See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/304994146

Improvement in failure prediction algorithm for textile composites

Article  in  International Journal of Damage Mechanics · July 2016

DOI: 10.1177/1056789516658505

CITATIONS READS

2 474

2 authors:

Ali Hallal Rafic Younes

Lebanese International University Lebanese University
30 PUBLICATIONS   262 CITATIONS    148 PUBLICATIONS   1,323 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

AirStorage : Wind-Diesel Systems with Compressed Air Energy Storage View project

HyperDetect : Hyperspectral imaging detection. View project

All content following this page was uploaded by Rafic Younes on 24 October 2017.

The user has requested enhancement of the downloaded file.

Article
International Journal of Damage
Mechanics
0(0) 1–29
Improvement in failure prediction ! The Author(s) 2016
Reprints and permissions:
algorithm for textile composites sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1056789516658505
ijd.sagepub.com

Ali Hallal1 and Rafic Younes2

Abstract
This study presents an analytical modeling of the mechanical behavior of textile composites. The main
objective of this study is to evaluate in-plane and out of plane ultimate strengths for different types of 2D
and 3D fabric-reinforced polymer. The proposed analytical model consists of a homogenization method, a
failure criterion, and a damaged stiffness model. An improved failure prediction algorithm is presented,
which tends to enhance the prediction of ultimate strengths. It is shown that assuming a final failure of
undulated yarns for a failure of 90% of subdivisions rather than 100%, alongside assigning a multi-failure
mode for the failed subdivision rather than single failure mode is yielding much better results. The
predicted elastic properties and ultimate strengths are compared to available experimental and numerical
data, where a very good agreement is shown. The proposed model offers a reliable, simple, and easy to
use analytical tool.

Keywords
Composite material, ultimate strength, failure, homogenization, analytical model

Introduction
In recent years, the importance of textile composites has increased due to their physical, thermal,
and mechanical properties, alongside light weight (Ayranci and Carey, 2008; Hu, 2008; Kaw, 2006;
Mouritz et al., 1999; Tong et al., 2002). The necessity of using textile composites in advanced
industries imposes a modeling of their mechanical behavior, in order to provide reliable data for
designers and engineers. Nowadays it is considered as a great challenge for many researchers. The
evaluation of mechanical behavior of textile composites could be done using three methods: experi-
mental tests, analytical modeling, or numerical finite element (FE) modeling.
As known experimental tests are expensive and time costly. However, the implementation of
numerical models based on FE method offer a good solution but they consume a lot of time,

1
SDM Research Group, School of Engineering, Lebanese International University, Beirut, Lebanon
2
Faculty of Engineering, Third Branch, Lebanese University, Rafic Harriri Campus, Hadath, Beirut, Lebanon
Corresponding author:
Ali Hallal, Department of Mechanical Engineering, School of Engineering, Lebanese International University, Beirut, Lebanon.
Email: ali.hallal@liu.edu.lb
2 International Journal of Damage Mechanics 0(0)

beside the need of super computers. On the other hand, analytical models offer straightforward
solutions and are considered as more flexible tools. Moreover, they need less computation time
without the necessity to use super computers, as well as the ability of simple integration of analytical
models in a large optimization study. Concerning predicted results, more studies and investigations
are to be done to enhance the predictive capabilities of such methods (Hallal et al., 2013; Hinton and
Kaddour, 2012; Kaddour et al, 2013).
From a literature review, it is shown that a major work is done concerning the analytical model-
ing of ultimate strengths for unidirectional and laminated composites (Hinton and Kaddour, 2012;
Kaddour et al., 2013) Concerning textile fabrics (woven, braided and knitted), most of the works
focus on developing homogenization methods to determine the stiffness (elastic properties) of com-
posites. However, the final failure of textile composites, the failure modes, and the damaged stiffness
matrix are rarely discussed (Christiane, 2006; Li et al. 2009; Lomov et al., 2009; Scida et al., 1999)
This study tries to improve the failure prediction algorithm in order to enhance the predictive
capabilities of analytical modeling by dealing with three major problematic. An accurate evaluation
of the stress and strain fields throughout the unit cell, assigning a correct failure mode for failed
subdivisions-UD composites and the prediction of the composite final failure presents the main parts
of the proposed analytical modeling. A detailed discussion for the proposed algorithm is presented
in the second section.
A brief review for available analytical models is presented, while for more detailed review study,
the readers could refer to Ansar et al. (2011), Byun (1989), Crookston et al. (2005), Hallal et al.
(2013), and Tan et al. (1997). As known, the classical laminated theory (CLT) method widely used
for 2D laminated composites is yielding good predictions of in-plane stiffness and ultimate strengths
comparing with experimental results. Based on CLT method, three initial and basic models are
presented by Ishikawa and Chou (1982a, 1982b, 1983a, 1983b, 1983c): the mosaic model, the
fiber undulation model, and the bridging model. These models used CLT assuming iso-strain or
iso-stress conditions in assembling subdivided layers. Later, many attempts were done in order to
extend these models to be applied on new types of composites (Byun et al., 1991; Ramakrishna,
1997; Whitney and Chou, 1989; Yang et al., 1986). It is noticed that the application of such models
for textile fabrics with more complex architecture is very complicated. In addition, these models are
in general limited for the prediction of in-plane properties.
Otherwise to overcome these difficulties, models based only on an iso-strain assumption for
assembling yarns subdivisions are introduced. They need only a discretization of undulated yarns
into sub-volumes (subdivisions), with a calculation of volumes and orientations of fibers for each
sub-volume. Then, the calculation of stiffness matrix using the iso-strain assumption is done. The
iso-strain model, which provides all in-plane and out of plane elastic properties, proves relative
effectiveness in modeling some types of braided and woven composites (Chang et al., 2006; Cox and
Dadkhah, 1995; El Hage et al., 2009; Hahn and Pandey, 1994; Quek et al., 2003; Shokrieh and
Mazloomi, 2010) . Even to its simplicity, easy to use and applicability with almost all textiles
architectures, the main disadvantage of the model remains in the assumption of only an iso-strain
condition, which will lead to inaccurate prediction of ultimate strengths (El Hage et al.,
2009).Moreover, due to the lack of experimental data, the capability of iso-strain model in predict-
ing elastic properties in through thickness direction is still in question.
Known that iso-strain and iso-stress models yield the upper and lower bounds of elastic proper-
ties respectively, many studies tend to develop a model based on mixed iso-strain and iso-stress
assumptions. These models represent an important attempt to predict the mechanical behavior of
textile composites, which try to reflect the real stress and strain fields throughout the representative
elementary volume (REV). A simple and basic idea in combining iso-strain and iso-stress models
Hallal and Younes 3
was proposed by Kalidindi and Abusafieh (1996) and Kalidindi and Franco (1997) by introducing a
weighted factor. The main challenge remains in giving a significant physical meaning and generalized
form of such factor to be used with different kinds of architectures.
Other models try to find a geometrical representation of the REV, where layers and sub-volumes
are assembled in parallel or in series along the loading or assembling direction, according to their
relative positions. Two different approaches were noticed. The first needs a fully discretized REV
where sub-volumes are assembled along loading direction in series or in parallel, such as Naik and
Kuchibhotla (2002a), Naik et al. (2002b) and Naik and Ridevi (2002c) model. The second approach
(Bystrom et al., 2000; Jiang et al., 2000; Lee et al., 2003; Pochiraju and Chou, 1999a, 1999b; Ruan
and Chou, 1996; Sankar and Marrey, 1997; Tan et al., 1999a, 2000a, 2000b, 2001; Vandeurzen et al.,
1996a, 1996b) adopts the assemblage by components, i.e. sub-volumes of undulated parts of yarns
are assembled initially, and then, homogenized yarns with remaining pure matrix are assembled to
predict the stiffness of the REV.
Furthermore, in order to find a generalized model that could be employed for different textile
types, semi-numerical models based on inclusion methods and the method of cells are introduced.
An inclusion model based on Mori–Tanaka (M–T) method was presented (Gommers et al., 1998).
This model is supposed to describe the stress–strain fields in the REV better than the iso-strain
model. It yielded similar and better results for 2D woven, braided, and knitted composites
(Gommers et al., 1998). Another approach based on the method of cells was proposed
(Prodromou, 2004; Prodromou et al., 2011)This model proves to yield better results than the
M–T model (Prodromou et al., 2011).To apply such a model, a full 3D discretization of the REV
into smaller cells is required using a geometrical modeling software. It is shown in Prodromou (2004)
and Prodromou et al. (2011) that the required computational time is less than that with FE methods
but much greater than conventional analytical models, especially those based on iso-strain and iso-
stress assumptions.
Recently, the authors proposed a generalized analytical model, the three stages homogenization
method ‘‘3SHM,’’ in order to evaluate the elastic properties of 2D and 3D textile composites (Hallal
et al., 2013). That model was developed during many studies (Hallal et al., 2011b, 2012; Nehme
et al., 2011). The model has shown that analytical modeling of elastic properties is yielding good
results and could be applied for all kinds of 2D and 3D textile composites.
In this study, the main objective is to predict ultimate strengths and evaluate the stress–strain
diagrams for pure axial and shear loadings for different types of 2D and 3D composites. The
proposed model tends to present a reliable and flexible tool to be used with different types of textiles
even those having complex architecture as 3D angle interlock woven composites.

