Reference 6

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Available online at www.sciencedirect.

com

International Journal of Solids and Structures 45 (2008) 4114–4129


www.elsevier.com/locate/ijsolstr

Three-dimensional continuum analysis for a


unidirectional composite with a broken fiber
Michael Ryvkin *, Jacob Aboudi
Department of Solid Mechanics, Materials & Systems, Faculty of Engineering, Tel Aviv University, Ramat Aviv 69978, Israel

Received 5 November 2007; received in revised form 6 February 2008


Available online 6 March 2008

Abstract

The three-dimensional problem of a periodic unidirectional composite with a penny-shaped crack traversing one of the
fibers is analyzed by the continuum equations of elasticity. The solution of the crack problem is represented by a super-
position of weighted unit normal displacement jump solutions, everyone of which forms a Green’s function. The Green’s
functions for the unbounded periodic composite are obtained by the combined use of the representative cell method and
the higher-order theory. The representative cell method, based on the triple discrete Fourier transform, allows the reduc-
tion of the problem of an infinite domain to a problem of a finite one in the transform space. This problem is solved by the
higher-order theory according to which the transformed displacement vector is expressed by a second order expansion in
terms of local coordinates, in conjunction with the equilibrium equations and the relevant boundary conditions. The actual
elastic field is obtained by a numerical evaluation of the inverse transform. The accuracy of the suggested approach is ver-
ified by a comparison with the exact analytical solution for a penny-shaped crack embedded in a homogeneous medium.
Results for a unidirectional composite with a broken fiber are given for various fiber volume fractions and fiber-to-matrix
stiffness ratios. It is shown that for certain parameter combinations the use of the average stress in the fiber, as it is
employed in the framework of the shear lag approach, for the prediction of composite’s strength, leads to an over estima-
tion. To this end, the concept of ‘‘point stress concentration factor” is introduced to characterize the strength of the com-
posite with a broken fiber. Several generalizations of the proposed approach are offered.
Ó 2008 Elsevier Ltd. All rights reserved.

Keywords: Periodic unidirectional composites; Representative cell method; Higher-order theory; Broken fiber; Penny-shaped crack

1. Introduction

The three-dimensional analysis of a periodic unidirectional fiber-reinforced material with several broken
fibers has received a considerable attention by various investigators. This three-dimensional problem is quite
complicated due to the loss of periodicity caused by the fiber breaks. In the classical publication by Hedgepeth
and Van Dyke (1967), the authors presented an analysis based on the shear lag approximation to this problem.

*
Corresponding author. Tel.: +972 3 6408130; fax: +972 3 6407617.
E-mail address: arikr@eng.tau.ac.il (M. Ryvkin).

0020-7683/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijsolstr.2008.02.022
M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129 4115

The shear lag approach for the modeling of two and three-dimensional cracked composites, in conjunction
with various improvements, was subsequently employed by Goree and Gross (1979), Suemasu (1984), Rosset-
tos and Shishesaz (1987), Ibnabdeljalil and Curtin (1997), He et al. (1998), Curtin and Takeda (1998), Landis
and McMeeking (1999), Beyerlein and Landis (1999), Landis et al. (1999), Landis et al. (2000), Mahesh et al.
(2002), and references cited there. In the shear lag analysis, the model includes an infinite number of fibers but
several rather restrictive assumptions regarding the behavior of the fiber and matrix materials are made.
When the finite element approach is employed for the analysis of a composite with broken fibers, both
phases are modeled by a continuum constitutive law but the total size of the model is finite. This necessitates
a three-dimensional analysis of a quite extensive region. In practice, due to the computational cost, one needs
to limit the finite element model to several intact fibers neighboring the broken one. Such analyses were carried
out by Van den Heuvel et al. (2004), Nedele and Wisnom (1994), and Reedy (1984), for example. A different
approach was employed by Case and Reifsnider (1996). Here, the periodic composite with a broken fiber was
replaced by four concentric homogeneous elastic cylinders representing the fiber and matrix phases.
In the present investigation, we offer a novel continuum approach for the three-dimensional analysis of a
fiber-reinforced periodic composite with a broken fiber. Contrary to the aforementioned approaches, in the
present model the fiber-reinforced periodic composite includes an infinite number of fibers embedded in the
matrix, and both constituents are considered in the framework of the continuum elasticity theory.
The broken fiber forms a penny-shaped crack embedded in the periodic composite subjected to uniaxial
tension. The solution of the crack problem is represented by a weighted superposition of unit normal displace-
ment jump solutions everyone of which forms a Green’s function. The unknown coefficients are adjusted such
that traction-free conditions at the crack’s faces are satisfied. The problem in which a unit displacement jump
is applied in the infinite periodic three-dimensional composite domain is solved by the combined use of the
representative cell method and the higher-order theory. In the representative cell method (Nuller and Ryvkin,
1980; Ryvkin and Nuller, 1997), the infinite periodic domain is reduced to a finite one (the representative cell)
by the application of the triple discrete Fourier transform. The resulting boundary value problem for this rep-
resentative cell is characterized by specific type (Born-von Karman) of boundary conditions that relate the
opposite sides of the cell. This representative cell 3D problem (in the transformed space) is solved by the
higher-order theory according to which the displacements, which are governed by the continuum equations,
are expanded into second order in terms of local coordinates (Aboudi et al., 1999). The actual elastic field is
determined by a numerical evaluation of the inverse transform integration. The idea of combining these two
methods was recently suggested by Ryvkin and Aboudi (2007a) and successfully employed for the solution of
the fiber loss problem and the two-dimensional problem of a transverse crack in layered composites (Ryvkin
and Aboudi, 2007b,c).
The present paper is organized as follows. In Section 2, the solution of a penny-shaped crack in a unidirec-
tional periodic composite is derived. To this end, Green’s functions that correspond to unit jumps applied in
the crack’s plane are generated. These functions are appropriately superposed to obtain the sought solution.
The accuracy of the offered methodology is verified by a comparison with the exact analytical solution of a
penny-shaped crack embedded in a homogeneous material. In Section 3, the present approach is applied to
investigate the behavior of a unidirectional fiber-reinforced composite with a broken fibers. To this end, an
extensive parametric study (various ratios of the fiber-to matrix-stiffness and fiber volume fractions) is carried
out to investigate the stress field in the crack’s plane as well as along the fiber direction in the broken and intact
fibers. A comparison between the behavior of composites with a square and hexagonal fiber configurations is
given. In the Section 4, several generalizations to the present approach are offered.

