Mechanics of Composite Materials and Laminated
Mechanics of Composite Materials and Laminated
Mechanics of Composite Materials and Laminated
Advanced composite materials usually consist of two constituents, i.e., the reinforcing
element and the supporting matrix. The reinforcing element being mush stiffer and
stronger than the matrix carries the load, and the matrix provides protection and
lateral support to the reinforcing element. Since materials are much stronger in fibrous
form than in bulk form, many reinforcing elements in advanced composites are high
performance fibers. It is conceivable that properties of fiber-reinforced composites
can be tailored by the selection of fiber and matrix systems and by varying fiber
volume fraction. Therefore, it is crucial to have a good understanding of available
reinforcing fibers and matrices and their properties.
1.1 FIBERS
The principal reinforcement fibers used in advanced composites include boron, car-
bon (graphite), glass, ceramic, and aramid (Kevlar) fibers. They are used in the
continuous form or chopped into various lengths. Boron, carbon, and ceramic fibers
offer excellent stiffness and strength properties, while glass and aramid fibers offer
high strengths and moderate stiffnesses. Table 1.1 presents typical properties of these
fibers. The properties of some steel and aluminum alloys are also listed for compari-
son.
Boron Fibers are manufactured by depositing boron on a tungsten fila-
ment. Currently, boron fibers are produced by chemical vapor deposition from a
gaseous mixture of hydrogen and boron trichloride (BCh) on an electrically heated
(to 1250C) tungsten substrate of 0.5 mil. (12.5 /-Lm) diameter. The tungsten sub-
strate is continuously pulled through the reactor. By varying the speed, the desired
boron coating thickness can be achieved. Currently boron fibers are produced in sizes
of 4 mils. (100 /-Lm), 5.6 mils. (140 /-Lm), and 8 mils. (200/-Lm) diameter. Boron fibers
have been modified by coating them with silicon carbide (SiC) and boron carbide
(B4 C) to improve the mechanical properties of their metal matrix composites.
Boron fiber exhibits very high strength and modulus properties. Unfortu-
nately, at this time, its cost is still very high due mainly to the high cost of tungsten
filament and boron trichloride. In order to lower the cost of boron fibers, the tung-
sten filament is replaced by a carbon monofilament. In addition, the drawing speed
of the boron carbon deposition reactor is substantially increased, leading to a higher
4 FIBERS, MATRICES, AND COMPOSITE MATERIALS
diameters of the glass fibers range from 0.1 to 0.8 mil. (3-20 jtm).
Aramid Fibers are the generic name for fibers formed from polymers. Du
Pont's Kevlar fibers are most important in composites applications. Kevlar fibers are
light and possess very high strength and rather high modulus. They are inherently
resistant to flame and high temperature; they do not melt but will decompose at
approximately 500C. Kevlar fibers are not brittle, and yarns of these fibers can be
woven into fabrics. They have the unique property in a negative coefficient of thermal
expansion in the longitudinal direction.
Kevlar fibers have a relatively high moisture absorption capability (up to 5-6
percent). They are subject to degradation by ultraviolet light with the consequence
of a substantial reduction in strength. However, this problem is minimized by the
fact that the matrix material in the composite may shield the fibers from ultraviolet
light. In general, Kevlars have good resistance to lubricants, oils, and solvents, except
for some strong acids.
A few other reinforcing fibers worth mentioning are silicon carbide (SiC) fiber,
silicon carbide whisker, and aluminum oxide (Ab03) fiber. These fibers possess
common quality in their inherent high temperature stability, thus are most suitable
for metal matrix and ceramic composites which require high temperature fabrication
processes.
Table 1.1 Mechanical Properties of Fibers, Matrix Materials and
Conventional Metals
Tensile Tensile
Modulus Strength Density
E (J",. P
Material GPa (rnsi) GPa(ksi) gjcm3
Fiber
E-Glass 77.0(11) 3.5(500) 2.54
S-Glass 85.0(12) 4.5(650) 2.48
Silicon Carbide (Nicalon) 190.0(27) 2.8(400) 2.55
Carbon (Hercules AS4) 240.0(35) 3.6(510) 1.80
Carbon (Hercules IM7) 280.0(40) 5.2(750) 1.80
Carbon (Toray T300) 240.0(35) 3.5(500) 1.80
Boron 385.0(55) 3.5(500) 2.65
Aramid (Kevlar-49) 130.0(18) 3.8(550) 1.45
Aramid (Kevlar-29) 65.0(9.5) 3.8(550) 1.45
Polymeric Resin
Epoxy 3.5-5.0(0.5-0.7) 0.05-0.13(7-20) 1.20
Polyester 2-4.5(0.3-0.07) 0.04-0.07(6-10) 1.10-1.40
Thermoplastic (PEEK) 5.0(0.7)
Metal
Steel 210.0(30) 0.34-2.10(50-300) 7.80
Aluminum Alloys 70.0(10) 0.14-0.62(20-90) 2.70
Titanium (Ti-6Al-4V) 110(16) 0.92(134) 4.46
6 FIBERS, MATRICES, AND COMPOSITE MATERIALS
1.3 COMPOSITES
Composites are formed by combining reinforcing fibers and matrix materials. The
properties of a composite obviously depend on the type of fiber and matrix used as
well as the fiber volume fraction. Mixing different types of fibers in forming hybrid
composites is also very attractive since hybrid composites may offer better properties
and potential savings.
In general, the reinforcements in a composite take the form of continuous fibers
or short fibers. For the former, fibers may be unidirectional or in the form of woven
fabrics. Multi-directional reinforcement preforms can also be produced using various
COMPOSITES 7
Nylon v..cuumBag
Glass Cloth Breather
(2.4)
where the two dummy indices i and j must be distinguished. A term with three or
more repeated indexes has no meaning unless it is used to denote a specific term with
no summation.
10 ELASTICITY OF ANISOTROPIC MATERIALS
Coordinate Transformation
Consider two Cartesian coordinates (Xl, X2, X3) and (x~, x~, x;). A base vector
is a unit vector parallel to a coordinate axis. Let el, e2, e3 be the base vectors for
the Xl, X2,X3 coordinate system, and e~,e~,e; be the base vectors for the x~, x~,x;
system. Since the coordinate axes are mutually orthogonal, we have
.(2.5)
. and
(2.6)
where a dot indicates scalar product and
if i = j
(2.7)
if i =F)
is the Kronecker delta.
A vector x can be projected onto the two coordinate systems with the result:
,,
x = xjej = xjej (2.8)
Taking the scalar product of (2.8) with e;, we obtain
I'
xje j . eIi = xjej . e Ii (2.9)
By using the relation (2.6), (2.9) yields
(2.10)
By defining
(3ij = e~ . ej (2.11)
(2.10) can be written as
(2.12)
A similar procedure by taking a scalar product with ej leads to the inverse
relation
(2.13)
Equation (2.12) or (2.13) gives the coordinate transformations between the
two sets of components of a vector x in the two coordinate systems.
Substitution of (2.13) into (2.12) yields
(2.14)
INDEX NOTATION AND TENSORlAL TRANSFORMATION 11
Example 2.1
Figure 1
In the two dimensional space, the base vectors for (Xl, X2) and (x~, x~) coor-
dinate systems shown in Fig. 2.1 are given by
and
e~ = (cosO ,sinO) , e; = (-sinO, cosO)
respectively. The coordination transformation matrix can be computed from (2.11)
as
Thus,
The above equations are readily recognized as the coordinate transformation between
the original system and the system obtained by rotating the original system counter
clockwise with an angle e.
(2.17)
A system is called a vector field or a tensor field of order one if it has three com-
e x:
ponents in the variables and if the components are related by the transformation
law
(2.18)
~i (Xl, X2, X3) = (3ki~(x~, X2' x 3)
The tensor field of order two is a system which has nine components tij in
t:
the variables Xl, X2, X3 and nine components j in the variables x~, x2' x3' and the
components are related by the characteristic transformation law
(2.19)
t ij = (3mi(3nfmn
(2.20)
tijk ... = f3mif3njf3pk ... t'mnp ...
Note that if all components of a tensor vanish in one coordinate system, then they
vanish in all other coordinate systems.
Contraction
Contraction is an operation on tensors that equates two indexes and sums over
that index, e.g.,
This result indicates that A.;i is a scalar and, thus, is invariant with respect to coor-
dinate transformation.
Partial Derivatives
If x; = f3 ij Xj, then for a vector Vj we have
OV l
_1
OXl,
(2.24)
This says that partial derivatives of any tensor field behave like the components of a
Cartesian tensor. It should be noted that this is not true in curvalinear coordinate
systems.
Quotient Rule
14 ELASTICITY OF ANISOTROPIC MATERIALS
The quotient rule states that if any two quantities in the above equation are
tensors, then the third quantity must also be a tensor.
2.2 STRAIN
When a solid is subjected to external loads, material points in the body are displaced.
If two material points experience a change in the distance between them, then defor-
mation is present. Displacements that do not result in changes in distance between
any two material points are called rigid body motions or rotations.
initial deformed
Figure 2
Consider a body in the initial state and the deformed state (see Fig. 2.2). The
positions of a material point in these two states are given by the position vectors x~
and Xi, respectively. The relation between these two positions is given by
i = 1,2,3 (2.25)
If the body remains as an integral body, the relation given by the above equations
must be unique and can be inverted to
(2.26)
The displacement that the material point travels from the initial state to the deformed
state is .given by the displacement vector
STRAIN 15
(2.27)
The displacement components Ui can be expressed as functions of the Lagrangian
variables (x?) or the Eulerian variables (Xi).
Let Po be a material point in the neighborhood of Po. The position of Po is
x? + dx? where dx? represents the difference in positions of Po and Po as shown in
Fig. 2.3. After deformation, the difference in positions of these two material points
is denoted by dXi. The initial and final distances between these two material points
are given by
~-----X1
Figure 3
Noting that
we can write
16 ELASTICITY OF ANISOTROPIC MATERIALS
(2.30)
where
. (2.31)
Infinitesimal Strain
In structural materials, the range of material stretching is usually limited.
This condition leads to small displacement gradients, i.e.,
Consequently, the product terrns in the Lagrangian strain tensor and in the Eulerian
strain tensor can be neglected with the result,
(2.34)
(2.35)
Noting that
(2.36)
we conclude that the two infinitesimal strain tensors have identical numerical values
and that the distinction between the Lagrangian and Eulerian descriptions' of the
deformed state vanishes.
Henceforth, we will use eij to denote the infinitesimal strain components,
Figure 4
1 ds 2 - ds~ ~ ds - ds o
ell =2 ds 2 = ds (2.38)
Thus, ell has the same meaning as the elementary definition of uniaxial strain. In-
terpretations for e22 and e33 can be obtained in a similar manner.
18 ELASTICITY OF ANISOTROPIC MATERIALS
dx~ - dS Q
After deformation, the corresponding material elements are denoted by dXi and axi,
respectively. Denoting the angle between dXi and axi by <P12 (see Fig. 2.4) and taking
the scalar product of these two vectors, we obtain
(2.39)
(2.40)
Thus, el2 can be interpreted as half the change of angle between the two material
elements originally parallel to the Xl - and X2- axis, respectively.
Henceforth, unless otherwise specified, eij will denote the infinitesimal strain
tensor. Since eij is a second order tensor, its components follow the coordinate
transformation law as
(2.41a)
The equations in (2.41a) can also be written in matrix form as
el e11
e2 e22
e3 II e33
{e} II (2.42)
e4 e23
e5 el3
e6 el2
in which the (x, y, z) coordinate system is used. Collectively, the engineering strain
components do not form a second order tensor, and, thus, do not follow the tensorial
coordinate transformation law. However, since they are related to the components of
the strain tensor, their coordinate transformation can be performed via the compo-
nents of the strain tensor.
If the engineering strain components are used, we define
20 ELASTICITY OF ANISOTROPIC MATERIALS
C1 Cxx
c2 Cyy
By using the transformation law (2.43) and then (2.45), the coordinate transformation
for the engineering strain components is obtained. We have
~F
ilS
Figure 5
Figure 6
Take a free body in the form of a small rectangular prism with the stresses
on the six faces shown in Fig. 2.7. Figure 2.8 shows a side view of this element.
The body force b (force per unit volume) is represented by the three components
bx , by and bz . If the body is in a state of static or dynamic equilibrium, the resultant
moment and force on the element must vanish. It is quite straightforward to show
that the moment condition can be satisfied if
and that the balance of forces can be assured if the stresses satisfy the following
differential equations:
22 ELASTICITY OF ANISOTROPIC MATERIALS
Y a
Ciyy+Ciyydy
aY
,,
aGyz razz
ITyz+-""'aY-dy
, azx .... _ , ,-
f--c; , CJXy+ ax dx
x;r ~ aazv dz
C"xx .... - -( k'Z'J az
I
azy
'<rzx aa dz
aOXx
0Xx+--
I
ax d
C"f.v ~ - - - - _~- az_ -0Xz+ OXzdx_ _
,'aazz
aiz+--dz ITyz
_ 8X7 x
, az ,-
/ C"yx"'" _ -(
,
z
Figure 7
To show that the stress components 0- xx, 0- yy, 0- z., 0- yz = 0- zy, 0- xz = 0- zx, and
o-xy = o-yx are sufficient to describe the state of stress at a point, consider the tetrahe-
dron shown in Fig. 2.9. On the three faces perpendicular to the coordinate axes. The
components of the three stress vectors are denoted by o-ij' The stress vector acting
on the inclined surface ABC is t, and the unit normal vector is n. The equilibrium
of the tetrahedron requires that the resultant force acting on it must vanish leading
to the following relations.
(2.51)
From the quotation rule and (2.51), we note that o-ij is a second order tensor. Thus,
we conclude that the state of stress in a body is completely given by the stress
STRAIN 23
Figure 8
components Uij. That is, given any surface with the unit normal vector ni, one is
able to determine the stress vector if the stress components are given.
