Académique Documents
Professionnel Documents
Culture Documents
COMPOSITE LAMINATES
3.1
In this chapter, we discuss some fundamental problems concerning fiberreinforced composite laminates; i.e. the classical part of the general theory of
composite materials.
The basic results existing in this field can be found, for instance, in the
monographies due to Ashton and Whitney [3.1], Jones [3.2], Christensen [3.3],
Tsai and Hahn [3.4], Cristescu [3.5], Whitney [3.6] and Gibson [3.7].
The fiber-reinforced composite laminates are made of fiber-reinforced laminae. The fibers considered here are long and continuous. A lamina is a plane
arrangement of unidirectional fibers strongly bounded in a matrix. In Figure 3.1
is shown a typical lamina together with its material symmetry axis, named also
principal material axes or directions.
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
112
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
113
we suppose that the fibers and the matrix are firmly bounded together. The same
assumption will be made concerning the laminae forming a laminate.
At the macro-mechanical level, the fiber-reinforced lamina will be assumed
to be an orthotropic linearly elastic material. The symmetry axis are parallel and
perpendicular to the fibers direction as shown in Figure 3.2. The most advantageous description of the stress-strain relation involves the (macro-mechanical or
effective or equivalent or overall ) technical or engineering constants of the lamina, considered as a homogeneous body. These constants are particulary helpful in
describing material behavior since they are determined by obvious and relatively
simple mechanical tests.
In the following, our attention will be focused on stress-strain relation for
orthotropic materials in a plane stress state, the most common condition satisfied
by a loaded composite lamina. The constitutive relations, initially formulated using
the material symmetry axes, will be expressed later by using coordinate systems
that are not aligned along the principal material directions. Such a change is
necessary in order to describe the global behavior of various laminates, composed
of laminae with various orientations of the reinforcing fibers.
Let us consider now a lamina in the 1-2 plane as shown in Figure 3.1. Here
the axes 1, 2, 3 are the principal material directions of the laminae, assumed to
be (macroscopically) orthotropic.
As usual, we say that the lamina is in a plane stress state relative to its
symmetry plane 1-2 if the components of the stress tensor satisfy the following
relations:
31 = 32 = 33 = 0.
(3.1.1)
Since the material is orthotropic, according to the constitutive equation
(2.2.70), from the above relation, it follows that the components of the strain
tensor satisfy the equations
31 = 32 = 0 , 33 = S13 11 + S23 22 ,
and, thus, the stress-strain relation (2.2.70) reduces to
1
1
S11 S12 0
2 = S12 S22 0 2 .
0
0
S66
6
6
(3.1.2)
We recall that in the above matrix form of the remaining constitutive equation,
we have used the Voigts convention; i.e.
1 = 11 , 2 = 22 , 6 = 212 , 1 = 11 , 2 = 22 , 6 = 12 .
Also, we note again that the axis 1, 2, 3 are the principal material directions
of the lamina, axis 1 being parallel to the fibers, axis 2 being perpendicular to the
fibers and situated in the plane of the lamina and axis 3 being perpendicular to
this plane.
114
The general relation (2.2.71) shows that the involved components S11 , S12 , S22
and S66 of the compliance matrix [S] can be expressed in terms of the technical
constants of the orthotropic lamina by the following equations:
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
S11 =
12
21
1
,
=
, S12 =
E2
E1
E1
1
1
.
, S66 =
S22 =
G12
E2
(3.1.3)
Since the matrix [S] is positive definite, the relation (3.1.2) can be inverted
to obtain the inverse stress-strain relations
1
Q11 Q12 0
1
1
2 = Q12 Q22 0
2 = [Q] 2 .
(3.1.4)
6
0
0
Q66
6
6
The quantities Q11 , Q12 , Q22 and Q66 are named reduced stiffnesses. They
have the following expressions:
Q11 =
S11
S12
S22
2
,
, Q66 = G12 , S = S11 S22 S12
, Q22 =
, Q12 =
S
S
S
(3.1.5)
E2
21 E1
12 E2
E1
, Q66 = G12 .
, Q22 =
=
, Q12 =
1 12 21
1 12 21
1 12 21
1 12 21
(3.1.6)
The reduced constitutive equations (3.1.4) represent the basis for the analysis
of the behavior of an individual lamina subjected to forces acting in its own plane.
For such special loading, the orthotropic lamina is indeed in a plane stress state.
We stress again that E1 is Youngs modulus in the fibers direction, E2 is
Youngs modulus in the direction perpendicular to the fibers and situated in the
lamina plane, 12 and 21 are Poissons ratios in the same plane, and G12 is the
shear modulus in the lamina plane.
We now present some numerical values of the involved material parameters for
laminae frequently used in applications. The values are taken from the monograph
[3.4] by Tsai and Hahn (see pp. 19 and 20). The material constants having physical
dimensions (such as E1 , E2 , G12 , S11 , ..., S66 , Q11 , ..., Q66 ) are expressed in GP a =
109 N m2 . Obviously, if E1 , E2 , 12 and G12 are known from experimental data
S11 , ..., S66 and Q11 , ..., Q66 can be calculated using the relation (3.1.3) and (3.1.6).
The data given in Tables 3.1, 3.2 and 3.3 show that for fiber-reinforced laminae, generally
E2 << E1 and G12 << E1
Q11 =
and
Q22 << Q11
and
115
Type
T300/5208
B(4)/5505
AS/3501
Material
Graphite/Epoxy
Boron/Epoxy
Graphite/Epoxy
E1
181
204
138
E2
10.3
18.5
8.96
12
0.28
0.23
0.30
G12
7.17
5.59
7.1
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
Type
T300/5208
B(4)/5505
AS/3501
S11
5.525
4.902
7.246
S22
97.09
54.05
111.6
S12
-1.547
-1.128
-2.174
S66
139.5
172.7
140.8
Type
T300/5208
B(4)/5505
AS/3501
Q11
181.8
205.0
138.8
Q22
10.34
18.58
9.013
Q12
2.897
4.275
2.704
Q66
7.17
5.75
7.1
We shall see in the Section 5, that the above large differences between the
magnitudes of the different rigidity moduli of a fiber-reinforced composite material have essential implications on the stability behavior of these bodies, having
obviously an internal structure.
We recall that the reduced constitutive relations (3.1.4) are expressed using
the stress and strain components corresponding to the material symmetry direction
of the lamina. These special directions often do not coincide with the coordinate
direction which are geometrically related to a given problem. Hence, we must
be able to express the reduced stress-strain relations using arbitrary systems of
coordinates x1 = x, x2 = y, x3 = z. For our needs, we assume that the principal
material direction 3 and the direction of the axis x3 = z coincide. Also, we
suppose that the planes x, y and 1, 2 coincide, and the principal directions 1, 2 are
obtained by rotating the axes x, y with an angle about the axis z, as shown in
Figure 3.3.
In the above mentioned case, the orthogonal matrix [qkr ], present in the
general lows (1.1.14) characterizing the connections between the components of a
tensor in the old and new axes have, according to the relations (1.1.8), the following
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
116
Figure 3.3: Positive rotation of principal material axes 1, 2 from arbitrary axes
x, y.
form:
cos
[qkr ] = sin
0
sin
cos
0
0
0
1
(3.1.7)
1
x
1
x
2 = [T ()] y , 2
= [T ()] y ,
(3.1.8)
6
xy
6 /2
xy
where the 3 3 square matrix [T ()]
cos2
[T ()] = sin2
sin cos
1
2 sin cos
2 sin cos .
cos2 sin2
x
1
x
y = [T ()]1 2 , y = [T ()]1
xy
6
xy
(3.1.9)
(3.1.8), we get
2
(3.1.10)
6 /2
cos2
sin2
2 sin cos
1
, (3.1.11)
[T ()] = [T ()] = sin2
cos2
2 sin cos
2
2
sin cos sin cos cos sin
117
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
x
1
y = [T ()] 2 ,
xy
6
Introducing the Reuters
1 0
[R] = 0 1
0 0
we have
1
1
2 = [R] 2
,
6
6 /2
x
1
y = [T ()] 2
.
xy
6 /2
matrices
0
0 ,
2
[R]
1
= 0
0
0
1
0
0
1/2
x
x
y = [R]1 y , since
xy
xy
(3.1.12)
(3.1.13)
xy = xy /2.
Now, returning to the primary stress-strain relation (3.1.4) and using the
above equations, we successively get
x
1
1
1
y = [T ()] 2 = [T ()] [Q] 2 = [T ()] [Q] [R] 2
xy
6
6
6 /2
x
x
1
= [T ()] [Q] [R] [T ()] y = [T ()] [Q] [R] [T ()] [R] y .
xy
xy
sin cos
= [T ()]T .
[R] [T ()] [R]
sin cos
cos2 sin2
(3.1.14)
Consequently, the needed stress-strain relation becomes
1
cos2
= sin2
2 sin cos
sin2
cos2
2 sin cos
Q11
x
x
y = Q() y = Q12
xy
xy
Q16
with
Q12
Q22
Q26
Q16
x
Q26 y ,
xy
Q66
T
Q() = [T ()] [Q] [T ()]
(3.1.15)
(3.1.16)
Finally, using the relations (3.1.11), (3.1.14) and the last equation, after long,
but elementary computations, we get for the components of the matrix Q() the
118
following expressions:
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
Q16 () = (Q11 Q12 2Q66 ) sin cos3 + (Q12 Q22 + 2Q66 ) sin3 cos ,
Q26 () = (Q11 Q12 2Q66 ) sin3 cos + (Q12 Q22 + 2Q66 ) sin cos3 ,
Q66 () = (Q11 + Q22 2Q12 2Q66 ) sin2 cos2 + Q66 (sin4 + cos4 ).
(3.1.17)
The matrix Q() is named the transformed reduced stiffness matrix, and its
components Q11 (), ..., Q66 () are the transformed reduced stiffness of the fiberreinforced lamina.
Note that the transformed reduced stiffness matrix has non-vanishing coefficients in all nine positions in contrast to the zeros existing in the primary
reduced stiffness matrix [Q]. However, there are still only four independent material constants since the lamina is orthotropic and it is in a plane stress state. The
stress-strain relation (3.1.15) shows that in general, with arbitrary x, y axis, there
is coupling between normal stresses and shear strains and between shear stresses
and normal strains. Thus, in the coordinates x, y, named in the following body
coordinates, even an orthotropic lamina behaves as would a general anisotropic.
That is the reason why such a lamina is called general orthotropic lamina, even if
it is actually orthotropic.
We observe now that, as an alternative to the foregoing procedure, we can
express in the body coordinates the strains in terms of stresses, by inverting the
relation (3.1.15) and by using the property
[Q]
S11
[S] = S12
0
S12
S22
0
0
0 .
S66
S 11
x
x
y = S() y = S 12
xy
xy
S 16
with
S 12
S 22
S 26
S 16
x
S 26 y ,
xy
S 66
T
S() = [T ()] [S] [T ()] .