Modeling methodology
In this section, the general scheme of the proposed analytical model, which provides all 3D elastic
properties, ultimate strengths and a stress and strain diagrams for axial and shear loadings, is
presented. The proposed modeling scheme is composed from a homogenization method, a failure
criterion and a damaged stiffness model (Figure 1). A geometrical modeling relying on a sinusoidal
modeling of the undulated yarns is adopted. Chamis micromechanical model (Chamis, 1989; Chamis
et al., 2013) is used to determine the elastic properties and ultimate strengths of subdivisions-UD
composites. Then, an analytical homogenization method developed by the authors (Hallal et al.,
2013) is employed to predict the stiffness matrix of the composite and to evaluate the stress–strain
fields throughout the REV. The fourth component of the model is the failure criterion. Tsai–Wu
failure criterion will be adopted for UD composites subdivisions, while Christensen failure criterion
4 International Journal of Damage Mechanics 0(0)

•calculaon of volumes (REV, yarns, matrix)

geometrical •discrizaon of undulated yarns and calculaon of volumes of subdivisions
modeling

•calculaon of elasc properes of UD composites

micromecha-
-nical
•calcualon of the ulmate strengths of UD composites
modeling

•Evaluaon of the elasc properes of the composite

analycal
homogenizat •Evaluaon of the stress-strain fileds throughout the composite
-ion method

•Christensen failure critrion used to dtermine the failure of the isotropic polymeric matrix
Failure •Tsai-Wu failure critrion is used to dtermine the failure of UD composites
criteria

damaged •A sffness damage model taking itno account the different failure modes
sffenss
model

Figure 1. General scheme of analytical modeling of the mechanical behavior of composites.

is used for the pure matrix part. The fifth component used is a damaged stiffness model for sub-
divisions-UD composites (Figure 1).
It should be noticed that this study deals more specifically with the algorithm used to predict the
ultimate properties (Figure 2), since the prediction of composite stiffness is done previously by the
authors (Hallal et al., 2013). The proposed algorithm shown in Figure 2 tries to predict the failure of
the composite based on the following conditions:

(1) Final failure is assumed when all components (all yarns and pure matrix) are considered at
failure.
(2) Undulated yarns are considered at complete failure only if 90% of subdivisions-UD composites
are damaged.
(3) A multi-failure mode is assigned to the failed subdivisions-UD composites.
(4) The applied stress on the REV is increased only if no new failure is occurred.

In order to propose a reliable strength model, our methodology is based on solving three major
problematic. The first problematic is related with the multi scale homogenization method used. That
method proposed by the authors (Hallal et al., 2011b, 2012, 2013; Nehme et al., 2011b) has been used
in previous studies to evaluate only the elastic properties of textile composites, where it shows better
predictions to available experimental data compared to other analytical and numerical models.
However, in this study, the capabilities of that method in the evaluation of stress and strain fields
throughout the REV are crucial in order to yield good predictions of ultimate properties.
The second problematic, which is rarely discussed in other works, is related to the assumption of
final failure. When modeling a textile composite, a braided or a woven composite, the main attention is
given for modeling of undulated yarns. An accurate modeling of failure for undulated yarns will affect
greatly the final failure of the composite. In most works, the failure of undulated yarns is assumed to
be happened when all subdivisions-UD composites are at failure. There are no studies that deal with
Hallal and Younes 5

σ = σ + dσ
(for the first increment we take σ = 0)

3SHM to calculate the

composite sffness

3SHM to calculate the stress ans strains throughout the

composite

Failure criteria
(Tsai-Wu for UD and Christensen for the matrix)

Checking the failure of all yarns and matrix

if YES
if NO
σ is the ulmate strength

subdivisions-UD composite or matrix

no new failure
new failure

substuon of sffness
matrix for damaged part

Figure 2. General scheme of the evaluation of the ultimate strength under static loading of textile composites.

that assumption. Whether failure of all subdivisions-UD composites (100% of them) of an undu-
lated yarn or the failure of a defined number of subdivisions-UD composites (0 < percentage of
damaged subdivisions < 100%) will indicate the complete failure of the yarn? In this study, this
assumption is discussed where a new assumption of failure of undulated yarn (less than 100%)
is introduced. This assumption will show that it well enhance the predictive capability of the
proposed model.
The last problematic appears with the stiffness matrix assigned for a damaged subdivision, the
damaged stiffness matrix. In most works (Prodromou, 2004; Prodromou et al., 2011), the damaged
6 International Journal of Damage Mechanics 0(0)

stiffness matrix used corresponds to a chosen unique single failure mode. The first problem to be
discussed is how to choose that failure mode. Secondly, assigning a single failure mode, means
neglecting other failure modes that could appear, that will lead to a specific damaged stiffness
matrix that only corresponds to the chosen failure mode. When textile composites are subjected
to a loading, even uniaxial tensile loading, a complex stress and strain fields are subjected on each
subdivision-UD composite of undulated yarns. It means that different failure modes are affecting the
final failure of each subdivision. Consequently, neglecting some failure modes could well influence
the choice of the correct damaged stiffness matrix of the failed subdivision-UD composite. This will
greatly affect the evaluation of the ultimate strengths. In this study, a modified stiffness damaged
model based on assigning multi-failure modes are introduced and investigated. In other words,
a damaged stiffness matrix that takes into account all failure modes is proposed. A comparison
between analytical modeling based on a single failure mode and proposed multi-failure modes is
shown and discussed.
In order to validate the proposed model, the predicted results are compared to available experi-
mental data. It should be noticed that all experimental data used are found in the literature,
imposing more challenge to the proposed model. Thus, a comparison that relies on comparing
predicted results with different experimental tests done by different scientists in order will show
better the reliability of the proposed model, as making an exercise for the proposed model.

Modeling
Geometrical modeling
Defining the REV is the first step in the geometrical modeling for any analytical model (Figure 3).
The geometrical parameters, such as length, width, and height of the REV, also the height and width
of each yarn’s cross-sectional area are to be determined. These parameters are involved in the
calculation of volumes of yarns, layers, and REV, as well as other needed elements as the orientation
of yarn’s sub-volumes, since yarn volumes are calculated by multiplying the length of longitudinal
section of yarns by its transverse section area. The modeling of longitudinal profile, transverse cross-
sections, and the calculation of fiber volume fraction is given in (Appendix 1).

Figure 3. REV for a plain weave and 8H-Satin weave fabrics.

Hallal and Younes 7

Analytical homogenization method – The 3SHM model

In this section, the homogenization method that evaluates the elastic properties and composite
stiffness matrix is briefly presented. Moreover, this method is also used to predict the stress and
strain fields throughout the REV. The homogenization method was developed earlier in order to
predict the elastic properties of textile composites.
The used homogenization method, the 3SHM model (Hallal et al., 2013),consists of three stages
of homogenization levels: micro, meso, and macro homogenization stages. At first stage, the micro
homogenization stage, the stiffness matrices of sub-volumes are calculated. Then, at meso homogen-
ization stage, the stiffness matrices of yarns are evaluated by assembling sub-volumes using both iso-
strain and iso-stress methods. Finally, at macro stage, the stiffness of the REV is evaluated in terms
of homogenized yarns and the matrix stiffness matrices under an iso-strain assumption. The detailed
homogenization procedure is given in the Appendix 1.