Since aij is a second order tensor, its coordinate transformation law is identical
to that for the strain tensor eij. Thus,
(2.52a)
Figure 9
where the summation convention over the repeated indexes is used, and [C] is a
6 x 6 matrix whose elements are the elastic constants of the material. The relations
given by (2.56) do not distinguish the tensile behavior from the compressive behavior.
That is, when (2.56) is used, the material is assumed to have the same stiffness under
tension as under compression.
The matrix of elastic constants, [C], can be shown to be symmetric from
the strain energy consideration. The proof will be presented in Section 2.6. Thus,
there exist 21 independent elastic constants for the most anisotropic materials. Note
that, if the stress-strain relations are written in terms of components of the strain
tensor eij, i.e.,
then [c] is not symmetric for general anisotropic materials. This may explain why
engineering strains have often been used for anisotropic solids.
As the strain energy in a linear elastic material must be positive, the matrix
STRESS-STRAiN RELATIONS 25
where [BJ is called the matrix of elastic compliances. The symmetric property of
[GJ is inherited by [BJ.
The stress-strain relations in the (x', y', Zl) system can be obtained by using
(2.46), (2.53), (2.56), and (2.58). We have
(2.59)
(2.60)
where [G/J and [B/J are the elastic constants and compliance matrices, respectively, in
the primed coordinate system.
In Section 2.4, the following relations are shown to be valid from the consid-
eration of strain energy density.
Thus. the relation between {cr/}and {e /} are also expressed in the form
Engineering Moduli
It is a common practice to express elastic material properties in terms of the
so called engineering moduli which can be measured through the use of simple
tension and simple shear tests in which only a single stress component is present,
The Young's modulus is defined as the slope of the normal stress - normal strain
curve produced in simple tension; the Poisson's ratio is the ratio of the lateral strain
and the longitudinal strain in simple tension; and the shear modulus is the slope of
the shear stress - shear strain curve produced by a simple shear.
For anisotropic materials, a uniaxial tension may produce shear strains in ad-
dition to normal strains, and a shear stress applied in the x-y plane may induce shear
strains in the y - z and x - z planes. These extension-shear and shear-shear couplings
require additional moduli beyond those normally used for isotropic materials.
For generally anisotropic linearly elastic materials, the compliance matrix can
26 ELASTICITY OF ANISOTROPIC MATERIALS
Ex Ey Ez G yz G xz G xy
Ex Ey Ez G yz G xz G xy
Ex Ey Ey G yz Gxz G xy
where
Ei = Young's modulus in the i-direction
G ij = shear modulus in the i-j plane
Vij = Poisson's ratio measuring contraction in the j-direction due to uniaxial
loading in the i-direction .
'fIij,k = coefficient of mutual influence of the first kind which characterizes
normal strain in the k-direction caused by shear in the i-j plane
'fIk,ji = coefficient of mutual influence of the first kind which characterizes
shear the i- j plane caused by normal stress in the k direction
fJ.ij,kl = Chentsov's coefficients which characterize shear strain in the kcl plane
caused by shear stress in the i-j plane
Material Symmetries
If the internal composition of a material possesses symmetry of any kind, then
symmetry can be observed in its elastic properties. The presence of symmetry further
reduces the number of independent elastic constants.
Let (Xl,X2,X3) be a coordinate system and (x~,x2'x~) be the second system
which is symmetric to the first in accordance with the form of its elastic symmetry.
Since both systems are equivalent with respect to elastic properties, the stress-strain
relations should be identical in both coordinate systems: In other words, the matrix
[Cj should remain unchanged in the equivalent coordinate systems.
STRESS-STRAIN RELATIONS 27
Monoclinic Material
To illustrate the invariant property of [G], let us consider a material with one
elastic symmetry plane, say, the Xl - X2 plane. The equivalent coordinate system
(x~, x~,x~) arranged as shown in Fig. 2.10 is obtained from the (Xl, X2, X3) system by
reversing the direction of the x3-axis. The base vectors for the unprimed and primed
systems are
e1 = (1,0,0) , e2 = (0,1,0) , e3 = (0,0,1)
and
e'l=(I,O,O), e~=(O,I,O), e~=(O,O,-I)
1 0 0 ]
[.8]= 010 (2.66)
[ o 0 -1
Noting that [Tu] = [Te], the transformation matrices [Tu] and [T,] can be calculated
using (2.44) and (2.47), respectively. With this [Tu], the stress components in the
primed system are related to those in the unprimed system as
all
a22
a33
- (2.67)
~a23
-a13
a12
Cll
C22
C33
- (2.68)
-123
-113
112
, =G
all +G
llcll ' 12 C22'+G
13c33'+G '+
14123 G 'G
15113 + 16112
' (2.70)
28 ELASTICITY OF ANISOTROPIC MATERIALS
X3
"------7'---- X1 I X'
I
I
I
I
X3
I
Figure 10
(2.71)
To satisfy the invariant property of [C] in these two coordinate systems, it is necessary
that
direction. Thus, if a new coordinate system obtained by reversing at least one of the
coordinate axes is used to describe the stresses and strains, then the stress-strain re-
lations should remain unchanged; i.e., the matrix [C] should be invariant with respect
to such changes in the coordinate system. This condition on [C] requires certain re-
lationships among the elastic constants as discussed earlier. Explicitly, this condition
leads to the following
Cu C12 C13 0 0 0
C12 C22 C23 0 0 0
C13 C23 C33 0 0 0
(2.73)
0 0 0 C44 0 0
0 0 0 0 C55 0
0 0 0 0 0 C66
It is noted that the number of independent moduli reduces to nine. In terms of the
engineering elastic moduli, the stress-strain relations for an orthotropic material can
be expressed in the form
1 V21 V31
0 0 0
E1 E2 E3
Vl2 1 V32
eu 0 0 0 au
E1 E2 E3
e22 Vl3 V23 1 a22
0 0 0
e33 E1 E2 E3 a33
- 1 (2.74)
123 0 0 0 0 0 a23
113 G23 a13
1
112 0 0 0 0 0 a12
G13
1
0 0 0 0 0
G12
Due to the symmetric property of the compliance matrix, the following relationships
are obtained
(2.75)
The relations between the stiffnesses Cij and the compliances 8ij are
8 22 8 33 - 8?3
Cu =
A
30 ELASTICITY OF ANISOTROPIC MATERIALS
8 u 8 33 - 8f3
C 22 -
.6.
8 u 8 22 - 8f2
C 33 -
.6.
8 13 8 23 - 8 12 8 33
C 12 - (2.76)
.6.
8 12 8 23 - 8 13 8 22
C 13 -
.6.
8 128 13 - 8 23 8 u
C23 -
.6.
1 1 1
C 44
8 44
, C 55 =8 ' C 66 =-
8
55 66
where
8 u 8 12 8 13
.6.= 8 12 8 22 8 23 = 8u822833 + 2812823813 - 8;3822 - 8u8~3 - 8;2833
8 13 823 8 33
The expressions for 8 ij in terms of Cij are similar to (2.76) with 8 ij and C ij
interchanged. From (2.74), the engineering moduli can be expressed in terms of the
compliances as
111
E1 - , E2 = - , E3 = -8 '
8u 8 22 33
8 21 8 31 8 32
V12 - V13 = - - V23 = - - (2.77)
8u ' 8u 8 22
1 1 1
G23 - - G13 = - , G12 = -
8 44 ' 8 55 8 66
The elastic constants Cij can also be expressed in terms of the engineering
moduli as
1 - V23V32
Cu E 2 E 3 .6.'
1 - V13V31
C 22 -
E 1E 3 .6.'
1 - V12V21
C 33 -
E 1E 2 .6.'
C 12 -
V21 +
V31 V 23
(2.78)
E 2 E3.6.'
C 13
V31 +
V21V32
E 2 E3.6.'
C23 -
V32 +
V12V31
E 1E3.6.'
C4 4 G23 , C 55 = G13 , C 66 = G12
STRESS-STRAIN RELATIONS 31
where
b.' = 1 - Vl2 V21 - V23 V 32 - V3l Vl3 - 2V2lV32Vl3
EIE2 E 3
The expression for the engineering moduli in terms of elastic constants Gij can be
obtained from (2.77) by using the relations between 8ij and Gij . An alternate ap-
proach is to consider simple tension and shear deformations and obtain the moduli
from their definitions. The results are
(2.79)
The other Young's moduli (E2 and E 3 ) and Poisson's ratios (V23 and VI3). can be
obtained by permuting the indices. The relations between the shear moduli and the
elastic constants are obvious from (2.76).
Table 2.1 lists the engineering moduli and thermal expansion coefficients of
some typical advanced composite materials.
Composite EI E2 G l2 Vl2
Type GPa (msi) GPa (msi) GPa (msi)
82 Glass/Epoxy 43.3 (6.2) 12.7 (1.8) 4.5 (0.65) 0.29
with respect to a rotation of the X2- and X3-axes about the xraxis. As a result of
this invariant condition, [C] reduces to
Cu C 12 C 12 0 0 0
C 12 C 22 C 23 0 0 0
C 12 C23 C 22 0 0 0
(2.80)
0 0 0 HC22 - C 23 ) 0 0
0 0 0 0 C 66 0
0 0 0 0 0 C66
The number of independent elastic moduli is five.
Transversely isotropic solids can be considered as a subset of orthotropic mate-
rials. If an orthotropic material also possesses transverse isotropy, then its engineering
moduli have the following relations
E3 , V12 = V13 , C 12 = C 13
E2
2(1 + V23)
Many unidirectional fiber composites can be modeled as a transversely isotropic solid
with reasonable accuracy.
Isotropic Material
In an isotropic material, every plane is a plane of symmetry and every direction
is an axis of symmetry. The matrix [C] reduces to
Cu C 12 C 12 0 0 0
C 12 C u C12 0 0 0
C 12 C 12 C11 0 0 0
(2.81)
0 0 0 !(Cu - C 12 ) 0 0
0 0 0 0 HCu - C 12 ) 0
0 0 0 0 0 ~(Cu - C 12 )
where
(1 - v)E
Cu -
(1 + v)(l - 2v)
vE
C 12 -
(1 + v)(l - 2v)
It is evident that two elastic. constants are sufficient to describe the stress-strain
relationship in an isotropic material. It is also noted that the shear modulus
1 E
C = "2(Cu - C12 ) = 2(1 +v)
(2.82)
If the material is linearly elastic, i.e., ai = CijCj, where Cij are constants, then
(2.83)
lIT 1 T
W= = -{c} [C] {c} = -{c} {a}
-CijCiCj (2.84)
222
where the superscript T denotes the transposed matrix.
Alternatively, the strain energy function can be expressed in terms of stress
components and the compliances Sij as
lIT 1 T
W = -SijO"iaj = - {a} [S] {a} = - {a} {c} (2.85)
2 2 2
Since the strain energy function is a scalar and is invariant with respect to
coordinate transformation, we can write
W - ~{cf[Turl{a'}
_ ~ {a}T [Terl {c'}
~ {c}T [Te]T {a'}
~ {a}T [Tu]T {c'}
The relations of (2.61) and (2.62) are thus obvious.
34 ELASTICITY OF ANISOTROPIC MATERIALS
Since the strain energy density W is always positive, from (2.84) and (2.85),
we conclude that both [C] and [8] are positive. Thus, their diagonal terms C11 , C22 ,
... , C66 , and 8 11 , 8 22 , ... , 8 66 are positive quantities, and the determinants of all the
principal diagonal submatrices of both matrices are also positive.
Consider orthotropic materials. From (2.76) and the fact that A > 0, we
obtain
8 22 8 aa > 8~a
8 11 8 aa > 8fa (2.88)
8 11 8 22 > 8f2
These inequalities can be expressed in terms of the engineering moduli given by (2.74)
with the result:
Ea 2
> V32
E2
EI 2
- > VIa (2.89a)
Ea
El
> V2l2
E2
Similar inequalities involving V2a, Val, and V2l can be obtained from (2.75)
and (2.89a) as
E2 2
> V2a
Ea
Ea 2
> Val (2.89b)
El
E2
> V22l
El
Using the conditions A > 0 or A' > 0, we obtain the additional inequality on the
engineering moduli:
(2.90a)
or equivalently,
(2.90b)
of motion. In the dynamic analysis of solids, it is often more convenient to write the
equations of motion in terms of displacement components, ul, U2, and U3. By using
the stress-strain relations and the strain-displacement relations, the displacement
equations of motion can easily be obtained from (2.50). For an orthotropic solid, the
displacement equations of motion with respect to the material principal axes are
{)2U1 {)2u1 {)2u1 {) [ {)U1
C44~
uX1
+ C66~
uX2
+ C55~
uX3
+,,--
UX1 .
(Cl l - C44 ),,--
UX1
(2.91)
In (2.91), p is the mass density; b1 , b2 , and b3 are the components of the body force
(excluding the inertia force); and Ull U2, and U3 are the displacement components in
the material principal directions, Xl, X2, and X3, respectively.
PROBLEMS
1. Verify that t;fjij = tj when 8ij is the Kronecker delta, i.e., 8ij = 1 if i = j and
8ij =0 if i =f j.
2. The base vectors for two Cartesian coordinate systems are given by
and
e~ = (0,0,1)
X' = (3ij Xj
36 ELASTICITY OF ANISOTROPIC MATERIALS
3. Verify that (3ik (3jk = {iij. Find the components of the vector
Find the components Tljk in the primed system. Use the two coordinate systems
in Problem 2.2.
!]
b) The deformation in a body is given by the strain tensor
21 13 00] 10- 2
[eij] =
[000 X
DISPLACEMENT-EQUATIONS OF MOTION 37
Find the elongation (per unit length) of a material element parallel to the
direction (~, - ~, 0) .
7. For infinitesimal strain components eij , show that eii represents the volume
change (per unit volume)
Find the three components of the stress vector t on the surface ABeD as shown
in Figure 2.11. Find the normal component O'n (perpendicular to the surface)
of the stress vector.
X2
, --x
, C
, ,,
,
,,
Fig. 2.11
[0'] = [~o ~o }J
Show that on any surface the force (or stress vector) is always perpendicular
to the surface and that the magnitude of the stress vector is equal to 0'0 ..
