860967

research-article2019
JEF0010.1177/1558925019860967Journal of Engineered Fibers and FabricsŠomođi et al.

Original Article

Journal of Engineered Fibers and Fabrics

Analysis of woven fabric in asymmetric Volume 14: 1­–8

© The Author(s) 2019
https://doi.org/10.1177/1558925019860967
DOI: 10.1177/1558925019860967
tensile loading using parabolic journals.sagepub.com/home/jef

approximation of tensile nonlinearity

Željko Šomođi, Emilija Zdraveva and Snježana Brnada

Abstract
Unlike many other engineering materials, deformational behaviour of fabrics is marked by specific nonlinearities. For
the purpose of certain engineering analyses, nonlinearity can be approximately described by means of appropriate
models. A number of possibilities in approximation of tensile nonlinearity are statistically analysed and compared for
the representative selection of woven fabrics. Second-order parabolic approximation is estimated to combine simplicity
and good accuracy for a selected woven fabric. It is then included into deformational analysis of specimen in asymmetric
tensile loading, as the case representative for structural application of textile, where geometric conditions combined with
material properties define the mechanical behaviour of the body. The results indicate the factors of stress concentration
due to load eccentricity. Simulation of tensile test gives the theoretical prediction of apparent reduction in stiffness and
strength of the specimen in terms of the load eccentricity.

Keywords
Woven fabric, tensile nonlinearity, out-of-centre tensile test, numerical simulation

Date received: 16 April 2019; accepted: 11 June 2019

Introduction prediction of the mechanical properties of woven fabrics by

Woven fabrics structures are quite complex, and even at
low deformations, there is a large number of intra-displace-
ments. The tensile properties of woven fabrics are very sig-
nificant as they can indicate the fabrics’ behaviour during
wear. In tensile testing, the main parameters include the
tensile stiffness, the tensile stress and tensile strain. A typi-
cal tensile stress–strain curve of a woven fabric confirms
the nonlinear behaviour of the material, which is present in
the initial stage during yarns alignment, which results in
lower stress increase and in the final stage as the applied
force overcomes the frictional force, thus, the yarn’s fibre
bonds become weak. Understanding the mechanical behav-
iour of woven fabrics will help in their design, fabrication
and fabrication process control.1 Within this respect, many
authors have reported in different studies dealing with the
theoretical modelling of the changes in their structure under
loading or by the usage of advanced simulation methods
whose numerical models are usually compared with mate-
rials real (experimental) behaviour.
Barburski and Masajtis2 discussed the modelling of
woven fabric structure change under static load by the
modification of the Pierce’s model and Painter nomogram
thus including the changes in yarns cross-sections changes,
stiffness in threads, one-system resistance area and the
contact area of two crossing threads. Malik et al. reported
Faculty of Textile Technology, University of Zagreb, Zagreb, Croatia
Corresponding author:
Željko Šomođi, Faculty of Textile Technology, University of Zagreb,
Prilaz baruna Filipovića 28a, Zagreb 10000, Croatia.
Email: zeljko.somodi@ttf.hr

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0
License (http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of
the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages
(https://us.sagepub.com/en-us/nam/open-access-at-sage).
2 Journal of Engineered Fibers and Fabrics 

