You are on page 1of 9

Composites Science and Technology 60 (2000) 1825±1833

www.elsevier.com/locate/compscitech

A novel micromechanics-based approach to the derivation


of constitutive equations for local/global analysis
of a plain-weave fabric composite
Yiwei Jiang 1, Ala Tabiei *,2, George J. Simitses
Aerospace Engineering and Engineering Mechanics Department, University of Cincinnati, OH 45221-0070, USA

Received 7 May 1999; received in revised form 26 January 2000; accepted 13 March 2000

Abstract
A stress and strain averaging procedure is presented in this paper for local/global analysis of plain-weave fabric composites.
Within a representative volume cell, using uniform stress and uniform strain assumptions, the constitutive equations are averaged
along the thickness direction. The cell is then divided into many subcells and an averaging is performed again by assuming uniform
stress distribution in each subcell to obtain the e€ective stress/strain relations of the subcell. As a result, a linear system of equations
is formed and subsequently solved to distribute the global average strains to each subcell. In this manner, given the average strains,
the global average stresses can be determined. Numerical results are generated for engineering constants and compared to other
numerical predictions and experimental data. Good agreements are achieved with respect to experimental data and ®nite-element
results. # 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Micromechanics; Woven composites; Micro/macro analysis

1. Introduction volume fraction, yarn undulation, weave style and


properties of the constituent materials.
Woven-fabric composites have long been recognized The ®nite-element method can be used to study the
as being more competitive than unidirectional compo- overall behavior of composite structures on the macro
sites for their good stability in the mutually orthogonal level and the material behavior on the constituent level.
warp and ®ll directions. This is attributed to more Whitcomb [1], Zhang [2] and Chapman [3] have studied
balanced properties in the fabric plane and enhanced the elastic material properties of woven composites by
impact resistance. These advantages have resulted in a the ®nite element method. A more practical issue may
growing interest in the use of woven-fabric composites arise when studying the global behavior of structures
for structural applications and development of analytical with consideration of the material non-linearity. Conse-
procedures for prediction of the thermo-mechanical quently, application of a micromechanics-based mate-
properties of these woven-fabric composites. Analytical rial model into FEA solvers provides a feasible way in
models for determination of mechanical properties of dealing with the global behavior of woven composite
woven composites provide a cost-e€ective tool for structures.
determining the e€ects of several parameters on the A number of analytical models were presented by
mechanical properties of woven-fabric composites. many investigators. All of these models studied the
These parameters include fabric weight, constituent average performance of a periodic representative volume
cell. Ishikawa and Chou suggested a variety of models
[4±6] to handle the in-plane behavior of woven compo-
* Corresponding author. Tel.: +1-513-556-3548; fax: +1-513-556- sites. These are the mosaic model, the ®ber undulation
5083.
model, and the bridging model. The basic assumption for
E-mail address: atabiei@uceng.uc.edu (A. Tabiei).
1
Current address: Sundstrand Aerospace, 4747 Harrison Ave., these models is that the classical lamination theory is
Rockford, IL 61125-7002, USA. valid for every in®nitesimal strip of a representative cell.
2
Director of Center of Excellence in DYNA3D Analysis. Ishikawa and Chou [7] also expanded their models to
0266-3538/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0266-3538(00)00064-6
1826 Y. Jiang et al. / Composites Science and Technology 60 (2000) 1825±1833

Nomenclature f local angle between the ®ll yarn and global


coordinate system
a half length of the representative cell w local angle between the warp yarn and global
b half width of the representative cell coordinate system
E Young's modulus …ÿ†… ; † average quantities of a subcell
G shear modulus …† average quantities of representative volume cell
H height of the representative cell … †t quantities of tangential sti€ness matrix or
Ht height of ®ber tows compliance matrix
lx length of a subcell … † quantities of in-plane or out-of-plane strains
ly width of a subcell or stresses in principal material coordinate
Lx length of the representative cell system
Ly width of the representative cell … †i in-plane stress or strain components
r radius of a circle thread … †o out-of-plane stress or strain components
Sij compliance components in principal material … †f quantities of the ®ll tow
system … †w quantities of the warp tow
Sij compliance components of constituents in … †m quantities of the matrix
ÿ
global coordinate system
Sij compliance components of the whole cell in Subscripts
ˆ
global coordinate system x, y, z quantities in global coordinate system
V volume of a representative cell 1,2,3 quantities in principal material coordinates

