Composite Structures
journal homepage: www.elsevier.com/locate/compstruct
a r t i c l e
i n f o
Article history:
Available online 21 April 2012
Keywords:
Woven
Interlock
Analytical
Homogenization
Modeling
a b s t r a c t
An improved analytical modeling, three stages homogenization method 3SHM, of a 2.5D interlock
woven composite is proposed. The development of the analytical model based on mixed iso-strain and
iso-stress assembling models (the stiffness and the compliance averaging models) is presented. A nite
element (FE) modeling is carried out in order to use its results in the development of the analytical model.
It is proved that a model based only on an iso-strain condition could not give accurate results, while a
mixed iso-strain and iso-stress model yields more accurate estimations. the developed homogenization
method as well as the geometrical modeling, that takes into account the real geometry of undulated
yarns, proposed by the analytical model leads to very good agreement in comparing with results obtained
from FE models and available experimental data from the literature.
2012 Elsevier Ltd. All rights reserved.
1. Introduction
The use of 3D ber reinforced in high-tech industrial domains
(spatial, aeronautic, automotive, naval, etc.) has been expanded
in recent years. While the 2D ber reinforced composites, such
as 2D laminate composites and the 2D woven composites, show
very good in plane mechanical properties, they suffer from weakness in the out of plane direction, as known by the through thickness direction. The better physical, thermal and mechanical
properties, such as interlaminar shearing force, mechanical and
thermal stability along all three axes of space of 3D ber reinforced
fabrics, had given it more advantages over the unidirectional and
bi-dimensional composites. However, the 3D woven composites
can produce complex near-net shaped performs, also with a complex geometry they can be less expensive to produce. They show
better delamination resistance, damage tolerance, high impact
resistance and higher tensile strain-to-failure values. Also, 3D woven composites exhibit higher interlaminar fracture toughness
properties [13]. In addition, 3D weaving allows the tailoring of
properties for specic applications. 3D woven composites are divided into two main types: the 3D orthogonal fabrics and the
2.5D angle interlock fabrics (Fig. 1). The latter which are the main
objective of this work could be divided into two types: throughthe-thickness fabrics and the layer-to-layer fabrics.
Corresponding author.
E-mail address: rac.younes@lsis.org (R. Younes).
0263-8223/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compstruct.2012.03.019
This kind of composite is characterized by its complex architecture. The unit cell contains warp weaver and weft yarns interlaced
at 90 in the (xy) plane (Fig. 1). The 2.5D angle interlock woven
composite involves the binding of straight warp yarns by interlocking warp yarns. Warp yarns can be bound to different depth where
various arrangements of yarns placement which can be used to
produce a wide range of these types of composites. The throughthe-thickness fabric is a multilayered fabric in which warp weavers
travel from one surface of the fabric to the other, holding together
all the layers. The layer-to-layer fabric is a multilayered fabric in
which warp weavers travel from one layer to the adjacent layer,
and back. A set of warp weaves together hold all the layers of the
fabric. In addition of the complex geometry, the volume fraction
of bers, volumes of yarns and the inclination angle of the undulated warp yarns allow the tailoring of properties for specic applications. In other words, designers could produce the optimal
perform of fabrics for a required mechanical properties [4].
While experimental test made to determine the effective elastic
properties Youngs and Shear moduli (Ex, Ey, Ez, Gxy, Gxz, Gyz) and
Poissons ratios (mxy, mxz and myz) of composites are expensive in
terms of money and time, analytical and numerical modeling could
present a good alternative solution for industries. In which concern
numerical modeling, the obtained results are supposed to be more
accurate. These methods have the disadvantages of the time consumed on the tasks of dening the geometry of the composite,
meshing and the calculation procedures, besides the need of super
computers to make the computation. Analytical modeling presents
more exible and easier tool in terms of geometric modeling and
3010
Fig. 1. Different types of 3D woven fabrics: layer-to-layer angle interlock, through-the-thickness angle interlock and 3D orthogonal woven composites.
3011
Table 1
Mechanical properties of carbon bers and matrix [9]. Carbon bers are assumed to
be transversly isotropic material which gives the following assumption:
Ef33 Ef22 ; Gf13 Gf12 ;
Carbon bers
v f13 v f12
and Gf23
E22
f
21v 23
Ef11 (GPa)
Ef22 (GPa)
Gf12 (GPa)
v f12
v f23
T300-J
230
15
50
0.278
0.3
Resin RTM6
Em (GPa)
vm
2.89
0.35
3012
Fig. 3. Perspective and Top views of the REV of three composites: (a) composite-71, (b) composite-H2 and (c) composite-69.
have a mean inclination angle h equal to 24. The weft yarns have
linear longitudinal part while its cross-section is a attened ellipse.
The REV of composite-71 (Figs. 2 and 3) has 12 warp yarns and
12 weft yarns. Warp and weft yarns have similar geometrical shape
of those for composite-H2. Warp yarns have the mean inclination
angle h equal to 12 of the undulated part of warp yarns. Both warp
and weft yarns have attened elliptical cross-section.
The REV of composite-69 (Figs. 2 and 3) contains 6 warp yarns
and 6 weft yarns. While warp yarns have a mean inclination angle
h equal to 29 for the longitudinal part of weft yarns is linear. Both
warp and weft yarns have attened elliptical cross-section.
The geometric parameters (Fig. 4) needed in the geometrical
modeling of the numerical and analytical models are dened as
follow:
Lx, Ly and Lz: the length, width and thickness of the REV.
aw and af: are the widths of the warp and weft yarns.
hw and hf: the thickness of the warp and weft yarns.
V fwc
V ffc
V fwy V w
Vc
V ffy V f
Vc
V fc V fwc V ffc
where
V fw=fy the bers volume fraction in warp or weft yarns.
Fig. 4. Geometrical parameters on the REV and the transversal section (cross-section) of the warp and weft yarns.