Evaluation of the ultimate strengths of textile composites

Evaluation of the stress–strain fields throughout the composite. In order to apply a failure cri-
terion and a damaged stiffness model, the stress and strain fields for each subdivision-UD composite
should be calculated. It will be done using the 3SHM model. An inverse procedure of that used to
determine the elastic properties is applied. Let’s consider a textile composite consists of n yarns and a
polymeric matrix. Textile composites are characterized by undulated yarns which are discretized to
small subdivisions considered as UD composites. The generalized homogenization method, the
3SHM method, allows the calculation of stress and strain fields throughout the composite as follow:
The average stress tensor throughout the composite is given by
ZZZ
1
fsc g ¼ fsc g:dV ð1Þ
Vc

The Hook’s law gives

fsc g ¼ ½Cc :f"c g ð2Þ
where ½Cc  is the stiffness matrix of the composite
The iso-strain conditions gives
     
f"c g ¼ "y1 ¼ "y2 ¼ . . . ¼ "yn ¼ f"m g ð3Þ
     
where f"c g, "y1 , "y2 ,. . ., "yn and f"m g are the average strain in the composite and the n yarns
and matrix
The stress throughout the matrix is given by
fsm g ¼ ½Cm :f"m g ð4Þ
where ½Cm  is the stiffness matrix of the matrix
The stress throughout the n yarns is given by
    
syi ¼ Cyi : "yi ð5Þ
 
where Cyi is the stiffness matrix of the i-th yarn.
8 International Journal of Damage Mechanics 0(0)

Then, the stress throughout each subdivision of each yarn is calculated as follow:
An iso-stress condition is considered for the left and right parts of the undulated yarns
     
sright part yi ¼ sleft part yi ¼ syi ð6Þ

Then, to calculate the stress throughout each subdivision, a mixed iso-stress and iso-strain con-
ditions are considered. The iso-strain condition means that the strain in the undulated yarns equals
to that for each subdivision:
    
"yi ¼ Syi : syi ð7Þ
 
where Syi is the compliance matrix of the i-th yarn
   
"yi ¼ f"1 g ¼ f"2 g ¼ . . . ¼ "yn ð8Þ
 
where "yi , f"1 g, f"2 g,. . . and f"n g are the average strain in the undulated right/left part of the i-th
yarn.
 
s0 i ¼ ½Ci :f"i g ð9Þ
   
where s0 i , f"i g and Cyi are the stress, strain, and the stiffness matrix, respectively, of the i-th yarn
Under iso-stress condition
       
syi ¼ s00 1 ¼ s00 2 ¼ . . . ¼ s00 yn ð10Þ

The stress throughout the subdivision in the (xyz) global coordinate system is given by
   
fsi g ¼ Pc: s0 i þ ð1  PcÞ: s00 i ð11Þ

Finally, the stress in local coordinate system is calculated using a transformation matrix as follow
fsi gxyz ¼ Tc :fsi g123 ð12Þ

Failure criteria for subdivision-UD composite and matrix. The failure criterion proposed to pre-
dict the failure of UD composites is a tensor polynomial criterion proposed by Tsai and Wu (1971)
This criterion is used to predict the failure of subdivision-UD composite and straight yarns. This
criterion may be expressed in tensor notation as

Fi si þ Fij si sj þ Fijk si sj sk  1 ð13Þ

where i, j, k ¼ 1,. . ., 6 for a 3D case. The parameters Fi, Fij, and Fijk are related to the lamina
strengths in the principal directions. For practical proposes, and due to the large number of material
constants required, the third-order tensor Fijk is usually neglected. Therefore, the general polyno-
mial criterion reduces to a general quadratic expression given by

Fi si þ Fij si sj  1 ð14Þ
Hallal and Younes 9
where i, j ¼ 1,. . ., 6. Considering that the failure of the material is insensitive to a change of sign in
shear stresses, all terms containing a shear stress to first power must vanish: F4 ¼ F5 ¼F6 ¼ 0. Then,
the explicit form of the general expression is
F1 s1 þ F2 s2 þ F3 s3 þ 2F12 s1 s2 þ 2F13 s1 s3 þ 2F23 s2 s3 þ F11 s21
ð15Þ
þ F22 s22 þ F33 s23 þ F44 s24 þ F55 s25 þ F66 s26  1

where
1 1 1 1 1 1
F1 ¼  , F2 ¼  , F3 ¼ 
Xt Xc Yt Yc Zt Zc
1 1 1
F11 ¼ , F22 ¼ , F11 ¼
Xt :Xc Yt :Yc Zt :Zc
1 1 1
F44 ¼ 2 , F55 ¼ 2 , F66 ¼ 2
Q R S

where
Xt, Xc, Yt, Yc, Zt, and Zc are the ultimate axial strengths for UD composites calculated by
Chamis model (Chamis, 1989; Chamis et al., 2013) where t is used to indicate a tensile case and c
for the compression case, for x, y, and z directions, respectively. Q, R, and S represent the ultimate
shear strengths for UD composites, while Q and R represent the out of plane shear stress and S is
used to indicate the in-plane shear stress.
Concerning the failure for the pure matrix domain, the Christensen’s failure criterion is adopted.
The Christensen failure criterion (Christensen RM, 2007) is given as follow
 
1 1 1
 :ðs1 þ s2 þ s3 Þ þ
Xtm Xcm Xtm :Xcm

 ð16Þ
1  2 2 2
  2 2 2

 : ðs1  s2 Þ þðs2  s3 Þ þðs3  s1 Þ þ 3: s12 þ s13 þ s23  1
2

If
Xcm
Xtm  ð17Þ
2

sI  Xtm ð18Þ

where sI is the maximum tensile stress at a point which is the maximum value of the three roots of
the following cubic equation

s3I  ðs1 þ s2 þ s3 Þ:s2I þ s1 s2 þ s2 s3 þ s3 s1  s212  s223  s213 :sI
 ð19Þ
 s1 s2 s3  s1 s223  s2 s213  s3 s212 þ 2s12 s23 s13 ¼ 0

Damaged stiffness model for UD composites and matrix

The proposed damaged stiffness model is used to replace the stiffness matrix of a failed component
(straight yarn, subdivision-UD composites and matrix) by a reduced stiffness matrix. In recent works,
10 International Journal of Damage Mechanics 0(0)

the damaged stiffness matrix is chosen according to the major loading that causes the failure. If per
example the subdivision-UD composite is subjected to axial tensile stress and in-plane shear stress,
where the ratio of applied tensile stress to ultimate tensile strength of the material is higher than that
for the shear stress, the failure mode will be that of the axial stress. Consequently, a reduced damaged
stiffness matrix according to a tensile stress failure mode will be given for that failed subdivision.
This concept will lead to a huge error in failure prediction of textile composites, as it will be
shown in the next section. In this study, where modeling of textile composites with complex undu-
lated yarns is one of the objectives, assigning different failure modes for each failed subdivision will
be crucial in order to enhance the estimation of ultimate properties. Thus, it is proposed that a
different failure mode will contribute in the damaged stiffness matrix given for failed subdivision-
UD composite. The failure modes, as given by (Hallal et al., 2011b)consist of: Tensile and com-
pressive axial failure modes along x, y, and z direction, and shear failure modes for both positive and
negative applied shear stress. The failure modes are (Figure 4):

Axial tensile stress failure mode (11t), axial compressive stress failure mode (11c).
Transverse tensile stress failure mode (22t), transverse compressive stress failure mode (22c).
Transverse out of plane tensile stress failure mode (33t), transverse out of plane compressive stress
failure mode (33c).
Positive in-plane shear stress failure mode (12t), negative in-plane shear stress failure mode (12c).
Positive in-plane shear stress failure mode (13t), negative in-plane shear stress failure mode (13c).
Positive in-plane shear stress failure mode (23t), negative in-plane shear stress failure mode (23c).

The damaged stiffness is considered as composed from different damage stiffness matrices cor-
responding to each failure mode. The damaged stiffness matrix is given as follow
CUDdamaged ¼ R11t :C11t þ R11c :C11c þ R22t :C22t þ R22c :C22c
þ R33t :C33t þ R33c :C33c þ R23t :C23t þ R23c :C23c ð20Þ
þ R13t :C13t þ R13c :C13c þ R12t :C12t þ R12c :C12c

Figure 4. Failure modes.