(3.1.18)
, we obtain
(3.1.19)
(3.1.20)
119
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
S 16 () = (2S11 2S12 S66 ) sin cos3 (2S22 2S12 S66 ) sin3 cos ,
S 26 () = (2S11 2S12 S66 ) sin3 cos (2S22 2S12 S66 ) sin cos3 ,
1
1
yx
xy
1
,
, S 66 =
, S 22 =
=
, S 12 =
Gxy
Ey
Ey
Ex
Ex
xy,y
x,xy
y,xy
xy,x
, S 26 =
=
=
=
.
(3.1.22)
Gxy
Ex
Ey
Gxy
S 11 =
S 16
The mechanical significance of the apparent Young moduli Ex , Ey , the Poisson ratios xy , yx and the shear modulus Gxy is the same as in the case of an
orthotropic material. Obviously, their usual significance must be related to the
coordinate axes x and y.
As can be seen, we have also introduced new engineering coefficients xy,x ,
x,xy , xy,y and y,xy . These material constants are named by Lekhnitski coefficients of mutual influence and are defined as:
i,ij = coefficient of mutual influence of the first kind which characterizes
the stretching in the idirection caused by shear in the ij plane, that is i,ij =
ii /2ij , for ij = , all other stresses being zero, and i 6= j;
ij,i = coefficient of mutual influence of the second kind which characterizes
the shearing in the ij plane caused by a normal stress in the idirection, that
is ij,i = ij /i , for ii = , all other stresses being zero, and i 6= j.
Obviously, the apparent technical moduli depend on the angle by which the
principal mutual directions were rotated.
Using the relation (3.1.22) and the equations (3.1.21), the apparent moduli
can be expressed in terms of the primary engineering moduli of the lamina and
120
cos
+
=
E1
G12
E1
Ex
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
xy = Ex
1
Ey
1
Gxy
1
E1
=2
sin4 + cos4 E11 +
12
E1
sin4 +
2
E1
xy,x = Ex
xy,y = Ey
n
n
2
E2
1
G12
212
E1
412
E1
212
E1
1
G12
2
E1
212
E1
1
G12
sin2 cos2 +
1
G12
2
E1
1
E2
1
E2
1
G12
1
E2
sin2 cos2 +
sin cos3
sin3 cos
sin4 ,
o
sin2 cos2 ,
cos4 ,
1
G12
sin4 + cos4 ,
2
E2
212
E1
1
G12
2
E2
212
E1
1
G12
o
sin3 cos ,
o
sin cos3 .
(3.1.23)
An important consequence of the presence of the coefficients xy,x and xy,y
is that traction tests in non principal material directions result, not only in axial
extensions and lateral contractions, but also in shear deformations.
Following Jones (see [3.2], Chapter 2), values typical for a glass/epoxy composite (E1 = 3E2 , E2 = 8.27GP a, G12 = 0.5E2 , 12 = 0.25) are plotted in Figure
3.4. In Figure 3.4, Ex is divided by E2 and Gxy by G12 . This normalization permits
an easier analysis of the behavior of the apparent technical moduli as a function
of .
3,0
3,0
2,5
E1
=3
E2
G12
=0.5
E
EX
E2
EX
E2
2,0
xy
=0.25
1,5
2,0
12
xy
Gxy
G12
xy,x
2,5
1,5
G xy
G12
1,0
1,0
0,5
0,5
xy,x
0
15
30
45
60
75
0
90
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
121
be expected, since the laminae actually is orthotropic. Also it can be seen that
at intermediate angles, this coefficient of mutual influence achieves large values as
compared to the apparent Poisson ratio xy . Also, as the first two equations (3.1.23)
show, the transverse axial modulus Ey behaves essentially like the longitudinal one
Ex , with the exception that Ey is small for near 00 and large when is near 900 .
Similar comments can be made for yx and xy,y .
We observe that the behavior presented in the Figure 3.4 is not always typical for all composites, fiber-reinforced laminae. For the considered glass/epoxy
composite, the maximal value of Ex is just E1 . There exist cases where Ex can
actually exceed both E1 and E2 , or can be smaller than both E1 and E2 , for some
orthotropic laminae and some intermediate values of the angle (see P.3.8).
The reduced stiffnesses given in relation (3.1.17) are relatively complicated
functions of the four primary material characteristics E1 , E2 , 12 , G12 , as well as of
the angle of rotation . There exists an ingenious recasting of the stiffness transformations equations that enables a more clear understanding of the consequences of
rotating a lamina in a laminate (see Jones [3.2], Chapter 2). By using elementary
trigonometric identities, the transformed reduced stiffnesses can be expressed in
the following way:
(3.1.24)
where
U1 =
U2 =
U3 =
U4 =
U5 =
1
(3Q11 + 3Q22 + 2Q12 + 4Q66 ) ,
8
1
(Q11 Q22 ) ,
2
1
(Q11 + Q22 2Q12 4Q66 ) ,
8
1
(Q11 + Q22 + 6Q12 4Q66 ) ,
8
1
(Q11 + Q22 2Q12 + 4Q66 ) .
8
(3.1.25)
122
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
stiffness property. For example, the first equation (3.1.24) shows that the value
of Q11 is determined by a fixed constant, U1 , plus a quantity of low frequency
variation with , U2 cos 2, plus a third quantity, U3 cos 4, of higher frequency
variation with . Hence, U1 is an effective measure of lamina stiffness in a design
application, and it is not being affected by the orientation of the lamina.
3.2
In the Section 3.1, our approach was macromechanical or macroscopic considering the overall properties of a lamina. That is, a large enough piece of the
lamina has been considered as being (macroscopically) homogeneous. The fact
that the lamina is piece-wise homogeneous, being made of two constituent materials (the matrix and the fibers) was neglected. In this sense, we were able to
say that a boron/epoxy composite lamina with unidirectional boron fibers has
certain elasticities and stiffnesses which were experimentally determined. In this
homogenized situation, the following question cannot be asked and cannot be
answered: how can the (effective, equivalent, overall) stiffness of the composite be
varied by changing the amount of boron fibers in the lamina? Because there must
be some rationales (reasons) for selecting a particular stiffness for a particular design application, there must also exist a rationale for determining how to find the
best procedure to achieve that stiffness for a fiber-reinforced lamina. That is, how
can the percentage or the concentration or the volume fraction of the constituent
materials be varied so as to arrive at the desired (overall, macroscopic, equivalent)
stiffness?
There are two methods to answer the above questions which can be characterized as being either micromechanical or macromechanical. In micromechanics,
the composite material behavior is studied taking into account the interaction
of the constituent materials, that is the composite is analyzed as being a (piecewise) heterogeneous body. In macromechanics, the composite material behavior
is analyzed assuming the body as being homogeneous, and the effects of the actual non-homogeneities are taken into account only as averaged apparent, overall,
equivalent properties of the composite.
When using micromechanical methods, the properties of a lamina can be
mathematically derived on the basis of the properties of the constituent materials.
When using macromechanical methods, the properties of a lamina can be experimentally determined is the as mate state. That is, we can predict the lamina
properties by the procedures of micromechanics and we can measure the lamina
properties by mechanical experiments and use the properties obtained by one of
the above methods in a macroscopic analysis of the structure.
Knowledge of how to predict properties is essential in order to construct
composites that must have certain apparent, overall, equivalent or macroscopical
properties. Consequently, micromechanics is a natural approach beside macromechanics when viewed from a design rather than an analysis point of view. Obvi-
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
123
ously, the real design efficiency is evidenced when the micromechanical predictions
of the properties of the composite agree with the measured properties. Unfortunately, the micromechanical approach has inherent limitation. For example, a
perfect bound between fibers and matrix is a usual analysis restriction that might
not be satisfied by some composites. Thus, the micromechanical predictions must
be validated by careful experimental work.
Nowadays there exist two basic approaches in the micromechanics of composite materials: (i) mechanics (strength) of materials; (ii) elasticity.
The mechanics of materials approach contains simplifying assumptions concerning the hypothesized behavior of the mechanical system.
The elasticity approach is actually: (i) bounding principles; (ii) exact solutions; (iii) approximate solutions. Some of these approaches will be discussed in
detail, for some important cases, in Section 4 devoted to macroscopically homogeneous composites. We shall present bounds for the overall moduli, obtained by Hill,
Hashin and Shtrikman for macroscopically isotropic and transversally isotropic
composites. Exact solutions will also be presented due to Hill and one, derived by
Budiansky and Hill. Also we shall discuss briefly some results obtained by taking
into account various geometrical models of different composite materials.
The final objective of all micromechanical approaches is to determine the
overall (equivalent, macroscopic, effective) elastic moduli or stiffness of a composite material in terms of the elastic moduli and concentrations of the constituent
bij of
materials or phases. For example, the overall elastic moduli, designed by C
a fiber-reinforced composite lamina must be expressed in terms of the fibers and
matrix moduli and their concentrations
bij = C
bij (Em , m , Ef , f , cm , cf ) ,
C
where Em , m and Ef , f are Youngs moduli and Poissons ratios of the matrix
and of the fibers, respectively, and
cm = vm /v , cf = vf /v
represent the concentration or volume fractions of the matrix and of the fibers,
respectively, v, vm , vf being the volumes occupied by the lamina, the matrix and
the fibers, respectively.
As we shall see, the above problem generally cannot be solved without introducing unrealistic assumptions, used in the strengths of materials. The overall
properties obtained in this way, generally do not agree with the measured ones.
This is the main reason why the much powerful approach formulated on the base of
elasticity and on the theory of macroscopically homogeneous composite materials
must be involved. In this way, generally, we can derive lower and upper bounds
for the overall moduli, and if these bounds are close, the obtained results can be
used in the design.
According to the micromechanical approach used, we must impose some basic restrictions on the composite material that can be treated, using the methods
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
124
4L
,
L
where 1 is the axial strains for both the fibers and the matrix, according to the
basic Voigt type assumption. Then, the axial stresses m and f in the matrix and
in the fiber are
m = E m 1 , f = E f 1 .
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
125
The average axial stress 1 acts on the cross sectional area S, m acts on the cross
sectional area Sm of the matrix, and f acts on the cross sectional area Sf of the
fiber. Thus, the resultant axial force F on the RVE is
F = 1 S = m Sm + f Sf .
Using the obtained results and taking into account that according to the
b1 , we have
definition of the overall axial moduli E
we get
b1 1 ,
1 = E
b1 = Sm Em + Sf Ef .
E
S
S
But the concentrations or volume fractions cm = vm /v and cf = vf /v of the
matrix and of the fibers can be expressed as
cm =
Sf
Sm
,
, cf =
S
S
where v, vm and vf are the volumes occupied by the RVE, by the matrix and by
the fiber, respectively. In this way, finally we get
b1 = cm Em + cf Ef .
E
(3.2.1)
This expression for the overall (apparent, equivalent, macroscopical) Young modulus in direction of the fibers is known as the rule of mixture or as the Voigt type
estimate. This rule leads to a simple linear variation of the overall Young modulus
b1 from Em to Ef as the fibers concentration cf varies from 0 to 1.
E
We stress the fact that, according to its definition, the overall axial modulus
b1 connects the mean stress and the mean strain, evaluated on the RVE of the
E
126
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
composite material. In the elasticity approach of the problem, the overall moduli
will be introduced in the same way!
b2
Determination of E
b2 , in the direction transverse
We now consider the overall Young modulus E
to the fibers. In the mechanics of the material approach, the same transverse stress
2 is assumed to be applied to both the matrix and the fiber, as shown in Figure
3.6. Such kind of hypotheses was first made by Reuss in 1929.
2
2
.