3013
Table 3
Fiber volume fractions of warp and weft yarns and of the composite, for composites
71, H2 and 69.
Fiber volume fraction
Composite-71
Composite-H2
Composite-69
Warp yarns
Weft yarns
REV
27.94%
11.04%
38.98%
17.92%
21.68%
39.60%
31.62%
13.48%
42.08%
U 1 x; a2 ; z U 1 x; a2 ; z 0;
U 2 x; a2 ; z U 2 x; a2 ; z 0;
U 3 x; a2 ; z U 3 x; a2 ; z 0
On the faces z = a3 we used tensor notation for stain, as
follows:
U 1 x; y; a3 U 1 x; y; a3 0;
U 2 x; y; a3 U 2 x; y; a3 0;
U 3 x; y; a3 U 3 x; y; a3 0
With U2 = U3 = 0 at the centre of the face x = a1.
Table 2
Geometric properties of the composites-71, H2 and 69 [9], in addition to the calculated mean inclination angle of undulated parts of warp yarns.
Composite
71
H2
69
Width
aw (mm)
Warp yarns
Thickness
hw (mm)
Weft yarns
Width
af (mm)
Thickness
hf (mm)
Weft yarns
Cl (mm)
Warp yarns
Ct (mm)
h ()
2.8 0.123
1.9 0.11
4.4706 0.3533
0.49 0.059
0.18 0.03
0.6471 0.0133
3.5 0.288
1.7 0.06
3.163 0.427
0.35 0.037
0.28 0.03
0.5882 0.1084
2.822 0.2007
0.4 0.12
2.8235 0.5087
0 0.069
0.2 0.08
0.00 0.06
11.61
24.09
29.21
3014
Fig. 5. Meshing of composites: (a) composite-71, (b) composite-H2 and (c) composite-69.
U 1 a1 ; y; z U 1 a1 ; y; z 0;
U 2 a1 ; y; z U 2 a1 ; y; z 0;
U 3 a1 ; y; z U 3 a1 ; y; z 0
On the faces y = a2 we used tensor notation for stains, as
follows:
U 1 x; a2 ; z K; U 1 x; a2 ; z 0;
Ex
F x a1
2 a2 a3 K
Gxy
F x a2
2 a1 a3 K
U 2 x; a2 ; z U 2 x; a2 ; z 0;
U 3 x; a2 ; z U 3 x; a2 ; z 0
On the faces z = a3 we used tensor notation for stain, as
follows:
U 1 x; y; a3 U 1 x; y; a3 0;
U 2 x; y; a3 U 2 x; y; a3 0;
U 3 x; y; a3 U 3 x; y; a3 0
With U2 = U3 = 0 at the centre of the face y = a2.
Where Ui (i = 1, 2, 3) are the displacement components along x,
y and z directions respectively, and K an arbitrary constant. The
constraints described by thus equations can be applied using constraint equations in ANSYS. However, periodicity conditions such
as described by thus equations are not easy to apply to FEM discretizations because nodes on opposite faces of the RVE cannot
be found in pairs with two identical coordinates (yz on warp faces
and xz on ll faces), but rather they are arbitrarily located as dictated by the free-mesh generation process.
5. Analytical modeling
The main objective of the proposed analytical model is to nd a
simple modeling method that could deal with woven composites,
especially the angle interlock type, giving accurate results in comparing with those numerically and experimentally obtained. The
proposed analytical model is based on geometrical modeling that
takes into account the undulated shape of yarns and a homogenization method based on mixed iso-strain and iso-stress assumption. A description of the geometric modeling is given in the
following section. In the homogenization method section an extensive work is made in order to develop a new assembly homogenization scheme. Numerical Finite Element modeling results are used
in order to develop and improve the analytical model. It will be
shown that a homogenization method based on both iso-strain
and iso-stress assumption, or a combined parallelseries model,
could give better results. In addition involving the effect of the
mean inclination angle of warp yarns could improve the prediction
of the longitudinal Youngs modulus.
5.1. Geometrical modeling
3015
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Composite-71
Composite-H2
Composite-69
40.20
25.51
28.86
33.07
58.76
41.06
7.82
7.78
9.37
2.75
2.76
2.83
2.97
2.87
4.65
2.19
2.36
1.95
0.058
0.033
0.085
0.463
0.475
0.437
0.506
0.435
0.490
Table 5
Finite Element numerical results of three composites 71, H2 and 69, using Chamis micromechanical model.
Numerical results
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Composite-71
Composite-H2
Composite-69
40.70
25.93
28.98
31.21
54.82
37.23
7.57
7.87
8.99
3.14
3.22
3.41
2.93
2.94
4.81
2.21
2.24
2.16
0.065
0.027
0.051
0.411
0.427
0.238
0.508
0.468
0.350
Fig. 7. Von-Mises stress. (a) Tensile traction along x direction of composite-H2. (b) Tensile traction along y direction of composite-H2. (c) Tensile traction along x direction of
composite-71. (d) Tensile traction along y direction of composite-71.
Aw=f
8
2
Zx a cos
where
2px
T
3016
V undw Aw Lundw
The volume for each micro volume:
V iundw Aw Liundw
With the length of the centerline of the undulated part:
Lundw
T=2
Lundw
s
2
dz
1
dx
dx
s
2
Z T
dz
1
dx
dx
T=2
With
2
dz
dx
2
a 2Tp sin 2Tpx .
Liundw
frc g
1
Vc
ZZZ
frc g dV
With {rc} is the true stress tensor in is the composite of volume Vc.
The average stress tensor can be expressed in terms of the average stresses tensors in sub-volumes:
frc g
n
1X
fr1 g V 1 fr2 gV 2 frn g V n
Vc 1
s
2
Z xi1
dz
1
dx
dx
xi
V linw Aw Llinw
With Llinw is length of linear part.