on the prediction of the tensile strength in cotton woven

fabrics by multiple linear regression models based on warp
and weft yarn strength, ends and picks per 25 mm and float
lengths as predictors. It was noted that main factors affect-
ing the fabric strength in the warp and weft direction were
the warp yarn strength and ends per 25 mm, and the weft
yarn strength and picks per 25 mm, respectively.3 The ten-
sile behaviour of different weave-patterned fabrics was
simulated by 3D geometrical models using finite element
analysis. Difference between experimental data and
numerical model was reported to be 20%. Among the fab-
rics used, the highest and least rupture loads were noted for
the plain and satin weaves.4 Significant contributions in
the field of textile fabrics’ mechanical behaviour analysis
were noticed in Postle et al.’s research works.5 Further
contributions are found in the study of Boisse et al.,6 where Figure 1.  Linear, polynomial, cosine and sigmoidal fitted
virtual tests or 3D finite element analysis of the unit woven curves of the plain weave fabric of 20 threads/cm density.
cells considered the biaxial tension and in-plane shear of
woven reinforcements. Lomov et al.7 considered optical
and eccentric loading by Wysmulski and Debski. It was
full-field strain measuring techniques of the deformability
reported that the eccentric load had significant effect on
in textiles during tensile or shear tests. In the analysis of
the column buckling depending on the eccentricity direc-
the stress–strain state of yarns and fabrics, some studies
tion. The significance was further confirmed for the lami-
were based on the principles of the elastic energy theory or
nate ply orientation on the load critical values.12 Other
probabilistic methods.8,9
authors considered analysis of cementitious materials
Apart from the standard tensile testing, another interest-
under modified eccentric compact tension tests,13 steel-
ing aspect would be the eccentric axial or the out-of-centre
fibre-reinforced composite bars (SFCB) under eccentric
loading test, where the applied force is at a distance from
compression tests,14 and so on. Forth-mentioned studies
the material centre. Such test results in certain phenomena
have important contribution in the design of load-bearing
that depend on materials tested. To the best of knowledge,
engineering structures.
forth-mentioned analysis and resulting phenomena con-
This study considers two aspects. First, curve fitting
cerning woven fabrics are scarce, while existing studies
models were set for the tensile curves of selected woven
mostly deal with composite materials based on concrete or
fabrics with different weaves by using OriginPro 8 soft-
epoxy resin matrices used in automotive, aerospace, or
ware tool. Second, a newly developed numerical model
civil engineering applications.
that simulates the tensile behaviour of woven fabrics under
A stress–strain model considering load eccentricity in
eccentric loading was presented. This study considers
carbon-fibre-reinforced polymer (CFRP)-confined con-
more realistic nonlinear approximation as compared with
crete columns was suggested by Wu and Jiang. The
authors’ previous work where the linear elastic law in ten-
increase in the eccentric loadings resulted in the stress–
sile deformation was used for the prediction of the fabrics
strain stiffness increase. When compared with the concen-
tensile behaviour under out-of-centre load.15 The newly
tric loaded columns, the out-of-centre-loaded columns
developed numerical model contributes in the mechanics
showed 50% higher ultimate failure strain.10 Khashaba
of woven textile materials.
et al. evaluated the mechanical behaviour woven glass-
fibre-reinforced polyester resin (GFRP) composites under
monotonic or combined (tension/bending) loading. In the Tensile test curve fitting models of
combined tests, the eccentric tension load was carried by
offset steel shims and the specimens were mounted eccen-
selected woven fabrics
trically. The load-displacement curve was characterised by In this section, the force-deformation curve fitting models
initial non-linearity due to high initial eccentricity and a are given for the selected woven fabrics including the plain
second linear behaviour up to failure. The decrease of the weaves with the warp-weft densities of 20 and 24 threads/
maximum tensile loads results from the increase in the ini- cm, respectively, the satin 4/1(3) and twill weave with the
tial eccentricity, and this effect is more significant in case warp-weft densities of 24 threads/cm (Figures 1–4). The
of the [0]8 samples than the quasi-isotropic woven GFRP chosen curve fitting models were linear, polynomial for
laminate [0/±45/90]s.11 Thin-walled column structure the power of 2 and 3, cosine and sigmoidal curves.
CFRP laminates, with eight plies symmetric to the mid- Table 1 lists the adjusted R-squared values of each of
plane of the laminate, were tested under compression axial the models for the selected woven fabrics and the
Šomođi et al. 3

Figure 2.  Linear, polynomial, cosine and sigmoidal fitted Figure 4.  Linear, polynomial, cosine and sigmoidal fitted
curves of the plain weave fabric of 24 threads/cm density. curves of the twill weave 2/2 fabric of 24 threads/cm density.

for all woven fabrics. Cosine and especially sigmoidal

curves provide overall excellent accuracy, but it is worth
noting that considerably simpler parabolic approximation
also gives very good fit for plane weave, and even linear
approximation for satin and twill weave. Maybe cubic
curve would fit better for knitted fabric.