2. A penny-shaped crack in a unidirectional periodic composite

Consider a periodic unidirectional fiber-reinforced material that is subjected to a remote tensile loading r
11 ,
in which the circular fibers are oriented in the x1 -direction. Assume that there is a broken fiber such that a
penny-shaped crack perpendicular to the x1 -direction exists, see Fig. 1(a). This problem is solved by combining
the analyses of the representative cell method and higher-order theory. Similar to the two-dimensional case
(Ryvkin and Aboudi, 2007c), the solution is achieved by constructing a set of Green’s functions as described
in the following.
4116 M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129

a b

c d

Fig. 1. (a) A periodically fiber-reinforced composite with a broken fiber subjected to a tensile loading r 11 at infinity (referred to the global
coordinates ðx1 ; x2 ; x3 Þ). The composite region is divided into identical cells denoted by ðK 1 ; K 2 ; K 3 Þ where K 1 ; K 2 ; K 3 ¼ 0; 1; 2; . . .. The
cell’s dimensions are: Dp , p ¼ 1; 2; 3. (b) A cross-section in the horizontal plane x1 ¼ 0 of the repeating cells ðK 1 ¼ 0; K 2 ; K 3 Þ. The penny-
shaped crack of radius a is located within the cell ðK 1 ¼ 0; K 2 ¼ 0; K 3 ¼ 0Þ. (c) A characteristic cell ðK 1 ; K 2 ; K 3 Þ in which local coordinates
ðx01 ; x02 ; x03 Þ are introduced. (d) The division of the characteristic (magnified) cell into subcells.

2.1. Modeling of the penny-shaped crack using unit jump Green’s functions

The circular crack region in the x2  x3 plane at x1 ¼ 0 is discretized into N rectangular domains Ri the cen-
ters of which are located at points ðxi2 ; xi3 Þ, i ¼ 1; . . . ; N . On everyone of these rectangles a single unit displace-
ment jump is applied. As a result, N elastic field distributions are generated. Let Gðxi2 ; xi3 ; xj2 ; xj3 Þ denote the
normal stress r11 at x2 ¼ xi2 ; x3 ¼ xi3 due to the application of the unit displacement jump at the rectangle j.
In order to obtain a traction-free crack surface these Green functions are superposed in the following form
X
N
f ð1Þ
ci Gðxi2 ; xi3 ; xj2 ; xj3 Þ ¼ r11 ; j ¼ 1; . . . N ð1Þ
i¼1
f ð1Þ
where r11 is the remote normal stress in the fiber which is given by
f ð1Þ Ef
r11 ¼ r
11 ð2Þ
E
where Ef is the Young’s modulus of the fiber and E denotes the longitudinal effective Young’s modulus of the
composite: E ¼ vf Ef þ ð1  vf ÞEm , with Em and vf denoting the Young’s modulus of the matrix and the fiber
volume fraction, respectively.
Eqs. (1) form a linear system of N algebraic equations for the determination of the coefficients ci ,
i ¼ 1; . . . ; N . It should be noted that for the present Mode I problem, symmetry guarantees the vanishing
M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129 4117

of the shear traction at the crack’s domain. Once these coefficients have been determined, the elastic field
U ðx1 ; x2 ; x3 Þ at any point of the composite can be obtained as follows
X
N
U ðx1 ; x2 ; x3 Þ ¼ ci U i ðx1 ; x2 ; x3 Þ þ U 0 ðx1 ; x2 ; x3 Þ ð3Þ
i¼1

where U i ðx1 ; x2 ; x3 Þ is the elastic field that generated at point ðx1 ; x2 ; x3 Þ by the application of the displacement
jump at segment i, and U 0 ðx1 ; x2 ; x3 Þ is the periodic elastic field in the undamaged composite caused by the
remote loading. It should by mentioned that in the framework of the shear lag theory, a similar approach
to the crack modeling by the construction of Green’s functions was employed, see Sastry and Phoenix
(1993), Landis and McMeeking (1999), Mahesh et al. (1999) and Landis et al. (2000), for example.

2.2. The determination of the Green’s functions

The determination of these Green’s functions is obtained by the combined use of the representative cell
method and higher-order theory. In accordance with the representative cell method (Ryvkin and Nuller,
1997) the composite is viewed as an assemblage of bonded identical cells labeled by the three indices
ðK 1 ; K 2 ; K 3 Þ where K 1 ; K 2 ; K 3 ¼ 0; 1; 2; . . ., see Fig. 1(b) which shows the cross-section in the x2 ; x3 -plane.
The composite’s region has been described with respect to global coordinates ðx1 ; x2 ; x3 Þ. In addition, in each
cell local coordinates ðx01 ; x02 ; x03 Þ are introduced whose origin is located in its center, see Fig. 1(c), such that the
cell extends over Dp =2 6 x0p 6 Dp =2, p ¼ 1; 2; 3. It should be noted that the sizes of the cell D2 ¼ D3 are deter-
mined by the periodicity of the unidirectional composite in the x2 and x3 -directions, whereas the size D1 can,
due to the translational symmetry in the x1 -direction, be arbitrarily chosen. The formulation of aforemen-
tioned Green’s function problem of applied unit displacement jump is presented as follows.
In the absence of body forces, the equilibrium equations in any cell ðK 1 ; K 2 ; K 3 Þ are given by
ðK 1 ;K 2 ;K 3 Þ
½rjk;k  ¼0 j; k ¼ 1; 2; 3; K 1 ; K 2 ; K 3 ¼ 0; 1; 2; . . . ð4Þ
ðK ;K ;K Þ
where rjk 1 2 3 are the stress components. For an elastic material they are given by
½rjk ðK 1 ;K 2 ;K 3 Þ ¼ C jklm ½lm ðK 1 ;K 2 ;K 3 Þ ð5Þ
ðK ;K ;K Þ
where C jklm the elements of the stiffness tensor of the material and jk 1 2 3 are the strain tensor components.
The strains are related to the displacement gradients in the standard form
 ðK ;K ;K Þ
ðK 1 ;K 2 ;K 3 Þ 1 ouj ouk 1 2 3
½jk  ¼ þ ð6Þ
2 oxk oxj
Furthermore, continuity of displacements and tractions between the fibers and matrix must imposed. In addi-
tion, continuity of displacements and tractions between adjacent cells should be imposed. To this end let us
define the vector
VpðK 1 ;K 2 ;K 3 Þ ðx01 ; x02 ; x03 Þ ¼ fu1 ; u2 ; u3 ; rp1 ; rp2 ; rp3 gðK 1 ;K 2 ;K 3 Þ ; p ¼ 1; 2; 3 ð7Þ