deal with constituent materials with shear non-linearity data. Our procedure is employed accordingly and com-
and initial failure (known as knee phenomenon). Their pared to those limited results. The real value of the
work basically considers one-dimensional strip of a methodology is in its application to trace the response
representative cell. As a result, these models cannot of real world situations.
represent the material behavior of woven composites
under bi-directional loading. Naik et al. [8±10] proposed
a series-parallel model and a parallel-series model for 2. Formulation
considering the two-dimensional undulation geometry
of plain woven composites. 2.1. Representative volume cells
The developed two-dimensional analytical models to
date provide a reasonable interpretation of material Any non-hybrid plain-weave fabric composite lamina
property prediction. In this research, a homogenization can be represented by double-periodic representative
method is presented that provides the constitutive volume cells, as shown in Fig. 1. The representative
equations of plain woven-fabric composites from the volume cell typically consists of two sets of interlaced
constitutive relations at the constituent level. It is prac- yarns, known as ®ll and warp threads, and resin. The ®ll
tically important to use the proposed method when and warp threads, in the form of mixture of ®bers and
global analysis of structures may directly require the resin, are assumed to be homogenous and transversely
constitutive equations of the constituents. Global ana- isotropic. The resin is homogenous and isotropic. Given
lysis such as material non-linear behavior, strength a representative
 volume cell, the average stresses f gand
modeling, temperature e€ect, and strain rate e€ect strains " can be written as:
would bene®t greatly from such formulation [11]. The
paper deals primarily with a micromechanics-based ……… ………
1  1
method of obtaining e€ective engineering constants for f g ˆ f gdv and " ˆ f"gdv …1†
V V
a woven composite. The reason for the new method is to
be able to evaluate these constants in a local/global
incremental analysis. The procedure is needed because where the stress and strain vectors are f g ˆ f x ;
T
at each load step the elastic constants (obtained y ; xy ; z ; yz ; xz gT and f"g ˆ "x ; "y ; xy ; "z ; yz ; xz ,
through the local constitutive equations) may change respectively. The question that must be answered is that,
as a result of damage and non-linearities in the matrix given the average stresses (or strains), what are the global
contribution. strains (or stresses). For this purpose, a micromechanics-
The only results available are those of Refs. [12,13], based model is developed on the basis of assumptions
which are based on linear ®nite-element and experimental related to local stresses and strains.
Y. Jiang et al. / Composites Science and Technology 60 (2000) 1825±1833 1827

2.2. E€ective stress/strain relations for an in®nitesimal f " x g ˆ f " x g m ˆ f " x gf ˆ f " x gw …3a†
element    
"y ˆ "y m ˆ "y f ˆ "y w …3b†
Consider a typical in®nitesimal block of the cell with
   
dx in length, dy in width and H in height, as shown in xy ˆ xy m ˆ xy f ˆ xy w …3c†
Fig. 2. This in®nitesimal element generally consists of
three di€erent materials with tm , tf and tw volume frac- and
tions of matrix, ®ll and warp, respectively. The material
coordinate system of ®lls and warps may not coincide fz g ˆ fz gm ˆ fz gf ˆ fz gw …3d†
with the global coordinate system. The e€ective average    
stress and strain vector can be expressed as: yz ˆ yz m ˆ yz f ˆ yz w …3e†
X X
f" g ˆ tk f"gk and f g ˆ tk f gk …2† fxz g ˆ fxz gm ˆ fxz gf ˆ fxz gw …3f†
kˆm;f;w kˆm;f;w
The above compatibility conditions are based on the
where the vectors with subscript k denote constituents, of rules of mixture, i.e. the in-plane relations for these
with m for matrix, f for ®ll and w for warp. It is three materials are in a parallel (Voigt) arrangement and
assumed that strain and stress compatibility conditions the out-of-plane relations in a series (Reuss) arrange-
exist such that: ment. Those assumptions are independent of stress/
strain relations and have been widely used in estimation
of material properties of uniaxial composite materials.
Combining Eqs. (2) and (3), the following equations are
derived for incremental strains and stresses:

df"i g ˆ df"i gk …4a†


X ÿ
df"o g ˆ tk df"o gk …4b†
kˆm;f;w

X ÿ
dfi g ˆ tk dfi gk …4c†
kˆm;f;w

dfo g ˆ dfo gk …4d†

where f"i g, f"o g, fi g, and fo g are the in-plane strains, out-
of-plane strains, in-plane stresses and out-of-plane stres-
ses, respectively, and given by the following equations:
8 9 8 9
< d"x = < d"z =
Fig. 1. Representative cell of a plain-woven composite. df"i g ˆ d"y ; df"o g ˆ d xz ; dfi g
: ; : ;
d xy d yz
8 9 8 9
< dx = < dz =
ˆ dy ; dfo g ˆ dxz …5†
: ; : ;
dxy dyz

In the material coordinate system, the constitutive


laws for matrix, ®ll and warp are all written as:

8 9 2 3 8 9
>
> d"11 >
> St11 St12 0 St13 St23 0 > d11
> >
>
>
> >
> 6 St12 7 >
> d22 >
>
>
> d"22 >
> 6 St22 0 0 0 0 7 >
> >
>
>
< d 12 >
= 6 0 7 >
< d12 >
=
6 0 St66 0 0 0 7
ˆ6 7
>
> d"33 >
> 6 St 0 0 St33 0 0 7 > d11 >
>
> >
> 6 13 7 >
>
>
>
>
>
>
> d >
> 4 St 0 0 0 St55 0 5 >
> d >
>
>
: 13 >
; 23 >
: 22 >
;
d 23 t
k 0 0 0 0 0 S44 k d12 k
Fig. 2. An in®nitesimal block of the representative cell. …6†
1828 Y. Jiang et al. / Composites Science and Technology 60 (2000) 1825±1833

where the compliance components with superscript t dfi g ˆ ‰Ci Šdf"i g ‡ ‰Cio Šdfo g …11a†
represent the tangential ones. For linear orthotropic
materials, the expressions for Stij are the same as Sij and df"o g ˆ ‰Coi Šdf"i g ‡ ‰Co Šdfo g …11b†
can be found in any text book on mechanics of compo-
site materials. For derivation purpose, Eqs. (6) are where
arranged by in-plane and out-of-plane components and
X ÿ h iÿ1
are denoted by ‰Ci Š ˆ tk Sti …12a†
ÿ k
( ) " t # ( ) k
d"i Si Stio di

ˆ …7† X ÿ h iÿ1 h i
d"o k Soi So k do k
t t
‰Cio Š ˆ ÿ tk Sti Stio …12b†
ÿ k ÿ k
k

where X ÿ h i h iÿ1
8 9 8 9 ‰Coi Š ˆ tk Stoi Sti …12c†
ÿ k ÿ k
 < d"11 =  < d"33 =  k

d "i ˆ d"22 ; d "o ˆ d 13 ; d i X ÿ h i h iÿ1 h i h i 


: ; : ;
d 12 d 23 ‰Co Š ˆ tk Stoi Sti Stio ÿ Sto …12d†
8 9 8 9 ÿ k ÿ k ÿ k ÿ k
d
< 11 =  < d33 = k