Thus the volume of the warp yarn is calculated from the volumes of left and right undulated parts and the linear part.
where fec g; fe1 g; fe2 g; . . . ; fen g are the average strain in the composites and the n sub-volumes, and [Cc], [C1], [C2], . . . , [Cn] are the effective stiffness matrices of the composite and the n sub-volumes.
The assumption of iso-strain condition throughout the composite fec g fe1 g fe2 g fen g leads to the following equation
known by the Voigt model
Pin
5.1.2. Volume of weft yarn
The volume of the linear part of weft yarns is:
V linf Af Llinf
With Llinf is length of linear part.
C cVoigt
i1 C i
Vi
Vc
Fig. 9. undulated part simulated by a sinusoidal function and subdivided into sub-volumes.
3017
1
fec g
Vc
ZZZ
fec g dV
With {ec} is the true strain tensor in the composite of volume Vc.
The average strain tensor can be expressed in terms of the average strain tensors in sub-volumes:
n
1X
fec g
fe1 g V 1 fe2 gV 2 fen g V n
Vc 1
Pin
ScReuss
i1 Si
Vi
Vc
C xyz T c C 123 T 0c
Sxyz T s S123 T 0s
In a rst attempt in the development procedure, the iso-strain
model is used to evaluate the effective elastic properties of the
composite. In other words, the sub-volumes of undulated warp
yarns, the weft yarns and the matrix are assembled in parallel.
Thus, the stiffness matrix of the woven composite is given by the
Voigt-model:
Pin
C cVoigt
i1 C ixyz
Vi
Vc
C C Pc C CVoigt 1 Pc S1
CReuss
where
Pc is a macro weighting parameter vary between 0 and 1,
Pn
Pn
C V
i1 wi wi
Vw
C v oigt
S V
i1 wi wi
SReuss
Vw
Cf V f Cm V m
Vc
Sf V f Sm V m
Vc
3018
Table 6
Analytical results of the iso-strain model compared to numerical results for the three composites 71, H2 and 69.
Effective elastic properties
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Composite-71
Numerical
Iso-strain model
Percentage difference
40.2
61.22
52.29%
33.07
30.60
7.47%
7.82
8.19
4.73%
2.75
2.92
6.18%
2.97
4.63
55.89%
2.19
2.27
3.65%
0.058
0.031
46.55%
0.463
0.693
49.68%
0.506
0.529
36.96%
Composite-H2
Numerical
Iso-strain model
Percentage difference
25.51
33.90
32.28%
58.76
54.35
7.51%
7.78
8.68
11.57%
2.76
2.93
6.16%
2.86
5.75
101.05%
2.36
2.58
9.32%
0.033
0.014
57.58%
0.475
0.778
63.79%
0.435
0.445
2.3%
Composite-69
Numerical
Iso-strain model
Percentage difference
28.86
39.67
37.46%
41.06
37.06
9.74%
9.37
10.61
13.23%
2.83
2.84
0.35%
4.65
10.78
131.82%
1.95
2.438
25.02%
0.085
0.0136
84%
0.437
0.92479
111.62%
0.49
0.375
23.47%
on well modeling of the warp weaver yarn. thus, a more sophisticated approach for assembling sub-volumes in parallel and in series
could lead to better predictions for these and the other elastic properties, which will lead to the next development procedure of the
model.
5.2.4. 3SHM (3 Stages Homogenization Method)
After analyzing analytical results in the previous section, some
enhanced remarks could be summarized as follows:
Using a mixed VoigtReuss model yield good results.
For Ex, Pc value for best agreement P ba
c increases with h, while
it decreases for Ez.
Better evaluation of the stiffness of the warp yarn is thought to
give better results.
Ey is better predicted with only Voigt model.
Based on these remarks, a better evaluation of the stiffness of
the warp weaver yarn, by assembling of sub-volumes using mixed
Voigt and Reuss models, is believed to yield more precise evaluation of all effective elastic properties of the angle interlock composite. A homogenization scheme, the 3SHM, is then suggested and
presented in this section.
In the 3SHM, the homogenization scheme is done in three
stages: microlevel stage, mesolevel stage and macrolevel stage. In
the rst stage the stiffness matrices of sub-volumes, of sub-divided
yarns, in the local coordinate system are calculated. The stiffness
matrices of warp and weft yarns as well as the stiffness of the matrix are evaluated in the global coordinate system in the second
stage. At the third stage, the stiffness of the composite is evaluated
Table 7
Analytical results of the mixed VoigtReuss model for composite-71 in terms of Pc compared to numerical results.
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Numerical
40.2
33.07
7.82
2.75
2.972
2.192
0.058
0.463
0.506
Pc
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
P ba
c for best agreement
2.924
2.834
2.745
2.655
2.565
2.475
2.385
2.295
2.205
2.115
2.025
0.8
4.631
4.3636
4.095
3.827
3.558
3.290
3.022
2.753
2.485
2.217
1.949
0.4
2.2686
2.2198
2.171
2.122
2.0736
2.024
1.976
1.927
1.878
1.830
1.781
0.8
0.0309
0.0364
0.04303
0.05102
0.0609
0.07355
0.09014
0.1129
0.1461
0.1989
0.2961
0.6
0.693
0.6777
0.6605
0.6420
0.6218
0.5995
0.5743
0.5449
0.5089
0.4611
0.3876
0.1
0.5293
0.5287
0.5278
0.5264
0.5245
0.5217
0.5175
0.5108
0.4994
0.4772
0.4188
0.3
1.59%
0.18%
1.68%
0.96%
5%
0.41%
0.95%
Composite-71
7.47%
0.9%
3019
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Composite-H2
Numerical
25.51
58,76
7,78
2.76
2.866
2.362
0.033
0.475
0.435
Pc
1
0
Best Pc
Result for best Pc
Percentage difference
8.68
4.96
0.7
7.85
0.87%
2.93
2.04
0.8
2.75
0.36%
5.75
1.890
0.3
3.049
6.39%
2.58
1.903
0.7
2.377
0.64%
0.014
0.260
0.6
0.0321
2.73%
0.778
0.413
0.1
0.509
7.2%
0.445
0.389
0.1
0.429
1.4%
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Composite-69
Numerical
28.86
41.06
9.37
2.83
4.65
1.95
0.085
0.437
0.49
Pc
1
0
Pc
Result for best Pc
Percentage differences
10.61
5.39
0.7
9.25
1.28%
2.84
2.18
1
2.84
0.35%
10.78
2.27
0.3
4.82
3.66%
2.438
1.95
0
1.95
0%
0.0136
0.2855
0.3
0.0733
13.76%
0.92479
0.408
0.1
0.534
6.64%
0.375
0.4109
0.2
0.435
11.22%
Table 9
Comparison of P ba
c for the three composites 71, H2 and 69.