Hallal and Younes 11
where ‘‘Rijt/c’’ are the ratios of applied stress to corresponding ultimate strength. These ratios are
calculated as shown in Appendix 1. ‘‘Cijt/c’’ represent the damaged stiffness matrices that corres-
ponds for each specific failure mode (Appendix 1). For a damaged subdivision, elastic properties are
recalculated in terms of those given by Chamis model (Chamis, 1989; Chamis et al., 2013) according
to the failure mode as shown (in Apprndix 1) (Lomov et al., 2009). Then the corresponding damaged
stiffness matrices are composed using the above reduced elastic properties.
In which concern the pure polymeric matrix that is considered as isotropic material, a damaged
stiffness matrix is used when the pure matrix is at failure. The damaged stiffness matrix is recalcu-
lated in terms of reduced elastic properties as follow (Blackketter et al., 1993)

Em ¼ 0:01 Em and Gm ¼ 0:2 Gm

Model’s discussion and improvements

In this section, the final failure of undulated yarns and the assigning of single or multi-failure modes
are investigated. A modeling of 8H-Satin weave will be done. This fabric is modeled because of the
influence of its undulated yarns in its final failure. In other words, the final failure of the undulated
yarn will indicate the final failure of that composite. In the first part of this section, the influence of
assigning single failure mode versus a multi-failure mode will be investigated.
In the second part, the number of damaged subdivisions that indicate the final failure of undu-
lated yarn is investigated. Using experimental data for the ultimate tensile strength, a determination
of the number of subdivisions that indicates the final failure is done. Then, the proposed modifica-
tions will be validated by modeling another type of composites, a 2  2 twill woven composite.

Single failure mode versus multi-failure mode

Almost all proposed analytical models that deals with strength prediction of composites used a
single failure mode to describe the failure of subdivisions-UD composites. This assumption was
adopted at first in this study; however, that assumption seems illogical where the effect of only one
stress is taken into account for assigning the corresponding failure mode, while the influences of
other stresses are neglected. This assumption is also more complicated to apply when a failure
criterion as Tsai and Wu (1971) or Hoffman (1967) is used. Assigning a single failure mode using
these criteria is based on the largest ratio of applied stress to the strength of the damaged subdivi-
sions-UD composite.
A modeling of 8H-Satin weave E-glass/epoxy composite (Scida et al., 1999) is performed in order
to investigate the effect of assigning one failure mode versus multi-failure modes. Figure 5 shows
the variation of the calculated ultimate tensile strength with the fiber volume fraction of warp and
weft yarns. It should be well noticed that this Vf is that for yarns and not for the entire
composite. Figure 5 shows clearly for both single and multi-failure modes, the ultimate tensile
strength increases regularly with Vf starting at 0.4 until 0.5. However, for a value of Vf ¼ 0.52, a
dramatic increase in tensile strength for the single failure mode is shown. A large increase from 490
Mpa at Vf ¼ 0.5, to 1675 Mpa at Vf ¼ 0.52 is shown. The value of 1675 MPa is definitely inaccep-
table and incorrect knowing that the experimental value is around 470 MPa. However, the multi-
failure modes method yields acceptable and more stable results, where no peak values are shown.
The tensile strength has increased gradually from 490 Mpa to around 600 Mpa, for an increase of Vf
from 0.4 to 0.7.
12 International Journal of Damage Mechanics 0(0)

2000
Sigma-X
(Mpa)
1800

1600

1400

1200

1000

800

600

400 Vf
0.4 0.42 0.44 0.46 0.48 0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 0.7
Sx-Single failure mode

Figure 5. The ultimate tensile strength obtained from single and multi-failure modes for different fiber volume
fraction of yarns Vf.

Final failure of undulated yarns

In this part, the final failure of undulated yarns is investigated. A problem that faces analytical
modeling of textile composite failure is related with assigning the complete failure of undulated
yarns. That problem did not appear in modeling the failure of laminated composites due to the
absence of undulated yarns. In literature, the first assumption says that, an undulated yarn divided
into subdivisions-UD composites is at final failure when all its subdivisions are at failure. However,
another assumption could be used, which claims that the final failure will occur when at least one
subdivision-UD composite is at failure. In order to understand that problem, the same 8H-Satin
weave fabric is modeled because its final failure under tensile loading along warp direction (x dir-
ection) is due to the final failure of the warp undulated yarn. Thus, the effect of final failure deter-
mination could have much influence on the predicted result.
In the above analysis, the first assumption for final failure is used. It is noticed that even when the
multi-failure mode is used, the calculated strength for (Vf yarns ¼ 0.634 equivalent to Vf composite ¼
0.501 similar to that experimental (Scida et al., 1999) is about 635 MPa and still far than that given by
experimental tests which is 470 MPa.
In order to solve that problem, it is proposed that the final failure of undulated yarn will be
considered for a specific number of failed subdivisions. Thus, a prediction of the ultimate tensile
strength is performed of the 8H-Satin weave composite with a fiber volume fraction similar to
experimental data (Scida et al., 1999)for different cases of assumptions of ratio of failed subdivisions
to total subdivisions that indicates the final failure of the yarn. The predicted results are plotted
against that ratio (Figure 6). It is noticed that a 0.9 ratio (which means a 90% of failed subdivisions)
is showing the best agreement with the experimental results.

Model improvements validation

In this part of the study, the final improvements of the proposed model are to be used to
determine the stress strain diagrams of a pure tensile loading for the previously studied
Hallal and Younes 13
Sigma-X (MPa)
650
625
600
575
550
analycal
525
Experimental 470 Mpa Sx
500
475
450
425
400
375
350
0.85 0.86 0.87 0.88 0.89 0.9 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1

Figure 6. The ultimate tensile strength for the proposed model with varying values of fraction that define the final
failure of an undulated yarn.

Table 1. Mechanical properties of fibers and matrices used in this study (Christiane 2006; Kaddour and Hinton,
2012).

Material E1 (Gpa) E2 (Gpa) G12 (Gpa) G23 (Gpa) NU12 NU23 Xt (Mpa) Xc (Mpa) S (Mpa)

Fibers
E-glass (Kaddour and 73 73 30.4 30.4 0.2 0.2 2150 1450 –
Hinton, 2012)
Carbon fiber AS4 231 15 15 0.2 0.2 3500 3000 –
(Kaddour and Hinton, 2012)
T300J (Christiane, 2006) 230 15 50 0.28 0.32 3500 3000 –

Matrix
Epoxy (Kaddour and 3.2 3.2 1.16 1.16 0.38 0.38 73 120 52
Hinton, 2012)
RTM6 (Christiane, 2006) 2.89 2.89 1.07 1.07 0.35 0.35 85 120 50

8H and a 2  2 twill E-glass reinforced polymer (Tables 1 and 2). The results in this part will be
used as first validation of the proposed model with the above improvements discussed in sections
single failure mode versus multi-failure mode and final failure of undulated yarns.
Figures 7 and 8 show the stress–strain diagrams for a tensile test obtained from experimental and
analytical modeling from Scida et al. (1999) compared to that predicted by the proposed analytical
model. The results for the 8H-Satin weave show great agreement for both ultimate tensile strength
and related ultimate strain (Figure 7). Concerning the 2  2 twill weave, the ultimate tensile strength
is also well predicted, while less agreement is found for the ultimate strain (Figure 8). For a stress-
based analytical model, both results seem acceptable and prove the potential of the proposed model.
14 International Journal of Damage Mechanics 0(0)

Table 2. Geometrical parameters for woven composites (Christiane, 2006; Scida et al., 1999).

Composite aw (mm) hw (mm) gw (mm) af (mm) hf (mm) gf (mm)

8H E-glass/epoxy 0.6 0.09 0 0.6 0.09 0

(Scida et al., 1999)
I-71 (Christiane, 2006) 2.8  0.123 0.49  0.059 2.822  0.2007 3.5  0.288 0.35  0.037 0  0.069
Twill (Scida et al., 1999) 0.83 0.09 0 0.83 0.09 0
Plain weave 1.8884 0.1116 0 1.8884 0.1116 0
(Tan et al., 1999b)

Figure 7. Pure tensile loading stress–strain diagrams for the 8H-sation weave: (a) obtained from analytical and
experimental data from Scida et al. (1999), (b) obtained from the proposed analytical model.

Figure 8. Pure tensile loading stress–strain diagrams for the Twill weave: (a) obtained from analytical and experi-
mental data from Scida et al. (1999) (b) obtained from the proposed analytical model.
Hallal and Younes 15
In the next section, the proposed model will adopt the proposed modifications and improvements in
order to predict the ultimate properties of different kinds of composites.