Computational mechanical model of

Figure 3.  Linear, polynomial, cosine and sigmoidal fitted
curves of the satin 4/1(3) fabric of 24 threads/cm density.
appropriate equations for the chosen fitting models. The
results suggest that the fitting level depends on the type of
fabric and type of chosen model. In this article, the high
fitting levels (highest R2) are noticed for the plain weave
(24 threads/cm) fabric for all chosen fitted models except
linear (values close to 1), while the high level of fitting
was provided by the cosine and sigmoidal curves models
fabric based on parabolic tensile law
In this section, let us implement previously considered
parabolic tensile approximation to a ‘structural’ situation
involving fabric. As a representative situation of such
type, in which there is an uneven distribution of stress,
consider the tensile load on fabric specimen applied by
means of a rigid crossbar with certain eccentricity, that
is, non-symmetrically, Figure 5(a). Let the material be
assumed as nonlinearly elastic with parabolic law, but
only in the tensile range, while in the compression it has
virtually zero stiffness, it is flexible and buckles easily
and any compressive stress is cut to zero. The adopted
stress–strain diagram is shown in Figure 5(b).
If stress and strain at the breaking point, as the maxi-
mum possible stress and strain in the fabric, are denoted as
σM and εM, respectively (Figure 5(b)), the assumed tensile
law can be written in the form

Table 1.  Curve fit equations and adj. R2 values of the selected woven fabrics.

Woven fabric Curve fit

Equation y = A + B × x y = A × x2 y = A × x3 y = B – B × cos(x × A) y = A1 + (A2 – A1)/

(1 + 10((logx –x)  × p))
0

R2
Plain weave (20 threads/cm) 0.96869 0.97200 0.81640 0.99902 0.99982
Plain weave (24 threads/cm) 0.91830 0.99842 0.93803 0.99835 0.99999
Satin 4/1(3) (24 threads/cm) 0.99334 0.88941 0.63095 0.98917 0.99967
Twill weave (24 threads/cm) 0.98097 0.94913 0.75735 0.99728 0.99956
4 Journal of Engineered Fibers and Fabrics 

Figure 5.  Specimen in (a) tensile loading and (b) assumed stress–strain diagram.

σM In equation (5), t is introduced as the thickness of

σ = kε 2 , k = (1) the body. Let us first consider the case of small
ε M2
eccentricity, when v(0) >  0, that is, aϕ<vc – the
deformed shape and stress distribution are depicted in
Due to non-symmetric load, deformation of the spec- Figure 6(a).
imen can be described by two parameters: elongation at If equation (4) is introduced to equation (5), integration
central point vc and angle of rotation of the crossbar ϕ in the limits from x = 0 to x = 2a gives the expressions for
(see Figure 6). First, let us try to establish the relation- force F and eccentricity e in the form
ship between the load parameters (tensile force F and
eccentricity e) and displacements vc and ϕ. To do so, 2atk  2 1 2 2 2 vc aϕ
like in any stress–strain analysis, we need geometric
F=
l2 vc + 3 a ϕ  ; e = 3 a 2 1 2 2 (6)
  vc + a ϕ
and static conditions, on top of the material law. The 3
displacement v of a given point of the crossbar, that is,
elongation at the corresponding point of the specimen This is valid only until aϕ=vc or e = a/2. This corre-
cross section, can be approximately expressed, intro- sponds to the case when the point of action of the tensile
ducing coordinate x as in Figure 5(a) or 6, in the form force reaches the edge of the ‘core’ of the cross section
(central region in which normal force produces the stress
v = vc + ϕ ( x − a ) (2) of equal sign in all points of the cross section). For higher
values of e, we talk of large eccentricity and the situation
is depicted in Figure 6(b). Again, the static conditions
The strain in the specimen is then, denoting by l the
require that the tensile force be equal to the area of the
length of the specimen
stress diagram in Figure 6(b), and that it acts in the centre
of gravity of the diagram. The following equations are
1 derived from these conditions
ε = vc + ϕ ( x − a )  (3)
l
 e   e
b = 2a  2 − 1 ; vc =  3 − 4  aϕ ;
The stress in the specimen is assumed to be uniaxial,  a   a
3 (7)
that is, the effect of transverse constraint of the rigid cross- 32 2atk  e  2 2
F= 1− aϕ
bar and the fixed upper clamp is neglected and the stress is 3 l 2  a 
then
Note that b represents the width of the specimen portion