which defines the displacements and tractions on a plane perpendicular to the xp -axis. Thus, these conditions
take the form
ðK 1 ;K 2 ;K 3 Þ ðK 1 þ1;K 2 ;K 3 Þ
V1 ðD1 =2; x02 ; x03 Þ ¼ V1 ðD1 =2; x02 ; x03 Þ
ðK 1 ;K 2 ;K 3 Þ ðK 1 ;K 2 þ1;K 3 Þ
V2 ðx01 ; D2 =2; x03 Þ ¼ V2 ðx01 ; D2 =2; x03 Þ ð8Þ
ðK ;K ;K Þ ðK ;K ;K þ1Þ
V3 1 2 3 ðx01 ; x02 ; D3 =2Þ ¼ V3 1 2 3 ðx01 ; x02 ; D3 =2Þ
The applied unit displacement jump in the x01 -direction in the rectangle Ri , located at the plane x01 ¼ 0 at cell
K 1 ¼ K 2 ¼ K 3 ¼ 0, can be represented by the relations
 ðK ;K ;K Þ
u1 ðþ0; x02 ; x03 Þ  u1 ð0; x02 ; x03 Þ 1 2 3 ¼ dK 1 ;0 dK 2 ;0 dK 3 ;0 ; x02 ; x03 2 Ri ð9Þ
4118 M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129

where dj;k is the Kronecker delta.


The formulated problem, Eqs. (4)–(6) and (8),(9), for the infinite composite is reduced to a single represen-
tative cell problem by the application of the triple discrete Fourier transform. This provides the transformed j-
th component of the displacement, for example, in the form:
X
1 X
1 X
1
ðK 1 ;K 2 ;K 3 Þ
uj ðx01 ; x02 ; x03 ; /1 ; /2 ; /3 Þ ¼
^ uj ðx01 ; x02 ; x03 Þ exp½iðK 1 /1 þ K 2 /2 þ K 3 /3 Þ ð10Þ
K 1 ¼1 K 2 ¼1 K 3 ¼1

where p 6 /p 6 p, p ¼ 1; 2; 3. As a result, we obtain a representative cell problem for the transforms of the
field variables in the finite region Dp =2 6 x0p 6 Dp =2, p ¼ 1; 2; 3. The governing equations in this region that
correspond to Eqs. (4)–(9) are
^jk;k ¼ 0 j; k ¼ 1; 2; 3
r ð11Þ
^jk ¼ C jklm^lm
r ð12Þ
 
1 o^ uj o^ uk
^jk ¼ þ ð13Þ
2 oxk oxj
b b 1 ðD1 =2; x0 ; x0 Þ
V 1 ðD1 =2; x0 ; x0 Þ ¼ expði/1 Þ V
2 3 2 3
b 2 ðx0 ; D2 =2; x0 Þ ¼ expði/2 Þ V
V b 2 ðx0 ; D2 =2; x0 Þ
1 3 1 3
b 0 0 b 0 0
V 3 ðx ; x ; D3 =2Þ ¼ expði/3 Þ V 3 ðx ; x ; D3 =2Þ ð14Þ
1 2 1 2
u1 ðþ0; x02 ; x03 Þ  ^
^ u1 ð0; x02 ; x03 Þ ¼ 1 ð15Þ
where all field variables in the transform domain (hat-quantities) form complex quantities. Eq. (14) are re-
ferred to as the Born-von Karman type boundary conditions.
The solution of this problem in the transform domain is carried out by employing the higher-order theory
that was described by Aboudi et al. (1999) and was similarly employed by Ryvkin and Aboudi (2007c) for the
solution of the two-dimensional crack problem. In the framework of this theory, the representative cell
domain Dp =2 6 x0p 6 Dp =2, p ¼ 1; 2; 3, is divided into parallelepiped subcells as shown in Fig. 1(d), in every-
one of which the transformed displacements are represented by a second-order polynomial. The transformed
equilibrium equation, displacements and tractions continuity conditions between the subcells and the Born-
von Karman boundary conditions are imposed in the average (integral) sense.
Once this solution has been achieved, the actual elastic field can be readily determined at every point of any
desired cell ðK 1 ; K 2 ; K 3 Þ of the infinite composite region by the inverse transform formula:
Z p Z p Z p
ðK ;K ;K Þ 1
uj 1 2 3 ðx01 ; x02 ; x03 Þ ¼ 3 uj ðx01 ; x02 ; x03 ; /1 ; /2 ; /3 Þ exp½iðK 1 /1 þ K 2 /2 þ K 3 /3 Þd/1 d/2 d/3
^
8p p p p
ð16Þ
The triple integration in this formula can be carried out by employing, for example, the Gauss’s numerical
integration. This requires the evaluation of the integrand at sufficient number of points within the integration
domain p 6 /p 6 p, p ¼ 1; 2; 3. Furthermore, in the vicinity of the point ð/1 ; /2 ; /3 Þ ¼ ð0; 0; 0Þ the integrand
possesses high gradients thus requiring dense integration points. Contrary to the two-dimensional case where
this numerical integration procedure was successfully performed (Ryvkin and Aboudi, 2007c), in the present
three-dimensional problem the evaluation of the triple integrals in Eq. (16) requires a quite large amount of
computer time.
Therefore, we propose herein an alternative efficient procedure for the analysis of the present three-dimen-
sional problem which is based on a shifted finite discrete Fourier transform. To this end, the infinite composite
domain is replaced by a sufficiently large finite domain on the boundaries of which the elastic field perturba-
tion caused by the application of a displacement jump is negligibly small. This follows from the St. Venant
principle since the force resultant caused by the applied displacement jump is equal to zero. The finite domain
consists this time of M 1  M 2  M 3 cells such that K p ¼ 0; 1; 2; . . . ; M p with p ¼ 1; 2; 3. Accordingly, Eq.
(10) is replaced by
M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129 4119

ðm Þ ðm Þ ðm Þ
X
M1 X
M2 X
M3
ðK 1 ;K 2 ;K 3 Þ ðm1 Þ ðm2 Þ ðm Þ
uj ðx01 ; x02 ; x03 ; /1 1 ; /2 2 ; /3 3 Þ ¼
^ uj ðx01 ; x02 ; x03 Þ exp½iðK 1 /1 þ K 2 /2 þ K 3 /3 3 Þ
K 1 ¼M 1 K 2 ¼M 2 K 3 ¼M 3