ˆ d22 ; d o ˆ d13


: ; : ; Finally, the e€ective stress/strain relations for the
d12 d23
in®nitesimal element are obtained by rearranging Eqs.
(11) as follows:
and ( ) 2 St St 3( )
2 t 3 2 t 3 d"i ˆi ˆ io di
 t S11 St12 0  t S13 0 0   ˆ4 t 5 …13†
Si ˆ 4 St12 St22 0 5; Sio ˆ 4 St23 0 0 5; Stoi d"o Soi Sto do
ˆ ˆ
0 0 St66 0 0 0
2 t 3 2 t 3
S13 St23 0 S33 0 0 where
 t
ˆ4 0 0 0 5; Si ˆ 4 0 St55 0 5 h i
0 0 0 0 0 St44 Sti ˆ ‰Ci Šÿ1 …14a†
ˆ

h i
In the global coordinate system, the constitutive rela- Stio ˆ ÿ‰Ci Šÿ1 ‰Cio Š …14b†
ˆ
tions can be written as:
h i
( ) 2 t 3 ( ) Stoi ˆ ‰Coi Š‰Ci Šÿ1 …14c†
d"i Si Stio di ˆ
ˆ4 t ÿ ÿ 5 …8† h i
d"o k Soi Sto d
ÿ ÿ o k k Sto ˆ ‰Co Š ÿ ‰Coi Š‰Ci Šÿ1 ‰Cio Š …14d†
ˆ

with
2 3 " #
Sti Stio Sti Stio 2.3. Coupling phenomena
4ÿ ÿ 5 ˆ ‰T1 Šk ‰T1 ŠTk …9†
Stoi Sto Stoi Sto
ÿ ÿ k k In Eq. (13), in-plane normal components are coupled
with transverse shear components. This stems from the
The expression of the transformation matrix ‰T1 Šk can fact that …stio †21 , …stio †22 , …Stio †23 and …Stio †33 are not always
be found in Appendix I of Ref. [1]. Eqs. (8) can be 
equal to zero. 
However,
ˆ ˆ
since either the cell is balanced
written in the partitioned matrix form as: or the structure is balanced, these coupling terms are
h iÿ1 h iÿ1 h i zero on the average. Therefore, coupling components of
dfi gk ˆ Sti df"i gk ÿ Sti Stio dfo gk …10a† the compliance matrix in Eq. (11) are always ignored in
ÿ k ÿ k ÿ k subsequent formulation.
h i h i
df"o gk ˆ Stoi dfi gk ÿ Sto dfo gk …10b† 2.4. Average stress/strain relations for a subcell
ÿ k ÿ k

Incorporating Eqs. (4) and (10), the following equa- The whole representative volume cell is divided into
tions can be derived many subcells, as shown in Fig. 4. It is assumed that
Y. Jiang et al. / Composites Science and Technology 60 (2000) 1825±1833 1829

stresses are uniformly distributed within each subcell.  


S12 ˆ c2 …S12 †f ‡s2 …S23 †f
The incremental average stress/strain relations for a ÿ f
subcell can be obtained by integrating Eqs. (13), i.e.:
 
8 9 2ÿ ÿ
38 9 S22 ˆ …S22 †f
< d"… ; † = S t
S t < d … ; † = ÿ f
i 6 i io 7 i
ˆ4 5 …15†  
: … ; † ; ÿ ÿ : … ; † ;
d"o t
Soi So t d o S66 ˆ c2 …S66 †f ‡s2 …S44 †f
ÿ f
with
2ÿ 3 2 3 with c ˆ cos…f † and s ˆ sin…f †, where f is the undula-
ÿ
St Stio …… Sti Stio tion angle, i.e. the angle between the ®ber and the global
6 i 7 1 6 ˆ ˆ 7
4ÿ 5ˆ 4 5dxdy …16† xÿy plane.
ÿ l l Stoi Sto
Stoi Sto ˆ ˆ
 O€-axis warp
    ÿ 
… ; † denote the number of a subcell. l and l are the …"x †w ˆ S11 …x †w ‡ S12 y w
ÿ w ÿ w
length and width of subcell … ; †, respectively. ÿ      ÿ 
"y w ˆ S12 …x †w ‡ S22 y w
ÿ w ÿ w
ÿ    ÿ 
3. In-plane stress/strain relations xy w ˆ S66 xy w …18†
ÿ w