k
σ= 2  ( c )
ϕ 2 x 2 + 2v ϕ − 2aϕ 2 x + v 2 
c
 (4)
in which strain is negative and thus stress is equal to zero
 −2av ϕ + ϕ 2 a 2  (Figure 6(b)).
l
 c  Now let us consider the impact of eccentricity e to the
degree of increase in maximum stress at the tensile load by
Static conditions are in fact the equilibrium equations a given constant value of the force F. This can be referred
of the crossbar to as stress concentration due to load eccentricity. In the
case without eccentricity, we have σ = F/2at = const., and
this value of stress can be kept for later reference. In the
F= ∫ σ dA = t ∫ σ dx; F ( a + e ) = t ∫ xσ dA (5) limit case between small and large eccentricity, that is, at
ε >0 ε >0 ε >0 e = a/2, use of the formula for the area limited by parabolic
Šomođi et al. 5

Figure 6.  Deformed shape and stress distributions for (a) small and (b) large eccentricity.

Table 2.  Maximum stress in terms of force eccentricity.

e/a 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.85 0.9 0.95
σMAX/σref 1 1.315 1.662 2.046 2.479 3 3.75 5 7.5 10 15 30

arc in equilibrium of forces on the crossbar delivers the when e = a/2. The expression for maximum stress is then
expression σMAX = 3 F/2at. In the case of small eccentricity, rearranged and reduced to
equation (6) upon elimination of aϕ lead to the following
quadratic equation in vc2  2
F  e e 
σ MAX = 2 + 3 − 1 − 3   (11)
2 2at  a a 
Fl 2 3  e  F 2l 4  
vc4 − vc2 +   2 2 2 = 0 (8)
2atk 16  a  a t k
In the case of large eccentricity (Figure 6(b)), the static
The solution of equation (8) is introduced, using again condition can be written in the form
the relations (equation (6)), in the expression for maxi-
mum stress (equation (9))
t
σ MAX ( 2a − b ) = F (12)
k 3
σ MAX = 2 ( vc + aϕ )2 (9)
l
The expression for maximum stress, using first of the
The following expression for maximum stress is then expressions (equation (7)), takes the form
obtained
F 3
  2   σ MAX =
3 e + 1  1 ± 1 − 3  e   + 2at  e  (13)
  2 1 − 
 a 2 a    a
F    
σ MAX =  2  (10) The results (equations (11) and (13)) are shown in Table 2
2at  9  e  1 
 2  a  2  and Figure 7. Maximum stress is normalised with reference
 e  to the uniform stress with zero eccentricity σref = F/2at, while
1 ± 1 − 3 
 a  the eccentricity is normalised with respect to the half width
of the specimen a.
Physically meaningful solution is extracted from equa- Solutions (equations (11) and (13)) are checked to pro-
tion (10) to comply with the mentioned σMAX = 3F/2at vide continuity in the transition point e = a/2 not only in
6 Journal of Engineered Fibers and Fabrics 

Finally, let us look for the computational prediction of

the curve F – Δl of tensile test with eccentric load. The
apparent elongation (displacement of the point of action of
the force) can be approximately taken as

∆l * = vc + eϕ (16)

In the case of small eccentricity, again the solution of

equation (8) for vc is required, along with the expression
for the angle ϕ following from equation (6)