ð17Þ

where
2pmp
/pðmp Þ ¼ p  ; mp ¼ 0; 1; 2; . . . ; M p ; p ¼ 1; 2; 3 ð18Þ
2M p þ 1

It should be emphasized that the equations that define the representative cell problem in conjunction with
the finite discrete Fourier transform are still given by Eqs. (11)–(15).
The inverse transform that corresponds to Eq. (17) is given by the triple sum:
ðK 1 ;K 2 ;K 3 Þ 1
uj ðx01 ; x02 ; x03 Þ ¼
ð2M 1 þ 1Þð2M 2 þ 1Þð2M 3 þ 1Þ
X
M1 X
M2 X
M3
ðm Þ ðm Þ ðm Þ ðm1 Þ ðm2 Þ ðm Þ
uj ðx01 ; x02 ; x03 ; /1 1 ; /2 2 ; /3 3 Þ exp½iðK 1 /1
^ þ K 2 /2 þ K 3 /3 3 Þ ð19Þ
m1 ¼M 1 m2 ¼M 2 m3 ¼M 3

It is worth mentioning that the employed form of the shifted finite Fourier transform eliminates the point
(0, 0, 0). The employment of the finite discrete Fourier transform in the present approach implies that the
problem of a crack in an infinite composite domain has been replaced by a crack in a finite domain that
extends over the region d p 6 xp 6 d p where d p ¼ ð2M p þ 1ÞDp , p ¼ 1; 2; 3. The size of this region has to
be chosen sufficiently large to ensure the decay of the perturbed field far away from the crack.
In order to check the above derivation, let us consider the case of a penny-shape crack in a homogeneous
isotropic material (i.e., the material properties of fiber and matrix coincide). It turns out that in this case it is
sufficient to take d 1 ¼ 12:5a, d 2 ¼ d 3 ¼ 18a (which corresponds to M 1 ¼ 4; M 2 ¼ M 3 ¼ 3), where a is the
crack’s radius, in order to get accurate results, such that a further increase of the domain size does not affect
the obtained results. Fig. 2 shows a comparison between the present solution and the analytical one for a
penny-shaped crack in a homogeneous isotropic material (Sneddon, 1951). In this figure, the normal stress
r11 in front of the crack is depicted. The numerical results in this figure were obtained by discretizing the
representative cell whose dimensions are D1 ¼ 1:4a, D2 ¼ D3 ¼ 2:5a into 2  12  12 subcells in the x01 , x02
and x03 -directions, respectively, see Fig. 1(d). It is readily observed that a very good agreement between the
two solutions exists. It should be noted that the analytical solution predicts a rather rapid decay which in
the plane of the crack yields the asymptotic formula

Fig. 2. The normal stress r11 variation ahead of the crack tip in the plane x1 ¼ 0 as predicted by the analytical solution and the present
approach. The radius of the crack is a=D2 ¼ 0:4.
4120 M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129
     qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 a3 a 4
r11 ðrÞ ¼ r
11 1 þ þO ; r¼ x22 þ x33 ð20Þ
3p r r

In summary, since in the framework of the higher-order theory (Aboudi et al., 1999), every subcell involves
21 unknown complex coefficients in the quadratic displacement expansions, one needs to solve a system of
ðm Þ ðm Þ ðm Þ
2  12  12  42 ¼ 6912 sparse algebraic equations for every triplet ð/1 1 ; /2 2 ; /3 3 Þ. The inversion formula
(19) requires the repeating of this procedure M ¼ ð2M 1 þ 1Þð2M 2 þ 1Þð2M 3 þ 1Þ times. The above chosen
parameters: M 1 ¼ 4; M 2 ¼ M 3 ¼ 3, result in M ¼ 441. It should be noted that the use of a direct approach with-
out employing the finite discrete Fourier transform (in conjunction with representative cell method) to obtain
this numerical result with identical resolution would require the solution of one system that corresponds to 441
cells thus forming more than one and a half million algebraic equations (an enormous numerical effort).
Since the effective response of fiber-reinforced materials exhibits an anisotropic isotropic behavior, it should
be interesting to verify the accuracy of the present method of solution by considering the problem of a penny-

a b

Fig. 3. A comparison between the analytical and present solution for the axial stress along the x1 -axis which passes through the center of
the penny-shaped crack. (a) Homogeneous isotopic material. (b) Homogeneous transversely isotropic material in which the axis of
symmetry is oriented in the x1 -direction.

Fig. p A comparison between the analytical and present solution for the shear stress along the x1 -axis at distance r=D2 ¼ 0:6
4. ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðr ¼ x22 þ x22 Þ from the crack’s center.
M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129 4121

shaped crack embedded in a homogeneous anisotropic material. To this end, consider a homogeneous
transversely isotropic material that effectively represents a homogenized unidirectional composite. The
crack plane x2  x3 coincides with the plane of elastic symmetry. The analytical solution for this problem
is given by Sih et al. (1981) and is summarized in the Appendix. The five elements of the stiffness matrix of
the homogenized material are: c11 ¼ 38:5 GPa, c22 ¼ c33 ¼ 12:5 GPa, c23 ¼ 4:19 GPa, c12 ¼ c13 ¼ 4:35 GPa,
c55 ¼ c66 ¼ 3:55 GPa, which correspond to the effective properties of a unidirectional glass/epoxy composite
in which the fibers are oriented in the x1 -direction, with a fiber volume fraction of 50%. Fig. 3 shows the
comparison between the analytical and the present solution for the normal stress r11 along the axis of sym-
metry x1 passing through the crack center. The results for isotropic and transversely isotropic materials are
presented. It is clearly seen that good agreement between the two approaches exists. As expected, the
response of the cracked solid in the isotropic case tends more rapidly to the applied far field. This effect
will be addressed in the following section where actual (rather than homogenized) fiber reinforced compos-
ite with a broken fiber is considered. Similarly, a good accuracy can be observed in the comparison
between the present and analytical solution for the shear stress r12 . This comparison is shown in Fig. 4
where the shear stress is plotted along the x1 -axis that passes at a radial distance r=D2 ¼ 0:6 from the crack
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
center ðr ¼ x22 þ x23 Þ.