In application of shell elements to analyze thin-walled


composite structures by ®nite element methods, the where
required constitutive relations can be partitioned into  
two parts. One part is the in-plane components and the S22 ˆ c4 …S11 †w ‡2c2 s2 …S13 †w ‡s4 …S33 †w ‡c2 s2 …S55 †w
ÿ w
other is the transverse shear components. In these rela-
tions the global normal stress in thickness direction is  
ignored. During the incremental/iterative solution S12 ˆ c2 …S12 †w ‡s2 …S23 †w
ÿ w
scheme, as used in ®nite element analysis of nonlinear
problems, a change of the nodal displacements takes  
place. The displacement increment causes an increment S11 ˆ …S22 †w
ÿ w
of strain  " (on an average basis for the micro mate-
rial model) at a material point. The material model is  
S66 ˆ c2 …S66 †w ‡s2 …S44 †w
required to calculate the tangential sti€ness matrix and ÿ w
the incremental average stress f g. In this investiga-
tion the averaging procedure employed yields the tan- with c ˆ cos…w † and s ˆ sin…w †, where w has the same
gential sti€ness matrix and incremental stress under an meaning as f .
increment of strains. Neglecting normal stress in the
thickness direction and the coupling among in-plane  Matrix
normal components and transverse shear components,
the in-plane stress/strain relations for o€-axis ®ll tows, …"11 †m ˆ …S11 †m …11 †m ‡ …S12 †m …22 †m …19†
warp tows and matrix can be written as:
…"22 †m ˆ …S12 †m …11 †m ‡ …S22 †m …22 †m
 O€-axis ®ll
   ÿ  … 12 †m ˆ …S66 †m …12 †m
…"x †f ˆ S11 …x †f ‡ S12 y f From Eqs. (17)±(19), the incremental form of in-plane
ÿ f ÿ
ÿ     f ÿ  stress/strain relations of constituents may be written as:
"y f ˆ S12 …x †f ‡ S22 y f 8 9 2ÿ t  ÿ t  3 8 9
ÿ ÿ
ÿ   ÿ 
f f
< ÿ…x†k = C k C12 k 0 < ÿ…"x†k =
ÿ 11  ÿ 
xy f ˆ S66 xy f …17† d ÿy k ˆ 4 Ct12 k Ct22 k 0 5d ÿ "y k
ÿ f : ; ÿ t  : ;
xy k 0 0 C66 k xy k
where …20†

  Consider a quarter cell of the representative cell of Fig.


S11 ˆ c4 …S11 †f ‡2c2 s2 …S13 †f ‡s4 …S33 †f ‡c2 s2 …S55 †f 1, as shown in Fig. 3. Because of in-plane symmetry, this
ÿ f
quarter cell represents the same mechanical properties
1830 Y. Jiang et al. / Composites Science and Technology 60 (2000) 1825±1833

d … ; † ˆ dx… ; †
ÿ x  …22a†
ˆ 1; . . . ; Nx ÿ 1;  ˆ ‡ 1; ˆ 1; . . . ; Ny

dy… ; † ˆ dy… ; †
ÿ  …22b†
ˆ 1; . . . ; Nx ; ˆ 1; . . . ; Ny ÿ 1;  ˆ ‡ 1

… ; † ÿ ÿ 
dxy ˆ dxy ˆ 1; . . . ; Nx ; ˆ 1; . . . ; Ny …22c†

X
Nx … ; †
l ÿ ÿ 
x
d"… ; †
x ˆ d" x ˆ 1; . . . ; Ny …22d†
ˆ1
Lx

Nyx … ; †
X ly ÿ
d"… ; †
y ˆ d" y … ˆ 1; . . . ; Nx † …22e†
ˆ1
Ly

Ny X
X Nx … ; † l… ; †
l xy … ; † ÿ ÿ 
d xy ˆ d xy ˆ 1; . . . ; Ny …22f†
ˆ1 ˆ1
Lx Ly
Fig. 3. A quarter cell of the representative cell.