1 3 e Fl 2
aϕ = (17)
Figure 7.  Diagram maximum stress–eccentricity for given vc 2 a 2atk
force F = const.
When the physically meaningful solution for vc is
value, but also in slope, so in the curve in Figure 7 there is inserted in equations (17) and (16), the apparent elonga-
C1 continuity in that point. Obviously, stress concentration tion takes the form
factor defined by equation (13) tends to infinity as e
approaches a.
2 2
e e
1 + 1 − 3  + 3 
Numerical simulation of tensile test Fl 2 a
  a
∆l * = (18)
with load eccentricity 2atk  2 
e
2 1 + 1 − 3   
In this section, let us try to computationally predict the out-  a 
come of the eccentric tensile test (force–displacement  
curve, ultimate force and elongation) based on the assumed
This can be squared to produce
parabolic tensile approximation, using the expressions
derived in the previous section. Let the reference ultimate
2
force be the one recorded by the usual symmetric tensile F  ∆l * 
test FM = 2atσM. The reduced ultimate force due to eccen- =  k ′ (19)
2at  l 
tricity, or as we refer to it apparent ultimate force FM* , fol-
lows from the condition that the maximum tensile stress be
equal to the ultimate stress σM. The apparent reduction of This means that the apparent stress–strain curve is again
ultimate force follows readily from equations (11) and parabolic, only instead of parameter k of equation (1) here
(13). In the case of small eccentricity e < a/2 it reads we have the corrected parameter k' due to eccentricity,
which has the form
FM
FM* = 2
e
2
 e  (14) e
2 + 3 − 1 − 3  1 + 1 − 3 
a a a
k ′ = 2k
 2 2
2 (20)
 e   e 
1 + 1 − 3   + 3   
In the case of large eccentricity e > a/2, the apparent
 a a 
ultimate force follows from equation (13) in the form  

2  e In the case of large eccentricity, apparent elongation

FM* = FM 1 − a  (15) (equation (16)) using equation (7) is obtained in the
3  
form

Note that the factor of ultimate force reduction would

3F
form a table similar to Table 1, in which the numbers ∆l * = 3l
would be reciprocal (inverse) to the factor of increase in  e  (21)
64atk 1 − 
maximum stress.  a
Šomođi et al. 7

