.M.; Bednarcyk, B.A. Micromechanics of Composite Materials; ELSEVIER: Oxford, UK, 2013.
72. Huberth, F.; Hiermaier, S.; Neumann, M. Material Models for Polymers under Crash Loads Existing LS-DYNA
Models and Perspective; Fraunhofer: Bamberg, Germany, 2005.
73. More, S.T.; Bindu, R.S. Effect of Mesh Size on Finite Element Analysis of Plate Structure. Int. J. Eng. Sci.
Innov. Technol. 2015, 4, 181–185.
74. Chandrupatla, T.R.; Belegundu, A.D. Introduction to Finite Elements in Engineering, 4th ed.; Pearson: London,
UK, 2012.
75. Aboudi, J.; Arnold, S.M.; Bednarcyk, B.A. Micromechanics of Composite Materials: A Generalized Multiscale
Analysis Approach; Butterworth-Heinemann: Oxford, UK, 2012.
76. Shin, D.K.; Kim, H.C.; Lee, J.J. Numerical analysis of the damage behavior of an aluminum/CFRP hybrid
beam under three point bending. Compos. Part B Eng. 2014, 56, 397–407. [CrossRef]
77. Pineda, E.; Waas, A.; Bednarcyk, B.; Collier, C.; Yarrington, P. Progressive damage and failure modeling in
notched laminated fiber reinforced composites. Int. J. Fract. 2009, 158, 125–143. [CrossRef]
78. Germain, N.; Besson, J.; Feyel, F. Composite layered materials: Anisotropic nonlocal damage models.
Comput. Methods Appl. Mech. Eng. 2007, 196, 4272–4282. [CrossRef]
79. Giner, E.; Sukumar, N.; Tarancon, J.E.; Fuenmayor, F.J. An Abaqus implementation of the extended finite
element method. Eng. Fract. Mech. 2009, 76, 347–368. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).