Composite
results
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
h ()
Composite-71
Num.
40.2
0.6
33.07
1
7.82
0.8
2.75
0.8
2.972
0.4
2.192
0.8
0.058
0.6
0.463
0.1
0.506
0.3
11.61
1.59%
7.47%
0.9%
0.18%
1.68%
0.96%
5%
0.41%
0.95%
25.51
0.7
58.76
1
7.78
0.7
2.76
0.8
2.866
0.3
2.362
0.7
0.033
0.6
0.475
0.1
0.435
0.1
6.4%
7.51%
0.87%
0.36%
6.39%
0.64%
2.73%
7.2%
1.4%
28.86
0.7
41.06
1
9.37
0.7
2.83
1
4.65
0.3
1.95
0
0.085
0.3
0.437
0.1
0.49
0.2
4.95%
9.74%
1.28%
0.35%
3.66%
0%
13.76%
6.64%
11.22%
P ba
c
Diff.
Composite-H2
Num.
P ba
c
Diff.
Composite-69
Num.
P ba
c
Diff.
h
i h
i1
C 123
S123
wi
wi
h
i h
i1
C 123
S123
f
f
1=E11
6 m =E
6 12 11
6
6 m12 =E11
S 6
6
0
6
6
4
0
0
m12 =E11
m12 =E11
1=E22
m23 =E22
m23 =E22
1=E22
0
0
0
0
1=G23
0
0
1=G12
0
0
1=G12
3
7
7
7
7
7
7
7
7
5
29.21
0
0
0
0
E
m
m
6
6
6
6
0
0
0
0
Em =21 v m
4
0
24.09
0
0
0
0
0
3
7
7
7
7
7
7
7
7
7
7
7
7
7
5
Em =2 1 v m
5.2.4.4. Warp yarn stiffness. The warp yarns in the angle interlock
woven composites have left and right undulated parts and often
an additional linear longitudinal parts (Fig. 11a).
The stiffness of the warp yarn is evaluated using iso-strain and
iso-stress assumptions when assembling these sub-volumes in
parallel or in series respecting the following scheme:
3020
C xyz
wLund
C xyz
wRund
Pn
i1 V wLundi
V Lundw
Pn
i1 V wRundi
C xyz
w
C xyz
wRundi
xyz
2 V wund C xyz
wund V wlin C wlin
V warp
2 Pn
C xyz
wLundi
6
2 V wund 6
4
C xyz
w
xyz
V
C
i1 wLundi wLundi
V wund
1 Pn
xyz
V
C
i1 wRundi wRundi
V wund
1 31
7
7
5
V wlin C xyz
wlin
V warp
V Rwund
Sxyz
wund
4. In the nal step, the homogenized undulated part (left + right) and the linear parts form a parallel system when
assembled along the warp direction (Fig. 11.d). The Voigt
model is used to evaluate the stiffness of the system; undulated (left + right) and linear parts.
1
1
C xyz
C xyz
wLund
wRund
2
1
xyz
C xyz
wund Swund
5.2.4.5. Weft yarn stiffness. The stiffness of the weft yarns is easier
to be evaluated considering their straight prole. The stiffness of
weft yarns in the global coordinate system is given in terms of that
in the local coordinate system, which has been already calculated
in the micro level homogenization stage.
C xyz
T c C 123
T tc
f
f
Noting that in some cases, weft yarns could have a quite low
undulated prole. In this case, the assembling of sub-volumes is
made along the x direction, where the sub-volumes are assumed
to form a parallel system. Voigt model is used to evaluate the stiffness of the weft yarns.
5.2.4.6. Stage 3 Macrolevel homogenization. At the macro level
homogenization stage the stiffness of the composite is evaluated
3021
Method
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Composite-71
Numerical
Analytical
Percentage difference
40.2
43.944
9.31%
33.07
30.471
7.86%
7.82
8.190
4.73%
2.75
2.913
5.93%
2.972
3.142
5.72%
2.192
2.261
3.15%
0.058
0.0636
9.66%
0.463
0.5194
12.18%
0.506
0.5249
3.74%
Composite-H2
Numerical
Analytical
Percentage difference
25.51
27.393
7.38%
58.76
53.913
8.25%
7.78
8.497
9.22%
2.76
2.907
5.33%
2.866
3.134
9.35%
2.362
2.566
8.64%
0.033
0.039
18.18%
0.475
0.512
7.79%
0.435
0.474
8.97%
Composite-69
Numerical
Analytical
Percentage difference
28.86
31.72
9.91%
41.06
35.99
12.35%
9.37
9.317
0.57%
2.83
2.79
0.141%
4.65
4.14
10.97%
1.95
2.40
23.08%
0.085
0.0716
15.76%
0.437
0.451
3.2%
0.49
0.481
1.84%
Table 11
Analytical results of the 3SHM, using Chamis micromechanical model, in comparing with numerical results for three composites 71, H2 and 69.