Results and discussion

While many experimental test are done in order to evaluate the mechanical properties of textile
composites, few studies had provided complete and detailed information about geometrical and
mechanical properties of the fabric, fibers, and matrix that enable accurate analytical or FE numer-
ical modeling. Selected experimental data are used in this section to provide a comparison with
different kinds of textiles (2D and 3D woven and braided fabrics), composed either from carbon or
glass fibers-reinforced polymers.
A modeling of elastic and ultimate strengths properties for different woven and braided compos-
ites is performed. The studied composites consist of one E-glass fiber/epoxy eight harness satin
weave (Scida et al., 1999) (Figure 9(a)), one E-glass fiber/epoxy twill weave (Scida et al., 1999)
(Figure 9(b)), one T300/epoxy 934 plain weave (Tan et al., 1999b), three AS4 carbon fiber/epoxy
tri-axial braided composites (Falzon and Herszberg, 1998) (Figure 9(c)), and one T300J carbon
fiber/epoxy 2.5D interlock woven composite (Christiane, 2006) (Figure 9(d)). The predicted analyt-
ical results are compared to available experimental data, predicted results from FE numerical
models and other numerical model, the asymptotic expansion homogenization method (AEHM).
In Table 1, the mechanical properties of fibers and matrices are shown. Moreover in Tables 2 and 3,
the required geometrical parameters for different fabrics are found.

Figure 9. Representative elementary volume (REV) for: (a) 8H-Satin weave, (b) Twill weave, (c) Tri-axially braided,
and (d) 3D angle interlock woven composites.
16 International Journal of Damage Mechanics 0(0)

Table 3. Geometrical parameters for braided composites (Falzon and Herszberg, 1998).

Braider
Composite L (mm) H (mm) amp (mm) Vfc angle

Br 30 (Falzon and Herszberg, 1998) 10.6 0.525 0.35 0.57 30

Br 45 a (Falzon and Herszberg, 1998) 19.796 0.42 0.35 0.57 45
Br 45 b (Falzon and Herszberg, 1998) 14.847 0.525 0.35 0.57 45

Table 4. Elastic properties for different woven composites predicted by the proposed model compared to available
experimental data (Christiane, 2006; Scida et al., 1999).

Ex Ey Ez Gxy Gxz Gyz

Composite Method (Gpa) (Gpa) (Gpa) (Gpa) (Gpa) (Gpa) vxy v xz vyz

8H Analytical 27.92 27.92 15.19 5.47 5.0 5.0 0.165 0.352 0.352
AEHM (Angioni et al., 2011) 27.85 28.50 16.70 6.21 5.68 5.68 0.14 0.3 0.57
Exp (Scida et al., 1999) 25.6 25.6 5.7 0.13
Diff% 9% 9% 4% 27%
Twill Analytical 18.77 18.77 8.8 3 3 3 0.134 0.373 0.373
Exp. (Scida et al., 1999) 19.2 19.2
Diff% 2% 2%
I-71 Analytical 49.04 33.75 7.32 3.50 3.38 2.73 0.056 0.408 0.37
FE (Nehme, et al., 2011; 49.20 33.07 7.82 2.75 2.97 2.19 0.058 0.463 0.506
Hallal et al., 2012)
Exp. (Christiane, 2006) 49.02 35.28 0.082
Diff% 0.04% 4% 32%
Plain weave FE numerical 56.5 56.5 10.18 4.28 2.86 2.86 0.043 0.045 0.045
(Tan et al., (Tan et al., 1999b)
1999b) Analytical 58.9 58.9 10 5.09 4.13 4.13 0.045 0.040 0.040
Exp. (Christiane, 2006) 59.4 59.4 4.353 0.049
Diff% 0.84% 0.84% 17% 7.23%

Tables 4 and 5 show the predicted results for the elastic properties; however, Tables 6 and 7 show
the ultimate tensile and shear strengths evaluated analytically by the above proposed model. Figures
10 to 12 show a comparison between analytical predictions and experimental measurements for in-
plane Young’s modulus Ex and Ey and tensile strength along warp direction ‘‘Sigma-x’’. The pre-
dicted results are compared to available experimental data and numerical models are found in the
literature. Unfortunately, there is no available experimental data for all elastic properties and ultim-
ate strengths. However, the available data could give a clear idea about the ability and potential of
the proposed model in evaluating the mechanical properties for different kinds of textile composites.
Concerning the elastic properties, it is well shown that for in-plane Young’s moduli Ex and Ey,
the error is not exceeding 10% and it is around 5% as average error. However, for in-plane shear
modulus Gxy, the error is also small with an error below 10%. For in-plane Poisson’s ratio, a good
Hallal and Younes 17
Table 5. Elastic properties for different braided composites predicted by the proposed model compared to
available experimental data (Tan et al., 1999b) and FE numerical modeling results Tsai et al. (2008).

Ex Ey Ez Gxy Gxz Gyz

Composite Method (Gpa) (Gpa) (Gpa) (Gpa) (Gpa) (Gpa) v xy v xz vyz

Br-30 Analytical 71.4 12.6 9.87 15.9 5.7 3.9 0.844 0.042 0.298
FE (Tsai et al., 2008) 68.574 9.639 8.594 19.002 3.114 2.889 1.394 0.019 0.330
Exp. (Tan et al., 68.9 14.98 1.23
1999b)
Diff% 4% 7% 31%
Br-45a Analytical 47.51 22.56 9.45 23.3 4.23 3.81 0.702 0.103 0.237
FE (Tsai et al., 2008) 42.934 20.873 9.395 25.969 2.781 2.782 0.824 0.017 0.123
Exp. (Tan et al., 44.1  2.3 21.4  1.2 25.03 0.75  0.02
1999b)
Diff% 8% 5% 7% 6%
Br-45b Analytical 47.22 21.61 9.42 21.3 4.24 3.82 0.681 0.111 0.247
FE (Tsai et al., 2008) 42.834 20.662 9.317 25.691 3.039 3.054 0.821 0.019 0.132
Exp. (Tan et al., 46.4  2.3 0.73  0.01
1999b)
Diff% 2% 6.7%

Table 6. Ultimate tensile and shear strengths for different woven composites predicted by the proposed model
compared to available experimental data (Christiane, 2006; Scida et al., 1999).

sx sy sz  xy  xz  yz
Composite Method (Mpa) (Mpa) (Mpa) (Mpa) (Mpa) (Mpa)

8H Analytical 465 465 75 55 55 55

Exp. (Scida et al., 1999) 470 470
Diff% 1.06% 1.06%
Twill analytical 500 500 75 40 55 55
Exp. (Scida et al., 1999) 480 480
Diff% 4.17% 4.17%
I-71 Analytical 689 210
Exp. (Christiane, 2006) 672.99 238.2
Diff% 2.38% 11.84 %
Plain weave FE numerical 651 651
(Tan et al., 1999b) Analytical 420 420
Exp. (Christiane, 2006) 515  45 515  45
Diff% 18% 18%
18 International Journal of Damage Mechanics 0(0)

Table 7. Ultimate tensile and shear strengths for different braided composites predicted by the proposed model
compared to available experimental data (Falzon and Herszberg, 1998).

sx sy sz  xy  xz  yz
Composite Method (Mpa) (Mpa) (Mpa) (Mpa) (Mpa) (Mpa)

Br-30 Analytical 635 90 90 90 60 60

Exp. (Falzon and Herszberg, 1998) 710
Diff% 11%
Br-45a Analytical 510 170 90 180 60 60
Exp. (Falzon and Herszberg, 1998) 451 178
Diff% 13% 4%
Br-45b Analytical 515 160 90 165 60 60
Exp. (Falzon and Herszberg, 1998) 458
Diff% 12%

80
70
60
50
EX (GPa)

40 Ex -ana

30 Ex -exp

20
10
0
8H Twill I-71 PW Br-30 Br -45a Br-45b

Figure 10. Analytical predictions compared to those experimental measurements and numerical predictions of
Young’s modulus Ex.

agreement is found between predicted analytical results and those experimental data. For the ultim-
ate strengths, beside the 8H-satin weave and the 2x2 twill weave composites, very good agreement
was found with an error not exceeding 13% for braided composite (Tables 6 and 7, and Figure 12).
Better results are shown for woven composites compared to those for tri-axially braided composites,
could be related to more accurate geometrical parameters used.
A fair discussion of the results predicted by proposed model shows that an error around 10% is
well acceptable, knowing that it is caused by many factors. That error is a result of errors in
geometrical parameters used, mechanical properties for fibers and matrices and the error of each
sub-model used for the calculation of the mechanical properties of subdivisions-UD composites, and
the error caused by the proposed model itself. Thus, any enhancement in these factors could lead to
a better prediction of the mechanical properties. An advantage of the proposed model is the
Hallal and Younes 19
70

60

50
Ey (GPa)

40
Ey -ana
30 Ey - exp

20

10

0
8H Twill I-71 PW Br-30 Br -45a Br-45b

Figure 11. Analytical predictions compared to those experimental measurements of Young’s modulus Ey.