models offer good balance between accuracy and simplicity

Table 3.  Relative values of k', F M* and ε M* in terms of for some woven fabrics. The second-order parabolic approxi-
eccentricity.
mation is applied in the example of stress–deformation analy-
e/a 0.2 0.4 0.6 0.8 sis where the uniaxial tensile stress is distributed unevenly in
the body. Tensile loading with deviation from symmetry
k'/k 0.915 0.7105 0.474 0.237
applied by means of a rigid crossbar is considered as the rep-
F M* /FM 0.617 0.4034 0.2667 0.1333
resentative example of such case. The analysis is performed
ε M* /εM 0.811 0.7535 0.75 0.75
using static and geometric conditions and the assumed
approximate material law. Unlike in solid bodies, where this
type of analysis is considered as combination of pure tension
and pure bending, in the case of large eccentricity the carrying
portion of the specimen is reduced due to the lack of compres-
sive stress capacity, which represents the separate source of
nonlinearity. The analysis can be subdivided into two separate
situations: at small eccentricity tensile stress is distributed
over the whole cross section, while at large eccentricity, the
carrying portion is reduced. Factor of maximum stress
increase theoretically tends to infinity as the relative eccen-
tricity approaches unity. In the load–elongation simulation of
tensile test with out-of-centre force, tensile diagrams indicate
the reduced apparent stiffness and reduced ultimate loads due
to eccentricity.
It should be noted that the present analysis does not pre-
tend to offer general solutions, but merely a specific solu-
Figure 8.  Computed curves for tensile test with tion for the case when parabolic approximation is adequate
eccentricities. (plain weave). Numbers in Table 1 suggest that trigonomet-
ric and sigmoidal approximations offer better overall fit –
This again produces the force–elongation relationship they could be included in the analysis in a similar way as
of the form equation (19), only now the corrected param- the considered parabolic approximation, with surely more
eter k' has the form elaborate mathematical background. Solutions obtained in
such a way could be considered closer to general.
32  e  One should also keep in mind the limitations due to the
k′ = k 1−
27  a 
(22) approximation of small rotations (equations (2) and (16)).
Remember that, for example, sinϕ = ϕ is in error by
under 5% at ϕ = 30o = 0.5236 rad and by over 20% at
The computed relative values of correction parameter, ϕ = 60o = 1.0476 rad. Therefore, the present analysis only
apparent breaking force and apparent ultimate strain are holds for moderate rotations consistent with limited exten-
given in Table 3 for four values of eccentricity. sibility – the unlimited growth of stress as relative eccen-
Computed curves relating tensile force and apparent tricity approaches unity (Figure 7) must anyway be stopped
elongation are shown in the tensile diagram of Figure 8. by the onset of breaking. Possible geometric redistribution
Note that at large eccentricities, apparent ultimate elonga- of load at large rotations is also neglected. Still, the authors
tion remains the same regardless of the value of eccentric- do believe that within these limitations the mechanical
ity, since ultimate force and parameter k are proportionally behaviour of fabric as ‘tension only material’ as opposed to
reduced, in fact, a simple calculation based on equations solid bodies is adequately described by the present model.
(15), (19) and (22) leads to εM*/εM = 0.75 (see Table 3).
This is similar to the recently reported result for simpler Declaration of conflicting interests
linear approximation in the tensile range.15
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Conclusion
Nonlinearity in tensile response of woven fabrics can be Funding
approximately described by a number of theoretical models. The author(s) disclosed receipt of the following financial support
Trigonometric and especially sigmoidal curves provide excel- for the research, authorship, and/or publication of this article:
lent accuracy in fitting experimental tensile curves, while This work has been fully supported by Croatian Science
considerably simpler linear and second-order polynomial Foundation under the project number IP-2018-01-3170.
8 Journal of Engineered Fibers and Fabrics 
References
1. Brnada S. Deformacije tkanina uvjetovane anizotropnošću
[Deformation of woven fabrics caused by anisotropy, in
Croatian]. PhD Thesis, Faculty of Textile Technology,
University of Zagreb, Zagreb, 2017.
2. Barburski M and Masajtis J. Modeling of the change in
structure of woven fabric under mechanical loading. Fibers
Text East Eur 2009; 17: 39–44.
3. Malik ZA, Malik MH, Hussain T, et al. Development of
models to predict tensile strength of cotton woven fabrics. J
Eng Fiber Fabr 2011; 6: 46–53.
4. Tehrani-Dehkordi M and Nosraty H. Tensile behavior simu-
lation of woven fabric with different weave pattern based on
finite element method. J Text Polym 2015; 3: 34–39.
5. Postle R, Carnaby GA and De Jong S. The mechanics of
wool structures. Chichester: Ellis Horwood, 1988.
6. Boisse Ph Gasser A, Hagege B and Billoet J-L. Analysis of
the mechanical behavior of woven fibrous material using
virtual tests at the unit cell level. J Mater Sci 2005; 40:
5955–5962.
7. Lomov SV, Boisse Ph, de Luycker E, et al. Full-field strain
measurements in textile deformability studies. Compos Part
A: Appl Sci Manuf 2008; 39: 1232–1244.
8. Rahnev I. Virtual energy evaluation of the fibers stress-
strain state. In: Proceedings of the 14th AUTEX world tex-
tile conference, Bursa, 26–28 May 2014.
9. Rahnev I. Probabilistic localization of the yield point into
the rheogram of the textile thread. In: Proceedings of the
15th AUTEX world textile conference, Bucharest, 10–12
June 2015.
10. Wu Y and Jiang C. Effect of load eccentricity on the stress-
strain relationship of FRP-confined concrete columns.
Compos Mater 2013; 98: 228–241.
11. Khashaba UA, Aldousari SM and Najjar IMR. Behavior of
[0]8 woven composites under combined bending and ten-
sion loading: part-I – experimental and analytical. J Compos
Mater 2012; 46: 1345–1355.
12. Wysmulski P and Debski H. Stability analysis of composite
columns under eccentric load. Appl Compos Mater 2019;
26: 9743–9748.
13. Seitl S, Rios JD, Cifuentes H, et al. Effect of the load eccen-
tricity on fracture behaviour of cementitious materials
subjected to the modified compact tension test. Solid State
Phenomena 2017; 258: 518–521.
14. Tang Y, Sun Z and Wu G. Compressive behavior of sustain-
able steel-FRP composite bars with different slenderness
ratios. Sustainability 2019; 11: 1118–1133.
15. Šomođi Ž, Brnada S, Zdraveva E, et al. Numerical and
experimental analysis of woven fabric in out-of-center ten-
sile loading. In: Proceedings of the 2nd international con-
gress of innovative textile – ICONTEX, Çorlu, 17–18 April
2019.