38 ELASTICITY OF ANISOTROPIC MATERIALS
!J
(10. Consider a 2-D isotropic solid. The stress-strain relations can be expressed as
1 v
E
o
v If O.
E E
1
o o G
E
Show that G = 2(1 + v) by using the invariant property of the stress-strain
relations with respect to coordinate transformation.
11. Show that the range of Poisson's ratios for 3-D isotropic solids is -1 ::; v ::; 1/2.
Find the corresponding elastic constants Cn , C22 , and C33 . Also consider the
case with VI2 = VI3 = V23 = O. Compare the values of El, E 2 , and E 3 . with
those of C n , C22 , and C33 , respectively.
13. A principal direction ni of the stress tensor O'ij satisfies
where 0' is the associated principal stress. Show that the principal directions of
O'ijand the corresponding strain tensor eij coincide in isotropic solids but not
in anisotropic solids.
Chapter 3
ANALYSIS OF A LAMINA
cll 0"11
{ C22
'Y12
}=
V12
- - =821
E1
1
- = 822
E2
0=826
{ } 0"22
0"12
(3.1)
1
0=861 0=862 -=866
G12
in which the relations given by (2.75) have been used. Note that there are four
independent elastic constants involved.
Inverting (3.1) we obtain
E1 V12 E 2
0
1- V12V21 1- V12V21
{ O"n } =
0"22 V12 E 2 E2 { Cll
C22 } (3.2)
0
0"12 1- V12V21 1- V12V21 'Y12
0 0 G12
The 3x3 matrix in the above relationship is usually denoted by
Q~J
Q12
[ Qll
[Q] = Q~l Q22 (3.3)
0
40 ANALYSIS OF A LAMINA
X2
()
Figure 1
in which the elements Qij are called reduced stiffnesses which should not be confused
with elastic constants Gij .
In stress analyses, sometimes a coordinate system x - y is set up which does
not always coincide with the material principal axes, x! and X2 as illustrated in Fig.
3.1. The two sets of stress components with respect to these two coordinates systems
are related by the reduced transformation matrix [TaJ:
O"u O"xx
{ 0"22 } = [TaJ { O"yy
0"!2 0" xy
} (3.4)
where
2 2
sin 0 2 sin 0 cos 0 ]
[ cos2 0
[TaJ = sin 0 cos2 0 -2sin 0 cos 0 (3.5)
-sinOcosO sin 0 cos 0 cos 20 - sin2 0
The elements of [TaJ can be obtained from (2.44) by noting [TaJ = [TeJ
In the same marmer, the strains with respect to the two coordinate systems
are related by
(3.6)
where
cos 2 0 sin2 0 sin 0 cos 0 ]
2
[TeJ = sin 0 cos2 0 -sinOcosO (3.7)
[ -2sinOcosO 2sinOcosO cos 2 0 - sin2 0
Note that the inverses [Tar! and [Tet! can be obtained by replacing 0 in (3.5) and
(3.7) with -B.
PLANE STRESS EQUATIONS FOR ORTHOTROPIC MATERIALS 41
Thus, the stress-strain relations for the state of plane stress parallel to x - Y (XI - X2)
plane become
xx
axx} .
ayy = [Q] {eeyy } (3.8)
{
a xy Txy
where
[Q] l
= [Tar [Ql [T.l = [T.r [Ql [T.l (3.9)
The explicit expressions for the elements in [Ql are given by
Ql1 - Ql1Cos4{1+2(QI2+2Q66)sin2{1cos2{1+Q22sin4{1
Q I2 (Ql1 + Q22 - 4Q66) sin2 {I cos 2 {I + QI2 (sin4 (I + cos4 (I)
Q 22 - Ql1 sin4 {I + 2 (Q12 + 2Q66) sin2 {I cos 2 {I + Q22 cos4 {I (3.10)
QI6 (Ql1 - QI2 - 2Q66) sin {I cos3 {I + (Q12 - Q22 + 2Q66) sin3 {I cos {I
Q26 - (Ql1 - QI2 - 2Q66) sin3 {I cos {I + (Q12 - Q22 + 2Q66) sin {I cos3 {I
Q66 - (Ql1 + Q22 - 2QI2 - 2Q66) sin2 {I cos 2 {I + Q66 (sin4 (I + cos 4 (I)
xx
{
eX.,
eyy } = [8] ayy }
{ a (3.11)
'Yxy a xy
where
(3.12)
and
{ }
a xx
{ }=
Cxx 1
_Vxy TJxy,y
Cyy a yy (3.14)
Ex Ey Gxy
a xy
"(Xy 'T}x,xy 7J y ,xy 1
Ex Ey Gxy
For an orthotropic material, the apparent engineering moduli can be expressed
in terms ofthe principal engineering elastic constants through the use of (3.14), (3.13),
and (3.1). The relations are
1 1
- -cos 4
11+ (1
--- 2l112). 2
- sm IIcos2 II+-sm1 . 4 II
Ex E1 G12 E1 E2
lI xy - - - (1
Ex [lI12 -+- 1 +21112
- - -1)- sin2 IIcos 2II]
E1 E1 E2 E1 G12
1 1 . 4 11+ (1 1 4 II
- -sm a-E 2l112). 2
sm IIcos 2 II+-cos (3.15)
Ey E1 12 1 E2
1
-1+ 4 ( -1+-1 +21112
- - - 1)
- sin2 IIcos 2 II
Gxy G12 E1 E2 E1 G12
- Ex [( - 2 + - - - - 1) sinllcos3 11- ( -2 + -
21112 21112
-- - 1) sin3 II cos II]
'T}x,xy
E1 E1 G12 . .E2 E1 G12
- Ey [( - 2 + - - - - 1) sin3 IIcosll- ( -2 + -
21112 21112
-- - 1) sinllcos 3 II]
'TJy,xy
E1 E1 G 12 E2 E1 G 12
Variations of the apparent moduli against fiber orientation II for some com-
posites are given in Fig. 3.2. Note that for the composites considered in Fig. 3.2, the
maximum coupling between extension and shear occurs between II = 10 and 20.
3.2 INVARlANTS
By substituting the following trigonometric identities
1
cos4 11 - "8 (3 + 4 cos 211 + cos 411)
cos 3 II sin II - ~ (2 sin 211 + sin 411)
1 .
cos2 11 sin2 II - "8 (1 - cos 411) (3.16)
6 0.6
4 0.4
0.2
o ~~~~~~~~~~~ ol....i....L.J....l...I.....L.......L......L........I-...LJ...l.......J
o 10 20 30 40 50 60 70 80 90 o 10 20 30 40 50 60 70 80 90
eO eO
0.6 1
-e- CarbonlEpoxy -e- CarbonlEpoxy
~ Boron/Aluminum """"l!t- Boron/Aluminum
0.5 -e- GlasslEpoxy
0.5 -e- Glass/Epoxy
0.4 o
V'f lJx,xy
0.3 -0.5
0.2 -1
0.1 -1.5
o , i , I -2 ~~~~~~~I~.~~~~
o 10 20 30 40 50 60 70 80 90 o 10 20 30 40 50 60 70 80 90
Figure 2
44 ANALYSIS OF A LAMINA
Qn - U1 + U2 cos2B + U3 cos4B
Q22 U1 - U2 cos2B + U3 cos4B
Q 12 U4 - U3 cos4B
1 .
Q16 - 2U2 sin2B + U3 sin4B . (3.17)
1
Q26 - 2U2sin2B - U3 sin4B
Q66 U5 - U3 cos4B
where
1
Ul - "8 (3Qn + 3Q22 + 2Q12 + 4Q66)
1
U2 - 2 (Qn - Q22)
1
U3 - "8 (Qn + Q22 - 2Q12 - 4Q66) (3.18)
1
U4 - "8 (Qn + Q22 + 6Q12 - 4Q66)
1 .
U5 "8 (Qn + Q22 - 2Q12 + 4Q66)
are independent of coordinate transformation. From (3.18) we note that
(3.19)
- -
12 - Q66 - Q12= Q66 - Q12
are two "invariants"; i.e., their relations with Qij and Qij are not affected by a
coordinate rotation in the XI-X2 plane. By combining (3.18) and (3.19), we obtain
1
U1 - "8 (311 + 412)
1
U4 - - (h - 412)
8
1
U5 "8 (II +4h)
However, among the three only two are independent as
1
U5 = 2 (U1 - U4 )
These invariants were first introduced by Tsai and Pagano [3.1]. In the form
of (3.18), the transformed stiffnesses Qij depend only on four invariants U1, U2, U3
OFF-AXIS LOADING 45
and U4 and the fiber orientation 8, while in the form of (3.10), Qij depend on six
constants Qij. The expressions of (3.17) also identify the quantities that vary with 8.
In a similar manner, the transformed compliances 8ij can be written as
811 U; + U~ cos 28 + U~ cos 48
822 - U; - U~cos28 + U~cos48
812 - U~ - U~cos48
816 - U~ sin 28 + 2U~ sin 48 (3.20)
826 - U~sin28 - 2U~ sin 48
866 - 4U~ - 4U~ cos 48
where
1
U'1 - 8 (3811 + 3822 + 2812 + 8 66 )
1
U~ - 2" (811-8d
1
u'3 - 8 (811 + 8 22 - 2812 - 8 66 ) (3.21)
1
u'4 - 8 (811 + 8 22 + 6812 - 8 66 )
1
u'5 8 (811 + 822 - 2812 + 8 66 )
\-- ----
\ e-x \
\
\CO CO \
\ -_\
'---------
Figure 3
Cxx =811 aD
Cyy =812 aD (3.22)
"(xy =816 aD
46 ANALYSIS OF A LAMINA
It is seen that shear deformation can result from application of a normal load, except
when x- and y-axes coincide with the material principal axes, XI and X2. Coupling
between normal deformation and shear deformation does not exist in isotropic solids.
Integrating the strain-displacement relations for exx and eyy yields the dis-
placement components:
where f(y) and g(x) are arbitrary functions. Substituting (3.24) and (3.25) into
8ux 8u y
"(xy = 8y + 8x (3.25)
we obtain
_
"(xy = l' (y) + g' (x) =816 0"0 L
(3.26)
where a prime indicates differentiation with respect to the argument. From (3.26), it
is obvious that f(y) and g(x) must be linear functions of y and X, respectively, i.e.,
Thus, the displacement field in the composite panel under the uniform stress 0" xx = 0"0
IS
The deformed shape of the panel can be determined from the displacement field
(3.33-3.34), which is depicted in Fig. 3.3.
Figure 4
stress parallel to the x-y plane is assumed. The tractions on the longitudinal edges
are absent. We assume that a yy = 0 everywhere. Thus,
(3.36)
Using (3.36) to eliminate eyy from the stress-strain relations (3.8), we obtain
(3.37)
where
-2
- Q 12
Qll ----
Q22
Q16 Q26Q12 . (3.38)
Q22
-2
- Q26
- Q66 - - - -
Q
22
where
duo d'I/J
co (x) = dx ' K,(x) = dx ' (3.43)
Substitution of (3.41-3.42) into (3.37) yields
axx } =
{ a xy
[<?il <?i6] { 'Yoco } + y { 0K, }
Q16 Q 66
(3.44)
The resultant extensional force N, shear force Q, and moment M are defined
as
h/2
h/2
Q - J
-h/2
CT xydy = al6eo + a66/0 (3.45)
h/2
M - J
-h/2
CTxxydy = DK-
where
(3.46)
Assume that loads are applied only at the two ends of the off-axis composite
beam as shown in Fig. 3.4. From the consideration of force and moment equilibria,
we easily derive the following:
Simple Tension
Consider the case of simple tension with N = No, Qo = QL = 0, Mo = ML =
0, and the ends are not restrained. From (3.47) we have
No
eo - 2 (3.54)
a l6
an--
a66
al6
/0 - --eo . (3.55)
a66
which leads to
a16 + 1f; ) x
Vo = - ( -co +c . (3.57)
a66
It is easy to show that
in order to satisfy the end conditions vo(O) = vo(L) = o. This implies that Vo = 0
everywhere.
Since co is constant, thus Uo = cox, and the horizontal displacement Ux is given
by
Substituting (3.54) into (3.59) together with the definitions of aij, we obtain
where 0"0 = No/h is the uniform applied stress. It can be shown that the solutions
given by (3.33) and (3.60) are identical.
1f; - 0 at x = 0, L (3.61)
Vo 0 atx = 0, L (3.62)
1 2
D1f; = 2Qox + Mox + C (3.63)
1
C=O and Mo = --LQo
2
Thus,
1 2 1
D1f; = -Qox - -LQox (3.64)
2 2
A BEAM THEORY FOR ANALYSIS OF OFF-AXIS SPECIMENS 51
b16 = (3.66)
.( 1
vo= -6D x
3 L
+ 4D x
2 bu)
+h x QO~hNox
b16 (3.67)
in which the integration constant is set equal to zero because of the boundary condi-
tions vo = 0 at x = O.The other condition (vo = 0 at x = L) leads to
-*
Qo = b16 N. - b16 Q n N. (3.68)
L2h 0 - _* L2 0
bu + 12D bu Qn + h 2
Note that the deflection is S-shaped as depicted in Fig. 3.5, and that vo = 0 at
x = L/2.
Fig. 3. 5
52 ANALYSIS OF A LAMINA
t
The induced bending moments at the two ends are
-*
Mo = -ML = -!LQo = -
2
u
Q L
-*
N~)
L
(3.70)
2 bu Qu + h2
Solutions of (3.68) and (3.70) indicate that the induced shear force and bending
moment would diminish as the length (L) becomes much larger than the depth (h)
of the beam.
Using two-dimensional plane stress elasticity theory, Pagano and Halpin [3.2J
solved a similar problem but with different end conditions to approximate clamped
off-axis specimens. Instead of satisfying the clamped end condition, their solution
satisfies the approximate boundary conditions,
aux
= 0, uy=-=O
Ux
ay
at x = 0, y = 0 and x = L, y = o.