3. Applications

Figs. 1(a) and (b) show a unidirectional composite with a broken fiber (which forms a penny-shaped crack)
subjected to a far-field tensile loading r 11 . Both fiber and matrix constituents are taken as isotropic materials
with Youngs’s moduli Ef , Em and Poisson’s ratios mf , mm , respectively. Figs. 1(c) and (d) exhibit the represen-
tative cell and its division into subcells. The dimensions of the representative cell were taken as in the case of a
penny-shaped crack in a homogeneous material namely, D1 ¼ 1:4a; D2 ¼ D3 ¼ 2:5a. As shown in Fig. 1(d),
this representative cell was discretized into 2  12  12 subcells. This choice corresponds to a fiber volume
fraction vf ¼ pa2 =D22 ¼ 0:5. In the case of a crack in a homogeneous medium it was found that a total size
of the finite domain of d 1 ¼ 12:5a; d 2 ¼ d 3 ¼ 18a was sufficient to provide accurate results. In the present case
however, due to the existence of the stiff fibers, the size of the region over which the perturbed field is appre-
ciable would change. It will increase in the fiber direction with the increase of the fiber-to-matrix stiffness ratio,
whereas it will decrease in the transverse direction (cf. Ryvkin and Aboudi (2007c) for the 2D case). To this
end, a total size of the finite domain in the x1 -direction in this case has been increased to d 1 ¼ 68a, while
d 2 ¼ d 3 ¼ 18a remained unchanged. This choice was verified in providing sufficiently good accuracy for a
fiber-to-matrix Young’s moduli Ef =Em 6 20.
Fig. 5 exhibits the normal stress distribution in the crack’s plane x1 ¼ 0 along the x2 -axis for a fiber volume
fraction vf ¼ 0:5 (which corresponds to a crack’s radius of a=D2 ¼ 0:4) with Ef =Em ¼ 20 and 4. In addition,
this figure shows the variation of this stress in the case of vf ¼ 0:2 and Ef =Em ¼ 20. In all cases presented
herein, the Poisson’s ratios of the fiber and matrix materials were taken as mf ¼ mm ¼ 0:2. The figure clearly
shows the locations of the fibers in which the stresses are quite higher that those in the matrix. In the vicinity
of the crack’s tip however the stress singularity in the matrix region is well exhibited. The figure shows that the
effect of fiber breakage in the horizontal direction is localized in the sense that a significant stress change can
be observed just in the neighboring intact fibers. Similar results was reported by Mahesh et al. (1999) who
investigated the stress concentration at the boundary of a penny-shaped cluster of broken fibers. The stress
distribution within this first intact fiber is significantly non-uniform. It falls from a maximum value which
is obtained at the closest point to the crack. A comparison between the plots of Ef =Em ¼ 20 and
Ef =Em ¼ 4 (Fig. 5a and b) both of which with vf ¼ 0:5 shows that this variation depends on the ratio of
the fiber-to-matrix stiffnesses. Namely, the contrast between the maximum and average stresses in the intact
fiber is higher in the case of higher fiber-to-matrix Young’s moduli ratio. Similarly, the increase of fiber vol-
ume fraction (Fig. 5a and c) results in an increase of this contrast. A detailed study of these effects is given in
the following.
It is interesting to examine the amount of localization of the elastic stress due to the fiber breakage. To this
f ð1Þ
end, let us investigate the redistribution of the force pa2 r11 which was carried by the fiber before its break.
This redistribution can be measured by the following ratio:
4122 M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129

a b

Fig. 5. The normal stress r11 variation along x2 with x3 ¼ 0 in the plane x1 ¼ 0 of the cracked fiber-reinforced material subjected to a
11 , with a fiber volume ratio vf ¼ 0:5. (a) Ef =Em ¼ 20, (b) Ef =Em ¼ 4, both with vf ¼ 0:5, and (c) Ef =Em ¼ 20 with vf ¼ 0:2.
remote stress r

Z Z
1 ð0Þ
d¼ f ð1Þ
½r11 ð0; x2 ; x3 Þ  r11 ð0; x2 ; x3 ÞdA ð21Þ
pa2 r11 A
2
where A is the area of a square containing ð2N þ 1Þ cells in the crack plane: N 6 K 2 6 N , N 6 K 3 6 N ,
ð0Þ
excluding the area of the broken fiber. In this equation, r11 ð0; x2 ; x3 Þ is the undisturbed stress field before fiber
f ð1Þ mð1Þ
breakage, given by either r11 or r11 which are the stresses in the fiber and matrix constituents, respectively.
These stresses are obviously identical to the remote stresses in the fiber and matrix phases. It should be noted
that the expression in the square bracket in Eq. (21) is the difference between the normal stress after and before
fiber breakage. For a composite with Ef =Em ¼ 20 and a fiber volume fraction vf ¼ 0:5, the obtained results for
the ratio d are given in Table 1, for different sizes of area A. It is clearly seen that the fiber breakage phenom-
enon is quite localized and the force which was carried by the fiber before its break is redistributed in the close
vicinity of this fiber. It should be emphasized that the previously mentioned requirement that the perturbed
elastic field region caused by the fiber break is much smaller than the total size of the finite domain of the
model is fulfilled. It is worth mentioning that Curtin and Takeda (1998) obtained, in the framework of the
shear lag analysis, a localization effect that extends over a rather wide region than the presently confined
one. This is probably due the interface sliding and yielding phenomena that were taken into consideration
by these authors.
M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129 4123

Table 1
The ratio of force redistribution as a result of the fiber breakage, Eq. (21)
N Size of the square domain Number of intact fibers d
1 33 8 78%
2 55 24 90%
3 77 48 96%
4 99 80 99%

It should be interesting to exhibit the entire axial stress distribution in the first intact fiber at the plane of the
crack. In Fig. 6, such distributions are shown, due to symmetry of the stress state, over one-half of the fiber for
a composite with a fiber volume fraction vf ¼ 0:2 and two values: Ef =Em ¼ 10 and 20. As was shown in Fig. 5,
the maximum normal stress occurs at the nearest point to the crack at the fiber-matrix interface and decays
with increasing the distance from the crack (x2 -direction). On the other hand, its variation in the perpendicular
x3 -direction is quite weak. This latter observation meets the results reported by Nedele and Wisnom (1994). A
comparison between the two cases shown in this figure reveals that for Ef =Em ¼ 20, both the magnitude of the
jump and the value of the stress concentration are more significant.
The effect of fiber breakage on the strength of the composite is usually (in particular in the framework of the
shear lag theory) characterized by the stress concentration factor (SCF) which is given by the normalized aver-
age normal stress in the first intact fiber:

Fig. 6. The axial stress distribution in the crack’s plane within the nearest to the crack intact fiber. (a) Ef =Em ¼ 10, (b) Ef =Em ¼ 20. The
fiber volume fraction is 0.2.
4124 M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129
Z Z
1
SCF ¼ f ð1Þ
r11 ðx2 ; x3 ÞdAf ð22Þ
pa2 r11 Af

where Af is the cross-section of the first intact fiber in the crack’s plane x1 ¼ 0. The results given in Fig. 5, on
the other hand, show that for the considered composite’s parameters such a characterization in terms of SCF
may not provide an adequate information about the stress distribution between and into the fibers adjacent to
the broken one. On the other hand, such a distribution obtained in the present study can serve for the devel-
opment of accurate load-sharing models which will be employed in the statistical modeling of composite fail-
ure. In order to characterize the maximal stress appearing in the fiber phase as a result from the single fiber
break, let us introduce the point stress concentration factor (PSCF) defined by the normalized maximum nor-
mal stress in the first intact fiber
1
PSCF ¼ f ð1Þ
max r11 ðx2 ; x3 Þ ð23Þ
r11 Af

Fig. 7(a) presents the dependence of SCF and PSCF on the matrix-to-fiber Young’s moduli ratio for a fiber
volume ratio of vf ¼ 0:5. It is readily observed that SCF is weakly dependent upon Em =Ef ratio. It should be
mentioned that in the simple version of the shear lag model where the axial stiffness of the matrix is not
accounted to, the SCF of the composite does not depend on the fiber to matrix moduli ratio. In the improved
shear lag model however, SCF in the nearest neighbor fiber diminishes as the axial stiffness of the matrix
increases. This feature can be well observed in Fig. 7(a). The figure also shows the variation of PSCF with
Em =Ef . Here, this variation is quite significant such that for a very compliant matrix this stress concentration
indicator is high as compared to SCF. Furthermore, the figure shows that the maximum stress in the first
intact fiber is significantly higher that the average stress especially for low values of Em =Ef . In the case of a
crack in a homogeneous material ðEf ¼ Em Þ, the values of SCF and PSCF obtained from the analytical solu-
tion are: 1.02 and 1.11, respectively. The relative differences of these values from the corresponding ones
obtained by the present analysis are 0.78% and 1.6%, respectively. It is worth mentioning that Fig. 7(a) shows
that SCF approaches a value in the vicinity of 1.1 for a very small ratio of Em =Ef . This result meets the clas-
sical value of 1.104 that was obtained by Hedgepeth and Van Dyke (1967) in the framework of the shear lag
approximation.
Fig. 7(b) illustrates the dependence of the normal stress distribution in the intact fiber upon the fiber volume
fraction vf for Ef =Em ¼ 20 in terms of SCF and PSCF. Here too, SCF exhibits a weak dependence upon vf
whereas PSCF is strongly dependent. The contrast between the values of SCF and PSCF is very high for high
values of fiber volume fractions. On the other hand, the values of SCF and PSCF are quite close for low values

a b

Fig. 7. (a) Variation of the stress concentration factors SCF and PSCF at the intact fiber against the ratio Em =Ef for a fiber volume ratio
vf ¼ 0:5. (b) Variation of the stress concentration factors SCF and PSCF at the intact fiber against the fiber volume ratio vf for
Ef =Em ¼ 20.
M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129 4125

of fiber volume fraction vf (i.e., when the fiber diameter is small with respect to the distance between the
fibers). This justifies the characterization of the strength of a composite with a broken fiber in this case by
the stress concentration factor (SCF) as it done in framework of the shear lag approach.
In the present paper, a particular attention is given to the stress variation within the fiber cross-section
rather than its average value. To this end, a comparison between the axial stress variation along the center
of the broken fiber and the average value is shown in Fig. 8(a) for Ef =Em ¼ 10 and vf ¼ 0:2. It is well observed
that there is no significant difference between the two values. This is in contradistinction with the situation in
the nearest intact fiber as shown in Fig. 8(b). This figure shows the axial stress variation along a line in the x1 -
direction that passes through the point A within the intact fiber. This variation is compared with the average
distribution of the axial stress in this fiber. It is clearly seen that a significant difference between the point and
average stress exists in the vicinity of the crack plane.
It is interesting to investigate the normal stress distribution along the fiber direction. This allows the estab-
lishment of the ineffective length defined by the distance over which the perturbed stress field is appreciable. To
this end, let us consider the normal stress distribution at the center of the broken fiber. In Fig. 9(a), this normal
stress distribution along x1 -direction is shown for several values of the fiber-to-matrix Young’s moduli ratios:
Ef =Em ¼ 20; 10; 4; 1. It should me mentioned that in the special case of a crack in a homogeneous material
ðEf ¼ Em Þ, the stress distribution as determined by the present method coincides up to the scale of the graph
by the one provided by the analytical solution (Sneddon, 1951). As expected, the asymptotic normal stresses
f ð1Þ
for large values of x1 =D2 approach r11 given by Eq. (2). The figure shows that the ineffective length increases
with the stiffness ratio Ef =Em . For the ratio of Ef =Em ¼ 20, 90% (say) of the asymptotic stress is obtained at
about x1 =D2 ¼ 3:5. This result is smaller than the ineffective length value of 6.8 as was obtained by Landis and
McMeeking (1999) who employed the approximate shear lag model in which the matrix cannot carry axial
load. It is important to examine the axial stress r11 distribution along the intact fiber. As discussed in conjunc-
tion with Fig. 5, the maximum axial stress in the crack’s plane is obtained within the fiber region at the inter-
face. Fig. 9(b) shows the axial stress distribution in the intact fiber along the line 0 6 x1 =D2 6 15, x2 =D2 ¼ 0:6
and x3 ¼ 0 for various values of Ef =Em and vf ¼ 0:5. The intersection of this line with the crack plane is shown
by point A at the inset to this figure. It is clearly observed that the maxima of the axial stresses occur at the
crack’s plane with a rapid decay with increasing x1 . It should be noted that perturbed region is significantly less
than the corresponding one in the broken fiber. The overall form of these curves (with dips) is similar to the
graphs given by Landis and McMeeking (1999), Case and Reifsnider (1996) and Nedele and Wisnom (1994).
In the latter two investigations however, the axial stress was calculated at the center line of the intact fiber
rather than near the interface where the maximum stress is obtained. In the framework of the shear lag model
employed by Landis and McMeeking (1999), only the average axial stress in the intact fiber can be examined.

a b

Fig. 8. A comparison between the point and average axial stress within the fiber cross-section. (a) Along the x1 -axis passing through the
center of the broken fiber, (b) along a line in the x1 -direction that passes through point A within the first intact fiber.
4126 M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129

a b

Fig. 9. (a) The variation of the axial fiber stress along the center line of the broken fiber. (b) variation of the axial fiber stress along the line
passing through point A, at x2 =D2 ¼ 0:6, x3 ¼ 0 for various values of Ef =Em . The fiber volume fraction is vf ¼ 0:5.