+ cf
Ef
Em
b2 must satisfy
Recognizing that according to its definition the overall moduli E
the material law
1
,
2 =
b2 2
E
Copyright 2004 by Chapman & Hall/CRC
127
finally we get
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
1
1
1
.
+ cf
= cm
b
E
E
f
m
E2
(3.2.2)
for the stress state 1 6= 0 and all over stresses are zero.
According to the Figure 3.7, the transverse deformation 2 is
2 =
W
= b
12 1 .
W
We also have
W = Wm + Wf ,
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
128
Wm and Wf being the transverse displacements of the matrix and of the fiber,
respectively. Consequently
Wf
Wm
= b12 1 .
+
W
W
Following the same procedure as in analysis for the overall transverse Young
b2 , we assume that the transverse displacements Wm and Wf are
modulus E
approximately
Wm = W cm m 1 , Wf = W cf f 1 ,
(3.2.3)
The strength of materials rule leads to the mixture rule or to the Voigt type
estimate of the overall Poisson ratio b12 .
b 12
Determination of G
b 12 of a lamina is estimated in the meThe overall in-plane shear modulus G
chanics of materials approach by assuming a Reuss type hypothesis. It is supposed
that the same shear stress acts in the matrix and in the fiber. Denoting by m
and f , the shear strains in the matrix and fiber, respectively, we get
m =
1
1
.
, f =
f
m
The loading is shown in Figure 3.8 and the deformations on microscopic scale in
Figure 3.9. The total (mean) shear deformation is
=
,
W
129
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
m /2
f
MATRIX
1
.
b 12
G
1
1
1
+ cf .
= cm
b 12
f
m
G
Copyright 2004 by Chapman & Hall/CRC
(3.2.4)
130
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
Material
Ef
f
Em
m
cf
Methods
Experimental
Mixture rules
Material
Ef
f
Em
m
cf
Methods
Experimental
Mixture rules
b1
E
49.40
47.16
Glass/Epoxy
70
0.17
2.85
0.33
0.66
b2
b 12 b12
E
G
18.00
7.77
7.80
2.95
Boron/Epoxy
413
0.2
4.10
0.35
0.70
b 12
b2
b1
G
E
E
237.8
290
13.3
26.7
5.5
12.2
0.22
0.224
b1
E
Carbon/Epoxy
234
0.2
3.8
0.33
0.6
b2 G
b 12 b12
E
151
141.8
9.3
9.2
6.2
3.5
0.32
0.25
b12
0.245
Table 3.4: Experimental and calculated values of the overall elastic coefficients.
The strength of materials approach leads to a Reuss type estimation for the
b 12 .
overall in-plane shear modulus G
b2 , only for a fiber volume greater than 50 percent of the
As in the case of E
b 12 rise to above twice m even if f = 10m .
total volume does G
Using the data given by Barran and Loroze [3.9], we present in Table 3.4 the
mechanical characteristics of three fiber-reinforced composite materials, giving also
the fiber concentrations. We also give the overall elastic coefficients experimentally
determined and the values of the overall moduli calculated using the mixture rules
b1 and
obtained by the strength of materials approach. The axial Young modulus E
the transverse Poisson ratio b12 are evaluated taking into account the Voigt type
b2 and the inmixture rules (3.2.1), (3.2.2), and the transverse Young modulus E
b
plane transverse shear modulus G12 are obtained using the Reuss type mixture
rules (2.3.3), (3.2.4). The axial and transverse Young moduli, as well as the shear
modulus are expressed in GP a = 109 P a.
Examining the above data, we can see that the calculated values of the overall
b1 and those of the overall transverse Poisson ratio b12 are
axial Young modulus E
acceptable as first approximations. However, the calculated values of the overall
b2 and those of the overall transverse shear modulus
transverse Young modulus E
b
G12 are not acceptable, and cannot be used as a first approximation. Generally,
we can say that much more powerful methods are necessary to evaluate and/or
to bound the overall moduli as those obtained with the strength of materials
approach.
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
131
The above estimations are only examples of the type of mechanics of materials
approaches that can be used to obtain approximate expression for the overall
moduli. Other assumptions of mechanical behavior lead to different estimations
for the overall elastic moduli of the lamina.
The true significance of the Voigt and Reuss type estimates can be clarified
only by using the elasticity approach to get the overall stiffnesses. As we shall
see in Section 4.1, the Voigt and Reuss type estimations give universal-bounds for
the overall moduli. Generally, these estimates are the worst bounds that can be
derived by the elastic approach.
We end this Section with some words concerning the approach named netting
analysis (see Jones [3.2], Chapter 3, Section 3.3.1). The basic assumption in netting
analysis is that the fibers provide all the longitudinal stiffness and the matrix
provide all the transverse and shear stiffness as well as the Poisson effect. Even
on the base of the above results furnished by the mechanics of material, we can
see that the assumptions made by the netting analysis must generally be rejected.
In turn, the results due to the strength of materials approach must be carefully
analyzed in light of the elasticity approach. Some important results of this analysis
will be presented in Chapter 4.
3.3
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
132
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
133
theory. This observation concerns all of the internal contradiction existing between
the assumptions (7), (8) and (9): though 13 , 23 and 33 are not vanishing, we
use the reduced stress-strain relations corresponding to vanishing 13 , 23 and 33 .
Obviously, the seriousness and consequences of these inconveniences can be established only by studying the implication of the theory based on the assumptions
(1)-(9). For this purpose, we must first develop the classical lamination theory,
using the supposed hypothesis.
From assumption (6), we obtain
33 (x1 , x2 , x3 ) =
u3
(x1 , x2 , x3 ) = 0.
x3
u3
u1
(x1 , x2 , x3 ) = 0,
(x1 , x2 , x3 ) +
x1
x3
223 (x1 , x2 , x3 ) =
u3
u2
(x1 , x2 , x3 ) = 0.
(x1 , x2 , x3 ) +
x2
x3
U3 (x1 , x2 )
U3 (x1 , x2 )
.
, u2 = U2 (x1 , x2 ) x3
x2
x1
(3.3.2)
In the above relations, U3 (x1 , x2 ) is the normal displacement of the middle surface,
and U1 (x1 , x2 ), U2 (x1 , x2 ) characterize the tangential displacement of the same
surface.
From (3.3.2), we obtain the following expressions for the non-vanishing strain
components 11 , 22 , 33 :
= e + x3 k , , = 1, 2,
where
e = e (x1 , x2 ) =
1
U
1 U
) = (U, + U, )
+
(
2
x
2 x
(3.3.3)
(3.3.4)
U3
= U3,
x x
(3.3.5)
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
134
x
11
x
11
[] = y = 22 , [] = y = 22 ,
xy
12
xy
212
e11
[e] = e22 ,
2e12
k11
[k] = k22
2k12
[] = [e] + x3 [k] .
(3.3.6)
(3.3.7)
xy k
Q16 Q26 Q66 k
xy
k
(3.3.9)
135
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
Since according to (3.3.7), e and k depend on x1, x2 only, the last equation becomes
[]k = Q k [e] + x3 Q k [k] for zk1 < x3 = z < zk and k = 1, .., N. (3.3.10)
The last equation expresses the plane stress 11 , 22 , 12 in the k-th layer, in
terms of the laminate middle surface strains and curvatures.
Expanded, the equation (3.3.10) becomes
k11
Q11 Q12 Q16
e11
Q11 Q12 Q16
11
22 = Q12 Q22 Q26 e22 +x3 Q12 Q22 Q26 k22
12 k
Q16 Q26 Q66 k 2k12
Q16 Q26 Q66 k 2e12
(3.3.11)
for zk1 < x3 = z < zk and k = 1, .., N.
In the above equations, k denotes the k-th lamina, ( )k , , = 1, 2 are the
stress in the k-th lamina, (Qij )k , i, j = 1, 2, 6 are the transformed reduced stiffness
of the k-th lamina, zk1 and zk are the distances from the middle surface to the
inner and to the outer surfaces of the k-th lamina, respectively, and N is the total
number of the laminae.
We recall that the reduced stiffness Qij , i, j = 1, 2, 6 depend on , the angle
made by the fibers with axis Ox1 , and we have
(3.3.12)
k representing the angle made by the fibers in the k-th lamina and the body axis
x1 .
Since (Qij )k can be different for each lamina of the laminate, the stress
variation through the thickness is not necessarily linear, even though the strain
variation is linear, as can be seen by examining equation (3.3.3).
In the laminated plate analysis, it is convenient to use the forces N and
the moments M per unit length, defined by the following relations:
N = N =
h
2
h
2
dx3 , M = M =
h
2
h
2
x3 dx3 , , = 1, 2. (3.3.13)
Let us observe that usually N11 , N12 = N21 , N22 are denoted by Nxx , Nxy =
Nyx and Nyy , respectively, and also M11 , M12 = M21 , M22 are denoted by Mxx , Mxy
= Myx , Myy , respectively.
According the relation (3.3.13)1 N11 , N12 , N22 are forces per unit length of
the cross-section. The mechanical meaning of these force resultants are shown in
the Figure 3.12.
Similarly, equation (3.3.13)2 shows that M11 , M12 , M22 are moments per unit
length of the cross-section. In Figure 3.13 is shown the mechanical meaning of
these moment resultants.
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
136
The relations (3.3.13) show that these force and moments resultants do not
depend on x3 , but are functions of x1 and x2 , the in the plane coordinates of the
laminate middle surface.
In more detail, the defining equations (3.3.13) can be written as
N Z zk
N11
11
X
22 dz;
[N ] = N22 =
z
k1
N12
12 k
k=1
Z
N
M11
11
zk
X
[M ] = M22 =
z 22 dz.
z
k1
M12
12 k
k=1
(3.3.14)
The integrations indicated in these equations can be rearranged to take advantage of the fact that the stiffness matrix for a lamina is constant within each
lamina. Thus, substituting the stress-strain relations (3.3.11) and taking into ac-
137
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
[M ] =
zk1
k=1
N
X
[Q]k
k=1
zk
zdz
zk1
[e] +
zk1
k=1
N
X
k=1
[Q]k
zk
z dz
zk1
N11
A11 A12 A16
e11
B11
[N ] = N22 = A21 A22 A26 e22 + B21
N12
A61 A62 A66
2e12
B61
M11
B11
[M ] = M22 = B21
M12
B61
B12
B22
B62
B12
B22
B62
B16
e11
D11
B26 e22 + D21
B66
2e12
D61
D12
D22
D62
[k].
B16
k11
B26 k22 ,
B66
2k12
D16
k11
D26 k22 ,
D66
2k12
(3.3.15)
Aij = Aji =
(3.3.16)
D12
D22
D26
D16
D26
D66
(3.3.17)
(3.3.18)
N11
A11 A12 A16 B11 B12 B16
e11
N22 A21 A22 A26 B21 B22 B26 e22
138
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
This equation describing the global behavior of the laminate, can be expressed
in the concentrated form
N
A B
e
e
=
= [E]
.
(3.3.20)
M
B D
k
k
The 6 6 symmetric matrix [E] is the global laminate stiffness matrix.