800
Sigma-x (MPa)
700

600

500

400 Sx -ana
Sx -exp
300

200

100

0
8H Twill I-71 PW Br-30 Br -45a Br-45b

Figure 12. Analytical predictions compared to those experimental measurements of tensile strength along
x-direction (warp direction) ‘‘Sigma-x’’.

flexibility it has, to be used for all kinds of textiles, and in the same time its reliability. The proposed
model allows a preliminary study of mechanical properties for different kinds of fabrics with dif-
ferent architectures, fibers, and polymeric matrices. Another important advantage is the calculation
of time required, which is very short where it did not exceed couple of seconds to evaluate all elastic
properties and shear stresses, and about 20 seconds for ultimate axial stresses, performed on a
portable computer (Intel Õ core TM i5-2410M CPU @2.30 GHz 2.30 GHz) with an installed
memory (RAM) of 8 Gb.
20 International Journal of Damage Mechanics 0(0)

Conclusion
In this study, an analytical modeling of the mechanical behavior for textile composites is presented.
The proposed model consists of different sub-models: the 3SHM homogenization method, the Tsai–
Wu and Christensen failure criteria and the multi-failure damaged model.
The presented modeling technique is able to predict the 3D elastic properties and the ultimate
axial and shear strengths of any textile composites. Two problems that face analytical models con-
cerning the strength prediction are discussed. It is found that assigning a multi-failure mode is much
better than assigning a single failure mode for a failed subdivisions-UD composites of an undulated
yarn. In addition, the final failure of undulated yarns is investigated in comparison with available
experimental data. It is shown that it is better to consider the final failure of an undulated yarn when
90% of subdivisions-UD composites are at failure rather than 100%.
In order to investigate the potential of the proposed model, a modeling of seven different textile
composites, consist of four woven and three braided composites, is performed. The comparison of
predicted results with experimental data, FE numerical, and AEHM numerical models shows very
good agreement. The model provides a simple, easy to use, and generalized, as well as reliable
solution for the prediction of elastic properties and ultimate axial tensile and shear strengths of
textile composites.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or pub-
lication of this article.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.

References
Angioni SL, Meoa M and Foreman A (2011) A comparison of homogenization methods for 2-D woven
composites. Composites: Part B 42: 181–189.
Ansar M, Xinwei W and Chouwei Z (2011) Modeling strategies of 3D woven composites: A review. Composite
Structures 93: 1947–1963.
Ayranci C and Carey J (2008) 2D braided composites: A review for stiffness critical applications. Composite
Structures 85: 43–58.
Blackketter DM, Wlarath DE and Hansen AC (1993) Modeling damage in a plain weave fabric reinforced
composite material. Journal of Composites Technology and Research 15: 136–142.
Bystrom J, Jekabsons N and Varna J (2000) An evaluation of different models for prediction of elastic proper-
ties of woven composites. Composites: Part B 31: 7–20.
Byun JH (1989) Chou TW. Modeling and characterization of textile structural composites: A review. The
Journal of Strain Analysis for Engineering Design 24: 253.
Byun JH, Whitney TJ, Du GW, et al. (1991) Analytical characterization of two-step braided composites.
Journal of Composite Materials 25(12): 1599–1618.
Chamis CC (1989) Mechanics of composite materials: Past, present, and future. Journal of Composites,
Technology and Research 11: 3–14.
Chamis CC, Abdi F, Garg M, et al. (2013) Micromechanics-based progressive failure analysis prediction for
WWFE-III composite coupon test cases. Journal of Composite Materials 47(20–21): 2695–2712.
Chang Y, Jiao G, Wang B, et al. (2006) Elastic behavior analysis of 3D angle-interlock woven ceramic com-
posites. Acta Mechanica Solida Sinica 19(2): 152–159.
Hallal and Younes 21
Christiane ELH (2006) Mode´lisation du comportement e´lastique endommageable de mate´riaux composites à
renfort tridimensionnel. PhD Thesis, Université de Technologie de Compiègne, France.
Christensen RM (2007) Yield and failure criteria for isotropic materials. Available at: www.failurecriteria.com/
Media/Yield_and_Failure_Criteria_for_Isotropic_Materials.pdf (2007, accessed on 14 December 2008)..
Cox BN and Dadkhah MS (1995) The macroscopic elasticity of 3D woven composites. Journal of Composite
Materials 29(6): 785–819.
Crookston JJ, Long AC and Jones IA (2005) A summary review of mechanical properties prediction methods
for textile reinforced polymer composites. Proceedings of the Institution of Mechanical Engineers Part L
Journal of Materials Design and Applications 219: 91–109.
El Hage C, Younes R, Aboura Z, et al. (2009) Analytical and numerical modeling of mechanical properties of
orthogonal 3D CFRP. Composites Science and Technology 69: 111–116.
Falzon PJ and Herszberg I (1998) Mechanical performance of 2D braided carbon/epoxy composites.
Composires Science and Technology 58: 253–265.
Gommers B, Verpoest IV and Houtte P (1998) The Mori-Tanaka method applied to textile composite materiels.
Acta Materialia 46(6): 2223–2235.
Hahn HT and Pandey R (1994) A micromechanics model for thermoelastic properties of plain weave fabric
composites. Journal of Engineering Materials and Technology 116: 517–523.
Hallal A, Fardoun F, Younes R, et al. (2011a) Evaluation of longitudinal and transversal Young’s moduli for
unidirectional composite material with long fibers. Advanced Materials Research 324: 189–192.
Hallal A, Younes R, Nehme S, et al. (2011b) A corrective function for the estimation of the longitudinal
Young’s modulus in a developed analytical model for 2.5D woven composites. Journal of Composite
Materials 45(17): 1793–1804.
Hallal A, Younes R, Fardoun F, et al. (2012) Improved analytical model to predict the effective elastic proper-
ties of 2.5D interlock woven fabrics composite. Composite Structures 94: 3009–3028.
Hallal A, Younes R and Fardoun F (2013) Review and comparative study of analytical modeling for the elastic
properties of textile composites. Composites: Part B 50: 22–31.
Hinton MJ and Kaddour AS (2012) The background to the second world-wide failure exercise. Journal of
Composite Materials 46(19–20): 2283–2294.
Hoffman O (1967) The brittle strength of orthotropic materials. Journal of Composite Materials 1: 200–206.
Hu J (2008) 3-D Fibrous Assemblies Properties Applications and Modelling of Three-Dimensional Textile
Structures. Abington Hall, Granta Park, Great Abington, Cambridge CB21 6AH, England: Woodhead
Publishing Limited in association with The Textile Institute Woodhead Publishing Limited.
Ishikawa T and Chou TW (1982a) Elastic behaviour of woven hybrid composites. Journal of Composite
Materials 16(1): 2–19.
Ishikawa T and Chou TW (1982b) Stiffness and strength behaviour of woven fabric composites. Journal of
Materials Science 17: 3211–3220.
Ishikawa T and Chou TW (1983a) One-dimensional micromechanical analysis of woven fabric composites.
AIAA Journal 21: 1714–1721.
Ishikawa T and Chou TW (1983b) In-plane thermal expansion and thermal bending coefficients of fabric
composites. Journal of Composite Materials 17(2): 92–104.
Ishikawa T and Chou TW (1983c) Nonlinear behaviour of woven fabric composites. Journal of Composite
Materials 17(5): 399–413.
Jiang Y, Tabiei A and Simitses GJ (2000) A novel micromechanics-based approach to the derivation of con-
stitutive equations for local/global analysis of a plain-weave fabric composite. Composites Science and
Technology 60: 1825–1833.
Kaddour AS and Hinton MJ (2012) Input data for test cases used in benchmarking triaxial failure theories of
composites. Journal of Composite Materials 46(19–20): 2295–2312.
Kaddour AS, Hinton MJ, Smith PA, et al. (2013) The background to the third world-wide failure exercise.
Journal of Composite Materials 47(20–21): 2417–2426.
Kalidindi SR and Abusafieh A (1996) Longitudinal and transverse moduli and strengths of 3-D braided textile
composites with low braid angles. Journal of Composite Materials 30: 885–905.
22 International Journal of Damage Mechanics 0(0)