3.5 OBLIQUE TAB FOR OFF-AXIS TESTING
As discussed in the previous section, tension testing of an off-axis composite coupon
specimen using rigid grips would induce bending, and a simple uniaxial stress field
cannot be obtained. Conceptually, the simple tension stress field could be realized if
the grip could allow the associated shear strain to take place freely. Such a grip is
not yet available. Recently, Sun and Chung [3.3J developed an oblique end tab that
can generate a state of uniform tension in off-axis specimens using rigid grips.
Consider an off-axis coupon specimen under the uniaxial stress a xx = aD. The
displacement U x is given by (3.33) as
- -
Ux =8u aox+ 816 aoy
An alternative expression of (3.33) is given by (3.60).
From (3.33) we see that the positions (x,y) that have the same longitudinal
displacement must satisfy the following equation.
(811 x+ 816 y) aD = constant (3.71)
Equation (3.71) represents a straight line making an angle rjJ (see Fig. 3.6)against the
x-axis with
cotrjJ = --_-
816
(3.72)
811
If (3.59) is used, then we obtain an alternative expression as
-*
cot rjJ = ~~6 (3.73)
Q66
OBLIQUE TAB FOR OFF-AXIS TESTING 53
For () = 0 - -.
and 90, S16=Q16= 0 and ; = 90.
The plot of ; versus fiber orientation () for the AS4/3501-6 graphite/epoxy
composite is given in Fig. 3.7. In view of this unique property, an end tab can be
designed so that its end is oblique, making an angle ; with respect to the x-axis
as shown in Fig. 3.6. If made more rigid, this end tab would produce a uniform
longitudinal displacement along its oblique edge that conforms with the displacement
field produced by uniform tension.
+~ ~pX~ ___N.~
__
Fig. 3.6
100
90
80
70
60
50
40
30
20
o 10 20 30 40 50 60 70 80 90
9'
Fig. 3. 7
REFERENCES
3.1 Tsai, S.W. and Pagano, N.J., "Invariant Properties of Composite Materials,"
in Composite Materials Workshop, S.W. Tsai, J.C. Halpin, and N.J. ;Pagano
(eds.), Technomic Publishing Co., Lancaster, PA, 1968, pp. 233-253.
54 ANALYSIS OF A LAMINA
3.2 Pagano, N.J. and Halpin, J.C., "Influence of End Constraint in the Testing of
Anisotropic Bodies," Journal of Composite Materials, Vol. 2, No.1, 1968, pp.
18-31.
3.3 Sun, C.T. and Chung, 1., "An Oblique End Tab Design for Testing Off-Axis
Composite Specimens," Composites, Vol. 24, No.8, 1993, pp. 619-623:
PROBLEMS
V3.1 Given a carbon/epoxy composite panel under uniaxial loading, i.e.,O"xx = 0"0" O"yy =
O"xy = 0, plot "(xy as a function of the fiber orientation (J. The composite prop-
erties are
EI = 140 GPa, E2 = 10 GPa, G l2 = 7 GPa, 1I12 = 0.3
Fiber composites are heterogeneous media with distinct phases of fibers and matrix.
Due to the presence of large numbers of fibers, it is impractical to analyze a composite
by retaining its distinct phases and their exact geometries. If a composite appears
statistically homogeneous at a large scale, then it may be effectively represenwctiJy
arnacro homogeneous solid with certain effective moduli t
~
describe th " average"
..:nat~ial prope.!!les~~ 0e composl e. nce these effective moduli are derived, a
composite is then an yzed as a homogeneous arllsotropic solid.
Many methods have been proposed for evaluating effective moduli of a com-
posite. Some are primitive, such as the so called rule of mixtures, while some attempt
to account for the local deformation in the fiber and matrix, and thus are called micro
mechanics approaches. Common to these approaches is the consideration of a repre-
sentative volume element (RVE) or a typical unit cell which is a subregion of the
composite that repeats itself over the entire body. The effective material properties
are obtained from this representative volume element through various assumed defor-
mations. Since in a homogeneous medium the field quantities such as displacement,
stress and strain must be described by continuous functions, the state of stress or
strain in the RVE's in a small neighborhood must vary slowly. In other words, the
characteristic length of the macro deformation must be relatively large as compared
with the characteristic dimension of the heterogeneity of the original medium.
and
i,j=1,2,3 (4.2)
56 EFFECTIVE ELASTIC MODULI
respectively, where a;j and I':;j are the actual stress and strain tensor components in
the RVE. In contracted notations, we write
- 11
a;= V a;(x,y,z)dV i = 1,,6 (4.3)
v
and
e;= ~ 1v c;(x, y, z)dV i = 1,,6 (4.4)
respectively. For a fiber reinforced composite, the average stress and strain can be
written as
and
respectively. In (4.54.6), VI denotes the region occupied by the fiber, Vm the region
occupied by the matrix, and a{ (I':n and a;" (1':;") are the stress (strain) fields in the
fiber and matrix regions, respectively.
The effective elastic constants G;j are defined from the following average
stress-average strain relations:
or in matrix notation,
{a} = [GJ {e}
The inverse relations are given by introducing the effective compliances B;j, i.e.,
(4.8) .
or
{e} = [BJ {a}
The equivalence between the actual heterogeneous composite medium and the
homogeneous medium given by the average stresses, strains, and the effective elastic
constants needs to be examined. For this purpose, we consider a macroscopically
homogeneous state in which an RVE is subjected to appropriate boundary tractions
ti or boundary displacements Ui that would produce Uniform stress (a;j) and strain
(e;j) in a homogeneous medium, i.e.,
(4.9)
THE EQUIVALENT HOMOGENEOUS SOLID 57
or
(4.10)
The total strain energy stored in the RVE is equal to the work done at the boundary,
i.e.,
!Jt.U.dS
2 "
- !2 J fy 'J.. nJ ek
'
XkdS = !2 J ~
ax. (fyek Xk) dV
'J'
s S V J
- - - J -dV
-1 aek aXk
2 'J'
v axJ
V _ <
"2 aij eik . Ujk (4.11)
V ( __ )
- "2 aijeij
~J tiu,dS - !J~
2 ax. (au)dV
'JJ
s v J
(4.14)
In the above derivation, the equilibrium equations aaij/aXj = 0 have been invoked.
Comparing (4.14) with (4.11), we have
v21.3 cc=
-G 'l
J
-1 -3 2-1 ' lacdV (4.15)
V
58 EFFECTIVE ELASTIC MODuLI
From (4.15) we conclude that the equivalent homogeneous medium stores the same
amount of elastic strain energy as in the original body. The effective moduli thus
defined ensure the equivalence in strain energy between the equivalent medium and
the original heterogeneous material.
In deriving the effective elastic constants or moduli, one must select the RYE
and find the actual stress and strain distributions corresponding to the appropriate
boundary conditions associated with uniform stress and strain in the equivalent ho-
mogeneous medium. This usually is not a simple task. Approximations are often
employed in idealizing the geometry and the stress and strain fields in the RYE in
order to simplify the mathematics involved. Voigt and Reuss models are two such
examples.
(4.16)
Ui ~ (1 cl ej dV + 1CF; ej dV)
Vf Vnl
(4.18)
which have the same form as that for the rule of mixtures. This uniform strain
assumption was first proposed by Voigt [4.2J in connection with the related polycrystal
problem.
The dual assumption is that the stress is uniform in the RYE, i.e.,
(4.19)
(4.20)
VOIGT AND REUSS MODELS 59
The uniform stress assumption was originally proposed by Reuss [4.3] for application
to a polycrystaL A good discussion on these two models was given by Hill [4.4].
<. In general, neither assumption is correct. The implied stresses according to the
Voigt assumption of constant strain cannot satisfy the stress continuity condition at
the phase boundaries, and the implied Reuss strains cannot satisfy the displacement
continuity condition)It should be noted that the elastic constants Cr;and compli-
ances 8~ given by (4.18) and (4.20), respectively, are not inverses of each other as
they are estimated based on different deformation and stress fields in the composite.
Consider a representative volume element (RYE) of a macroscopically homo-
geneous and isotropic solid which is subjected to a state of hydrostatic pressure p.
The average stresses are
-
-
CT yy= CT- zz = -p (4.21)
o-xz = iiyz= 0 (4.22)
Using the stress strain relations for isotropic solids, (2.81), we have
where
is the volume change per unit volume of the representative volume element. The bulk
modulus K is defined as
K = _E- = C11 + 2C12 (4.24)
E:o 3
If the strain stress relations of (4.8) are used, then
(4.25)
K= 1 (4.26)
3 (811 + 2812 )
Consider a particulate composite which is to be represented by an equivalent
isotropic solid. Using the Voigt assumption, we obtain
K
V
= i (C{l + 2C(2) + ~ (C;;' + 2C~) = cfKf + cmKm (4.27)
If the Reuss uniform stress is imposed throughout the RVE, then we obtain
~ ,~
, . .
v-
S11 Sf
- cf 11
+ Cm sm11
Substitution of the above equations into (4.26) yields
1
3cf (S{l + 2S(2) + 3Cm (SIT + 2S~)
1
(4.28)
A similar examination of simple shear leads to the Voigt and ReUss estimates
of the effective shear modulus as
(4.29)
and
respectively.
The difference between the estimates can be put as
(4.30)
with a similar expression for the shear moduli. Thus, the Voigt values always ex-
ceed the Reuss ones. The difference becomes large if the rigidities between the two
constituent phases differ substantially.
(4.31)
or
1 _/.112 0
E1 E1
/.112 1
E1 E2
o (4.32)
1
0, 0
,~v--IJ-..' ;h4lV""' wi.-I.<,"", Ge:~
Within the RVE, th~ following are assumed:
m -
En = En: conStant strain (4.34a)
m
0"22 = 0"22: constant stress (4.34b)
m
0"12 = 0"12: constant stress (4.34c)
for any appropriate general deformation. Equation (4.34a) indicates non-separation
between the fiber and the matrix, while (4.34b, c) satisfy the continulty of stresses
at the interface of the fiber and the matrix. We arrive at these conditions often via
our physical intuition.
Since the stress and strain are constant in each constituent, integrations in
(4.3) and (4.4) can be carried out to yield the following relations.
(4.35a)
(4.35b)
(4.35c)
Equations (4.34) and (4.35) and the stress-strain relations of the fiber and
matrix can be used to obtain the stress-strain relations. for the equivalent orthotropic
solid. Alternatively, the special deformation associated with each effective engineering
modulus can be considered from which the particular modulus is obtained. The latter
approach is taken in the following derivations.
62 EFFECTIVE ELASTIC MODULI
To determine the effective longitudinal modulus EI and the Poisson's ratio VI2
we consider a macroscopically simple tension, i.e.,
(4.36)
In view of (4.34) it is evident that both fiber and matrix are in simple tension. Thus,
f
"11 - Efc{1 = E f E'll (4.37a)
f f -
c22 - -vfcll = -vf cll (4.37b)
m
"11 EmEu = Em E'll (4.38a)
f - m
-VmEll = -Vm Cll-
(4.38b)
c22
Thus,
EI = CfEf + emEm ---.---------~,
. (4.39)
Substitution of (4.37b) and (4.38b) into (4.3515)leads to
"
---.---------
Hence,
(4.40)
Applying the macroscopic simple shear U12i' 0, Ull =U22= 0, and using the
constant stress condition given by (4.34c), we obtain from (4.35c)
1'12 - Cj/'12
f + eml'12m
- -
"12 "12
- Cf-+em-
Gf Gm
- - +Gem
( Cf
Gf
-m ) -
"12
(4.41)
A HYBRID VOIGT-REUSS MODEL 63
(4.42)
and
(4.43)
respectively, in which the constant stress assumption, (4.34b), has been invoked.
Substituting (4.42) and (4.43) into (4.35b), we obtain the average strain E22as
(4.44)
(4.45)
(4.46)
From the strain-stress relations of the fiber and the matrix, we have
(4.47)
(4.48)
Using the constant strain assumption e{l = ell in (4.48) together with (4.45), e{l can
be eliminated from (4.47) with the result
f Cm (vfEm - vmEf) -
E (4.49)
0"11 =
Cf f + Cm E m
0"22
64 EFFECTIVE ELASTIC MODULI
Substitution of (4.49) into (4.46) yields the strain-stress relation for simple
tension in the x2-direction:
(4.50)
it
a
B
17
l1 a AF
I+-JCfa-+!
I_ a
AM
-I
x1
Fignre 4.2 An improved RVE for fiber reinforced composites
The effective engineering moduli for the combined regions AF and AM can be
obtained from (4.39), (4.40), (4.41), and (4.50). The corresponding reduced stiffnesses
are denoted by Q;j .
A SQUARE FIBER MODEL 65
The next step is to consider the composite mediwn consisting of region B and
region A with the latter already represented by an equivalent homogeneous mediwn
with Qj. This can be accomplished by adopting the following asswnptions regarding
the three modes of deformation.
A B
cll - cll = cll: constant strain (4.51a)
A B
c22 - c22 = c22: constant strain (4.51b)
A B
112 - 112 = 'Y12: constant strain (4.51c)
(4.52)
where QZ are the plane stress reduced stiffnesses for the matrix, and CA and CB are
the volwne fractions of regions A and B, respectively. The relations between the
plane stress reduced stiffnesses Qij and the elastic moduli are given by (2.93).
The effective elastic moduli for the composites are obtained from the effective
Q ij using the following relations
(4.53a)
(4.53b)
(4.53c)
(4.53d)
References
4.3 Reuss, A., "Berechnung der Fliessgrenze von Mischkristallen auf Grund der
Plastizitatsbedingung ftir Einkristalle," Zeitschrijt fUr angewandte Mathematik
und Mechanik, Vo!' 2, 1929, pp. 49-58.
4.4 Hill, R., "Elastic Properties of Reinforced Solides: Some Theoretical Princi-
ples," J.Mech.Phys.Solids, Vo!' 11, 1963, pp.357-372.