The coefficients Aij are called extensional stiffnesses, the coefficients Bij describe the coupling stiffness, and the coefficients Dij are called bending stiffnesses.
The presence of the coefficients Bij implies coupling between bending and extension
of a laminate. That is, it is impossible to pull on a laminate that has non-vanishing
Bij terms, without at the same time bending and/or twisting the laminate. Thus,
an extensional force results not only in extensional deformation of the middle surface, but also in twisting and/or bending of the laminate. Also, such a laminate
cannot be subjected to a moment without at the same time being subjected to
an extension of the middle surface. The experiments made with laminates confirm
these theoretical predictions. In spite of this fact, in the stability analysis of laminates, this coupling is generally neglected and we shall discuss this question in
Chapter 7.
It is easy to see that the matrix equation (3.3.19) can be written in the
following tensorial or component form, very useful in many problems
N = A e + B k ,
M = B e + D k , , , , = 1, 2.
(3.3.21)
(3.3.22)
(3.3.23)
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
139
(3.3.24)
(3.3.25)
(3.3.26)
Equation (3.3.25) and (3.3.26) give a partially inverted form of the equation
(3.3.20)
e
A B
N
=
,
(3.3.27)
M
C D
k
with
[A ] = [A]
, [B ] = [A]
(3.3.29)
(3.3.30)
Finally, (3.3.29) and (3.3.30) lead to the following inverted global constitutive
equation.
0
e
A B0
N
N
1
=
=
[E]
,
(3.3.31)
k
C 0 D0
M
M
where
[A0 ] = [A ] [B ][D ]1 [C ],
[B 0 ] = [B ][D ]1 ,
[C 0 ] = [D ]1 [C ] = [B 0 ]| = [B 0 ],
[D0 ] = [D ]1 .
(3.3.32)
The last results show that the 6 6 global compliance matrix [E]1 is symmetric. This is an obvious result, since if the global stiffness matrix [E] being
symmetric, its inverse, if it exists, must also be symmetric.
140
3.4
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
This Section considers special classes of laminates for which the stiffnesses
can easily be calculated. The special classes will be presented in increasing order
of complexity.
1. Single - layered configurations
For a single isotropic layer with material properties E, and thickness h,
equations (3.3.16) give
A11 = A22 =
1
Eh
A , Bij = 0 ,
A, A12 = A , A16 = A26 = 0 , A66 =
2
2
1
1
Eh3
D.
D , D12 = D , D16 = D26 = 0 , D66 =
2
2
12(1 )
(3.4.1)
In order to obtain the above relations, we must use the equations (3.1.6),
supposing an isotropic material.
From (3.4.1), we can conclude that the resultant forces depend only on the
in-plane strains of the laminate middle surface, and the resultant moments depend
only on the curvatures of the middle surface. There is no coupling. The constitutive
equations become
A A
0
e11
N11
e22 ,
N22 = A A
0
2e12
N12
0
0 (1 ) A2
(3.4.2)
D D
0
k11
M11
k22 .
M22 = D D
0
D
2k12
M12
0
0 (1 ) 2
D11 = D22 =
In particular, we have
h2
A.
(3.4.3)
12
For a simple specially orthotropic layer of thickness h the lamina stiffnesses
are given by equation (3.1.6). Hence, according to (3.3.10), the laminate stiffnesses
are
D=
A11 = hQ11 , A12 = hQ12 , A22 = hQ22 , A16 = A26 = 0 , A66 = hQ66 ,
Bij = 0,
D11 =
h3
h3
h3
h3
Q66 .
Q22 , D16 = D26 = 0 , D66 =
Q12 , D22 =
Q11 , D12 =
12
12
12
12
(3.4.4)
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
141
Again, the resultant forces depend only on the in-plane strains, and the resultant moments depend only on the curvatures. There is no coupling. The constitutive equation becomes
e11
N11
A11 A12
0
N22 = A12 A22
0 e22 ,
0
0
A66
2e12
N12
(3.4.5)
M11
D11 D12
0
k11
M22 = D12 D22
0 k22 .
0
0
D66
M12
2k12
2. Symmetric laminate
For laminates that are symmetric in both geometry and material properties
about the middle surface, the general stiffness equations (3.3.16) simplify considerably. Because of the symmetry of the transformed stiffnesses (Qij )k and of the
thicknesses hk , it can be shown that all coupling stiffness Bij of the laminate are
zero. There is no coupling. Obviously such laminates are much easier to analyze
than laminates with coupling. Consequently, symmetric laminates are commonly
used unless special circumstances require an unsymmetrical laminate possessing
the coupling property.
The constitutive equations for a symmetric laminate are
N11
A11 A12 A16
e11
N22 = A12 A22 A26 e22 ,
N12
A16 A26 A66
2e12
(3.4.6)
M11
D11 D12 D16
k11
M22 = D12 D22 D26 k22 .
M12
D16 D26 D66
2k12
Ek
, (Q16 )k = (Q26 )k = 0 ,
1 k2
Ek
k Ek
.
, (Q66 )k =
(Q12 )k =
2(1 + k2 )
1 k2
(Q11 )k = (Q22 )k =
(3.4.7)
142
In these equations, Ek and k are the Youngs modulus and the Poissons
ratio for the k-th lamina.
It is easy to see that
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
N11
A11 A12
0
e11
N22 = A12 A11
0 e22 ,
0
0
A66
N12
2e12
D11
M11
M22 = D12
0
M12
D12
D11
0
k11
0
0 k22 .
2k12
D66
(3.4.8)
(3.4.9)
k
E2k
21
E1k
E1k
,
, (Q22 )k =
, (Q12 )k =
k k
k
k
k
k
1 12
1 12 21
1 12 21
21
(3.4.10)
k
k
E1k , E2k , 12
, 21
and Gk12 being the engineering material constants of the k-th specially orthotropic lamina.
Because (Q16 )k and (Q26 )k are zero, it is easy to see that A16 , A26 , D16 and
D26 vanish; i.e.
A16 = A26 = 0 , D16 = D26 = 0.
(3.4.11)
Also, because of symmetry, the coupling stiffnesses Bij are all zero; i.e.
Bij = 0.
(3.4.12)
Hence, the constitutive equation for the laminate takes the form
N11
A11 A12
0
e11
N22 = A12 A22
0 e22 ,
N12
0
0
A66
2e12
M11
D11
M22 = D12
0
M12
D12
D22
0
k11
0
0 k22 .
D66
2k12
(3.4.13)
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
143
Taking into account the above equations, this type of laminate could be called
specially orthotropic laminate in analogy to a special orthotopic lamina.
A regular symmetric cross-ply laminate represents a very common special case
of symmetric laminates with multiple specially orthotropic laminae (layers). The
regular symmetric cross-ply laminate occurs when the laminae are all of the same
thickness and material properties, and their major principal material direction
(that is, the fiber directions) alternate at 00 or 900 with respect to the laminate
(body) axes, for examples (00 /900 /00 ) as in Figure 3.14.
N
X
k=1
Qij
(zk zk1 ) ,
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
144
trigonometrical functions are involved with even powers and Q11 , Q12 , Q22 , Q66
are positive. Thus, A11 , A12 , A22 and A66 are all positive since the thicknesses
of the laminae are obviously positive. However, (Q16 )k and (Q26 )k are zero for
lamina orientation of 00 or 900 to the laminate axes. Thus, A16 and A26 are zero
for laminates made of orthotropic laminae oriented at either 00 or 900 to the
laminate axes.
Next, we consider the coupling stiffnesses
N
Bij =
1X
2
,
(Qij )k zk2 zk1
2
k=1
Dij =
1X
3
,
(Qij )k zk3 zk1
3
k=1
145
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
be largest when N = 3 (the lowest N for which this class of laminates exists) and
decrease in proportion to 1/N as N increases
A16 =
N
X
k=1
1X
3
,
(Q16 )k zk3 zk1
3
k=1
obviously, A16 and D16 are sums of terms of alternating signs since
(Q16 )+ = (Q16 ) .
(3.4.14)
(3.4.15)
(3.4.16)
since
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
146
where the above symbols represent the corresponding transformed reduced stiffnesses in the laminae with the orientation and of the fibers.
For general antisymmetric laminates, all coupling stiffnesses are non-vanishing.
Hence, the constitutive equation (3.3.15) become
N11
A11 A12
0
e11
B11 B12 B16
k11
N22 = A12 A22
0 e22 + B12 B22 B26 k22 ,
N12
0
0
A66
2e12
B16 B26 B66
2k12
M11
B11
M22 = B12
M12
B16
k11
0
0 k22 .
D66
2k12
(3.4.17)
We discuss now two important subclasses of antisymmetric laminates.
An antisymmetric cross-ply laminate consists of an even number of orthotropic
laminae laid down on each other with principal material directions alternating at
00 and 900 to the laminate axes, as in the example given in Figure 3.16.
B12
B22
B26
B16
e11
D11
B26 e22 + D12
0
B66
2e12
D12
D22
0
k11
e11
B11 0
0
N11
A11 A12
0
N22 = A12 A22
B11 0 k22 ,
0 e22 + 0
0
0
0
0
0
A66
2k12
2e12
N12
M11
B11
M22 = 0
M12
0
0
B11
0
0
e11
D11
0 e22 + D12
0
2e12
0
D12
D22
0
0
k11
0 k22 .
D66
2k12
(3.4.19)
147
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
A regular antisymmetric cross-ply laminate is a special case, which has laminae of equal thickness. This type of laminate is common because of simplicity of
fabrication.
It can be shown that the coupling stiffness B11 of an antisymmetric cross-ply
laminate approaches zero as the number of layers increases, for a fixed laminate
thickness.
An antisymmetric angle-ply laminate has laminae oriented at + degrees to
the laminate coordinate axes on one side of the middle surface and the corresponding equal thickness laminae on the other side is oriented at degrees. A simple
example is given in Figure 3.17.
y = x2
x = x1
(3.4.20)
N11
A11 A12
0
e11
0
0
B16
k11
N22 = A12 A22
0 e22 + 0
0
B26 k22 ,
0
0
A66
N12
2e12
B16 B26 0
2k12
M11
0
M22 = 0
M12
B16
B16
e11
D11
B26 e22 + D12
0
2e12
0
0
k11
0 k22 .
D66
2k12
(3.4.21)
It can be shown for a fixed laminate thickness that the coupling stiffnesses
B16 and B26 tend towards zero as the number of layers in the laminate increases.
Summing up the presented results, concerning some special classes of fiberreinforced composite laminates, we can say the following:
0
0
B26
D12
D22
0
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
148
(1) Simple layer laminates with a reference surface, at the middle surface do
not exhibit coupling between extension and bending.
(2) Multilayered laminates, in general, develop coupling between extension
and bending.
(3) The coupling is influenced by the geometrical as well as by the material
properties of the laminae.
(4) There exist combinations of the material properties and geometrical characteristics for which there is no coupling between extension and bending.
(5) The elastic symmetry of the laminae (transversally isotropy, orthotropy)
is not necessarily conserved in the laminate.
(6) Moreover, the symmetries can be increased, decreased or remain the same.
(7) The symmetry properties of the three stiffness matrices [A], [B] and [D]
need not be the same, and are generally different.