Fig. 10 presents the shear stress r12 distribution along the line mentioned above at the intact fiber-matrix
interface at x2 =D2 ¼ 0:6 and x3 ¼ 0 for various values of Ef =Em and vf ¼ 0:5. This figure shows that the max-
imum shear stress which may initiate a fiber-matrix debonding, occurs at a distance about the fiber diameter
from the crack plane. The value of this maximum slightly increases with the increase of the fiber-to-matrix
Young’s modulus ratio.
Thus far, the periodic composite has been assumed to possess a square fiber arrangement as shown by
Fig. 1(b). It should be important to compare the obtained results with the case where the periodic fiber-rein-
forced composite possesses the more realistic hexagonal configuration, see Fig. 11(a). This figure illustrates the
division of this hexagonal configuration into identical rectangular parallelepiped cells every one of which is
labeled by the tripletpffiffiffiðK 1 ; K 2 ; K 3 Þ. The dimensions of the cell are D1 ; D2 ; D3 where contrary to the square
arrangement, D2 ¼ 3D3 . Fig. 11(b) and (c) show the axial stress distribution in the longitudinal direction
within the broken and first intact fiber, respectively. Fig. 11(b) shows this distribution at the center of the bro-
ken fiber for both fiber arrangements. Since it was not possible to design discretized square and hexagonal
configurations with completely identical fiber volume fraction, results in this figure are given for vf ¼ 0:232
and vf ¼ 0:226 which are very close. It can be concluded that both configurations give identical longitudinal
stress distributions since the slight difference in the asymptotic values away from the crack is due to the small

Fig. 10. The variation of the shear fiber stress in the intact fiber along the line passing through point A (shown in the inset of Fig. 9(b)).
M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129 4127

a b

Fig. 11. (a) A fiber-reinforced material in which the fiber arrangement forms a hexagonal array. The broken fiber is located at cell (0, 0, 0).
(b) The variation of the axial fiber stress along the center line of the broken fibers in composites with square and hexagonal arrays. (c) The
variation of the axial fiber stress along the line in the x1 -direction passing through point A of the intact fiber in the square array (see inset to
Fig. 9(b)) and through point A of the intact fiber in the hexagonal array (see inst to (a)).

difference in the fiber volume fraction. Fig. 11(c) shows the distributions of the axial stresses in the closest
intact fiber to the broken one for the square and hexagonal configurations. The axial stresses due to these
two configurations were recorded along lines in the axial directions. The locations of these lines are indicated
by points A shown in Figs. 9(b) and 11(a) for square and hexagonal configurations, respectively. It is readily
observed that the resulting two axial stress distributions are very close with a less than 1% relative difference.
This difference which is also observed in Fig. 11(b) is of the order of the difference between the two fiber vol-
ume fractions. In conclusion, the comparisons shown in Fig. 11 indicate that the results obtained for a periodic
fiber-reinforced composite with a square configuration are applicable to a composite in which the fiber
arrangement is of a hexagonal type, as long as the fiber volume fraction and the stiffness ratio of the fiber
to matrix are identical.

4. Conclusions

An accurate three-dimensional modeling of the behavior of unidirectional periodic composites with a bro-
ken fiber has been presented. The model is based on the elasticity equations for both fiber and matrix constit-
uents. This model allows the consideration of a penny-shaped crack that represents the fiber breakage in the
unbounded periodic composite. The method of solution is based on the combined application of the represen-
tative cell approach and the higher-order theory.
4128 M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129

It was found that the present approach is a reliable tool for the analysis of damaged periodic composites.
The computational efficiency of the present method enabled the conduction of extensive parametric study of
the composite behavior. It was found that the normal stress distribution in the plane of the crack within the
intact neighboring fiber to the crack may be significantly non-uniform. As a result, it was possible to introduce
the concept of the point stress concentration factor which enables an accurate prediction of the strength of the
cracked composite. This is in contrast to the shear lag approach according to which the strength prediction is
based on the average stress in the intact fiber rather than its maximal value.
The dependence of the point stress concentration factor upon the fiber volume fractions vf and the stiffness
ratios Ef =Em of fiber to matrix phases were investigated. For high values Ef =Em and low values of vf , the pres-
ent predictions approach the corresponding values obtained by the shear lag approximation. Furthermore, a
comparison between the behaviors of composites with square and hexagonal fiber packing with the same fiber
volume fraction and fiber-to-matrix stiffness ratio reveals that their strengths are very close.
Since our modeling is based on the construction of unit jump Green’s functions, the present continuum
approach can generalized in several aspects. Some of these generalizations follow the corresponding ones that
were developed in the framework of the shear lag theory and its elaborations. The circular fiber geometry can
be readily extended to an arbitrary shape. To this end, one should construct unit jump Green’s functions in
accordance with the given fiber cross-section. Furthermore, it is possible to analyze the problem of a crack
penetrating the matrix phase by an appropriate construction of such Green’s functions over the cracked
region. In addition, a composite with several arbitrarily located broken fibers can be considered. Here, since
mirror symmetry with respect to the cracks planes does not exist, one needs to generate Green’s functions that
correspond to normal as well as to shear unit displacements jumps. Periodic composites with arbitrary multi-
directional fibers (e.g., woven composites) some of which are broken can be also considered by the suggested
approach. Finally, the present approach can be also extended to fiber and matrix constituents with arbitrary
anisotropic behavior.