Problems
Prob 4.1 Derive the effective moduli E 1 , E 2 , G12,and 1112 for a unidirectionally fiber-
reinforced composite with transversely isotropic fibers and an isotropic matrix.
Use the hybrid Voigt-Reuss approach.
Prob 4.2 Consider the boron/aluminum with the following elastic properties.
Boron fiber E = 380 GPa G = 172 Gpa v = 0.1
Aluminum : E = 68 GPa , G = 26 Gpa , v = 0.3
b) Reuss assumption
c) Hybrid model
Unidirectionally reinforced fiber composites have superior properties only in the fiber
direction. In practical applications, laminae with various fiber orientations are com-
bined together to form laminated composites which are capable of carrying loads of
multiple directions. Due to the lamination, the material properties of a laminate
become heterogeneous over the thickness. Further, due to the arbitrary fiber orien-
tations of the laminae, the laminate may not possess orthotropy as each constituent
lamina does.
00
u(x, y, z) - I>iUi(X,y)
i=O
00
w(x,y, z) - LZiWi(X,y)
i=O
CLASSICAL LAMINATED PLATE THEORY 69
hl2
.1-----------
/
,,/ I hl2
// ./
./
x
Figure 5.2 Coordinate system for laminated plate
For laminated plates with thicknesses that are small as compared with the
lateral dimensions, the variation of displacement over the thickness is small. Conse-
quently, good approximation can be achieved by retaining the first few terms in the
series expansion given by (5.1). In view of this, we consider the following approximate
plate displacements (using 1/Ix and 1/Iy to replace Ul and Vl, respectively):
If we further require that the deformed midplane remains perpendicular to the plane
sections, i.e., the transverse shear strains vanish, then
awo
1/Ix ax
awo
1/Iy - (5.4)
ay
70 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
Substitution of (5.4) into (5.2) yields the displacements in classical plate theory:
8wo
u - uo-z--
8x
8wo
v - vo-z-- (5.5)
8y
w - Wo
where
8u o
8x
2wo
"'x = -8- - (5.7)
8x 2
Thus, the strains of the laminate are continuous over the thickness, and are described
by the in-plane strains c~, c~, ,~yand the curvatures K.x , K. y , K. xy of the mid-surface.
(5.8)
{ ~:
Nxy
} = ?{::: }
-h/2 (Y xy
dz (5.9)
PLATE CONSTITUTIVE EQUATIONS 71
and
{
~:
Mxy
} = T{::: }
-hj2 o-xy
zdz (5.10)
where h denotes the thickness ofthe plate, and the stress components O-xx, o-yy and o-xy
assume the values of o-~~ ,o-~ and o-~~ if z is located in the kth layer. The resultant
forces and moments are depicted in Figs. 5.3 and 5.4.
~--------------~----y
x
Figure 5.3 Resultant forces
x
Figure 5.4 Bending and twisting moments
If the kth layer occupies the region from z = Zk_lto Z = Zk (see Fig. 5.5), the
integrals in (5.9) and (5.10) can be expressed as
{
Nx}
::y = 'fzl
n Zk { o-xx }
~:: k dz
(5.11)
72 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
and
. (5.12)
where n is the total number of layers in the laminate. Substituting (5.8) into (5.11)
and (5.12), we obtain
(5.13)
and
(5.14)
- ---- ' - - - - -
h/2
Zo II
1 1
Figure 5.5 Lamina coordinates
Note that the quantities c~, c~, I'~y, "'x, "'y, and "'xy are independent of z. Hence,
the integrations in (5.13) and (5.14) can be performed. The results can be combined
PLATE CONSTITUTIVE EQUATIONS 73
where
h/2
(A;j, B ij , Dij) = J
-h/2
- (k) (
Qij
. 2)
1,z,z dz . (5.16)
or, explicitly,
- :21 ~ - (k)
L. Qij
( 2 2)
Zk - Zk_l (5.17)
k=l
1 ~ - (k)
3" L. Qij
(3 3)
Zk - Zk_l
k=l
Denoting the thickness and the distance to the centroid of the kth lamina by
tk and Zk, respectively, (see Fig. 5.5) we can also write (5.17) as
(5.18)
A66 - VoAUS - V3 A U 3
where
VoA - h
h/2
h/2
V2A J sin20dz
-h/2
= ttksin20k
k=l
(5.20)
h/2
V3A - J cos40dz
-h/2
= ttkcOS40k
k=l
h/2
V4A - J sin40dz
-h/2
= t
k=l
tk sin40 k
Similar expressions for Bij and Dij are obtained from (5.19) by replacing ViA (i = 0,
1, .... 4) with ViB and ViD, respectively. The coefficients ViB and ViD are defined as
h/2
V[OB,lB,2B,3B,4Bj = J [1, cos20, sin20, cos40, sin40] zdz (5.21)
-h/2
h/2
V[OD,lD,2D,3D,4Dj = J [1,cos20,sin20,cos40,sin40] z dz2
(5.22)
-h/2
(5.23)
{N} = [A]{cO}
{M} = [D] {,,;}
As a result, the in-plane forces {N} and strains {cO} are uncoupled from the bending
moments {M} and the out of plane deflection associated with the curvatures {,,;}.
SPECIAL CLASSES OF LAMINATES 75
In other words, when the plate is subjected only to in-plane forces, no out-cif-plane
deformations would occur. Similarly, when only bending moments are applied, no
extensions of the mid-surface of the plate would be induced.
Obviously, if [B] is not null, then coupling between extension and bending
exists. Thus, stretching a laminated plate by in-plane forces may also produce bending
and twisting in the plate and vice versa.
Aij - Qij h
1 - 3
Dij - 12 Qij h (5.24)
Eh A _ vEh
All - A22 = 1 -v 2 12 - 1 -v 2
A66 - Gh , AI6 = A 26 = 0
Eh3
Du - D 22 -- 12 (1 _ v2) DI2 = vD ll (5.25)
(5.26)
(5.27)
76 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
Equation (5.27) indicates that the laminate is effectively a 2-D anisotropic solid in-
plane stress and [A]/h is the effective elastic constant matrix. The inverse relation of
(5.27) is
{CO} = h [A'l{ o-} (5.28)
where
[A'] = [Ar l
The components A;j are given by
Comparing (5.28) with (3.12), we can relate the components A;j to the effective
engineering moduli for the laminate as
1
hAll
A~2 . Ai2
l/xy. - Vyx = - A~2 . (5.30)
All
Al6 A~6
'T/xy,x -
A66 'f)xy,y = A'66
1
hA66
If a symmetric laminate also possesses the property Al6 = A 26 = 0 (e g ,
[0/90]. and [45].), then the effective moduli can be explicitly expressed as
Ex - (AnA22 - Ai2)/hA22
Ey - (AllA22 - Ai2)/hAll
V xy A12/A22 (5.31)
Vyx - AldAn
Gxy A 66 /h
SPECIAL CLASSES OF LAMINATES 77
The apparent Poisson's ratio lI xy can be calculated using (5.30). Figure 5.6 shows lI xy
as a function of O. It is seen that negative values of 1Ixy are possible. Also note that
unusually high positive values of lI xy can be produced.
0.8
0.6
\;y
0.4
0.2
.Q.2
0 45 90 135 l8C
f1
Figure 5.6 Apparent Poisson's ratio lI xy in [0/0 + 25Js laminates
Antisymmetric Laminates
Nonsymmetric laminates may be used for fabricating pre-twisted components
such as turbine blades. The material properties of an antisymmetric laminate is anti-
symmetric about the midplane. For a laminate consisting of similar fibrous plies to be
antisymmetric, the number of plies must be even. For example, [+45/+30/-30/-45J
is an antisymmetric laminate. It is then easy to see that a +0 layer always accom-
panies a _0 layer located at the symmetric position with respect to the midplane.
From (3.10), we have
(Ql6to - - (QI6L
(Q26to - (Q26Lo
which lead to
Al6 = A26 = Dl6 = D 26 = 0
Thus, neither extension-shear coupling nor bending-twisting coupling is present in
antisymmetric laminates.
78 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
Cross-ply Laminates
A cross-ply laminate consists of an arbitrary number of plies each with fiber
orientation of either 0 or 90 to the x- axis. Since QI6=Q26= 0 for both plies, thus,
Al6 = A 26 = Bl6 = B 26 = Dl6 = D26 = O. An antisymmetric cross ply laminate is
one that has a 0 ply for every 90 ply at the symmetric position and vice versa. For
such laminates, we have the additional properties
(5.32)
Balanced Laminates
If for every +{I ply there is a -{I ply in a laminate, then the laminate is called
a balanced laminate. For laminates of this type, Al6 . A 26 = 0, but, in general, Dl6
#OandD 26 #0. ,C -1-4/'fJI"N/-~>J'-:=;
5.5 QUASI-ISOTROPIC LAMINATES
For symmetric laminates, the in-plane deformation {cD} and bending [x;] are uncou-
pled. The in-plane load-deformation relation is given by [A] as
(5.33)
If [A] is invariant with respect to coordinate transformation, then the laminate would
have the same longitudinal stiffness in every direction; Le., the in-plane laminate stiff-
ness is isotropic. However, a laminate possessing an isotropic [A] does not necessarily
imply an isotropic [DJ, thus the name quasi-isotropic laminate.
Consider a symmetric laminate having N fiber orientations each of which
consists of an equal number of composite plies. The total thickness of plies with the
same fiber orientation is
t=h/N
Using the expressions of Qij given in (3.15), we have
A11 h { NUl
N E E
+ U2 Ncos 2{1k + Ua N }
COS4{1k
N N}
h { NUI-U2Ecos2{1k+UaEcos4{1k
A22
Al6 -
N
h
N r N
2U2 ESin2{1k + Ua N E }
sin 4{1k (5.34)
A 26 -
h
N
{I2 EN
U2 sin2{1k -
N}
UaEsin4{1k
QUASI-ISOTROPIC LAMINATES 79
k=l
N
V3 + iV4 = Le 4 (k-l)7fi/N
(5.38)
k=l
The above are two geometric series whose sums are easily obtained as
1 - e 27fi
Vi +iV2 - 1 - e27fi / N
1 - e47fi
V3 + iV4 - 1 - e41Ti / N (5.39)
Using the expressions, it is easy to verify Vi = V3 = 1, 112 = V4 = 0, for N = 1 and
Vi = 112 = V4 = 0, V3 = 2 for N = 2. For N <:: 3, conditions (5.35) are satisfied
simultaneously, and we have
A11 - hU1
A12 - hU4
A22 - hU1 (5.40)
A16 -
A66 -
80 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
which indicate that the stiffness matrix [A] is invariant with respect to coordinate
transformation.
Apparently, the minimum number of fiber orientations to form a laminate with
isotropic in-plane stiffness properties is N = 3. An example is the [0/60]s laminate
which is called the 7r/3 laminate. The laminate [45/0/90]s (the 7r/4 laminate)
belongs to the group of N = 4. Other quasi-isotropic laminates corresponding to
higher values of N can be constructed in the same manner.
The in-plane stiffness of a quasi-isotropic laminate is similar to that of an
isotropic solid, i.e.,
Au
[A] = A~2 (5.41)
[
(5.42)
The stresses a;',,), a1~' and a;,~, in the kth lamina are calculated according to (5.8),
i.e.,
(5.43)
If the plate displacements UO, va, and Wo are known, then {eO} and {K} can
be calculated from their relations with the plate displacements. The laminar stresses
are then obtained from (5.43).
Example 5.2 LaminaT Stresses in a Quasi Isotmpic Symmetric Laminate (45/0/90js Subjected
to a Uniaxial Load N x
Thecomp=os~n~e~~ __ ~~~~__________________~
El - 20 X 106 psi E2 = 1.4 X 106 psi (a)
G 12 - 0.8 X 106 psi V12 = 0.3
~
0.42
1.41 ] X 106 psi (b)
o 0.8
LAMINAR STRESSES 81
[Q]90 o -
[1.41 0.42
0.42 20.13~ ] X 106 psi (c)
o 0 0.8
[6.4 4.8 4.7 ]
[Q]4S o - 4.8 6.4 4.7 x 106 psi (d)
4.7 4.7 5.2
[ 0.343 0.104 0]
[AJ - 0.104 0.3453 0. X 106 (e)
0 o 0.119
{Yxx }
{Yyy
{
G'xy 45 0
The distributions of the laminar normal stresses are shown in Fig. 5.7. It is
evident that the stress distributions from lamina to lamina are not continuous. Also
note the large compressive stress {Yyy developed in the 90 lamina due to its inability
to contract in the y-direction.
82 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
+45"
_45
=
f=
0
90
90 u x
0
_45 ~
+45 ~
+45 f::
_45 f-
0
90 ~
90 r- y
O
-45 f-
+45 f::
Figure 5.7 Normal stress distribution in the laminas
1.28 0.96 0 ]
[AJ = 0.96 1.28 0 x 105 lb/in (5.44)
[ o 0 1.03
By treating this laminate as an equivalent homogeneous orthotropic solid in plane
stress, the equivalent elastic moduli can be obtained from (5.31). We have
Ex = Ey = 2.8 x 106psi , G xy = 5.2 x 106psi
. (5.45)
v xy = Vyx = 0.75 1/x,xy = 1/xy,x = 0
Comparing these moduli with those of the unidirectional composite, we note
a significant increase in the shear rigidity. However, this is a.chieved at the expense
of the longitudinal modulus Ex.
-45 0 laminae, ~
O"xx = O"O,O"yy = O,O"xy ~o (5.46)
and that (O"xy)-450 = -(O"xy)+450. From the coordinate transformation for stress, we
have
0"12 = - sin 0 cos 00"xx + sin 0 cos OO"yy + (cos 2 0 - sin2 0) 0" xy (5.47)
For 0 = -450 , it is obvious that
1
0"12 = 20"0 (5.48)
Under uniaxial tension, the strains in both +45 0 and -45 0 are identical and
'Yxy = O. From the coordinate transformation law on strains,
en }
10 22 = [T.] { exx }
eyy (5.49)
{
"1 ,2 'Yxy
we obtain
"1 ,2 = -2sinOcosO(exx - eyy) (5.50)
In the -45 0 lamina, (5.50) gives
Thus,
(5.52)
. (5.53)
The relation (5.53) can be used to determine G 12 from the tension test of a [45]s
laminate specimen [5.1-5.2].