Composite
Method
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Composite-71
Numerical
Analytical
Percentage difference
40.7
44.94
10.42%
31.21
30.49
2.31%
7.57
6.95
8.19%
3.14
3.21
2.23%
2.936
3.42
16.49%
2.21
2.55
15.38%
0.065
0.059
9.23%
0.411
0.414
0.73%
0.508
0.375
26.18%
Composite-H2
Numerical
Analytical
Percentage difference
25.93
27.54
6.21%
54.82
53.92
1.64%
7.87
7.23
8.13%
3.22
3.212
0.25%
2.94
3.379
14.93%
2.242
2.868
27.92%
0.027
0.037
37.04%
0.427
0.3616
15.32%
0.468
0.3691
21.13%
Composite-69
Numerical
Analytical
Percentage difference
28.98
31.81
9.77%
37.23
36.07
3.12%
8.99
7.98
11.23%
3.41
3.15
7.62%
4.81
4.41
8.32%
2.16
2.78
28.7%
0.051
0.058
13.73%
0.238
0.298
25.21%
0.35
0.324
7.43%
Fig. 13. The assembling of sub-volumes of warp yarn is affected by the mean inclination angle
h.
warp and weft yarns. Knowing the volumes and the stiffness matrices of each phase in the xyz system, the stiffness matrix of the composite could be calculated. The three volumes are assumed to
3022
Table 12
Analytical results, for the 3SHM with Pcw using GSCM micromechanical model, in comparing with numerical results for composite-71.
Composite-71
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Numerical
Pcw = 1
Pcw = 0.9
Pcw = 0.8
Pcw = 0.7
Pcw = 0.6
Pcw = 0.5
Pcw = 0.4
Pcw = 0.3
Pcw = 0.25
Pcw = 0.2
Pcw = 0.1
Pcw = 0
Percentage difference
40.2
43.944
43.453
42.962
42.472
41.981
41.491
41.001
40.510
40.282
40.020
39.529
39.039
0.20%
33.07
30.471
30.469
3.046
30.466
30.465
30.463
30.461
30.459
30.463
30.458
30.4561
30.454
7.88%
7.82
8.190
8.183
8.177
8.170
8.162
8.155
8.148
8.140
8.377
8.132
8.1242
8.115
7.12%
2.75
2.913
2.913
2.912
2.912
2.912
2.912
2.912
2.912
2.910
2.912
2.9119
2.9117
5.82%
2.972
3.142
3.107
3.072
3.037
3.002
2.967
2.932
2.896
2.870
2.861
2.8253
2.789
3.43%
2.192
2.261
2.260
2.260
2.260
2.260
2.260
2.260
2.260
2.238
2.260
2.260169
2.260
2.09%
0.058
0.0636
0.06372
0.06380
0.06388
0.0639
0.0640
0.0641
0.0642
0.0650
0.0642
0.06436
0.06444
12.07%
0.463
0.5194
0.5197
0.5201
0.5205
0.5209
0.5213
0.5218
0.5222
0.5358
0.5226
0.5230
0.5234
15.72%
0.506
0.5249
0.5247
0.5244
0.5242
0.5240
0.5237
0.5235
0.5232
0.5420
0.5230
0.5227
0.5225
7.11%
Table 13
Analytical results, for the 3SHM with Pcw using GSCM micromechanical model, in comparing with numerical results for composite-H2.
composite-H2
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Numerical
Pcw = 1
Pcw = 0.45
Pcw = 0
Percentage difference
25.51
27.393
25.472
23.896
0.15%
58.76
53.913
53.899
53.883
8.27%
7.78
8.497
8.357
8.221
7.42%
2.76
2.907
2.905
2.903
5.25%
2.866
3.134
2.908
2.675
1.46%
2.362
2.566
2.564
2.563
8.55%
0.033
0.039
0.03836
0.0373
16.24%
0.475
0.512
0.5291
0.5428
11.39%
0.435
0.474
0.4741
0.4742
8.99%
Table 14
Analytical results, for the 3SHM with Pcw using GSCM micromechanical model, in comparing with numerical results for composite-69.
Composite-69
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Numerical
Pcw = 1
Pcw = 0.5
Pcw = 0
Percentage difference
28.86
32.28
28.77
25.21
0.31%
41.06
36.08
36.05
35.99
12.20%
9.37
9.45
9.11
8.69
2.77%
2.83
2.84
2.83
28.30
0%
4.65
4.215
3.85
3.33
17.20%
1.95
2.43
2.43
2.42
24.61%
0.085
0.0725
0.066
0.0611
22.35%
0.437
0.450
0.495
0.5409
13.27%
0.49
0.481
0.484
0.489
1.22%
Table 15
Analytical results, for the 3SHM with Pcw using Chamis micromechanical model, in comparing with numerical results for composite-71.
Composite-71
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Numerical
Pcw = 1
Pcw = 0.1
Pcw = 0
Percentage differences
40.7
44.958
40.73
40.27
0.07%
31.21
30.490
30.481
30.48
2.33%
7.57
6.973
6.939
6.93
8.33%
3.14
3.212
3.210
3.210
2.22%
2.936
3.417
3.110
3.075
5.93%
2.21
2.549
2.548
2.548
15.29%
0.065
0.060
0.05975
0.05971
8.08%
0.411
0.417
0.4136
0.4132
0.63%
0.508
0.379
0.37940
0.3793
25.31%
Table 16
Analytical results, for the 3SHM with Pcw using Chamis micromechanical model, in comparing with numerical results for composite-H2.