as the whole cell. The quarter cell is further divided into The quantities with ``bar'' denote the incremental
many subcells, as shown in Fig. 4, for performing the average strain and stress components of the whole cell.
averaging procedure. Once the incremental forms of the Nx and Ny are the total number of subcells in x and y
stress-strain relations of the constituents are obtained directions, respectively. The incremental average stresses
from Eqs. (20), the in-plane relations for a subcell can induced by incremental average strains are expressed as:
be derived from Eqs. (15) and are denoted by:
Ny … ; † … ; †
8 … ; † 9 2 t… ; † 3 8 9  ÿ XNx X lx ly 
t… ; † … ; † d  ˆ d  … ; † …23†
< x = C11 C12 0 < "x = Lx Ly
6 7 ˆ1 ˆ1
d y… ; † ˆ 4 Ct… ; † C t… ; †
0 5d "… ; †
: … ; † ; 12 22 : y… ; † ;
xy 0 0 Ct… ; † "xy
66
where Lx and Ly denote the length and width of the
…21† quarter cell, respectively.
In this procedure, it is assumed that the average in-plane Eqs. (22) together with Eqs. (21) provide sucient
strains and stresses among subcells have the following information to distribute the incremental average
relationships: strains to each subcell. Once the average strains in each
subcell are known, the incremental average stresses of
the cell can be obtained by Eqs. (21) and (23). Simulta-
neously, one can obtain the tangential sti€ness matrix.
In the next section, the calculation procedure will be
outlined. Now, by replacing the stress terms in Eqs.
(22a)±(22f) by using Eqs. (21) and moving the terms in
the right hand side at the equality sign to the left, a
simultaneous linear system of equations is formed with
incremental strains of each subcell as unknown and
incremental average strains of the cell as known. The
system is denoted by the following:
nÿo
‰BŠfdf"i gg ˆ ‰KŠd "i …24†

where
 n oT n oT n oT n o 
…N ;N † T
fdf"i ggT ˆ d "i…1;1† d "i…2;1† d "i…3;1† . . . d "i x y

Fig. 4. Division of a quarter cell into subcells. …25†


Y. Jiang et al. / Composites Science and Technology 60 (2000) 1825±1833 1831

‰BŠ is a 3Nx  Ny by 3Nx  Ny matrix, and ‰KŠ is a 3Nx 4. The total stresses of each subcell are then obtained

Ny by 3 matrix. From Eq. (24), the incremental strains from the following equation:
of each subcell can be solved for from the following:  n‡1 ÿ  n‡1 ÿ  n‡1 ÿ  n‡1
nÿo  … ; † ˆ t m  … ; † m ‡ t f  … ; † f ‡ t w  … ; † w
fdf"i gg ˆ ‰BŠÿ1 ‰KŠd "i …26†
…32†
The partitioned form of Eq. (26) is where the subscripts m, f and w denote matrix, ®ll and
n o   nÿo warp respectively.
d "… ; †
i ˆ A… ; † d "i
ÿ  …27†
ˆ 1; . . . ; Nx and ˆ 1; . . . ; Ny 4. Veri®cation examples

By combining Eqs. (21), (23) and (27), the incremental Two veri®cation examples are presented. The ®rst is a
average stresses can be obtained: plain-weave composite with circular cross-section ®ber
tows, and the second has sinusoidal cross-section ®ber
ÿ X Nx X
lx ly  … ; †  … ; †  nÿo
Ny … ; † … ; †
tows. For the ®rst example, ®nite-element predictions
d  ˆ C A d " …28† are available for comparison. The second example con-
ˆ1 ˆ1
Lx Ly
sidered, an experimental data is available. A parametric
study is performed to determine the number of subcells
or necessary for convergence to the prediction of the ®nite-
element and the experimental results.
ÿ h ÿ i nÿo
d  ˆ C d " …29†
4.1. A plain-woven composite lamina with circular cross-
section ®ber tows
with
The prediction of woven composite lamina properties
hÿi X Nx XNy … ; † … ; †
lx ly  … ; †  … ; †  using the presented methodology is compared to the ®nite
C ˆ C A …30† element prediction of [12]. The shape of the undulation of
ˆ1 ˆ1
Lx Ly
the ®ber tows in [12] are taken as sinusoidal form of a cir-
cular cross-section. The shape of the ®ll is assumed to
Eq. (30) provides the tangential sti€ness matrix for have the following form:
the average in-plane stress/strain relations. One can  
calculate the e€ective engineering constants, Ex , Ey , vxy f H x
h1 ˆ … ÿ r† sin… † ‡ 1 …33†
and Gxy , from this equation if desired. For other engi- 2 Lx
neering constants such as Ex , Gyz , Gxz , vyz and vxz ,
three-dimensional constitutive relations are required in Similarly, the shape of the warp undulation is
the above procedure.    
w H Ly 
h1 ˆ … ÿ r† sin …y ÿ † ‡1 …34†
3.1. Solution algorithm 2 2 Ly