{lOT} = T
1022
={~~}l!.T (5.54)
T
"1 ,2
84 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
The mechanical strains that produce stresses are the total strains less the
thermal strains. The stress strain relations for a plane stress unidirectional fiber
composite are modified to
{a} = [Q]({e} - {e T }) (5.55)
in reference to the material principal axes.
In view of (5.54), the vector of thermal expansion coefficients (0<) should trans-
form identically as the strains. That is,
(5.56)
0 + 0<2sin20
2
O<x - 0<1 cos
. (5.58)
Using the plate strain displacement relations given by (5.6), the stress strain relations
(5.55) in the x - y coordinate system can be expressed as
(5.59)
Following the procedure in developing the plate theory, we obtain the plate
constitutive equations as
THERMAL STRESSES 85
are thermal in plane and bending loads, respectively. After performing integration
over the laminate thickness, the expressions in (5.61) and (5.62) can be written ex-
plicitlyas
n
- b..TI:[QL {ahtk (5.63a)
k=l
For free thermal expansions, no mechanical loads are applied, i.e., {N} = {M}. = O.
We have
(5.64)
(5.65)
For symmetric laminates, it is not difficult to see that {MT} = {O}. Thus, with [B]
= {O}, we obtain from (5.65)
(5.66)
The matrices [Q]oo, [Q]900, [Q]450 and [A] are given in Example 5.2. Using (5.63) we
obtain
1.37 }
N: - 1.~7.6.T lb/in
{
NIu
{MT} = {O}
In the absence of mechanical loads, {N} = {O} and {M} = {O} . It follows
that
{I~} - {O}
{ch - {cO} = [At1{NT} (a)
{ :::} =
U xy k
[Q"]k { ;~ =~~~: }
IXy - .6.To<xy
(b)
To estimate the stresses in the laminate after curing, the temperature change
.6. T must first be determined. Since thermal expansion coefficients as well as elastic
moduli are temperature-dependent, the evaluation of cUring stresses using the present
formulation is not precise. An approach to this problem is to use an effective' tem-
perature drop .6.T. For a 350F cure epoxy system, the range of .6.T is usually taken
between -250F and -300F.
TAILORING OF LAMINATE THERMAL EXPANSION 87
ilA. \
~
{ X 10-
3
. \ .~ ~'~
The corresponding thermal residual stresses &e obtained from (b). We ll: ~e .<, TIt, erf~
{ :::} ={ :~~} ={ ~~~;2 } psi I
a xy 00 0"12 00 0 \
{
:~~} = { ~~~;2 } psi
(712 45' 0
The curing stresses can be quite significant as compared with the trans~erse strength
of the unidirectional composite.
(5.67)
In deriving (5.67), (5.61 and (5.66) have been used. From (5.67), it is evident that the
effective laminate thermal expansion coefficients {a} depend on the lamina thermal
expansion coefficients as well as laminate stiffnesses.
<
88 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
(5.68)
where
(5.69)
(5.71)
where N is the number of fiber orientations in the laminate, and t = h/N is the total
thickness of all the laminae with the same fiber orientation. In the last step of (5.71),
the relation (5.56) has been used. .
Using the results of (5.35) with N :::: 3 and Ok = k1r/N, we have
N n N
- + Lcos20k =-2
2 k=l
N n N
- - - Lcos20 k =- (5.72)
2 k=l 2
TAILORING OF LAMINATE THERMAL EXPANSION 89
N 1 n
I.: cos Ih sin Ih = 2 I.: sin 2(1 k = 0
k=1 k=1
With the above relations, one can easily verify the following
I.:
NT N[110]
[Telk = - 1 1 0 (5.73)
k=1 2 0 0 0
(5.74)
(5.75)
The effective coefficients of thermal expansion ax and ay are computed using (5.67).
Figure 5.8 shows the variation of ax and a y as a function of (I. It is interesting to note
that negative thermal expansion can be produced by selecting a (I that falls within
16 and 37.
90 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
x 1O-6 rF
25
20
15
a 10
-5
0 15 30 45 60 75 90
eO
Figure 5.8 Effective thermal expansion coefficients of [BJ graphite/epoxy laminates
One of the methods used for deriving the plate equations of motion is the
integration of the 3-D equations of motion over the plate thickness. Specifically,
equations (5.76-5.78) are first integrated over the laminate thickness to ensure the
global balance of forces in the x, y, and zdirections, respectively. To ensure balance
of moments about x and y axes, equations (5.76) and (5.77) are multiplied by z and
then integrated over the plate thicknesses, respectively. Thus, five plate equations of
motion are obtained.
To illustrate this procedure, we now proceed to integrate (5.76) with Tespect
to z:
h/2
J 8axx
8x dz + J
h/2
8a xy +
Tydz J Tz
h/2
8a xz dz = J (
h/2
w
p Uo - z 88xo) dz (5.79)
-h/2 -h/2 -h/2 -h/2
h/2
P - J
-h/2
pdz
h/2
R J
-h/2
pzdz
It is obvious that when the laminae have the same mass density (e.g., all laminae are
of the same composite), R vanishes automatically.
In a similar manner, (5.77) can be integrated with the result
8Nxy + 8Ny = PV
o
_ R8wo (5.81)
8x 8y 8y
Integration of (5.78) yields
8Qx 8Qy p ..
--+-+q=
8x 8y
Wo (5.82)
where
h/2
(Qx, Qy) = J(ax.. ayz)dz
-h/2
92 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
and
The quantities Qx and Qy are the shear forces acting on the x face and y face,
respectively; q is the net applied transverse load (N/m 2), and q+ and q- denote the
transverse tractions on the top and bottom surfaces, respectively. .
Multiplying (5.76) by z and integrating with respect to z over the plate thick-
ness yield
{)M
-
x
- + {)M
{)x
--+
xy
{)y
Jh/2
{)u-xzz d z=
-
{)z
J(.
h/2
{)W o
PUoz-p--z
{)x
2) dz (5.83)
-h/2 -h/2
Noting that
we have
h/2
-J - - z
h/2
{)u xz
{)z
dz= - Q x
-h/2
where
h/2
1= J pz dz2
-h/2
is the mass moment of inertia.
A similar procedure on (5.77) leads to
{)Mxy {)My Q _ R
- - + -{)y- - y - Vo- I{)wo
{)x
-{)y (5.85)
Equations (5.80), (5.81), and (5.86) are the plate equations of motion in terms
of the resultant forces and moments. These equations can also be expressed in plate
displacement components uo, vo, and woo From the plate constitutive relations, the
resultant forces and moments can be written in terms of the plate displacements uo,
vo, and Wo as
8wo 82wo
- 2B66-- - B 26 - -
8x8y 8y2
8uo
Mx - Bu -
8x
+ BI6 ( -8uo
8y
+ -8vo)
8x
+ B 12 -8vo
8y
82wo
- Du - -
8x 2
82wo 82wo
-2D I6 - - - D I2 - -
8x8y 8y2
8uo ( 8uo 8vo)
My - B 12 -
8x
+ B 26 -
8y
+ -8x + B 22 -8vo
8y
-
82wo.
D I2 - -
8x 2
82wo 82wo
-2D26 - - - D 22 - -
8x8y 8y2
8uo ( 8uo 8vo)
Mxy - B 16 -
8x
+ B66 -
8y
+ -8x + B 26 -8vo
8y
82wo
- D I6 - -
8x 2
82wo 82wo
- 2D66-- - D 26 - -
8x8y 8y2
Substituting the above expressions into the plate equations of motion, (5.80),
(5.81) and (5.86), we obtain the plate equations of motion in displacements:
94 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
(5.88c)
Symmetric Laminates
For thin laminates undergoing transverse deflections, the mass coupling inertia
coefficient R and the rotatory inertia I are usually negligible and can be set equal to
zero. In addition, if the laminate possesses a symmetric stacking sequence, then Bij
= 0 and the equations of motion are simplified to
(5.89a)
(5.89b)
BOUNDARY CONDITIONS 95
(5.89c)
It is noted that, for symmetric laminates, the in-plane motion and the flexural
motion are uncoupled.
For an isotropic plate, D16 = D 26 = 0, and
Eh3
- D22 = D = 12(1 _ v 2 )
- vD
_ (1- v) D
2
and the transverse plate equation of motion reduces to the well known classical plate
equation
(5.90)
where
Un = Us = Wo = Mn = (5.92)
o Hinged-free in the normal direction
(5.93)
96 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
(5.94)
o Clamped
awo
Un = Us = Wo = - = 0 (5.95)
an
o Free
(5.96)
The displacement components Un and Us follow the coordinate transformation
of the components of a vector, i.e.,
Un - uocosa + vosina
uosina - vocosa (5.97)
where a is the angle between the x axis and the normal direction to the boundary
contour (see Fig. 5.9).
Figure. 5.9-
The resultant forces N n and N ns and the resultant moments Mn and Mns
follow the coordinate transformation of the components of the stress tensor. We have
Note that in the free edge boundary conditions (5.96) the last condition
aMns Q _ 0
~+ n-
Z q
~lllfffffffffllllff1!Jftlltttt~
a a
I- I- .1
Figure 5.10 Hinged-hinged laminate under uniform transverse load
AI6 - A 26 = 0
B22 - -Bll' BI2 = B66 = BI6 = B 26 = 0
DI6 D 26 = 0
Using the above properties and the assumption of cylindrical bending, the governing
equations (5.89) reduce to
(a)
(b)
(c)
where
B~I
D=Dn--
All
The general solution for the ordinary differential equation given by (c) is easily
obtained as
(d)
98 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
Since the deflection should be symmetric with respect to the center, we readily con-
clude that
C1 = C3 = 0
From (5.87) we note that, for cylindrical bending,
(e)
N x = constant =N;
This implies that the in-plane force in the laminate is uniform. From (e) we have
1 0 B11 dwo
uo(x)=-A Nxx+- -d (f)
11 A 11 x
The moment free boundary condition can be expressed using the following relation
Using the general solutions given by (d) and (f), the boundary conditions
become
NO Bnq 2 C 0
x + 6D a - 2B11 2 -
q 4 2C C 0
24Da +a 2+ 4 -
B11 0 1 2
- N - -qa + 2DC2 - 0
A11 x 2
MINDLIN PLATE THEORY 99
3AuD + Erl 2
C2 - qa
12AUDllD
C4 - -a2C2 - _q_a4
24D
En 2
Nx - 2En C 2 - 6D qa
o
N =--qa
En 2
x 3D n (i)
It is obvious from (h) that extension-bending coupling increases the deflection and
thus reduces the effective bending stiffness of the laminate Also note that increasing
bending stiffness Dn has the same effect as reducing extension-bending coupling.
It is interesting to note from (i) that the in-plane force N x is proportional to Eu.
For [0/90]' En is negative, and for [90/0]' En is positive. Thus, for a positive
(q > 0) load, the [0/90) laminate would be under in-plane compression, while the
[90/0) laminate would be under in-plane tension. Further discussion on bending of
unsymmetric laminates is given in Chapter 7.
is not small (to be discussed later) as compared to the characteristic length (wave
length) of deformation. A more accurate plate theory can be developed by a.J.low-
ing transverse shear strains Ixz and Iyz and considering uo, vo, wo, '!f;x and '!f;y as
independent kinematic variables. Such an approach was used by Mindlin [5.4] to
develop a shear-deformable theory for plates of isotropic materials. A similar theory
was developed by Reissner [5.5]. Later, several authors [5.6-5.7] extended Mindlin's
theory to composite laminates. The governing equations of the Mindlin plate theory
for composite laminates can be derived in exactly the same manner as discussed in
the previous sections. Results for the isothermal case are surmnarized as follows.
(5.99)
where
(5.100)
The resultant forces N x , Ny, N xy , the bending moments Mx, My, M xy , and A,
B, D matrices are the same as in classical plate theory.
Transverse shear forces:
Qy }
{ Qx = k
[~4 ~5] I~z
~5 A55
{I~z }
(5.101)
where
(5.102)
are the transverse shear strains (assumed constant over the plate thickness) in the
y - z and x - z planes, respectively; k is a shear correction coefficient introduced to
compensate for the error resulting from the constant shear strain assumption; and
Aij (i,j= 4,5) are the plate transverse shear stiffnesses defined as
h/2
A;j = J
-h/2
Cijdz i,j = 4,5 (5.103)
In (5.104), Cij are the transformed elastic constants in reference to the x-y-z coordi-
nate system.
MINDLIN PLATE THEORY 101
For plates of isotropic materials, various ways were suggested to determine the
value of the shear correction coefficient k. Basically, k was chosen to match certain
exact elasticity solutions. For example, Mindlin [5.4] matched the plate solution with
the exact elasticity solution for the natural frequency of the first antisymmetric mode
of thickness shear vibration (such as 'I/lx = eiwt , 'I/ly = Wo = 0) and obtained
k = 71'2/12. On the other hand, Reissner [5.5] obtained k = 5/6 based on consideration
of the exact transverse shear stress distribution. For general composite laIninates with
many possibilities of different stacking sequences, the above methods cannot yield a
unique value of k. Based on a numerical study of some laminates, Whitney and
Pagano [5.7] found that k = 5/6 was suitable for composite plates.
Equations of Motion
In terms of resultant forces and bending moments, the plate equations of .
motion are given by
fJNx fJN
xy ..