Composite-H2
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Numerical
Pcw = 1
Pcw = 0.5
Pcw = 0
Percentage differences
25.93
27.556
25.868
24.177
0.24%
54.82
53.92
53.91
53.90
1.66%
7.87
7.25
7.173
7.073
8.85%
3.22
3.211
3.209
3.207
0.34%
2.94
3.373
3.183
2.946
8.26%
2.242
2.864
2.863
2.861
27.70%
0.027
0.03701
0.0361
0.0351
33.70%
0.427
0.366
0.380
0.395
11.01%
0.468
0.372
0.372
0.373
20.51%
Table 17
Analytical results, for the 3SHM with Pcw using Chamis micromechanical model, in comparing with numerical results for composite-69.
Composite-69
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Numerical
Pcw = 1
Pcw = 0.6
Pcw = 0
Percentage differences
28.98
31.836
29.199
25.198
0.756%
37.23
36.07528
36.063
36.035
3.13%
8.99
8.0056
7.821
7.457
13.00%
3.41
3.155
3.152
3.147
7.56%
4.81
4.409
4.146
3.586
13.80%
2.16
2.780
2.777
2.77
28.56%
0.051
0.0584
0.0552
0.0504
8.23%
0.238
0.3018
0.338
0.3940
42.02%
0.35
0.3282
0.3304
0.336
5.60%
3023
represent a parallel system when they are assembled along x direction (Fig. 12). The Voigt model is used to evaluate the stiffness of
the composite as follows:
C REV
V w Cw V f Cf V m Cm
V REV
S11
6
6 S12
6
6
6 S13
6
SREV 6
6 0
6
6
6 0
4
0
2
S12
S13
S22
S23
S23
S33
S44
S55
1=Ex
6
6 myx =E2
6
6
6 mzx =Ez
6
6
6
0
6
6
6
0
4
1
;
S11
v xy
Gxy
Ey
S12
;
S11
1
;
S66
7
0 7
7
7
0 7
7
7
0 7
7
7
0 7
5
S66
mxz =Ex
1=Ey
myz =Ey
mzy =Ez
1=Ez
1=Gyz
1=Gxz
1=Gxy
1
;
S22
v xz
Gxz
mxy =Ex
0
Ex
1
;
S55
Ez
S13
;
S11
3
7
7
7
7
7
7
7
7
7
7
7
5
1
;
S33
v yz
Gyz
S23
;
S22
1
;
S44
5.2.4.7. Comparison with numerical modeling results. Results of analytical modeling using the 3SHM model of the three angle interlock
composites described in the material section are compared to
those of numerical modeling. The results are evaluated with a
two different micromechanical models; the GSCM of Christensen
and Chamis model.
Tables 10 and 11 show good agreement between analytical and
numerical results. Concerning the Young moduli, the percentage
difference is around 10%. The shear moduli and the Poissons ratio
are also very well estimated with a small error, in some cases below 5%. It is shown that micro models could have an important
inuence on predicting the elastic properties. With Chamis model
Ey is better estimated, while shear moduli are better evaluated
using the GSCM model.
The results (Tables 10 and 11) show that the longitudinal Young
modulus Ex is always overestimated while the transversal modulus
Ey is always underestimated. It is believed that, the overestimation
of Ex is due of only using Voigt model in the homogenization of
sub-volumes of the undulated parts. This case is discussed on the
next section where an improvement of the proposed model is
suggested. Moreover, its shown that when Chamis model is used,
Table 18
Best agreement weighting parameter for weaver warp yarn Pba
cw in terms of the
mean inclination angle
h for GSCM and Chamis micromechanical models.
h ()
P ba
cw (GSCM)
0
0
11.61
0.25
24.09
0.45
29.21
0.5
90
1
P ba
cw (Chamis)
0.1
0.5
0.6
Average P ba
cw
0.175
0.5
0.55
1
xyz
xyz
i1 V wLundi C wLundi
i1 V wRundi SwLundi
C xyz
1 P cw
wLund P cw
V Lundw
V Rundw
C xyz
wRund P cw
Pn
xyz
i1 V wRundi C wRundi
V Rundw
1 P cw
Pn
1
xyz
i1 V wRundi SwRundi
V Rundw
3024
Fig. 14. Best agreement weighting parameter for weaver warp yarn P ba
cw in terms of the mean inclination angle h for GSCM and Chamis micromechanical models.
Table 19
Fiber volume fractions of warp, stuffer and weft yarns and the composite, for
composite LTL1.
Warp
Stuffers
Weft
Composite
4.9%
25.6%
27.2%
57.6%
Pba
cw 0:000137;
h2 0:023724;
h 0:025432:
yarns have straight prole while the warp weaver yarns have an
undulated prole. While the direct values of the geometrical
parameters of yarns, the width and the thickness, are not given.
The available data of lineal spacing of yarns and the macro-cell
dimensions [7,8] are used to derive the needed dimensions. Knowing that the composite contains 5 stuffer, 6 warp weaver and 4
weft yarns in the thickness direction and by determining of number of yarns in Ly and Lx directions, a geometrical representation of
the composite is given (Fig. 15).
Ns = W Sl = 1; there is approximately 1 stuffer in a distance W.
Nw = W Wl = 0.4; there is approximately a half warp weaver
yarn in a distance W.
Nf = P Fl = 1.8; there is approximately 2 weft yarns in a distance
P.
where Ns, Nw and Nf are the numbers of stuffer, warp and weft
yarns in Ly and Lx directions respectively. P and W are the Pitch and
Width of the macro-cell given in [7,8]. Sl, Wl and Fl are the lineal
spacing of Stuffer, warp and weft yarns respectively [7,8]. Then,
the dimensions of the REV dened by the length Lx (along stuffer
direction) and Ly (along weft direction) and Lz (the thickness direction) are determined (Fig. 15).
The thickness of yarns are assumed to be equal:
hs hf hw
with
3025
aw
hw
as
hs
af
hf
Lx
Ly
Lz
V fw=y
V fs=y
V ff =y
Case A
Case B
0.74
0.54
0.45
0.45
1.1
1.2
0.45
0.45
1.5
1.5
0.45
0.45
3
3
1.47
1.47
4.064
4.064
0.6880
0.8915
0.8664
0.7511
0.6879
0.6879
Table 21
Mechanical properties of carbon ber and epoxy.