The steps necessary


 to calculate the constituent stresses where Lx equals Ly in the model. The o€-axis angles of
of each subcell  … ; † in Eq. (23) can be summarized by the ®ll and warp are
the following:  f
ÿ1 dh1
f ˆ tan
1. For
 the n ‡ 1st increment of global average strains dx
n‡1
d " , the corresponding incremental strains of
each subcell are determined by Eq. (27); and
 
2. The total strains of each subcell are updated by the dhw
w ˆ tanÿ1 1
;
following equation: dy

 n‡1  n  n respectively. The engineering constants of constituents


"… ; † ˆ "… ; † ‡d "… ; † …31†
are listed in Table 1. The volume fraction of ®ber tows is
expressed as:
3. The total stresses for each constituent correspond-
ing to the total average strains of each subcell are 2r2
Vf ˆ …35†
determined by solving Eqs. (17)±(19) HLx
1832 Y. Jiang et al. / Composites Science and Technology 60 (2000) 1825±1833

Table 1 Table 4
Properties of graphite/epoxy lamina and resin (GPa for moduli) Prediction of In-Plane E€ective Engineering Constants (GPa for
moduli)
E1 E2 E3 12 13 23 G12 G13 G23
Ex Ey xy Gxy
Unidirectional 113.0 8.82 8.82 0.3 0.3 0.495 4.46 4.46 2.95
lamina Present No. 1 50.98 50.98 0.0504 4.29
Resin 4.0 4.0 4.0 0.35 0.35 0.35 1.48 1.48 1.48 of 4 47.56 47.56 0.0530 3.95
Subcells 9 46.95 46.95 0.0534 3.88
16 46.74 46.74 0.0536 3.86
The engineering constants of a woven composite with 25 46.64 46.64 0.0538 3.85
two di€erent tow volume fractions are calculated and 100 46.38 46.38 0.0538 3.84
225 46.35 46.35 0.0538 3.83
compared to the FEA results of [12] in Table 2. The Experimental 48.3a/49.8b 48.3a/49.8b 0.062a/0.068b 5.41a/3.83b
prediction of the presented formulation with 9, 25, 36, Data [13]
and 100 subcells is presented in the table. Results show a
Number of plies=1.
that even with only nine subcells the methodology yield b
Number of plies=4.
good results as compared to the more expensive and
time consuming ®nite-element prediction.

4.2. A plain woven composite lamina with sinusoidal case of previous example the prediction of the presented
cross-section ®ber tows methodology yield an acceptable results.

In this example, sinusoidal cross-section of yarns is


considered. The following geometry is used for the cal- 5. Conclusion
culation: a ˆ b ˆ 2:0 (mm), H ˆ Ht ˆ 0:196 (mm).
Material properties for this example are listed in Table An averaging procedure is presented for homo-
3. The calculated in-plane engineering constants using genization of the stress/strain relations of plain-weave
the presented methodology is compared with the composites. A stress update procedure is developed. The
experimental data from [13] and presented in Table 4. presented methodology can be directly applied to ®nite-
The experimental data consisted of two sets of data in element software for structural analysis. Various calcu-
which a single ply and four plies are tested. As in the lations of elastic properties demonstrate the adequacy
of the method. The power and the usefulness of the
developed formulation lies in the fact that stresses and
strains can be found in each constituents of the subcell.
Table 2 This will help in de®ning the e€ects of many parameters
Prediction of in-plane e€ective engineering constants (GPa for moduli) such as temperature, strain softening, strain rate, etc.,
Ex Ey xy Gxy on the mechanical properties of the woven composite.