-fJx+ -fJy- Piio + R'I/lx
fJN fJN ..
xy
-fJx- +fJy-y - Pvo+R'I/ly
(5.105)
Boundary Conditions
For simplicity, consider the edge parallel to the y-axis. Various boundary
conditions associated with the Mindlin plate theory are given as follows.
o Simply-supported-movable in the plane of the plate
(5.106)
MINDLIN PLATE THEORY 103
Uo = Vo = Wo = Mx = Mxy = 0 (5.107)
N x = Vo = Wo = Mx = Mxy = 0 (5.108)
Uo = N xy = Wo = Mx = Mxy = 0 .(5.109)
o Clamped
Uo = Vo = Wo = 'l/Jx = 'l/Jy = 0 .(5.110)
o Free
N x = N xy = Mx = Mxy = Qx = 0 (5.111)
Example 5.6 Cylindrical Bending
Consider a symmetric cross ply laminate [0/90]s with a span L in the x-
direction and infinite in the y-direction. The static transverse loading q is a function
of x only, and the edges are uniformly supported in the y-direction. In addition, a
state of plane strain is assumed so that Vo = 'l/J y = o. Thus, the plate displacements
and rotations must be functions of x only, and the deflected surface is cylindrical. For
the symmetric cross ply laminate, we have
Bij = 0 (a)
Thus, the in-plane displacements U o and Vo are uncoupled from w o , 'l/Jx and 'l/Jy. If the
edges (x = O,L) are simply supported, it can be shown that Uo = o. The equilibrium
equations (5.105) then reduce to .
(b)
Du a (aw
2
o
ax'I/Jx 2 - + ax )
kAss 'l/Jx = 0 (c)
Wo = Mx = 0 at x = O,L (d)
(e)
104 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
qo
o (h)
B = L3 qO (i)
71'3 Dl1
') qOL2
(Wo) max = (Wo max + 2kA (j)
71' 55
where
'( ') L4qo
Wo max = 4D
71' 11
is the maximum deflection according to classical plate theory. The second term on the
right hand side of (j) is the additional deflection due to transverse shear deformation.
Consider the ratio
(Wo)max 1 71'2 Dl1
.;---::.;-::.:= = +-=--=- (k)
(W~)max L2kA55
For the [0/90]s laminate, it can easily be shown that
h
A55 = Z(G 13 + G23 )
in which
material system 1:
G 23 = 0.4 , 1112 = 02
.5
E2
material system 2:
, 1112 = 0.25
--MS1
4 .......... MS2
---MS3
(w.) .... 3
(w~) .... \,
2
\~-.R-.::
\\.......
.......~u.:".u.: ~
............ ........
1
o~~~~~~~~~~
o 5 10 15 20 25 30
Uh
Figure 5.11 Maximum deflection of [0/90]s laminates of three material systems
under cylindrical bending
106 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
References
5.1 Petit, P.H., "A Simplified Method of Determining the Inplane Shear Stress
Strain Response of Unidirectional Composites," Composite Materials: Testing
and Design, ASTM STP 460, 1969, American Society for Testing and Materials,
pp. 83-93.
5.2 Rosen, B.W., "A Simple Procedure for Experimental Determination of the Lon-
gitudinal Shear Modulus of Unidirectional Composites," Journal of Composite
Materials, Vol. 6, 1972, pp. 552-554.
5.3 Timoshenko, S. and Woinowsky-Krieger, S., Theory of Plates and Shells, 2nd
Edition, McGraw-Hill Book Company, New York, 1959.
5.4 Mindlin, R.D., "Influence of Rotatory Inertia and Shear Deformation on the
Bending of Elastic Plates, Journal of Applied Mechanics, Vol. 18, No.1, 1951,
pp. 31-38.
5.5 Reissner, E., "The Effect of Transverse Shear Deformation on the Bending of
Elastic Plates," Journal of Applied Mechanics, Vol. 12, No.2, 1945, pp. 69-77.
5.6 Yang, P.C., Norris, C.H. and Stavsky, Y., "Elastic Wave Propagation in Hetero-
geneous Plates," International Journal of Solids and Structures, Vol. 2, 1966,
pp. 665-684.
5.7 Whitney, J .M. and Pagano, N.J., "Shear Deformation in Heterogeneous Anisotropic
Plates," Journal of Applied Mechanics, Vol. 37, No.4, 1970, pp. 1031-1036.
Problems
5.1 Compute the A, B, ,D matrices for [02/902]' [902/02]' [0/90/0/90]' and [0/90]s
laminates. Summarize the characteristics of these matrices. The moduli of the
composite are
El = 140GPa, E2 = 10GPa, G 12 = 7GPa, V12 = 0.3, ply thi<;kness = 0.127mm
5.2 Consider the [0 2 /902], [902/0 2], and [0/90]s laminates in Prob. 5.1. Find the
inplane strains and curvatures produced by uniaxial loading N x = 5000N/m.
Compare the deformed shapes of [02/90 2] and [902/0 2] laminates. If the laminate.
in-plane stiffness in the x-direction is defined as
K _Nx
x - 0
ex
which laminate among the three has the highest in-plane stiffness?
MINDLIN PLATE THEORY 107
5.3 Find the bending moments that are needed to suppress out-of-plane deflection
in the [02/90 2J laminate (see Prob. 5.1 for composite properties) subjected to
a uniaxial load N x = 5000N/m and Ny = N xy = O.
5.4 Consider a [45Js laminate. If the constituent composite material is highly
anisotropic, i.e.,
E1 E2 and E1 G 12
show that the effective engineering moduli for the laminate can be expressed
approximately as
Ex ~ 4Q66 ~ 4G 12
G ~ Qu ~ E1
xy- 4 - 4
Qu - 4Q66 E1 - 4G 12
l/xy ~ ~
Qu + 4Q66 E1 + 4G 12
Compare these approximate values with the exact values for AS4/3501-6 graphite/epoxy
composite. .
5.5 Derive approximate expressions for the effective engineering moduli for the lam-
inates [0/90Js and [0/90/ 45Js.
5.6 Compare the in-plane longitudinal stiffnesses in the x-direction for [30/0Js
and [30 2 /OJs laminates of AS4/3501-6 graphite/epoxy composite. Which is
stiffer?
5.7 Plot the effective moduli Ex, G xy , and v xy versus (J for the angle-ply laminate
. [IiJs of AS4/3501-6 graphite/epoxy composite.
5.8 Find the shear strains CYXy) in the AS4/3501 - 6 graphite/epoxy composite
[45Js and [0/90Js laminates subjected to the shear loading N xy = 1000N/m.
Also find the lamina stresses O"u, 0"22, and 0"12. If the maximum shear strength
of the composite is 100d = 100MPa, what are the shear loads (Nxy ) the two
laminates can carry?
5.9 Use I:!..T = -150C to estimate the curing stresses in the [0/90Js laminates of
the following two composite materials.
AS4/3501 - 6 graphite/epoxy: E1 = 140GPa, E2 = lOGPa, G 12 = 7GPa .
V12 = 0.3, 0<1 = -1 X 10-6 rC, 0<2 = 26 X 10-6 rC
S glass/epoxy: E1 = 45GPa, E2 = 9GPa, G 12 = 4.5GPa
V12 = 0.3, 0<1 = 5 X 10-6 rC, .0<2 = 26 x 1O-6 / oC
5.10 A [0/90/0J laminate of AS4/3501-6 graphite/epoxy composite is confined in the
0 direction. (i.e., c~ = 0) but free to expand (or contract) in the 90 direction
(i.e., Ny = 0). Find the lamina stresses for I:!..T = -100C.
108 ELASTIC ANALYSIS OF COMPOSITE LAMINATES
5.11 A [0/90]. laminate of A84/3501-6 composite is fixed on all sides. Find the
reaction forces N x and Ny and the lamina stresses for fiT = -100 0 e.
5.12 Do Prob. 5.11 for the [30]s laminate. Also find the effective thermal expansion
coefficients ax and ay.
5.13 Due to the presence of curing stresses, the [0/90] laminate would warp after
curing. If the laminate is constrained so the ~y = ~xy = 0 but ~x oF 0, find the
curvature ~x after curing. Assume fiT = -150C and N x = Ny = N xy = Mx -'-
O. The composite is I84)8? ?1MO )KA.? Y I@, iiZl3B Nlrl composite plates.
What are the relations between these two sets of stresses?
5.16 A [0/90] unsymmetric crossply laminate is subjected to the in-plane loads N x =
N, Ny = N xy = O. (see Fig. 5.12.) The plate is simply supported along x = a.
Assume that the laminate is very long in the y-direction (so that Vo = 0,
~y = 0). Find the transverse deflection wo(x) in terms of A, B, D matrices.
90
0 X
Chapter 6
failure processes are not described. Further, they are all based on linear elastic analysis.
plane shear strength. Since the compressive strengths in composite materials are usually
different from the tensile strengths, X' and Y' will be used to denote the compressive
strengths.
Jan 1993 AAE555 132
Maximum stress theory states that a lamina fails if a state of stress (au, a22, (12)
(6.1)
or
(6.2)
or
(6.3)
If the normal stresses are negative, then the conditions given by (6.1) and (6.2) are
replaced by
(6.4)
(6.5)
If, instead of the maximum stresses, the maximum strains are registered at the
points of failure of a lamina, a criterion similar to the maximum stress theory can be
established. Specifically, a state of deformation would cause failure of a lamina if any of
the following inequalities is satisfied:
Jan 1993 AAE555 133
(6.6)
(6.7)
(6.8)
when Xg, Yg, and Sg are the ultimate tensile strains in the xI- and x2-directions, and the
maximum shear strain in the Xl - X2 plane, respectively. Again, to account for different
(6.9)
and
(6.10)
respectively, where X'g and Y'g are the ultimate compressive strains (negative values).
then
S
Sg=--
0 12
(6.11)
However, actual stress-strain curves for a composite usually exhibit some nonlinearity,
especially the in-plane shear deformation as shown in Fig. 6.2. It is obvious that
_ X
X --
g EI
Thus, using the measured strength S and that obtained from the measured strain Sg,
namely 012Sg, the maximum stress criterion will predict different results. In general, the
maximum stress and the maximum strain criteria yield different results for strength
predictions even if linear stress-strain relations are imposed and the relations of (6.11) are
Jan 1993 AAE555 134
assumed true.
One of the obvious weaknesses of the maximum stress and maximum strain
criteria is the fact that they disregard the combined effects of stress on failure. Such
strains are present. The failure criteria discussed in the following provide the remedy for
this deficiency.
HilI-Tsai Criterion
For isotropic materials, a popular yield criterion is the von Mises yield criterion
which states that yielding of a material begins when the distortion energy density equals
the distortion energy density at yield in simple tension. The distortion energy is the total
strain energy density less the dilational strain energy. Inherent in the theory is the
assumption that yielding is due solely to shear deformation and that pure dilation (a
volume change) such as deformation due to a hydrostatic pressure would not produce
yielding.
Hill [6.1] extended von Mises' isotropic yield criterion to account for orthotropy of
Tsai [6.2] used the expression (6.12) as a failure criterion for unidirectional
F, G, H, L, M, and N are determined from tests of the longitudinal strengths in the three
material principal directions and the shear strengths in the three orthogonal planes of
2F = _1_ + _1___1_
y2 Z2 X2
2G = _1_ + _1___1_
Z2 X2 y2
2L=_I_
sb
1
2M=-
S~I
2N =-2-
1
SI2
in which X, y, Z are the uniaxial tensile strengths in the material's principal directions
XI, x2, 'and X3 , respectively, and S23, S31, SI2 are the shear strengths in the x2 - X3,
If, in addition, we assume that the composite is transversely isotropic, i.e., Y = Z, then
(6.15)
In Hill's original yield criterion, tensile and compressive yield properties were
assumed the same. For fiber composites, the tensile strength may be significantly
different from the compressive strength. The strength criterion given by (6.15) can be
used in the four quadrants formed by the normal stresses with the proper strengths values.
For example, in the third quadrant where 0"11 and 0"22 are both negative, X' and Y' are to
Jan 1993 AAE555 136
Tsai-Wu Criterion
A general stress failure criterion can be expressed in the fonn
i = 1, 2, ... , 6 (6.16)
Since strength is an inherent material property, the failure criterion must be invariant
with respect to the choice of coordinate systems. An explicit fonn which satisfies this
requirement was proposed by Tsai and Wu [6.3] as
11 + Faa
Fa IJ1J + Fjkaaak
IlJ + '" =1 (6.17)
where Fi, Fij, ... , etc., are material parameters. In (6.17) the summation convention over
repeated indices is used.
To ensure the said invariant property, each tenn in (6.17) must be a scalar. When
the stress components ai are viewed as a second order tensor, then according to the
quotient rule, the six independent components Fi can also be expanded to fonn a second
order tensor. By the same token, Fij can be expanded to fonn a 4th order tensor. The
tensorial properties of the expanded fonns Fi , Fij ... provide the coordinate
transfonnation laws for the material constants Fi, Fij ... , .
For orthotropic materials, it is more convenient to work with the coordinates which
are parallel to the material principal directions, x1> X2, and X3. With respect to this
particular coordinate system, extension and shear are uncoupled. By using the physical
argument that strength should not depend on the sign of the shear stresses, a4, as, and
a6, the following conditions are obtained from (6.17),
(6.18)
F14 = F 15 =F16 = 0
F24=F25=F26=0
in quadratic tenns. Similar conditions on the higher order coefficients can be obtained.
If we retain up the linear and quadratic terms in (6.17), the failure criterion is given
explicitly as
+ F550"~ + F660"g = 1
If a state of plane stress parallel to the xI - X2 plane exists, Le., 0"3 = 0"4 = 0"5 = 0,
then the failure criterion further reduces to
(6.21)
Thus, six material constants are to be determined in order to complete the failure
criterion. This task can be carried out in part by performing simple tension, compression,
and shear tests up to the failure point. Using these test results and (6.21), we obtain
1 1
Fl =x-+ X'
-1 -1 1
Fu = XX' , F22= yy' , F66=Z (6.22)
S
The remaining constant F12 can be determined only from a state of stress in which both
0"1 and 0"2 are present. Such a state can easily be produced by an off-axis test. Let O"u be
the ultimate stress for a unidirectional composite in plane stress under off-axis loading.
0"12 =0"6 =- .l
2
O"u sin 28 (6.23)
At first glance the expression given by (6.24) suggests that F12 depends on the angle 8.