We stress the fact that the fundamental concept of coupling between extension and bending must be well understood, because there exists many applications
of the composite laminates where the neglect of the coupling can be catastrophic.
Considering coupling is the key to the correct analysis of eccentrically stiffened
plates.
The procedure to describe a laminate by use of individual layer thicknesses,
principal material property orientations, and overall sequence can be quite involved. However, all pertinent parameters can be represented in a simple way if
one uses the following stacking sequence terminology.
For regular (equal thickness layers) laminates, a listing of layers and their orientation suffices, for example, [00 /900 /450 ]. Let us observe that only the principal
material direction orientations need to be given.
For irregular (when the layers do not have the same thickness) laminates, a
notation of layers thickness must be added to the previous notation, for example,
[00 /h1 /900 /h2 /450 /h3 ].
For symmetric laminates, the simplest representation of the laminate [0 0 /900 /
0
45 /450 /900 /00 ] is [00 /900 /450 ] sym.
We shall discuss now in greater details the relation by which the cross-ply
laminate stiffnesses can be expressed.
We recall that a cross-ply laminate has N unidirectionally reinforced orthotropic layers with the principal material directions alternatingly oriented at
00 and 900 with respect to the laminate coordinate axes. The fiber direction of
the odd -numbered layers is the x1 direction of the laminate. The fiber direction
of the even-numbered layers is the x2 direction of the laminate. We assume that
all odd-numbered layers have the same thickness, all even-numbered layers have
also equal thicknesses but the odd and even numbered layers do not necessarily
have the same thickness. For this special but important case for applications, two
geometrical parameters are important:
(1) N, the total number of layers,
and
149
(2) m, the ratio of the total thicknesses of odd numbered layers to the total
thickness of the even-numbered layers, called cross-ply ratio. Hence,
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
m=
hk
k=odd
hk
(3.4.22)
k=even
For instance, for a five-layered cross-ply laminate, which has a lamination or stacking sequence [00 /h1 /900 /2h1 /00 /2h1 /900 /2h1 /00 /h1 ], we get
m=
h1 + h 1 + h 1
= 3/4.
2h1 + 2h1
Let us observe that the cross-ply ratio m has a specific meaning only when
the laminae (layers) have alternating 00 and 900 orientation!
Tsai (see Jones [3.2], Chapter 4) has shown that the laminate stiffnesses
Aij , Bij and Dij for cross-ply laminates with odd or even numbers of layers, can
be expressed in terms of m and N . In addition, Tsai uses the stiffness ratio
f = Q22 /Q11 = E1 /E2 6 1
(3.4.23)
1 + mf
1 + mf
m+f
A11 ,
hQ11 =
hQ11 , A12 = hQ12 , A22 =
m+f
1+m
1+m
A16 = A26 = 0 , A66 = hQ66 , Bij = 0,
h3
h2
1+m
(f 1)p + 1 3
Q12 ,
{(f 1)p + 1} A11 , D12 =
h Q11 =
D11 =
12
12
m+f
12
h2
1+m
(1 f )p + f 3
{(1 f )p + f } A11 ,
h Q11 =
D22 =
12
m+f
12
h3
Q66 ,
(3.4.24)
D16 = D26 = 0 , D66 =
12
A11 =
where
p=
150
1 + mf
1 + mf
m+f
A11 ,
hQ11 =
hQ11 , A12 = hQ12 , A22 =
m+f
1+m
1+m
A16 = A26 = 0, A66 = hQ66 ,
m(f 1)
m(f 1) 2
hA11 , B22 = B11 ,
h Q11 =
B11 =
N (1 + m)(f + m)
N (1 + m)2
B12 = B16 = B26 = B66 = 0,
h3
h2
1+m
(f 1)r + 1 3
Q12 ,
{(f 1)r + 1} A11 , D12 =
h Q11 =
D11 =
12
12
m+f
12
h2
1+m
(1 f )r + f 3
{(1 f )r + f } A11 ,
h Q11 =
D22 =
12
m+f
12
3
h
Q66 ,
(3.4.25)
D16 = D26 = 0 , D66 =
12
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
A11 =
where
r=
8m(m 1)
1
.
+
1 + m N 2 (1 + m)3
151
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
A11 = A22 =
and
B22 > 0.
(3.4.27)
In the following, we shall indicate the way in which Tsais formulas (3.4.26)
for a regular antisymmetric cross-ply laminate can be obtained. Such a laminate
is shown in the Figure 3.18.
h
h
for k = 0, 1, ..., N.
(3.4.28)
+k
N
2
According to the definition of a regular antisymmetric cross-ply laminate, we have
also
k = 00 if k is odd and k = 900 if k is even .
(3.4.29)
zk =
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
152
Since all layers have the same thickness h/N , we obviously have
zk zk1 = h/N for any k.
(3.4.30)
Hence,
N
h X
(Q11 )k .
N
A11 =
k=1
A11 =
h
(Q11 + Q22 ).
2
1+f
hQ11 .
2
Hence, we get the first Tsai formula (3.4.26), giving the extensional stiffness A 11 .
In the same way all Tsai formulas can be deduced concerning the extensional
stiffnesses Aij .
Let us analyze now the coupling stiffnesses. According to (3.3.16)2 , we have
(x3 = z)
N
1X
2
).
(Qij )k (zk2 zk1
Bij =
2
k=1
h2
{(1 + N ) + 2k}.
N2
153
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
B11 =
k=1
k=1
X
X
h2
k(Q11 )k }.
(Q11 )k + 2
{(1 + N )
2N
We recall again that (Q11 )k = Q11 if k is odd, and (Q11 )k = Q22 if k is even.
Thus we obtain
B11 =
X
X
N (1 + N )
h2
(Q11 + Q22 ) + 2(Q11
k + Q22
k)}.
{
2
2N
k=odd
k=even
l=
l=0
n(n + 1)
.
2
(3.4.31)
k=
N (N + 2)
N2 X
.
,
k=
4
4
(3.4.32)
k=even
k=odd
Introducing these values in the last expression of B11 , after some algebra, it results
B11 =
h2
(Q11 + Q22 ).
4N
f 1 2
h Q11 .
4N
Therefore, we have obtained the first Tsai formula (3.4.26), giving the coupling
stiffness B11 .
In the same way, all Tsai formulas can be deduced (3.4.26) giving the coupling
stiffness Bij.
According to (3.3.16)3 , for the bending stiffness D11 , we have (x3 = z)
N
D11 =
1X
3
).
(Q11 )k (zk3 zk1
3
k=1
h3
N3
3N 2 + 6N + 4
3(N + 1)k + 3k 2 .
4
154
Also, we know that (Q11 )k takes N/2 times the value Q11 , if k is odd, and
N/2-times the value Q22 , if k is even. Hence, the equation giving D11 becomes
D11 =
X
X
h3 3N 2 + 6N + 4 N
(Q
+
Q
)
3(N
+
1)(Q
k
+
Q
k)
{
11
22
11
22
2
N
3N 3
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
k=odd
+3(Q11
k 2 + Q22
k=odd
k=even
k 2 )}.
k=even
We use again the relation (3.4.31) and the well known formula
n
X
l2 =
l=o
n(n + 1)(2n + 1)
.
6
(3.4.33)
k=odd
k2 =
N (N + 1)(N + 2)
N (N 2 1) X 2
.
,
k =
6
6
(3.4.34)
k=even
Introducing (3.4.32) and (3.4.34) in the last expression of D11 , elementary computations leads to
D11 =
1+f 3
h2
h Q11 ,
(Q11 + Q22 ) =
24
24
3.5
In the theory of thin plate, the Cauchys equilibrium equations are replaced
by global equilibrium conditions satisfied by the force resultants N , moment
resultants M and by the resultant shear forces Q defined by the relations
Q =
h/2
3 dx3 , = 1, 2.
(3.5.1)
h/2
To get these equations, we start with the Cauchys (local) equilibrium conditions, assuming absence of body forces
, + 3,3 = 0 , 3, + 33,3 = 0, , = 1, 2.
(3.5.2)
155
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
(3.5.3)
h/2
x3 3,3 dx3 = 0.
h/2
We have
x3 3,3 = (x3 3,3 ) 3 .
Consequently,
Z
h/2
x3 3,3 dx3 =
h/2
h
h
3 (h/2) 3 (h/2)
2
2
h/2
3 dx3 .
h/2
We recall that 3 (h/2) = 0 and we will use the definition (3.5.1) of the resultant
shear forces Q .
156
Thus, we finally get the equilibrium condition which must be satisfied by the
shear forces resultants Q and moment resultants M , in the plane domain D
occupied by the middle surface of the laminate
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
M, Q = 0 in D, , = 1, 2.
(3.5.4)
From (3.5.3)2 and (3.5.4), we can now see that the resultant moments M
must satisfy the following second order (global) equilibrium equation
M, + q = 0 in D , = 1, 2.
(3.5.5)
1
(U, + U, ), k = U3, ;
2
(3.5.6)
Constitutive equations:
N = A e + B k , M = B e + D k ;
(3.5.7)
Equilibrium equations:
N, = 0 , Q, + q = 0 , M, Q = 0 , M, + q = 0.
(3.5.8)
We recall also that according to (3.3.23), the constitutive coefficients have the
following symmetry properties :
A = A = A = A ,
B = B = B = B ,
(3.5.9)
D = D = D = D .
Obviously the Greek indices take the values 1, 2 and the Einsteins summation
convention applies.
157
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
Figure 3.19: The plane domain occupied by the middle surface of the laminate.
Denoting by n and the (plane) components of n and , we get
n1 = 2 , n2 = 1 .
(3.5.10)
(3.5.11)
(3.5.12)
In the same way, the normal and tangential components Un and U of the
in-plane displacement on the boundary D can be obtained using the equations
Un = n U , U = U .
(3.5.13)
Also, for a later use, we introduce on the boundary D the normal and
tangential derivatives U3,n and U3, of the normal displacement U3 . According to
158
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
the general definition of the directional derivative of a scalar field, given in the
Section 1.2, we have
U3,n = n U3, , U3, = U3, .
(3.5.14)
Using these equations, we can express the partial derivatives U3, through
the normal and tangential derivatives U3,n , U3, . From (3.5.10) and (3.5.14), we
get
U3,1 = 2 U3,n + 1 U3, , U3,2 = 1 U3,n + 2 U3, .
(3.5.15)
(3.5.17)
(3.5.18)
In order to transform the right hand side term of this equation, we now use
the equilibrium condition (3.5.8)3 . We get
Z
Z
Q U3, da =
U3, M, da.
(3.5.20)
D
Using again the Greens theorem, the geometrical equations (3.5.6)2 and the
Cauchys type relations (3.5.11), from (3.5.20) we obtain
Z
Z
Z
Q U3, da =
U3, Mn ds +
M k da.
(3.5.21)
D
(3.5.22)
159
where
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
Qn = Q n
(3.5.23)
=
=
=
(3.5.24)
Transforming adequately the first term in the right hand side of (3.5.24), we obtain
U3, Mn = (U3 Mn ), U3 Mn, + U3,n Mnn .
The tangential derivatives involved in this equation have the following expressions:
(U3 Mn ), = (U3 Mn ),1 1 + (U3 Mn ),2 2 ,
Mn, = Mn,1 1 + Mn,2 2 .