Kalidindi SR and Franco E (1997) Numerical evaluation of isostrain and weighted-average models for elastic
properties of three-dimensional composites. Composites Science and Technology 57: 293–305.
Kaw AK (2006) Mechanic of Composite Materials, 2nd ed. LLC, New York: Taylor & Francis Group.
Li DS, Lu ZX, Chen L, et al. (2009) Microstructure and mechanical properties of three-dimensional five-
directional braided composites. International Journal of Solids and Structures 46: 3422–3432.
Lee SK, Byun JH and Hong SH (2003) Effect of fiber geometry on the elastic constants of the plain woven
fabric reinforced aluminum matrix composites. Materials Science and Engineering A347: 346–358.
Lomov S, Bogdanovich A, Ivanova D, et al. (2009) Finite Element modeling of progressive damage in non-
crimp 3D orthogonal weave and plain weave E-glass composites. In: 2nd World conference on 3D fabrics,
Greenville, South Carolina, USA.
Mouritz AP, Bannister MK, Falzon PJ, et al. (1999) Review of applications for advanced three-dimensional
fibre textile composites. Composites Part A 30: 1445–1461.
Naik NK and Kuchibhotla R (2002a) Analytical study of strength and failure bahaviour of plain weave fabric
composites made of twisted yarns. Composites: Part A 33: 697–708.
Naik NK, Sk NMA and Durga Prasad P (2002b) Stress and failure analysis of 3D Angle interlock woven
composites. Journal of Composite Materials 36: 93.
Naik NK and Ridevi E (2002c) An analytical method for thermoelastic analysis of 3D orthogonal interlock
woven composites. Journal of Reinforced Plastics and Composites 21: 1149.
Nehme S, Hallal A, Fardoun F, et al. (2011) Numerical/analytical methods to evaluate the mechanical behavior
of interlock composites. Journal of Composite Materials 45(16): 1699–1716.
Pochiraju K and Chou TW (1999a) Three-dimensionally woven and braided composites. I: A Model for
Anisotropic Stiffness Prediction. Polymer Composites 20(4): 565–580.
Pochiraju K and Chou TW (1999b) Three-dimensionally woven and braided composites. II: An experimental
characterization. Polymer Composites 20(6): 733–747.
Prodromou A (2004) Mechanical modelling of textile composites utilizing the method of cells. PhD Thesis,
Katholieke Universiteit Leuven, Belgium.
Prodromou A, Lomov S and Verpoest I (2011) The method of cells and the mechanical properties of textile
composites. Composite Structures 93: 1290–1299.
Quek SC, Waas AM, Shahwan KW, et al. (2003) Analysis of 2D triaxial flat braided textile composites.
International Journal of Mechanical Sciences 45: 1077–1096.
Ramakrishna S (1997) Analysis and modeling of plain knitted fabric reinforced composites. Journal of
Composite Materials 31(1): 52–70.
Ruan X and Chou TW (1996) Experimental and theoretical studies of the elastic behavior of knitted-fabric
composites. Composites Science and Technology 56: 1391–1403.
Sankar BV and Marrey RV (1997) Analytical method for micromechanics of textile composites. Composites
Science and Technology 57: 703–713.
Scida D, Aboura Z, Benzeggagh M, et al. (1999) A micromechanics model for 3D elasticity and failure of
woven-fibre composite materials. Composites Science Technology 59(4): 505–517.
Shokrieh M and Mazloomi M (2010) An analytical method for calculating stiffness of two-dimensional tri-axial
braided composites. Composite Structures 92: 2901–2905.
Tan P, Tong L and Steven GP (1997) Modelling for predicting the mechanical properties of textile composites –
A review. Composites Part A: Applied Science and Manufacturing 28A: 903–922.
Tan P, Tong L and Steven GP (1999a) Micromechanics models for mechanical and thermomechanical proper-
ties of 3D through-the-thickness angle interlock woven composites. Composites: Part A 30: 637–648.
Tan P, Tong L and Steven GP (1999b) Micromechanics models for the elastic constants and failure strengths of
plain weave composites. Composite Structures 47: 797–804.
Tan P, Tong L and Steven GP (2000a) Behavior of 3D orthogonal woven CFRP composites. Part I.
Experimental investigation. Composites: Part A 31: 259–271.
Tan P, Tong L and Steven GP (2000b) Behavior of 3D orthogonal woven CFRP composites. Part II. FEA and
analytical modeling approaches. Composites: Part A 31: 273–281.
Hallal and Younes 23
Tan P, Tong L and Steven GP (2001) Mechanical behavior for 3-D orthogonal woven E-glass/epoxy compos-
ites. Journal of Reinforced Plastics and composites 20(4): 274–303.
Tong L, Moritz AP and Bannister MK (2002) 3D Fiber Reinforced Polymer. Oxford: Elsevier.
Tsai KH, Hwan CL, Chen WL, et al. (2008) A parallelogram spring model for predicting the effective elastic
properties of 2D braided composites. Composite Structures 83: 273–283.
Tsai SW and Wu EM (1971) A general theory of strength for anisotropic materials. Journal of Composite
Materials 5: 58–80.
Vandeurzen Ph, Ivens J and Verpoest I (1996a) A three-dimensional micromechanical analysis of woven-fabric
composites: I. Geometric analysis. Composites Science and Technology 56: 1303–1315.
Vandeurzen Ph, Ivens J and Verpoest I (1996b) A three-dimensional micromechanical of woven-fabric com-
posites: II. Elastic. Composites Science anal Technology 56: 1317–1327.
Whitney TJ and Chou TW (1989) Modeling of 3-D angle-interlock textile structural composites. Journal of
Composite Materials 23: 890–911.
Yang JM, Ma CL and Chou TW (1986) Fiber inclination model of three-dimensional textile structural com-
posites. Journal of Composite Materials 20(5): 472–484.

Appendix 1
Modeling of longitudinal section
Undulated yarns of 2D woven, 3D woven, and 2D braided composites will be modeled by a
sinusoidal shape.
An undulated yarn in the (XZ) plane is taken as example (Figure 13). The undulated part of yarn
is modeled by its centerline with a sinusoidal function
 
2x
ZðxÞ ¼ a:cos ð21Þ
T

where

Left undulated section : T=2  x  T ð22Þ

Right undulated section : 0  x  T=2 ð23Þ

where ‘‘a’’ and ‘‘T’’ are the amplitude and the period, respectively. They are calculated in terms of
geometrical parameters according to the composite architecture.
The volume of the entire undulated part:

Vund ¼ A:Lund ð24Þ

Figure 13. Sinusoidal shape of an undulated yarn.

24 International Journal of Damage Mechanics 0(0)

where A is the area of transverse section and Lund is the length of the centerline of the undulated part
is given by
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z T=2  2
dz
Lund ¼ 1þ dx ðfor right undulated partÞ ð25Þ
0 dx
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z T  2
dz
Lund ¼ 1þ dx ðfor left undulated partÞ ð26Þ
T=2 d x

 2   
dz 2 2x 2
where ¼ a: :sin ð27Þ
dx T T

Modeling of the transverse section

The transverse section of yarns could be represented in different shapes: rectangular, circular,
lenticular, elliptical, and racetrack (ellipse þ rectangle) shapes (Figure 14). In order to determine
their areas, the accuracy in determining the width and the height of yarns is crucial. Then, the area of
each cross-section is easily calculated in terms of width and height of the yarn’s transverse section.
Noting that, the geometrical parameters such as the width ‘‘a,’’ the height ‘‘h,’’ in addition to the
yarns gap ‘‘g’’ serve in the calculation of the longitudinal section.

Fiber volume fraction

The calculation of fiber volume fractions in the composite Vfc and in each yarn Vfy is very important
in determining the stiffness matrix of the composite. There are many analytical methods where Vfy
could be determined, dependent on available experimental data of the studied composite.
Having the number of fiber filaments ‘‘n’’ and its diameter ‘‘d,’’ Vfy could be estimated after the
calculation of the filaments cross-section area ‘‘Af ’’ and the yarn cross-section area ‘‘A’’ as follow

Af
Vfy ¼ ð28Þ
A

Figure 14. Geometrical parameters and different yarn cross-sections: (a) rectangular, (b) circular, (c) lenticular, (d)
elliptical, (e) racetrack.
Hallal and Younes 25
where

d2
Af ¼ n:: ð29Þ
4
Another method could be adopted if the density and the weight/length ratio ‘‘WL’’ are available;
in this method, the area of the cross-section is also needed.