Since F12 is presumably a material constant and should not vary with 8, its invariance is
expected to be achieved by the compensating effect of the 8-dependent O"u. It has been
found that the failure criterion is not sensitive to the value of F12 , and it is suggested to
set
Hence, the tensor polynomial criterion given by (6.21) reduces to the Hill-Tsai criterion
if X =-X' is assumed.
If, as in Hill's yield criterion, the tensile strengths are assumed to be the same as
then the linear terms in (6.20) drop out. The Hill's yield criterion can be deduced from
this resulting equation by funher requiring that the criterion is independent of hydrostatic
(6.26)
a ~ a ~ a ~
-acr1 wcren - + -(Fcrcr)
lJ 1 J acr acr2 -+-(P.crcr)
lJ 1 J acr acr3 -=0
lJ 1 J acr
(6.27)
p p p
The result is
Since this condition must hold for arbitrary stresses, it requires that
Fn +F12 +F13 =0
From these equations, the coefficients associated with the cross-product terms can be
expressed as
With conditions given by (6.25) and (6.29) it is quite straightforward to show that
(6.31)
and that Tsai-Wu criterion reduces to Hill's yield criterion for orthotropic materials.
Jan 1993 AAE555 140
There are two major modes of failure in unidirectional fiber composites, i.e., fiber
breakage and matrix cracking. The failure condition for each mode is governed by the
individual state of stress in the fiber and matrix, respectively. If Hill-Tsai or Tsai-Wu
failure criterion is used, these two modes cannot be distinguished. Usually, the largest
ratio among cru/X, cr22[Y, and cr12/S is used to identify the associated failure mode.
Consider failures of the fiber and matrix separately. Using a quadratic form in
stress, similar to the Hill's criterion, as the failure criterion for the fiber, we have
(6.32)
when crt are the stresses in the fiber, and Xf, Yf, and Sf are the major strengths in the
fiber. From the simple micromechanical model discussed in Section 3.3, it is seen that
cr~2 and cr{z are approximately the same as the corresponding stresses in the matrix.
Thus, cr~2 and crh cannot be raised to a significant level as compared with Y f and Sf
without causing matrix failure. In view of this, (6.32) can be simplified to
[~: r=1
(6.33)
Since crL is proportional to the composite stress cru, (6.33) suggests the suitable failure
or
0"11
-=1 for tension
X
Jan 1993 AAE555 141
and
all
-=1 for compression
X'
where X and X' are the longitudinal tensile and compressive strengths of the composite,
respectively.
A quadratic polynomial in stress similar to (6.32) can be assumed to predict failure
of the matrix. If we assume that aft is small (since the fiber takes the load in proportion
to its Young's modulus), then the failure criterion reduces to
(6.35)
This leads to the criterion for matrix failure in the unidirectional composite as
(6.36)
In this fonn, the transverse strength Y and the shear strength S of the composite may
Rotem [6.4].
this state of stress, the stress components referred to the x I - x2 system are obtained by
coordinate transformation as
. 29
0"22 = 0"0 sm (6.37)
0"0 2 . 2
11 = - (cos 9 -V12 sm 9)
EI
0"0 . Z Z
2Z = Ez (sm 9 - V21 cos 9) (6.38)
0"0 .
Y12=---sm9cos9
IZ
The uniaxial strength is the smallest value among the following three
0"0 = X/cosz 9
If the maximum strain criterion is used, then the uniaxial strength is determined by
It should be noted that then the compressive strengths X', y', and X'E' Y'E must be used
The prediction of the uniaxial strength of the panel using the Hill-Tsai criterion
can be obtained by substituting the stresses given by (6.37) into (6.15). The maximum
uniaxial strength is found to satisfy the following equation
-1= - c
1 o s4 S+ [1 1] cos2
--- 1,4 S
. 2 S+-sm
Ssm (6.41)
O"~ X2 S2 X2 y2
situe lamina strength in the laminate could be substantially higher than that measured in
unidirectional composites. In addition, failure modes such as delamination may occur in
laminates which often require three-dimensional stress analysis rather than the laminated
plate theory discussed in ChapterjY.'S-
The analysis of laminate strength presented in this section is basically a laminar
failure analysis in conjunction with laminate stiffness reduction due to laminar failure.
The procedure is quite straightforward and, to a certain extent, arbitrary. Two major
modes of failure are assumed, i.e., the fiber mode and the matrix mode. The former is
signified by fiber breakage; the latter, by matrix cracking. The mode of failure can be
identified easily if the maximum stress criterion and the maximum strain criterion are
used. For example, when the failure condition is satisfied by 0"11 <: X , then it is a fiber
failure mode; otherwise, it would be a matrix failure mode. When the Hill-Tsai or Tsai-
Wu criterion is employed, the mode of failure is identified by comparing the ratios
O"ll/X, 0"22/Y, and O"12/S and assuming that the stress that produces the highest ratio
For a given load, the stresses in each lamina are first calculated and examined with
a chosen laminar failure criterion. After a lamina is found to have failed, the laminar
properties are modified to reflect the mode of failure. There are several ways to modify
the laminar stiffnesses, each of which has a certain degree of arbitrariness. The amount of
stiffness reduction is still an open question. The following are several methods
commouly used in this type of laminate strength prediction.
reduced and the subsequent loading will follow the path determined by the new laminate
stiffness.
It is important to note that, after the first ply failure, the loading path is governed
by the current stiffness (see the dashed lines in Fig. 6.5). Thus, between two ply failure
events, the laminate stress-strain relation is linear, although the entire progressive failure
process may produce a highly nonlinear stress-strain relation. This greatly simplifies the
progressive failure analysis in a laminate.
In reality, one seldom observes distinct jumps in strain as illustrated in Fig. 6.5.
The post first-ply failure stress-strain curve appears more or less smooth. This does not
necessarily rule out the scenario given by the parallel spring model. Experimental results
have shown that matrix cracking in a lamina of a laminate is a stochastic process, and it
Jan 1993 AAE555 145
never occurs completely and at the same time as assumed by the strength model. 1n fact,
the matrix crack density increases gradually as the load increases. Figure 6.6 is a
schematic illustration of the matrix cracking process associated with a single failure
mode in a lamina. This stair case type response may appear as a smooth curve in the
global scale when the jumps are small and numerous.
The following example illustrates the procedure in laminate strength prediction
using the parallel spring model. The stiffnesses corresponding to a failure mode are
reduced to zero once it occurs.
20.13 0.44 0 ]
[Q]O" = 0.44 1.51 0 x 106 psi
[o 0 1.00
1.51 0.44 0 ]
[Q]900 = 0.44 20.13 0 x 106 psi
[o 0 1.00
The [A] is
3.50 1.01 0 ]
[A] = 1.01 3.50 0 x 105 Ib/in
[o 0 1.24
3.13 -0.91 0 ]
[Ar l = -0.91 3.13 0 x 10-6 in/lb
[
o 0 8.08
For a uniaxial load N x (Ny =N xy =0), the laminate strains are
The laminar stresses o"xx, O"yy, and O"xy can be computed using
Jan 1993 AAE555 147
::} = [Q]k
{O"xy k
{:~
"/iy
}
By using the coordinate transfonnation law for stress, we next obtain the stress
components O"n, 0"22, and 0"12 in each lamina. The results are
O"n} {62.6 }
{: : 0 = N 0.~3
x (b)
O"n}
0"22 = Nx
{22.82}
2.18 (c)
{0"12 4.04
45
O"n} {-1690}
{::~ 900 = N ~33
x (d)
To detennine the failure load for each lamina, we choose the Hill-Tsai criterion.
By substituting (b)-(d) into the failure criterion, we can obtain the critical N xcr for failing
each lamina. For the 90-ply, we have
2
N xcr {[
16.9
310000 ]
2
+ 9000[~]2 + 16.9x4.33
3100002
}= 1
from which
By using a similar procedure, the critical loads for failing other laminas can be
calculated. We obtain Nxcr =4950 Ib/in for the 0 ply and Nxcr =27111b/in for the
45 -plies. In this case, the critical load for the 90-ply turns out to be the smallest and,
thus, is the critical load for the laminate. This load is the first ply-failure load.
au 16.90 N xcr
0.11
X' = 310000
a22 4.33 Nxcr
y = 9000 =0.99
3.34 0.97 0 j.
[AJ =
[
0.97 3.48 0 x las lb/in
o 0 1.14
Though not necessary, we continue to load the laminate with Nx =2063 Ib/in. At
this load, the stresses in the laminas are obtained as
Jan 1993 AAE555 149
an} {134.55}
{ : : 00 = 0;1 x 103 psi
{:~~} {;~::}x
al2 45 0
= 103 psi
The corresponding failure load for each lamina is calculated using the Hill-Tsai
criterion as before. The results ate
(Nxcr)O" =47561b/in
(Nxcr )45 =2605 Ib/in
Thus, the second ply-failure-load is equal to 2605 lb/in. The mode of failure is easily
identified as matrix failure (in-plane sheat) in the 45 -plies.
After modifying the <'1j for the 45 -plies, a new [AJ matrix is obtained, and the
laminat stress analysis continues. At Nx =2605 lb/in, the stresses in the laminas ate
found as
Jan 1993 AAE555 150
195.16}
--O~61 x 103 psi
{
Examination of these laminar stresses with the failure criterion indicates that the
Nxcr =41291b/in
and the mode of failure is that of fiber failure. Once there is fiber failure, the laminate is
assumed to have failed. Thus, the laminate under consideration has the uniaxial tensile
To avoid the sudden jump in strain at ply failures, a model resembling the bilinear
hardening rule in classical plasticity can be formulated. It is assumed that the reduced
B
= (6.42)
~M. B D
where I'>. indicates the increment, and A, B, and D are the reduced laminate stiffnesses
Jan 1993 AAE555 151
after the i-th ply-failure. For i = 0, A, B, and D of the virgin laminate stiffnesses are
recovered. Thus, between the i-th and (i+ l)th ply-failures, the incremental laminate
deformation can be obtained from (6.42) for a given load increment. The total load is
obtained from adding the load increment to the critical load causing the ith ply-failure,
that is,
(6.43)
Similarly, the total strains and curvatures of the laminate are given by
(6.44)
in Fig. 6.7. For monotonic loading, the stress-strain curve produced by this model seems
unloading path. If classical plasticity is followed, then unloading should follow the
elastic unloading path AB (see Fig. 6.7) which is parallel to the initial slope of the
stress-strain curve. In most fiber composites with brittle matrices, however, the unloading
path more or less follows the path AO. Thus, the loading curve beyond the first
ply-failure cannot be regarded as the hardened yield surface in the sense of continuum
plasticity.
Jan 1993 AAE555 152
progressive matrix failures is usually insignificant. This suggests that, in such laminates,
the progressive stiffness reduction may be unnecessary and the laminate failure may be
taken to coincide with the failure of the load carrying ply. Figure 6.S shows the stress-
strain curves of a few fiber-dominated T300/520S graphite/epoxy composite laminates
under simple tension [6.5]. These results confirm our conjectures.
Consider Example 6.1 in which the 0-ply in the [45/0/90]s laminate is the
controlling ply. From the laminate analysis results given by equation (b) the sudden
failure model in conjunction with the Hill-Tsai failure criterion predicts the strength of
the [45/0/90]s laminate to be
Comparing this with Nxcr = 41291b/in predicted by the parallel spring model, it is
evident that the sudden failure model tends to predict a higher laminate strength.
This model was evaluated experimentally by Kim [6.5] using T300/520S
graphite/epoxy composite laminates in conjunction with Tsai-Wu failure criterion. The
unidirectional composite properties are
V12 =0.3
X = -X' = 200 ksi (13S0 MPa)
Y=6.96ksi(4SMPa) , Y'=-27.Sksi(-192MPa)
Figure 6.9 presents the comparison between the theoretical and experimental
strengths of [02/9]s laminates. Fairly good agreement is noted. For these laminates, no
delamination prior to laminate failure was observed.
Figure 6.10 shows similar results for [0/9/90]s and [0/90/9]s laminates. The
theoretical results are fairly good except for 9 S; 45 in [0/9/90]s laminates. In these
laminates with 9 S; 45 extensive delamination is present which is an additional failure
mode that is not accounted for by any of the three laminate failure models.
------------..
, I
, I
" I
w...--I-_Y
! !
Fig. 6.3 Coordinate system for off-axis composite panel
Jan 1993 AAE555 158
z- --z
x Laminate Failure
Ply Failure
CJ)
CJ)
W
~
f-
CJ)
Presumed Unloading
Loading Path
STRAIN
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
strain
/1
II
/ I
/ I
/ I
/ I
/ I
/ I
/ I
/ I
/
/" I
~ Ply Failure
/ I
/ I
/ I
/ I
/ I
/ I
/ I
V BI
strain
[0/90/:l:15j. [0/90/:l:30j.
1000 o [0/:l:IS/901. 1000 o [01:!:30/90j.
:.2i
U'>
U'>
l:!5oo
I-
U'>
o [0/90/:!:4S~ [0/90/:l:60j.
600 o [0/:!:45/90~ 600 Q [0/:!:60/9~.
.
~ 400 ~
.
Q.
400
en U'>
...a::
en ...a::
U'>
I- I-
U'>2oo en 200
O~----~~------~--
0.5 1.0
0L-----~~----_7~--
0.5 1.0
STRAIN,% STRAIN, %
T300/5208 GrIEp
1400 0 [Oz/8]s
1200
1000
~
:=: o o
800
I-
CI
Z o
W
a: 600 O-DEG
l-
I/)
400
200
00 15 30 45 60 75 90
8. DEGREE
Fig. 6.9 E)(perimental and calculated strengths using the sudden failure model.
The calculated laminate strength is the solid line associated with the
failure of the 0 -ply (Ref. 6.5).
131
200
0~--~15~-=:;Q~~4~~~-60~!~~75~!---J90
S.DEGREE
Fig. 6.10 Experimental and calculated strengths using the sudden failure model.
The calculated laminate strength is the solid line associated
with the failure of the 0 -ply (Ref. 6.5).