We assume that U3 and Mn are uniform function on D. Hence,
Z
(U3 Mn ), ds = 0.
(3.5.25)
since D is a closed curve. Thus from (3.5.24) and (3.5.25) we finally get
Z
Z
U3, Mn ds =
(U3 Mn, + U3,n Mnn )ds.
D
(3.5.26)
Now, equations (3.5.18) and (3.5.26) lead to the following work relation:
Z
Z
{Nnn Un + Nn U + (Qn + Mn, )U3 U3,n Mnn }ds +
qU3 da
D
D
Z
= 2
wda,
(3.5.27)
D
160
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
2w = N e + M k .
(3.5.28)
1
1
e A e + e B k + k D k .
2
2
(3.5.29)
As in the usual elasticity theory, the left-hand side of the work relation (3.5.27) represents the total work of the external force acting on the laminate. Consequently,
w is the specific elastic energy stored in the deformed laminate. Hence, the total
elastic energy W stored in the laminate is
Z
W =
wda.
(3.5.30)
D
The above interpretation and equation (3.5.27) represents the content of the
announced and proved theorem of work and energy.
This theorem tells us what kind of boundary conditions can be given
on the boundary D, in order to assume the uniqueness of the solution of various
boundary value problems. The structure of the left-hand side of the work relation
(3.5.27) shows that the following fields can be prescribed on the boundary D:
Un or Nnn , U or Nn , U3 or Qn + Mn, and U3,n or Mnn .
(3.5.31)
The same work relations show that the following theorem takes place.
Uniqueness theorem. If the specific elastic energy w is a positive definite
quadratic form, the various boundary value problems can have no more than one
regular solution, modulo a rigid displacement.
3.6
We observe that due to the symmetry relations (3.5.9), the differential system
(3.5.6)(3.5.8) is self-adjoint. Due to this fact we can establish various variational
and extreme principles, corresponding to various boundary value problem.
In order to do this, we introduce first the three-dimensional vectors U and
U having the components (U , U3 ) and (U , U3 ), respectively. Also we consider
the energy functional
Z
W (U) =
w(U)da
(3.6.1)
D
where w is the quadratic form given by the equation (3.5.28) and (3.5.29).
161
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
W =
d
W (U + U)|=0 , being a real variable.
d
(3.6.2)
In order to evaluate the above variation, we take into account the geometric
relation (3.5.6), the constitutive equations (3.5.7), the symmetry properties (3.5.8)
and the expression (3.5.29) of the specific elastic (strain) energy w. Thus, after
some elementary computations, we get
Z
W =
(N U, M U3, )da.
(3.6.3)
D
(U N n + U3 M, n U3, M n )ds.
(3.6.4)
(3.6.5)
Un and U being the normal and tangential components of the in-plane displacement field (U1 , U2 ).
Similarly, using the equilibrium condition (3.5.8)4 and the equation (3.5.23),
we obtain
U3 M, n = U3 Q n = U3 Qn ,
(3.6.6)
Qn = Q n being the normal component of the resultant shear force (Q1 , Q2 ).
Using a relation analogous to (3.5.27), we get
U3, M n = U3, Mn = (U3 Mn ), U3 Mn, + U3,n Mnn .
(3.6.7)
Introducing (3.6.5), (3.6.6), (3.6.7) in the line integral of equation (3.6.4), and
observing that
Z
(U3 Mn ), ds = 0,
D
162
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
(3.6.8)
(3.6.10)
Taking into account equation (3.6.8), for the variation I of I in U, in the direction
U, we get
Z
I =
{N, U + (M, + q)U3 } da
D
Z
+
{(Nnn )Un + (Nn )U
D
(3.6.12)
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
163
(3.6.15)
(3.6.16)
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
164
The proof of the second variational principle is similar to that given for the
first one, and is entirely based on equation (3.6.16).
Let us observe that in the second variational principle, U and U are not
arbitrary on the boundary D; U must satisfy the given boundary conditions and
U must satisfy homogeneous boundary conditions. In this sense, we consider the
boundary conditions (3.6.13) as being essential boundary conditions. The situation is entirely analogous to that existing in the usual elasticity theory where the
displacement is prescribed on the boundary of the body.
The above boundary conditions are not typical either in the classical plate
theory or in the composite laminate theory. In these domains, the most frequently
encountered boundary value problems have a mixed character. That is, on boundary D are given some data concerning the displacements and complementary information concerning the force and moment resultants. Jones shows ([3.2], Chapter
5) that the most frequently used boundary conditions in laminate theory are formulated as a choice between simply supported, clamped or free edges. The situation
for laminate plates (laminates) is complex because there are actually four types
of boundary conditions that can be called simply supported edges and four types
of boundary conditions that can be called clamped edges. According to Jones,
the eight types of simply supported (prefix S) and clamped (prefix C) boundary
conditions are commonly classified as
S1 : U3 = 0 , Mnn = 0 , Un = , U = ,
S2 : U3 = 0 , Mnn = 0 , Nnn = , U = ,
S3 : U3 = 0 , Mnn = 0 , Un = , Nn = ,
S4 : U3 = 0 , Mnn = 0 , Nnn = , Nn = ,
(3.6.17)
and
C1 : U3 = 0 , U3,n = 0 , Un = , U = ,
C2 : U3 = 0 , U3,n = 0 , Nnn = , U = ,
C3 : U3 = 0 , U3,n = 0 , Un = , Nn = ,
C4 : U3 = 0 , U3,n = 0 , Nnn = , Nn = , on D.
(3.6.18)
165
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
1
0
0
0
)}da.
)D (k + k
) + (k + k
+(e + e0 )B (k + k
2
Taking into account the symmetry relations (3.5.9) and the constitutive equations
(3.5.7), we get
Z
0
M )da,
W (U + U0 ) = W (U) + W (U0 ) +
(e0 N + k
D
(3.6.19)
In order to obtain minimum principles, this equation plays the same role as
that played by equation (3.6.8) used to obtain variational principles.
Let us consider first the boundary value problem (3.6.10) and the functional
I(U) defined by equation (3.6.11). As in the usual elasticity theory, we call I(U)
the potential energy of the laminate, corresponding to the traction boundary
value problem (3.6.10). We shall denote by A the set of all regular displacement
fields U.
Using equation (3.6.19), we can formulate and prove a principle of minimum
potential energy appropriate to the composite laminate theory and to the boundary
value problem (3.6.10).
In order to prove the minimum principles, we assume that the specific elastic
(strain) energy w of the laminate, defined by equation (3.5.30), is a positive definite
quadratic form. We have:
The first principle of minimum potential energy. Let A denote the set
of all regular displacement field U and let I(U) be the functional defined on A by
equation (3.6.11). Let U be the solution of the traction boundary value problem
(3.6.10). Then
e
I(U) I(U)
(3.6.20)
e A, and equality holds only if U
e = U modulo a rigid displacement.
for every U
166
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
e U A and define U0 =U
e U. Using the definition (3.6.11) of the
Let U,
functional I(U), the equation (3.6.19), the equilibrium equation (3.5.8) 1,4 and the
fact that U is a solution of the traction boundary value problem (3.6.10), after
some simple computations, we get
e = I(U) + W (U0 ).
I(U)
(3.6.21)
e
J(U) J(U)
(3.6.22)
(3.6.23)
Since w is positive definite, from (3.6.23) we can conclude that (3.6.22) is true.
0
e
If J(U) = J(U), we obtain e0 = k
= 0 on D, hence U0 is a rigid displacement
0
field. Moreover, U must satisfy the homogeneous boundary conditions on D.
Hence U0 = 0 on D and the proof is complete.
We note again that now the admissible displacement fields for which the
functional J is defined must satisfy the given displacement boundary conditions
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
167
on the boundary line D. This is due to the fact that the displacement boundary
conditions (3.6.13) are essential boundary conditions.
We recall that the most frequent boundary conditions in the laminate theory
correspond to simple supported or clamped edges and are of mixed type. Using
equation (3.6.19), appropriate potential energies can be found, and can be formulated and proved appropriate minimum principles, corresponding to various
boundary conditions listed in the relations (3.6.17) and (3.6.18).
As in the usual elasticity theory, we can prove the conversers of the given
minimum principles.
Converse of the first principle of minimum potential energy. Let
U A and suppose that
e
I(U) I(U)
(3.6.24)
0
+Mn, ) U3,n
(Mnn )}ds 0.
0
(Mnn )}ds = 0,
+Mn, ) U3,n
since W (U0 ) = 0.
Since U0 is an arbitrary field, from the above equation and from the supplementary constitutive relation (3.6.5), it follows that the equilibrium equations
(3.5.8) and the boundary conditions (3.6.10) are satisfied. Hence, U is a regular
solution of the traction boundary value problem (3.6.10). Analogously, we have
the following.
Converse of the second principle of minimum potential energy. Let
U B. Suppose that
e
J(U) J(U)
(3.6.25)
e B. Then U is a solution of the displacement boundary value
for every U
problem (3.6.13).
168
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
Since U is arbitrary in D, from the last equation and from the supplementary
constitutive relations (3.6.5), it follows that the equilibrium equations (3.5.8) are
verified by U. Moreover, since U B, U satisfies also the boundary conditions
(3.6.13). Hence, U is a regular solution of the displacement boundary value
problem (3.6.13) and the demonstration is complete.
In a similar way, the converses of principles of minimum potential energies
corresponding to various, possible boundary value problems can be proved.
We note also that, as in the usual elasticity theory, the principles of minimum potential energy can be used to prove uniqueness theorems for the involved
boundary value problems.
As we have seen, in order to prove uniqueness theorems and minimum principles, we have assumed that the specific deformation (strain) energy w is a positive
definite quadratic form.
We recall also that from the beginning we have supposed that the elasticity
tensor c of any linearly elastic body is positive definite. Using this hypotheses, in
the Section 2.2 we have derived various restrictions which must be satisfied by the
elasticities in order to ensure the positiveness of c.
In a natural way, the following question is raised. If the elasticities ck , k =
1, ..., N of the laminate are positive definite, does the specific strain energy w of
the laminate have the same property?
Since the reduced transformed stiffnesses (Qij )k of the laminae and the laminate stiffnesses Aij , Bij , Dij are complicated functions of the laminae mechanical
and geometrical characteristics, it is difficult to find an answer for the above question for a laminate having arbitrary structure.
This is the reason why we shall analyze the problem only for the particular,
but important case of the regular antisymmetric cross-ply laminates, for which
the extensional, coupling and bending stiffnesses of the composite laminate can
be expressed by relatively simple relations in terms of the primary mechanical
characteristics of the laminae, using Tsais formulas (3.4.26).
We recall that from the positive definiteness of the elasticity c of an orthotropic elastic material, it follows that the technical constants of the body must
satisfy the restrictions (2.2.74)(2.2.79). In particular, we have
(3.6.26)
169
E1
, |21 | <
E2
E2
.
E1
(3.6.27)
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
(3.6.28)
Hence in this case, the Poissons ratios 12 and 21 satisfy the restrictions
0 < 12 <
E1
, 0 < 21 <
E2
E2
.