Vf ˆ 0:40 No. of 9 8.6150 8.6150 0.2231 2.2683


Subcells
25 8.5955 8.5955 0.2230 2.2559
Acknowledgements
36 8.5588 8.5588 0.2229 2.2490
100 8.5446 8.5446 0.2229 2.2434 The research was sponsored by the NASA Langley
FEA [12] 8.4446 8.4446 0.2124 2.1949 Research Center under grants NAG-1-1280 and Nag-1-
Vf ˆ 0:52 No. of 9 8.9371 8.9371 0.2378 2.5306 2071. Computing support was provided by the Ohio
Subcells 25 8.8821 8.8821 0.2385 2.5102 Supercomputer Center. Their support is gratefully
36 8.8432 8.8432 0.2383 2.5003 acknowledged.
100 8.8123 8.8123 0.2385 2.4915
FEA [12] 8.4181 8.4181 0.2334 2.396

References

[1] Whitcomb JD. Three-dimensional stress analysis of plain weave


composites. In: O'Brien TK, editor. Composite materials: fatigue
Table 3 and fracture, vol 3. (ASTM STP 1110). Philadelphia: American
Properties of CFRP lamina and epoxy resin (GPa for moduli) Society for Testing and Materials, 1991. p. 417±38.
[2] Zhang YC, Harding J. A numerical micromechanics analysis of
CFRP Resin the mechanical properties of a plain weave composite. Computers
and Structures 1990;36(5):839±44.
EL ET GLT LT TT Em m
[3] Chapman CD, Whitcomb JD. Strategy for modeling eight-har-
137.3 10.79 5.394 0.26 0.46 4.511 0.38 ness satin weave carbon/carbon composites subjected to thermal
loads. AIAA-9601520-CP, 1996. p. 1774±80.
Y. Jiang et al. / Composites Science and Technology 60 (2000) 1825±1833 1833

[4] Ishikawa T, Chou TW. One-dimensional micromechanical analysis [10] Naik NK, Shembekar PS. Elastic behavior of woven fabric com-
of woven fabric composites. AIAA Journal 1983;21(12):1714±21. posites: III Ð laminate design. Journal of Composite Materials
[5] Ishikawa T, Chou TW. Elastic behavior of woven hybrid com- 1992;26(17):2522±41.
posites. Journal of Composite Materials 1982;16:2±19. [11] Tabiei A, Jiang Y. Woven fabric composite material model with
[6] Ishikawa T, Chou TW. Thermoplastic analysis of hybrid com- material nonlinearity for the ®nite element simulation. In: Pro-
posites. Journal of Material Science 1983;18:2260±8. ceedings of The Crashworthiness, Occupant Protection Sympo-
[7] Ishikawa T, Chou TW. Nonlinear behavior of woven fabric sium (AMD-Vol 225). ASME, Winter Annual Meeting, Dallas,
composites. Journal of Composite Materials 1983;17:399±413. TX, 1997. p. 175±90.
[8] Naik NK, Shembekar. Elastic behavior of woven fabric compo- [12] Al-Huniti N., Micromechaincal modeling of woven and textile
sites: I Ð lamina analysis. Journal of Composite Materials composites. PhD thesis, Dept. of Aerospace Eng. & Eng.
1992;26(15):2196±225. Mechanics, University of Cincinnati, 1996.
[9] Shembekar PS, Naik NK. Elastic behavior of woven fabric com- [13] Ishikwa T, Matsushima M, Hayashi Y. Experimental con®rma-
posites: II Ð laminate analysis. Journal of Composite Materials tion of the theory of elastic moduli of fabric composites. Journal
1992;26(15):2126±246. of Composite Materials 1985;19:443±58.

You might also like