Carbon bers
Ef11 (GPa)
Ef22 (GPa)
Gf12 (GPa)
v f12
v f23
Hercules-AS4
235
17
55
Em (GPa)
vm
0.25
0.27
3.45
0.3
V fw;s;f =y
V fw;s;f V c
V w;s;f
where
hf hs Lz
T9 (T is the thickness of the composite [7,8]).
9
for weft yarns the width af is calculated as follows:
af 2p (assuming that the spacing between adjacent weft yarns
is negligible).
The problem encountered in the determination of the exact
geometrical parameters, width values of the stuffer and warp
yarns, makes necessary to take some assumptions. Noting that as
and aw are related with the width of the REV by this relation:
W as 1=2aw
Ly as aw
aw Ly as
Different values have been suggested for both as and aw in such
a way that conserves the ber volume fraction of the yarns in the
REV as calculated by Chou, and in the same time gives reasonable
Table 22
Comparison between analytical and experimental results for composite LTL1 for the two cases A and B.
Composite-LTL1
Ex (GPa)
Ey (GPa)
Exp. [8]
Analytical [8]
Iso-strain model case A
MT case A
3SHM Case A
Percentage difference (3SHM case A vs Exp)
Iso-strain model case B
MT case B
3SHM Case B
Percentage difference (3SHM case B vs Exp)
71.32
73.73
69.56
78.96
68.08
4.54%
69.72
80.89
69.09
3.13%
66.33
65.64
69.91
84.87
70.18
5.80%
69.79
87.22
69.97
5.20%
Ez (GPa)
11.38
12.15
10.59
11.29
12.66
10.58
Gxy (GPa)
7.74
6.97
5.64
8.76
7.75
0.13%
5.28
8.28
6.56
15.24%
Gxz (GPa)
Gyz (GPa)
6.92
5.28
6.86
4.33
3.44
4.94
6.506
5.17
5.59
4.32
3.44
4.89
vxy
0.046
0.080
0.028
0.049
0.037
19.56%
0.029
0.046
0.038
17.39%
vxz
vyz
0.433
0.56
0.297
0.258
0.01
0.304
0.426
0.461
0.267
0.255
0.19
0.299
Table 23
Comparison between analytical and experimental results for composite 71.
Composite-71
Ex (GPa)
Ey (GPa)
Ez (GPa)
Gxy (GPa)
Gxz (GPa)
Gyz (GPa)
vxy
vxz
vyz
Exp.
Iso-strain model
MT
Analytical Chamis (3SHM)
Percentage difference (3SHM vs Exp.)
49.02
61.22
69.59
49.04
0.04%
35.28
30.60
37.86
33.75
4.33%
8.19
6.79
7.323
2.92
5.62
3.498
4.63
3.26
3.379
2.27
2.30
2.73
0.081
0.031
0.015
0.056
29.62%
0.693
1.061
0.408
0.529
0.185
0.372
Table 24
Sensitivity of effective elastic stiffness constants with models 1 and 2.
Composite
Model
DC11
DC22
DC33
DC44
DC55
DC66
Composite-71
Model 1
Model 2
0.137
0.009
0.07
0.003
0.02
0.02
0.02
0.015
0.09
0.010
0.032
0.02
Composite-H2
Model 1
Model 2
0.117
0.019
0.080
0.000006
0.031
0.0003
0.028
0.00007
0.134
0.01
0.032
0.00009
3026
7. Conclusion
In this paper, the development of an analytical model was
presented. It is shown that an analytical model based only on
an iso-strain assumption for the whole composite will not yield
the best results, especially with the angle interlock types. The
contribution of the iso-stress assumption in the homogenization
method proved to be crucial in evaluating the effective elastic
properties. A geometrical modeling with sinusoidal function is
adopted to describe the undulated yarns. Finite Element numerical modeling is employed for three angle interlock woven composites to assist the development of the homogenization method
in the proposed analytical model. A mixed iso-strain and isostress model is used at rst, to show the signicant inuence
of the iso-stress assumption that could affect the estimation of
the elastic properties. The comparison between numerical and
analytical results reveals that better evaluation of the stiffness
of the warp yarn is vital in estimating Ex, while using different
micromechanical models could affect, to a certain extent, the
estimation of Ey and shear moduli. Thus, a homogenization
method, the three stages homogenization method 3SHM, is
developed and proposed. An assembling scheme, in parallel or
DC
DC
1
P c e C cVoigt 1 P c e S1
CReuss P c C cVoigt 1 P c SCReuss
DC
C non perturbed
e C cVoigt S1
CReuss
C non
perturbed
02h
2V wund
V REV
B6
@4
i1 31
i1 h
cxyz
wLundp
C xyz
7
5
wRundp
DC
C non
1
xyz 1 xyz 1
1
cwLund C wRund
C
A
2
perturbed
with
C xyz
wLund P cw
C xyz
wLund P cw
Pn
xyz
i1 V wLundi C wLundi
V Lundw
1 P cw
Pn
xyz
i1 V wLundi C wLundi
C xyz
wLundp P cw e
C xyz
wLundp P cw e
V Lundw
Pn
i1 V wLundi
1 P cw
C xyz
wLundi
V Lundw
Pn
i1 V wLundi
C xyz
wLundi
V Lundw
Pn
1
xyz
i1 V wLundi C wLundi
V Rundw
Pn
1
xyz
i1 V wLundi C wLundi
V Rundw
1 Pcw e
1 Pcw e
Pn
i1 V wLundi
C xyz
wLundi
1
V Rundw
Pn
i1 V wLundi
C xyz
wLundi
1
V Rundw
The comparison shows that the effective elastic stiffness constants are much less sensitive for 3SHM with Pcw than for the
mixed VoigtReuss model with P c (Table 24). This means that
3SHM is more stable when using the Pba
cw values evaluated in terms
of the mean inclination angle.