Af
Vfy ¼ ð30Þ
A

where
WL
Af ¼ ð31Þ

Knowing that almost all experimental data of composites give Vfc , where Vfy could be calculated
in terms of REV and yarns volumes

Vfc :VREV
Vfy ¼ ð32Þ
VYarns

In this method, it is assumed that all yarns have the same fiber volume fraction. It is better used
with non-hybrid textiles and with yarns that have the same Tex.
It could be noticed from above equations, the importance of well defining the transversal section
type of yarns and the calculation of its cross-sectional area. In addition, the calculation of the length
of longitudinal section, in the third method, is important to estimate a correct value og Vfy.

Homogenization
At the micro level homogenization, the discretized small sub-volumes of undulated yarn, in addition
to the straight yarns, are considered as unidirectional lamina with long fibers. They represent
transversely isotropic materials (Figure 15). Their effective elastic properties are found using a
micromechanical analytical model as described in the previous section. Then, stiffness matrices of
yarns sub-volumes ½C123
ui  are known in the local coordinate system (123) in terms of compliance
matrix

½C123 123 1
ui  ¼ ½Sui  ð33Þ
2 3
1=E11  12 =E11 12 =E11 0 0 0
6 7
6  12 =E11 1=E22  23 =E22 0 0 0 7
6 7
6 7
6  12 =E11  23 =E22 1=E22 0 0 0 7
½S123 6
ui  ¼ 6
7 ð34Þ
0 0 0 1=G23 0 0 7
6 7
6 7
6 0 0 0 0 1=G12 0 7
4 5
0 0 0 0 0 1=G12

½S123
ui  is the compliance matrix of a transversely isotropic material.
26 International Journal of Damage Mechanics 0(0)

Figure 15. The homogenization scheme of the 3SHM model.

There are many analytical micromechanical models used to predict the elastic properties of
unidirectional lamina with long fibers. Some micromechanical models are also used to evaluate
the ultimate strength of UD composites. Based on a comparative study done previously by the
authors (Hallal et al., 2011a), the Chamis micromechanical model (Chamis, 1989; Chamis et al.,
2013) is used to evaluate the elastic properties and ultimate strength for a UD composite.
The matrix is considered to be an isotropic materiel where its stiffness matrix ½Cm  can be simply
derived from the compliance matrix ½Sm  in terms of Young modulus Em and the Poisson’sratio m .
2 3
1=Em  m =Em m =Em 0 0 0
6 7
6  m =Em 1=Em  m =Em 0 0 0 7
6 7
6 7
6  m =Em  m =Em 1=Em 0 0 0 7
½Sm  ¼ 6
6
7
7 ð35Þ
6 0 0 0 Em =2:ð1 þ vm Þ 0 0 7
6 7
6 0 0 0 0 Em =2:ð1 þ vm Þ 0 7
4 5
0 0 0 0 0 Em =2:ð1 þ vm Þ

At the meso level homogenization, the stiffness matrices of the undulated yarns are evaluated.
The stiffness of an undulated yarn is evaluated using both iso-strain and iso-stress assumptions when
assembling sub-volumes respecting the following scheme:

(1) The stiffness and compliance matrices of sub-volumes in the (XYZ) coordinate system are
given by

Cxyz 123 t
ui ¼ Tc :Cui :Tc ð36Þ
Hallal and Younes 27
and

Sxyz 123 t
ui ¼ Ts :Sui :Ts ð37Þ

where Tc and Ts are the 3D transformation matrices.

(2) The sub-volumes of left and right undulated parts (referred by L and R, respectively) (Figure 16)
are assembled in a mixed iso-strain /iso-stress model parallel using a weighted parameter Pcu ,
where the stiffness matrices of left and right undulated parts are given respectively by

Pn Pn 1
  VLui :Cxyz VLui :Sxyz
Cxyz
Lu ¼ Pcu : i¼1 Lui
þ ð1  Pcu Þ: i¼1 Lui
ð38Þ
VLu VLu
Pn Pn 1
  VRui :Cxyz VRui :Sxyz
Cxyz
Ru ¼ Pcu :
i¼1 Rui
þ ð1  Pcu Þ: i¼1 Rui
ð39Þ
VRu VRu

where

Pcu ¼ 2: ð40Þ

where
VLui : Volume of the i-th sub-volume of the left undulated part
VRui : Volume of the i-th sub-volume of the right undulated part

Figure 16. (a) Parallel system, (b) series system, (c) sub-volumes of actual undulated yarn.
28 International Journal of Damage Mechanics 0(0)

VLu : Volume of the left undulated part

VRu : Volume of the right undulated part
 The mean value of the inclination angle of sub-volumes ‘‘’’
:  of a subdivided undulated yarn
P
i
( ¼ n ; with i the inclination angle of each sub-volume and n the number of sub-volumes)
Noting that in general and due to the symmetry of left and right undulated parts, we have

VLu ¼ VRu ð41Þ

And the same value of Pcu for both parts.

(3) Now the yarn is composed of homogenized left and right undulated parts which form a series
system when assembled along undulation direction (Figure 15). The iso-stress model is then used
to evaluate the stiffness of the undulated yarn (left þ right):

 1  1
  VLu : Cxyz
Lu þVRu : Cxyz
Ru
Sxyz
u ¼ ð42Þ
Vu
   xyz 1
Cxyz
u ¼ Su ð43Þ

where Vu is the volume of the yarn

At the macro level homogenization stage, the stiffness of the REV is evaluated in terms of the
previously evaluated stiffness matrices of the matrix and the n yarns. Therefore, the composite is
treated as nþ1 homogenized blocks, which are assumed to represent a parallel system and are
assembled under an iso-strain condition
Pk¼n
k¼1 Vku :Cku þ Vm :Cm
CREV ¼ ð44Þ
VREV

where Vm and VREV are the volumes of the matrix and the REV, respectively.

Damaged stiffness model

Ratios
R11t ¼ sXxt (ratio of stresses along x direction); R11c ¼ X sx
(ratio of stresses along x direction)
c s
sy
R22t ¼ Yt (ratio of stresses along y direction); R22c ¼ Yyc (ratio of stresses along y direction)
sz
R33t ¼ Y t
(ratio of stresses along z direction); R33c ¼ Yszc (ratio of stresses along z direction)
 xy  xy
R12t ¼  xyu (ratio of in-plane (xy) shear stresses for positive applied shear); R12c ¼  xyu (ratio of in-
plane (xy) shear stresses for negative applied shear)
R13t ¼ xzuxz
(ratio of in-plane (xz) shear stresses for positive applied shear); R13c ¼ xzu
xz
(ratio of in-
plane (xz) shear stresses for negative applied shear)
yz yz
R23t ¼ yzu (ratio of in-plane (yz) shear stresses for positive applied shear); R23c ¼ yzu (ratio of in-
plane (yz) shear stresses for negative applied shear)
 Hallal and Younes 29
Reduced elastic properties
Axial tensile stress failure mode (11t): E11 ¼ 0:01E11 , G12 ¼ 0:01G12 , G13 ¼ 0:01G13 and
v12 ¼ 0:01v12 .
Axial compressive stress failure mode (11c): G12 ¼ 0:01G12 , G13 ¼ 0:01G13 .
Transversal tensile stress failure mode (22t): E12 ¼ 0:01E12 , G12 ¼ 0:01G12 , G23 ¼ 0:01G23 and
v23 ¼ 0:01v23 .
Transversal compressive stress failure mode (22c): G12 ¼ 0:01G12 , G23 ¼ 0:01G23 .
Transversal out of plane tensile stress failure mode (33t): E33 ¼ 0:01E33 , G13 ¼ 0:01G13 ,
G23 ¼ 0:01G23 and v13 ¼ 0:01v13 .
Transversal out of plane compressive stress failure mode (33c): G13 ¼ 0:01G13 , G23 ¼ 0:01G23 .
Positive in-plane shear stress failure mode (12t): same as failure mode (22t).
Negative in-plane shear stress failure mode (12c): same as failure mode (22c).
Positive in-plane shear stress failure mode (13t): same as failure mode (33t).
Negative in-plane shear stress failure mode (13c): same as failure mode (33c).
Positive in-plane shear stress failure mode (23t): E22 ¼ 0:01E22 , G12 ¼ 0:01G12 , G13 ¼ 0:01G13 ,
G23 ¼ 0:01G23 and v23 ¼ 0:01v23 .
Negative in-plane shear stress failure mode (23c): E22 ¼ 0:01E22 , G12 ¼ 0:01G12 , G13 ¼ 0:01G13 ,
G23 ¼ 0:01G23 and v23 ¼ 0:01v23 .

View publication stats