Appendix

The stress field in an elastic transversely isotropic homogeneous material with an embedded penny-shaped
crack was established by Sih et al. (1981). The expressions for the stress components that were utilized in the
present paper for checking the accuracy of the proposed method of solution are given below. The transversely
isotropic material, in which the axis of symmetry is oriented in the x1 -direction, is subjected to a far-field stress
loading r11 , and the radius of the embedded penny-shaped crack is a.
Let
c22 nj  c66
mj ¼ ; j ¼ 1; 2
c13 þ c66

where nj are the roots of the characteristic equation

c22 c66 n2 þ ½c13 ðc13 þ 2c66 Þ  c22 c11 n þ c11 c66 ¼ 0

and ckl are p


theffiffiffiffiffiffiffiffiffiffiffiffiffiffi
elements of the stiffness matrix of the material. Then the axial stress r11 and the shear stress r1r
where r ¼ x22 þ x23 , being the radial distance from the x1 -axis, are given by

o2
r11 ðx1 ; rÞ ¼ r
11 þ c66 ½ð1 þ m1 Þn1 /1 þ ð1 þ m2 Þn2 /2 
ox21
o2
r1r ðx1 ; rÞ ¼ c66 ½ð1 þ m1 Þ/1 þ ð1 þ m2 Þ/2 
ox1 or

where the functions /j ðx1 ; rÞ are:


M. Ryvkin, J. Aboudi / International Journal of Solids and Structures 45 (2008) 4114–4129 4129

pffiffiffiffiffi Z 1  
2 n1 r 11 a2 1 cosðsaÞ sinðsaÞ sx1
/1 ðx1 ; rÞ ¼ pffiffiffiffiffi pffiffiffiffiffi  2 2 J 0 ðsrÞ exp  pffiffiffiffiffi ds
pc66 n1  n2 ð1 þ m1 Þ 0 s sa sa n1
pffiffiffiffiffi Z 1  
2 n2 r 11 a2 1 cosðsaÞ sinðsaÞ sx1
/2 ðx1 ; rÞ ¼ pffiffiffiffiffi pffiffiffiffiffi  2 2 J 0 ðsrÞ exp  pffiffiffiffiffi ds
pc66 n1  n2 ð1 þ m2 Þ 0 s sa sa n2

References

Aboudi, J., Pindera, M.-J., Arnold, S.M., 1999. Higher-order theory for functionally graded materials. Composites: Part B: (Engineering)
30, 777–832.
Beyerlein, I.J., Landis, C.M., 1999. Shear-lag model for failure simulations of unidirectional fiber composites including matrix stiffness.
Mech. Mater. 31, 331–350.
Case, S.W., Reifsnider, K.L., 1996. Micromechanical analysis of fiber fracture in unidirectional composite materials. Int. J. Solids Struct.
33, 3795–3812.
Curtin, W.A., Takeda, N., 1998. Tensile strength of fiber-reinforced composites: I. Model and effects of local fiber geometry. J. Compos.
Mater. 32, 2042–2059.
Goree, J.G., Gross, R.S., 1979. Analysis of a unidirectional composite containing broken fibers and matrix damage. Eng. Fract. Mech. 13,
563–578.
He, J., Beyerlein, I.J., Clarke, D.R., 1998. Load transfer from broken fibers in continuous fiber Al2 O3  Al composites and dependence on
local volume fraction. J. Mech. Phys. Solids 47, 465–502.
Hedgepeth, J.M., Van Dyke, P., 1967. Local stress concentrations in imperfect filamentary composite materials. J. Compos. Mater. 1, 294–
309.
Ibnabdeljalil, M., Curtin, W.A., 1997. Strength and reliability of fiber-reinforced composites: localized load sharing and associated size
effects. Int. J. Solids Struct. 34, 2649–2668.
Landis, C.M., McMeeking, R.M., 1999. Stress concentrations in composites with interface sliding, matrix stiffness and uneven fiber
spacing using shear lag theory. Int. J. Solids Struct. 36, 4333–4361.
Landis, C.M., Beyerlein, I.J., McMeeking, R.M., 2000. Micromechanical simulation of the failure of fiber reinforced composites. J. Mech.
Phys. Solids 48, 621–648.
Landis, C.M., McGlockton, M.A., McMeeking, R.M., 1999. An improved shear lag model for broken fibers in composite materials. J.
Compos. Mater. 33, 667–680.
Mahesh, S., Beyerlein, I.J., Phoenix, S.L., 1999. Size and heterogeneity effects on the strength of fiberous composites. Physica D 133, 371–
389.
Mahesh, S., Phoenix, S.L., Beyerlein, I.J., 2002. Strength distributions and size effects for 2D and 3D composites with Weibull fibers in an
elastic matrix. Int. J. Fract. 115, 41–85.
Nedele, M.R., Wisnom, M.R., 1994. Three-dimensional finite element analysis of the stress concentration at a single fibre break. Compos.
Sci. Technol. 51, 517–524.
Nuller, B., Ryvkin, M., 1980. On boundary value problems for elastic domains of a periodic structure deformed by arbitrary loads. In:
Proceedings of the State Hydraulic Institute, Energia, Leningrad (in Russian), vol. 136, pp. 49–55.
Reedy Jr., E.D., 1984. Fiber stress in a cracked monolayer: comparison of shear-lag and 3D finite element predictions. J. Compos. Mater.
18, 595–607.
Rossettos, J.N., Shishesaz, M., 1987. Stress concentration in fiber composite sheets including matrix extension. J. Appl. Mech. 54, 723–
724.
Ryvkin, M., Aboudi, J., 2007a. Analysis of local thermomechanical effects in fiber-reinforced periodic composites. Int. J. Fract. 145, 229–
236.
Ryvkin, M., Aboudi, J., 2007b. The effect of a fiber loss in periodic composites. Int. J. Solids Struct. 44, 3497–3513.
Ryvkin, M., Aboudi, J., 2007c. A continuum approach to the analysis of the stress field in a fiber reinforced composite with a transverse
crack. Int. J. Solids Struct. 44, 6826–6841.
Ryvkin, M., Nuller, B., 1997. Solution of quasi-periodic fracture problems by the representative cell method. Comput. Mech. 20, 145–149.
Sastry, A.M., Phoenix, S.L., 1993. Load redistribution near non-aligned fiber breaks in a two-dimensional unidirectional composite using
break-influence superposition. J. Mater. Sci. Lett. 12, 1596–1599.
Sih, G.C., Chen, E.P., 1981. Mechanics of fracture, vol. 6. Cracks in Composite Materials, Martinus Nijhoff, The Hague.
Sneddon, I.N., 1951. Fourier Transforms. McGraw Hill, New York.
Suemasu, H., 1984. Probabilistic aspects of strength of unidirectional fibre-reinforced composites with matrix failure. J. Mater. Sci. 19,
574–584.
Van den Heuvel, P.W.J., Goutianos, S., Young, R.J., Peijs, T., 2004. Failure phenomena in fibre-reinforced composites. Part 6: A finite
element study of stress concentrations in unidirectional carbon fibre-reinforced epoxy composites. Compos. Sci. Technol. 64, 645–656.

You might also like