E1
(3.6.29)
Using the Tsais stiffness ratio f , we can express the inequalities (3.6.29) satisfied
by Poissons ratios 12 and 21 in the equivalent form
0 < 12 <
p
1
, 0 < 21 < f .
f
(3.6.31)
We return to equations (3.1.6) giving the (primary) reduced stiffness Q11 , Q12 ,
Q22 and Q66 in terms of the engineering constants of the lamina; we have
21 E1
E1
,
, Q12 =
1 12 21
1 12 21
E2
and Q66 = G12 .
=
1 12 21
Q11 =
Q22
(3.6.32)
Thus, from (3.6.26) and (3.6.28) we can conclude that all these stiffnesses are
positive; i.e.
Q11 , Q12 , Q22 , Q66 > 0.
(3.6.33)
The positive definiteness of the elasticity c of the lamina assures only the
positivity of the reduced stiffnesses Q11 , Q22 and Q66 . For the positivity of the
stiffness Q12 some supplementary assumption, leading to the inequalities (3.6.28),
must be also used.
We recall now Tsais formulas (3.4.26) giving the extensional, coupling and
bending stiffnesses of a regular antisymmetric cross-ply laminate. Using also the
170
221
E1
1+f
A11 , A66 = hG12 ,
, A12 =
h
1+f
1 12 21
2
1 f 1
hA11 ,
B11 =
2N f + 1
h2
h2 212
h2
A66 .
(3.6.34)
A11 , D66 =
A11 , D12 =
D11 =
12
12 1 + f
12
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
A11 =
(3.6.35)
We observe now that the positive definiteness of the specific strain energy w of
a laminate, defined by equation (3.5.29), is equivalent with the positive definiteness
of the global stiffness matrix [E] of the laminate, introduced by equation (3.3.20).
According to Tsais relations (3.4.26), the stiffness matrix [E] of a regular
antisymmetric cross-ply laminate has the following expression:
A11 A12
0
B11 0
0
A12 A11
0
0
B11 0
0
0
A66 0
0
0
.
(3.6.36)
[E] =
B11 0
0
D11 D12
0
0
B11 0
D12 D11
0
0
0
0
0
0
D66
A11
A12
0
B11
A12
A11
0
0
0
0
A66
0
6 = det [E] .
B11
0
0
D11
A11
A12
, 5 = 0
B11
0
A12
A11
0
0
B11
0
0
A66
0
0
B11
0
0
D11
D12
0
B11
0
D12
D11
(3.6.37)
Long, but elementary computations lead to the following expression of the
above determinants:
171
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
2
421
A211 ,
1 = A11 , 2 = 1
(f + 1)2
(
2
421
1
h2
3
1
A66 A11
4 =
(f + 1)2
3
2
(
2
421
1
h4 4
1
A11 A66
5 =
(f + 1)2
3
16
3 = 2 A66 ,
2 )
f 1
1
,
N2 f + 1
2 )
f 1
1
,
N2 f + 1
6 = 5 D66 .
(3.6.38)
(3.6.39)
(3.6.40)
The third relation (3.6.38) and the above equality show that
3 > 0,
(3.6.41)
=
f +1
3 N
2
421
1
(f + 1)2
1
2
N
f 1
f +1
2
> 0.
(3.6.42)
172
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
5 > 0 , 6 > 0
(3.6.43)
3.7
Rectangular laminates
173
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
Generally, the relation (3.7.1) is a complicated and coupled differential system. In-plane deformation and distortions of the laminate appear together, in an
inseparable way. But, obvious and sometimes important simplifications result
when the laminate is symmetric about the middle surface (Bij = 0), specially
orthotropic (all the terms with 16 and 26 indices vanish in addition to the B ij ),
2
homogeneous (Bij = 0 and Dij = Aij h12 ), or isotropic. In all these cases, equations
(3.7.1)1,2 are uncoupled from the equations (3.7.1)3 . In such situations, the first
two equations contain only the in-plane displacements U1 , U2 , and the third one
contains only the normal displacement U3 . Accordingly, equation (3.7.1)3 must
be solved to obtain the deflections of a plate and the system (3.7.1)1,2 must be
integrated to obtain the in-plane deformations of the laminate. Unfortunately, the
more general case of nonsymmetric laminates (there exist nonvanishing coupling
stiffnesses Bij ) requires the simultaneous integration of the coupled system (3.7.1).
Currently, many methods exist to solve the equilibrium equations (3.7.1).
The methods range from exact solutions to approximate numerical integration,
using finite element or finite difference approaches and various approximate energy methods of Rayliegh-Ritz or Galerkin type, based on the given minimum or
variational principles. In what follows, we shall consider only the case of a rectangular laminate, taking into account a small number of special examples.
Let us consider first the general class of laminated rectangular plates, as
shown in Figure 3.20, that are simply supported along edges x1 = 0, x1 = a1 , x2 =
0, x2 = a2 and subjected to a given normal load q = q(x1 , x2 ).
We suppose that the given normal load q = q(x1 , x2 ) can be expanded in a
double Fourier series; i.e.
q(x1 , x2 ) =
m=1 n=1
qmn sin
nx2
mx1
.
sin
a2
a1
(3.7.2)
Figure 3.20: Simply supported laminated rectangular plate under distributed normal load.
In what follows, the various types of possible laminates, such as specially orthotropic, symmetric angle-ply, antisymmetric cross-ply and antisymmetric angleply will be analyzed for the loading q = q(x1 , x2 ) given by equation (3.7.2).
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
174
As we know, a specially orthotropic laminate has either a single layer of specially orthotropic material or multiple specially orthotropic layers that are symmetrically placed about the laminate middle surface. For this special case considered, the nonvanishing laminate stiffnesses are A11 , A12 , A22 , A66 , D11 , D12 , D22 ,
D66 . In other words, neither shear or twist coupling, nor bending-extension coupling exists. Also, the in-plane deformations and bending and twisting are decoupled.Thus, for the laminate problem, the vertical deflection are described only by
one differential equation of equilibrium, resulting from the third equation (3.7.1),
D11 U3,1111 + 2(D12 + 2D66 )U3,1122 + D22 U3,2222 = q.
(3.7.3)
(3.7.4)
m=1 n=1
Amn sin
nx2
mx1
.
sin
a2
a1
(3.7.5)
This normal displacement field satisfies the equilibrium equation (3.7.3) only
if the Fourier coefficients amn are given by the relation
Amn =
qmn
1
.
m 4
4
D11 ( a1 ) + 2(D12 + 2D66 )( am1 )2 ( an2 )2 + D22 ( an2 )4
(3.7.6)
Once the normal displacement or deflection U3 is known, all force and moment
resultants can be obtained using the corresponding constitutive equations.
The case of a symmetric angle-ply laminate is much more complicated, even
if, for these structures, the coupling stiffnesses Bij are also vanishing. But now the
shear coupling stiffnesses A16 , A26 and the twist coupling stiffnesses D16 , D26 are
not zero. The equilibrium equation (3.7.1)3 , describing the deflection of the plate,
is decoupled and takes the form
D11 U3,1111 + 4D16 U3,1112 + 2(D12 + 2D66 )U3,1122 + 4D26 U3,1222 + D22 U3,2222 = q.
(3.7.7)
The simply supported edge condition becomes
U3 = 0, M11 = D11 U3,11 D12 U3,22 2D16 U3,12 = 0 for x1 = 0 and x1 = a1 ,
U3 = 0, M22 = D12 U3,11 D22 U3,22 2D26 U3,12 = 0 for x2 = 0 and x2 = a2 .
(3.7.8)
175
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
Now the solution of the equilibrium equation (3.7.7) is not as simple as before,
because of the presence of D16 and D26 . Due to the terms involving these nonvanishing coefficients, the method using the separation of variables cannot be applied and a Fourier type expression (3.7.5) does not satisfy the governing equation
(3.7.7). Moreover, the expansion (3.7.5) also does not satisfy the boundary conditions (3.7.8), again since the terms involving D16 and D26 are present. Actually,
the variables x1 and x2 cannot be separated in the expression of the normal displacement U3 . This is the reason why Ashton (see Jones [3.2], Chapter 5) has
solved the problem using the second principle of the minimum potential energy.
The problem being decoupled, the involved functional J(U), given in the equation
(3.6.14), takes the following simplified form:
1
2
2
2
+ 2D12 U3,11 U3,22 + D22 U3,22
(D11 U3,11
Z
2
qU3 da.
+4D66 U3,12 + 4D16 U3,11 U3,12 + 4D26 U3,22 U3,12 )da
J = J(U3 ) =
(3.7.9)
U3 max = U3 (0, 0) =
0.00425a4 q
.
D11
If D16 and D26 are ignored, that is the symmetric angle-ply laminate is approximated by a special orthotropic laminate having
(D12 + 2D66 )
D11
= 1 , D16 = D26 = 0,
= 1,
D11
D22
0.0032a4 q
.
D11
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
176
Hence, the error in neglecting the twist coupling terms D16 and D26 , is about
24 percent, which represents a no negligible error. Thus, generally, a specially
orthotropic laminate is an unacceptable approximation for a symmetric angle-ply
laminate.
Let us consider now a regular antisymmetric cross-ply laminate. Such a
laminate has nonvanishing extensional stiffnesses A11 , A12 , A11 = A22 and A66 ,
bending-extensional coupling stiffnesses B11 and B22 = B11 , and bending stiffnesses D11 , D12 , D22 = D11 and D16 . Since B11 and B22 are not vanishing, the
displacement equilibrium equations are coupled. The general system (3.7.1) takes
the following simplified form:
A11 U1,11 + A66 U1,22 + (A12 + A66 )U2,12 B11 U3,111 = 0,
(A12 + A66 )U1,12 + A66 U2,11 + A11 U2,22 + B11 U3,222 = 0,
(3.7.10)
U2 =
U3 =
m=1 n=1
X
m=1 n=1
X
m=1 n=1
mn cos
nx2
mx1
,
sin
a2
a1
mn sin
nx2
mx1
,
cos
a2
a1
mn sin
nx2
mx1
.
sin
a2
a1
(3.7.12)
It is easy to see that the boundary conditions (3.7.11) are satisfied. Obviously,
the Fourier coefficients mn , mn and mn can be obtained taking into account the
equilibrium equation (3.7.10).
The results obtained by Whitney and Leissa are replotted in Figure 3.21 for
the special situation in which the normal load is one term of the Fourier series;
177
x2
x1
, q0 = const.
sin
a2
a1
15
U3,MAX E 2 h3 3
10
q 0 a24
E1
G
=40 E12=0.5
E2
2
=0.25
12
NUMBER OF LAYERS
10
4
6
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
q = q(x1 , x2 ) = q0 sin
x
q(x1,x2)=q0sin ax1 sin a22
1
2
a
ASPECT RATIO, a 1
2
178
In Figure 3.22 are given the maximum deflection of a square regulate antisymmetric cross-ply laminate, under sinusoidal transverse load, in terms of the
E1
1
orthotropic modulus ratio E
E2 . At E2 = 1, the effect of coupling is nonexistent. As
E1
E2 increases, the effect of coupling between bending and extension increases.