Appendix B. Micromechanical models
Generalized self-consistenet
Christensen [47]:
method
(GSCM)
model
of
3027
E11 V f Ef11 V m Em
4 V f V m v f12 v m 2
Vf
Km
f
v 12 V f v f12 V m v m
G12 Gm
V V v
1
Gm
f
12
Vf
km
Vm
Kf
v
1
Gm
1
km
1
kf
m
v
kf
Gf 1 V f Gm V m
Gf V m Gm 1 V f
2
G23
G23
C 0
A
2B
Gm
Gm
with
!
!
Gf23
Gf23
A 3V 1 V
1
gf
Vm
Gm
"
! #
3
Gf23
Gf23
g gf gm m gm gf V f
Gm m
G
"
!
!#
Gf
Gf23
1
g
1
V f gm 23
Gm
Gm m
f 2
B 3V f V m
Gf23
1
Gm
Gf23
gf
Gm
!
vf
!"
! #
Gf23
Gf23
Gf23
f3
m 1
m gf
m gm gf V
G
G
G
!
! #)
Gf
Gf
f3
1 23
2 23
m gf
m gm gf V
G
G
gm 1
2
# "
( "
1 Gf23
Gf
g 23
1 V f 1 gf
2 Gm m
Gm
!
!
Gf23
Gf23
C 3V V
1
gf
Gm
Gm
! #
"
!
#"
Gf
Gf23
Gf23
Gf23
f
f3
23
g
1
V
g
g
g
f
f V
Gm m
Gm
Gm
Gm m
f
m2
with
gm 3 v m ; gf 3 v f23
E
v 23
K m G23
;
K m G23
with m 1 4K
v 212
E11
K
E22
K m K f Gm V m K f K m Gm V f
K f Gm V m K m Gm V f
2 1 v 23 G23
E11 V f Ef11 V m Em
Em
E22
p
1 V f 1 Em =Ef22
V 12 V f v f11 V m v m
Gm
p
1 V f 1 Gm =Gf12
Gm
p
1 V f 1 Gm =Gf23
G12 G13
G23
l 2 m3 l 3 m 2
l1 m3 l3 m1
l1 m2 l2 m1
K 1 1;
K 2 2;
K 3 3:
References
[1] Tong L, Moritz AP, Bannister MK. 3D ber reinforced polymer. Oxford: Elsevier;
2002.
[2] Kaw AK. Mechanic of composite materials. 2nd ed. LLC, New York: Taylor &
Francis Group; 2006.
[3] Hu J. 3-D brous assemblies properties applications and modelling of threedimensional textile structures. Woodhead Publishing Limited; 2008.
[4] Younes R, Zaki W. Optimal weaving for 2.5D interlocks. Compos Struct
2011;93:125564.
[5] Naik NK, Azad Sk NM, Durga Prasad P. Stress and failure analysis of 3D angle
interlock woven composites. J Compos Mater 2002;36:93.
[6] Tan P, Tong L, Steven GP. Micromechanics models for mechanical and
thermomechanical properties of 3D through-the-thickness angle interlock
woven composites. Composites: Part A 1999;30:63748.
[7] Pochiraju K, Chou TW. Three-dimensionally woven and braided composites. I:
A model for anisotropic stiffness prediction. Polym Compos 1999;20(4).
[8] Pochiraju K, Chou TW. Three-dimensionally woven and braided composites. II:
An experimental characterization. Polym Compos 1999;20(6).
[9] El Hage C. Modlisation du comportement lastique endommageable de
matriaux composites renfort tridimensionnel. PhD thesis. Universit de
Technologie de Compigne; 2006.
[10] Ishikawa T, Chou TW. Elastic behaviour of woven hybrid composites. JComput
Mater 1982:1629.
[11] Ishikawa T, Chou TW. Stiffness and strength behaviour of woven fabric
composites. J Mater Sci 1982;17:321120.
[12] Ishikawa T, Chou TW. One-dimensional micromechanical analysis of woven
fabric composites. AIAA J 1983;21:171421.
[13] Ishikawa T, Chou TW. In-plane thermal expansion and thermal bending
coefcients of fabric composites. J Comput Mater 1983;17:92104.
[14] Ishikawa T, Chou TW. Nonlinear behaviour of woven fabric composites. J
Comput Mater 1983;17:399413.
[15] Ishikawa T, Matsushima M, Hayashi Y. Experimental conrmation of the
theory of elastic moduli of fabric composites. J Comput Mater
1985;19:44358.
[16] Naik NK, Ganesh VK. Prediction of on-axes elastic properties of plain weave
fabric composites. Compos Sci Technol 1992;45:13552.
[17] Naik NK, Shembekar PS. Elastic behaviour of woven fabric composites: I Naik
lamina analysis. J Comput Mater 1992;26:2197225.
[18] Naik NK, Shembekar PS. Elastic behaviour of woven fabric composites:
111-laminate design. J Comput Mater 1992;26:252341.
[19] Naik NK, Sridevi E. An analytical method for thermoelastic analysis of 3D
orthogonal interlock woven composites. J Reinf Plastics Compos
2002;21:1149.
[20] Naik NK, Kuchibhotla R. Analytical study of strength and failure behavior of
plain weave fabric composites made of twisted yarns. Composites: Part A
2002;33:697708.
[21] Sankar BV, Marrey R. Analytical method for micromechanics of textile
composites. PII:SO266.3538(97)00030-4.
[22] Hahn H T, Pandey R. A micromechanics model for thermoelastic properties of
plain weave fabric composites. J Eng Mater Technol 1994:116517423.
3028