25
3
U 3,MAX E 2 h 2
10
q0a 4
G12
=0.5
E2
20
=0.25
12
15
2
10
4
NUMBER
OF
LAYERS
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
30
x
q(x1,x2)=q sin ax1 sin a 2
0
10
20
30
MODULUS RATIO,
E1
E2
40
50
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
3.8. PROBLEMS
179
3.8
Problems
P3.1 Show that for an isotropic lamina and in plane stress state, the strainstress relation are
1
1
S11 S12 0
2
2 = S12 S11 0
0
0
2(S11 S12 )
6
6
with S11 = E1 , S12 = E , E and being Youngs modulus and the Poissons
ratio of the laminae, respectively.
P3.2 Show that for an isotropic lamina and in plane stress state, the stressstrain relations are
1
Q11 Q12 0
1
2 = Q12 Q11 0
2
0
0
Q66
6
6
with
Q11 =
E
E
E
= = G.
, Q66 =
, Q12 =
2
2
2(1 + )
(1 )
(1 )
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
180
E1 = 10E2 , G12 =
1
E2 , 12 = 0.3 , E2 = 18.5GP a.
3
with
E1
and b =
a=
Ex
E1
G12
212
P3.10 Use the above expression of Ex to find its maxima and minima. Show
that if
E1
,
G12 >
2 (1 + 12 )
E1
E1
E2
+ 12
,
181
3.8. PROBLEMS
Q11
U1 U2
0
U3
0
1
Q22 U1 U2 0
U3
0
cos 2
Q12 U4 0
0
U
0
3
sin 2 ,
=
U5 0
Q
0
4U3 0
cos 4
66
2Q 0
0
U2 0
2U3
16
sin 4
0
0
U2 0
2U3
2Q26
where U1 , U2 , U3 , U4 , U5 are given by the equations (3.1.25).
P3.14 The engineering constants of a T300/5208 graphyte/epoxy fiber-reinforced lamina are given in Table 3.1. Find the coefficients U1 , U2 , U3 , U4 and U5
for this material.
P3.15 A boron/epoxy fiber-reinforced composite lamina has the following
engineering constants:
E1 = 206.85GP a , E2 = 20.68GP a , 12 = 0.3 , G12 = 6.86GP a.
(a) Plot the reduced transformed
stiffnesses Q11 (), Q22 (), Q12 () and Q66 ()
for this lamina for 00 , 900 .
(b) Plot the reduced
transformed
stiffnesses Q16 () and Q26 () for the same
lamina and for 00 , 900 .
(c) Analyze the results obtained in (a) and (b).
P3.16 Show that the functions S 11 (), ..., S 66 (), given by equations (3.1.21)
can be expressed in the following matrix form:
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
S 11
S 22
S 12
S 66
S 16
S 26
V1
V1
V4
V5
0
0
V2
V2
0
0
0
0
0
0
0
0
V2
V2
V3
V3
V3
4V3
0
0
0
0
0
0
2V3
2V3
1
cos 2
sin 2
cos 4
sin 4
where
V1 =
V2 =
V3 =
V4 =
V5 =
1
(3S11 + 3S22 + 2S12 + S66 ) ,
8
1
(S11 S22 ) ,
2
1
(S11 + S22 2S12 S66 ) ,
8
1
(S11 + S22 + 6S12 S66 ) ,
8
1
(S11 + S22 2S12 + S66 ) .
2
182
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
(a)
Ef
= 10;
Em
(b)
b12
G
Gm
Ef
= 100.
Em
P3.19 What conclusion do you get if you assume for the determination of
b2 equal strains in both the fiber and the matrix,
the overall transverse modulus E
instead of equal stresses.
P3.20 Show that if the curvatures k (x1 , x2 ), , = 1, 2 are vanishing, the
middle surface of a composite laminate rests plane after the deformation of the
laminate. In other words, show that the curvatures characterizes the bending and
the twisting of the laminate.
P3.21 Let us assume that the in-plane deformations e (x1 , x2 ) and the curvatures k (x1 , x2 ), , = 1, 2 of a laminate are vanishing.
Find in that case the components uk (x1 , x2 , x3 ), k = 1, 2, 3 of the displacement field and give the geometrical meaning of the obtained result.
P3.22 Assuming that a laminate is submitted to a rigid displacement, find
the components of the in-plane deformation and the components of the curvature.
P3.23 Show that the extensional stiffnesses Aij , i, j = 1, 2, 6 of a composite
laminate can be expressed by the following relations:
A11 = U1 V0A + U2 V1A + U3 V3A ,
A22 = U1 V0A U2 V1A + U3 V3A ,
A12 = U4 V0A U3 V3A ,
A66 = U5 V0A U3 V3A ,
1
A16 = U2 V2A U3 V4A ,
2
1
A26 = U2 V2A + U3 V4A ,
2
where
V0A = h,
N
X
V1A =
(zk zk1 ) cos 2k ,
k=1
V2A =
N
X
k=1
V3A =
N
X
k=1
V4A =
N
X
k=1
183
3.8. PROBLEMS
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
B12 = U3 V3B ,
B66 = U3 V3B ,
1
B16 = U2 V2B U3 V4B ,
2
1
B26 = U2 V2B + U3 V4B ,
2
where
V0B = 0,
N
V1B
1X 2
2
(zk zk1
) cos 2k ,
=
2
k=1
V2B =
1X 2
2
(zk zk1
) sin 2k ,
2
k=1
V3B =
1X 2
2
) cos 4k ,
(zk zk1
2
k=1
V4B =
1X 2
2
) sin 4k .
(zk zk1
2
k=1
184
where
V0D =
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
V1D =
h3
,
12
N
1X
k=1
3
) cos 2k ,
(zk3 zk1
V2D =
1X 3
3
) sin 2k ,
(zk zk1
3
k=1
V3D =
1X 3
3
) cos 4k ,
(zk zk1
3
k=1
V4D
1X 3
3
) sin 4k .
(zk zk1
=
3
k=1
U1 U2
0
U3
0
[A11 , B11 , D11 ]
V0[A,B,D]
in the
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
3.8. PROBLEMS
185
P3.31 Show that if a cross-ply laminate is symmetric about its middle surface,
all of its coupling stiffnesses are vanishing.
P3.32 Starting with Tsais formulas (3.4.24) and using the definition of a
regular symmetric cross-ply laminate, express the stiffnesses of the laminate in
2
terms of the reduced stiffness Q11 , the thickness h and the stiffness ratio f = E
E1 .
P3.33 Give a direct proof to the results obtained in P3.32.
P3.34 Prove that the global stiffness matrix [E] of a regular symmetric crossply laminate is positive definite if the elasticity tensor of the composing laminae
are positive definite, and the Poissons ratio satisfy the inequalities ij > 0 for
i, j = 1, 2, 3.
P3.35 An expanded view of a [+45/45/45/+45] regular angle-ply laminate
consisting of 0.25-mm thick unidirectional AS/3501 graphite/epoxy laminae is
shown in Figure 3.23. Determine the stiffness matrix [E] of this laminate.The
engineering constants of the laminae are
E2
= 0.0196.
E1
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
186
Q11 + = Q11 , Q12 + = Q12 ,
Q22 + = Q22 , Q66 + = Q66 ,
Q16 + = Q16 , Q26 + = Q26 .
P3.40 Show that for a symmetric (N odd!) regular angle-ply laminate, the
187
3.8. PROBLEMS
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
(A11 , A12 , A22 , A66 ) = h Q11 , Q12 , Q22 , Q66 ,
h
Q16 , Q26 , Bij = 0, i, j = 1, 2, 6,
(A16 , A26 ) =
N
h3
Q11 , Q12 , Q22 , Q66 ,
(D11 , D12 , D22 , D66 ) =
12
h3 2N 2 2
Q16 , Q26 .
(D16 , D26 ) =
3
12 N
P3.42 Show that A16 , A26 and D16 , D26 for a symmetric regular angle-ply
laminate approach zero as the number of layers increases, while the total thickness
is held constant. What happens if equal thickness layers are added so that the
total laminate thickness increases, too?
P3.43 Show that B16 and B26 for an antisymmetric regular angle-ply laminate
approach zero as the total number of layers increases, while the total thickness is
held constant. What happens if equal thickness layers are added so that the total
laminate thickness increases, too?
P3.44 In what conditions are the stiffnesses A16 , A26 , D16 , D26 of a symmetric regular angle-ply laminate vanishing? In what conditions are the stiffnesses
B16 , B26 of an antisymmetric regular angle-ply laminate vanishing?
P3.45 Using the notations introduced in the Section 3.5 shows that
U Nn = U1 Nn1 + U2 Nn2 = Un Nnn + U Nn .
P3.46 Show that the specific strain energy w of a composite laminate can be
expressed in the following matrix form:
w=
1 T
[d] [E] [d] ,
2
where
P3.47 Using the work theorem for composite laminates and assuming the positive definiteness of the specific elastic energy w, formulate and prove the uniqueness
theorem corresponding to the classical plate theory of composite laminates. More
exactly, prove that if Un or Nnn , U or Mn , U3 or Qn + Mn, and U3,n or Mnn
Downloaded by [Hong Kong University of Science and Technology] at 01:44 17 November 2014
188
are vanishing on the boundary D of the plane domain D occupied by the middle
surface of the laminate, then e and k are vanishing on D, assuming usual regularity conditions. Hence, the solution is vanishing, modulo a rigid displacement
of the plate. Give conditions in which this displacement is also vanishing!
P3.48 Formulate and prove a variational principle corresponding to the simply
supported edge boundary conditions S1.
P3.49 Formulate and prove a variational principle corresponding to the clamped edge boundary conditions C4.
P3.50 Find the appropriate potential energy and prove its minimum property
for a simply supported laminate, submitted to the boundary conditions S2.
P3.51 Find the appropriate potential energy and prove its minimum property
for a clamped laminate, submitted to the boundary conditions C2.
P3.52 Using the corresponding principles of minimum potential energy prove
the uniqueness theorems corresponding to the boundary value problems S2 and
C2, respectively.
P3.53 Formulate and prove the converses of the principles of minimum potential energy corresponding to the boundary value problems S2 and C2, respectively.
Bibliography
[3.1] Ashton J.E., Whitney J.M., Theory of laminated plates, Progress in Material Science Series, Vol. IV, Technomic Publishing Co., Stanford, 1970.
[3.2] Jones, R.M., Mechanics of composite materials, Hemisphere Publishing Co.,
New York, 1975.
[3.3] Christensen, R.M., Mechanics of composite materials, John Wiley and Sons,
1979.
[3.4] Tsai, W., Hahn, M.T., Introduction to composite materials, Technomic Publishing Co., Westport, Conneticut, 1980.
[3.5] Cristescu, N., Mechanics of composite materials, University of Bucharest,
Bucharest, 1983 (in Romanian).
[3.6] Whitney, J.M., Analysis of laminated anisotropic plates, Technomic, Lancaster, PA, 1987.
[3.7] Gibson, R.F., Principles of composite material mechanics, McGraw-Hill Inc.,
New York, 1994.
[3.8] Lekhnitski, S.G., Theory of elasticity of an anisotropic elastic body, HoldenDay, San Francisco, 1963.
[3.9] Barran, J.J., Laroze, S., Calcul des structures en materiaux composite, Ecole
National Superieur de lAeronautique et de lespace, Dept. structures, materiaux, technologie, 1987.